Objekttyp: Journal

Zeitschrift: IABSE reports = Rapports AIPC = IVBH Berichte

Band (Jahr): 67 (1993)

PDF erstellt am: 18.10.2021

Nutzungsbedingungen
Die ETH-Bibliothek ist Anbieterin der digitalisierten Zeitschriften. Sie besitzt keine Urheberrechte an
den Inhalten der Zeitschriften. Die Rechte liegen in der Regel bei den Herausgebern.
Die auf der Plattform e-periodica veröffentlichten Dokumente stehen für nicht-kommerzielle Zwecke in
Lehre und Forschung sowie für die private Nutzung frei zur Verfügung. Einzelne Dateien oder
Ausdrucke aus diesem Angebot können zusammen mit diesen Nutzungsbedingungen und den
korrekten Herkunftsbezeichnungen weitergegeben werden.
Das Veröffentlichen von Bildern in Print- und Online-Publikationen ist nur mit vorheriger Genehmigung
der Rechteinhaber erlaubt. Die systematische Speicherung von Teilen des elektronischen Angebots
auf anderen Servern bedarf ebenfalls des schriftlichen Einverständnisses der Rechteinhaber.

Haftungsausschluss
Alle Angaben erfolgen ohne Gewähr für Vollständigkeit oder Richtigkeit. Es wird keine Haftung
übernommen für Schäden durch die Verwendung von Informationen aus diesem Online-Angebot oder
durch das Fehlen von Informationen. Dies gilt auch für Inhalte Dritter, die über dieses Angebot
zugänglich sind.

Ein Dienst der ETH-Bibliothek

ETH Zürich, Rämistrasse 101, 8092 Zürich, Schweiz, www.library.ethz.ch

http://www.e-periodica.ch
 IABSE REPORT - RAPPORTS AIPC - IVBH REPORTS

Volume 67

IABSE COLLOQUIUM
COPENHAGEN 1993

Remaining Structural Capacity

Capacitä restante d'une structure
Rest-Tragvermögen von Bauwerken

REPORT RAPPORT BERICHT

Organized by
Danish Group of IABSE
 III

Preface

Over the last Century, structural engineers have covered our globe with structures that are essential to
sustain the life of more than 5 billion people. The present and future need to maintain and adapt these
structures is evident; however, our ability to maintain and adapt the structural capacity is pitifully
inadequate.

Structures can operate long after their design life has been reached, indicating that structural capacity still
remains. Increased loads, exceeding design loads, are often carried by structures without deterioration.
One problem is our limited capability in predicting future loading patterns over the years in which the
structure shall serve us. The question arises how the reliability level of the structure can be defined or
evaluated.

The papers in this colloquium explore the structural capacity of existing structures by uncovering and
utilizing the often considerable structural reserves. Use has been made of recent advances in
computational methods in structural analysis and in reliability methods. In addition, Information obtained
from load testing helps to confirm the validity of analysis. Finally, a number of cases demonstrate what
is and can be done in practice.

More studies are necessary to fully understand how risk levels for adapted structures or existing
structures under increased loading must be chosen. Yet this colloquium report clearly shows the direction
in which to go. It is hoped that structural engineers will delve deeper and more thoroughly into these
specialized areas of structural assessment and concepts of reliability.

Delft, February 1993 Johan Blaauwendraad

Chairman, Scientific Committee
IV

Präface

Au cours des cent dernieres annöes, les ingänieurs civils ont couvert notre globe avec des constructions
essentielles ä la vie de plus de 5 milliards d'etres humains. Les besoins actuels et futurs d'entretien et
d'adaptation de ces constructions apparaissent clairement, mais notre capacite* a les satisfaire est
malheureusement inadäquate.

Les structures de gönie civil peuvent etre utilisöes bien au-delä de leur duröe de vie projetöe, montrant
ainsi qu'une aptitude a rösister et a bien se comporter subsiste. Des charges accrues sont souvent
supportöes par les structures sans que celles-ci ne subbisent de dommages. II est aujourd'hui difficile de
prövoir les cas de charges futurs pour lesquels la construction devra ötre maintenue en service. II est
cependant nöcessaire de döfinir ou d'övaluer le degre" de fiabilite* de telles structures.

Les contributions ä ce Colloque concernent la capacitö restante de constructions existantes, en cherchant

a döterminer et utiliser les röserves de rösistance, souvent importantes, de ces constructions. Les röcents
progres dans les möthodes informatiques de l'analyse des structures et les möthodes de fiabilitö ont 6t6
appliquöes dans ce domaine. De plus les rösultats d'essais permettent de confirmer la valeur de cette
analyse. Enfin un nombre d'ötudes de cas montre ce qui a 6t6 fait en pratique, et ce qui peut l'etre encore.

De plus amples ötudes sont nöcessaires pour comprendre parfaitement comment choisir les niveaux de
risques pour des structures röparöes ou pour des structures existantes sous des charges plus 6lev6es. Le
Rapport du Colloque indique la direction a suivre. II faut espörer que les ingönieurs des structures
s'intöresseront de facon plus approfondie a ces domaines sp£cialis£s de l'övaluation structurale et des
concepts de fiabilitö.

Delft, fövrier 1993 Johan Blaauwendraad

Präsident du Comite' Scientifique
Vorwort

Im zurückliegenden Jahrhundert entstanden überall auf unserem Erdball Tragwerke des konstruktiven
Ingenieursbaus, die für über 5 Milliarden Menschen lebenswichtig sind. Offensichtlich müssen sie heute
und in Zukunft unterhalten und neuen Auforderungen angepasst werden, doch unsere Fähigkeit dazu ist
beklagenswert unzulänglich.

Tragwerke können noch in Betrieb stehen, wenn ihre veranschlagte Nutzungsdauer schon lange
überschritten ist; ein Zeichen, dass ihre Tragfähigkeit noch gegeben ist. Einwirkungen, die die
Bemessungslastfälle übersteigen, werden oft ohne Anzeichen einer Schwächung aufgenommen. Was uns
fehlt, sind ausreichende Möglichkeiten zur Vorhersage zukünftiger Beanspruchungen in all den Jahren,
während denen das Trag werk uns dienen soll. So erhebt sich die Frage, wie die Höhe der Zuverlässigkeit
des Tragwerks definiert oder bestimmt werden kann.

Die Beiträge dieses Kolloquiums ergründen die Tragfähigkeit bestehender Ingenieurbauwerke, indem sie
die oft beträchtlichen Tragreserven aufdecken und nutzbar machen. Dabei bedienen sie sich jüngster
Fortschritte in statischen Berechnungsverfahren und Zuverlässigkeitsmethoden. Zusätzlich werden durch
Belastungsversuche Informationen gewonnen, die die Gültigkeit der Berechnung zu überprüfen gestatten.
An einigen Beispielen wird schliesslich aufgezeigt, was in der Praxis gemacht wird oder gemacht werden
kann.

Weitere Studien sind nötig, um völlig zu verstehen, wie Risikoniveaus für abgeänderte oder einer höheren
Belastung unterworfene Tragwerke zu wählen sind. Immerhin gibt dieser Kolloquiumsbericht die Richtung
vor, die wir einschlagen müssen. Es ist zu hoffen, dass konstruktiv tätige Bauingenieure sich tiefer und
gründlicher in dieses spezielle Feld der Tragfähigkeitsbestimmung und Zuverlässigkeitsanalyse einarbeiten
werden.

Delft, Februar 1993 Johan Blaauwendraad

Vorsitzender des
Wissenschaftlichen Komitees
VI

Colloquium Committees

Scientific Committee

J. Blaauwendraad, Netherlands (Chairman)

E.Y. Andersen, Denmark
B. Bakht, Canada
R. Favre, Switzerland
B. Godart, France
B. Hillemeier, Germany
A.G. Lichtenstein, USA
H.O. Madsen, Denmark
A.S. Nowak, USA
A. Pakvor, Yugoslavia
A.C.W.M. Vrouwenvelder, Netherlands
E. Watanabe, Japan

Organizing Committee (All Danish)

P. Clausen (Chairman)
E.Y. Andersen
P. Kristensen
L. Lovgren
J.E. Nielsen
J. Skov
0. Olsen
J. Vejlby Thomsen (Secretary)

Sponsored by:

Carl Bro Group A/S

COWIconsult
Ramboll, Hannemann & Hojlund A/S
 VII

List of Authors Liste des auteurs Autorenverzeichnis

Abdunur C. 353 Halchini C. 283-443

AbeM. 369 Harre W. 385
Allen D.E. 77 Hassan M. 319
Helou A. 127
Hirose T. 301
Hirt M.A. 53
Hohberg J. M. 157
Bakht B. 261 Horiguchi T 369
Balendra T. 149
Baumgärtner W. 427
Beranger M. 417
Billington C.J. 291
Blaauwendraad J. 185 Ichinose L.H. 207
Bolt H.M. 291 Ishida M. 309
Borst R. de 185 Ishihara Y. 207
Brühwiler E. 375
Bürdet 0. 319

Jacob B. 37
Jaeger L.G. 261
Casas J.R. 241 Jensen B.L. 335
Cervenka V. 165 Johannesen J.M. 19
Chan G.K. 149 Jokubaitis V. 141
Ciampoli M. 69 Juhäs P. 327
Cullimore M.S.G. 197 Juhäsovä E. 327

Diamantidis D. 417 Kajitani Y. 249

Ditlevsen 0. 19 Kamata N. 93
Kamei M. 207
Kaminskas K.A. 141
Karlsson M. 19
Kawai Y. 215
Faber M.H. 61 Kemp A. R. 105
Favre R. 319 Kendall M.H. 393
Filippou F.C. 173 Koshiba A. 369
Fujimoto H. 93 Kroon I. 61
Fujiwara M. 301 Kunz P. 53-375
Furuta H. 341
Furuya T. 93

G Labbe* J. 443
Lao Y. 223
Gao L. 233 Lin M.S. 149
VIII

M Sobrino J.A. 241

Sokolfk A. 85
Manfredini G.M. 417 Somerville G. 29
Margoldovä J. 165 Spacone E. 173
Mason P.J. 197 Sugidate M. 369
Massicotte B. 443 Sorensen J.D. 61
Miesseler H.-J. 435
Mine Y. 341
Mistry Z. 361
Miyamoto A. 249-309
Molins C. 401 Takeuchi K. 309
Morikawa H. 249-309 Talbot M. 283
Murakoshi J. 301 Taucer F.F. 173
Toorn A. van der 119

Nakade 0. 207
Nakamura S. 215 Veen C. van der 185
Nanjyo A. 341 Vrouwenvelder T. 5
Napoli P. 69
Natuaki Y. 93
Nielsen S.G. 335 W
Nishikawa K. 301
Nishimura N. 341 Ward J.K. 291
Noordlander K. 409 Watanabe E. 207
Nowak A.S. 275 Waubke H. 427
Wolff R. 435
WuS. 223

Pei Y. 223 Y
Pukelis P. 141
Pukl R. 165 Yamao T. 215

Ranganathan R. 45-361 Zhou X. 133

Ravi G. 45
Roca P. 401

Sakimoto T. 215
Savard M. 283
Scordelis A.C. 401
Shen D. 223
Shimamura M. 369
Sluszka M. 393
Smith J.W. 197
 IX

Table of Contents • Table des matteres • Inhaltsverzeichnis

Preface III
Pröface IV
Vorwort V

Committees-Comitäs-Komitees VI
List of Authors-Liste des auteurs-AutorenVerzeichnis VII
Table of Contents-Table des matieres-lnhaltsverzeichnis IX

Session 1:
Probabilistic Concepts in Structural Evaluation
Concepts probabilistes dans l'6valuation des structures
Wahrscheinlichkeitskonzepte für die Tragwerksbeurteilung 1

Keynote Speaker 3

T. VROUWENVELDER
Codes of Practice for the Assessment of Existing Structures
Normes pour l'övaluation de structures existantes
Normen für die Beurteilung existierender Bauwerke 5

Selected Papers 17

M. KARLSSON, J.M. JOHANNESEN and 0. DITLEVESEN

Reliability Analysis of an Existing Bridge
Analyse de la fiabilite" d'un pont existant
Zuverlässigkeitsanalyse einer existierenden Brücke 19

G. SOMERVILLE
Residual Service Life of Concreto Structures
Vie rgsiduelle des structures en bäton arm6
Restnutzungsdauer von Betontragwerken 29

B.JACOB
Definition of Load Spectra
Definition de spectres de Charge
Definition von Einwirkungsspektren 37

G. RAVI and R. RANGANATHAN

Fatigue Reliability Analysis of Early Steel Railway Bridges in India
Söcurite* a la fatigue d'anciens ponts-rails mötalliques en Inde
Ermüdungssicherheit früher Eisenbahn-Stahlbrücken in Indien 45
P. KUNZ and M.A. HIRT
Reliability Analysis of Steel Railway Bridges under Fatigue Loading
Analyse de fiabilite* des ponts raii en acier sous charges de fatigue
Zuverlässigkeitsanalyse von Bahnbrücken aus Stahl unter Ermüdungslasten 53

M.H. FABER, J.D. S0RENSEN and I. KROON

Optimal Fatigue Testing - a Reassessment Tool
Essais de fatigue optimalisös - une nouvelle estimation
Optimale Ermüdungsversuche als Hilfsmittel zur Überprüfung 61

M. CIAMPOLI and P. NAPOLI

Upgrading Reliability Assessment of Degraded Structures
Meilleure Evaluation de la fiabilite' de structures endommagöes
Neue Sicherheitsbewertung geschädigter Tragwerke 69

D.E. ALLEN
Safety Criteria for the Evaluation of Existing Structures
Criteres de söcuritö pour I'Evaluation de structures existantes
Sicherheitskriterien für die Bewertung bestehender Tragwerke 77

A. SOKOLfK
Experimental Investigation of Traffic Load on Highway Bridges
Recherche expErimentale de l'influence de la Charge de trafic sur les ponts routiers
Experimentelle Untersuchungen der Verkehrslast auf Strassenbrücken 85

T. FURUYA, N. KAMATA, H. FUJIMOTO and Y. NATUAKI

Remaining Structural Capacity of Brick-built Piers and Abutments
CapacitE portante rEsiduelle des piles et des culEes en maconnerie
Resttragfähigkeit von Pfeilern und Wiederlagern aus dauerhaftem Mauerwerk 93

Session 2:
Analytical Evaluation of Structures
Analyse de la fiabilite des structures
Analytische Beurteilung von Trag werken 101

Keynote Speaker 103

A.R. KEMP
Evaluation of Reserve Capacity of Frames
Determination de la rEserve de capacitE portante des cadres
Bestimmung der Tragreserve von Rahmen 105

Selected Papers 117

A. VAN DER TOORN

Is there still Life after the Lifetime of Sheetpiling?
Est-il possible de prolonger la durEe de vie des rideaux de palplanches?
Gibt es lebensverlängernde Massnahmen für Spundwände? 119

A. HELOU
Retrieval of System Properties of Existing Structures
Determination des caractEristiques de systemes de structures existantes
Ermittlung der Systemeigenschaften bestehender Tragwerke 127
 XI

XINGGANG ZHOU
Strength Evaluation of Existing Masonry Structures
Evaluation de la rEsistance de constructions en briques
Festigkeitsermittlung für bestehende Mauerwerksbauten 133

P. PUKELIS, V. JOKUBAITIS and K.A. KAMINSKAS

Stress Assessment of Reinforced Concreto Structures with Cracks
Estimation des contraintes des constructions en bEton armE fissurEes
Spannungermittlung in gerissenen Stahlbetontrag werken 141

G.K. CHAN, M.S. LIN and T. BALENDRA

Detection of Local Stiffness Changes of Buildings
Mesure des variations locales de rigiditE dans des bätiments
Messung örtlicher Steifigkeitsänderungen in Hochbauten 149

J.-M. HOHBERG
Planes of Weakness in Finite Element Analysis
Surfaces de rupture en analyses aux EIEments finis
Versagensflächen in Finite-Element-Berechnungen 157

J. MARGOLDOVÄ, R. PUKL and V. CERVENKA

Remaining Structural Capacity of Power Plant Frame
CapacitE portante rEsiduelle d'une fondation sous portiques de turbines
Resttragfähigkeit eines Turbinen-Rahmenfundaments 165

F.C. FILIPPOU, F.F. TAUCER and E. SPACONE

Analytical Models for Strength and Stiffness Evaluation of Concreto Structures
Modöles pour l'Evaluation de la rEsistance et la rigiditE des structures en bEton
Modelle für die Berechnung von Tragfähigkeit und Steifigkeit von Betontragwerken 173

Session 3:
Analytical Evaluation of bridges
Analyse de la fiabilitE des ponts
Analytische Beurteilung von Brücken 181

Keynote Speaker 183

R. DE BORST, C. VAN DER VEEN and J. BLAAUWENDRAAD

Finite Element Prediction of a Destructive Field Test of a Bridge
PrEdiction de la Charge de rupture d'un pont au moyen des EIEments finis
Vorausberechnung einer Betonbrücke im zerstörenden Feldversuch mittels Finiter Elemente 185

Selected Papers 19 5

M.S.G. CULLIMORE, P.J. MASON and J.W. SMITH

Analytical Modelling for Fatigue Assessment of the Clifton Suspension Bridge
Modele analytique d'Evaluation de la fatigue du pont suspendu de Clifton
Analytisches Modell zum Ermüdungszustand der Clifton-Hängebrücke 197

E. WATANABE, M. KAMEI, Y. ISHIHARA, 0. NAKADE and L.H.ICHINOSE

Long-term Prediction of Behaviour of Cable-Stayed Bridges
PrEdiction du comportement ä long terme de ponts ä haubans
Voraussage des langfristigen Verhaltens von Schrägseilbrücken 207
XII

S. NAKAMURA, T. YAMAO, Y. KAWAI and T. SAKIMOTO

Spatial Behaviour of Plate Girder Bridges under Redecking
Comportement spatial des ponts ä poutres ä äme pleine lors de la rEnovation de la chaussEe
Räumliche Wirkung von Plattenbalken-Brücken bei Fahrbahnerneuerung 215

S. WU, Y. LAO, D. SHEN and Y. PEI

Method for Evaluating the Load Carrying Capacity of Existing Bridges
MEthode d'Evaluation de la capacitE portante des ponts
Eine Methode zur Tragfähigkeitsbewertung bestehender Brücken 223

L. GAO
Damage Assessment and Remaining Life Prediction for Structures
Evaluation des dommages et de la vie restante des structures
Schadensbewertung- und Restlebensdauer-Prognose für Tragwerke 233

J.R. CASAS and J.A. SOBRINO

Probabilistic Response of Reinforced and Prestressed Bridge Cross-Sections
REponse probabiliste de sections de ponts en bEton armE et prEcontraint
Wahrscheinliche Belastungsantwort von vorgespannten Stahlbetonbrückenquerschnitten 241

A. MIYAMOTO, H. MORIKAWA and Y. KAJITANI

Dynamic Behaviour in Existing Concrete Bridges and Damage Detection
Comportement dynamique des ponts en bEton et dEtection des dommages
Dynamisches Verhalten bestehender Betonbrücken zur Schadenerkennung 249

Session 4:
Structural Evaluation by Testing
Contröle par des essais de Charge
Tragwerksbeurteilung durch Versuche 257

Keynote Speaker 259

B. BAKHT and L.G. JAEGER

Analytical and Observed Responses of Steel Girder Bridges
Comportement analytique et observE des ponts ä poutres d'acier
Berechnetes und beobachtetes Verhalten von Stahlträgerbrücken 261

Selected Papers 273

A.S. NOWAK
Reliability-Based Evaluation of Existing Bridges
ApprEciation de la fiabilitE des ponts existants
Zuverlässigkeitsuntersuchung bestehender Brücken 275

M. TALBOT, C. HALCHINI and M. SAVARD

Load Testing and Numerical Modelling of Quebec Bridges
Essais de Charge et modeles numEriques de ponts au QuEbec
Belastungsversuche und ihre numeriche Modellierung bei Brücken Quebecs 283

C.J. BILLINGTON, H.M. BOLT and J.K. WARD

Redundancy and the Reserve and Residual Strength of Frames
Redondance, rEserve de rEsistance et capacitE restante de cadres
Redundanz, Reserve- und Restfestigkeit von Rahmen 291
 XIII

M. FUJIWARA, K. NISHIKAWA, J. MURAKOSHI and T. HIROSE

Stress Histogramme and Fatigue Life Evaluation of Highway Bridges
Histogrammes des contraintes et Evaluation de la durEe de vie a la fatigue des ponts-routes
Spannungshistogramme und Ermüdungslebensdauer von Autobahnbrücken 301

A. MIYAMOTO, H. MORIKAWA, K. TAKEUCHI and M. ISHIDA

Field Testing and Safety Evaluation of Concrete Bridges
Essai sur site et Evaluation de la sEcuritE des ponts en bEton
Feldtest und Sicherheitsbegutachtung von Betonbrücken 309

M. HASSAN, O. BÜRDET and R. FAVRE

Interpretation of 200 Load Tests of Swiss Bridges
Interpretation de 200 essais de Charge de ponts en Suisse
Auswertung von 200 Belastungsversuchen an Schweizer Brücken 319

P. JUHÄS and E. JUHÄSOVÄ

Remaining Carrying Capacity of Temporary Steel Bridges
CapacitE restante des ponts provisoires en acier
Resttragfähigkeit provisorischer Stahlbrücken 327

B.L. JENSEN and S.G. NIELSEN

Load Tests on Acute-Angled Motorway Bridge
Essai de Charge sur un pont d'autoroute biais
Belastungsversuch an einer spitzwinkeligen Autobahnbrücke 335

N. NISHIMURA, H. FURUTA, A. NANJYO and Y. MINE

Field Testing of Existing Bridge on Remaining Load Capacity
Essais de Charge et capacitE restande d'un pont existant
Feldversuch zur Resttragfähigkeit einer bestehenden Brücke 341

Session 5:
Case Histories
Etudes de cas
Fallstudien 349

Selected Papers 351

C. ABDUNUR
Testing and Modelling to Assess the Capacity of Prestressed Bridges
ExpErimentation et modElisation pour Evaluer la capacitE de ponts prEcontraints
Modellierung und Versuche zur Einschätzung der Kapazität vorgespannter Brücken 353

R.RANGANATHAN and Z. MISTRY

Reliability Analysis of a Highway Bridge
Analyse de la fiabilitE d'un pont-route
Zuverlässigkeitsanalyse für eine Autobahnbrücke 361

M. SHIMAMURA, M. SUGIDATE, T. HORIGUCHI, M. ABE and A. KOSHIBA

Remaining Structural Capacities of Steel Railway Bridges
Evaluation de la capacitE restante de ponts de chemin de fer
Verbleibende Tragwerksfestigkeit von Eisenbahnbrücken aus Stahl 369
XIV

E. BRÜHWILER and P. KUNZ

Remaining Service Life of a Riveted Railway Bridge
DurEe de vie restante d'un pont-rail rivetE
Restnutzungsdauer einer genieteten Eisenbahnbrücke 375

W. HARRE
Test Results from Suspension Cables Preloaded in Practice
REsultats d'essais sur des cäbles de Suspension prEchargEs en pratique
Erkenntnisse aus der Prüfung baupraktisch vorbelasteter Brückenseile 385

M. SLUSZKA and M.H. KENDALL

Safe Capacity of Suspension Bridge Cables
SEcuritE de la capacitE portante des cäbles de ponts suspendus
Sichere Traglast von Hängebrückenseilen 393

P. ROCA, A.C. SCORDELIS and C. MOLINS

Nonlinear Analysis of an Existing Prestressed Shell
Analyse non linEaire d'une coque prEcontrainte existante
Nichtlineare Berechnung einer bestehenden vorgespannten Schale 401

K. NOORLANDER
Remaining Strength of Bridges in Rotterdam
CapacitE restante des ponts ä Rotterdam
Resttragfähigkeit von Brücken in Rotterdam 409

M. BERANGER, D. DIAMANTIDIS and G.M. MANFREDINI

Probabilistic Re-Analysis of Existing Offshore Structures
Contröle probabilistique de plates-formes marines existantes
Probabilistische Überprüfung existierender Offshore-Konstruktionen 417

H. WAUBKE and W. BAUMGÄRTNER

Traffic Load Estimation by Long-Term Strain Measurements
Estimation des charges due au trafic au moyen de mesures de contraintes ä long terme
Abschätzung von Verkehrslasten mittels Langzeit-Dehnungsmessungen 427

R. WOLFF and H.-J. MIESSELER

Monitoring of Trial Loadings of Bridges using Optical Fiber Sensors
Surveillance des charges d'Epreuve des ponts ä l'aide de fibres optiques
Probelastungen von Brückenbauwerken überwacht mit optischen Sensoren 435

B. MASSICOTTE, C. HALCHINI and J. LABB£

Evaluation of the Capacity of an Existing Steel Truss Bridge
Evaluation de la capacitE d'un pont existant ä treillis mEtallique
Ermittlung der Traglast einer bestehenden Stahlfachwerkbrücke 443

IABSE Reports
Rapports AIPC
IVBH Reports 451
 SESSION 1

PROBABILISTIC CONCEPTS IN STRUCTURAL EVALUATION

KEYNOTE SPEAKER
J%

Codes of Practice for the Assessm

Normes pour l'övaluation de
Normen für die Beurteilung e

Ton VROUWENVELDER
Prof. of Structural Mechanics
Delft University of Technology
Delft, The Netherlands

* s
V
 CODES OF PRACTICE FOR THE ASSESSMENT OF EXISTING STRUCTURES

1. INTRODUCTION

Most building codes deal with the design of new structures only. In many cases these codes
are even quite explicitly referred to as design guides. The problems which are encountered
when assessing existing structures are often not mentioned. In the historical and social
context this is fully understandable. Especially in the period after the second world war, most
countries experienced an enormous growth in building production of all kinds and there was
a clear need to have this production guided by means of a System of codes.

This period of large building activity seems, at least to some extend, to have come to an
end. On the other hand, there is an increasing number of existing structures that are
questioned with respect to their fitness for use. As a result, assessment of existing structures
is no longer an occasional job for some specialists, but becomes more and more a part of
common engineering practice, which requires guidance in the same way as design.

2. PRESENT SITUATION

to distinguish between
In discussing codification related to exisiting structures, it is useful
code type documents and guidelines. At present, only a few countries have a general
applicable and real code type document for the assessment of existing structures. As far as
the author is aware, only in Czechoslowakia [1] and in The Netherlands [2] such a
document exists. In Canada and USA codes are in preparation [21, 22]. Guidelines exist in
a larger number of countries. Examples are references [3-5].

The typical difference between a code and a guideline is that codes essentially deal with
minimum requirements on the structures, while guidelines primarily provide Information about
how to plan and carry out the assessment activities in a systematic and economic way. A
guideline does not touch the subject of possible reductions in safety targets or load levels,
or gives recommendations only. A code, on the other hand, has to teil explicitly what the
acceptance criteria are. As a rule, it will not inform the engineer what particular inspections
or calculations might be useful for an economic assessment within certain circumstances.
As it is, there seems to be a need for both types of documents.

In the first paragraphs of this section only general documents were discussed. Many
countries may have documents which deal with the assessment of particular aspects or
special catagories of structures, like bridges [6-10], towers [11], seismic aspects [12, 13],
offshore structures, remodelling [14], and so on. These documents are sometimes of a
guideline and sometimes of a code type nature. It is clear that these documents have been
developed from a special need. Many bridges, for instance, are confronted with much
heavier loads than anticipated during design. Both with respect ot ultimate load capacity as
to fatigue resistance, this has caused much concern. Apart from that, the maintenance of
bridges is an activity where budgets are always limited, and priorities have to be set in a very
careful way.

Considering the present State of codification, it is also interesting to have a look at

international codes. In the revision of ISO-2394 [15] on Reliability of Structures some general
Statements will be given, indicating on what type of arguments criteria and rules should be
based. Some of the recommendations in the further sections of this paper will be based on
the present draft of this document.
 T. VROUWENVELDER

In the present draft of Eurocode 1, Basis of Design [16], the Statement is present that
Eurocodes, without appropriate modification, cannot be used for the assessment of existing
structures. It was decided that further guidance would require too much prenormative
research to be ready for the present edition.

In this respect it is also worthwile to point to prenormative research that has been carried
out by CEB [17-19] and that is under progress by the Joint Committee on Structural Safety
[20]. This may raise the expectation that in the next editions of Eurocode and ISO the
Situation will be improved. Also the present paper is intended to stimulate this process.

3. FUNDAMENTAL DIFFERENCES BETWEEN DESIGN AND ASSESSMENT

In the design Situation, the engineer has many degrees of freedom to adapt the structural
dimensions or even the concept of the structural System. This way he may, with relatively
little effort, prove that the structure meets all the design requirements. If necessary he can,
without much additional costs, strengthen a structure which does not fulfil the (calculatory)
requirements.

This Situation, however, changes fundamentally once the structure has been build. Compare
for example the possibility to add a Single reinforcement bar to a concrete beam in the
design stage to the same modification in an existing structure. In the first case the additional
costs are very small, in the second case they may prove to be prohibitive. Furthermore, the
history of a building and all the changes and damages that have occurred, may lead to a
complex and often fuzzy structural System. In addition there may be a great uncertainty with
respect to the geometrical and material properties in the structure. This means that even if
an existing structure meets the requirements, it may be very difficult to prove it. The
conclusion should be that all requirements that are used in the design Situation, and with
good reason, are not automatically applicable in the 'as built' Situation. We will come back
to this in chapter 5.5.

On the other hand, the advantage of existing structures compared to structures to be

designed is the possibility to measure their properties: one can measure the geometrical
dimensions, the material properties, some of the loads and loading parameters, the
structural behaviour, the structural response, it's degree of deterioration, and so on. In
practice these possibilities of course are limited, because of the costs involved. But even
Visual inspection or the Observation that a structure has survived some heavy load situations
without any damage, may help to get a better view on the properties of the structure than
is possible in the design stage.

Finally, in the design stage one has to prove that a structure is fit for a certain intended
period of use, the design working life. This may lead to special and often implicit durability
requirements. In some design concepts even the reference period for the design loads is
related to the design working life. If a designed structure does not meet these requirements
the design should be changed. It will be clear that the same requirements, implicit or explicit,
do not hold for an existing structure: if an existing structure does not fulfil a long term
durability requirement, it does not mean that the structure should be rejected immediately.
One should consider the costs: it might be more economical to leave the structure as it is
and accept a (possible) shorter remaining life time.

Summarizing this section, the following fundamental differences between structures to be

 CODES OF PRACTICE FOR THE ASSESSMENT OF EXISTING STRUCTURES &A

designed and existing structures have been observed:

- the increased cost to strengthen the structure

- the increased difficulty of structural analysis
- the possibility to do measurements and observations
- the possibility to reduce the reference period

What further may be concluded is that the process of assessment the existing structure is
a far more diffuse process than the process of design. The design process is much more
universal, while problems encountered in the appraisal of exising structures seems to be
more of a unique nature. This means that, for the case of assessment, it will be more difficult
to give detailed guidance to the engineer. As an example, consider the following Statement
from the Czech code [1]:

"The extend of tests depends on the type of materials, structural System, execution
method, homogeneity of the material, technical possibilities of sampling and also on
the purpose of the tests."

A Statement of this type certainly may help the engineer to find the right way of thinking, but
it leaves on the other hand many possiblities where he has to find the answers on the basis
of his own judgement. Design guides may be expected to give more specific information.

4. GENERAL ASPECTS OF CODIFICATION

The conclusion of the previous discussion is that a design code can not be used directly for
the assessment of existing structures: some clauses will need modification, some may not
apply at all and a number of additional clauses may prove to be necessary. This seems to
justify the writing of a special assessment code. In order to be complete and operational,
such a code should adress at least the following aspects:

1. Criteria to do an assessment
2. Structural properties and loads
3. Evaluation of inspection results
4. Structural analysis
5. Acceptance criteria

The first item already indicates a typical difference between a design code and a code for
existing structures: every new structure should simply be accounted for in one way or
another but when exactly an existing structure should be assessed and to what detail is
already a difficult matter in itself. The most essential difference, however, between the two
types of codes becomes manifest in the last item: to what degree can it be justified to
release the design requirements and to accept reduced criteria in the case of assessment.
All five items mentioned above will be discussed to some detail in section 5.

Additional to the specification of a set of minimum requirements in a code type document,

there is, as stated in section 2, a need for giving some guidance to the engineer how to
tackle the assessment. Such a guidance will, according to section 3, necessarily be of some
global nature. Most problems related to existing structures are of a rather unique nature and
guidance for that reason cannot be very specific. It will leave more detailed decisions to the
judgement of the engineer. What guidance can be given will be discussed in section 6.
 T. VROUWENVELDER

5. DISCUSSION OF A CODE TYPE DOCUMENT

In the previous section five aspects have been mentioned that should be covered by a code
on existing structures. In this sections these five items will be discussed one by one.

5.1 Criteria to do an assessment

The code should list the various reasons that might exist to Start an assessment procedure
for an existing structure. A distinction should be made between situations where the
assessment is required by the code and situations where the need to assess comes form
other sources, but the code nevertheless applies. The most typical cases where a code
could require an assessment are:

- Routine

A routine assessment can as a rule be relatively simple: the structure is inspected and if the
result is within predefined limits, the assessment is that the structural capacity is still
sufficient. If the result is outside the predefined limits a further investigation might follow, or
it might be decided to repair or maintain the structure immediately.

Nowadays, routine inspections are performed by owners on their own initiative (public
bodies for instance) or on the basis of a requirement by insurance and Classification
companies (offshore platforms). The degree and nature of the inspections might differ
substantionally, varying from looking into appearance aspects only to intensive searches for
fatigue cracks.

A country, having a code on existing structures, could require routine assessments for all
types of buildings, or alternatively for certain classes, for instance public buildings. The
degree in which this is possible, of course, depends on the legal Status of the code.

- Deterioration or (suspected) damage

When there is obvious damage to the structure, for instance corrosion, spalling of concrete,
cracks, leakage, heavy Settlements, and so on, a structural appraisal should be required by
a code. The difficult point of course is to indicate the markation between "innocent
deterioration" and "deterioraton demanding further investigation". A code should provide as
much guidance as possible, but this typically will remain one of the cases where engineering
judgement will be decisive.

The code may also require investigations in the case of "suspected damage". Suppose that
a structure has been loaded by some extreme load (earth quake, fire, tornado, explosion).
In those cases there might be a request for further investigation, even if this particular
structure does not show any visible damage at first sight. The same may hold for other
types of shortcomings, resulting for instance from construction errors, bad workmanship,
unexpected behaviour and so on. In those cases the defects observed on a Single structure
often brings a large group of similar structures under suspicion. The interesting advantage
in such cases is that the available budget for research and analysis is usually relatively high,
enabling the use of advanced engineering tools in the assessment.

Finally, a code should specify that only the parts suffering from real or suspected damage
(deterioration) should be checked. There is no reason to inspect all parts if only some parts
10 CODES OF PRACTICE FOR THE ASSESSMENT OF EXISTING STRUCTURES

or aspects are under suspicion.

- Increased loads/extended life/change of use

Structures loaded by higher loads than anticipated during design or structures being at the
end of their intended service life, should be investigated. This is an actual problem in many
countries as far as bridges are concerned [23, 24]. Many bridges have been designed for
traffic loads that are much smaller than the present day traffic loads. Many engineers have
the feeling that no further investigation is required if the bridge does not give signs of
distress. This is a dangerous way of reasoning. If all structures give proper signs of distress
before collapsing, the argument would be valid, but this certainly is not the case. It should
at least be verified that the structure indeed will show signs of distress in due time.

- Reconstruction

If a structure is to be reconstructed for whatever reason, an assessment should be

mandatory. The mere fact that the building is reconstructed makes it necessary to assess
the present Situation: the dimensions and properties of the present structure should be
known in order to be able to make an analysis of the reconstructed building.

5.2 Structural properties and loads

If the original drawings and specifications still exist and if there is no reason to doubt them,
the code should specify that these can be used to assess the geometrical and material
properties. Deterioration effects, of course, should be taken into account. Reasons for doubt
could be: damage under otherwise normal circumstances, premature aging, bad
Performance of the structure or of similar structures elsewhere. In cases of doubt
measurements are always necessary.

Values for material properties should be taken according to the present day codes. If a
material is no longer used, actual codes may not provide relevant Information. In that case
properties as specified in old codes can be used. Sometimes it will be necessary to make
some transformation, for instance from an allowable stress value to a design value or
characteristic value. For the sake of analysis it may be sometimes worthwile to do additional
tests. For instance in the case that use of plastic properties is made, but for the original
material only the elastic properties are known. In such a case ductility tests may be
necessary.

If no data is available with respect to the grade of the material used, one may do tests, but
it might turn out that it is more economical to take the lowest possible grade, used at the
time of construction. A code could allow such an approach.

Loads should always be taken according to the new codes and according to the Situation
to be expected in the reference time for the assessment. Load reductions may follow from
measuring some load model parameters (see 5.3), from economic criteria (see 5.5), orfrom
a reduced reference period.

5.3 Evaluation of inspection results

Inspections can be of various types:

 T. VROUWENVELDER 11

- Visual
- direct measurement
- non destructive testing
- response measurements
- proof load

The code should specify how data obtained from inspection can be used in the subsequent
analysis. A starting point should be that all available data is of value: all information can lead
to a better estimate of the structural capacity and to a reduction of the uncertainties. In
principle one should combine all information: Visual observations, Performance in the past,
measurements of various kinds and so on. This may require the use of an expert System
[25]. From a theoretical point of view, probabilistic methods [26,27] offer an ideal framework
for such a procedure (see Figure 1). In an operational code these formal procedures should
be translated into operational methods within a load and resistance factor design approach.

Note that the conclusions of an inspection should not only be concerned with the structure
or structural part under consideration. Inspection of one part of a structure always teils
something about other parts which are in similar circumstances. Finding a bad concrete
quality in one column may increase the probability of finding a bad concrete in another one.
In order to use information from tests and measurememts, the accuracy of the inspection
method should be known. This is a weak point in the present State of the art. There is a
great variety of inspection techniques, but only in a limited number of cases (for instance
crack detection in offshore structures) investigations have been done into the so called
probability of detection curves and into the quantification of measurement errors.

Figure 1: Original and updated probability density function for an inspected variable x.

fx(x)
M

updated distribution

initial distribution

Also inspections on the loading side may be of help in the assessment. The research will
depend on the type of loading. For offshore structures one may check the local wave
climate. For wind loading on special shaped structures one may measure the shape
coefficients. For industrial loads measurements may indicate differences from the original
design assumptions. Of course one should be careful: loads in codes are intended to
repesent the maximum in say 50 year values. In general it is of course not possible to
measure this directly.
12 CODES OF PRACTICE FOR THE ASSESSMENT OF EXISTING STRUCTURES

5.4. Structural analysis

The analysis of existing structures may differ form the analysis of a structure under design.
Especially if a structure has been damaged, some of the assumptions which are normally
correct in design might be no longer true and other failure mechanisme might become
important. As an example, consider a Standard steel section. For a new section, the check
on local buckling is normally not necessary, but if dimensions are reduced because of
corrosion this might be decisive. In the case of repair and reconstruction, the Cooperation
between old and new members should be given extra attention. One should keep in mind
that strengthening members in general will only help for the additional load. For füll profit of
the strengthening the ductility requirement for the old structure is more important than
normal. These items require a detailed and material oriented approach.

In cases of damage it is necessary that the damage can be explained by analysis. Only if
the present Status of the structure can be simulated from construction, observed loads and
structural data, further meaningful Steps can be set [28].

If the analysis of an existing structure leads to the conclusion that the structure is not safe
enough, a more advanced analysis may be of help. Sometimes the question is raised wether
this can be justified. The argument is that the addtional safety hidden in normal design
should be considered as an integral part of the required structural safety. This argument has
some truth. If all structures would be analysed on the edge of the theoretical possibilies, the
average safety would probably be lower. On the other hand, a carefull analysis may be
expected to indicate the weak spots in the analysis which are normally overlooked. Anyway,
the code has to come up with some Statement about this matter. It is the authors' conviction
that both in design and in assessment simple and advanced modeis can be used to meet
the safety requirements. The errors made in both modeis should be taken care of by
introducing the proper model uncertainties.

Quite another point is it wether it is allowed to include nonstructural compontens like

separating walls in the analysis. This, to the authors opinion, should be done only with great
care and as a kind of emergency measure.

5.5 Acceptance criteria

Criteria for acceptance of an existing structure should be based on present day codes. The
mere fact that the structure fulfils the code of its time of construction can never be decisive.
Codes have changed and in general for good reasons. This of course does not mean that
a if a new code comes into practice with on some point more severe requirements than the
old one, this should lead to immediate disapproval of all old buildings. Reasons for possible
reductions of the requirments for existing structures will be discussed in this paragraph.

The possible reduction of requirements for existing structures can best be discussed in
terms of a probability based code. In a probability based code the design is based on partial
factors that depend on the degree of uncertainty and on the required life time reliability.
Three items can be taken into consideration:

- the information (inspections, observations) that is available

- the reference period
- the cost benefit ratio of safety measures
 T. VROUWENVELDER 13

If the assessment of an existing structure is based upon the results of extensive

measurements, one may reason that the uncertainty has been reduced (see figure 1). This
reduction in uncertainty may lead to the use of lower partial factors. In fact this does not
mean any reduction of the real safety at all. Note further that measurements not only affect
the partial factor, but also the characteristic value. This may lead to much higher values for
the material properties than those to be assured in design.

A similar reasoning may be set up for the reference period. New structures are designed for
a period of at least 50 years. This long period may lead to a number of requirements which
are useless in the assessment of existing structures. There is no point in rejecting an existing
structure because we do not expect it to last a period of 50 years from now on. Again this
does not really mean a reduction in reliability.

Finally there is the item of cost. It has already been explained that the cost of improving a
new structure is generally much lower than the costs involved in upgrading an existing one.
This means that if the upgrading of an old structure is considered to be uneconomical, this
need not be the case for a new one and vice versa. This argument may lead to a difference
in requirements between new and existing structures. Of course, if the safety of human lifes
is at stake, some care must be taken with this argument. We will not go into the details, but
for instance see [29].

There are many examples from practice where the above type of reasoning has been
followed: in many guides for existing structures reduced requirements have been introduced,
for instance for fire, earth quake, and so on.

A final fundamental question is: what is an existing structure? The only logical answer can
be that a structure should be regarded and judged as an existing structure, the very day that
it has been erected. The criteria might even be applied for those parts that have been
constructed, even if the total building is not yet completed. In combination with the proposed
reduction in the requirements, this might lead to the conclusion that the contractor of a new
building can produce a reduced quality without punishment. This of course cannot be true:
the contractor is obliged by a contract to deliver the design quality. But if he fails, it might
be better that he pays a fine than that he Starts to make repairs.

6. GUIDANCE TO THE DECISION PROCESS

In section 5 of this paper the code type assessment requirements have been adressed.
More than in normal cases, however, the engineer also wants guidance in the way of
attacking the problems, in the decisions he has to make, even if it is of a very global nature.
Typically two types of decisions have to be made by the engineer:

-one with respect to the depth of the assessment itself

-one with respect to the measures to be taken

In both cases the engineer should look for the most economical Solution, keeping in mind
that the total costs include assessment costs, costs of measures and the costs or benefits
in the subsequent exploration of the building.

Normally it makes sense to Start with a global assessment. Refinements should be done
only if (1) the global assessment leads to unfavourable results and (2) the costs of the
14 CODES OF PRACTICE FOR THE ASSESSMENT OF EXISTING STRUCTURES

refinement are not prohibitive compared to the expected profits. Note that the expected
profits depend on the probability that the outcome of the refined assessment is positive.

In the choice of possible measures, given a certain amount of information, the time aspect
is very important. In general, the decison to build a new structure will lead to higher direct
costs than a repair decision. In those cases, however, one should also look for the long
term costs: a new building may be good for 50 years while the repair may only only solve
the problem for about 10 years.

In giving guidance to these problems, a distinction should be made between the "individual
assessment" and a "long term inspection and maintenance program".

The individual assessment occurs in case of damage, remodelling, and so on. A typical
guidance for such cases is often provided in the form of a flowchart for activities [3, 5].
Figure 2 gives an example. The Chart is based on the principles stated above.

In the long term inspection and maintenance planning one formulates the criteria for repair,
maintenance, inspection periods, etc in advance. The criteria should be formulated in such
a way that the total cost expectation is a minimum. To get a grip on this problem, a
presentation on the basis of an event tree could be of help, see figure 3. Of course, for a
quantitative optimization, the costs of inspections, repair and failures should be known, as
well as the probability of failure between one inspection and the next one. Of course, in most
cases the costs, let alone the failure probabilities, are only vagely known. Nevertheless a
clear picture of what the ideal decision criteria are can help to improve the decision quality,
even in the absence of exact data.

Figure 2: Simple flow Chart for decision making in the case of individual assessment

global inspection
assemble documents
> f and other information
direct measures

\f
explain and assess
the condition of the ^
structure

> <

more information needed?

yes
advanced analysis needed?

> rno
decisions:
normal use
reduced use
rebuild
repair
 T. VROUWENVELDER 15

Figure 3: Event tree for a structure subject to a long term inspection and repair
program; after every inspection the structure may be o.k. or need repair;
during the inspection intervals the structure may fail

At At At
¦*-
failure
failure
repair repair
0
t
Start
no failure >
no failure t i
o.k.
failure
o.k. repair
i^ no failure ti
o.k.
inspection
> k

inspection

7. CONCLUSIONS

The codification of assessment procedures for existing structures is a relatively new field of
development. Although an increasing number of codes and guidelines is published, the
feeling in most countries is that making a code for existing stuctures still requires additional
prenormative research. To stimulate this research this paper has tried to sum up the basic
items that should be adressed by such a code. These items are:

1. Criteria to do an assessment
2. Structural properties and loads
3. Evaluation of inspection results
4. Structural analysis
5. Acceptance criteria

The most difficult items probably are the incorporation of inspection results into the total
analysis and the possible reductions in Performance requirements. Rational ways to deal
with these problems seems possible only within the framework of probability based methods.
Of course, for the application in every day practice, the results of those methods need to
be translated into Standard load and resistance factor procedures.

In addition to the typical code type items mentioned above, an engineer needs some
guidance to judge in what circumstances what inspections or what measures should be
chosen. Here a cost optimization approach should be followed. This requires a guideline
type of document, telling the engineer what decisions are the most economical, given the
requirements in the code and the costs of the various alternatives.

REFERENCES

1. CSN 73-0083 1986, Czechoslovak Code for the Design and Assessment of Building Structures
16 CODES OF PRACTICE FOR THE ASSESSMENT OF EXISTING STRUCTURES

Subjected to Reconstruction. (English Translation available), Prague.

2. Regeling Bouwbesluit Constructieve Veiligheid. (Legal Rules for Structural Safety) Minlstry of Houslng
and Environment. The Netherlands, 1992.
3. Structural appraisal of existing buildings for change of use, Building Research Establishment. United
Kingdom, 1991.
4. ISE 1980, Appraisal of Existing Structures. The Institution of Structural Engineers, London, 1980.
5. ASCE 11 -90, Guideline for Structural Condition Assessment of Existing Buildings.
6. CSA S6-1990, Supplement No. 1 (Existing Bridge Evaluation) to CSA Standard CAN/CSA-S6-88: Design
of Highway Bridges. Canadian Standards Association, Rexdale, January 1990.
7. AASHTO, Manual for Maintenance Inspection and Bridges, American Association of State Highway and
Transportation Officials, Washington D.C., 1983.
8. OHBDC 1983, Ontario Highway Bridge Design Code: Clause 14-Evaluation of Existing Bridges. Ontario
Ministry of Transportation, Downsview, Ontario.
9. AASHTO, Guide Specifications for Strength Evaluation of Existing Steel and Concrete Bridges,
American Association of State Highway and Transportation Officials, Washington D.C., 1983.
10. AASHTO, Manual for Maintenance Inspection of Bridges (draft), American Association of State Highway
and Transportation Officials, Washington D.C., 1991.
11. CSA S37 1990, Antennas, Towers and Antenna-Supporting Structures, New Clause 3.2: Evaluation of
Existing Towers. Canadian Standards Association, Rexdale Ontario.
12. FEMA, U.S. Handbook of Seismic Evaluation of Exisiting Buildings (Preliminary). FEMA-178, Federal
Emergency Management Agency, Washington, D.C., 1989.
13. NZNSEE, Earthquake Risk Buildings: Recommendations and Guidelines for Classifying, Interim
Securing and Strengthening. New Zealand National Society for Earthquake Engineering, Wellington,
New Zealand, 1985.
14. CIRIA, Construction Industry Research and Information Association, Structural renovation of traditional
buildings. Report 111, CIRIA, 1986.
15. ISO, Reliability of Structures, IS 2394, 7th Draft, 1992.
16. CEN, Eurocode 1, Basis of Design and Actions, draft, 1992.
17. CEB, Assessment of Concrete Structures and Design Procedures for Upgrading. Comitö Euro-
International du Beton. Bulletin d'information No. 162, Lausanne, 1982.
18. CEB, Assessment of Concrete Structures and Design Procedures for Upgrading. Comit6 Euro-
International du B&on. Bulletin d'information No. 163, Lausanne, 1982.
19. CEB, Diagnosis and Assessment of Concrete Structures. Comite Euro-International du B6ton. Bulletin
d'information No. 192, Lausanna, 1989.
20. JCSS, Project on Existing Structures, Discussion Note, 1991.
21. ALLEN D.E., Criteria for Structural Evaluation and Upgrading of Existing Buildings. Institute for Research
in Construction, National Research Council Canada, Ottawa, Ontario. Can. T. of Civil Engineering,
December 1991.
22. TURKSTRA C.J., Evaluation of Existing Buildings and Design by Testing. A Review of Related North
American Codes Recommendations and Research Work. Department of Civil Engineering, Polytechnic
University, Brooklyn, New York, April 1990.
23. MOSES F. and VERMA D., Load Capacity Evaluation of Existing Bridges. Report No. 301, National
Cooperative Highway Research Program, Transportation Research Board, Washington, DC, 1987.
24. ALLEN D.E., Reliability Index for Bridge Evaluation. Institute for Research in Construction, National
Research Council Canada, Ottawa, Ontario.
25. ZHANG X.J. and YAO J.T.P., The Development of Speril Expert Systems for Damage Assessment.
Structural Engineering, Ce-STR-86-29, 1983.
26. MADSEN H.O., Model Updating in Reliability Theory. Proceedings, ICASP5, Vancouver, B.C., May 1987.
27. DIAMANTIDIS D., Reliability assessment of existing structures. A.S. Veritas Research, Hovik, Norway,
September 1986.
28. STEPHENS J.E. and YAO J.T.P., Survey of Available Structural Response Data for Damage
Assessment. Structural Engineering, CE-STR-83-23.
29. DITLEVSEN O., Decision Criteria in Reliability Updating of Existing Structures. Appendix B, Structural
Reliability of Existing Bridges, Report no. 1, The Road Directorate, Ministry of Transport, Denmark,
1986.
 17

SELECTED PAPERS
18
 19

Reliability Analysis of an Existing Bridge

Analyse de la fiabilitö d'un pont existant
Zuverlässigkeitsanalyse einer existierenden Brücke

Max KARLSSON Johannes M. JOHANNESEN Ove DITLEVSEN

Civil Eng. Civil Eng. Prof. Dr.
AEC Consult. Ltd. A/S AEC Consult. Ltd. A/S Techn. Univ. of Denmark
Vedbaek, Denmark Vedbaek, Denmark Lyngby, Denmark

Max Karlsson has worked with
practical applications of structural
reliability since 1980. He is a
member of the permanent
Committee of the Danish Society of
Engineers on Actions on Structures
and he is chairman of the
working group of the Danish
Society of Engineers on Structures in
Glass.
Johannes Morsing Johannesen got
his Master's Degree at the Technical
University of Denmark in
1990 and is currently PhD
Student. He is studying methods of
reliability analysis of damaged
structures.
Ove Ditlevsen has contributed to
the development of the theory of
structural reliability since 1961
with a large number of publica-
tions dealing with a large variety
of topics. Ove Ditlevsen is a member
of the Joint Committee on
Structural safety and of CIB, W81:
Actions on Structures.

SUMMARY
The assessment is described of the remaining structural capacity of an existing concrete bridge. A
probabilistic reliability analysis is applied to a simple conventional carrying capacity model for the bridge.
This simplified reliability analysis is calibrated by a random effectivity factor to give realistic results. The
calibration uses some particularly chosen deterministic analyses of the bridge. These analyses are based
on a refined FEM-model of the failure behaviour taking into account that the observed strength throughout
the structure differs from what was assumed at the design stage. The cases for deterministic analysis are
obtained through the reliability analyses of the simple model.

r£sum£
L'article traite de 1'Evaluation de la Resistance restante d'un pont en bEton armE. L'analyse probabiliste
de la fiabilite* du pont est rEalisE sur la base d'un modele simple de la rEsistance ultime du pont. Cette
analyse simplifiEe de la fiabilite* est calibrEe au moyen d'un facteur d'efficacitE pour obtenir des rEsultats
exacts et r6alist.es. Le calibrage utilise des rEsultats de certaines analyses dEterministes des structures du
pont. Ces analyses ont EtE faites en utilissant un modele trEs dEtaillE, par EIEments finis du comportement
du pont en tenant compte que la rEsistance observEe en certaines parties du pont est diffErente de Celles
supposEes lors de I'Etablissement du projet. L'analyse dEterministe est Etablie sur la base de l'analyse de
la fiabilitE du modele simplifiE.

ZUSAMMENFASSUNG
Die Bewertung der Resttragfähigkeit einer Betonbrücke wird beschrieben. Eine Wahrscheinlichkeitsanalyse
der Sicherheit ist auf der Grundlage eines einfachen Standardmodells der Tragfähigkeit der Brücke
durchgeführt. Diese einfache Wahrscheinlichkeitsanalyse ist mit einem Effektivitätsfaktor kalibriert, um
ein realistiches Ergebnis zu erreichen. Die Kalibrierung nutzt speziell ausgewählte deterministiche Analysen
der Brücke. Diese basieren auf einem verfeinerten FEM-Modell bezüglich der Brückenkonstruktion unter
Berücksichtigung von Festigkeitsabweichungen gegenüber den Bemessungsannahmen. Die betreffenden
deterministischen Untersuchungen werden aufgrund des einfachen Zuverlässigkeitsmodells bestimmt.
20 RELIABILITY ANALYSIS OF AN EXISTING BRIDGE

1. Introduction

Well—developed rational reliability based methods for designing new concrete bridges are
available today. However, for a number of reasons this is not the case concerning the assessment
of the remaining structural capacity of an existing and deteriorated bridge.
A reliability analysis of a bridge in the design State is a formal procedure based on common
practice. The modeis have to a certain extent become Standard so that the target safety levels
together with associated and selected failure modes give structural dimensions which are
known to be satisfactory for normal structures. Moreover, the analytical modeis used in design
are practically manageable in size and complexity and they are assumed to model the structural
carrying properties of the bridge in a sufficiently realistic way.
A similar Standard procedure for analysing existing bridge structures has not yet been
developed. When considering an existing bridge the reliability analysis is no longer just a
formal procedure. The potential failure modes have to be modelled realistically taking into
aecount the available knowledge on geometry, strengths, etc. This raises the problem of how
to set up such probabilistic modeis that are sufficiently rieh in concepts to take the available
information into aecount and at the same time can be standardized to an extent that makes
the reliability analysis result comparable with the result from a similar reliability analysis of
another existing bridge or with speeified target safety levels.
Another problem is that the reliability analysis of an existing bridge due to observed dete-
riorations, errors etc. often becomes more complicated than the analysis during the design
State. However, such information must be considered seriously and it makes it more difficult
to set up modeis that are practically manageable.
In the following a method is demonstrated by which these problems can be overcome for
concrete bridges suffering from severe damages.

2. The considered existing bridge

The bridge across Salpetermosevej in Hiller0d, Denmark, was construeted in 1977. It is designed
as a reinforced concrete frame structure. The length of the free span is approximately 6 m.
The concrete used for casting the bridge was supplied by a local plant for ready—mixed
concrete and delivered by truck mixers. The workability of the concrete was very poor and too
stiff for the contractor to obtain a satisfactory compaction. Thus, the hardened concrete
obtained gross porosity, showing high intensity of honeycomb at the finished concrete surface
and a high content of entrapped air in the interior of the structure. Furthermore, the fresh
concrete even contained fractions of hardened concrete due to insufficient cleaning of the
mixer. The compressive strength of the concrete was determined by cast cylinders. The test
results indicated that the potential compressive strength of the concrete in the structure
would be lower than prescribed, mainly due to large variability. An investigation made in
1977 with tests on drilled cores from the bridge verified this suspicion.
 M. KARLSSON - J.M. JOHANNESEN - 0. DITLEVSEN 21

The present appearance of the bridge shows concrete which is seriously disintegrated by
cracks and other signs of deterioration especially in the bridge deck. Due to the extensive
porosity of the concrete the influence of aggressive substances from the environment is signifi-
cant. The carbonation, the Chloride ingress (de—icing salt) and the leaching by rainwater
seeping through the concrete have been the dominating environmental actions on the concrete
bridge. The effect of these attacks is a decrease of the compressive strength and a loss of
protection against corrosion of the rebars in various parts of the structure. In this paper we
will be content with a study of the reliability analysis of the deteriorated bridge deck.

3. The effectivity factor reliability analysis method

The probabilistic reliability analysis of the bridge deck can be made on the basis of a simple
conventional upper bound yield line collapse model. In the following we will denote this model
as the simple model. In order to make the results of the analysis "realistic" an effectivity
factor reduction is introduced on some of the variables from which the yield moments are
calculated. The effectivity factor is calculated from one or more carrying capacity results that
correspond to certain optimized statically admissible stress fields. These statically admissible
stress fields are obtained in afinite element model that represents a refined model of the local
strength properties of the bridge deck taking into aecount that the yield moments vary over
the deck according to some stochastic field model. This stochastic field model reflects the
observed deteriorations of the bridge deck. Moreover, by using the lower bound theorem of the
ideal plasticity theory searching an optimal admissible stress field it is automatically ensured
that the reliability is obtained for the most critical failure mode. This finite element model
will in the following be denoted as the elaborate model.
The inverted commas around the word realistic are put there because the ideal plasticity
theory is not necessarily particularly realistic. However, the ideal plasticity theory is used
herein in order to illustrate that a simple model by use of an effectivity factor modification
can be made reliability equivalent to a far more elaborate model of the same phenomenon.
The effectivity factor is obtained by a Single or some few calculations with the elaborate
model. The set of input values for these calculations with the elaborate model is obtained by a
reliability analysis carried out by use of the simple model.
In this way an elaborate (and possibly realistic) model for the carrying properties of a
structure can be reliability analysed by use of a suitably calibrated simple conventional carrying
capacity model. This reliability equivalence may be the key to a rational codification of
methods to evaluate remaining structural capacity. The theoretical considerations leading to
the method are given in Ditlevsen and Arnbjerg-Nielsen (1992,1992). Here the method will be
summarized without the argumentations for the validity of the method.
Let (x^XpwXpJ be the total vector of basic variables (input variables) that are contained
in the elaborate model. The subvectors Xq and xR are the vectors of load variables and
strength variables respectively. These variables are with sufficient generality defined such that
22 RELIABILITY ANALYSIS OF AN EXISTING BRIDGE

they all have physical units that are proportional to the unit of force. The subvectors x^ is
the vector of all the remaining basic variables (of type as geometrical and dimensionless basic
variables).
Two limit State equations

gr(xs,xR,xD) 0 gi(xs,xR,xD) 0 (1)

are given representing the "realistic" model and the idealized model respectively. It is assum-
ed that for each fixed (xq,Xp,Xß) the equations

SAXS'XR'XD) ° ' Si(*ixS'xR'xD) ° (2)

can be solved uniquely with respect to n respectively. The Solutions are

and k.
k (xq,xR,Xpv) and K.(xq,Xo,X|0. By using the physical property of dimension homogeneity it
can be shown (Ditlevsen and Arnbjerg—Nielsen 1993) that the two equations

gr(xs,xR,xD) 0 gi(xS^(xgix^Xp)xR'xD) ° (3)

are equivalent in the sense that the two set of points they define are identical. The idea of the
effectivity factor method is to use a suitable simple approximation to the last equation in (3)
in the reliability analysis in place of the first equation in (3). The point is to approximate the
effectivity factor function ^(xq,Xo,XjO k (xq,Xp,xT^)//c.(xCj,Xo,xT.) by a constant or at
most an inhomogeneous linear function of (xq,Xrj,XjA The approximation is made such that
it is particularly good within the region of the Space that contributes the most to the failure
probability. Let (x£,x£,x£) be a point of this region and let v* ^(x£,x£,xß) The equation

g.(xg,i/*xR,xD) 0 (4)

then defines an approximating limit State in the important region. The problem is now
reduced to the problem of how to choose the point of approximation (x£,x£,x£) The answer
to this problem is given in the reliability theory. With a judgmentally chosen value v~ of i/*
a first or second order reliability analysis (FORM or SORM, see e.g. Madsen, Krenk, and
Lind (1986) or Ditlevsen and Madsen (1991)) is made with (4) as limit State. This analysis
determines the most central point (the design point) (xq-pX^pX^-,) and an approximate
failure probability p, Using that ^(xqpXrjpXy^-,) l/i/ft an improved value u, pc\Kr\
of v* is calculated where n Kr(xQpxppxr)i)• Then a new FORM or SORM analysis is
made with (4) as limit State. This gives the most central point
(xq^Xp^XT^o) and the
approximate failure probability p2 Proceeding iteratively in this way we get a sequence
(«ri jPi),(Kr2>Po)v tlmt; may or may not be convergent. If the sequence is convergent in the
 M. KARLSSON - J.M. JOHANNESEN - 0. DITLEVSEN 23

first component it is also convergent in the second component and we have

k,,^,.. -> 1
PpP2,.. -> p where p will be denoted as the zero order approximation to the probability of
the failure event of the elaborate model.
If the sequence is not convergent we still can define the zero order approximation by simple

interpolation to the value k 1

among points («,/?) (ß=—$~ (p) $ standardized
normal distribution function) corresponding to the sequence or simply obtained for a series of
different values of v~
A check of the goodness of the zero order approximation is made by replacing the effectivity
factor function j/(xq,XTpXT)) by its first order Taylor expansion

*/(xg,xR,xD) ^(xs,xR,xD) ~i/* + a'(xg-x*) + b'(xR-x£) + c'(xD-x£) (5)

at the most central point (x£,x£,x£) corresponding to the limit State (4) with iA being the
effectivity factor value corresponding to k 1 The numerical determination of the coeffi-
cients a, b, c requires that the values of z/(xq,XjpXjO are known at least at as many points
in the vicinity of (x£,xÄ,x£) as the number of variables in (xq,xR,Xpx) These values of v
are obtained by solving the equations (2) with respect to «r and ^ respectively at each
chosen point (xq,Xrj,x^)
With (5) substituted for k /k. into the last equation in (3) we get a limit State for which
both the probability of failure and the value of k in general will be different from the
probability p and the value k obtained by the zero order approximation. However, by
1 as
a unique scaling factor k on the load vector Xq we can achieve that the limit State defined
by g.(k Xq, v(k Xq,xR,xD)xR,xD) 0 corresponds to the failure probability p With the
Taylor expansion (5) substituted into this equation we get the limit State equation g.(krXg,
z/(kXq,Xrj,x^))xo,Xp)) 0 for which we can determine k by iterative application of FORM

or SORM analysis such that the corresponding failure probability becomes p The pair
(k ,p) will be called the first order approximation. The size of the deviation of kf from 1
can then be used to judge the accuracy of the zero order approximation. Also the change of
the most central point contributes to this judgment.

4. Reliability analysis of deteriorated concrete bridge

The slab structure of the Salpetermosevej bridge is shown in Figure 1. The slab is one span
and clamped in both ends. Actually the slab is skew with skew reinforcement, but the skew-
ness is relatively small — the angle between a free edge and a clamped edge is 82°. Finite
element calculations verify that assuming orthogonal reinforcement and a rectangular slab
shape gives a small relative error on the load carrying capacity. Only one load case with fixed
load is considered in the present study. According to the rules for loads on Danish road
bridges, Vejdirektoratet (Danish Road Directorate) (1984), the critical truck load on the
24 RELIABILITY ANALYSIS OF AN EXISTING BRIDGE

undamaged slab structure is found to consist of two trucks as shown in Figure 1. A uniformly
distributed load is also prescribed but is found to be negligible in this case. For reference later
(when reliability index versus load parameter curves are found), it is mentioned here that the
load parameter value corresponds to the prescribed characteristic load including corrections
for dynamic loading. In order to reduce the computational efforts in this illustration the
analysis is made solely on one half part of the slab structure utilizing the geometrical and
loading symmetry, Figure 1 This is made possible by prescribing the torsional moment to be
zero along the symmetry line in the finite element model. (From a stochastic modelling point
of view this symmetrization is not necessarily correct).

Geometrv Loading
Tiuck 1 0 130 MN per wheel
SHKWXKM^
Free edge
^ 3 0
Truck 2 0 065 MN per wheel

1 5 <L
1 5
3 0 B t?
Clamped edge 1 5

3 6 / 5 7
14 3
o o o
Wheel load Truck 1
Truck 2

Figure 1. The slab structure ofthe Salpetermosevej bridge. Length unüs [m].

In the elaborate model a lower bound Solution is used. The analysis method is based on the
lower bound theorem, which states that stress fields in equilibrium not violating the yield
criterion are admissible Solutions. The Solution method is to find the stress distribution that
maximizes the load obtained by proportional loading. Polygonization of the yield condition
leads to a linear programming problem. A stress based finite element code is used, H0yer
(1989). The FE code is described shortly in the following. The stress State is given by a set of
stress parameters that always satisfy the local equilibrium conditions in an element, here a
triangulär 3—noded element. Global equilibrium is obtained in the System of nodal forces that
correspond to the stress parameters and which are in equilibrium with the external nodal
forces. The polygonized yield condition is checked at each node. The polygonized yield
condition is given as

m
xy
± min{mF -m -mi +m mp -m -m' +m } (6)

in which m 6 [—mp ,mp ] is the moment per length unit in a cross section perpendicular to
x
xx
the x—axis and corresponding to compression in the upper side, m ] is defined
y
e [—mp ,mp
y y
analogously, while m is the torsional moment per length unit. Lower index F indicates
xy
yield capacity (absolute value) and prime indicates compression in the lower side of the slab.
It is noted that the polygonized yield surface (6) is inside the yield surface defined by
 M. KARLSSON - J.M. JOHANNE§EN - 0. DITLEVSEN 25

m min{(mp —m )(mp —m (mp +m )(mp +m )} the latter being the Standard yield
xx y^ xx y
surface for reinforced concrete slabs. Thus the polygonized yield surface leads to a lower
bound Solution as compared to the usual Solution.
The simple model is based on the upper bound theorem in the theory of plasticity for ideal
plastic materials. The work equation method is used, e.g. Nielsen (1984). A simple expression
for the load carrying capacity of the undamaged homogeneous slab structure is set up as
follows. The yield line pattern is shown in Figure 2. The fixed length of the positive yield line
in the middle of the slab and the assumptions d/x < 1/2 and x < d+b™ are found to be
reasonable for the strength values of the undamaged slab structure. The load parameter A

is then given by (for Symbols see Figure 2) (ax +/?x+7)/(10x—d) where a

AP

8(mp +m^ )/b ß 4bT(mp +m^ )/b 7 2b(mFx+m£x) The optimal value of x is
9
the relevant Solution to the equation llox -4o;dx—ll'^-2ad 0

Truck 1: 0.130MN 2P per wheel; Truck 2: 0.065MN P per wheel

Yield line
1
b/4
b/4 <L
b/4
b/4

Wheel load Truck V Truck 2

Figure 2. Yield line pattern in upper bound calculation.

For the damaged structure the slab is nonhomogeneous. However, the same yield line pattern
is used for optimization of the load parameter with respect to x The internal work is
calculated approximately corresponding to the moment capacities in the different zones that
model the damages of the slab structure.
The concrete strength is assumed constant over the thickness of the slab. Concrete Covers
and reinforcement areas are considered deterministic before the occurrence of damages. The
yielding force of the reinforcement is used directly in the reliability analysis. The variables
are: f concrete strength, F F yield force per length unit in lower reinforcement in the
c x y
x-direction and the y—direction respectively, F' F1 yield force per length unit in upper
x y
reinforcement in the x-direction and y-direction respectively, d d effective depth of
lower reinforcement in the x—direction and the y—direction respectively, d\ d1 effective
x y
depth of upper reinforcement in the x—direction and the y—direction respectively.
As it earlier, the finite element code is formulated in cross sectional moment
is stated

capacities (per length unit). The bending moment capacities are calculated as for a normally
reinforced beam, i.e. it is assumed that the reinforcement in tension is yielding at failure. The
26 RELIABILITY ANALYSIS OF AN EXISTING BRIDGE Ä
if^

assumption is reasonable in a deterministic analysis considering fixed values only, for example
characteristic values. In a reliability analysis one or more of the input variables can take such
values in the tails of their respective distributions that other than the assumed modes of
bending failure can occur. This matter is not persued further in this study. Neglecting
reinforcement in compression, the moment capacities
mp ,mp ,mp ,mp are given by a for-
x x y y
2
mula of the form mp (1—$/2)<I>d f $ F/(df with the relevant indices x or y and
no prime or prime put on all the Symbols mp, $, d, F
In the treated problem, the expressions (2) become —*; Xq + A (xrpXj-.) 0 —/s.Xq +

A.(xR,Xp) 0 where A (•) and A.(-) are the carrying capacity functions corresponding

to the elaborate and simple model, respectively. Hence the effectivity factor function simpli-
fies to z/(xq,Xp,Xj~v)(xr,Xq)/A.(xü,Xj-v) showing that the effectivity factor is inde-
A

pendent of the load. This gives the simplification relative to the general problem that solving
the equation in (2) with respect to n and ac. requires only one calculation of A (xjpX|0
and A^XppXjO respectively. Furthermore, derivatives with respect to the load variable need
not to be calculated.

Example 1: Corroded reinforcement. Minor Cracking of the concrete Corroded reinforcement

and minor Cracking of the concrete can be caused by Chloride ingress. If only minor Cracking
with no sign of corrosion at the surface of the concrete has occurred, the corrosion is normally
either limited or it has the character of pitting. Pitting can lead to a total loss of strength in a
section. It is assumed in this example that severe corrosion damages are observed in a
relatively large zone.

<l
mm
w\\\\\\\\\\\ /i\\\w\\\±

Figure 3. Finite element mesh and zones corresponding to damages (half of the structure).

The damage zones are shown in Figure 3. The damage zones are chosen to be the same as the
damage zones for the actual slab of the Salpetermosvej bridge, treated in the next example,
except that a much larger reduction of the lower reinforcement is assumed in zone 3. Zone 1

along the clamped edge is considered to be undamaged. The concrete strengths in the zones 2

and 3 f
by the random variables Rr 2 and Rr o respectively.
are reduced by multiplying
Analogously the reinforcement areas and thus the yielding forces in the lower side in zone 3
are reduced by the factors Rp ~ and Rp ~
The variables entering the problem and distribution assumptions are shown in Table 1. The
 M. KARLSSON - J.M. JOHANNESEN - O. DITLEVSEN 27

units correspond to [m] and [MN]. Other geometrical properties of the slab are taken to be
constant.

Name Distribution Fixed value Mean C.o.v.

Load parameter Fixed Varying -

fc Lognormal - 30.0 0.15

Fx>Fx Lognormal - 0.3848 0.05
F ,F' Lognormal - 0.8747 0.05
y y
Fixed
x',d'X 0.245
d - -
d ,df Fixed 0.261 - -
y' y
Rfc2 Lognormal 0.9 0.20
Rfc3 Lognormal 0.8 0.20
RFx3 Uniform:[0.4,0.6] - -

RFy3 Uniform:[0.3,0.5] - -

all random variables are assumed to be mutually independent

Table 1 Data for the reliability analysis. (C.o.v. coefficient of Variation)

no 11 0
ooooo \ k, ooooo \=k1
100
90 o
DODDD \ kr0
ooooo A=krl
10 0

9 0
DDDDD \=kr0
ooooo A=krl
X ksORM X kSORM
80 8 0

70 7 0

6 0 H 0 -

50 5 0

40 4 0

30 3 0

20 2 0

1 0 1 0

00 0 0
1 5 2 0 2 5 3 0 3 5 2 0 2 5 3 0 3 5 4 0

Figure Reliability index versus load parameter in Example

4- 1 (left) and in the Example 2:
Salpetermosevej (right).

Corrections of the upper bound Solution for a homogeneous slab are obtained by replacing a,
/?, and 7 by (1, 2 and 3 refer to the zones as defined in Figure 3) a
4(2m£xl+mFy2+mpy3)/b ß 4bT(m£y+mpy3)/b 7 b(mFx2-fmFx2-fmFx3+mFx3) '
The results from the reliability analysis are shown in Figure 4 (left). The load parameter k.
1/v* corresponds to the simple model, whereas the load parameters kQ and k correspond -.
28 RELIABILITY ANALYSIS OF AN EXISTING BRIDGE ¦f!!4

to the effectivity factor method calculation with a constant and a first order Taylor expansion
of the effectivity factor, respectively. The fully drawn curve comes from a direct SORM
analysis for the elaborate model, i.e. the finite element model. It is seen that even with the
large deviations between the results of the simple model and the elaborate model the
effectivity factor method yields a quite good agreement between the zero order approximation
results and the results from the elaborate model. Furthermore it is seen that the agreement is
improved by using the first order approximation.

Example 2 Salpetermosevej Example 1 corresponds to the slab structure of the Salpetermose-

vej bridge with deliberately overestimated reinforcement reductions. In this example the data
are the same as in Example 1except for the random variables Rp and Rp o i«e. the
<>

reduction factors of the yield forces in the lower reinforcement in the zone 3 in the directions
x and y respectively. Here these reduction factors are assumed to be uniformly distributed
between 0.9 and 1.0, that is, less severe reductions are assumed. Measurements of the present
properties of the bridge have not been carried out but the bridge has been inspected visually.
Based on engineering judgments it is anticipated that the assumed data very well can be valid
for the bridge. The results from the reliability analysis are shown in Figure 4 (right). As in
Example 1 there is good agreement between the effectivity factor method and results from the
direct SORM analysis of the elaborate model.

Acknowledgement

This work has been fmancially supported by the Danish Technical Research Council.

References

Ditlevsen, O, and Madsen, H.O.: Beerende konstruktioners sikkerhed (in Danish, under translation to
English). SBI—rapport 122. 1990.

Ditlevsen, O. and Ambjerg—Nielsen, T.: Effectivity Factor Method in Structural Reliability. Proc.
of 4th IFIP WG7.5 Working Conference on Reliability and Optimization of Structural Systems. Munich,
Sept. 11—13, 1991. (Rackwitz, and P. Thoft—Christensen, Eds.) Lecture Notes in Engineering, Springer
Verlag, 1992.

Ditlevsen, O. and Ambjerg—Nielsen, T.: Model Correction Factor in Structural Reliability.

Submitted to Journal of Engineering Mechanics, ASCE, Sept. 1992.

H0yer, O.: EDB—orienterede brudstadieberegninger (in Danish). M.Sc—thesis. Department of Structural

Engineering, Technical University of Denmark, DK 2800 Lyngby, Denmark, 1989.

Madsen, H.O., Krenk, S., and Lind, N.: Methods of Structural Safety. Prentice—Hall, Inc. 1986.

Nielsen, M.P.: Limit Analysis and Concrete Plasticity. Prentice—Hall Series in Civil Engineering
and Engineering Mechanics 1984.

Vejdirektoratet—Vejregeludvalget (Danish Road Directorate) Belastninger. Beregnings— Og belast-

ningsregler for vejbroer (in Danish, English title: Loads. Rules for Analysis of Road Bridges and for
Loads on Road Bridges). October 1984.
 Residual Service Life of
Vie residuelle des struc
Restnutzungsdauer vo

George SOMERVILLE
Dir. of Engineering
British Cement Assoc.
Wexham Springs, GB
30 RESIDUAL SERVICE LIFE OF CONCRETE STRUCTURES

1. INTRODUCTION
When a decision is taken to appra ise a deteriorating structure - perhaps based
on input from routine inspection the client will generally want a simple
answer to the simple question "Is the structure safe and serviceable now, and
how long will it remain so in the future ?". In posing this question, he will
most probably have established hi s minimum acceptable requirements for technical
Performance - based on his future plans for the structure, primarily for
financial and functional reasons and will want to know how long it will be
before those may be reached [with or without interim repairs], i.e. the residual
Service life.
Rate of deterioration
® The Situation is shown
Present Performance schematically in Figure 1. The
problem is to establish the
Minimum acceptable I
Performance at Point A, and how
Performance long it will take to reach point
B. Point B itself requires
definition; this will depend on
Residual future client needs, while also
line taking aecount of the levels of
safety and serviceability
I provided in the original design.
Time

Fig. 1 Schematic illustration of assumed behaviour

The essential steps in the process, and the links between them, are shown in
Figure 2. In Short, where one or more specific deterioration mechanisms are
involved, these have first to be identified, their effects quantified, and the
results integrated into the procedures usually adopted for conventional
structural apprai sal.
A number of differences between assessment work and original design should be
noted. The most important of these are:
- material properties can be measured, not assumed
- dead loads can either be measured, or otherwise determined with some
aecuraey
- more realistic estimates of live load can be made
- more rigorous analytical methods can be used, while taking aecount of
interactions between all elements, and possibly establishing alternative
load bearing mechanisms
- the relative importance of load effects may be different. While basic
effects such as bending, shear and compression will have to be checked,
actions such as anchorage and bond can become more important when
deterioration is involved.
- the establishment of the likely future environment [especially micro-climate]
is of critical importance, since it will influence future deterioration
rates.
- lower factors of safety, and smaller "margins", can be used, because of the
greater knowledge of the structure - compared with the assumptions made in
design.
 G. SOMERVILLE 31

START

"
Data on initial Decision to appraise • Routine j
l
I Structure (design, ." structure ^ "~ inspection and

L matenals construction)
L-^tT^TL-J
v
Define current
State of
deterioration

/ \
(surveys, measurements)

Identify Evaluate Define

deterioration material environment

\ //\
mechanisms properties

<r

Define current r i
r levels of safety
& servicafaility

/
\aggressive/ inspections
1 |

X i i

\'
Yes \s
Estimate rate of
future deterioration

^ '

j Establish j Estimate time to

minimum technical
I
Performance Performance

1I 1'

r n
Discussions
n Decisions on
--?
1

'
1
and owners &f jture m anagement
1
J

Fig. 2 Steps involved in assessing structures and residual Service life

2. BRIEF DESCRIPTION OF BRITE PROJECT BREU-CT92-0591

This 3-year project, which started officially on 1/3/92, is concerned with
developing a quantifiable System to enable residual Service life to be evaluated
in accordance with Figures 1 and 2. The specific deterioration mechanisms to be
covered are: reinforcement corrosion; freeze-thaw; alkali-silica reaction; with
their effects considered singly and in combination.
While the project is concerned with all the steps identified in Figure 2,
particular emphasis is placed on certain factors, where new data or a new
approach is required; these are:
(a) definition of the aggressivity of the environment. This has to be
appropriate for each deterioration mechanism, while allowing for both macro-
micro-climate.
and
(b) assessment of current material properties in the structure. Basically,
this is concerned with test methods for all relevant material properties -
and the interpretation of data from these.
32 RESIDUAL SERVICE LIFE OF CONCRETE STRUCTURES

(c) definition of current levels of safety and serviceability. This is a

modelling process, taking aecount of the output from (b) above, other known
characteristics, and modern methods of analysis.
(d) assessment of future deterioration rates. Using stage (c) above as a
starting point, while augmenting the data from stages (a) and (b), future
deterioration rates are assessed.
(e) definition of minimum acceptable technical Performance. As indicated in
Figure 2, this depends not only on the output from stages (a)-(d) above,
but also on input from clients regarding future Performance requirements.
Extensive experimental work is planned for stages (a)-(d) above, but probably
the key to the whole project is stage (e), where discussions are necessary with
clients/owners to establish a ränge of future needs for different types of
structure [e.g. nuclear reactors v bridges v domestic housing]. Finally, to
verify the applicability of the whole approach in practice, a series of case
studies are planned; this means that, with the co-operation of clients, our
System of assessment will be used on structures which have been examined
previously and decisions taken on their residual life and future management.
When all of this work has been completed, the entire approach will be produced
in the form of a Manual, for convenient every-day use in practice.

3. CURRENT STATUS OF THE PROJECT

Initially, most effort is being concentrated on obtainin g experimental data,

For each deterioration mechanism, research is well advan ced in relating changes
in material properties to a wide ränge of environmental conditions, since the
realistic Classification of aggressive environments is s een as a high priority.
The starting point is a Classification for reinforcement corrosion,
concentrating on local conditions close to, and within, the structural concrete
cover. Separate conditions are foreseen for corrosion d ue to carbonation and to
Chlorides, with Chlorides being further sub-divided into de-icing salts and sea
water. Conditions for corrosion Initiation and propagat ion will be treated
separately.
Moisture conditions in the concrete are seen as being eritieal. When the
Classification is subsequently extended to cover freeze-thaw and ASR, this will
become all-important, since different moisture conditions will be eritieal (see,
eg. Table 1, taken from the CEB Recommendations on durability[l]).
Nevertheless, quantitative approach is considered feasible, and the proposed
a
Classification for corrosion will be finalised before the end of 1992.
Process
Effective
relative Corrosion of steel in
humidity Carbonation concrete which is Frost Chemical
attack attack
carbonated Chloride
contaminated
Very low [< 45%] 1 0 0 0 0
Low [45-65%] 3 1 1 0 0
Medium [65-85%] 2 3 3 0 0
High [85-98%] 1 2 3 2 1

Saturated [> 98%] 0 1 1 3 3

Risk : 0 significant
not slight 1 2 medium 3 high
Table 1
Significance of moisture State in influencing different durability
processes
 G. SOMERVILLE 33

Further experimental work is in hand on corrosion, concentrating initially on

the relationships between levels of corrosion and both Cracking and bond;
information here is important in assessing levels of structural safety and
serviceability.
Corresponding experimental programmes already exist for the other two
deterioration mechanisms - freeze-thaw and ASR. These data will be obtained
early in the 3-year Programme, since considerable effort will then be required,
in deriving assessment procedures for the combined effects of the three
deterioration mechanisms - an important, and original, part of the project.
In parallel to this activity, preliminary ideas are being formulated for the
overall framework necessary for the assessment of residual service life, while
satisfying the basic need in Figure 1 and concentrating on the Steps in the
bottom half of Figure 2. The factors involved, and preliminary development work
are described below.

4. PRELIMINARY DEVELOPMENT OF OVERALL PROCEDURES FOR ASSESSING RESIDUAL LIFE

4.1 Some basic principles
(a) A viable assessment method is made up of a number of essential and
interacting elements; these are:
- a behavioural model
- criteria defining satisfactory Performance
- loads under which these criteria should be satisfied
- relevant representative material properties
- factors or margins to take aecount of vagaries and variability in the
system, and to simplify its application in practice.
There has to be a proper balance between all the elements, in deriving the
reliability of the method. That overall reliability will generally [but
not always] be the same as that for the original design - but calculated in
a different way, using more accurately defined input.

(b) In principle, material deterioration should not be regarded as a state [to

which limits might be set] from a structural point of view, but rather as
something which can lead to an unsafe or unserviceable state. Its
influence on individual elements and structures is what has to be assessed,
in relation to defined action effects [e.g. bending, shear, compression,
for the different regions shown in Figure 3, for a simple beam.]
bond
(c) Element characteristics change at different rates with time, as shown
schematically in Figure 4. In determining current levels of safety - e.g.
at point A in Figure 4 - and in predicting future deterioration, there is a
need for awareness that other mechanisms may subsequently interfere and
become more eritieal, e.g. anchorage and bond in relation to Figures 3
and 4.
(d) Different load carrying mechanisms [see Figure 3] may have different
deterioration rates, and different levels of minimum technical Performance,
depending on the acceptable risks for each [see Figure 5].
(e) A viable assessment method for residual service life should satisfy three
essential criteria:
- the prediction of residual strength, with known acceptable safety
factors, while properly representing the effects of the deterioration
mechanisms involved,
- the meeting of future client needs, in terms of function and
serviceability;
34 RESIDUAL SERVICE LIFE OF CONCRETE STRUCTURES

incorporation of sufficient flexibility, to address the characteristics

-
of individual structures, while being underpinned by a robust
theoretical approach.
The attraction of a simple and pragmatic approach is perhaps a fourth criterion.

Region 1
Bond anchorage
e^
^ö Flexural compression

Shear
©
7^
Region 3
(cracking, flexural tension)

Fig. 3 Possible eritieal regions in a simple beam, due to deterioration

SU« ngw
K&*
co!2

Ä
bondstreng^ ,€>
6*2
o^

Time

Fig. 4 Schematic change in element characteristics with time

*> Pm4
5 Pm2
~ Pml
Pm3
©

0
® ©

Time

Fig. 5 Possible different deterioration rates for the mechanisms shown in

Figure 3. Each may have different minimum technical Performance
requirements, depending on risks [Pm1 - P^]
 G. SOMERVILLE 35

4.2 Available general methods of assessment

Most theoretical approaches can be put in one of two categories:
- damage Classification methods, e.g. [2]
- reliability analysis, e.g. [3] [4], [5]
4.2.1 Damage Classification methods
Based on condition surveys, the usual approach is to classify individual
structures into one of a number of defined categories, having looked at the
materials and elements, as well as the structure as a whole. The cause of the
deterioration is determined, and the effects [in mechanical, physical, chemical
and biological terms]. Normally, the end result is a rating System, with
recommended actions associated with each rating. The discipline associated with
this System has considerable merit, in terms of testing and evaluation of data.
4.2.2 Reliability methods
This favoured approach involves extending existing techniques, used in
developing design methods for Codes: here, safety in assessment terms can be
more easily related to that in the original design. Engineering perspectives
are provided by Schneider[6] and Fagerlund[7].
4.2.3 General
Few papers and reports address the difficult question of minimum acceptable
technical Performance in a quantitative way: there are exceptions, e.g. [8] [9].

4.3 Practica! considerations

The service lives of essentially similar structures, under apparently similar
conditions, may vary considerably. There may be many reasons for this, but two
key points emerge from the literature:
- some structures are inherently more vulnerable to deterioration due to their
design concept and the quality of the detailing. For example, in making
recommendations on the structural effects of ASR, the Institution of
Structural Engineers[10] found it necessary to define the sensitivity of the
structure and individual elements in three classes which reflected the
potential resistance to any defined level of "load", and its associated
expansion.
- if structural collapse is considered to be an extreme case of a deteriorating
structure, then a study of these cases where lack of durability has
contributed, e.g.[11] reveals that failure is due to a combination of factors
some of which are not amenable to mathematical modelling. Again, this
suggests the inclusion of a factor which reflects the quality of the original
design and construction.

5. CONCLUDING REMARKS

It will be obvious that BRITE PROJECT BREU-CT92-0591 is still at an early stage.

Experimental work is not finished; the development of an approach to assessment
is only at a preliminary stage. What is important are the principles involved
in that approach. Priority is given to meeting the needs of individual clients
for individual structures, in a practica! way. Nevertheless, the approach has
to be soundly based technically, in accounting for the effects of the
deterioration mechanisms, singly and in combination. The BRITE partners would
particularly welcome comments, especially from clients and fellow Professionals,
both now and as the work progresses.
36 RESIDUAL SERVICE LIFE OF CONCRETE STRUCTURES

ACKNOWLEDGEMENTS

The author is especially grateful for the input from his fellow partners, in
preparing this Paper. The partners are:
Dr C Andrade, Torroja Institute, Madrid, Spain
Professor G Fagerlund, Lund Institute of Technology, Lund, Sweden
Dr B Lagerblad, CBI, Stockholm, Sweden
Dr J Rodriguez, Geocisa Madrid, Spain
Dr K Tuutti, Cementa AB, Danderyd, Sweden.

The partners would like to thank the EEC, and the BRITE/EURAM Organisation in
particular, for permission to publish this Paper.

REFERENCES

1. COMITE EURO-INTERNATIONAL DU BETON [CEB], Durabie concrete structures. CEB

Design Guide, Second Edition 1989. Bulletin d'information No. 182, June
1989.

2. RILEM, Materials and Structures, Chapman & Hall, 1991, Vol. 24, No. 142,
320 pp.

3. JOINT COMMITTEE ON STRUCTURAL SAFETY, Working Documents, May 1991.

4. OSTLUND L, Durability and safety, Studies on Concrete Technology, December

14, 1979, pp 277-286.

5. SIEMES A J M, VROUWENVELDER A J M & Van den BEUKEL A, Durability of

buildings: a reliability analysis, Heron, Vol 30, 1985, No. 3, 48 pp.
6. SCHNEIDER J S, Some thoughts on the reliability assessment of existing
structures, Structural Engineering International, 1/92, pp 13/18.
7. FAGERLUND G, Service life of concrete structures, Keynote lecture at
International Symposium, Nanjing, September 1991.

8. DEPARTMENT OF TRANSPORT, The assessment of highway bridges and structures,

Departmental Standard BD 44/90, October 1990 and advice note BA 44/90,
October 1990.
9. THE INSTITUTION 0F STRUCTURAL ENGINEERS, Appraisal of existing structures,
July 1980, 60 pp

10. AD HOC COMMITTEE OF THE INSTITUTION 0F STRUCTURAL ENGINEERS, Structural

effects of the alkali-sil ica reaction, 2nd Edition, 1992.
11. WOODWARD R F W, Collapse of Ynys-y-Gwas bridge, West
J & WILLIAMS
Glamorgan, Proceedings of the Institution of Civil Engineers, Part 1, Paper
9357, August 1988, pp 635-699.
 Definition of Lo
Definition de spect
Definition von Einwi

Bernard JACOB
Civil Eng.
Lab Central Ponts et Chaussäes
Paris, France

::,: :¦::-
38 DEFINITION OF LOAD SPECTRA

1. INTRODUCTION

Knowledge and modelling of the actions on structures are a basic step in the
structural safety and reliability assessment. Most of the actions on
structures, either natural or due to human activities, are random and
impossible to forecast. For about 20 years, many research works were
undertaken on probabilistic modeis of actions, in order to describe and
quantify their occurrences and intensities or to compute the load effects,
stresses and strains induced in the structures. Many papers were published by
the ASCE, WCSE, OMAE, IFIP, Structural Safety, CEB, ICOSSAR and ICASP, IABSE
etc. and some state of the art reports are prepared by the CIB, Commission
W81 "Actions on Structures" II. a to 1. j].
Then engineers discovered the great advantages of using such probabilistic
modeis for the design and reliability evaluation of bridges, off-shore
structures, buildings, nuclear power plants, and other structures. Also the
experts who are writing or rewriting the new semi-probabilistic codes, such as
the Eurocodes in the EC countries or the Ontario bridge code, are using such
advanced methods for the calibration of the load modeis and the partial loads
safety factors. As an example, an accurate definition of the load spectra is
very important for fatigue assessment of bridges under variable vehicular
loads or of off-shore jackets under wave loads.
But if the advanced probabilistic load modeis, based on the use of random
variables or stochastic processes, are useful for accurate reliability
calculations, such as those made by the FORM-SORM methods to evaluate
ß-indices and probabilities of failure, they are not always easy for
engineers, companies, or Consultants to use. Common structural design and
checking are by application of codes, which mainly cont ains conventional load
modeis. These load modeis must be well calibrated, with the help of convenient
data and random modeis, and are often presented as load cases or load spectra.
For a given action and type of structure, the definition of load spectra
generally follows the diagram:

data gathering > probabilistic model » load spectra

load effects reliability evaluation load model calibration

2. DEFINITIONS - ACTION CLASSIFICATION AND MODELLING

2. 1 Definitions
An action is an external phenomenon that induces stresses or strains in a
structure. It is often expressed by a set of concentrated or distributed loads
which produce load effects. An action is defined by its occurrences, the time
periods of application, its spatial and directional properties, and its
intensities. There are normal, abnormal or accidental actions. The correlations
between several actions must be known.

The JCSS and ISO consider, depending on their occurrences:

- permanent actions: the period of application is the whole structure life-
time and the intensity variations are small or rare, and mainly due to
changes in the structure use or affectation (e.g.: dead load).
 B. JACOB 39

- variable actions: the occurrences are discrete, the durations, intensities

and other characteristics vary with time. There are some "high-frequency"
variable actions (wind, wave or traffic loads), and some "low-frequency" or
semi-permanent actions with slow variations (floor live loads, temperature
and snow loads).
- accidental loads: the occurrences are exceptional, unpredictable and almost
unavoidable, and they have Short periods of application (e.g.: shocks, fire,
explosions).
From the viewpoint dynamic load effect or structural response, there are:
- static actions: which do not induce significant accelerations in the
structure (snow, temperature, furniture loads),
- dynamic actions: which produce large accelerations, vibrations, or impact
forces in the structure (vehicle loads, earthquakes, wind).
2.2 Classification
Actions on structures may be classified, according to their origin, into:
- environmental actions: these result from natural phenomena, climatic or not,
and are generally very little controlled if at all. The main climatic
actions are: wind, snow, ice, rain and water currents, temperature and wave
loads. The others are earthquakes and other ground motions, avalanches, etc.
- actions due to human activities: these result from the use of the structure
and are gnerally at least partially controlled. In this category are:
traffic loads on bridges and pavements, live loads in buildings, fluid
pressures in tanks, etc.
- accidental loads: these result from human error, misuse of the structure or
very exceptional and unpredictable climatic events. The vehicle collisions
on parapets, ship collisions on piers, falling aircraft, explosions, fire,
tornados, typhoons and tidal waves are accidental actions.
2.3 Data and modeis
2.3.1 General remarks
Knowledge of any action and the definition of load spectra must be based on
both data and modeis. The data obtained by measurements of samples are
necessary to choose and calibrate the modeis and to check their hypothesis.
But they are not sufficient in most cases to fully describe an action and its
characteristic values. For a time-variable random action, a set of data only
gives a specified sample time history but does not represents its scattering
and randomness. For example, the famous record of the El Centro earthquake
provides an accelerogram which is very often used for structural design or
checking, but italways reproduces the same frequency spectrum and the same
magnitude and intensity.
In most problems, except for fatigue calculations, the maxima of the loads are
relevant, and these maxima have to be considered over long time periods, close
to the expected structure lifetimes (e.g. 50 to 100 years). It is rare to get
measurements of an action time history over such long periods at the right
location and under the same conditions. And even in such a case it would be
not enough to derive reliable extreme value distributions or far fractiles.
Then it is necessary to build good modeis based on verifiable assumptions with
parameters which can be fitted on the available data. These data may also be
used to check some of the model assumptions, e.g. by Statistical tests.
40 DEFINITION OF LOAD SPECTRA ,il4

2.3.2 Probabilistic modeis

Most of the common actions and loads on structures are random or may be
considered as random because of lack of knowledge. The probabilistic modeis
give an aecount for the Statistical uncertainties and make possible the
calculation of mean values, Standard deviations, fractiles, probability of
level crossing, extreme value distributions and finally probability of failure
or safety indices. A "good" model must be simple enough to be tractable for
calculations, compatible with the available data - and less informative, i.e.
not create any additional and unverifiable information not included in the
data -, but nevertheless detailed enough to give an aecount of the physical
properties of the action. Special attention must be paid to the correlations
between various loads, such as wind and waves or temperature and snow, in
order to avoid gross errors in load combinations. In particular a set of n
correlated random variables is only well described by an n-dimensional density
called the cross-correlation probability distribution. The convenient assump-
tion of independence must always be carefully checked.

Random time varying loads are modelled by stochastic processes, generally

stationary, and some of them by random fields if there are significant space
variations to be taken into aecount. The time-independent loads and some other
loads are represented by random variables, if the time influence may be
removed, above all in static or quasi-static analysis. In this case the load
effects are more easily calculated, with no need for stochastic differential
equations. In this common case, the notion of return period is very useful in
choosing the characteristic load values:
if X X ...X are independent random variables with the same law as X (i.e.
are the random samples of X at times 1, 2, ...n), the return period T(x) of
the value x of X is the mean time interval between two level crossings of x by
the X If F(x) is the cumulative distribution function of X, we get:
T(x) [l-F(x)]"1. If P(X^x) p (small) during n time units, we have:
T(x) ^ -n/[Ln(l-p)] for any X. One must take care not to mistake the return
period T(x) for the reference period n: in this way a 5% upper fractile over
a reference period of 100 years has a return period of about 2000 years!
For most of the actions the characteristic values generally adopted are those
having a return period between 50 and 200 years (climatic actions) and up to
1000 to 2000 years (traffic loads on bridges).

The modeis of extreme values commonly require the three types of asymptotic
distributions [2]: type I (Gumbel) for exponentially decreasing probability
distribution funetions (PDF), type II (Freenet) for geometrically decreasing
PDFs and the type III (Weibull) for upper-bounded PDFs (maxima). One must be
careful with multiple-peak PDF representing a mixture of several distributions,
as in the case of snow loads in a maritime climate.

2.3.3 Data gathering and use

Some data are necessary to choose the PDF type of any random variable and its
parameters. They may provide the basis for any Statistical test of hypothesis,
such as normality (Shappiro-Wilk), the choiee of some fitted parameters (chi-
square test) or independence.
For time independent loads, a large enough sample of data may be sufficient to
derive reliable estimators of the characteristic parameters or a PDF. But for
random loads varying with time, stationarity must be first checked; if
the seasonal or periodic variations as well as the trends must be removed
not,
before any stationary model fitting. In addition the process must be ergodic
 B. JACOB 41

(i.e. the means on time history samples converge to the Statistical mean) in
order to make the Statistical use of the data meaningful. For periodic
actions, the sampling frequency f
must be adapted to the smallest period T,
according to Shannon's rule: f^2/T. A spectral analysis is often useful for
variable actions (wind, waves, earthquakes, traffic loads), in order to
identify the stationarity time periods and the frequency peaks of energy. A
spectral representation (by the power spectral density) of the stationary
processes is used in dynamic stochastic analysis (structures under seüsmic
loads or off-shore jackets in random waves).

In the data must be filtered when recorded to eliminate noise. The

many cases
passband of the filter must be matched to the frequencies of the action. The
sensitivity of the recording device must be sufficient and any bias must be
eliminated or corrected. In some cases the samples are truncated and the upper
tails of the PDF may be underestimated!

3. ACTION LOAD SPECTRA

3.1 Definition of a load spectrum

The improper but commonly used term of "load spectrum" means a load distribution,
either derived from a sample or measurements by a Statistical analysis
(histogram, fig. 1) or defined by the theoretical or fitted PDF of the load
parameter (e.g. intensity, fig. 2). In both cases, the load spectrum may be
used to simulate randomly a time history of the load (Monte-Carlo Simulation)
or to calculate a load effect distribution. It is especially useful for
fatigue calculations (steel bridges under traffic loads, jackets in waves,
tall buildings under wind load, etc), because the whole distribution of
stresses may contribute to the damage increase. A load spectrum may also be
conventional, instead of being the result of a Statistical or probabilistic
analysis. This is the case of the code load modeis used for structural design
or checking. Some such load modeis contain a set of load cases, with various
intensities, each modified by a frequency or a probability.
3.5e-3 0.14
Hl/3 - 10

0 100 200 300 400

Tw
500 600 700 0 5 10 15 20
(W (m)
Fig.1 Histogram of the lorry weights Fig. 2 Rayleigh's PDF of wave heights
An example of a conventional load spectrum is given by the so-called "fatigue
convoys" used for fatigue calculations of steel bridges. A "fatigue convoy" is
a set of a few lorries or concentrated loads, with given gross weights, axle
loads, axle and vehicle spacings and number of passages for each vehicle. For
any sub-structure of the bridge, such as a cross-beam or a stiffener, the
42 DEFINITION OF LOAD SPECTRA /*%

application of the load spectrum means the passage of the convoy on the bridge
deck, and provides a stress Variation time history. The stress cycle amplitu-
des may then be counted by the "rain-flow" method and the fatigue damage or
the expected lifetime of the structure may be computed.
A load spectrum is generally defined by using a large sample of data
(histogram) with a number of classes adapted to the sensitivity of the load
effect considered. But in some cases it is derived from a theoretical model,
in order to simplify the calculations.
3.2 Load spectra of climatic actions
2 Wind loads are applied as pressures on
structures. The pressure is proportional
to
the Square of the wind velocity. The conventional
wind velocity is measured at 10 m above
the ground level and increases with altitude.
The peaks of the frequency spectrum are
between 3 s and 10 mn and around 4 days.
Because of the nonstationarity of the wind,
the velocity is described by its mean values
¦> T or maxima over 10 mn to 1 hr periods. These
lmn lomn ihr 4days
stationary periods correspond to the lower
Fig.3 Wind frequency spectrum parts of the frequency spectrum (fig. 3). The
velocity PDF is often described by a Weibull
distribution [l.h], and the maxima over 1 to 100 years follow the type I
(Gumbel) extreme distribution. The successive mean values are independent.

The wave loads on an off-shore structure or a sea wall are linked to te wave
heights, which are themselves correlated with the period or wavelength.
Because of the uncertainty on the wave top due to breaking, the upper 1/3
height is considered. For a given period ränge, a Rayleigh function is
generally adopted as the PDF of this height. During Short time periods of about 20
mn, the sea state is considered as stationary, giving a specified parameter a
to the Rayleigh PDF, and the successive wave heights are independent.
3.3 Traffic loads on bridges
The most important design loads for bridges are the traffic loads. They are
quite complex and the definition of load spectra is difficult because of the
randomness of many parameters: vehicle occurrences and spacing, axle loads and
gross weights. The traffic processes are generally nonstationary, correlated
from one traffic lane to the next, and traffic jams or platoons of lorries
occur randomly and often govern the structural design. Theoretical and
numerical approaches and modeis have been proposed by various authors (CIB,
ASCE, Eurocode) [l.f], [3].

The simplest and most common vehicular load modeis are built with a set of
random variables; e.g. for a traffic lane: type of vehicle (from a given
Silhouette Classification), gross weight, axle loads, spacing between vehicles
and traffic flow. Vehicule length, speed, duration of jam, number of lorries
in a platoon, and spacing in congested traffic may also be considered, as well
as the distribution of the vehicles among lane. Correlations between some of
these variables are introduced (gross weight/axle loads/silhouette/vehicle
length; speed/spacing/flow, etc). Some authors have proposed more sophistica-
ted traffic load modeis using random processes (Poisson and marked Poisson
[4], Markov, renewal Markov [5]) in order to better describe the space
distribution of the loads.
 B. JACOB 43

Most of these modeis are based on detailed mesured traffic data recorded by
Weigh-In-Motion (WIM) techniques. Large samples of traffic are now available
[6] providing sufficient information for model definition and calibration.
Moreover these data are often rieh enough (more than 30,000 lorries conti-
nuously recorded on one location) to use them directly for load effect
calculations [7]. The results are then much more accurate than by Monte-Carlo
Simulation, especially for short and medium spans.

The PDFs proposed for lorry gross

Load length L Eurocode [8] TN0 [3] weights are in most cases
bimodal, with normal or Weibull
modes. The speed PDFs are often
axle - 290 kN 260 kN
normal while the spacing PDF is a
tandem - 440 410 gamma distribution. From these
modeis and data, characteristic
triple - 530 580
axle loads and gross weights have
lorry - 925 940 been derived, as well as distri-
buted loads on a lane length. Table
EUDL 20 m 47-56 kN/m 62 kN/m 1 gives some results from [4]
EUDL 50 38-43 36 [83. Based on the extrapolations
EUDL 100 33-38 28
of the load effects having a
return period of 1000 years, a
EUDL 200 25-36 20 conventional unified load model
was built and calibrated for
Table 1 Characteristic loads with a return Eurocode 1 (Bridge loading) for
the EC countries [9]. It will
period of 1000 years, slow lane. include also some fatigue loads.
3.4 Load combinations
The combination of random actions is a difficult problem because of the
possible correlations, and in any cases because of the very low probability to
observe simultaneous extreme values and the lack of cross-observations. The
codes contain \p coefficients to modify the characteristic values in order to
make load combinations, but they are largely "heuristic". In any load
combination it
is important to know the respective variations of each
component, and its frequencies. It _is quite easy to combine a permanent load G
with a variable one L: Max(G+L) G + Max(L). If G is a quasi-permanent load
or a slowly varying one,
G* is a frequent value of G.
it
is possile to write: Max(G+L) * G*+ Max(L), where

The load combinations must be studied case by case; for prestressed concrete
bridges the combination of temperature and traffic loads is relevant. The
temperature load (gradient) may be modelled by a sinusoidal function with a
period of 24 hrs and a random amplitude, with a distribution depending on the
season. The combination with the lorry loads is made by adding the bending
moments due to each load, because of non-linearities in the stress
calculations. A procedure using Simulation and traffic records was developed
for a bridge in France [10]. Some correlation exists between these loads, due
to the traffic density peaks which oeeur at specific times in the day, like
the temperature maxima or minima.

Traffic loads must be combined with the wind load for long cable-stayed or
Suspension bridges. This problem becomes more complicated than the previous
one because both loads vary considerabily, at high frequencies. The (negative)
correlation only concerns extreme wind velocities, when traffic restrictions
are imposed for safety reasons.
44 DEFINITION OF LOAD SPECTRA

REFERENCES

1. CIB - W81 (Actions on Structures):

l.a. General Principles. 5th draft, 1986.
l.b. Self-Weight Loads. CIB report n° 115, 1991.
I.e. Live Loads in Buildings. CIB report n° 116, 1991.
l.d. Snow Loads. 6th draft, 1989.
I.e. Loads in Car Parks. 5th draft, 1991.
l.f. Traffic Loads on Bridges. Ist draft, 1990.
l.g. Fire. 4th draft, 1992.
l.h. Wind Loads. 2nd draft, 1987.
l.i. Earthquake Loads. Ist draft, 1987.
1. j. Combinations of Actions. 4th draft, 1992.

2. ANG A.H. and TANG W.H., Probability Concepts in Engineering Planning and
Design. J. Wiley & Sons, New-York, 1984.
3. WAARTS P.H. and VROUWENVELDER A. Traffic Loads on Bridges. TNO Building
and Construction Research report B92-011, February 1992.

4. JACOB B., MAILLARD J.B., GORSE J.F., Probabilistic Traffic Load Models and
Extreme Loads on a Bridge. Structural Safety & Reliability, Proceedings of
ICOSSAR'89, ASCE, San Francisco, August 7-11, 1989.

5. GHOSN M., F., Markov Renewal Model for Maximum Bridge Loading.
MOSES ASCE,
J. of the Engineering Mechanics, vol 111 n°9, September 1985.
6. BRULS A., JACOB B., SEDLACEK G. Traffic Data of the European Countries.
Report of the WG 2, Eurocode on Actions, part 12, March 1989.
7. EYMARD R. and JACOB B., Programme CASTOR-LCPC pour le Calcul des Actions et
Sollicitations dans les Ouvrages Routiers. Bull, de liaison des LPC, n°165,
Novembre-Decembre 1989.

8. JACOB B. and al., Methods for the Prediction of Extreme Vehicular Loads and
Load Effects on Bridges. Report of the WG 8, Eurocode on Actions, part 12,
August 1991.
9. EUROCODE 1 (Basis of Design and Actions on Structures), Traffic Loads on
Bridges. Part 3, draft 7, June 1992.
10. JACOB B., CARRACILLI J. GODART B. TROUILLET P., Investigation of Fatigue
of Prestressed Concrete Bridge under Combined Actions of Traffic and
a
Temperature Gradient. 4th DBMC, Singapore, November 4-6, 1987.
in french in: Bull, de liaison des LPC, n°152, Novembre-Decembre 1987.
 Fatigue Reliability Analysis of Earl
S6curit6 ä la fatigue d'anciens p
Ermüdungssicherheit früher Eise

G. RAVI
Research scholar
Indian Inst, of Techology
Bombay, India
»TP

R. RANGANATHAN
Professor of Civil Eng.
Indian Inst, of Techology
Bombay, India
46 FATIGUE RELIABILITY ANALYSIS OF EARLY STEEL RAILWAY BRIDGES IN INDIA

1.0 INTRODUCTION

In Indian Railways, a large number of rivetted bridges which were built during
the early part of this Century are still in service and hence are in use for
more than ninety years. With the increase in the train axle loading and their
frequency of occurrence,these bridges are now subjected to relatively severe
conditions and some of thera have shown signs of distress. Even though most of
these bridges have crossed their design life, their replacement with the new
ones is not economically viable. With a rational method of evaluation,
inspection and subsequent repair ifneeded, at regulär intervals, their
service life can be further extended at an acceptable risk level. A research
study has been undertaken to develop a general method of reliability based
design and evaluation of railway bridges. Attempts are being made to arrive at
a rationally developed inspection strategy which would enable the designer to
check about the possibility of service life extension of bridges.
The two key parameters required for any fatigue reliability analysis are :
a) the load spectrum or the stress spectrum and
b) fatigue test results to establish the S-N curve. Here S Stands for stress
ränge and N for number of cycles,
For getting the load spectrum, field measurements have to be done under actual
traffic conditions for various members of different bridges. For conducting
the fatigue tests, samples have to be taken from early steel girder bridges.
An extensive study in this regard has been made by Research, Designs and
Standards Organisation( RDSO Lucknow, India, the results being published
as a report [11. Necessary data for this study has been obtained from this
report.
In this paper, the fatigue reliability analysis of Early Steel Girder Bridges
has been presented. Using the results from RDSO report [11, S-N curve
characteristics and equivalent stress ränge have been oomputed. A limit state
equation based on Miner's rule has been formulated. Using the Advanced First-
Order Second Moment AFOSM method, the reliability index, ß, has been found
out for various cases.

2.0 ANALYSIS OF RESULTS OF FATTGUE TESTS AND STRESS SPECTRUM

The results of fatigue tests and stress spectrum have l>een taken from
Reference [11.Fatigue tests have been performed on specimens obtained from
early steel girder bridges namely a)Mahanadi bridge b)Netravathi bridge, and
c)Koakhai bridge. The specimens contain five rivet holen and some specimens
with rivets also have been tested. Using the test results, a linear reqression
analysis [21 has been carried out to get the S-N curve parameters. The results
are presented in Table 1.
For most of the structural details, the value of slope of S-N curve, m is
taken as 3 However the results from Table 1 indicate that the ränge of m is
0.6954 to 4.6495 This could be due the Variation in chemical composition and
other factors like age of the structure, method of rivet hole preparation,
rivet clamping force ,Variation in loads and their frequency etc. In Table 2,
the computed values of mean and ooeffioient of Variation (COV) of m and K are
given.
Under the actual traffic conditions, stress records have been obtained over a
period of time and then using rain flow [31 method of oycle count ing stress
histograms are created [11.For the stress histograms available in Reference
 G. RAVI - R. RANG AN ÄTHAN 47

Sl.No. Details of the bridge S-N curve parameters

Inverse Intercept K
slope m (in MPa units)
A.Mahanadi bridge
1 Bottom chords BC 4.649 2.473 E + 15
2 Stringers ST 1.298 1.518 E + 08
3 Cross girders OG 4.522 1.303 E + 15

4 Top chords TC + BC 1.567 7.518 E + 08

5 End rakers ER + 2.651 6.233 E + 10
Diagonals DG

6 ER + DG + TC 2.628 5.389 E + 10

7 ST + CG 2.370 3.071 E + 10
8 All members 1.804 1.879 E + 09

B.Netravathi bridge
9 BC + DG + ER 1.033 6.780 E + 06

10 All members 0.695 1.815 E + 06

C.Inter bridge
combinations
11 Cross girders 1.128 8.251 E + 06

12 Bottom chords 1.308 2.467 E + 08

13 End rakers 3.248 1.052 E + 12

Table 1 Linear regression results

Sl.No. Description Mean Coefficient of

Variation 00V

1 Intercept K 2.906 E + 14 2.575

2 Slope m 2.223 0.578

Table 2 Statistics of S - N curve parameters

[11, equivalent stress ränge S has been eomputed as follows [41

1/m
=[y f sf 1 1
48 FATIGUE RELIABILITY ANALYSIS OF EARLY STEEL RAILWAY BRIDGES IN INDIA

where f. is the relative

frequency of the stress ränge S. m is the average
value of slope .The results obtained for various cases are listed in Table 3

Sl.No. Description S MPa No. of cycles / day

re
of bridge

Mahanadi
1 Stringer 17.881 1170
2 Cross girder 20.161 1160
3 Bottom chord 49.109 116
4 Diagonal 25.068 424
Baitarani
5 Stringer 28.144 1217
6 Cross girder 26.135 365
Krishna
7 Stringer 65.988 1966
8 Cross girder 74.273 1328
9 Bottom chord 104,799 188
10 Diagonal 52.831 36

Table 3 Eguivalent stress ränge for all members

3.0 LIMIT STATE FORMULATION AND RELIABILITY ANALYSIS

To carry out the reliability analysis, the limit state equation has to be
formulated. This equation is developed based on Miner 's model for cumulative
damage [5].
The cumulative damage, D, is given by

-ii
D yL n./N.
¦= 2
i j
number of cycles of stress
itois the
where n. J N.is the ranqe S. and number of
cycles failure at constant stress ränge S..The value of N.is to be chosen
from S-N curve. Failure state is reached when D is equal to one. If summation
is written for each cyole,
»-1 1/N(S
summation l^einq oarried out for each
3

cyele. From the S-N curve, the following

relation is written for constant stress ränge.
K N S 4
However, for variable amplitude stress
ränge, knowing the stress histogram,
equivalent stress ränge can be calculated from Eq.l Hence Eq.3 is written
as
 G RAVI - R. RANGANATHAN 49

D (N/K) y Sjn 5

Introducmq the equivalent stress ranqe, S the above equation reduces to

D (N/K) (S )m 6
re
Hence the limit state equation is
q D-(N/K) (S )m 7
re
and the failure surface equation is
D-(N/K) (S )m 0 8
re
The modified form of limit state equation is obtained by takinq loqarithm on
either side of Eq.6. Thus

re -loq
loq(D K)-m loq S N 0 9

is the equation for the failure surface and the modified limit state equation
is
q loq(D K)-m loq S -loq N 10

The random variables are D, K and m. The K and m have statisties of

already
l>een fjxed in Table
and qiven value and3 of D, obtained from
Mean
OOV
Referent e[61, ai e 1.044 and 0.3 resp*x t Lvely. Havi nq Eq.10 as the limit state
equation, jeliability
index, ß, is omputed for various cases usmq =i proqram
c

l>ased on AFOSM methex:!, N represent s the desired in cvoles. All the

variables are assumed to l>e loqnormall^ distrjbufed. The oomputed values of ß
life
are presented in Tables 4 and 5 for different bridqes.

4.0 DISCUSSION AND CONCTUSION

In this studv, f-tt ique reliability ai^lysis of early steel qnder bridqes wjth
S-N curve ipproac h has been presented. Usmq the fatique test results and
stress iK-asurement s obtained frort Reference [11, C values for various cases
have been presented in Tables 4 and 5 .Table 1 presents the S-N curve
chott ac t eri st ir,,

From Tables 4 and 5, it

can [>e seen that the averaqe value of ß is axound
which corresponds to a prolability of failure of 2.3 per cent. This averaqe
2

value of H is the same as ptoposed by Moses et cl] [71 for fatique evaluation
of hiqhway bridqes. Botfoni chotds l^mq the mam members of a fruss bridqe
have a hiqher stress ranqe than other neml>ers and henoe there is a reduction
in beta vaJue for t hf se memt>ers. However from T^ble 3, if is observed that the
number of c}< les per day for Ix^ttom chords is very low when rompared to other
members. 1* is also seen that t he*re exists * linear rc-lat lonship between tu and
loqarithm of numl>er of \c les, N.For \artous cases, this relationship has been
<

e*-toblished ar d the results of this analysis aie presented in TaLde 6 A

1ypi(al £)lot of 3 \/f j sij-» loci N values for st rinqer of Mahanadi bridqe is shown
jn Fiq.l .Considerinq all the jiernt>ers and bridqes foqethei an averaqe ß
versus loq N urv - is obtained as shown in Fiq.2.
c
50 FATIGUE RELIABILITY ANALYSIS OF EARLY STEEL RAILWAY BRIDGES IN INDIA

Sl. No. No. of cycles Reliability index ß

NxlO6
Stringer Cross Bottom Diagonal
girder chord
Mahanadi bridge
1 12.81 2.076 2.001 1.517 1.870
2 2.00 2.270 2.194 1.711 2.064
3 1.75 2.283 2.208 1.724 2.072
4 1.50 2.299 2.223 1.740 2.092
5 1.25 2.316 2.240 1.757 2.110
6 1.00 2.338 2.262 1.779 2.132
7 0.75 2.365 2.289 1.806 2.160
8 0.50 2.403 2.327 1.845 2.197
9 0.25 2.466 2.390 1.907 2.260
10 0.10 2.546 2.470 1.987 2.340
Krishna bridge
1 12.81 1.381 1.329 1.186 1.483
2 2.00 1.575 1.523 1.380 1.677
3 1.75 1.588 1.536 1.393 1,690
4 1.50 1.603 1.552 1.408 1.705
5 1.25 1.621 1.569 1.426 1.723
6 1.00 1.643 1.591 1.448 1.744
7 0.75 1.670 1.618 1.475 1.771
8 0.50 1.708 1.656 1.513 1.809
9 0.25 1.771 1.719 1.576 1.872
10 0.10 1.851 1.799 1.656 1.952

Table 4 Value of reliability index for members of

Mahanadi and Krishna bridges

S.No 1 2 3 4 5 6 7 8 9 10

xlO6 12.81 2.00 1.75 1.50 1.25 1.00 0.75 0.50 0.25 0.10
ß
1.805 1.998 2.012 2.027 2.044 2.066 2.093 2.131 2.194 2.274
ST
ß
for* 1.846 2.040 2.053 2.068 2.086 2.108 2.135 2.173 2.236 2.316
CG

* Note: ST- Stringer; CG- Cross girder

Table 5 ß values for members of Baitarani bridge
 G. RAVI - R. RANGANATHAN 51

Sl.No Description Equation to the? line

Mahanadi bridge
1 Stringer ß - 0.2219 logN + 3.6655
2 Cross girder ß - 0.2219 logN + 3.5895
3 Bottom chord ß - 0.2219 logN + 3.1065
4 Diagonal ß - 0.2219 logN + 3.4594

Baitarani bridge
5 Stringer ß - - 0.2220 logN + 3.3939
6 Cross girder ß - 0.2219 logN + 3.4354

Krishna bridge
7 Stringer ß - 0.2220 logN + 2.9706
8 Cross girder ß - 0.2220 logN + 2.9188
9 Bottom chord ß - 0.2220 logN + 2.7757
10 Diagonal ß - 0.2219 logN + 3.0719

11 All the cases ß - 0.2219 logN + 3.2387

Table 6 Relationship between /"? and log N for various cases

2-6 2-8

2-5 2-4

2-4
5 20
¦o
4*
•ü 2-3 c
>*1-6

r2'2
51*2
.j2 2-1 - 41
4> er
er
0-8
20-
1-9 J 0-4 rf:
Vi sni) 5-56-06-57-0TS 4*5 50 55 60 65 70 7-5
Log of cycles Log of cycles

Fig. 1 Relationship l>etween ß and Fig. 2 Relationship between ß and

log N for stringer of log N eonsidering all cases
Mahanadi bridge
52 FATIGUE RELIABILITY ANALYSIS OF EARLY STEEL RAILWAY BRIDGES IN INDIA

From Fig 2 it
This gives the
can be seen that for a ß value of 2
number of cycles as 0*3x10 m
log is equal
For the stringer of
N to 5.5.
Mahanadi
bridge undergoing 1170 cycles per day, this would mean that a detailed
inspection has to be done once in every 250 days. Similarly for the bottom
chord of Krishna bridge, the Optimum inspection interval would be four years.
Hence if
the inspection is accompanied with subsequent repair needed, the
service life of the merober of the bridge could be further extended at a
if
acceptable risk level Thus these results provide an insight to the fatigue
behaviour of early steel girder bridges and also help in deciding about the
optimum interval for inspection and evaluation These results also aid in
fixing target reliability
a index for the design of steel bridges.

AO(NOWLEDGEMENT

The authors wish to thank Director General, RDSO, India, and other officials
of the Organisation for the material support rendered du ring the preparation
of this work.

REFERENCES

1. ,Investigations on the Assessment of Residual Fatigue Life of

Early Steel/Wrought Iron Girder bridges. Civil Engg. report No. C -
245, RDSO, Ministry of Railways, Lucknow, India, July 1987.

2. LELAND BLANK, Statistical Procedures for Engg.,Management and Science

McGraw Hill Book Company, 1980.

3. SCHILLING,CG., Stress Cycles for Fatigue Design of Steel Bridges.

Journal of Struct. Engg., ASCE, Vol 11Ö, No.6, June 1984.
4. WHITE G.J., AYYUB,B.M., Reliability Based Fatigue Design of Ship
Structures» Naval Engineer1s Journal, May 1987.

5. BANNANTINE,J.A.,CQMER,J.J.,HANDROCK, J.L., Fundamental of Metal

Fatigue Analysis. Prentice - Hall publication 1990.
6. WIRSCHING,P.W.,Fatigue Reliability Analysis of Offshore Structures.
Journal of Struct. Engg., ASCE, Vol,110,No.l0,Oct. 1984.
7. MOSES,F.,SCHILLING,CG. ,RAJU,K.S.,Fatigue Evaluation Procexlures for
Steel Bridges NCHRP Rept. No. 299, Nov .1987.
 Reliability Analysis of Steel Railway
Analyse de fiabilite des ponts rail e
Zuverlässigkeitsanalyse von Bahnbrücke

Peter KUNZ
Research Associate
Swiss Federal Inst, of Technology
Lausanne, Switzerland

Manfred A. HIRT
Professor
Swiss Federal Inst, of Technology
Lausanne, Switzerland

1
54 RELIABILITY ANALYSIS OF STEEL RAILWAY BRIDGES UNDER FATIGUE LOADING

1 INTRODUCTION

Many railway bridges built at the end of the last or at the beginning of the present Century
are still in service today. The defined service lifes of these structures have been reached
or even exceeded [1]. Owners and Controlling authorities must decide how these bridges
can be used in the future. This involves a re-assessment of structural safety including
fatigue. An example of the evaluation of a riveted bridge built in 1875 is presented in [2].
Other reasons for an assessement of the remaining service life are new service conditions
or changed structural behaviour under service loads.

The method for the assessment of fatigue safety comprises three stages :

1. Identification of eritieal construction details.

2. Evaluation of fatigue safety by calculation of failure probability.
3. Monitoring of construction details by regulär inspection.

service / monitoring

assessment of fatigue safety

identification evaluation monitoring

of of of
eritieal failure construction
details probability details

Figure 1 Method for assessing the fatigue safety of eritieal construction details

This paper presents a method for assessing fatigue safety, in particular the procedure for
the calculation of failure probability is discussed. The relationship between the calculations
of crack propagation and of damage aecumulation is emphasized.

2 METHOD FOR ASSESSING THE FATIGUE SAFETY

The condition of a structure should be monitored by regulär inspection throughout its

service life. An assessment of fatigue safety may be required as a result of observations
(displacements, Vibration characteristics, corrosion, cracks) or because of changes in
service conditions (increase of axle or uniformly distributed loads) or for legal reasons
when a defined service life is reached. Fatigue safety is mainly dependent on the following
three parameters:

- Applied stress ranges: The applied stress ranges are a function of the service loads
and the structural behaviour of the bridge.
- Geometry of construction details: The stress concentrations caused by the
geometry of construction details and the crack shape may lead to an acceleration of
crack propagation and a decrease in fatigue category.
- Number of stress cycles: The number of stress cycles applied in the past directly
influences the remaining service life of a structure.
 P. KUNZ - M.A. HIRT 55

In the method for the assessment of fatigue safety (Figure 1) [1], based on the three
main parameters mentioned above, the following stages are identified assuming that the
bridge to be evaluated has been defined.

STAGE 1 For the structure to be assessed, the first task is the Identification of eritieal
:

construction details. In order to identify eritieal construction details, it is often sufficient to

carry out an expert inspection and an intensive study of the available documents.
Inspection and maintenance reports can be particularly helpful. The eritieal construction
details can be identified using a deterministic calculation appropriate to current design
methods. In this way, one can derive a list of priorities for later investigations.

STAGE 2 : In this paper a probabilistic method to evaluate the failure probability will be
presented. This approach enables an assessement of fatigue safety that is more
sophisticated than was previously possible.

STAGE 3: The monitoring of construction details makes it possible to maintain the

structure. To this end, the inspection intervals and techniques need to be selected.

In the following chapter, the discussion is focused on stage 2 relating to the evaluation of
failure probability.

3 EVALUATION OF FAILURE PROBABILITY

3.1 Relationship between crack propagation and damage aecumulation

calculations

The calculation of remaining fatigue life is improved by combining the advantages of both
damage aecumulation and crack propagation calculations.

The advantages of the damage aecumulation calculation, based on fatigue strength

curves (S-N curves, Wöhler curves), are: this procedure is widely aeeepted, many test
results are available and the probability density funetions of the fatigue strength of the
construction details are known. In addition, a Classification System [3] is available, which is
applicable to frequently used construction details.

The advantage of a crack propagation calculation, based on fracture mechanics, is that

crack propagation for each individual stress ränge can be calculated. Using this approach
the influence on crack propagation of each stress ränge can be considered. The important
disadvantage is that the fracture mechanics parameters for the individual construction
details are generally not well known.

Stable crack propagation can be calculated by using a fracture mechanics model

(equation 1), shown in Figure 2. Crack propagation depends on the difference between
stress intensity factors AK, also called stress intensity factor, defined by
AK Y(a).Aa-VJTä (1)

AK difference between stress intensity factors (stress intensity factor)

56 RELIABILITY ANALYSIS OF STEEL RAILWAY BRIDGES UNDER FATIGUE LOADING

AK* threshold value of stress intensity

Y(a) stress concentration factor
A<T applied stress ränge
a crack size

For a constant stress ränge Ao, the stress intensity factor AK will increase with increasing
crack size a. In addition, a threshold value can be observed, i.e. a limit below which no
crack propagation will occur. For an applied spectrum of different stress
ranges, the
number of the stress ranges which are greater than the threshold limit will increase as the
crack size increases.

On fatigue strength curves (Figure 3) for constant amplitude stress

ranges, a fatigue limit
AaD can be observed, i.e. a limit below which no crack propagation will occur. In the case
of variable amplitude stress ranges, damage will accumulate if a part of the spectrum is
above the fatigue limit. Therefore, the limit below which no crack propagation occurs is no
longer constant but decreases with increasing crack size. This limit is now called the
damage limit Aam, defined by
Aath AaD(1-D) (2)

AoD fatigue limit

Aa» damage limit
existing damage

da log Aa
log —
dN
m N C*- Aam
threshold m
transition region
^.=CAKm Ao,K
dN Aa fatigue limit Ao
transition region damage I. a<j
th

;
AKtn log AK NK NC logN
Figure 2 Crack propagation as a function Figure 3 Fatigue strength curve
of the stress intensity factor

The transition region from the threshold to the Paris law in the crack propagation curve
can be described as follows:

^ c.(ak--ak-) (3)

da/dN crack propagation per stress cycle

C crack propagation constant
m slope of the crack propagation curve (Paris law)

By analogy a transition region can also be considered in a damage aecumulation

calculation. The point (NK, AaK) is a reference point on the S-N curve having a constant
 P. KUNZ - M.A. HIRT 57

slope m (Figure 3). The position of the fatigue strength curve is defined by the fatigue
strength Aac at Nc cycles; this allows the correlation with reference to the detail category
in the tables of the ECCS recommendations [3]. The damage increase per cycle as a
function of the applied stress ränge Aat is [4]:

^r-^g.(i-Dr._L
AojJ-Aag.(1-D)m NK

damage increase per cycle

Aa, applied stress ränge
AaK reference stress ränge
N* number of cycles to failure for the reference stress ränge

The damage aecumulation calculation with equation (4) can be used to derive the
remaining service life by simulations with an aecuraey of 5 to 7 % (according to the
spectrum) with respect to a crack propagation calculation using fracture mechanics. This
corresponds to a significant improvement compared to the fatigue strength curves used
for design. The ECCS [3] and Eurocode Classification Systems may be used in this
calculation, bringing with them the possibility of calculating the design life on the basis of
the statisties of a large number of fatigue strength tests.

3.2 Evaluation of reliability

The remaining service life depends on the traffic loads used for the simulations. The traffic
load model used aecounts for the fatigue effect of service loads on the structure. Traffic in
the past can be represented approximately by the Standard UIC (International Union of
Railways) trains [5]. A corresponding model has also been developed for future traffic.
Dynamic coefficients are used for the evaluation of stress ranges for Standard trains,
thereby taking into aecount the dynamic characteristics of the structure and the vehicles.

A simplified Standard load, called the fatigue load Qfat, multiplied by a dynamic
coefficient O, is used as a reference value. With the fatigue correction factor a, the
relationship between the stress ränge due to the fatigue model Aa(OQfat) [6] and the
damage due to the stress ranges of the traffic model can be established.

Based on Simulation calculations, in which the Statistical scatter for each parameter is
considered, the Statistical distribution of the correction factor p(ot) as a function of a given
number of future trains Nfut can be calculated. The distribution of the required fatigue
strength for a construction detail in an existing bridge can be calculated as the produet of
the fatigue load stress ränge and the distribution of the fatigue correction factor [4]:

p(Aareq) p(cc) • Ao(0 • Qfat) (6)

Aa(0-Qfat): stress ränge due to the fatigue load multiplied by the dynamic coefficient
P(Aareq): probability distribution of the required fatigue strength
p(oc) probability distribution of the fatigue correction factor for a given number of
future trains N^
58 RELIABILITY ANALYSIS OF STEEL RAILWAY BRIDGES UNDER FATIGUE LOADING ,©%

Knowing the required fatigue strength, the available strength of a given detail has to be
determined. Probability distributions of the fatigue strength p(Aor) of typical construction
details have been obtained from tests [7, 8]. By comparison of the two distributions, the
probability of failure can be calculated and expressed in terms of the reliability index ß :

ß= ,R-S (8)
Vi
ß reliability index
R.S mean values of the resistance and the load effect, respectively
SR, SS Standard deviations of the resistance and the load effect, respectively

Figure 4 shows the relationship between the reliability index ß and the number of future
trains Nfut. For an increasing number of trains, the reliability index ß can be seen to
decrease. For a semi-logarithmic scale and using analytically derived expressions for R
and S, this results in a straight line. This simple relationship is very convenient for a
sensitivity analysis of the main parameters.

The number of trains for 1, 20 and 100 years are also identified in Figure 4. For the
assessement of fatigue safety of an existing bridge, the region between 105 and 106 trains
is of most interest. This corresponds to a future service period of between 5 and 25 years.

Assumptions:
Ao(<frQfat):
- number of trains per day: 120
lyear 40 [N/mm2]
6 ¦ - year of construction : 1900
50 - influence length : 20 m
60 - fatigue limit at: 7-106 cycles
20y 100y - coefficient of Variation of
the parameters with scatter: 10 %

104 105 106 107

Nfut

Figure 4 Variation of reliability index with number of future trains

3.3 Sensitivity analysis

Figure 4 shows the influence of the stress ränge due to fatigue load. Improving the
accuracy of the structural model directly improves the value of the stress ränge and thus
increases the calculated reliability. Site measurement of bridge behaviour is
recommended for calibration of the structural model. The assumptions of Figure 4 using a
stress ränge of 50 N/mm2 will be used for the following figures.
 P. KUNZ - M.A. HIRT 59

Figure 5 shows the effect of the fatigue limit. The Position of the fatigue limit AaD is
characterised by the corresponding number of cycles Np. The longer a bridge will remain
in service, the more important the determination of ND will be. To this end, the fatigue limit
should be defined for each category of construction detail.

The influence of the year of construction is shown in Figure 6. This parameter is normally
known. The small number of trains at the beginning of this Century does not influence the
reliability index. In addition, with increasing number of future trains this parameter is less
important. This means that the same correction factor oc can be used for plus or minus 20
years, for the values assumed for Figure 4.

year of construction
8 T 8 T 1900,1915
-1930
.1945
3-1
4- 5-106
7-106
2
10
IG
10 10 10 10 104 105 106 107

tut fut
Figure 5 Influence of the Position Figure 6 Influence of the year of
of the fatigue limit ND construction

ß number of trains 3
8 T per day: 8
[ /0[%]
60
120 6
180
4 4

2 2

n
104 105 106 107 10< 10£ 10< 10'
Nfut tut
Figure 7 Influence of the number Figure 8 Influence of model uncertainty
of trains per day

Figure 7 shows the influence of the number of trains. The number of trains in the past has
been fixed according to a proposal by UIC and is based on the actual number of trains as
a reference value. The figure shows that the number of future trains is significant for less
than 105; for a greater number this parameter has little importance.
60 RELIABILITY ANALYSIS OF STEEL RAILWAY BRIDGES UNDER FATIGUE LOADING J%

The influence of model uncertainty is shown in Figure 8. The scatter has a significant
effect on the reliability index. When the coefficient of Variation increases, the reliability
index decreases. Results are not conservative when the model uncertainty is neglected
(higher values of ß). However, it can be seen that the difference for ß due to a change of
uncertainty between 10 and 15% is small. The influence of the model uncertainty
decreases for increasing coefficents of Variation and with increasing number of trains.

4 CONCLUSIONS

A method for the evaluation of fatigue safety for existing railway bridges has been
presented. The following conclusions can be made :
- Based on the principles of fracture mechanics and especially on the concept of a
threshold value for crack propagation, a new damage limit for the fatigue strength
curves can be defined. This limit decreases with increasing damage. For calculation
purposes, it is assumed that below this damage limit no crack propagation and
therefore no damage will occur. Based on this assumption a good agreement between
damage aecumulation and the more complex crack propagation calculation is obtained.
- The probability of failure can be calculated by comparing the required and existing
fatigue strength. The most important parameters are stress ränge due to the fatigue
load and the position of the fatigue limit. Of lesser importance are the year of
construction, the year in which service began and the influence length, if it is greater
than 10 m.

The probability of failure can be related to the probability of crack detection [4]. A
construction detail with a theoretical probability of failure below the target value can
remain in service, when the probability of crack detection (based on a given inspection
technique) and the inspection intervals are taken into aecount.

LITERATUR
[1] SEDLACEK, G. et al. Verfahren zur Ermittlung der Sicherheit von alten Stahlbrücken unter
Verwendung neuester Erkenntnisse der Werkstofftechnik. Bauingenieur, 1992, Vol. 67, pp. 129 -136.
[2] BRÜHWILER, E. and KUNZ P. Remaining service life of a riveted railway bridge. Proceedings, IABSE
- Colloquium Copenhagen : Remaining Structural Capacity. Zürich : 1993.
[3] ECCS Nr. 43. Recommendations for the fatigue design of steel structures. European Convention for
constructional steelwork, 1985.
[4] KUNZ, P. Probabilistisches Verfahren zur Beurteilung der Ermüdungssicherheit bestehender Brücken
aus Stahl. Dissertation Nr. 1023 der ETH Lausanne. Lausanne: 1992.
[5] UIC 779-1 E. Recommendation for the evaluation of the load carrying capacltiy of existing steel
bridges. International Union of Railways, 1986.
[6] KUNZ, P und HIRT, M. Grundlagen und Annahmen für den Nachweis der Ermüdungssicherheit in den
Tragwerksnormen des SIA - Dokumentation D 076. Schweizerischer Ingenieur- und Architektenverein.
Zürich :SIA, 1991.
[7] UIC. Statistical analysis of fatigue tests on steel riveted connections. International Union of Railways,
1986.
[8] FISHER, J. et al. Fatigue and fracture evaluation for rating riveted bridges. National Cooperative
Highway Research Program Report 302. Washington, D.C.: 1987.
 Optimal Fatigue Testing
Essais de fatigue optimalisäs
Optimale Ermüdungsversuche a

Michael Havbro FABER John Dalsgaard S

Structural Engineer Structural Engine
Univ. of Aalborg Univ. of Aalborg
Aalborg, Denmark Aalborg, Denmar

Michael Havbro Faber, born 1961, John Dalsgaard S

Ph.D. from the University of 1956, received his
Aalborg, Denmark in 1989. After- the University of
wards he was empolyed at the Denmark in 1984. From
Technical University of Munich, he was a post do
Germany. At present he is a research fellow and since
associate at the University of been an associate
62 OPTIMAL FATIGUE TESTING - A REASSESSMENT TOOL

1. Introduction
Engineering structures subjected to environmental conditions such as time varying loading
and corrosion will fail when the accumulated damage of the structure reaches a certain eritieal
level. When a structure is designed and its design is adjusted such that the target safety of
the structure is maintained throughout its design lifetime. This is obtained either through
classical code based design or using modern probabilistic concepts. Typically, however, for
engineering structures the original use of the structure or the initial design conditions is
changed several times before it is taken out of service. Such changes are e.g. a Prolongation
of the design lifetime, changes in the loading conditions, but also imposed accidental damage
conditions have similar effects. In such cases it may be necessary to justify that the structure
is capable of fulfilling its requirements in terms of safety, i.e. to reassess the structural
safety. For this purpose information about the actual state of the structure is collected. Such
information obviously includes the damage state of the structure, but also information about
other important characteristics of the failure modes of the structure. Important examples
hereof are material parameters, loading characteristics and geometry. The collection of such
information can be rather expensive and cumbersome as is e.g. the case of inspection planning
for offshore structures or in the case of material fatigue life testing. Therfore, it is mandatory
to have access to a methodology which provides a rational decision basis on how to collect
such additional information taking into aecount the economic aspects. The framework of
modern reliability theory, see e.g. Madsen et al. [1] and classical decision theory, see e.g.
Raiffa & Schlaifer [2] provides such a tool. The scope of the present paper is to present this
tool and to illustrate its application in the case where the safety of a structure subjected
to fatigue failure is reassessed using additional fatigue life experiment data. Two different
cases of reassessment are considered, namely the case where the reliability of a structural
component subjeet to fatigue failure is updated using new fatigue data and the case where a
new experiment is planned for reassessment.

2. Experiment Planning as a Decision Tool

The use of experimental data for the purpose of modelling is recognized as one of the most
important tools in the design of engineering structures, see e.g. Ditlevsen [3]. Typical examples
hereof are the estimation of material characteristics such as yield stresses and modulus
of elasticity, but experimental results are also used for the estimation of parameters in para-
metric equations such as fatigue crack growth modeis. In the past experiments have normally
been performed such that the uncertainty associated with the measured quantity is adjusted
to some speeified acceptable ränge, see Viertl [4]. These methods disregard economic aspects
and the actual engineering application where the statistics of the considered quantity are
used. Experiment planning was i.e. seen as an isolated problem.
The increasingly aeeepted application of modern probabilistic methods such as FORM/SORM
methods in structural engineering allows for a more refined formulation of experimental
planning. This is due to realistic probabilistic modelling of loading, consistent representation of

experimental results together with efneient tools for the estimation of probabilities. Using
these tools it is possible to perform experiment planning from a more rational basis namely
to reduce the total expected costs for the considered engineering structure. This approach
is fundamentally different from the classical approach mentioned above as it allows to
perform experiment planning in a cost optimal fashion. Following results from classical decision

theory, see e.g. Ang & Tang [5] the optimal experiment plan is the experiment plan which
minimizes the expected total cost E[Ct] of the considered engineering structure. Here, total
expected costs include all costs associated with the planned experiments E[Ce]y the expected
 M.H. FABER - J.D. S0RENSEN - I. KROON 63

costs of the structural design E[Cd], the expected costs of maintenance E[Cm] together with
the expected costs of failure of the structure E[Cf]. Hence, the expected total costs for an
engineering structure can be written as

E[CT) E[Ce] + E[Ct] + E[Cf] + E[Cm] (1)

Experiment planning can in a wide sense be understood as the planning of any action revealing
information which has impact on the predicted Performance of the structure. Therefore, an
experiment can be the action of performing experiments for the estimation of the structural
material parameters but it can also be the action of measuring unknown (and uncertain)
quantities such as structural damage, structural dimensions and characteristics of the loading
environment. With this Interpretation of experiment planning it is seen that experiment
planning becomes an essential tool in decision making for engineering structures not only in
the design phase of the structure but also in the Situation of a reassessment of the structural
integrity.

3. Experiment Planning in Fatigue Testing

In the fatigue life assessment of engineering structures such as steel bridges and offshore
steel structures parameters of importance are among others the crack growth law material
parameters, the stress concentration factors, the actual geometry of the considered structural
detail and the damage state of the structure.
In the reassessment Situation the reliability of the structure is updated through experiments
revealing information about any of the above-mentioned quantities.
Assume that the failure probability can be estimated from

Pf P(g(X,N)<0) (2)

where #(•) is the limit state function, X is a vector including the basic uncertain variables
such as geometric parameters and stress concentration factors, see e.g. Dover et al. [6], and
N is the random fatigue lifetime. Then the failure probability can be updated through
experiments revealing the realizations of the basic uncertain variables and/or through experiments

revealing realizations of functional relationships of the basic uncertain variables.

In the following it is assumed that it is possible to perform fatigue experiments using a
material which is representative for the material used in the structure under consideration.
The specific problem of making a cost optimal experiment plan is treated.
Typically, fatigue life experiments are performed in order to estimate the generally uncertain
parameters P of the lifetime distribution i^(n|P) for a given material. Given a distribution
assumption of the fatigue lifetime in terms of the number of constants or equivalent stress
ränge load cycles to failure N the distribution parameters P are estimated using Standard
tools from the statistics, such as the maximum likelihood method, the method of moments
or Bayesian statistics. Fatigue life experiments can also be used to estimate parameters
in distribution free material lifetime modeis using e.g. regression analysis in the Statistical
analysis of SN data.
The SN curve and the data points illustrated in figure 1 are taken from Dover et al. [6]. The
curves represent the mean value and the two Standard deviation fractile of fatigue life tests
of offshore tubulär joints obtained in the study [7]. The data points in the figure are used
later in an example as new information in a reassessment Situation.
64 OPTIMAL FATIGUE TESTING - A REASSESSMENT TOOL

400 -
\\ N
\
V>

300 - mean —2a*

\\ N
N B4
N
200 -
\ \B3 N> BZ
N
N

100
\ \ ßl
V*
104 105 106 107
Figure 1. Illustration of typical representation of fatigue lifetime in terms of an SN diagram,
Dover et al. [6].

To define an experiment plan the number of experiments, the stress ränge levels for the
individual experiments and the maximum number of load cycles until termination are most
frequently used as decision parameters. When the number of experiments is increased the
uncertainty structure associated with the model parameters P is changed. The uncertainty
will in general decrease if the number of experiments is increased. Therefore, the expected
failure costs for the mechanical component considered are also changed.
The experimental costs due to additional experiments are obviously dependent on the stress
ränge levels for which the experiments are to be performed. Therefore, when deciding if
additional experiments should be performed the relevant failure criteria and all the information
about the uncertain variables involved in the problem have to be taken into consideration.
As stated above the optimal experiment plan is the plan which minimizes the total expected
experiment and failure costs caused by additional experiments at a given stress ränge level
and a given maximum number of load cycles before termination. As the design costs cannot
be changed in a reassessment Situation their part in the expected total costs can be omitted
in the present context. For simplicity, maintenance costs are not considered even though they
can play an important role in the case of experimental planning for future reassessments. It is
assumed that some prior information exists, for example in the form of existing experimental
results and the problem is to determine an optimal plan for additional experiments. The
existing experiments are assumed to be performed at M stress ränge levels s\, 32,..., &m- The
number of additional experiments are n (ni, n2,..., n\£)T at the M levels *i,42,..., $m•
As decision variables n and the number of load cycles to termination Nier can be used.
Because the number of load cycles to failure in an additional experiment is random the change
in total expected costs has to be integrated over all possible outcomes of load cycles to failure
weighted by their likelihood. This corresponds to a pre-posteriori analysis from the classical
decision theory see e.g. Raiffa & Schlaifer [2].
The corresponding optimization problem is written
min E[CT(Nur,n)] (3)
iV*er,n
 M.H. FABER - J.D. S0RENSEN - I. KROON 65

Constraints related to the failure probability can easily be incorporated into (3). The total
expected costs E[Ct(Nut, n)] associated with additional experiments are then

E[CT(NUr, n)] CfENu[Pfu(Nu, NUr, n)] + ENv[Ce(Nu, Nter, n)] (4)

where C/ is the cost of failure, PJ7(Nt/\Nttr,n) is the 'updated' probability of failure given
the additional unknown experimental results modelled by the number of load cycles Nu to
failure at the corresponding stress ränge levels. How to determine this probability is described
in the next section. N^ are modelled as random variables and Effu [•] denotes the expectation
Operation with respect toN17. Ce is the experimental costs. The expectation Operations in
(4) can be estimated by nested FORM/SORM, see e.g. Guers & Rackwitz [8].
The above-mentioned technique can without theoretical difficulties be generalized to the
Situation where no experimental results are available at the time where the test planning is
made. In this case subjective prior information can be used.

4. Probabilistic Reassessment
When new information becomes available the estimates of the probability of failure (and
the reliability) of structures can be reassessed. The information considered in this paper is
divided into two types
• information of functions of basic stochastic variables
• sample information of basic stochastic variables
The first type of information is related to information about events involving more than one
basic stochastic variable. Examples of this type of information are proof load tests, non-
failure observations, measurements of response quantities and inspection results related to
damage quantities such as fatigue crack sizes.
The information is generally modelled using a stochastic variable Y which is a function of
the basic stochastic variables, i.e. Y &(Xi,X2,...,Xn,iV). The actual measurements
are thus realisations (samples) of Y. The observations can be modelled as equality events
E {H 0} [Y ym] or inequality events / {H < 0} {Y < ym} where ym can be
some observed quantity.
The probability of failure of a single element with safety margin Mp </(X, N) < 0 can then
be updated, see e.g. Madsen [9] and Rackwitz & Schrupp [10].

jy-iws^o).«"',*,'™*0) <5)

or in the case of observations modelled by equality events

These conditional probabilities can be evaluated by Standard FORM, see e.g. Madsen [9].
The second type of information is related to situations where samples of one or more basic
stochastic variables are obtained. Examples of this type of information are measurements
of the geometrical quantities and test results for the fatigue life of a component. Bayesian
Statistical methods can be used to obtain updated (predictive) distribution functions of the
stochastic variables, see Lindley [11] and Aitchison & Dunsmore [12].
Based on prior information (subjective and/or test data) a density function /^(n|P) for a
single basic stochastic variable N is established. P are parameters defining the distribution
function for N. The initial (prior) density function of P is denoted /P(p).
66 OPTIMAL FATIGUE TESTING - A REASSESSMENT TOOL

Next it is assumed that an experiment or inspection is performed. m realisations of the

stochastic variable N are obtained and are denoted n* (nj, nj,..., nj^). The measurements
are assumed to be independent. The updated (posterior) density function /p(p|n*) of the
uncertain parameters P taking into aecount the realisations is

*"(n\n*\} ~ /m(n*lp)/p(p)
/p(p|n - <*>
//m(n*|p)/p(p)<fp
where/m(n*|p) n^iMn,|p).
The predictive density function (i.e. the updated density function) of the stochastic variables
N taking into aecount the realisation n* is obtained by

fN(n\n') JfN(n\p)£(p\n')dp (8)

An updated estimate of the probability of failure P^(n*) P(<j(X,iV) < 0) can then be
determined using the updated (predictive) density function /jy(n|n*) as density function for
N.
An updated estimate of Pf can also be obtained using the posterior density function of P.

Py{n') P(g(X,N(P))>0) (9)

In (9) N, X and P are stochastic variables. The density function for N is /jv(n|P) and the
density function for P can be the posterior density function /p(p|n*).
Instead of using the posterior density an updated stochastic model for P can also be obtained
using classical Statistical methods, e.g. the maximum likelihood method. In this case the
parameters P are treated as stochastic variables and the distribution parameters in the Joint
distribution function /p are determined by e.g. the maximum likelihood method.

5. Example
In the following example a reassessment Situation is considered for an offshore tubulär Joint
subjected to fatigue crack growth. The Joint considered in particular is the Joint also
considered in Dover et al. [6] where the fatigue life has been experimentally determined. It

is assumed that the prior information about the fatigue life of the considered Joint is given
through the SN curve in figure 1. Two problems are considered here. First the problem of
updating the reliability of the Joint for reassessment by introduetion of the four new data
points in figure 1 is considered. Thereafter the problem of planning an additional fatigue
experiment for the purpose of reassessment is considered.
In order to model the prior information of the fatigue life of the Joint the model from Madsen
et al. [1] is used with the modification that the slope of the SN curve m is assumed to be
a deterministic variable m — ß. Thereby the fatigue lifetime of the offshore Joint can be

given as

logiV y+^i±^±S(/+^)-/9(log5-x) (10)

where r is the total number of experiments, x ^ ^2i=i log s, and y £ 21=1 l°Sn?- The
parameters Sxx,b and D are different combinations of first and second moments of the r
I
experiments as defined in [1], T\ and are standardized normal stochastic variables. T3 has
a X2(r ~ I) distribution. The stochastic variables are assumed to be independent.
 M.H. FABER - J.D. S0RENSEN - I. KROON 67

As prior information the curves in figure 1 are used. It is assumed that the SN curve in figure
1 is based on
r 20 experiments, four experiments at five different levels of effective stress
ranges. The logarithm to the fatigue lifetime N given S is assumed to be normal distributed
with mean value equal to 29.69 — 3.0 log 5 and Standard deviation equal to 0.6/ y/\ + 1/r.
Based on this assumption the sample moments defining the parameters in (10) are estimated
using Simulation.
The reliability of the Joint can now be estimated by considering the following limit state
function
(7(x) logiV-lognc (11)
where it is assumed that the effective stress ränge S is log-normal distributed with expected
value 200 MPa and Standard deviation equal to 20 MPa, nc is assumed equal to 5 • 104. A
FORM analysis gives a reliability index ß 3.772. The reliability is next updated using the
four new experiments from figure 1. The results of this updating is shown in table 1. It is seen
that inclusion of experiment Bl and B2 gives an increased reliability index whereas B3 and
B4 decreases the reliability. If all four experiments are used, then the reliability increases.
experiment ß
Bl 3.857
B2 3.820
B3 3.765
B4 3.769
B1+B2+B3+B4 3.892

Table 1. Reassessed reliability indexes using

the new experimental data from figure 1.
Finally the problem of planning an additional experiment for reassessment of the reliability
is considered. Assuming that the expected cost of a fatigue life experiment is ^[Ce]
1 • 106 -+- E[N] and that the cost of failure of the offshore Joint is
Cf 1.1 1011 the expected
•

total costs E[Ct] of the Joint given one additional experiment are plotted in figure 2 as a
function of the stress ränge where the additional experiment is performed. Also, the total
costs corresponding to no experiment are plotted. It is seen that the largest utility is obtained
by performing an experiment at S 340MPa.
E[Cr 10"r

10 0 experiment

9 0" no experiment

8 0

7 0"
^
S
0 100 200 300 400 500 600 700
Figure 2. The total expected costs given 1 additional experiment and the total expected cost
if no experiment is performed.
68 OPTIMAL FATIGUE TESTING - A REASSESSMENT TOOL *m

6. Conclusions
Based on the modern reliability theory and the classical decision theory a methodology has
been proposed for the reassessment of the reliability of engineering structures subject to
fatigue failure. Two situations are considered in particular, namely the Situation where
the reliability is updated using new information about the fatigue life and the Situation
where a fatigue life experiment is being planned taking economic aspects into aecount. The
methodology is illustrated by an example where a tubulär offshore Joint is considered for
which experimental data are available in terms of SN data. The example clearly shows the
significance of additional experiments for the reliability of the Joint. It is also shown that the
proposed methodology for planning of future fatigue life experiments can be used to identify
the most cost-effective stress ranges for additional SN experiments.

7. Acknowledgements
Part of this paper is supported by the research project "Risk Analysis and Economic Decision
Theory for Structural Systems" sponsored by the Danish Technical Research Council which
is greatly acknowledged.

8. References
[I] Madsen, H.O., S. Krenk & N.C. Lind : Methods of Structural Safety. Prentice-Hall,
1986.
[2] Raiffa, H. & Schlaifer, R. : Applied Statistical Decision Theory. Harward University
Press, Cambridge, Mass., 1961.
[3] Ditlevsen, 0. : Uncertainty Modelling with Applications in Multidimensional Civil En¬
gineering Systems. McGraw-Hill, N.Y., 1981.
[4] Viertl, R.: Statistical Methods in Accelerated Life Testing. Vanderhoeck &; Ruprecht,
1988.
[5] Ang, A.H-S. & W.H. Tang: Probability Concepts in Engineering Planning and Design.
John Wiley & Sons, New York, 1975 (Vol I), 1984 (Vol II).
[6] Dover, W.D., C. Peet & W.P. Sham: Random Load Corrosion Fatigue of Tubulär
Joints and Tee Butt Welds. Department of Mechanical Engineering, University College
London, London, 1987.
[7] UK Department of Energy: Background to New Fatigue Design Guidelines for Steel
Welded Joints in Offshore Structures. HMSO, 1984.
[8] Guers, F. & R. Rackwitz: Time-Variant Reliability of Structural Systems Subject to
Fatigue. Proc. of the 5th Int. Conf. on Applications of Statistics and Probability in
Civil Engr., ICASP5, Vancouver, Canada, 1987.
[9] Madsen, H.O. : Model Updating in Reliability Theory. Proc. ICASP5, 1987, pp. 564-
577.
[10] Rackwitz, R. & K. Schrupp : Quality Control, Proof Testing and Structural Reliability.
Structural Safety, Vol. 2, 1985, pp. 239-244.
[II] Lindley, D.V. : Introduction to Probability and Statistics from a Bayesian Viewpoint,
Vol 1+2. Cambridge University Press, Cambridge 1976.
[12] Aitchison, J. & LR. Dunsmore : Statistical Prediction Analysis. Cambridge University
Press, Cambridge 1975.
 Upgrading Reliability Assessm
Meilleure 6valuation de la fiabilitö
Neue Sicherheitsbewertung

Marcello CIAMPOLI
Researcher
Universitä di Roma
Roma, Italy

Paolo NAPOLI
Assoc. Professor
Politecnico di Torino
Torino, Italy

SUMMARY
70 UPGRADING RELIABILITY ASSESSMENT OF DEGRADED STRUCTURES

1. INTRODUCTION

The reliability assessment of an existing structure and, eventually, the design of upgrading Operations
are processes demanding in-depth knowledge of the effective response of the structure
under realistic action scenarios. Moreover, the structural behaviour has to be determined
taking into
aecount the level of deterioration of the structural elements.
In achieving this objeetive one must always deal with the considerable uncertainty that arises in
defining both the action scenarios, and the structural model and the materials' mechanical
properties, which are closely tied to construction quality and generally deteriorate in random fashion

over time. Without considering the actions that actually involve the structure in its future Operation,
the other sources of uncertainty arise out of the variability in space and time of the geometric
and mechanical characteristics of the structural elements, and out of the need to adopt an
analytical model of their behaviour. They also arise when the available information on the basic
variable is incomplete or not wholly significant.

In most cases then, to deepen understanding of the structural behaviour, it becomes necessary to
work up information got from quality control, from proof testing, from experimental tests, and from
periodic inspection or continuous monitoring, this information being all that is available when the
original design is missing. By means of it structural safety can be assessed more reliably: in fact,
the additional information, if accurate and consistent, provides substance to the modeis assumed
for deriving the analytical reliability evaluations, which are made on the basis of estimates of the
materials' mechanical properties and of their deterioration, and on the basis of forecasts of col-
lapse mechanisms, which would otherwise be devoid of objeetive support.
A probabilistic approach to evaluating structural safety is made natural by the need to establish
stochastic modeis for each of the several sources of uncertainty. Therefore, the possibilities pro-
vided by probabilistic methods in assessing the safety of existing buildings, and in particular in
interpreting the results of tests made directly in situ, are set forth in the following, with special
reference to an example case of notable importance.

2. SOME REMARKS ON THE RELIABILITY ASSESSMENT PROCEDURE

The methodology examined, proposed in [1], and applied in [2], [3], [4], [5], [6], [7], is based on
coupling the FORM (or SORM) methods with the techniques of Bayesian Inference.
This theory appears the most useful approach for quantifying uncertainties in structural engineering
problems, especially when coupled with decision analysis. According to Bayesian Inference in
fact, and proeeeding in a consistent and explicit manner, design can deal with information on
events and propositions (qualitative information that is, or estimates based on expert judgment).
The Bayesian approach also provides a satisfactory way of explicitly introducing assumptions
about prior knowledge, the relevant experience being quantified by the prior distributions. Moreover
it does not break down where large amounts of data are absent, it providing a mechanism
for using experience, intuition and judgment productively and in a scientifically responsible fashion.
Finally, it is also compatible with first-order reliability methods, and is therefore suited to the
problem's numerical treatment within a unitary formal context.
The method's goals are manifold: to optimize inspection programmes for existing buildings, with a
view to more efficient repair or upgrading interventions, and according to the more likely deterioration
factors; to correctly interpret the results of these inspections; and to carry on parametric
sensitivity analyses. As concems the methods for using Bayesian Inference, a "direct use" may
be distinguished, aimed at updating the probability density functions of the basic random
variables, as may be an "extended" use, coming out of the coupling of the criteria lying at its base
with the techniques of reliability analysis of structural Systems peculiar to the advanced first-order
second-moment methods.

2.1 "Direct" use

Bayesian Inference can be directly applied to problems involving parameter estimation, that is,
problems in which additional information are available about the parameters of the probability
 M. CIAMPOLI - P. NAPOLI 71

density functions of the basic r.v.s, these parameters being considered as r.v.s having a prior
distribution that expresses the designer's prior belief in (or knowledge of) their values. The
method makes it possible to derive updated pdfs on the basis of all kind of additional information,
as, e.g., those derived by experimental tests.
The prior density can be fitted empirically to observations in past experience: occasionally,
subjective assignment has to be made. Depending on the probability density functions assumed for
the "a priori" and "a posteriori" modeis, the problem's Solution can be carried out in closed form or
numerically. In the example case, the updating of the pdfs of the material properties has been
carried out assuming that both the mean value and the variance are unknown.
Therefore, denoting by X a material property, by ju the mean value of X and by a2 its variance, the
Joint prior density of ja, and a2 is expressed as the product of a conditional log-normal density LN
([n],xa2) and an inverted gamma density IG (a,p). The calibration of the parameters of the posterior
pdf of X, conditional on the results of the experimental tests (represented by a vector r of
actual observations
upon X) can be performed by means of the updating procedure illustrated in [6]
or[8].

2.2 "Extended" use

The Bayesian approach can also be useful to deal with the results of experimental tests furnish-
ing more general information on structural behaviour, then calling into play a number of stochastic
parameters. In fact those results which form one or more conditions on the vector of the basic
r.v.s X, may be interpreted by defining "artificial" events corresponding to functiona! relationships
between the Xf. Such relations derive from the analytical model of structural response that is util-
ized to interpret the particular kind of test performed. The "artificial" events are expressed in the
form:
(a) Hr(y)<0 r=1,2, n
(b) HsQ) 0 s=1,2, ....m
where Y is the vector of the basic r.v.s X plus others variables. These others are called into play
by the particular type of test, or directly included in the analytical model to explicitly characterize
the uncertainty attributed to the experimental results and to the analytical model itself.
Examples of type (a) artificial events are represented by proof loading results, where it has been
ascertained that the strength of the structure is larger than the applied load; examples of type (b)
artificial events are represented by direct measurements of derived quantities, that must be
expressed by means of an analytical model as a function of the basic r.v.s. It is obvious that more
complex experimental tests may provide information which can be represented with several
events of type (a) and (b).
When put in this form, the additional information can be applied directly to the updating of the
failure probabilities estimated a priori. In fact, updating the estimate of the structural reliability with
respect to a given limit State by means of additional information is fairly simple if the analysis is
performed by means of the advanced First-Order Second-Moment methods. The updating procedure
requires the evaluation of the conditional probability expressed by the relation:
Pf PJH<0|B, <On nHn<0nHn+1 0n nHn+m=o}
where: {H(x)} is the limit state function corresponding to the limit State considered, and the
experimental tests furnish data that can be interpreted by means of n type (a) conditions and m type
(b) conditions. The methods for evaluating the conditional probability are set forth in [2] and [9],

3. AN EXAMPLE CASE

The basic concepts outlined in the preceding paragraphs have been applied to the reliability
assessment of an important R.C. building in Turin, designed by Italian engineer Giacomo Matte
Trucco to serve as an industrial plant for the production of cars and industrial vehicles, and now
re-analyzed in view of a change in its usage assignment. The main body of the complex, built
between 1916 and 1920, consists of two parallel identical 5 storey buildings, connected to each
72 UPGRADING RELIABILITY ASSESSMENT OF DEGRADED STRUCTURES

other by means of transversal elements located every 120 m; the total length of the complex is
about 556 m. At the ends of the main body are two ramps, built in 1925-1926 and of helical form,
which allow vehicles to reach the flat roof, where there is a test track with banked curves.
In order to assess the structure reliability, an extensive campaign of investigation was planned
and developed, this comprising, besides the search for and the analysis of the original drawings,
and the survey made of the effective shape and dimensions of the structural elements:
- compression tests on concrete samples, which were cored from the main columns (a total
number of 49 samples were tested);
- the measurement of ultrasonic pulse velocity in the column cores. These tests were performed
on 5638 columns, the final result of each test being taken equal to the mean of two measurements,
made near the bottom and near the top of each column;
- the measurement of the rebound Schmidt hammer index (in 3082 different positions);
- the measurement of the electrical potential in order to evaluate any corrosion of the reinforce¬
ment (a total number of 238 tests were developed);
- the evaluation of the depth of carbonation, by means of Phenolphthalein tests;
- tensile tests on reinforcing bars cut off from the structure (22 specimens);
- compression tests on entire columns, cut out of the structure where some demolition was re¬
quired by the architectural restructuring design (3 tests);
- load tests on beams and decks.
The preliminary structural analysis, made using the results of a first series of tests, led to the
conclusion that the horizontal elements (beams and decks) should have a satisfactory bearing
capacity, therefore requiring only the repair of local damage; but all columns located on the first
and second levels seemed to be eritieal, and some of those on the third level too.
A more refined analysis was then performed in order to evaluate the failure probability of the
columns. Taking into aecount the possibility that different contractors worked at the same time in
different parts of the structure, the safety check was performed independently for each building
portion delimited by two adjacent construction joints.
The main Steps of the analysis carried out in order to verify the need for any upgrading were:
• the evaluation of the mechanical properties of the materials (i.e., of the probability density
functions of the concrete compression strength and of the steel tensile strength);
• the safety check of the columns;
• the updating of the failure probabilities derived in the previous step, and according to the re¬
sults of the direct compression test made on a column sample.
The main results of the analysis are summarized in what follows, with special reference to two
different zones: the northem ramp and a zone of the main building, called Zone 1. The norchern
ramp was erected in 1925. It is helical in form, and is supported by columns located along the
internal and external ramp perimeter, and the structure of the main building is very regulär, and
consists of span equal to 6 m in both directions.

3.1 Evaluation of the probability density functions of the strenqths of concrete and steel

3.1.1 Concrete

A preliminary sensitivity analysis has shown that concrete compression strength is of major importance
to the safety check of the columns. Therefore, to obtain the most accurate evaluation of the
pdf of this variable, it is mandatory that proper aecount be taken of a number of available
information items.

Assumption of the prior probability density function

The evaluation of the parameters of the prior pdf of the concrete strength is very difficult because
no indications were found in the original design documents, nor were Italian Standards for
reinforced concrete buildings available at the time of the construction. Therefore, the prior pdf must
be derived only on the basis of the experimental results got from similar buildings of the same
age. In this case, the measurements made on an industrial building erected in the same years in
Venice were available [7], giving a mean value of the strength of 16 MPa. For safety, a mean
value of 15 Mpa and a coefficient of Variation of 0.5 were assumed. The prior pdf is thus (in Mpa):
 M. CIAMPOLI - P. NAPOLI 73

fc=LN(l5;56.25)
Updating on the basis ofcore tests
Compression strength tests were performed on 49 cores, taken from the buildings at different
levels. From the tests results, a mean value of 15.73 Mpa was obtained, together with a Standard
deviation of 5.82 MPa, corresponding to a coefficient of Variation (c.o.v.) of 37%. The large value
of the c.o.v. corroborates the assumption of poor homogeneity of concrete in different zones.
Applying the updating procedure illustrated in See. 2.1, the posterior density function is:
fc=LN(l5.72;35.58)

Correlation between the ultrasonic pulse velocity and concrete strength

From the 5638 ultrasonic pulse velocity tests, the 49 values obtained on the columns from which
the cores were taken are considered. In Fig. 1 the measurements on cores of pulse velocity and
of the compression strength are compared: the scatter appears to be quite large. Nevertheless,
the data was fit to a correlation function between the ultrasonic pulse velocity and the concrete
compression strength in the form:
fc=crexp(c2-V)-8
where: V is the ultrasonic pulse velocity; c1 and C2 are constants to be determined in order to fit
experimental data; e is a r.v. measuring the model uncertainty associated with the form of the
correlation curve and with the scatter around the mean of the results obtained by the correlation.
The values of c1 and C2 corresponding to the best fit of experimental data, derived by means of a
non-linear regression procedure, are equal to: 2.6792 and 0.0005, respectively. The above
relationship is represented by the solid line in Fig. 1. The Standard error of the correlation is about
0.29, so that the pdf of the r.v. e can be taken equal to:
e LN(1;0.45)

(MPa) fc MPA)

15-

0-

2000 2500 3000 3500 4000 4500 40

V (m/s) Rl
Fig. 1 Correlation between V and f Fig. 2 Correlation between Rl and f

Correlation between the rebound Schmidt hammer index and the concrete compression strength
The rebound Schmidt hammer test derives the concrete compression strength from the amount of
rebound at the surface of the structural element. Many calibration tests are available to validate
the results of this non destruetive technique; however its considerable uncertainty owes mainly to
the need to establish a relation between the Young's modulus of elasticity (conditioning the
amount of rebound) and the concrete compression strength. Unfortunately, in the case of very old
concretes, another source of uncertainty must be considered: in fact, the test refers exclusively to
the surface of the structural element, where carbonation produces a local increase in strength.
Consequently, the use of the correlation diagram accompanying the instrument would produce a
serious overestimation of the strength. Therefore, the original correlation diagram was disre-
garded, and a new one was sought directly using the results on cores, as was done in the case of
the ultrasonic pulse velocity tests.
74 UPGRADING RELIABILITY ASSESSMENT OF DEGRADED STRUCTURES

Considering the rebound Schmidt hammer results on the same columns from which the cores
were taken, the results of Fig. 2 are obtained. The same figure shows the best fit obtained with
the relationship:
fc=c3exp(c4.N)-8
where: N is the rebound Schmidt hammer index; c3 and c4 are constant coefficients, whose
values, derived by means of a non-linear regression procedure, are equal to: 3.023 and 0.0516,
respectively. The above relationship is represented by the solid line in Fig. 2. The Standard error of
the correlation is about 0.29, so that the pdf of the r.v. 8 can be taken as:
s LN(1;0.45)

Correlation between the pdf of the ultrasonic pulse velocity and the rebound Schmidt hammer
and the pdf of the concrete strength
index

Considering the local values of the ultrasonic pulse velocity, the parameters of the concrete
compression strength pdf are deduced as follows.
For the northern ramp, 38 measurements are avaliable, giving:

E[V] 3872m / s Var[ V] 44100(m / s)2

therefore, applying the procedure outlined in See. 2.1:

E[fc] =20.49MPa Var[fc] 90.56(Mpa)2

For the Ist level of Zone 1, 176 measurements are available, giving analogously:

E[ V] 3507m / s Var[ V] 71829(m / s)2

E[fc] 18.22MPa Var[fc] 127.7(MPa)2

Considering the local values of the rebound Schmidt hammer index, the parameters of the
concrete compression strength pdf are derived as follows.

For the northern ramp, 38 measurements are avaliable, giving:

E[R|] 37.45 Var[Rl] 28.51

E[fc] =23.78MPa Var[fc] 168.57(Mpa)2

For the Ist level of Zone 1, 38 measurements are available, giving:

E[RI] 29.04 Var[Rl] 24.38(m / s)2

E[fc] 15.49MPa Var[fc] =68.23(MPa)2

Combination of the results derived from ultrasonic pulse velocity and rebound hammer tests
The two resulting densities are then combined, a weighting being attributed to each of them,
whose value is subjectively set on the basis of the degree of confidence given to the various
tests. In this case, the parameter X,, representing the relative weight attributed to the first test
method [(1 - X) being the weigthing factor for the second one], is assumed equal to 0.6, in order
to take into aecount the greater uncertainty associated to the rebound Schmidt hammer test, due
to the effects of carbonation.
Consequently, the posterior pdf of the concrete compression strength is:
fc=LN(21.11; 48.34)
for the northern ramp, and:
fc=LN(l6.56;32.67)
for the st floor of Zone
1 1.
 M. CIAMPOLI - P. NAPOLI 75

3.2 Steel

Assumption ofthe prior pdf

As already discussed for the concrete compression strength, no useful indications can be derived
from the design documents or the Italian Standards. Therefore, and referring to the same industrial
building in Venice, the following prior pdf is assumed for the steel tensile strength:
fy=LN(320;6400)
Updating with results of tensile tests
Visual inspection of the reinforcing bars in the structural elements brought out the existence of
three kinds of rebars: round section rebars; elliptical folded section rebars; small rectangular section
rebars. To evaluate the mechanical properties of the reinforcing steel in the columns, 22 test
were available, giving a mean strength value equal to 352.6 MPa and a Standard deviation of
56.41 MPa. Applying the updating procedure, the posterior pdf is then:

fy =LN(347; 19710)

3.3 Local verification of columns

In the reliability asessment of the main columns, the basic random variables are* the concrete
compressive strength (fc); the reinforcement yield strength (fy); the cover thickness (c); the unin-
tentional eccentricity of the live load (e); the section heigth (n) and width (b); the intensity of the
permanent load (G); and the intensity of the live load (Q). The main characteristics of the input
variables, evaluated for the most loaded columns, are reported in Table 1.
Table 1 Parameters of the basic variables

TYPE Northern ramp Zone 1 Zone 1

Ist level 2nd level
Cover thickness (mm) N 50/0.40 50/0.40 50/0.40
Eccentricity (mm) N 80/0.50 50/0.50 50/0.50
Section heigth (mm) N 600/0.10 600/0.10 600/0.10
Section width (mm) N 800/0.10 600/0.10 600/0.10
Permanent load (kN) LN 1367/0.05 1928/0.05 1594/0.05
Variable load (kN) LN 690/0.15 810/0.15 594/0.15

The amount of reinforcement in the columns section is equal to 3768 mm2 in the northern ramp,
and 1848 mm2 in the Zone 1.
The limit State function is derived considering the ultimate limit State of the base section of the
columns subjected to bending and compression.
Assuming for the materials the prior pdfs, the values of the safety index ß and of the probability
of failure Pf reported in Table 2 are obtained. If, instead, the posterior pdfs corresponding to
updating according to core tests for concrete and tensile tests for steel are used, the values ß' and
Pf'are obtained.
A more precise evaluation can be performed using for concrete the local pdf resulting from the
combination of ultrasonic pulse velocity and rebound hammer tests with core tests: the
corresponding values of the safety index ß" and of the probability of failure Pf" are reported in the
same Table 2. This level of safety is satisfactory for the northern ramp, while it is doubtful for the
second level of Zone 1, and insufficient for the first level of the same Zone. A direct updating of Pf
on the basis of the results of destructive tests was then decided. Taking advantage of the ne-
cessity of demolishing one span to erect a staircase, three full-scale samples of column were
tested until collapse: the ultimate resistance for the columns of interest resulted equal to 6768 kN.
Characterizing this result as a normal r.v. (with a c.o.v. equal to 0.20, to aecount for measurements
uncertainty, and, mostly, for differences between the tested speeimen and the other
columns), the direct updating procedure of See. 2.2 has been applied.
76 UPGRADING RELIABILITY ASSESSMENT OF DEGRADED STRUCTURES

The corresponding "artificial" event (i.e., the comparison between the theoretical ultimate Nu and
the measured Nproof normal force) forms a type (a) condition on the entire set of the vector of the
r.v.s. The results of this updating are reported in the last column of Table 2.
The reliability of second level of Zone 1 resulted completely satisfactory, while it was decided to
Upgrade the most loaded columns of the first level of Zone 1.
Tabel 2 Values of the safety index ß and of the probability of failure Pf for the various updating

ß/Pf ß7 Pf' ß'V Pf" ß"7 Pfm

northern ramp 2.33/.99 KT2 3.31/.4710-3 4.92/.43-1 Or6 5.72/.54-10-8
zone 1 - Ist level 1.67/.5410-1 2.37 /.88-10-2 2.97/.15 10-2 /
3.90 .48-10-4
zone 1 - 2nd level 2.17/.1510T1 3.10/.95 10-3 4.29/.91-10-5 4.77 / .94-10~6

4. CONCLUSIONS

Bayesian Inference has proved to be a powerful procedure for improving knowledge of the
of materials in existing structures, and of the bearing capacity of the structural System. The
properties

coupling of Bayesian Inference with FORM or SORM methods provides a very straigthforward
process for directly updating the failure probability of a structure, taking advantage of load tests.
The cost of in situ testing a structure and the increased complexity of the calculations are usually
more than compensated for by the saving made possible by the improved knowledge of the ca-
pability of the structure. In the example case presented here, a very accurate in situ investigation
has been performed, producing a large amount of data. Owing to the Statistical processing within
the rational framework of Bayesian Inference, and using the FORM method, the amount of
strengthening required has been notably reduced.

ACKNOWLEDGMENT: The authors are grateful to the owner LINGOTTO S.r.l. and to the
Consulting firm FIAT ENGINEERING S.r.l., which co-ordinates the redesign process, for kindly having
made available the test results and the design data on the Lingotto building in Turin.

REFERENCES
1. MADSEN H.O., KRENK S., LIND N.C., Methods of Structural Safety Prentice-Hall, Inc.,
Englewood Cliffs, New Jersey, 1986.
2. MADSEN H.O., Model Updating in Reliability Theory, in Reliability and Risk Analysis in Civil
Engineering, Vol. 1, Proc. ICASP5, Vancouver, Canada, May 1987, pp. 565-577.
3. NAPOLI P., RADOGNA E.F., On Safety Evaluation of Existing R.C. Structures: Critical Con-
siderations About the Bayesian Approach Applied to Level 1 and Level 2 Reliability Methods,
Politecnico of Torino, Dep. of Structural Engineering, Publ. 16, October 1988.
4. CIAMPOLI M., PINTO P.E., Methods for the Reliability Assessment of Existing Structures,
Proc. AICAP'89, Naple, May 1989, pp. 21-36.
5. NAPOLI P., RADOGNA E.F., MATERAZZI A.L., Valutazione della sicurezza delle costruzioni
esistenti: utilizzazione delle tecniche Bayesiane ed impiego di metodi di livello 2 per la cali-
brazione dei coefficienti parziali di livello 1, Proc. AICAP'89, Naple, May 1989.
6. CIAMPOLI M., Reliability Evaluation of Existing Structures: Updating Technique to Account for
Experimental Data, Proc. of the Int. Conf. on 'Monitoring, Surveillance and Predictive Maintenance
ofPlants and Structures', AIPnD, Taormina, October 1989.
7. CIAMPOLI M., NAPOLI P., RADOGNA E.F., Criteri generali per la stima della sicurezza nella
scelta degli interventi sulle costruzioni esistenti, L'lndustria Italiana del Cemento, N. 649,
November 1990, pp. 908-922.
8. BERGER J.O., Statistical Decision Theory and Bayesian Analysis, 2nd Edition, Springer-Verlag,
New York, 1988.
9. DITLEVSEN O., Uncertainty Modeling with Applications to Multidimensionai Civil Engineering
Systems, McGraw-Hill, 1981.
 Safety Criteria for the Evaluati
Criteres de söcuritö pour 1'eValua
Sicherheitskriterien für die Bewer

D.E. ALLEN
Research Officer
National Research Council
Ottawa, ON, Canada
78 SAFETY CRITERIA FOR THE EVALUATION OF EXISTING STRUCTURES

1. INTRODUCTION

There are increasing pressures to preserve and maintain existing structures such as
buildings and bridges for as long as possible with a minimum of structural Intervention.
The pressures derive primarily from the cost of upgrading but include also user
disruption, energy conservation and heritage value. These pressures, along with the fact
that the structure exists, has performed satisfactorily and has been inspected for defects
means that the criteria for evaluation of existing structures for continued use need not be
as conservative as for the design of new structures. The following describes the basis for
minimum safety levels for the evaluation of existing structures under development in
Canada.

2. SAFETY CRITERION

The following safety concepts can be applied to determine appropriate safety levels for
civil engineering structures:

1. Probability of failure (Pf) or reliability index (ß). In Canada a reliability index of

3.5 is used for the design of bridges. For buildings it varies but is approximately the
same. An exception is connectors such as bolts and welds for which, ß is of order 5.

2. Life safety, or probability of death or injury for persons exposed to structural

hazards. This considers, in addition to the probability of failure, the likelihood of death
or injury if failure occurs, as well as other factors such as the activity and number of
people at risk.

3. Optimum hazard reduction. This concept applies to an inventory of existing

structures that contain structural hazards. The objeetive is to gradually reduce these
hazards in aecordance with benefit in reduced risk vs. cost. Such an approach is being
carried out to reduce the seismic hazards posed by existing buildings in Los Angeles.

4. Damage control. There may be life safety, economic and other reasons not only to
prevent collapse but to control structural damage as well. For hospitals, for example,
damage control becomes a life safety issue in the event of a disaster such as an
earthquake or hurricane.

Each concept has its applications depending on the project under consideration. It is,
however, possible to identify minimum safety levels for 'ordinary' structures based on
life safety. These minimum safety levels must be adjusted upwards for evaluation of
special structures such as hospitals, key bridges or communication towers, depending on
the consequences of failure or damage. Also, based on life cycle considerations, it often
becomes economical to follow current design criteria if structural upgrading is required.

The following life safety criterion is used to determine minimum safety levels for
structural evaluation [1]:
J[% D.E. ALLEN 79

where
Pf target probability of failure based on life safety (this is a notional
probability for setting technical criteria, not an actuarial one)

K calibration factor based on experience with existing criteria

A human activity factor which reflects what risk is acceptable in relation to

other non-structural hazards associated with the activity (taken as 1 for
buildings, 3 for bridges, 10 for certain work-related activities [1])

W warning factor corresponding to the likelihood that, given failure or

recognition of approaching failure, a person at risk will be killed or
seriously injured (W=1.0 for impact with no warning)

Vn^ importance factor based on the number of people likely to be at risk if

failure occurs, essentially an aversion factor based on public reaction to
high fatality hazards

T assumed reference period

3. CALIBRATION TO DESIGN CRITERIA

It is well known that life-threatening structural collapses are relatively rare, furthermore
most are due to human error or accidents not addressed by current design criteria.
Therefore current design criteria, if correctly applied, provide a safe upper bound to the
life safety criterion, Equation (1). This assumption can be used by considering the ratio
of the target probability of failure for evaluation to the target probability of failure for
design where, from Equation (1):

i=Ag_.VVi.V^ (2)
Pfd Ad we 7*7
where the subscripts d and e refer to design and evaluation respectively.

Because of the logarithmic relationship between Pf and ß, the ratio Pfe/Pfd can ^e
approximated by an adjustment in target reliability index, i.e.
A ßd-ße (3)
where ßd and ße are the target reliability indices corresponding to the target failure
probabilities Pfe and Pfd determined from the Standard normal distribution curve. For
example, A 0.5 corresponds to Pfe/Pfd of approximately 1/5 for ß^ in the ränge 2.5 to
3.5.
80 SAFETY CRITERIA FOR THE EVALUATION OF EXISTING STRUCTURES

4. EVALUATION FACTORS FOR DETERMINING SAFETY LEVELS

If the ratios Wd/We, V^d/V^T anc* Ae/Ad can ^e determined for evaluation as
compared to design then the target reliability index ße can be determined from Eqn. (3)
and safety factors determined by current reliability techniques. The factor W, however,
is not easy to assess in practice. Factors that can be assessed by the structural evaluator
which affect W include the following:

- component behaviour: If a component fails gradually then failure is likely to be

noticed before collapse takes place allowing. time to avoid life-threatening
consequences.

- System behaviour: If a component fails without collapse because of alternate paths of

support (redundancy) then the risk to life is considerably reduced.

- inspection: Inspection affects the warning factor W by providing clues of approaching

or potential failure in time to avoid life-threatening consequences.

These factors, along with risk category which is related to Vn and A,

are listed in Table 1
along with a comment as to whether or not they are taken into aecount in current
Canadian structural design codes for bridges and buildings.

Table 1 Structural Evaluation Factors Affecting Risk to Life

Factor Taken into Account by:

Evaluation Factor Parameter in Eqn (2) Bridge Code[2] Building Code[3]
Component Behaviour W no yes
System Behaviour W no no*
Inspection W no no
Risk Category Vn and A no no**
* partly, for earthquake only
** only on the basis of building use and oecupancy

5. APPLICATION TO BRIDGE EVALUATION

Minimum safety levels for bridge evaluation under traffic load have been developed
based on the above approach [4] and incorporated in the Canadian bridge code [2]. The
safety levels are expressed in terms of a target reliability index given in Table 2, adjusted
as a function of the four evaluation factors in Table 1. The reliability index adjustment, A
is made up of contributions from each of the four evaluation factors. The maximum
contribution for each factor is based partly on a consideration of the values of the life
safety factors in Eqn (2) and partly on existing criteria used in other codes. A maximum
A of 0.5 for component or System behaviour, for example, corresponds to an assumed
likelihood of death/injury if failure oecurs of approximately 1 in 5, or 1 in 25 for both
together. A A of 0.5 is applied for supervised passage of an overloaded vehicle, because
ßk D.E. ALLEN 81

all other traffic is kept off the bridge, which reduces the factor ^/n^ in Equation (2), and
only the driver is at risk, which increases the factor A^ in Equation (2).
Table 2 Reliability Index, ße, for Bridge Evaluation

ße 3.5 - [Ac + As + Aj + AR] > 2.0

where ße is based on a one-year time interval for all traffic categories except for supervised
overload, where ße is based on a single passage.

Adjustment for Component Behavior A^

Sudden loss of capacity with little or no warning 0.0
Sudden failure with little or no warning but retention of post failure capacity 0.25
Gradual failure with probable warning 0.5
Adjustment for System Behaviour A5
Element failure leads to total collapse 0.0
Element failure probably does not lead to total collapse 0.25
Element failure leads to local failure only 0.5
Adjustment for Inspection Level Aj
Component not inspectable -0.25
Component regularly inspected 0.0
Critical component inspected by evaluator 0.25
Adjustment for Risk Category Ar
All traffic categories except supervised overload 0.0
Supervised overload 0.5

The total ränge of ße in Table 2 is from 1.75 to 3.75, where the upper limit, 3.75,
corresponds to a safety equivalent to that assumed for design [2]. The lower limit, which
occurs only for supervised overload, represents an economic risk to the bridge authority
(theoretically 1/25 times the loss if failure occurs); a lower limit of ße=2.00 was therefore
imposed. Most traffic networks have considerable flexibility if a bridge failure takes
place but in some cases the effect of a bridge failure on the local economy can be severe.
In such cases the lower limit for ße should be increased.

The target reliability index in Table 2 was used to develop load and resistance factors for
the evaluation of bridges in the Canadian bridge code [5].

6. APPLICATION TO BUILDING EVALUATION

The same basic approach has recently been applied to buildings [6]. Although the basis
is the same as for bridges, the method was altered. The reason for this is that the
confidence in reliability methods is much greater for bridges under traffic load than for a
wide variety of buildings under a wide variety of loads, including earthquake. Instead
of recommending reduced target safety indices for building evaluation it is more
82 SAFETY CRITERIA FOR THE EVALUATION OF EXISTING STRUCTURES

practical to recommend reduced load factors. These were determined by use of the
following log normal relationship [6]

oce adexp [-aV(1 + VJ|)(1 + v!)] [4]

where a^ is the design load factor and oce is the evaluation load factor, A the target safety
index adjustment. Vr and Vs are the coefficients of Variation representing the
uncertainties of resistance and load respectively. Based on assumptions for Vr and Vs
given in Table 3, Figures 1-3 show the relationship between load factor and the target
reliability index adjustment A. Based on Figures 1-3, Table 4 contains recommended load
factors for building evaluation.

Table 3 Uncertainty Assumptions

for Estimating Load Factors for
Buildings

Uncertainty
Load vs
Dead 0.1
Variable* 0.3
Earthquake 1.1
Resistance Vr 09
Steel 0.1-0.15
r-03
Concrete 0.15-0.2 -0 2 00 02 04 06 08 10 12 14

Masonry 0.2-0.3 RELIABILITY INDEX ADJUSTMENT A

Wood 0.3

Occupancy, snow and wind loads Fig. 1 Dead load factor (Vs=0.1)

I 6

14 4- I 4 I- I
4 I- I

O I
4 rr
O
I

I
2

0
rV.
IT r I T I I

£08 4- I
^N^ I- I
4 I- I

Q
Q 2
I
306 r i "i r ^5fe,*,**v!- T ' '

04 iI
~
i i
I
^T****»*^
i ivr
^*S5!ia.
o 1

01!
015 I I I I I I I I
Vr»03
09 02 4~ I 4 l_ |
4 I- I

08 00
-0 2 00 02 04 06 06 -0 4 -0 2 00 02 04 06 08 10 12 14
RELIABILITY INDEX ADJUSTMENT A RELIABILITY INDEX ADJUSTMENT A

Fig. 2 Variable Load Factor (Vs=0.3) Fig. 3 Earthquake load factor (Vs=l.l)
 D.E. ALLEN 83

Table 4 Load Factors for Building Evaluation

Adjustment to
Design Safety Load Factor for:
Level
A (As + AR+AP)t Dead Load* Variable Loads Earthquake
0 1.25 (0.85) 1.50 1.00
0.25 1.20 (0.88) 1.40 0.80
0.5 1.15 (0.91) 1.30 0.63
0.75 1.11 (0.93) 1.20 0.50
1.0 or more 1.08 (0.95) 1.10 0.40

t Adjustment for System Behaviour As

-failure leads to collapse, likely to impact occupants 0.0
-failure is unlikely to lead to collapse, or unlikely to impact 0.25
occupants
-failure is local only, very unlikely to impact occupants 0.5

t Adjustment for Risk Category Ar

-high building importance or high occupancy exposed to 0.0
failure
-normal occupancy exposed to failure 0.25
-low occupancy exposed to failure 0.5

+ Adjustment for Past Performance Ap

-no record of satisfactory past Performance 0.0
-satisfactory past Performance** or dead load measured*** 0.25

The value in the brackets applies when dead load resists failure
Apply only to dead and variable load factors, age 50 years
or more, no significant deterioration.
Apply to dead load factor only.

Two evaluation factors in Table 1 were not included in Table 4, namely 'component
behaviour' because it is already taken into aecount in current design criteria, and
'inspection' because building structures are not inspected on a regulär basis and therefore
warning is not reliable. The risk category for occupancy in Table 4 (high, normal, low)
can be estimated on the basis of floor area exposed to potential collapse if the failure
occurs, oecupant density and duration of occupancy (hours per week).

A new evaluation factor 'past Performance' is included, however not because it affects the
life safety criterion Equation (1), but because it reduces the uncertainty in estimating
loads and resistance compared to design. Dead load parameters, for example, may be
measured, and the corresponding reduction on uncertainty (Vs from 0.1 to 0.05)
corresponds to a A of 0.25 [6]. More significant, however, is satisfactory past
Performance over many years under dead and variable loads such as wind and snow.
84 SAFETY CRITERIA FOR THE EVALUATION OF EXISTING STRUCTURES

Successful past Performance, however, is difficult to quantify in terms of reduced safety

coefficients. Table 4 contains a conservative adjustment, A=0.25, the same as for
measured dead load [5].

Besides the load factor adjustments contained in Table 4, there will also be adjustments
in the resistance factors for components such as bolts and welds.

7. REFERENCES

[1] CSA S408-1981 Guidelines for the Development of Limit States Design, Canadian
Standards Association, Rexdale, Ontario, Canada

[2] CSA S6-1990. Design of Highway Bridges: Supplement No. 1-Existing Bridge
Evaluation Canadian Standards Association, Rexdale, Ontario, Canada

[3] National Building Code of Canada 1990, Part 4, Structural Design, National
Research Council Canada, Ottawa, Ontario

[4] Allen, D.E., Canadian Highway Bridge Evaluation: Reliability Index. Canadian
Journal of Civil Engineering, Vol. 19, No. 6, December 1992

[5] Kennedy, D.J.L., Gagnon, D.P., Allen, D.E., and MacGregor, J.G., Canadian
Highway Bridge Evaluation: Load and Resistance Factors. Canadian Journal of
Civil Engineering, Vol. 19, No. 6, December 1992

[6] Allen, D.E., Limit states criteria for structural evaluation of existing buildings.
Canadian Journal of Civil Engineering, Vol. 18, No. 6, December 1991, p. 995-1004
 Experimental Investigation of Tra
V6rifications expörimentales de la Ch
Experimentelle Untersuchungen der V

Andrej SOKOLfK
Prof. Dr.
Univ. of Transport and Commun.
Zilina, Slovakia
86 EXPERIMENTAL INVESTIGATION OF TRAFFIC LOAD ON HIGHWAY BRIDGES

1. INTRODUCTION
From the point of view of service reliability
of a bridge construction
as well as an increase in traffic
intensity on highways together
with an effort to apply the knowledge of theory of reliability
to
the design of this type of structures we need to clarify many
partial problems connected with the Solution of the system bridge-
loading-environment. At present the main task is to obtain a true
picture of the load magnitude and its effects on the structure. For
this reason it
is clear that bridges which are subjected to
considerable dead and moving load as well as to a secondary load
deserve a maximum attenttion.

2. THE MEASURING AND COMPUTING SYSTEMS

A measuring system (Fig. 1) has been developed to provide measuring
time variations of the bridge deflection, enconding the passing
trucks into 8+2 categories (Table 1) and sensing their axles
together with the directions of the drive. Both, analog Signals
(deflections) and discrete data (traffic parameters) are stored by
an Instrumentation tape recorder. The deflection sensing device
consisting of a wire stretched by a spring and an inductive
displacement transducer or a stretched resistance wire between the
measurement point and the terrain are used to measure the bridge
deflections [1].

A minicomputer system (Fig. 2) real-time operating Software

with a
is exploited to digitize preprocess the recorded deflections
and
together with the discrete data on the vehicles. The results are
digital records containing just a relevant part of the digitized
deflection Signal and a block of the data on the trucks which have
evoked previous responses.

A Fortran IV Programme computes dynamic parameters (Fig. 3) as they

are defined by the Czechoslovak Standard 73 6209 [2]. The digital
filtering is used to obtain a mean deflection curve and to select a
pure dynamic portion of the signal. The computation results in the
dynamic parameters for each relevant response of the bridge to the
passing trucks together with the discrete data on the trucks.
 A. SOKOLIK 87

INSTRUMENJAHON VAN Mi«E

OPERATOR
INSTRUMENTATION
TAPE RECORDER TRUCK-TYPE
ENCODER

STORAGE
SCOPE
n in ADAPTER
AXLE
DETECTORS
BALANCING
VOLTAGES DVM
SUPPLY

L. VZ&Z' BRIDGE
DEFLECTION
SENSING
DEVICES

for monitoring bridge

% i*
Fig. 1 The measuring system
deflections and some traffic parameters

INSTRUMENTATION DISC
TAPE RECORDER MEMORY

TRUCK-TYPE
DECODER

CJ Q_
/- Z> MINICOM¬
ADAPTER PUTER
DECODER

CD CO
_J Lü CJ
CD C£ _J UJ Q
\
LOW-PASS DZ LU LJ X
I—I
ANALOG
FILTERS
Lü ü_
_J
Q
r—l

:>: Q-
I
<C Q_
DZ l-l
CJ r—
CO
c CRT
CN TERMINAL
cn <c

TT
Fig. 2 The Computing system for preprocessing the recorded
analog and discrete Signals
88 EXPERIMENTAL INVESTIGATION OF TRAFFIC LOAD ON HIGHWAY BRIDGES

t ^r ^^¦"'^^•*^*'*S»""BT^^" ^t
co CO

m m>10
f=
t« + t2 + +t
f
t, + t2+ *t„ nr
Ae tf
<r/ z=s

Ml2
1 tiT
Fig. 3 Illustration of Computing the natural frequencies and f
f, the damping *& and the dynamic coefficient S^^/S-
nid x. in

A nonrecursive low-pass filter produces a mean deflection curve. The

first and the second differences of this curve are continually
evaluated as to locate the mflection points xlf i2 and the minimum
Sm. At the interval <i1# i2> vibrations of the loaded bridge are
mvestigated and the Smax is located. To distinguish the natural
frequency f
from higher frequencies induced by dynamic wheel loads
of the vehicles a nonrecursive band-pass filter is applied on the
Signal at the interval. The same filter is used to find out the
natural frequency f
of free decayed vibrations and the logarithmic
decrement of damping & The Programme processes also the encoded da
ta about the vehicles, which exited the previously evaluated records
of the bridge deflections. Speeds, axle bases and the categories are
calculated and output as the traffic data.

3. THE LONG TERM OBSERVATION OF INFLUENCE OF THE TRAFFIC LOAD

ON THE BRIDGE
The long term Observation has been carried out on two one span
bridges on the first class highway. Both bridges are of the same
construction system. The structure is made of Standard
post-tensioned concrete (50 MPa) I-girders (Fig. 4). The former is
of 26 m span and its construction height is 1.25 m. The latter is of
21.26 m span and is 1.1 m high. The substructures are created by
massive abutment with the parallel wmgs. The expansion joints are
placed above one abutment only.
 A. SOKOLIK 89

13 500/2 13 600/2

1250 5250 3500 1500 1300

150 100
2°/ 2% 2% f
£öffl90 8

575. 1500 1500 1500 1500 1490 1490 1490 1490 575

1150 350 1150 350 1150 350 1150 350 I 340 1150 340 1150 340 1150 340 1150
f-f f-* +-^ f-f—I—*—* «r-f +-f *-*

i
575 ,575
(9x1150)^(8x350)] /2 [(9x1150) + (8x340)]/2
J*
150
'
265
;
Fig. 4 Crossection of the bridge near H. Hricov and K. Lhota

The position of the bridges on the highway network is characterized

by the transport intensity. We have recorded about 2.000 passing
trucks on the first bridge and about 2.400 on the second one during
the same period. The trucks have been encoded into 8 categories
according to the number of axles, type of a truck, axle distances
and bearing capacity, plus two extraordinary categories (Table 1).
The Operator observes the traffic
the recorder by the and operates
truck-type encoder. He selects the pushbutton corresponding with the
category and the drive direction of the passing truck or with an
extraordinary Situation on the bridge.

3.1 Histograms and their Statistical characterics

Having processed the whole group of the digitized deflection Signals
the obtained results were processed by the Fortran programme which
compiled the measured values into the histograms of the deflection
distribution for each category and drive direction of the trucks as
well as for the whole group.

Then the Statistical data of the histograms were computed and the
aproximation by the following theoretical probability distribution
was tried: Weibull's,
Gumbel' s, Raleigh's, Exponential, Normal and
Logarithmic-normal, Gama and Chi-model. We have carried out the
Statistical testing based on the assumption that at least one of the
theoretical probability distribution is realistic for distribution
obtained from the measurement and that at least one of the
introduceci theoretical distribution would correspond to this
90 EXPERIMENTAL INVESTIGATION OF TRAFFIC LOAD ON HIGHWAY BRIDGES

Ist BRIDGE 2nd BRIDGE

SCHEME
VEHICLE THEORETICAL MODEL
TYPES
No OF TRUCKS

1
c feO } Sa}
VOLVOl
TRAIL
WEIBULL 'S 57 61 NORMAL

2 c „X „n
T
T
813 \ TRAIL GUMBEL 'S
138J
57 IG WEIBULL 'S

/ T 111 T 813
CU~j
1

3 T
T
138
148
GUMBEL 'S 165 205 LOG-NORMAL

4 c^L_J § 7061

S
+SEMI- GUMBEL 'S
lOOj TRAIL
178 210 WEIBULL 'S

S 7061
5 c o—H-o—o- s ioofTRAILEF GUMBEL'S 195 265 NORMAL

S 706
6 \j \j" S 100
GUMBEL 'S 379 485 LOG-NORMAL

/~
gUSES

7 [o o
S
SL
706-RT0
IKARUS
11. SC 73^
GUMBEL 'S 153 217 LOG-NORMAL

>^-r V 35-R0MAN
8 I-oJ
o-l S5T, AVIA
ROBUR, IFA
WEIBULL 'S 371 410 FRECHET 'S

ALL VEHICLES DON'T

UNIVERSAL
9 GROUP
INC LUDED IN
GROUPS 1-8
GUMBEL 'S 94 124 NORMAL

UNIQUE PASSING AND

0 GROUP
MEETINGS
2 VEHICLES
GUMBEL 'S 344 350 WEIBULL 'S

ALL PASSING TRUCKS GUMBEL 'S 1989 2409 WEIBULL 'S

Tab. 1 The schemes occuring trucks plus

of the 8 most
2 extraordinary categories and the corresponding

optimal theoretical probability distribution modeis

of the deflections for each category as well as for
the whole group of the passing trucks
 A. SOKOLIK 91

distribution. The parameters ofthe competent theoretical distribution

for each group of deflection were calculated together with the
significance test of each distribution [4].

Analysing the results achieved on the first bridge we have found out
that the Gumbel' s model, at the 5% significance level, is the most
suitable in 72% and the one of Weibull' s in 28% of all the histograms
in which the measured deflections were divided according to
the category of trucks as well as according to the both directions
of drive. On the second bridge we have found out that the Weibull' s
model can be taken as an optimal theoretical one in 30% of all the
histograms. For 30% of them the Logaritmic-normal model is the most
suitable one as well as the normal one for other 30% of all the
histograms. For the rest (10%) the Frechet's model is optimal.

3.2 Dynamic characteristics of both bridges

In Table 2 there are natural frequencies of the loaded and the un-
loaded bridges. It can be seen that the mean natural frequency of
the second loaded bridge is higher approximately by 1.0 Hz than the
frequency of the first one. This difference confirms the reality that
with a growth of rigidity of construction its frequency grows up.

FREQUENCY 1 st BRIDGE 2 nd BRIDGE

(Hz) MIN MEAN MAX MIN MEAN MAX

LOADED
1,000 4,700 12,500 3,900 5,667 8,820
BRIDGE

UNLOADED
BRIDGE - - - 5,560 5,759 8,200

Tab. 2 The frequencies of vibrations of the bridges

Table 3 shows the values of the logarithmic decrement of damping

1/) the dynamic coefficient (8) as well as the dynamic increment
which have been measured according to the Czechoslovak Standard [2].
Here we would like to mention the fact that the dynamic coefficients
have been greater by 3 to 10% in one direction of drive than in the
opposite one. This difference of the dynamic influences is due to
92 EXPERIMENTAL INVESTIGATION OF TRAFFIC LOAD ON HIGHWAY BRIDGES

the expansion Joint which lies on the abutments of both bridges in

one direction only.

LOGARITHMIC DECREMENT DYNAMIC C0EFFICIENT DYNAMIC INCREMENT

OF DAMPING (i*0 (<f)
BRIDGE

MIN MEAN MAX MIN MEAN MAX MIN MEAN MAX

FIRST - - - 1,05 1,165 2,10 0,015 0,085 0,195

SECOND -0,134 0,170 0,500 1,00 1,118 1,875 0,010 0,072 0,180

Tab. 3 Some dynamic characteristics of second bridge

The analysis of the correlation between the maximum deflections and

the dynamic coefficients respectively, the dynamic increment has
shown that in all casses on both bridges the negative respectively
positive coefficient of correlation (for linear as well for
logarithmic and exponential dependance) is very high and its values
are very close to 1.0. The results have confirmed that with an
increasing traffic load the dynamic coefficient decreases but the
dynamic increment increaes.

4. CONCLUSION
In conclusion we can say that for further theoretical investigation
of durability as well as a reliability analysis of a bridge
construction the Weibull's or the Gumbel' s theoretical model of
probability distribution of the deflection could be considered as
a response of the bridges to the vehicular loads. Further we can say
that we have obtained average values of some dynamic characteristics
of that type of the bridge from ample Statistical data.

REFERENCES
[1] KYSKA R. - POLLAK T.: Measurement of dynamic properties of
bridges IMEKO X. Preprint, Vol. 3 Prague 1985 pp. 155-162
[2] CSN 73 6209 "Zatezovaci skousky mostü". Praha 1980.
[3] SOKOLIK A. at al." Dynamicke charakteristiky mostov z betönu.
Zäverecnä spräva VÜ P12-526/267/E04. VSDS Zilina 1983.
[4] CLARKE G.M. - COOKE D.: A Basic Course in Statistic. ELBS
LONDON 1983.
 Remaining Structural Capacity of
Capacitö portante rösiduelle des pi
Resttragfähigkeit von Pfeilern und Wide

Tokiharu FURUYA
Civil Engineer
East Japan Railway Co.
Tokyo, Japan

N. KAMATA
+w
H. FUJIMOTO
Civil Engineer Civil Engineer
East Japan Railway Co. East Japan Railw
Tokyo, Japan Tokyo, Japan
94 REMAINING STRUCTURAL CAPACITY OF A BRICK-BUILT PIER AND ABUTMENT

1. INITODUCTION
The number of piers and abutments on the JR lines has reached 132,000 (conventional
lines), of which about 110,000 (83%) are made of brick or stone masonry
units, or piain concrete [ 1 ]. The number of brick or stone masonry piersand
abutments reaches about 32,000 (24%).
About of these structures were built
70% 40 or more years ago. Surprisingly, 90%
of brick or stone masonry structures are 60 or more years old (see Fig. 1).
X 1,000)

j3 20
Stone/
brick
24% Concrete
Others 59%
Reinforced
(1%) Concrete

(16%)
M.
10 or "~20 2I 3I <" 5«
-30 -10 -50 -60 -70
6I 71 or
7i Ui
less nore
Age
Materials Used Age by Material
Fig. State of Substructure(132,000 units)
1
(from Journal of Structurel Design Journal,No.9,February 1983)
Some of these old structures have already deteriorated because of
years of use.
Although they are being replaced with reinforced concrete structures, takes it
huge amounts of cost and time to replace all of them.
to use the old structures until they are replaced.
is inevitable, therefore, It
Japan is one of the most seismically active countries in the world and has
suffered many large earthquakes, which have resulted in substantial structural
damages.
Recent damage-causing earthquakes ine lüde the Nankai Earthquake (M8.1, 1946),
Fukui Earthquake (M7.3, 1948), Tokachi-oki Earthquake (M8.1, 1952), Boso-oki
Earthquake (M7.5, 1953), Niigata Earthquake (M7.5, 1964), Tokachi-oki Earthquake
(M7.9, 1968), Miyagiken-oki Earthquake (M7.4, 1978), Nihonkai Chubu Earthquake
(M7.7, 1983), and Chibaken Toho-oki Earthquake (M6.2, 1987) [2].
Joints in brick piers and abutments are often in the deteriorated condition due
to years of use. These deteriorated joints are vulnerable particularly to lateral
loads like earthquake loads.
It is therefore important to
evaluate the soundness, particular ly earthquake
resistance, of those old structures
and take whatever necessary steps aecordingly.
In this paper, data obtained from model experiments, which can be used in evalu-
atmg the earthquake resistance of brick piers and abutments, is reviewed. is
believed that the data will be found useful in maintaining those old structures.
It

2. FACT-FINDING SURVEY

2.1 Strengths of Brick Structures [3]

Block samples were taken from an old pier( built in 1889), and the shear strength
of the samples was measured. Photo 1 shows a similar brick pier.
 T. FURUYA - N. KAMATA - H.

2.1.1 Shear Strength

Specimens used in the loading
test are cubes measuring about
80cm x 80cm x 80cm, and their
surfaces have been chiseled.
The loading test was conducted
in the form of a double shear
test, as shown in Fig. 2.
To introduce a stress of 0.19
MPa, which corresponds to the
dead load (6374 kN), bearing
plates (to cover overall
surfaces) were installed on
both sides of the specimens,
and six r/> 16mm steel bars
were used for tensioning. Upper
Shear failures occurred at two beari
Joint faces, but the failures plate
at the two faces did not occur Steel bar'SK
simultaneously. Thus the first t
shear failure and the next one
are referred to as the first
failure and the second failure,
respectively. The results of
the shear strength test are
shown in Table 1.
96 REMAINING STRUCTURAL CAPACITY OF A BRICK-BUILT PIER AND ABUTMENT

The configuration of the setup for elastic wave measurement is lllustrated m

Fig.3.

Impact Reception
serisor sensor Oscilloscope

Jt Elastic wave Wave form O

iAjT" LJ
receiver recorder

Object

Fig. Standard Configuration of Elastic Wave Measurement System [4]

3

Two sensors were used for the detection of elastic waves: an impact sensor and
a reception sensor. The impact sensor detects surface Vibration and converts
into Signals representing the occurrence of waves. The reception sensor detects
it
waves propagated through an elastic body. In the test, the Signal for the
Initiation of waves and an elastic wave Signal were chosen at the elastic wave receiver
and stored in the wave form recorder. The recorded wave forms were then
displayed on the oscilloscope.
2.2.2 Method for Measurement of Elastic Waves
The concepts of two Standard
methods for measuring elastic
waves are shown in Fig. 4.
Usually the transmission impact sensor
method is to measure the
used

tu
impact sensor
T
^
Impact Impact
Reception sensor
velocity of elastic waves u.
transmitted through media,
and the reflection method is
used to determine the locati-
ons and sizes of discontinui-
ties in horizontal joints.
Y,-Reception sensor
V
Transmission Method Reflection Method
Often these methods are used
in combination.
Fig. 4 Standard Measuring Method [4]

2.3 Nondestructive and Destructive Tests on Brick Specimens [5]

2.3.1 Relationship between Elastic Wave Velocity and Flexural-Tensile Strength
As shown in Fig. 5, specimens were taken from a pier of a bridge on a railway
line that went out of Operation, and were dressed manually. Using these specimens,
a nondestructive test as shown in Fig. 6 and a flexural failure test as
shown in Fig. 7 were conducted.

Trigger Reception sensor

<n w
Hammer
ooo In measuring reflected waves, the trigger
300 and the reception sensor are positioned at
the corresponding points on the opposite
faces.
Fig. 5 Dimensions of Brick Specimen
(unit: mm) Fig. 6 Nondestructive Test
 T. FURUYA - N. KAMATA - H. FUJIMOTO - Y. NATUAKI 97

Fig. 8 is a plot of the results of

the tests, with the axis of ordina-
Amsler
tes measuring the flexural-tensile Loading beam
strength and the axis of abscissas
the elastic wave velocity (where Loading point
the dotted line represents a multiple
regression formula). As a pra- 330
ctical formula for the relationship
between the elastic wave velocity
and the flexural-tensile 45r ° sa
^_
strength DisplacementJQ
based on engineering judgments, the gauge
following equation is proposed: 800

(J et „ 2.65 x 10" 4
V (1) Fl9- 7 Loading Device and Loading Points
where (unit: mm)

cj et „ :flexural-tensile strength(MPa)
V :velocity of elastic wave(m/sec) 1 o

2.3.2 Estimation of Residual Strength i—•

rV
and the Locations of Defective
Regions [4]
m 0 5
When the velocity of
elastic waves ,o;8 yc i c
in an actual pier or abutment is to
be measured, transmission velociti-
es are measured at multiple points
as shown in Fig. 9 (a), and the av-
erage value of measurements is taken 0 1000 2000 3000 4000
as the transmission velocity for
the structure. Then, the flexural- V Elastic wave velocity (m/ s
tensile strengtha ct„ is calculated Fl9- 8
Relationship between Elastic Wave
using Eq. (1). Velocity and Flexural-Tensile Strength
The locations of defective regions
can be determined by use of a time-
distance curve as shown in Fig. 9(b). **
A time-distance curve shows the $
time 1
elastic waves require to travel a ce- 1
*\
d a %

rtain distance. Since the gradient pb

U
H
of the time-distance curve represents 1 UV
> C last

velocity, a greater gradient of the 5d QJ

curve means a lower velocity. The

1

oe 1
lower velocity the steep gradient
shown by ^mr
in the section b-d in Fig. 9(b) Distance from top face
suggests the possibility of disconne- (a)
cted (b)
joints.
Fig. 9 Tnne-Distance Curve
2.3.3 Estimation of the Embedment Depth of Existing Structures

L^f TT1
SsefTTL6
it
cases
tÜat dSSlgn drawln9s °f aged structures such as brick piers
^
is necessary to Xff
dUrin9
^
W3rS °r for some other
determine the embedment depth m order «aaons.
^n Ich
to evaluate tne
the
soundness of particular structures.
98 REMAINING STRUCTURAL CAPACITY OF

The embedment depth of a particular struct

the velocity of elastic waves transmitted
tom reflection from reflected waves [4].
transmitted wave forms and reflected wave

mm m
&

Photo 2 Transmitted Wave Forms

3. JUDGMENT OF SOUNDNESS (MAINLY EARTHQUA

3.1 Failure Due to Lateral Seismic Loads o

From data on brick piers damaged by earthq
tests on abandoned brick piers, the failur
 T. FURUYA - N. KAMATA - H. FUJIMOTO - Y. NATUAKI 99

4. AN EXAMPLE EVALÜATING THE EARTHQUAKE RESISTANCE OF AN ACTUAL PIER

4,1 Pier
The earthquake resistance of
several piers has been evaluated. +
In this section, a pier (8P) of
the former Fuji River Bridge on 0.611 6.83

the Tokaido Line is considered.

Test Section
7<KW
0.40 8.20 0.40 0.30 400 0.30

<D

Fig. 11 Fuji River Bridge on the Tokaido Line

(out-of-use brick pier, 8P; unit:m)
Superstructure:Theodore truss W=1433KN
4.2 Results of Earthquake Resistance Evaluation
Tables 2 and 3 show the results of the earthquake resistance evaluation. The
shear strength used here is as per ACI's shear friction theory, and /z =1.0 is
assumed.

Ground surface Natural Response acceleration Response acceleration Equivalent Average elastic
acceleration period at structure' s center at center of gravity seismic wave velocity
of gravity in test section intensity
Bridge name ^^^^ a(gal) (see) /Sa(gal) ß' a(gal) K ß' all (m/s)

Former Fuji River bridge 2 2 1 0. 3 5 0 8 5 6 6 0.578 2 5 0 0

Table 2 Evaluation Data

Bending stress (Mpa) Shear stress (Mpa)

Bridge name
^^^^ Occurrence Tolerance j
Judgment Occurrence Tolerance Judgment

Former Fuji River bridge -0.6 9

-0.6 6 x 0. 1 0 0. 1 8 O

Table 3 Calculated Stresses

5. CONCLUSION
Findings from this study can be summarized as follows:
(1) A method for estimating the flexural-tensile strength of joints in brick
structures like piers has been developed on the basis of the results of a
nondestructive test using impulsive elastic waves.
(2) The embedment depths of existing structures can be estimated by use of the
transmission method and the ref lection method using impulsive elastic waves.
(3) The earthquake resistance of brick structures like bridge piers can be eval¬
uated by the combined use of (1) and (2) above.
100 REMAINING STRUCTURAL CAPACITY OF A BRICK-BUILT PIER AND ABUTMENT

REFERENCES

1. Atsushi Murakami, "Present State and Problems of Bridge Maintenance, "Journal

of Structural Design, JNR Structural Design Office, February 1988.
2. Tokyo Astronomical Observatory, Science Almanac, Maruzen Co., Ltd., 1992.
3. Akio Kobayashi, and Tokiharu Furuya, "Earthquake-Proofing of Old
Structures,"
Technical Report on Earthquake-Proofing of Railway Facilities, Japan Railway
Civil Engineer Association, March 1982, pp. 339-357.
4. Akio Kobayashi, and Tokiharu Furuya, "Methods for Evaluation of the Soundness
(Mainly Earthquake Resistance) of Brick Masonry Piers and Reinforcement,"Structural
Design Journal, No. 89, Japan Railway Civil Engineer Association.
5. Akio Kobayashi, and Tokiharu Furuya,"Improvement of the Earthquake Resistance
of Old Structures," Technical Report on Earthquake-Proofing of Railway Facilities,
Japan Railway Civil Engineer Association, March 1985, pp. 197-213.
 101

SESSION 2

ANALYTICAL EVALUATION OF STRUCTURES

102
 103

KEYNOTE SPEAKER
104
 Evaluation of Reserve C
Determination de la röserve de c
Bestimmung der Tragre

Alan R KEMP
Prof. of Civil Eng.
Univ. of the Witwatersrand
Johannesburg, South Africa

4*.
106 EVALUATION OF RESERVE CAPACITY OF FRAMES

1. INTRODUCTION

Structural frames often possess load-resisting capacity above that assessed in the original design
due to the following reasons:
• Semi-rigid end-connections that may provide continuity where simple-supports were assumed.
• Stress-strain properties of materials, including nonlinear effects, that differ from those
originally assumed (conservative properties and partial material factors may be adjusted after
in-situ testing).
• Partial composite action in structures where this was neglected.
• Benefits of limit states design codes allowing for redistribution of moments and ultimate
(stress-block) resistances compared to older allowable stress codes, but also requiring more
comprehensive analysis including non-linear P-A effects and ductility criteria.

If adequate analysis procedures are available, these factors will often lead to an assessed increase
in load capacity. This may be improved further by strengthening procedures that enhance flexural
resistance and stiffness, introduce additional continuity and load paths, or prevent secondary modes
of strain-weakening behaviour.

The first two sections of this paper describe moment-curvature and frame analysis algorithms that
link together to provide a computational method of allowing for all these nonlinear characteristics
without the need for finite-element analysis involving numerous elements both across the sections
and along the length. Features of this approach include:
1. In the frame analysis each "member" is represented by only two "sub-members", each
reflecting the integrated non-linear behaviour between an end and the internal section of
maximum moment (or the midspan if no maximum internal moment exists), without further
discretization in inelastic regions.
2. The frame analysis identifies not only the ultimate load capacity, but also the plastic rotations
at each eritieal section before loss of moment resistance, that are required to check the
ductility, as described in the third section of the paper.
3. The behaviour of each element in positive and negative bending is determined in the moment-
curvature algorithm allowing for nonlinear material behaviour, shrinkage, creep, interface slip,
residual stresses and other effects.
4. This moment-curvature algorithm minimises the number of "elements" representing the cross-
section because it is not necessary to subdivide for strain gradient through the depth.
5. Strain-hardening followed by strain-weakening behaviour beyond the elastic region is
represented by an idealised elastic-perfectly-plastic moment-curvature relationship for frame
analysis, together with expressions for determining the available plastic rotation prior to the
moment falling below the design resistance.

The moment-curvature and frame analysis algorithms and the limit states criterion for ductility are
illustrated by the example of a three-span composite beam in Fig. 1. Although this is a relatively
simple structure, the approach has been applied to more complex sway structures involving frame
instability [1,2]. This example is also used at the end of the paper to illustrate the reserve capacity
that can be mobilised by allowing for nonlinear characteristics, including continuity in a previously
simply-supported beam using the semi-rigid end detail reflected in Figs. lb and c.

2. MOMENT/CURVATURE-ROTATION-DEFLECTION RELATIONSHIPS

2.1 Stress-strain modeis of material behaviour

 A.R. KEMP 107

1 2 3 4 5 6 7 1250
A • I I i i
A o oAo
gjTx 500tQ, O
500mrrrof
0^ cm
L 6m 6m 6m
r i
T *\ t reinforcement
rein "~^

Beam spacing 2.5m

Original
a) 3 - span beam : converted to 305x102
continuous composite beam. cieat
X25UB connection
The moment-curvature method described in this paper has ——i New end
its origins in some of the many stress-strain modeis that plate
exist [3], but has been developed and tuned progressively
b) Section dimensions
over the last decade by the author to represent
consistently the inelastic behaviour of a wide ränge of
structural components. Importantly, the results may be
integrated to determine the element properties required
for nonlinear frame analysis, including the assessment of
required ductility using a mixed method of analysis that
is described in the next section. / n

Any stress-strain curve is subdivided into three regions as A Slope C 83000kNm/rad

I (Connection stiffness)
illustrated in Fig. 2. Each region is represented by a
local curve with an origin at the Start of the region and a
relationship in region i of the following form:
Ao, - EAe, + DfAe,)"1 (la) Rotation 0
c) End connection M - 0 curve
Ae - e - e (Tb)
Fig»! Composite beam example
which Aaj and £±z-x are the changes of stress
and strain at strain e from the origin of the
region,eio is the strain at the origin, E{ is the S»
slope of the stress-strain curve at the origin
and Dj and n} model the observed
Metal (reinforcing)
characteristics of the material.

The general Eqns. la and lb for each of the

three regions are particularly useful in
calculating the moment-curvature relationship
(co
because at any extreme fibre strain € the
&J
average stress under the curve a, the location Concrete (tension
of the centroid of this area c /aede/ae2 Strain
and the tangental slope Et may be expressed Region 1
Region 2 Region 3
?M *+*
Slope E
in simple algebraic form as follows: Slope

~ö
- E A /e (2a) Fig.2 Stress-strain curves
A - oioAet + EAe(2 / 2 + DtAe,)"1*1 /(n+1) (2b)
E [Ae + o, Ae? / 2 + EAe3 / 3 +
+
D(Ae)n, 2/(n + 2)] (2c)
c -
oe'
E +nD (Ae)1 (2d)
108 EVALUATION OF RESERVE CAPACITY OF FRAMES

in which the summation extends over all regions i up to and including the region containing the
strain e, aio is the stress at the Start of the region and A{ is the area under the stress-strain curve
in any region given by Eqn. 2b. The index n{ is calculated from a known change of slope (Et -
Et) over Ao{ and Aej from Eqns. la and 2d:

n - (E - E) Ae / (EAe - Ao) (2e)

The material properties that need to be specified to determine the unknowns Ei? D{ and n4 in the
three regions of the curve apart from continuity are shown in Fig. 2 and Table 1. This table also
includes in brackets typical design values, incorporating partial material factors, that are used in
the example of Fig. 1.

2.2 Moment-curvature relationships

Any cross-section of a structural element comprising one or more materials and subjected to
bending with or without coincident axial force, may be subdivided into an appropriate number of
equivalent rectangular elements or concentrated areas to represent the geometry of the cross-
section. This subdivision does not have to be sufficiently fine to neglect the strain gradient because
of the availability of expressions for average stress and lever-arm of the resultant force vector
given by Eqns. 2a to 2c. Thus a steel I-section may be represented by three equivalent rectangles
and a reinforced concrete beam by one rectangle and one concentrated area, whereas a circular
cross-section may require twelve rectangular elements.

A linear distribution of longitudinal strain is considered through the depth of the section and the
strain gradient is assumed equal to the curvature. Considering any rectangular element j shown
in Fig. 3 of breadth b, thickness t and depth of centroid yc from an arbitrary reference level (the
top surface), the axial force supported by the element and its moment of resistance may be
determined directly from the stress-strain properties of the material described previously. The
resistance of the rectangle under consideration ABCD may be represented by the difference in
resistance of the rectangles ABEF and CDEF in this figure. For the strain distribution shown,
representing a curvature or strain gradient <f>:

Material Mod. of Proportional Strain hardening/weakening

(Fig.2) Elasticity limit fp
E Onset Slope
1. Structural fp=f es=(5-15)ey Es=E/(30-100)
Steel (206 GPa) (fy=235MPa) (es=10ey) (ES=E/100)
2. Metal fp=0.8fy es=(1.5-2)fy/E Es=E/(30-200)
(reinforcing) (200 GPa) (fy=390MPa) (es=2fy/E) (Es=E/200)
3. Concrete E=20-35GPa fp (0.5-0.8)fc Es=-E/(3-50)*
(Compression) (E=24GPa) fc 15MPa 0.002-0.0025 (Es=-E/5)
fp=9MPa (em =0.0022)
4. Concrete E as for fp=ft=0.3(fc)2/3 €s=(3-8)fp/E Es 0
(tension) compression fp 1.8MPa (e.=4fp/E)
Depends on extent of triaxial restraint provided by reinforcement
Table 1. Material properties for stress-strain curves in Fig. 2
 A.R. KEMP 109

- *yu - <YC - 0,5t - yn) (3a) Reference Level

I
eu
e, - <|>y, - (yc + 0,5t - yn) (3b)
Vn Centroid of section
in which yu and ye are the depths from Neutral Axis_Leyel
'
the neutral axis to the upper and lower i i
"a
"f"
surfaces of the rectangle defined as Rectangular i
i

C' Element in
positive below the neutral axis, €u and
ee are the corresponding strains and yn
i. z '

is the depth of the neutral axis below

the reference level. Fig.3
v—M
Assessment of element ABCD

The axial force Fj supported by element j is given by the difference in axial forces on the two
rectangles ABEF and CDEF:

F - b (o,y# - o y (4)

in which oe and au are the average stresses under the stress-strain curve given by Eqns. 2a and
2b at strains ee and eu respectively.

The moment of resistance Mj of the element about the centroid of the complete section is given
by the difference in moments of resistance of the two rectangles ABEF and CDEF:

M b (otcty2t - oucuy2u) + F (yn - ycc) (5)

in which and cu are the centroidal ratios given by Eqn. 2c and ycc is the depth of the centroid
ce
of the section from the reference level. Similar expressions may be derived for concentrated areas
such as reinforcement [4].

In all of the above equations the assumed sign Convention is stress, strain and force positive in
tension, curvature and moment positive in sagging bending and depth y positive below the arbitrary
reference level. The expressions apply whether the element is above or below or cut by the neutral
axis. For materials such as concrete the form of these equations also enables different stress-strain
curves to be adopted under compression and tension by referring oe, ce and au, cu to the properties
of different materials if the neutral axis falls within the depth of the element.

In Eqns. 3 to 5 it is assumed that the depth of the neutral axis below the reference level yn is
known. At inelastic levels of stress and strain, yn is determined iteratively to satisfy equilibrium
between the net axial force resisted by the section, £Fj, and the external axial load N applied to
the section.

The method has been applied [4] to a wide ränge of structures and elements of different materials
at serviceability and ultimate load and has been adapted to model the following important aspects
of structural behaviour:
• Prestressing and shrinkage in concrete elements, by introducing initial values of strain in the
relevant elements.
• Residual rolling or welding stresses, by providing additional elements with different initial
strains to reflect the residual stresses approximately, based on proposals such as Young [5].
• Interface slip in composite beams with partial shear connection, by introducing slip-strain
increments at the interface that are a function of curvature and, when integrated over the half-
110 EVALUATION OF RESERVE CAPACITY OF FRAMES

span of the beam, provide a total end slip that is consistent with measured behaviour in push-
out or beam tests.
• A combination of differing section conflgurations and stress-strain properties under permanent
and imposed loads.
• Strain-weakening due to interactive local and lateral buckling of yielded steel sections based
on a semi-empirical model described by Kemp [6].

2.3 Idealisation for frame analysis

The moment-curvature relationships for the composite beam example shown in Fig. lb, with
material properties defined in brackets in Table 1, are illustrated by the solid lines in Fig.4 for
positive and negative bending. The ultimate design resistance Mp is an important parameter that
is determined from code rules in this case, or other considerations. These curves include the
effects of residual rolling stresses, interface slip due to a 50% partial shear connection and
interactive plastic local and lateral buckling at high moments in the negative moment region. The
region of the beam adjacent to the semi-rigid end connection shown in Figs. lb and c has been
modelled by assuming that the steel section cannot resist tension.

In the three-span continuous beam of Fig la, the first regions to develop plastic hinges are adjacent
to the internal Supports. Uncertainty exists over the moment-curvature path followed by sections
in the inelastic region adjacent to these Supports (represented by portion PM in Fig. 4b) once the
structure is loaded under displacement control beyond the maximum moment and experiences
strain-weakening with the section of maximum moment following portion MM1 in Fig. 4b. Based
on a qualitative assessment of test behaviour it is proposed that the distribution of curvature in this
inelastic region at the load level at which the maximum moment at the internal support falls to the
design resistance Mp may be modelled simply and approximately by the line P'M1. This implies
that curvature is a linear function of moment and the available inelastic rotation 8ap is given by the
shaded area in Fig. 4b as:

6 - 0.5M (m -1) [*' - * (m +1) / 2m]/mV (6)

M =200 M„=88.
M„=76
M=152
£££S£
Mter stra
"in_weakening

9ap=shaded area
Elastic slope Cracking Idealised
=EI=13000kNm2
"° 20
CD
CO
0m
0 ,e«

01 02 03 04 05 0 0 2 0 3 0.4 0 5

Curvature (m-1) Curvature (m-1)

a) Positive (sagging) bending b) Negative (hogging) bending

Fig- 4 Moment-curvature relationships for example of Fig. 1
 A.R. KEMP 111

in which Vp is the shear force at the section of maximum moment and m Mm/Mp in Fig. 4b.

On this basis the behaviour of elements of the structure between sections of zero moment and
adjacent sections of maximum positive or negative moment (or adjacent joints) may be modelled
in the frame analysis described subsequently as the superposition of the following effects:

1. Elastic behaviour is represented by the slope of the dashed elastic region of the idealised
moment-curvature relationship OP in positive bending EI Mp/<f>p.
2. Inelastic behaviour is modelled by the horizontal dashed line PMf representing a concentrated
ideally-plastic hinge of moment capacity Mp with an available rotation capacity 6ap equal to the
shaded area in Fig. 4b and given by Eqn. 6.
3. Differences in stiffness between positive and negative moment regions including cracking of
concrete adjacent to internal Supports, is represented by an additional component of available
plastic rotation 6acr given approximately by the integration of a linear difference in flexural
rigidity between the positive moment EI and the cracked negative moment EI1, as follows:
6
acr - 0 e [1 - (M er / M p )2] [ (El / El') -11 (7)

in which is the calculated elastic rotation in the negative moment reJgion and Mcr is the
6e
moment at which cracking occurs.
4. Semi-rigid end connections are represented by an idealised elastic stiffness C in Fig. lc and,
in cases where the connection rather than the adjacent member is the eritieal flexural element,
by identifying the available plastic rotation 8ac of the end connection prior to the moment
falling below the design value Mp as shown in this figure.

3. FRAME ANALYSIS

3.1 Mixed flexibility/sway-deflection method

The selection and development of a mixed flexibility/sway-deflection method of analysis arose

primarily from convenience in modelling elastic and inelastic behaviour, differences in negative
and positive moment characteristics and semi-rigid connections, as outlined in the previous section
[1,2]. By selecting end moments and independent sway-deflections as unknowns in frame analysis
all of these properties including the development of plastic hinges can be considered on a consistent
basis without changing the number or location of the unknowns. Additional benefits of this
approach include:
• In nonlinear analysis the use of end moments as unknowns rather than Joint displacements is
likely to lead to more stable and sensitive Solutions.
• Axial forces in members are determined from equilibrium considerations rather than as a
stiffness function of axial distortions.
• The unknown independent sway-deflections are all directly related to the plastic and instability
mode shapes of the structure and therefore relevant to inelastic P-A methods.

The method is described firstly in terms of the unknown end moments and then in terms of the
additional unknown, independent sway-deflections.

3.2 Solution equations for unknown end moments

In the substrueture shown in Fig. 5 that is used for illustrating this approach, ij is one of the two
112 EVALUATION OF RESERVE CAPACITY OF FRAMES

sub-members representing member ijn between sections of maximum moment (or midspan if there
is no internal section of maximum moment). The matrices relating end rotations 0 to end moments
M and relative end deflections due to sway 6S (all positive anti-clockwise) in ij are:
/ ^
0 L (b/3 + f) -b2/6 b IL
ij
" 31] (8)
0 El -b2/6 (b/3 + f) b IL
V \0

in which L and EI are the length and elastic flexural rigidity of the member, bx and b2 are Berry
stability functions [7] that allow for increased flexibility due to axial force and have values of unity
when second-order P-A effects are neglected and f is the non-dimensional flexibility ratio of the
end connection expressed in terms of the elastic stiffness of the connection, C in Fig. lc :

f - El / CL (9)

The following equations are used for solving the n unknown moments at the ends of the n members
meeting at Joint i in Fig. 5a (n 4 in Fig. 5a) :

1. One equation representing equilibrium of

the moments, M at the ends of the
members meeting at Joint i: --m
m

Vm-m+m+m+m
L-*t i ij ik i/ im - M' i
(10)

in which M,l is the magnitude of any

external moment applied to Joint i.
k -
2. (n-1) compatibility equations expressing
the equality of rotations at the ends of a) Substructure and sub - members
each pair of members meeting a Joint i
(three compatibility equations for Joint i
in Fig. 5a representing equal rotations at
1
\fcsü
end i of members ij and ik, ij and U, and
ij and im), as follows: xeo
Oefteo SögS
B
ij - Bkik ; B ij - ü,\i and Hij - B
im
(11)
v ' ' *A
\X0
in which, for example in sub-member ij,
% ö. + e'ij + 0rij and 0jj is the
superposition of B£j (the end rotation due
to unknown end moments given by Eqn.
8), 0'ij (the end rotation due to member b) Sub - member ij
loads with end moments equal to zero -
i.e. simply-supported) and 0rij (the
Fig.5 Sub-member ij at Joint i

required plastic rotation of a hinge if it

develops at this section, as discussed
subsequently). In considering P-A effects
the end rotations 0' may be amplified due
to axial force using Berry functions [7],
 A.R. KEMP 113

The only exception to this subdivision of equilibrium and compatibility equations occurs at a fixed
support where all n equations will be compatibility equations.

3.3 Solution equations for unknown sway deflections

The independent modes of sway are treated as additional unknowns and the approach therefore
becomes a mixed method combining unknown end moments with unknown independent sway
deflections.

In a structure with j joints, m members and r constraints to global translation at support joints
(horizontal and vertical components), there are s independent components of sway deflection given
by:
s - 2j - m - r (12)

A systematic approach has been developed by the author [1] for identifying the most appropriate
unknown Joint translations to represent these s unknown modes of sway and their relationship to
the other Joint translations. Typically there is one mode of sway representing plastic collapse of
each member and one representing each mode of sway instability. In this algorithm conventional
sway-equilibrium equations are used to solve for each unknown sway deflection and are derived
by the Principle of Virtual Displacements applied to a Virtual free body displacement of the
structure in the mode of sway. These sway equilibrium equations are expressed in terms of:
• the unknown end moments in the sub-members,
• P-A terms for elastic or inelastic stability analysis in which P are the axial forces extrapolated
from the previous iteration and A are the unknown sway deflections normal to P.

3.4 Application of this frame analysis method

This mixed flexibility/sway-deflection method may be used in the same form for elastic, elastic-
plastic and elastic-plastic-instability (P-A) analyses as follows:
1. Elastic analysis (including elastic connection flexibility) : The unknown end moments are
solved using the idealised (dashed) elastic properties in Fig. 4a and the Joint equilibrium and
compatibility Eqns. 10 & 11 and the unknown sway deflections using conventional sway-
equilibrium equations.
2. Elastic-plastic analysis : After the elastic analysis, plastic hinges may be introduced at eritieal
sections either by defining the ultimate design resistance Mp or by identifying a speeified ratio
between the moment capacities at two sections and the location at which a hinge is expected
to develop. In either case the moment at the plastic hinge is then known and is replaced as
unknown by the required plastic hinge rotation 0r (forming part of Eqn. 11) to aecommodate
the necessary redistribution of moments, as discussed subsequently.
3. Elastic-plastic-instability and P-A analyses : The elastic-plastic analysis approach is extended
to include the Berry stability functions in Eqns. 8 & 11 and the P-A term in the sway-
equilibrium equations. The member axial force is extrapolated from the previous analysis Step
introducing a limited iterative procedure.

4. REMAINING CAPACITY ASSOCIATED WITH REDUNDANCY

A significant source of reserve or remaining capacity may exist in redundant structures if they
possess sufficient ductility to redistribute moments from heavily stressed sections, at which the
114 EVALUATION OF RESERVE CAPACITY OF FRAMES Jt
ultimate stress-block moment Mp is achieved, to less heavily stressed sections as the load is
increased. The two preceding parts of this paper have described relatively simple algorithms for
determining results that are directly relevant to assessing ductility, namely:
1. The available inelastic rotation 0a at plastic hinges prior to the moment falling below the
ultimate resistance Mp and made up of components due to plastic behaviour 0ap (Eqn. 6),
inelastic rotation of end connections 0ac (Fig. lc) and reduced stiffness due to cracking of
negative moment regions, 0acr (Eqn.7).
2. The required inelastic rotation 0r at the same plastic hinges from the frame analysis of the
redundant structure for specified loads and ultimate resistances Mp.

A general limit states criterion of ductility has been proposed [2] for determining how much
capacity exists in indeterminate structures to redistribute moments. This requires that the available
inelastic rotation 0a at each plastic hinge should be greater than the plastic rotation 0r required to
achieve the specified level of moment redistribution or hinge development:

i.e. LIMIT STATE OF DUCTILITY : (0a /

yj > 0r (13)

in which Ymd is a partial material factor to allow for the considerable uncertainties in assessing 0a
and 0r, in the ränge 1.5 to 3 depending on whether it is a ductile or brittle mode of failure. This
criterion may be expressed non-dimensionally in terms of rotation capacity by dividing both sides
by the elastic rotation at Mp between the plastic hinge and the adjacent section of zero moment.
The criterion conforms to limit-states terminology by having a resistance on the left hand side and
an action effect on the right hand side, and may be applied to any mode of failure inhibiting
ductility for all types of structural materials.

5. EXAMPLE

Consider the three-span steel beam illustrated in Fig.l that was originally designed as non-
composite and simply supported with cleat connections only resisting shear, but is to be considered
for upgrading into a continuous composite beam. The slab contained 0.4% longitudinal
reinforcement over the internal Supports for crack control and this will be utilised. The spans are
about 2/3rds of a typical full-scale beam, but reflect the dimensions of a similar specimen tested
by the author for available plastic rotation [8]. The nominal load carrying capacity of the existing
beam is 6kN/m of permanent and 2 kN/m2 of imposed load on the basis of an existing allowable
stress code for steel, assuming the compression flange is restrained against lateral buckling.

Upgrading will be achieved by cutting out slots in the concrete above the steel beam to accomodate
a partial shear connection capable of mobilising 50% of the ultimate slab force in sagging bending
and the füll effective area of reinforcement over the internal Supports (500 mm2). Each beam will
be propped at midspan during casting of the grout around the shear connectors and welding of the
end plate to the bottom flange of the steel beam and adjacent web as shown in Fig. lb. This end
plate provides a semi-rigid end connection with moment resistance made up of the reinforcement
in tension and the bottom flange in compression when the prop is released [8].

The evaluation of the enhanced load-carrying capacity at the ultimate flexural limit State is
undertaken as follows:
• The properties of rows 1 to 4 of Table 2 are determined using the moment-curvature results
in Fig. 4 allowing for interactive plastic local and lateral buckling in negative bending and
concrete crushing in positive bending.
 A.R. KEMP 115

• The properties in rows 5 and 6 of this table are calculated from these values using Eqns. 7 and
6 respectively.
• The elastic flexibility ratio f of the end connection (row 7) is determined from experimental
results (Fig. lc) using Eqn.9 : no significant nonlinear component of connection rotation was
apparent at the level of moment developed in the adjacent member, so 0ac 0 in this case.
• An ultimate load capacity of the three-span beam of 41.8 kN/m associated with plastic hinges
at the internal Supports and midspan region of the outside spans, is determined from the frame
analysis, in two Steps (elastic and plastic) using 6 sub-members (Fig. la) and 12 unknown end
moments and 3 unknown sway deflections. This represents a more than three-fold increase
in the existing imposed load capacity of this beam.
• Required plastic hinge rotations in the outside spans adjacent to the internal Supports 0r
0.0135 rad. are obtained from the frame analysis and are used as the action effect in Eqn. 13
to check the ductility involved in plastic moment redistribution. A partial material factor of
Ymd 1-5 is adopted for ductile failure due to local and lateral buckling and the resistance
effect is the available plastic rotation 0a 0.0216 rad. made up of 0ac (equal to zero), 0acr and
0ap (from Table 2, rows 5 and 6).

Limit States Criterion (Eqn. 13) : (0.0216 / 1.5 > 0.0135)

The available plastic rotation 0a compares favourably with tests on composite beams with similar
semi-rigid end connections [8]. This excellent ductility is explained in this reference by the
location of the plastic neutral axis being close to the compression flange which severely inhibits
local and lateral buckling.

Moment-curvature properties (Fig.4) : +ve Moment Region -ve Moment Region

1. Elastic flexural rigidity EI (kNm2) 13000 13000 (+ve moment)
2. Ultimate design resistance Mp (kNm) 152 76 (semi-rigid)
3. Maximum moment Mm (kNm) 200 88
4. Falling branch curvature <r>'m (m"1) 0.56
Assessed properties:

5. Plastic rotation (Concrete cracking) N/A 0.0017

0acr Eqn.7 (rad.)
6. Plastic rotation (yielding) 0ap Eqn. 6 N/A 0.0199
(rad.)
Experimentally measured properties:
7. Elastic flexibility ratio of end
connection f EI/CL (Eqn.9) N/A 0.05

Table 2. Properties required for frame analysis

6. CONCLUSIONS

A significant source of reserve capacity exists in many structures if the implications of inelastic
material behaviour, continuity and plastic redistribution of moments as well as the ductility
requirements, can be assessed analytically without resorting to finite elements modeis that require
116 EVALUATION OF RESERVE CAPACITY OF FRAMES

numerous elements both across the section of the members and along their length. A twin
algorithm is illustrated in this paper for assessing firstly the inelastic section properties of two sub-
members per frame member, and secondly the elastic-plastic-instability analysis of frames
comprised of these sub-members. These analyses also identify the available and required plastic
rotations that are used to check adequate ductility using a simple limit states criterion involving
material plasticity, inelastic properties of end connections and differing flexural rigidities in
positive and negative bending.

7. REFERENCES

1. KEMP A.R., A Consistent Mixed Approach to Computer Analysis of Frames. Civil Engineer
in South Africa, 30(7), 317-322, July 1988.
2. KEMP A.R., Quantifying Limit States of Rotation Capacity in Flexural members. Proc.
Institution of Civil Engineers, 89(2), 387-406, Sept. 1990.
3. DESAI CS. and SIRIWARDANE HJ., Constitutive Laws for Engineering Materials.
Prentice-Hall, New York, 1984.
4. KEMP A.R., Simplified Modelling of Material Non-Linearity in Structural Frames. Civil
Engineer in South Africa, 30(9), 425-432, Sept. 1988.
5. YOUNG B. W., Residual Stresses in Hot-Roiled Sections. Dept. of Engin., Technical Report
No. CUED/C - Struc/TR8, University of Cambridge, 1971.
6. KEMP A.R. and DEKKER N., Available Rotation Capacity in Steel and Composite Beams.
The Structural Engineer, 69(5), 88-97, March 1991.
7. PIPPARD A.J.S. and BAKER J.F., The Analysis of Engineering Structures. 3rd Edition,
Edward Arnold, London, 550-555, 1957.
8. KEMP A.R., TRINCHERO P. and DEKKER N., Ductility Effects of End Details in
Composite Beams. Engineering Foundation Conference on Composite Construction, Potosi,
Missouri, June 1992.
 117

SELECTED PAPERS
118
 Is there still Life after the L
Est-il possible de prolonger la duröe d
Gibt es lebensverlängernde Mas

Ad van der TOORN

Civil Engineer
State Public Works
Utrecht, The Netherlands
120 IS THERE STILL LIFE AFTER THE LIFETIME OF SHEETPILING Jk
1. INTRODUCTION
All civil engineering structures detenorate, so don't trust the one who
tries to seil you a 'mamtenance-f ree' structure.
Only the scale of time in which the agemg-processes takes place can vary
and so can save or foolish us.
The ancient pyramides, though still
in function if
we have not robbed them,
suffer from a substantial surface-damage when given a nearby view.
For the more ordinairy structures of our times we don't have to wait so
long.
Civil engineering structures like roads have the shortest lifecycle (15 -
25 by the wear and tear of
year), primairy caused the traffic, but speeded
up by the ever growing intensity and a bad quality of subsoil.
Sheetpiling-structures, allthough designed for a fifty years or more, often
appeares to have a much shorter lifetime (10 - 25 year), because of the
much more aggressive environment.
Corrosion-velocities of 0.25 mm per year with maxima in the order of 0.5 mm
are measured, not only along the seashore but also for mland polluted
canals.
When design or buildmg-failures makes airsupply to the backside of the
sheetpiling possible, these velocities will nearly double.
So even heavy walls with a steel thickness of 10 mm or more are in that
case of 'a short breath'.
Most of our civil engineering structures have a protection layer on the
actual bearing construction. For example a pamt-coating on a steel bridge,
a concrete cover on buildings or the armour-blocks on the slope of a dike.
So degradation is from a so called 'two-stage-mechanism', in which the
damage of the first stage is a warnmg-bell for the starting attack of the
second underlaying structural more essential part.
In contrary a sheet-piling structure most of the time is from the type
'one-stage-mechanism', that is degradation (corrosion) allmost Starts from
the very early beginning.
It's a lucky circumstance that in contrast with others, this degradation
(corrosion) is rather easy to measure by way of ultrasonic waves or more
destructive by drilling or oxygen burmng followed by a normal thickness
measurement.
So on the side of 'the assessment of the conditlon-parameters' there are
less problems than on the side of 'the assessment of structural (reserve)
capacity' but above all 'the adjustment of acceptable risk'•

^w
\— 4
r
SS

Fig.1 Structural reserve and acceptable risk

 A. VAN DER TOORN 121

2. THE STANDARD DESIGN METHODS

In normal practice a sheetpiling wall is at best designed as a two-dimen-

sional structure, which has an infinite extent in the third dimension.
The Variation in this third dimension, with respect to loads, geometry,
soil and construction properties is considered to be small or to be brought
into aecount in the Variation of the other two dimensions.
Even local loads or anchors are translated in a kind of equivalent line-
loads, so the third dimension can be neglected.
For the two-dimensional computation of sheetpiling walls there are a few
analytical or grafical methods available, like Rowe, Brinch Hansen, Blum.
Because the mechanical problem is in fact staticaly indetermmed, these
methods give approximate Solutions under certam assumptions like infinite
rigid piling and only hydrostatic active or passive earth pressure.
More recently developed Computer programs (like the dutch programs MSHEET
or DAMWAND/3) do take into aecount the stiffness of the sheetpiling and the
stiffness of the soil by bi-lmear Springs (dependent on the horizontal
displacement they come at last in the active or passive plastic stage).
The input for these computations is in general as follows:
Given or assumed by experience: - GEOMETRY (H,h,a,/3)
- SOIL PROPERTIES ((f>,c,-y,8)
- LOADINGS (q,F)
Estimated by rules of thumb : - SHEETPILING (1,1)
Than computation results in : - BENDING MOMENTS M(z)
- SHEERFORCES D(z)
With the admissible stresses : - MODULUS OF RESISTANCE (W min.)
- THICKNESS OF BODY (t
min.)
If the estimated profile doesn't fit, design is repeated with a better one.
losses ->

I,W residue
Fig General eetp ng wa Fig.3 General corrosion profile
The actual sheetpiling profiles are mostly heavier than strictly needed
with respect to the computation.
If extensive corrosion is expected the designer will take a few millimeters
more (if he is aware of the phenomenon and in the possition to do so!).
But in fact at that moment he has first to answer the difficult question:
What could and should be the minimum thickness of flange and body at
the end of the designlife in relation to the function of the
sheetpiling, so the consequencies of failure, the influence of inspection
on this, the ability of a new (more plastic) equilibrium, etc.

If ram-ability of the sheetpiling in that specific soil is expected to be a

problem, this can result in a heavier profile too.
The sheetpiling that finally will be found at location still can differ
from the one selected above, because of delivery problems, sheetpiles in
stock of the contractor, problems with achieving the right depth etc.
This 'as build' data should be saved well in a kind of birth-register
because it is of great importance for the reassesment of the structure!
122 IS THERE STILL LIFE AFTER THE LIFETIME OF SHEETPILING

3. THE DEGENERATION PROCES

In the case of sheetpiling composed of steel profiles like Larssen, Hoesch

etc. the degeneration process is mainly ordinairy corrosion. Besides steel
there are two elements needed for the Initiation and the continuation of
this corrosion process, namely oxygm and water.
So in general a vertical corrosion profile is found with two maxima, one
just beneath the low water level and one in the splash zone (Fig.3).
The first is limited by the amount of oxygin, and the other by the amount
of water.
Also in horizontal direction there will be found a wide spread in loss of
material.
The first fluctuation is measured within one Single plane of a sheetpile
and may be in the order of 100 mm due to local steel properties.
The next fluctuation is found between flanges, bodies and edges. Especially
for cold-rolled profiles at the deformed outward corners, the grid is 'open
for corrosion', so warm-rolled profiles are prefered in cases of extensive
attack.
Over the sheetpiling wall there may be spots working like anodes and others
like cathodes, caused by the metal composition, deformations and soil
properties. The addition of copper and other more precious metals ment to
prevent or to decrease corrosion, after some times proves to intensify this
anodic and cathodic spots perhaps due to unequal alloys.
The largest fluctuation in horizontal direction may have its origin in the
location or use of the sheetpiling wall. Tidal streams in combination with
fresh water tonques, manoeuvrmg ships, etc. may cause a tendensy to vary
over distances in the order of a 100 meters.
It will be clear that there is a discision problem. On the one hand in the
case of too little Information (thickness measurements) it is impossible to
make distiction between the sources of above mentioned fluctuations.
So this 'all on one heap' approach will lead to an overestimated loss of
thickness and so to an underestimation of the remaining strength.
At the other side more Information will ask for money, but may lead to a
better understandmg and probable to a longer residual service life.
The smallest fluctuations over one plane are just of mterest for the
moment that minimum thickness will become zero and so loss of soil material
may Start, because strength will depend on the mean value.
The largest fluctuation in the order of the construction length, sometimes
will lead to a seperate consideration and perhaps measurements for parts of
the wall.
So in that case only the spread in the midrange-variations remains of
direct mterest for the reassesment of sufficient strength.
Besides the decreasing thickness (that influences strength), the geometry
and the loads may change in time too.

Geometry may differ from design because of dredgmg, scouring, additions,

so in case of doubt measurements like sounding the bottum can give insight.
Keep in mind taht the computation and so the behaviour of sheetpiling is
more sensitive to the retaining height H than to the thickness t!
Loads may differ from design because the destination of the adjacent Site
may be different (gravel storage is not covered by the often arbitrairy
chosen one ton per Square meter!).
Allthough soil properties will hardly change, original design assumptions
may be of an arbitrary or global level. Supplementary measurements may give
a better insight in the present Situation.

After the state of the sheetpiling wall is well mapped, the evaluation will
finally Start.
 A. VAN DER TOORN 123

THE REASSESMENT OF SHEETPILING WALLS

4.1 The general concept

In principal during the design stage of a structure there ought to be made
a weighing between the initial Investment plus expected maintenance cost of
this new structure on the one hand versus the risc involved with the loss
of functions on the other hand and so lookmg for the total cost Optimum in
the life-cycle.
Allthough this is a sound economic concept it is hardly been done.
As in most of the cases there are design codes or at least practical rules
that relieves the designer of this difficult economic approach.
These codes of pr actice prescribes certain safety-margins that Covers the
above mentioned b alance 'on the safe side', that is a rather low risc
gained by a little
bit more Investment (Fig.4).
The consequencies of this practical concept is that during lifetime the
risc won't dominat e so fast and we may say 'nature is mild'!
These safety-margi ns are historically grown and reflects the level of pros-
perity and the ave rsion of society agamst structural failure, because this
risc is extremely low in comparison with others [1].
For existing structures things have changed even without deterioration!
Design values for geometry, loads, soil and material properties may be
known better by measurements, deteriorated construction properties are more
scattered than before, consequencies of failure can be better estimated but
in comparison with design, cost of adjustment are now of higher order.
Yet for existing structures the engineer still tries to hide behmd design
codes, because that is the easiest and common way but unfortunately in many
cases no more practicable.
As special codes or rules for existing structures are hardly available at
this moment, every engineer has to do this unknown exercition himself.
First he tries to uncover all hidden structural reserves. Material reserves
like differences between 'as build' and 'as needed' profiles or mechanical
reserves applying more advanced modeis. But if deterioration is extensive
this won't be enough.
Then trymg to exploit reliability-reserves, he will be confrontated with
the basic questions about safety-margins and acceptable reliability-levels
for existing structures in relation with design-values.
The considerations may be:

Well known loads and resistance by way of measurements may lead to

less Variation so to a smaller safety-margin with equal probability
of failure (Fig.6)
A shorter residual lifetime may lead to lower extreme loads (if time
dependend) and less loss of material so to smaller safety-margins,
ending up with a margin for temporary sheetpilings (if known!).
More costly maintenance measures together with better known failure
consequencies may lead to a higher probability of failure (Fig.5) so
to a smaller safety-margin but should not exceed other social riscs.

¦r

/
codes othe*. msks
M o /V*ot0
/S .t |,>vrK-
<W
sea* opt ~^*
KtSc Ä/ic

SicM €*!* MA <^e&Ain^C£

Fig ,_4 Economic balance Fig.5 Economic balance Fig.6 Safety-factor vs.
in design stage for maintenance failure probability
124 IS THERE STILL LIFE AFTER THE LIFETIME OF SHEETPILING

4.2 The evaluation of sheetpiling wall

a

The reassesment of a sheetpiling wall follows the above given concept.

First the engineer tries to interprete all the thickness measurements in

terms of a 'representative value' at the different levels. For instance
using the 5%-value of the normal probability function. So overthickness
given in design or construction stage will be included. With this new
adjusted values and all others like in design he tries to passes through
the 'design-loop' with the ordinairy two-dimensional computation.
If this still doesn't satisfy the present-day design rules, the more hidden
and so less computable structural reserves will be taken into aecount.
In vertical direction there may be a certain redistribution of moments, if
a more plastic behaviour of the sheetpiling in the computation is possible.
There are Computer programs written for the design of concrete retaining
walls, that can handle yielding moments (for example the dutch program
DIEPWAND/1). Yet reduction of the fieldmoment leads to increasing (acceptable?)
moments and forces about tie level and the fixed end (Fig.7).

Also in horizontal direction there is a possibility of a certain redistribution

between the individual sheetpiles.
Less corroded piles will take over a part of the load on heavy attacked
sheetpiles. In fact in the third dimension the wall can be seen as a structure
with a certain amount of parallel elements. What exactly will be the
zone to be mobilized by a weak pile depends on the given local Situation.
The stiffness of the wale, the geometry and the soilproperties play an
important role in this.
A three dimensional computation (with the dutch program DIANA) has proved
that in particular for the case of an anchored sheetpiling this horizontal
redistribution may be considarable (Fig.8) [2].
Though this mobilized zone is in the order of the retaining height of the
sheetpiling it
has not been possible yet to derive a general rule of thumb.

The expectation is that within this zone Short and midrange fluetuations
may be ignored and only the mean value of the thickness have to be taken
into aecount.
So in opposite if combined measurements are always done
mobilized zone, the mean value of that thickness may be
for such a to be
of direct use in
the normal two-dimensional computation.

Y/x&mw*
WertMZffö -z°HE

{{? >

¦ Met**.

Fig.7 Redistribution of Fig.8 A 3-dimensional FEM-model

bending moments of a sheetpiling wall.
 A. VAN DER TOORN 125

Now a lot more hidden structural capacity is brought into sight it is

that the sheetpiling wall satisfies the present-day
possible design rules.
If not it has been argued that for existing structures and time passes by
safety-margins may decrease with relation to the original design-values.
The arguments already given above are shortly:

1. A better knowledge about strength and loads so less uncertainty.

2. A shorter residual lifetime so less extreme loads and losses.
3. More costly (maintenance) measures i.r.t. design changes.
4. Better known failure consequencies i.r.t. design Starting-points.
Quantifying these arguments leads to:
ad.l Main contributions to the failure of a sheetpiling are given by the
soil properties (<^,c), the retaining height (H), the thickness of
the sheetpiles (t) and if applied the anchors.
Measurements may give the actual coefficient of Variation (0.1-0.2),
so with reliability theory it may be possible to check if an other
safety-margin may be applied (Fig.6). But since design was based on
traditional building codes, first a calibration is needed to know
the hidden starting-points in terms of coefficients of Variation.
ad.2 As sheetpilings usually are not designed for time varying loads, the
only benefit could be the time dependend loss of material. But again
in the traditional design it is not clear which part of the total
safety margin was reserved for this.
ad.3 In general maintenance measures are from a higher order (factor 10)
in relation with design measures havmg the same effect on risc.
So the new cost Optimum will result in an higher probability of
failure (factor 5) and so m a reduced safety-margin (factor 1.1).
ad. 4 If failure consequencies are from a lower order (factor 10), the
probability of failure may rise with the same magnitude for constant
risk. This may lead to a reduced safety-margin (factor 1.2). But
again the original starting-points are not known.

Although tendencies are clear, the traditional safety-margin used like a

'dust-bin' makes it
hard to pay the individual aspects.
Only calibration of traditional designs based on probabilistic methods
taking into aecount al relevant parameters and used mechanical model may
give better insight! In Holland this study is now underway [3,4].
So in
now in
the meantime a more arbitrairy reduction factor up tili 1.3 is used
practice, mainly affected by the consequencies of failure [5].
126 IS THERE STILL LIFE AFTER THE LIFETIME OF SHEETPILING

5. STRETCHING THE REMAINING LIFETIME

Now
any
it'life
has been proven by measurements and calculations that
after death', the responsible authority should be
there is still
informed about
the best way to manage that structure in the future. So the next step is
the weighing of alternative scenarios.
There are a few technical possibilities to Stretch the residual lifetime of
sheetpiling by applying a (combination of) preventive maintenance action(s)
like the wellding of plates or beams to seal and strengthen the sheetpiles,
painting and cathodic protection to slow down the corrosion proces, ground-
injection to stop losses of soil etc.
Each alternative has his own cost and expected Stretch of lifetime.
Cost may be the direct cost of the maintenance action plus cost ahead to
sustain or maintain this action.
Cost also contains the risk involved with this Solution, by which risk is
the probability times the consequences of function loss (mserviceability
or failure). On its turn this probability is dependent on the frequency of
inspection, which again represents cost.
On the other hand the expected lifetime is inflenced by these inspection
and maintenance actions too.

It is up to the engineer to bring all this in the right weighing within one
scenario and next to balance these scenarios against the zero Option 'doing
nothing', so replacement after certain time [6].
Now this complex desicion may be sustained by some analytical or Markovian
modeis which brings into aecount cost, lifetime and interactions [7,8].
Nevertheless this rational approach there are often practical restrictions
like budget-shortage and traditional philosophies that dietates the real
life, especially when no one is responsible for the total life-cycle cost!
REFERENCES

1. CIRIA Report 63. Rationalisation of safety and serviceability

factors in structural codes. July 1977.
2. BIJNAGTE J. e.a., The consequencies of corrosion for the computation
of a sheetpiling wall (in dutch). Grondmechanica Delft. May 1992.
3. CALLE E., Safety of cantilever sheetpiling structures (in dutch).
A probabilistic method for the determination of partial safety
factors. Grondmechanica Delft. May 1987.

4. SPIERENBURG S., Safety of anchord sheetpiling structures (in dutch).

A probabilistic method for the determination of partial safety
factors. Grondmechanica Delft. June 1988.

5. TOORN van der A. The evaluation of existing sheetpiling walls (in

dutch). Rijkswaterstaat-Bouwspeurwerk. June 1989.
6. TOORN van der A. Management and maintenance of sheetpiling walls
(in dutch). CUR-Rijkswaterstaat. Jan.1992.
7. WIJNMALEN D. e.a., Minimisation of maintenance cost by flexible
strategies (in dutch). TNO-FEL. Dec.1990.
8. WALTMANS M. e.a., Computation of an optimal maintenance strategy for
a sheetpiling wall (in dutch). ICIM. July 1991.
 Retrieval of System Properti
Determination des caract6ristiques de
Ermittlung der Systemeigenscha

Amin HELOU
Prof. of Struct. Eng. _
jtKtK
Jr
AN-Najah University
West Bank - via Israel
¦

¦ mm
128 RETRIEVAL OF SYSTEM PROPERTIES OF EXISTING STRUCTURES

1. INTRODUCTION

Environmental attacks, corrosion and prolonged use of existing structures make their structural
evaluation rather limited because their members' properties may not conform to the design values.
Hence, classical methods of structural analysis become inadequate to tackle and overcome the
difficulty involved. Therefore, it is both necessary and prudent to improve such methods. In this
study, system identification techniques are introduced. In such techniques, the structural stiffness
is recovered from known forces and known associated displacements. Once the stiffness matrix of
a structure is determined, the internal design forces due to any loading condition can readily be
obtained.

2. STATEMENT OF THE PROBLEM AND THE SOLUTION

Present methods of structural analysis are primarily based upon the stiffness methods of analysis
in which the input is a family of stiffness coefficients presented in a matrix form and the loading
conditions entered in a vector form. The unknowns are displacements and subsequently internal
forces. The Standard mathematical representation of these three variables is:

{f|-MM (1>

in which
F is an Nx1 loading vector
K is an NxN stiffness matrix
x is an Nx1 system displacement vector
N is the number of degrees of freedom

In the traditional approach to structural analysis, {x} is the unknown, whereas in this study the
unknowns are the elements of K which in some sense represent the characteristics of the
structure. A process that has been developed for other engineering disciplines, but which is being
introduced in structural engineering, is generically referred to as "System Identification". It is an
attractive procedure to formulate and improve mathematical modeis.

To illustrate the derivation of the stiffness of the structure in terms of the applied force vector and
the associated and measured displacements, the following Situation is used:

—F —R
—>x?

\
Figure 1

Figure 1 shows a two-degree of freedom system in which two lumped masses are attached to three
linear Springs with stiffnesses kv k^ k3
 A. HELOU 129

The force displacement relation for this Situation is written in the following form:

(2)

For an exact Solution, the following Statement holds true, i.e.

(3)
~K *2+A3

However, when this is not the case equation error vectors can be defined as:

/q+Ag -l<2
(4)
-*2 V*3 *z

To obtain an error function the right hand side of equation (4) is squared and the result is then
summed over the number of degrees of freedom. For the present case, the squared error function
becomes

E*-E* + E> (5)

The problem now is reduced to that of minimizing the error function with respect to the unknown
stiffnesses. This is achieved by taking the derivative of E2 with respect to each unknown element
stiffness and setting it equal to zero. This leads to a set of linear equations equal in number to the
number of elements.

Taking the first derivative of equation (5) with respect to kv /% and k$ yields the following set
of equations written in matrix form

X< ~~X2

0
~~X<

x2
+Xg
F2^k2x2-(k2^kz)x2\ ¦fl (6)

which may be further reduced to

x, 0 to]
*1 X,-X2 0 F,
K \ (7)
X* ~X2 ~~X< +Ao <

- X* Xn Xa +J\n
ü J\a
t J\n Xn
0 x2 .V 0 %,
130 RETRIEVAL OF SYSTEM PROPERTIES OF EXISTING STRUCTURES

Defining a Jacobian matrix [ J] as follows

BEa BEa be*
Bk*, Bk^ Bk§
(8)
IJ]
BEz BEz BEz
Bk* Bkz Bk$

It is readily noticed that equation (7) can be written in the following form

MrM{*}-Mr{F} (9)

From which [k] can be solved for directly

(10)
{*}-[MW1M7{F}

The following example illustrates the Solution. In this example a determinate truss configuration is
chosen for simplicity in which displacements were actually computed using the Standard Direct
Stiffness Method. This is a numerical experiment meant to test the proposed method for the retrieval
of the structure's unknown element stiffnesses. It must be mentioned, however, that for a
determinate truss no such elaborate procedure is necessary because the problem in such a case
is reduced to the Solution of a system of linear equation.

For an indeterminate truss the inverse of [J]r[i/] upon which the Solution hinges is not
guaranteed. To circumvent such a Situation and to assure the existence of a Solution two or more
loading cases must be used and the squared error function given in equation (5) can be formally
written as
NLC N N
E2 - E 2 F," (H)
n-1 M «-1

in which N is the number of degrees of freedom and NLC is the number of loading conditions.

From which the Solution for the element stiffness may be written as
NLCr « NLC / 12
n-1
L J
n-1 " l

3. EXAMPLE
N e 14

.5
The determinate truss shown in figure 2 is used to test the
procedure. The truss is composed of 3 elements of cross 21m
sectional area equal to 25 cm2. The modulus of elasticity
is 200x10e kN The truss has 3 unrestrained degrees Fig 2
m2
 A. HELOU 131

of freedom with the following reduced stiffness matrix derived with the Standard Direct Stiffness
Method for pin jointed trusses and written in terms of the unknown element stiffnesses.

0.36/c,+0.36*2 0.48Ä,-0.48*2 -0.36fr,

[K\ 0.48/r.,-0.48*2 0.64*^ +0.64^ 0.48*2
-0.36*2 0.48*2 *3 +0.36*2

The applied loads are written in the following Standard load vector

kN

Therefore the error vector of equation can be written as

K F, 0.36
V 0.36*2 0.48fc,-0.48*2 -0.36fr,
*2 F2 •- 0.48*, - 0.48*2 0.64fr, + 0.64*2 0.48*2
kl kl -0.36*2 0.48*2 *3 +0.36*2 J Ixj

From which the following Jacobian matrix can readily be deduced

0.36*,+0.48*2 0.36x, - 0.48*2 - 0.36*3 °

M- 0.48*,+0.64X2 -0.48x, + 0.64*2 + 0.48x3 0
0 -0.36x, + 0.48x2 + 0.36x3 x3

Upon performing the Operation as defined by the derived formula (10) the element stiffness are
kN kN which
retrieved i.e. fr, - *g - 28571.4 -^ and *g - 23809.5 -^ are exactly the same as can
be computed using EAjL It must be reiterated that the displacements x^x\ and Xg supposed

to measured, were in this numerical experiment computed using Standard Computer programs.
132 RETRIEVAL OF SYSTEM PROPERTIES OF EXISTING STRUCTURES #%

4. REFERENCES

Matzen, VC. and McNiven, H.D., Investigation of the Inelastic Characteristics of the Single Story
Steel Structure Using System Identification and Shaking Table Experiments, Report No. EERC 76-20,
Earthquake Engineering Research Center, University of California, Berkeley, August 1974.

Shield, P.C., Elementary Linear Algebra, Worth Publishers Inc.

Touqan, A.R., Najah 91, An Engineering Analysis Program for Framed Structures, November 1990.

White, R.N., Gergely, R, Sexsmith, R.C., Structural Engineering, Combined Edition, John Wiley and
Sons.
 Strength Evaluation of Exist
Evaluation de la resistance de
Festigkeitsermittlung für beste

Xinggang ZHOU
Research Assist.
Yantai University
Yantai, China

'*

¦
134 STRENGTH EVALUATION OF EXISTING MASONRY STRUCTURE

1. INTRODUCTION
As for existing masonry structures, it is known that many of them are in illness State in
their serviceability life time, slightly or seriously. In fact, suffering various unfavourable
factors in construction and application, such as dispersity of masonry material,
temperature cracks etc. masonry structures are in insufficient strength or strength
deterioration state which is one of illness state concerned seriously. But the problem how to
pre-
cisely examine the real serviceability state of masonry structures have not been solved for
a long time. In this paper,mothod solving this problem was investigated,and the mothod
developed from experimental information and theoretical analysis.

2. STRENGTH AND RIGIDITY

2. 1 Compressive strength and rigidity
Compressive stress—strain curve of masonry envalope can be expressed []l][^ 2] :

a /«(l - c-^O (2 - 1)

where, a and e are compressive stress and strain respectively. fm is the compressive
strength of masonry envalope, a is a coefficient. Therefore, elastic modulus of masonry
envalope can be obtained;

E gUo afj> (2 - 2)
Based on experimental informationell ,the Statistical value of a is 370. As we know,the
(lateral)rigidity K is equal to the ratio of lateral load to displacement:

K 1
6 h3

12EI
1

~^~
GA
hj h3
1
1. 2h
Etb3 "*" 0. 3Etb
,h.v
V
b } +
Et
~*~
A<h.\
^ }
b
(2
- S)

Substituting equation 2 — 2 into formula (2 — 3 the ralation bettween rigidity and

compressive strength can be got:

K —1 ^^
(|)C(|-)2 + 4]
(2-4)
in which,t,h and b are the thickness,height and width of masonry wall respectively.

2. 2 Shear strength and rigidity

Researching the hysteresis characteristic of masonry wall [lHC2]][3], the Statistical
skeleton curve of hysteresis loops are shown in Figure 1, it is indicated that in the initial
stage the load — displacement relationship is linear and after cracking the displacement in-
creases significantly with appeared and developed cross cracks. Defined Pu as the ultimate
load, /\u as the displacement in regard to pu ,and ac is the normal stress of masonry wall.
The skeleton curve can be expressed as follow:

-£-
(1) P/Pu (0
/\u < 3)
< P/Pu < 0. 78,0 < — (2 — 5)
2. 6 0.
/\u
äk XINGGANG ZHOU 135

(2) P/Pu 0. 69 4- 0. 31 -£-

Am (0.78 < P/Pu<; 1,0. 3 < -£-
Am < 1)

(3) P/Pu =1 + 0. 44(1 — 0. 085a.) — 0. 44(1 — 0. 085ac) Am -£-

(2 - 6)

(2 - 7)
(P/Pa 1*1 A 0. 3)
(4) P/Pu 0. 55 + 0. 04a, (2 - 8)
From formula (2 — 5 the rigidify K can be written
as follow also:

o
K
P
-£-
A
2. 6
P
-g-
Am
2. 6
-f th
±p
Am
(2 - 9) rH

LO

where,fv is the shear strength of masonry wall. O

According Statistical analysis, the ralationship bettween
fv and Au is: O

Am (3 + 4. 5/„)1/2/(0. 45 + 0. 05a,)
0 l. 0 2. 0 3. 0 A/Au
(2 - 10) Fig.
~
1 Skeleton curve of
masonry wall
Thus,

(1. 17 + 0. 13oy,)/.&
K (3 + 4.5/.)1'2 (2 - 11)

Equation (2 — 11) illustrates the ralationship bettween the rigidity K and the shear
strength fv.

2. 3 Rigidity deterioration
Looking Fig. 1 again,it can be found that as increasing of displacement the stiffness of
masonry wall decrease obviously. The stiffness K' at any displacement many be calculated
from the following empirical formula:
K' 0. 0017(A/#)-°-91if CA/H > 1/1000) (2 - 12)

In other word, formula (2 — 12) show the deteriration of rigidity as increasing of displacement
A under later load.

3. ACCUMMLATED DEFORMATION ENERGY AND DAMAGE INDEX

3. 1 Input energy of perunit mass
The problem related to the strength and stiffness of masonry wall in serviceability State
are discussed above. In order to predict the damage degree of masonry structure,
accumulated deformation energy should be stuided, because any damage by earthquake is
the result of accumulated deformation in vabration process Assume that w is the
accumulated deformation energy[4,5]. As for one — freedom system,in general,

W ±mXL (3 - 1)
136 STRENGTH EVALUATION OF EXISTING MASONRY STRUCTURE #%

v^here,Xemaxis the maximum value of elastic vabration velocity of system,and,

X~ 2,w/21/2 (3-2)
ymax is the maxiumum velocity of earth's surface vabration. Substitute formula (3 — 2)
into formula (3 — 1) then,

W \mijL, mEo (3 — 4)
4

E0 can be defined as the input energy of per unit mass. From statitical analysis, E0 can be
f ormulated as follow:

E0
jyL, expil. 385/ — 6. 39) (3 — 5)

Where,I is the earthquake intensity.

3. 2 Damage index of masonry structure

The vabration equilibrium equation of ith storey for multistorey masonry structures can
be expressed (damp is neglected) :

rrini^Xj + y) + nin-xC^Xj + y) + + w*. (]>]£, + y) /,(a\)

(3 - 6)
Multiplying the equation by Xidt dxj, and integralling the equation in the whole vabration
time,thus
r k n

Ü^m^^x^dx, + Sxt^2mkydt Sfx(xx)xxdt (3 — 7)

k=i j=l ik=»

It is noted that:

fXjdXt fxtdxj fXjdxt lxxdxj 0

Jyxidt Jxidxi

xx hxX\/h\

therefore,
» »
J
* 7

Wx Sfi(x^dxt Sjmkl'ydxx T]™*/ y-xidxx T]ra* j-Eo (3 — 8)

TZ tTt hl T=x hl

hi, h of storey.
is the height
Assume that t\> and r\ are the ratio of ultimate deformation energy to elastic deformation
energy and the ratio of deformation energy to elastic deformation energy respectively.
According to experimental date,rfc is about 12 refer to masonry structures. t\ can be wrote
as follow.
 XINGGANG ZHOU 137

* *

Vt= Wx/Wx, 2YV

TZ
h
j-E*/P„&cr
"i TZ hi
h
2KTy.mk y-E0/Cyfv<tby (3 - 9)
Y is a statitical factor which is equal to 0. 78. Let ß express the damage index of masonry
structures by earthquake, then,

ß
W — W,
_ W,-W, _ 7j_
v, - 1
(3
- 10)

From experimental and earthquake damage information[^2, 3 ,6] ,it can be defined that:
ß ^ 1. 0 partly collapse
0. 90 ^/? < 1. 0 serious damage
0. 50 ^S/?< 0. 90 moderate damage

< 0. 50
0. 15 ^S/? slight damage
0
^^ < 0. 15 intact state

4. APPLICATION
4. 1 Application
Up to now, we discussed the strength, rigidity and accumulated deformation energy. In
this section,we will discuss how to evalute the strength of masonry structures in service.
Serviceability State and how to predict the damage of masonry structure by earthquake.
As we know that structure's natrual frequency and mode of vabration can be measured
and analyzed from ambient vabration. So the rigidity of structures can be identified using
the date of natrual frequency, mode of vabration and equibibrium equation of vabration.
Since the rigidity can be indentified, substituting the rigidity into formulae (2 — 4) and (2
— 11). Using formulae (2— l)and(2 — 3) ,the compressive and shear strength of masonry
structures can be evaluted. On the basis of these results, it can be found that which
storey is the weak part in serviceabilty State or under earthquake circumstance. Applying
(3 — 9) and (3 — 10) it can be pridicted that which degree of damage will be caused
under grant earthquake intensity.

4. 2 Example
A multistorey masonry structure. Seven storey,the height of storey is 2800mm. it's plane
figure refering to Fig2. The thickness of outer horizatal wall is 490mm, inter horizatal
wall are 370mm (the first floor) and 240mm (from the second to the seventh). The
thickness of outer transerve wall is 370mm. The results of measured date from ambient
vabration are show in table 1. Using the date of table 1 and the method discussed above.
The distribution of rigidity, strength and damage index under seven degree of earthquake
intersity etc are caculated and shown in table 2.
138 STRENGTH EVALUATION OF EXISTING MASONRY STRUCTURE

Table 1. The results of ambient vabration

Floor weight(kg) Ai(mm) Ki(KN/m)
1 6. 928X105 0.019 8. 75X106
2 6. 566X105 0.038 8. 53X106
3 6. 566X105 0.061 6. 74X106
4 6. 566X105 0.16 1. 42X106
5 6. 566X105 0.18 55. 43X106
6 6. 566X105 0.20 33. 50X106
7 4. 156X105 0.26 1. 49X106
note frenqvlence fi 2. 832HZ
Fig. 2 Plane figure of example

Table 2 The results of evalution

Compressive shear compressive earthquake Damage

Floor Ki(KN/m) stress (MPa) index
strength (MPa) strength (MPa) shear stress

1 8. 75X106 2.80 0.47 0.47 0.15 0

2 8. 53X106 3.62 0.599 0.55 0.29 0.44

3 6. 74X106 3.10 0.440 0.46 0.26 0.13
4 1. 42X106 1.10 0.076 0.36 0.23 7.10
5 5. 43X106 2.68 0.332 0.27 0. 18 0.20
6 3. 50X106 2.00 0.200 0.19 0. 122 0.30
7 1. 44X106 1.13 0.092 0.09 0.05 0.26
note: 1. earthquake intensity is seven degree.
2. earthquake shear stress caculated by equivalent base shear method.

XJ
t>- _ 1n
i
strengl
x:
CD *
1
CD -i i
i

h
strengl

LO - LO -
>
^mpressive
8 i
i shear

CO - l t0
0
to CO
-
i
i
i
X3

CM
-
\
\
l,0 ü
CM -
l

1
1

r
1

i—1 - 1

I
r—1 -
/
/ i i
1

• _L —t—
0 12and strength
stress
3 4. O(MPa) 0
stress
0.5 (MPa)
and strength
3 Distribution of compressive Fig. 4 Distribution of shear
Fig.
" strength and stress strength and stress
 XINGGANG ZHOU 139

6. CONCLUSION AND DISCUSSION

CD Above,the strength,the rigidity,the deformation energy and the damage index of
masonry structure are studied. Ralationship bettween them are given also.

(2) Using the ralationship and date from ambient vabration, the serviceiability State of
existing masonry structures can be asserted,and the damage index by earthquake can be
predict also.

(3) Based on experimental statisfical information and structural dynamic analysis. The
method reflect every aspect involved in masonry structure and suggest a way to examine
the existing masonry structure comprehensively.

(4) Example show that the compressive strength,the shear strength and the damage
index evaluted by the mathod are in good agreement.

(5) Examples show that the mathod are feasible. The assertment results of existing
masonry structure are reliability.

(6) Further investigation should be carried and make the mothod more perfecter.

REFERENCES
1. YiLiang Q. The paper slection of masonry structures, Huonan university publish
house,1989.

2. Jing qin,X. Some problem about seismic Resistance of Brick Masonry walls. The
proceding of Seismic Risistance Vol2. 1986.

3. Jumin S,XinggongZ,Shiping F,Seismic Behavior of Reinforced Brick Masonry Cav-

ity walls,IABSE SYMPOSIUM BRÜSSELS 1990,

4. G. W. Houser,Limit Design of Structures to Risist Earthquakes,Pros,ist. W,c. E. E.

Berkeleg 1986.

5. Haodong L,The evalution of Brick Masonry Structures after Earthquakes,The pro-

ceeding of Seismic Resistance vol 1. 1986
6. Yucheng,Y. The Analysis of Damage to Multistorey Brick Buildings During the Tangshang
Earthquake, Congress Earthquake Protection of Construction in Sersmic Area Ra-
mania,1978. 11
7. Bao zhiwen. etal. Ambient Vibration Survey of Some High Building in Hongkong.
Tsinghua Publish House,1985
140
i%

Stress Assessment of Reinforced C

Estimation des contraintes des const
Spannungsermittlung in gerisse

Petras PUKELIS Vidmantas JOKUB

Associate Professor Associate Profess
Vilnius Technical University Vilnius Technical
Vilnius, Lithuania Vilnius, Lithuania

1*1
Petras Pukelis, born 1932, received Vidmantas Jokubait
his diploma of civil engineer received his diplom
and candidate of science degree at and candidate
Kaunas Technological University, science degree at
Kaunas, Lithuania. His scientific Technological University
142 STRESS ASSESSMENT OF REINFORCED CONCRETE STRUCTURES WITH CRACKS

1. INTRODUCTION

Evaluation of existing reinforced concrete structures is very important for

maintenance, renovation and reconstruction of industrial and other in-service
buildings. There are several methods to evaluate stress and strains in structures
using their examination data. In the case of reinforced concrete structures with
normal cracks in tensile zone it is possible to evaluate stress by the crack parameters
measured during the examination of these structures. The degree of crack
development is considered as a result of existing stress - strain State in the
member [1]. In this paper a posibility to employ classical theory of crack propagation
in a brittle body for calculation of tensile reinforcement stresses at the cross
section through crack is investigated.

2. REINFORCEMENT AND CONCRETE STRESSES IN CROSS SECTION

THROUGH CRACK

Within the limits of assumed model of a brittle body in the case of macroscopic
cracks the following equation is valid [2]:

lim {Vis a }= — (1)

where s is the distance of body points situated in the crack plane from the top
of the crack; tJ is breaking stress for a member with macroscopic crack of 210
length and K is modulus of bond. It is shown in [2] that stresses in body outside
the crack (x>l) due to any normal pressure acting in the banks of this crack may
be expressed by formula, see Fig.l(a):

(2)

Due to limitation of stress er (x0) in the body for the case of x—>l the following
equation must be observed.

(3)
X—»1
[ ^~S

Boundary conditions for the function of the normal pressure pn (E) in the tensile
zone of a reinforced concrete member, see Fig. l(a), may be written as follows:

Pott) l£l<c
pn«) 0 1
S 1

* VC
"f, 10<-x<1 (4)

OL
10 < X < 1

where p0(%) is arbitrary function of pressure applied on banks of the crack and
estimated as a resultant of stresses acting in
 P. PUKELIS - V. JOKUBAITIS - K.A. KAMINSKAS 143

a
~K
As

M M

& -C

CM
CJ
<sj
CSJ
'V
et

p0_m
31

y
-D
*$
«^
^
P dl
OL.

-c 2*PP- 2?
¦ir-i
öf
r
*
w

Mtrt Mut

O^r ttut

T3
f
flJA

_w

Fig. 1 Models of normal crack in tensile zone of reinforced concrete

member: a - for evaluation of reinforcement stress intensity coefficient;
b - for assesment of tensile reinforcement stress in crack

continuous concrete of the crack zone \x\ < l\ fct is tensile strength of concrete;
crß/oc is stress at the level of tensile reinforcement; oc Es /Ec; ß 125 p and
is used when longitudinal reinforcement ratio p As /(bd) is small when
p > 0.008, ß =1.

Using equation (4) and considering that s x-L Vx2 - /

Vs Vl0 + s, equation (3) may be rearranged into the following form

lim
s->o l r ay(*v)yJ «
{Vs
2& ß
'
OC7T
r
Vt (5)

From equations (1) and (5) tensile reinforcement stress

Ka (6)
a 2ßVt~
144 STRESS ASSESSMENT OF REINFORCED CONCRETE STRUCTURES WITH CRACKS jrffm

and eritieal steel stress intensity factor

k -
er
1.6 f vr£
s fJC
<7)

where fs is strenght (yield limit) of reinforcement steel. Values greater than

eritieal Ker values leads to the fracture of not overreinforced concrete member
with tensile steel.

By similar method as in [2] assuming bodies as isotripic expression of the

modulus of bond K for reinforced concrete member has been deduced. A fictitious
cross section of the depth h} was assumed, see Fig. l(b), the centroid of which
eoineides with the central axis of the member. The depth h1 is obtained from the
following expression:

h+h^-^
r/Vj et
(8)

where S} is statical moment of ficticious cross - sectional area A} in respect to

the centroid of longitudinal tensile reinforcement.

Modulus

K
ML t -
m AM NfN
isi ^ r-
Vh
JE
/Q\
V?)
*> t Wj
-
A1 ; ' 2

where Mtot M - Pe op and is the total moment in relation to the centroid of

the real cross section of the member, see Fig. 1; P is longitudinal reinforcement
prestressing force; eo is the eccentricity of P\ AM Ntot (aQ-a0 and is the
increase in bending moment due to deviation of centroid of fictitious cross
section from the centroid of the real cross section of the member; Ntot P ± N
and is the total longitudinal force acting in the centroid of the real cross section.
Compressive TV force is positive; W1 =1/(0. 75hr +hct and is modulus of cross
section ; 11 is inertia moment of fictious cross section in relation to the neutral
axis situated at distance h r +h et from the tensile face of fictitious cross section;'
h 8crhr /w and is the depth of concrete tensile zone above the crack, see [ 1 ].

The eritieal width of the crack end

ocr=0.00012 a VejTr, ^ ; jgg % (10)

where 0 is bar diameter of tensile reinforcement in mm; 77 is factor for consideration

of bond between concrete and reinforcing steel; p} As / [bh + (b - b)hf] < 0.02
and p2 As / (bh) < 0.02 and they are longitudinal reinforcing factors; b and h
are width and the thickness of the flange in the tensile zone of the member.

Substitution of modulus K value by (9) into formule (6) gives the final expresion
of tensile reinforcement stress value
Jf%L P. PUKELIS - V. JOKUBAITIS - K.A. KAMINSKAS 145

+ AM)(0.75h+ht) N^-.tccc^TK"
q » - r(Mtot
L
Ix
"
A1
J ^
4ß t fin
^11;

If the neutral axis is located extreme compressed concrete fiber stress crc may be
calculated by the equation:

a
M (h -
Ired
h
-jy (12)

where Ired is inertia moment about neutral axis of transformed cross section by
reducing its area by Act hb.

For concrete coefficient

Ker =V8Eft (13>

er c et

where iscandy^are modulus at elasticity and tensile strength of concrete respectively.

3. RESULTS OF EXPERIMENTAL INVESTIGATION

Special test specimens, beams and eccentrically compressed reinforced concrete

members, were investigated. Prestressed concrete beams of rectangular cross
section bxh 100xl80 mm, span / 1800 mm, longitudinal prestressing steel
ratio p 0.8 - 0.95 %, prestressing force P 41.2 - 132.5 kN were subjected
to two concentrated forces at 1/3 distances from supports. Electrical resistance
gauges were used to measure steel strain in the cross section through crack. The
parameters hr and w of crack in pure bending zone of beams were measured by
24 times magnifying microscope and concroled by electrical resistance strain
gauges closely spaced on beam face along its height.

Results of special tests on beams reported in [3] also are employed to evaluate
theoretical proposotions of this paper. Reinforced concrete beams of rectangular
crosis section b x h=120 x 300 mm, span length / =2000 mm, longitudinal tensile
reinforcement ratio p 0.48 - 3.83%, without prestress were loaded by two
concentrated forces at 1/3 distances from supports. Special notehes were formed
in the test beams to locate the main cracks. Steel strain in the main crack and
parameters hr and w of this crack were measured.

Eccentrically compressed members of reetantgular cross section

b x h 100 x 150 mm, length / 1000 mm, tensile reinforcement rat io
p 0.46 - 1.26% were tested by the authors of this paper. Notehes in the middle
of the member length from tensile face up to longitudinal reinforcing bars deep
were formed. Crack parameters were measured by a microscope magnifying 24
times and by a dial gauge at the level of tensile reinforcement.
146 STRESS ASSESSMENT OF REINFORCED CONCRETE STRUCTURES WITH CRACKS

The values of tensile reinforcement stresses obtained by tests cjsoos and calculated
by equation (11) er are compared, see Fig.2. In the case of eccentrically compressed

£
\
^ 9
£

"5 3
0__
3
CT) 7%
8
oo • o

§
0.3 M 07 M/Ma(Me/Mu)

Fig.2 Comparison of measured in tests erfand theoretical er calculated by

formula (11) tensile reinforcement stress values: • - prestressed concrete
beams; o - reinforced concrete beams; o - eccentrically compressed
members

members ratio c?sohs/ er is plotted in respect to quantity Ne / Mu where N is

external force acting at distance e from centroid of tensile reinforcement and Mu
is the ultimate bending moment of the cross section.

Theoretical values er calculated by (11) are on the average 7% higher than asobs
measured in tests. The Variation coefficient of the ratio cJsobJ er is equal to 0.15.

Experimental values of concrete extreme fiber stress <Jcobs obtained from tests [3]
were compared with theoretical values Cc of this stress calculated by equation
(12). Theoretical values Oc on the average are 3% higher than experimental values
erc.obs of this stress. Coefficient of Variation of the ratio erc.oos / erc is equal to 0.11.
u

4.CONCLUSIONS
Contour of through normal crack in tension zone of a reinforced concrete member
always is continguous to tensile concrete and to reinforcement. The latter has
substantial influence on rupture strength of tensile concrete. Parameters of this
strength Kcr and 8cr expressed by (7), (10) and (13) define correctly character of
stable crack propagation observed in tested samples.
 P. PUKELIS - V. JOKUBAITIS - K.A. KAMINSKAS 147

Calculation of stresses in tensile reinforcement by (11) and in extreme fiber of

compressed concrete by (12) makes it possible to assess stress state of reinforced
concrete structures in service with sufficient accuracy.

REFERENCES

1. JOKUBAITIS V., PUKELIS P., Stress - strain State estimation in existing

reinforced concrete structures subjected to bending (in russian). Transactions
of Lithuanian universities. Reinforced concrete structures No.12, Vilnius, 1983.
2. PANASIUK V., Ultimate equilibrium of brittle bodies with cracks (in russian).
Kijev, 1968.

3. CHUBRIKOV V., Criteria of fracture mechanics for stress assesment in

reinforced
concrete bending member section with crack (in russian). Lvov,
1989.
148
 Detection of Local Stiffness
Mesure des variations locales de
Messung örtlicher Steifigkeitsän

Ghee Koh CHAN Ming See LIN

Senior Lecturer Research Assistan
National Univ. of Singapore National Univ. of S
Singapore Singapore

f^
Chan Ghee Koh, born in 1956, got
his Ph.D in Civil Engineering at the
University of California Berkeley.
He taught at the National University
of Singapore since 1986. His
150 DETECTION OF LOCAL STIFFNESS CHANGES OF BUILDINGS

INTRODÜCTION
In recent years, application of system identification (SI) to damage
assessment and safety evaluation of civil engineering structures has received
considerable attention (e.g. Natke and Yao 1987; Agbabian et al. 1991). Based
on input and output measurements of dynamically excited structures, structural
parameters such as stiffnesses are determined and then compared with intended
design values. In this manner, periodic monitoring of state of structures can
be performed for detection of structural changes due to damage or
deterioration. However, several problems have yet to be resolved before this
methodology can become viable for actual structures.
One of the problems reported by some researchers is that current SI
techniques have not been satisfactory in detecting local damages. Modal
parameters as determined by frequency domain analysis are not sensitive to*
local damages, except for small structural Systems or unless high modes are
taken into aecount. The aecuraey of high modes is, however, often difficult
to achieve because of measurement noise. Hence, there is apparently a trend
that researchers prefer time-domain SI approaches, among which the extended
Kaiman filtering (EKF) developed by Kaiman and Bucy (1961) is
originally
perhaps most widely used.
stiffness
Nevertheless, it
has been found that the change in
matrix due to member stiffness changes in the locality tends to
"spread out" or "diffuse" into adjacent structural members (herein referred to
as "stiffness diffusion"), thereby making local damage identification
difficult (Natke and Yao 1987).
In addition, from the viewpoint of structural safety evaluation,
important to estimate the confidence level (or reliability) of identified
is it
parameters taking into consideration measurement noise as well as modeling
errors. In this aspect, Agbabian et al. (1991) has applied least-squares
approximation methods to successive time Windows of input and output (I/O)
time histories. In their numerical studies, the effects of I/O noise have
been taken into aecount. However, to the authors* knowledge, modeling errors
have thus far not been considered in the confidence estimate of identified
parameters.
IMPROVED CONDENSATION METHOD

The problem of local damage detection is aggravated by the large number

of degrees of freedom (DOFs) in modeling an actual structure. When applied to
a complete structural model involving all DOFs, the EKF and other time-domain
SI approaches alike are often found to be numerically inefficient in terms of
aecuraey, convergence and computation speed. Alternatively, a "reduced" model
with a smaller number of DOFs can be considered. For instance, if
quantification of storey stiffness changes of a building suffice for the
purpose of damage detection, a simple lumped mass model can be used to reduce
DOFs. Unfortunately, as a result of considerable modeling errors, diffusion
problem of storey stiffness into adjacent storeys would render local damage
detection ineffective.
As an attempt to detect the location of damage and quantify the magnitude
of damage in terms of stiffness reduction of a building, an improved
condensation method (ICM) is proposed here. For illustration purpose, a
single-bay n-storey plane frame building as shown in Fig. Ha) is considered.
Axial and shear deformations are assumed to be negligible. Static
condensation is first condueted, reducing the complete structural model to a
"Condensed" model with a significantly smaller number of DOFs [Fig. Hb)]. In
this study, columns are considered to be the eritieal elements where damages
are likely to occur and affect the overall Performance of the building.
 G.K. CHAN - M.S. LIN - T. BALENDRA 151

Non-critical elements (beams) are assumed to be undamaged and any difference

in storey stiffness is solely due to changes in column flexural rigidities
(EI), In order to narrow the gap between the Condensed model and the actual
structure (upon which I/O measurements are taken), a "remedial" (or
correction) stiffness
3n-2 3n-1 matrix K
R
is derived
o_ — 3n o——n based on a lumped
o mass model as shown
in Fig. l(c). By
applying the EKF on a
1 time window of data,
the stiffnesses of
'£ — 2 '»
—2 this remedial model
are identified and
'll <>
— 1
then used to update
the Condensed model.
This process is
(b) (0
repeated for a
specified number of
FIG. 1. (a) Complete Model; (b) Condensed Model; time Windows (or
(c) Remedial model. (Numbers indicate DOFs). unt i1 convergence).
The procedure of the ICM for damage detection of the building considered
is described below.
(1) Divide the excitation and response time histories into time Windows.
(2) Form mass matrix M (3nx3n) for the complete model based on known mass
distribution. Given flexural rigidities of the "undamaged" building, form
complete stiffness matrix K (3nx3n) encompassing all DOFs, i.e. two Joint
rotations and one horizontal translation per floor [Fig. Ha)].
(3) Damping matrix C (3nx3n) is constituted by adopting Rayleigh damping
assuming that damping ratlos of two Vibration modes are known. (This
assumption can be relaxed by including damping ratios as additional
unknown parameters to be identified.)
(4) Perform static condensation to obtain Condensed mass matrix M »damping
matrix C c and stiffness matrix Kc all of size nxn [Fig. Hb)],
(5) The improved condensation model is derived by adding the remedial
stiffness matrix K (as explained earlier) to the Condensed stiffness
matrix K Elements in K are the unknown parameters to be identified by
the EKF, while K remains unchanged.
(6) Compute
N
stiffness correction factor tj
J
for the j-th storey as follows:

t). V (ei vK)/(ei

where denotes the undamaged quantity, R denotes the remedial quantity
U
and N
w
is the current time window number. Since the remedial model is
derived from a shear building, the remedial flexural rigidity is given by
(EI R =(KR 1
J
12, where (K R
J J /
is the corresponding storey stiffness in
J
K
R
and
J
is the column length
J The updated flexural rigidities
EI
J (EIu).(l +,j). J 1 n (2)
are then used to compute the complete stiffness matrix K.
152 DETECTION OF LOCAL STIFFNESS CHANGES OF BUILDINGS

(7) Repeat step 3 to step 6 for all time Windows considered. The severity of
the damage in each storey is finally given by the end result of tj
ADAPTIVE FILTER
EKF, uncertainties in terms of variances
In SI techniques employing the
of identified parameters are supposedly reflected in the error covariance
matrix. The error covariance is dependent on the output noise covariance and
the system noise covariance in the EKF algorithm. The variances of input and
output noise can be estimated from resolution and aecuraey of Instruments and
data acquisition system. The main problem is the difficulty in estimating the
variance of system noise which includes modeling errors as well as input
noise. In application of the EKF, the uncertainty in the system noise causes
the divergence phenomenon, especially when the input noise is small in
comparison with the modeling errors.
An adaptive filter
was developed by Jazwinski (1969) as an algorithmic
attempt to control divergence in Kaiman filtering of orbit determination
problems. In this paper, this adaptive filter is modified to suit SI problems
for the purpose of obtaining statistically consistent variances of identified
parameters.
Determination of System Error Covariance Q
Consider a time window beginning with k-th time step. The predicted
residual vector at p steps läter (i.e. time t is defined as
k+p *k+p k+p k+p k
where y*k+p is an Observation vector (mxl), x(t,K+P \t K is an expected state
vector (nxl), and H(t,k+p is an Observation matrix (mxn). The covariance of
the predicted residual vector can be derived by means of the EKF algorithm as

«WV - H^ic+p^(tk+p't)c)P(tkltk)$T(tk+p^lc)HT(tk+p)
P

k+p I«Wtw,Qttw-i)#IlVw) [nT(t,k+p

1=1
+ r,k+pn (4)

where E{ • > denotes expectation Operator, $ is a state transition matrix (nxn),

P is an error covariance matrix (nxn), Q is a system noise covariance matrix
(nxn), and R is a measurement noise covariance matrix (mxm). Q is determined
by ensuring consistency between the residuals and their statistics such that
P

~y I rk+irk+i ^ E{rk+prk+P} (5)

i=l
Equation 4 can thus be written as
P ^ P

H(tIrk+pJ I«ti.p-ttoi)«(tic.i.i)#T(Wti-i)V(W
1=1 y J
^lT
i=l
k+i k+i
T

|Q(t,k+p )=0} ^6a^

E{r,k+pr?"k+p1^
where

k+p
 G.K. CHAN - M.S. LIN - T. BALENDRA 153

In the present formulation, only diagonal elements of r, r. are considered

whereas all off-diagonal elements are assumed to be zero (i.e. no
cross-correlation between residuals). Let Q(£,k+p equal
^ to G,k+pQ,k+pG,k+p where
The left hand side of Eq. 6(a) can now be
k+p is
G. a distribution matrix.
expressed
* as A, diag[Q,
& k+p ] where
k+p
P

«•- =iüH(t,k+p )$(t,k+p ,t,k+iJG.k+i ->]' (7)

Hence, the diagonal terms of Eq. 6(a) can be obtained from
k+pdiag[Q,
A, ö k+p ] (8a)
k+p
where
diag[— Y
k+1.rl
r*
k+1 - E{r7k+P k+P |Q(t,k+p
)=0}1 (8b)
fe+P L P *^i *Y J

It is reasonable to assume that Q remains constant in a time window of N

sampling points, i.e. Q Q Q
k+N'
The system noise covariance in this
time window is given by
diag[Qk,tf] lAl,N Ak,N}'lAl,N Ck,N (9a)
where A is an Nmxn matrix and e is an Nmxl vector written as follows
Ak,tf {Ak*Ak+l >Ak+N} ek,N {ek'ek+l"-"ck+N} (9b,c)
The main procedure of adaptive filter is schematically explained in
Fig. 2. For purpose of discussion, we now let 9 denote a vector of parameters
to be identified, such as unknown stiffnesses, and x contains only response
variables (displacements and velocities). Initially the EKF may be carried
out once (not shown in Fig. 2) to obtain a better guess of G An adaptive
filter cycle comprises two processes: (a) determination of Q by enforcing
Statistical consistency of residuals using N sampling points, and (b)
determination of x, G and P by the EKF using M sampling points.
N samples delay x{k\k)
e(*IA)

x{k+M\k+M)
e{k+M\k+M)
x(k\k), 0(k\k), Q=0,P(k\k) Determine EKF P(/r+AM+Af) End of Yes
e(sls)
Start *""
Q Algorithm Data
P(sls)

k=k+M No

FIG. 2. The Adaptive Filter Procedure.

NUMERICAL EXAMPLES AND DISCUSSIONS

In our examples, the input is a force comprising several (five or more)

harmonics of frequencies covering the first few significant Vibration modes of
the structure. Added to input and output time histories are independent
Gaussian noises with zero mean and Standard deviations equal to certain
specified percentages of their respective unpolluted root-mean-square values.
154 DETECTION OF LOCAL STIFFNESS CHANGES OF BUILDINGS

Example 1 (ICM)
The procedure of ICM is illustrated by considering a 12-storey plane
frame building. The "complete" plane frame model has a total of 36 DOFs (two
Joint rotations and one horizontal translation at each floor). The mass
derived from a
matrix and stiffness matrix for the undamaged building are
small-scale steel laboratory model. The damping ratio is 0.5% as determined
by free Vibration tests of the laboratory model.
We now consider the building to be "damaged": the column stiffnesses in
the first, fourth, eighth, ninth and eleventh storeys are reduced by 10, 15,
30, 20 and 25 per cents, respectively. An excitation force is applied at the
top floor and horizontal responses at all floors are measured. Total
Observation time history of 2 s at a sampling rate of 0.0005 s is divided into
20 Windows. All rotational DOFs are eliminated in the Condensed model and the
remedial model thus has 12 DOFs.

Following the procedure of the ICM described earlier, storey stiffnesses

of the damaged building are identified without I/O noise and with different
noise levels. The ratios of damaged storey stiffnesses to the corresponding
values of the undamaged building are computed and summarized in Table 1. In
the ideal case of zero 1/0 noise, the identified stiffnesses are almost exact
(error < 1%) for all twelve storeys and the stiffness diffusionused problem is
negligibly small. In comparison,
the results
the
(not
lumped
shown
if
mass
in
model were
Table 1) would
instead
have been
for the same conditions,
1/0 noise level
disastrous with error as high as 60% at some storeys. For an
of 20% which is considerably high in practice, the6% identified results are
remarkably good (with error ranging from 0.2% to only) in view of the
fairly large system for identification.

Story
1/0 Ist 2nd 3rd 4th 5th 6th 7th 8th 9th lOth llth 12th
Noise
0% 90.0 99.6 100.5 85.0 99.9 99.8 100.0 70.1 80.3 99.9 75.0 99.6
10% 89.7 97.2 103.4 87.6 100.7 99.6 99.8 72.8 83.6 97.7 76.8 100.3
20% 90.9 97.1 95.8 86.9 100.2 101.9 100.7 74.1 83.4 97.1 76.0 101.2
207. 88.5 120.8 126.0 91.0 92.9 115.1 140.0 79.7 78.6 97.8 79.4 76.9
+
20% 90.6 99.6 105.5 86.9 97.5 104.5 101.0 72.3 81.0 97.5 78.4 94.2
Exact 90.0 100.0 100.0 85.0 100.0 100.0 100.0 70.0 80.0 100.0 75.0 100.0
* Six horizontal response measurements at alternate floors.
+ Averaged results based on twelve different time histories of excitation with same noise level
TABLE 1. Percentage Ratio of Damaged Storey Stiffness to Undamaged Storey Stiffness

requires only 20% more than the

In terms of computation time, the ICM
lumped mass approach in this whereas a complete structural
example,
identification with 36 DOFs would be very time consuming (easily ten times
more) if
convergence can be achieved at all. The ICM is hence a simple and
yet effective approach to determine local structural changes with virtually no
stiffness diffusion problem.
Example 2 (Adaptive Filter)
In this example, the adaptive filter procedure is applied to an 1-DOF
system to obtain error variances of identified parameters under the influence
of 1/0 noise. The mass is known and has a value of 1 whereas the stiffness
(K) and damping coefficient (C) are to be identified. Assuming independent
 G.K. CHAN - M.S. LIN - T. BALENDRA 155

system noises for all state variables, the distribution matrix G is simply a
unit matrix. Initial conditions are: x =y-'o x2=y
l Jo x3=100, x4=1, P 3,3
=400 and
p =0.1. The sampling numbers are J\T=20 and tf=3. Total Observation time IS
4,4
22.5 s at a sampling rate of 0.075 s.
To evaluate the Statistical consistency of estimation error, the
following Performance index is defined:
** { B-J1[xrUi}^(ti,ti),a ]
/ [pr,rWk) ]
where y =1, 2, 3 and 4 denote displacement, velocity, stiffness and damping,
respectively. Due to randomness, the Performance index fluctuates with k and
would be desirable
}

it
(12)

10 I ~~ to have the index 1

o Adaptive Filter (Q) averaging about one.

A EKF The Performance of
A the adaptive filter is
compared to that of the
_A
>—— Force
EKF with zero system
A noise(Q=0). The Performance
<fc index for stiffness
h -A is shown in
7/ 7"
Fig. 3a for 10% 1/0
A noise and in Fig. 3b
for 30% I/O noise. It
o°o 0°° „ooooooooo ^ooo can be seen that the
r\
0ü° 0o ö°o000° uoooocTuo0oU00 Performanceindex
n 1
l 1
i 1

diverges in the case of

10 the EKF. This means
that the error variance
(b) are
o Adaptive Filter
statistically
aEKF inconsistent and thus
do not truly reflect
the confidence level in
A
the identified
A parameter. The Performance
<k indices in the case of
_A the adaptive filter
Clusters around 1,
which is an indication
of good Statistical
oOQQoQOOOoO
onoom0 ooo0oOOO
consistency of error
*oo covariances determined
JL I JL in the SI process.
FIG. 3. Performance Indices for Stiffness (a) under 10% L/O
Noise and (b) under 30% 1/0 Noise.

Example 3 (ICM with Adaptive Filter)

The last example demonstrates the combined application of the ICM and the
adaptive filter for identification of local structural changes with confidence
estimates of a three-storey plane frame building which has the same damping
ratio, columns and beams as in Example 1. The excitation force is applied at
the top floor and horizontal responses at all three floors are measured.
Total Observation time is 2 s at a sampling rate of 0.0005 s. In the
application of the ICM, the Observation time history is divided into 20 time
Windows.
156 DETECTION OF LOCAL STIFFNESS CHANGES OF BUILDINGS

of structural changes, we consider a damaged building with

For Simulation
column stiffnesses at the first and third storeys reduced by 15% and 30%,
respectively, and the second storey undamaged. These results in terms of
percentage ratios relative to the "undamaged" values are summarized in Table
2. It can be seen that the effect of stiffness diffusion into the second
storey is very small. Specifically, under a 0% noise level, the identified
stiffness change of the supposedly undamaged second storey is only about 1%.
The severity of damage in the first and third storeys is accurately reflected
even for I/O noise level as high as 20%.
In the absence of I/O noise, the Standard deviation of each identified
stiffness ratio (in %) is about 0.8. The variability of identified results in
this case is primarily attributed to the modeling errors. With 20% I/O noise,
the Standard deviation increases and thus the reliability of identified
results is less. If randomness of an identified parameter is approximated by
a Gaussian distribution, the reliability can be translated into a maximum
likelihood ränge corresponding to a specified confidence level. As an
illustration, 95%
I/O
Storey Stiffness Standard 95% confidence ranges based on
Noise Ratio Deviation Confidence ränge ±1. 96<r are presented in
Ist 86.3 0.8 84.7 - 87.9 Table 3. Hence, with
0% 2nd 101.2 0.8 99.6 - 102.8 the determination of
3rd 69.7 0.8 68.1 - 71.3
identified stiffnesses
Ist 86.7 5.8 75.3 - 98.1 and their respective
20% 2nd 95.1 5.8 83.7 - 106.5 variances, the combined
3rd 70.5 5.8 59.1 - 81.9
application of the ICM
Ist 85.0 - - and the adaptive filter
Exact 2nd 100.0 - - would be useful for the
3rd 70.0 - -
reliability analysis
TABLE 2. Percentage Ratio of Damaged Story Stiffness to and safety evaluation
Undamaged Storey stiffness and their Confidence Estimates of buildings.
C0NCLUSI0NS
Two issues, namely (a) local detection and (b) confidence
damage
estimation of identified parameters, dealth with in this paper.
have been
Firstly, an "improved condensation" method is
proposed to identify the
locations and magnitudes of structural stiffness changes of buildings.
Secondly, confidence levels in identified parameters are estimated by means of
an adaptive filter which ensures Statistical consistency of error covariances
in the application of the EKF. The application of the proposed procedures to
numerical examples have shown their potential as an effective tool to identify
local structural changes of buildings with consistent confidence estimates.
APPENDIX I. REFERENCES

Agbabian, M. F., Miller, R. K. and Caughey, T. K. System

S. Masri, S.
identification approach to detection of structural changes. Journal of
Engineering Mechanics, ASCE, 117(2), 370-390, 1991.
Jazwinski, A. H. Adaptive Filter ing. Automatica, 5, 475-485, 1969.
Kaiman, R. E. and Bucy R. S. New
theory. Journal of Basic Engineering, ASME, 83, 95-108, 1961.
results in linear f iltering and prediction

Natke, H. G., and Yao, J. T. P., System identification approaches in

structural safety evaluation. Structural Safety Evaluation Based on System
Identification Approaches, Proceedings of Workshop at Lambrecht/Pfatz, 29 June
to 1 July 1987, Natke and Yao, eds., Frieder Vieweg and Sohn, Braunschweig,
Germany, 460-473, 1987.
„iA

Planes of Weakness in F
Surfaces de rupture en ana
Versagensflächen in Finite

Jörg-Martin HOHBERG
Dr. Eng.
IUB Engineering Ltd
Berne, Switzerland

,
1
58 PLANES OF WEAKNESS IN FINITE ELEMENT ANALYSIS

1. INTRODUCTION
The finite element method (FEM) is widely used for the assessment of material damage by following
the gradual development of deterioration in structures in a step-by-step procedure. Usual material
modeis are based on incremental plasticity, damage theory or smeared cracking, where for monotonic
loading the anisotropy of damage is often neglected to avoid over st ifF numerical results; such overstiff
behaviour is absent in discrete crack modeis [1]. Apart from distributed ageing phenomena as con-
tinuum deterioration, the inspection of deficient structures may reveal a number of existing fractures,
which are possibly oriented oblique to the present stress regime and need be modelled as to their effect
on the stress redistribution and the failure mode of the cracked structure.
Through the joining of finite elements at their common nodes, the conventional FEM is basically
a continuum method. At least with nodal displacements as primary variables, equilibrium is only
satisfied in an integral sense: Although the displacement fields are compatible along the element sides,
the stress fields exhibit finite jumps at interelement boundaries, thus precluding the computation
of strict lower bound limit loads [2], However, lines or planes of displacement discontinuity can be
introduced via double nodes with suitable constraint conditions and used to investigate the ultimate
bearing capacity of structures by means of postulated failure planes, a concept which is akin to the
kinematic approach of limit analysis in that one looks for the mechanism giving the smallest failure
load as upper bound to the true limit load.1 The presence of an elastic compliance below the onset of
plastic deformation does not invalidate the limit analysis theorems as long as displacements remain
small [4]. It is rather the behaviour of the weak planes which infringes on certain vital hypotheses.

2. KINEMATICAL DISCONTINUITIES
2.1 The Concept in Limit Analysis
The general idea is that arbitrary velocity fields can be introduced, which do not need to satisfy
equilibrium and may be discontinuous as long as they are kinematically compatible. For instance it
is permissible to assume that large parts of the structure move as rigid blocks, separated by narrow
plastic regions of thickness t. These are characterized by a high homogenous strain rate, which is the
relative velocity between blocks per thickness, 6/t. The discontinuities are supposed to consist of a
thin layer of material, which obeys a modified Mohr-Coulomb yield criterion (with associated flow
rule) and behaves just like a solid, except that the in-plane 'stretching' strain rate £S9 is zero because
of the adjacent rigid bodies. Computing principal strain rates with ins \^n» (Fig. 1),

«1,2 \tnn ± \y/i>nn+1n, — (sina ± 1) (1)

their directions are found to bisect the angle between the n-direction and the velocity vector, resp.
between the s-direction and the normal to the velocity. While i\ denotes a volume increase due to
shear dilatancy or opening, £2 corresponds to a compression field in the adjacent block [5]. The latter
would only disappear for a pure cleavage at a 90°, i.e. if the discontinuity were to coincide with a
mode-I crack (£1 > £2 0). Principal directions at 45° ± ^ (with respect to s) characterize slip lines
in a state of pure shear.
The internally dissipated work per unit area is that of a ductile homogenous material, the band
thickness droping out during Integration:

Wi (<7i£! -f a2e2) t \6crx (sina + 1) + \6<t2 (sina - 1) (2)

For a general a one obtains [6]
1 — sin a sin a — sin <f>
Wi S
ö + ft (3)
1 — sin^
1
As stated in [3]: "The structure will collapse if there is any compatible pattern of plastic deformation for which the
rate of work of the externa! loads exceeds the rate of internal dissipation."
 J.-M. HOHBERG 159

fr
M
/ fr
|
plane stress
h es

45°- % 90 +Oc
r^ (ft.H-fcl

plane strain
(o,-t

Figure 1: Strain rates and failure surface in band discontinuity

with the pure failure modi of shear and opening, using k — (1 + sin^)/(l — sin^):
shear (a <j>) : Wi \8 fc (1 - sin <j>) 6 c cos <j> opening (a ^) : W\ Sft (4)

-
From letting t —? 0 in eq. (1) for which £\y2 grows to ± infinity - it is concluded that the Joint
material needs to be formulated for plane-strain conditions [7]. Together with the associated flow rule
arising from von-Mises' postulate of maximum dissipation, this implies that a < is not permitted <f>

by this kind of model; it would become feasible only in plane stress where another corner stress state
allows for simultaneous shear and compression failure [8].

2.2 Interface Element Formulations

The FEM knows a similar concept of degenerating a solid to a layer of finite thickness *, assuming a
strain-formulated layer material model for a constant strain gradient across the thickness [9]:

{da} [D'--D"]-t{dS} (5)

The stretching strain component e9a is again assumed to vanish, because of the assumption
the length of the layer element [10]. In view of the fact that also ^De grows to hifinity with t -+ 0, a
Kl,
very thin layer would infact behave rigid-plastic if Dc were not corrected for the layer element aspect
ratio t/L [11]. With nodal displacements as primary unknowns, this is required by the finite-precision
arithmetic of the equation solver. Note that only the plastic strain components correspond to the
'kinematic strains' in limit analysis and dissipate energy on the stress state.
For a vanishing layer thickness the interface can directly be formulated in relative displacements
d6 — dSe + d6p. The elastic stiffness of the bonded state is given by local penalty parameters k, G/t
and Kn E/t, and the stress-strain constitutive model is just converted into a relationship between
tractions and relative displacements, the factor 1/t being virtually incorporated into De and Dp.
Because of the traction formulation plain-stress and plane-strain states can no longer be distinguished
in the Joint. Whether or not a thin-layer element approaches indeed the slip behaviour of a zero-
thickness interface, depends on the form of stress evaluation: Unless the inforniation of the interface
orientation is passed on to the constitutive routine, an ordinary principal stress criterion will result in
premature failure of the interface compared to Coulomb friction, because the shear planes in the layer
material are predicted according to eq. 1 as being inclined relative to the interface orientation [12].
160 PLANES OF WEAKNESS IN FINITE ELEMENT ANALYSIS

i 's

^ XS^
(D
A
t^ rgn fy ^
-f,4- ft+

Figure 2: Modified Coulomb modeis in strains and relative displacements

A populär failure surface for combined slip and opening is the hyperboloid, which differs not too
much from the cone with spherical cusp in limit analysis (Fig. 2). The continuous curvature simulates
the added geometrical strength component resulting from the inclination of asperities with a mean
roughness angle if>, which are overridden under small compression and become progressively sheared
off under high compression. With increasing \crn\ the surface approaches the asymptotic friction cone
of a plane interface with a 'basic' friction angle <j>^, and zero dilatancy. Since the 'mobilised' angle
of friction is of the form ^mo& <t>» + VK0"*^*)» the flow (or slip) rule can never be associated [13].
For a rough surface the truncated friction cone is but a linearization, where the geometric stiffness
component is simplified to an apparent cohesion intercept.
Angles of 6* larger than rp must contain an opening component. There is no reason why the
flow potential should display a smooth transition from shear to opening. More likely, the potential
surface for shear dilatancy forms a corner singularity with the n-axis. This allows to distinguish
irreversible opening due to override in shear from reversible gap displacements. If the interface is
initially cemented, a retractable tension cap extends into the tension/shear domain, furnishing a tensile
strength and a true cohesion. Both quantitites are destroyed together in any arbitrary combination
of tension and shear (area* 1? in Fig. 2 right) [14]. The roughness or apparent cohesion is treated
separately: As continued override wears the asperities down, the failure surface will shrink in function
of the accumulated sliding distance 6t or, alternatively, of the plastic slip work Wf (softening of area
'2').

3. LIMIT ANALYSIS VS. LIMIT EQUILIBRIUM METHODS

The theorems of limit analysis ofFer the great avantage that neither the initial conditions in the
structure nor the exact load path to failure need be known, provided the material is sufficiently
ductile and stable in Drucker's sense. Concrete departs from the assumption of unlimited deformability
already in compression, such that a hypothetical plateau need be fitted at a reduced average stress
over a particular strain ränge [3], this reduction being commonly termed the 'effectiveness factor' [15].
Since the strain history differs for each particular problem setting (bending, shear, etc.), this factor
varies and accounts for different influences in a global manner. To confuse the matter further, also
the effect of construction joints is sometimes subsumed in there [8] even though it could be aecounted
for by reduced material parameters in an explicitely modelled weak plane.
The definition of kinematical discontinuities ignores any dependence of the ductility on the angle
a, which would be valid only for the assumption of 'unlimited ductility' at zero tensile strength. Even
then a strain-capacity problem is present in the crack width across which shear stresses can still be
carried by aggregate interlock. Because the kinematical discontinuities are usually not identified with
cracks - except for socalled collapse cracks in pure tension [5] -
plasticity theory tacitly assumes
that the compression struts (£2 in Fig. 1) are not restrained by the crack pattern in their ability to
 J.-M. HOHBERG 161

For a Joint inclined under the angle ß to the horizontal, the externa!
work due to a uniaxial compressive stress <ry is given (per width B and UUliH. ol
unit thickness of the specimen) by

WE 6 <ry sin(/3
- a) B
and the internal plastic work in the line of discontinuity

WT=Sfe (l-sina)fl
2cos/3

Equating externa! and internal work and minimizing with respect to ß,

one finds simply o*y>Tmn /c, just as for the associated case of a </> [8],
only at a different eritieal orientation ß x/4 -f a/2. irm
Figure 3: Analysis of a prism test with construction Joint

adapt themselves to principal stress rotations during loading, cf. [16]. Interestingly, FEM interface
modeis may be liable to the same pitfall if the limitation of dilatancy by the height of asperities is
not incorporated in the constitutive model. This Information need be supplied explicitely to force the
stress point during continued plastic shear flow to the apex of the failure surface (Fig. 2), where alone
subsequent gap can take place [17].

Neglecting the cementitious cohesion, the nonassociated slip rule and shear softening still violate
the assumptions of limit analysis [13]. Only in statical determinate situations, where the amount
of dilatancy does not play a role, certain limit load formulae remain valid (Fig. 3); but for highly
confined situations as typical in geotechnics the limit load decreases substantially with rp < <p [18].
Limit load theorems in their classical sense - i.e. the maximum lower bound and the minimum upper
bound converging to a unique value - are no longer valid but need be recast in a weak form furnishing
'safer' lower and higher upper bound values [19,20]. On the basis of associated flow, Solutions with
finite element interfaces can still be obtained by optimization methods [21].
With the exception of blast loading and other energy-based design cases, upper-bound Solutions are
of little value in civil engineering practise anyhow. Through the prudent choiee of material parameters
one strives rather at obtaining conservative limit loads in spite of an underlying mechanism concept.
Very good results have been obtained with interface elements for difficult lhnit load problems [22].
Pre-inserting planes of discontinuity without tracking their formation (i.e. strain localisation) means
that part of the stress history is neglected in favour of a limit equilibrium analysis for a mechanism
which is not necessarily the one that would actually develop. As with plastic lhnit theory one must
therefore check also the yield criterion in the solid domains between the planes of weakness and the
strain limits and transient strength components, which - depending on the unknown stress history -
may undermine the füll mobilisation of the mechanism's resistance [23]. It seems thus very helpful if
interface element constitutive modeis dispose of an initial cementitious strength with the capability
for mixed-mode decohesion, so that they can be inserted in a mesh as 'sleeping discontinuities' in the
sense of Hillerborg's fictitious cracks.

4. EXAMPLES OF VOUSSOIR ARCHES AND BEAMS

To conclude this contribution, two simple applications to arches are given, which are either supposed
to be construeted from independent blocks or to be radially cracked. Such voussoir arches are a
classical application of rigid-plastic limit analysis even though the modification of the plastic-hinge
concept to aecommodate no-tension gaps between blocks seems rather bold [24]. The simplicity lies in
-
the fact that slipping of blocks is exeluded from the catalogue of allowable mechanisms postulating
a sufficiently steep orientation of the force resultant with respect to the interfaces -, that the stress
ränge is supposed to be low enough to avoid crushing of edges and that the joints have no tensile
strength. Therefore no energy will be dissipated in the mechanism, and the energy balance must be
162 PLANES OF WEAKNESS IN FINITE ELEMENT ANALYSIS

1152 4mm
l3M2Jn(io/9

3048mm

smeared cracking 8 Gauss points through thickness

joints 2x3 Gauss points through thickness

joints 3 Lobatto points through thickness

Figure 4: Discrete Joint model of a circular arch

maintained by passive external work of parts of the structure moving against the direction of loading.
According to lhnit analysis theory, any feasible thrust line which lies fully inside the arch would thus
give a lower-bound limit load, whereas any collapse mechanisms would give an upper-bound limit load
[25]. The added advantage of the FEM discretization of the joints is that the no-slip assumption is
checked automatically, i.e. La Hire's vision of 1695 of arches as an assembly of wedges (viz. [26]) is
300 years later turned into a practical method.

The example in Fig. 4 shows a circular arch, which was tested in 1951 by Pippard & Chitty and
previously analysed by mechanism and continuum FE methods [27]. Modelling every segment as a
finite element separat ed by interfaces, it can be seen that the computed extent ofJoint opening - shortly
before the fourth mechanism is formed - corresponds quite well to the prediction by the smared-crack
model. This may surprise as it is often anticipated that the discrete model will automatically lead
to a concentrated mechanism, but it finds an explanation in the tangential orientation of the thrust
line and indicates that not all the joints would have to be included to catch the failure mechanism
[25]. Observe also that the distance between the two outermost Integration points determines the
eccentricity of the pivot and hence the effective depth of the section in which the thrust line must
reside. Other Integration schemes - among them a socalled floating point scheine, which contracts
the Integration points into the remaining compression zone - have been tested [28], but the results for
only one Joint element across the thickness are seen to be rather satisfactory if a 3-pt. Lobatto rule
(nodal Integration) is used. Note also that the solids between partly open joints still exhibit tensile
stresses, due to the coupling of equivalent nodal forces through the shape functions; this emphasizes
once more the advantage of stress evaluation in discrete mechanisms.
The second case concerns the rather common problem of estimating the load carrying capacity
of an unreinforced concrete beam by considering a hidden arch, even though in this particular case
the 'beam' is a horizontal slice through a large concrete gravity dam under reservoir pressure [29].
According to the lower-boxmd theorem, any permissible stress field - i.e. not violating the yield limits
of the material anywhere - would give a safe estimate of the load carrying capacity, irrespective of the
 J.-M. HOHBERG 163

74cmi"r7i
I
* x
1 E= 20 GPa £
v=0 2 ',,

11 —
'i.

^~^_„
\
JL
„.
' )m
\ ©
(Prpstress) — Abutment Displacement

117cm

®
8 10 12 M* 16 cm
*- +25cm

Beam Deflection

Figure 5: Hidden horizontal arch in a 'tension-free' beam

strain compatibility [3]. The maximum load will thus result from the arch with the largest camber,
so that paradoxically the (elastic) beam seems the stiffer the deeper it is cracked in flexure. In terms
of stress resultants, the arch is only stable if the bending moment does not exceed the normal force
times half the depth of the cross-section, as otherwise the thrust line would pass outside the structure
[25]. The problem with this particular loading is that the bending moment is already active before a
sufficient normal force can build up. It would not arise if the Segments were wedges [30], but as the
joints are oriented parallel to the direction of loading, the thrust requires prying action in bending
which is unstable under small pressure (points 'x' in the graph). If one does not count on residual
prestress from Joint grouting or cementitious cohesion - but takes rather some foregoing Joint opening
due to shrinkage or cold temperature into aecount
-, the only way how such a voussoir beam could
work without shear keys would be by means of considerable dilatancy developing during the relative
slip between blocks [31]. Even then an absolute limit would be given by the height of asperities as
previously mentioned.

5. CLOSING REMARKS
Interface elements to model weak planes or existing discontinuities are a very useful tool for lhnit
equiUbrium calculations. The conceptual similarity to upper-bound limit analysis lies in the need
to perturbate prospective mechanisms for finding the most eritieal one, but fortunately there are
many situations (like well-shaped arches) which are rather insensitive to the assumed location of
discontinuities. However, phenomena of limited strain capacity and transient strength components
need be regarded if they are not to defy the analysis results. The influence of more realistic interface
constitutive modeis in the FEM may also be elueidating to limit analysis practise.

REFERENCES
1. Rots J.G. and Blaauwendraad J., Crack modeis of concrete: discrete or smeared? fixed, multi-directional or
rotating? fferon 54 (1), 1989.
 164 PLANES OF WEAKNESS IN FINITE ELEMENT ANALYSIS

2. PlAN T.H.H. and Tong P., Basis of the finite element method for solid continua. Int. J. Num. Mcth. Eng. 1,
1969, pp. 3-28.
3. Drucker D.C.,
On structural concrete and the theorems of limit analysis. IABSE Memoires 21, 1961, pp. 49-59.
4. SALENQON J., Calcul a Ja rupture et analyse limite. Presses ENPC, Paris 1983.
5. MÜLLER P., Plastische Berechnung ron Stahlbetonscheiben und -ballen. ETH Zürich IBK Rep. 83, Basle 1978.
6. CHEN W.-F. and DRUCKER D.C., Bearing capacity of concrete blocks or rock. J. Eng. Mech. Dir. ASCE 95 (4),
1969, pp. 955-9T8.
7. Marti P., Plastic analysis of reinforced concrete shear walls. IABSE Coli. Plasticity of Reinforced Concrete,
Copenhagen 1979. Introd. Rep. AK-28, pp. 51-69.
8. JENSEN B.C., Lines of discontinuity for displacements in the theory of plasticity of piain and reinforced concrete.
Mag. Concrete Res. 27 (92), 1975, pp. 143-150.
9. Des AI CS., ZAMAN M.M. et al., Thin-layer element for interfaces and joints. Int. J. Num. Anal. Meth. Geomech.
8 (1), 1984, pp. 19-43.
10. HOHBERG J.-M., discussion of 'Modeling of cyclic normal and shear behavior of interfaces' (by CS. Desai and
B.K. Nagaraj). J. Eng. Mech. ASCE 116 (8), 1990, pp. 1870-1880.
11. Hohberg J.-M. and Schweiger H.F., On the penalty behaviour of thin-layer elements. NUMOG IV, Swansea
1992 (in print).
12. HOHBERG J.-M., A review of Joint constitutive modeis. 2nd NUMEG, Prague 1992 (in print).
13. Drucker D.C, Coulomb friction, plasticity and limit loads. J. Appl. Mech. ASME 21 (1), 1954, pp. 71-74.
14. HOHBERG J.-M., Multimechanism plasticity with coupled damage in tension and shear. COMPLASIII, Barcelona
1992. Proc. 2, pp. 1503-1514.
15. Exner H., On the effectiveness factor in plastic analysis of concrete. IABSE Coli. Plasticity in Reinforced
Concrete, Copenhagen 1979. Final Rep. AK-29, pp. 35-42 + disc. pp. 67, 132/133, 349, 353.
16. Kupfer H. and Bulicek H., Comparison of fixed and rotating crack modeis in shear design of slender concrete
beams. Int. Workshop Progress in Structural Engineering, Brescia 1991 (typescript 10 p.)
17. HOHBERG J.-M, A Joint element for the nonlinear dynamic analysis of arch dams. ETH Zürich IBK Rep. 186,
Basle 1992.
18. De Borst R. and VERMEER P.A., Possibilities and limitations of finite elements for limit analysis. Geotechnique
34 (2), 1984, pp. 199-210 + disc. 35 (1), 1985, pp. 90-94.
19. De Josselin de Jong G., Lower bound collapse theorem and lack of normality of strain rate to yield surface
for soüs. IUTAM Symp. Role of Plasticity in Soil Mechanics, Grenoble 1964. Proc. (1966), pp. 69-75 + disc. pp.
75-78.
20. COLLINS I.F., The upper bound theorem for rigid/plastic solids generalized to include Coulomb friction. J. Mech.
Phys. Solids 17, 1969, pp. 323-338.
21. Sloan S.W., Upper bound limit analysis using finite elements and linear prograrnrning. Int. 3. Num. Anal. Meth.
Geomech. 13 (3), 1989, pp. 263-282.
22. Van Langen H. and Vermeer P.A., Interface elements for singular plasticity points. Int. J. Num. Anal. Meth.
Geomech. 15 (5), 1991, pp. 301-315.
23. Manfredini G., Martinetti S. and Rtbaccht R., Inadequacy of limiting equilibrium methods for rock slope
design. 16th US Symp. Rock Mechanics, Minneapolis/MN 1975. Proc. (1977), pp. 35-43.
24. Kooharian A., Limit analysis of voussoir (segmental) and concrete arches. ACI Journal 49, 1952, pp. 317-328
+ disc. pp. 328/1-4.
25. HEYMAN J., The stone skeleton.Int. J. Solids Struct. 2 (2), 1966, pp. 249-279.
26. Kurrer K.-E., Zur Entstehung der Stützlinientheorie. Bautechnik 68 (4), 1991, pp. 109-117.
27. Crisfebld M.A., Finite element and mechanism methods for the analysis of masonry and brickwork arches.
TRRL Res. Rep. 19, Crowthorne 1985.
28. Hohberg J.-M. and BACHMANN H., A macro Joint element for nonlinear arch dam analysis. 6th ICONMIG,
Innsbruck 1988. Proc. 2, pp. 829-834.
29. HOHBERG J.-M., discussion of 'Spatial action of straight gravity dams in narrow Valleys' (by M.A. Herzog). J.
Struct. Eng. ASCE 117 (2), 1991, pp. 637-641.
30. LrviSLEY R.K., Limit analysis of structures formed from rigid blocks. Int. J. Num. Meth. Eng. 12 (12), 1978,
pp. 1853-1871.
31. Pender M.J., Prefailure Joint dilatancy and the behaviour of a beam with vertical joints. Rock Mech. Rock Eng.
18 (4), 1985, pp. 253-266.
ßk

Remaining Structural Capaci

Capacitö portante residuelle d'une fon
Resttragfäahigkeit eines Turb

Jana MARGOLDOVÄ Radomlr PUKL

Dr. Eng. Dr. Eng.
Czech Techn. Univ. Czech Techn. Uni
Prague, Czech Rep. Prague, Czech Re

J. Margoldovä received her Pd.D. R. Pukl

\ m

received his
degree in 1987 from the Czech in 1985 from the Cz
Technical University in Prague. University in Prague
She works as a research engineer the field of computa
166 REMAINING STRUCTURAL CAPACITY OF POWER PLANT FRAME

INTRODUCTION

Heavy prefabricated reinforced concrete frames were typically used for construction of large
thermal electric power plants in Czechoslovakia in the 50s and 60s within an extensive energy-
production plan. These power plants burn low-quality brown coal and are the main source
of energy in Czechoslovakia. In the course of time number of problems have emerged in
connection with the servise of these plants. To mention only the most serious ones: uncontrolled
environmental polution and structural damages due to heavy service loading. The most
exposed structures are the reinforced concrete foundation frames of turbines. They are subjected
to large static and dynamic loadings, chemical and thermal effects. Today in many cases these
structures are also at the end of their designed life time. This life time is about half of that
of similar unexpposed structures. It is in the interest of the electric power industry to extend
the life service of these structures and thus to avoid building new plants. This tendency is
also evident world-wide. In this context the technical diagnostic is becoming the important
engineering branche. It is expoliting the reliability theory, the structural modeling and on
site investigations. The last two mentioned categories were used in the present report for the
diagnostic of the remaining structural capacity.

The concerned power plant is located in the North-West Bohemia and has been under recon-
struction. It had been subjected to the long term monitoring to determine the extent of wearing
after twenty years of service. In order to asses the remaining structural capacity of the turbine
foundation the structure was analyzed by the finite element program SB ETA. The damage and
failure states of the structure were simulated. The goal of this analysis was to simulate the
effects of the poorly manufactured joints of the precast members on the load carying capacity.
The results served to design the measures for the necessary reconstruction.

2. PROGRAM SBETA
Program SBETA was recently developed at the Institute of Material Science of the University
of Stuttgart in Cooperation with the Klokner Institute of the Czech Technical University in
Prague. It is a commercial program designed for the analysis of reinforced concrete structures
in the plane stress state. It can predict the response of complex concrete structures, with or
without reinforcement, in all stages of loading, including failure and post-failure. It can be used
to analyze the remaining structural capacity of existing structures. Details about the program
and its constitutive model can be found in papers [5,6] and documentation [8]. Here only a
brief description is given. The other applications of the program are reported in ref. [2,3,4,7]

The constitutive model in SBETA is based on the smeared material approach with isotropic
damage model in uncracked concrete and orthotropic damage after cracking. The behavior of
concrete is described by the stress-strain diagram, which is composed of the four branches:
nonlinear loading in compression, linear loading in tension, and linear softening in both, tension and

compression. The parameters of this diagram are adjusted according to the plane stress state
using the biaxial failure function of Kupfer for compression. The mechanics of cracked reinforced
concrete, which is relevant to this study case, includes: (a) reduction of compressive strength
in direction parallel with cracks; (b) variable shear retention factor; (c) tension stiffening. All
these properties are controlled by the tensile strain, which reflects the crack opening. The
nonlinear fracture mechanics is introduced by means of the Bazant's crack band theory [1].
The tension softening modulus is adjusted for each element according to the fracture enegry.
Both, fixed and rotated crack modeis are implemented. Reinforcement behavior is bi-linear.
 J. MARGOLDOVÄ - R. PUKL - V. CERVENKA 167

A four-node quadrilateral finite element is used for the concrete. The reinforcement can be
included either in a smeared form, as a part of the concrete element, or discrete, as a bar
element passing through the quadrilateral element. The updated Lagrangean formulation is
adopted allowing the modeling of a second order geometry effect. The non-linear Solution
is performed by means of a step-wise loading and by an equilibrium iteration within a load
step. Newton-Raphson and arc-length methods are the options for the Solution strategy in the
equilibrium iteration.

The program system SBETA includes a pre-processor, a Solution program, and an efficient
post-processor. The finite element analysis can be interactively controlled and runs in several
levels of real-time graphics. Thus, the Solution process can be observed and Solution parameters
can be adjusted by a user if necessary. A restart Option is available. The post-processor
generates automatically deformed shapes and images of stress, strain and damage fields (cracking,
crushing). All results of the analysis presented in this paper are produced by the SBETA
post-processors.

3. ANALYSIS OF DAMAGED POWER PLANT FRAME

3.1 Girder Joint in Precast Foundation Frame
The schematic plan view of the frame is in Fig.l. The precast reinforced concrete girders A and
B are supported by column. Each girder is formed by a couple of two identical girders, A\, A2
and JE?i, B2. The interaction of the coupling girders is assured by the slab, which is casted over
the girders. The Joint of girdes is located above the column. During the construction a space
of 350 mm had been left between the vertical end faces of adjacent girders, the reinforcement
was welded and the Space was filled by concrete. The location of the Joint is denoted in
Fig.l. There were doubts about the quality and correct casting of the filier concrete and
about füll interaction of parallel longitudinal girders. The diagnosis of the frame was done
by means of nondestructive testing methods combined with visual investigations supported by
endoscop. However, a detailed investigation of this Joint was impossible because of technological
obstructions. The main purpose of the structural analysis was to make an estimate of the
function of the Joint under various assumptions of quality of the filier concrete.

3.2 Finite Element Model

In the linear analysis of the whole space frame, which was performed also for other purposes,
the function of the structural detail of Joint in the global structural system was studied. On
the basis of this global analysis the region of damaged Joint was identified. From the complex
structure of the frame only a section adjacent to the Joint was modeled. The surrounding
structure was approximated by appropriate boundary conditions and artificial Springs.

The analytical model, its geometry, boundary conditions and reinforcing are shown in Fig.2.
The finite element mesh is shown in Fig.3. It has 387 quadrilateral elements. The vertical
stirrups were modeled by smeared reinforcement and all main horizontal and inclined bars were
modeled by discrete reinforcement. The concrete quality of girders identified from core-drilled
samples was 28 MPA. The nominal quality of filier concrete in the Joint was 34 MPA. However,
there were doubts about the quality of its casting and its actual state could not be reliably
verified. Therefore, variable properties of the filier concrete were considered in this study by
168 REMAINING STRUCTURAL CAPACITY OF POWER PLANT FRAME

values of 1%, 10%, 60% and 100% of its nominal value.

Two alternatives of boundary conditions were considered for the modelling of the surrounding
structure. In the first one the column under the Joint of the longitudinal girders was modeled
by Springs and the continuity of the structure was modeled by conditions of the symetry in
the middle of the girders. In the second case the support on the column was rigid and the
continuity of the girders was modeled by the Springs. This enabled the approximate Simulation
of the axial displacements of girders due to flexibility of the frame structure.

The loading is due to technological forces which are trasfered through numerous fastenings as
indicated by vertical arrows in Fig.2. The force in the load-displacement diagram refers to the
sum of the technological forces. Two loading cases were considered. In the first loading case
the füll interaction of girders A\, A2 and B\, B2 was assumed and the girders were loaded by
the half of the total technological load. (The structure was designed under this assumption.)
The second loading case assumed no interaction between the girders and the füll technological
loading was applied on the internal girders. In both cases the dead load of girders was included.

The study was performed on the personal Computer 286 under MS-DOS operating system.
Solution of one case on this Computer took about 10 hours of Computer time. (Of course in
case of PC486 the time would be much shorter.) In addition to the ultimate load capacity each
analysis provided ample of informations on stress' and strain state, crack patterns and failure
mode. The fixed crack approach has been used in all cases.

3.3 Discussion of Results

The behavior of analyzed system is illustrated in Fig.4., which shows the crack patterns in
three load stages and the failure state with cracks, crushing and amplified deformations. The
load displacement diagram with the load levels corresponding to the stages in Fig.4 is shown
in Fig.5. The yielding of reinforcement is also graphically indicated (but apparent only from
coloured output). Strong shear behavior is evident from inclined cracking. The failure mode was
dependent on the degree of lateral constraint and the quality of the filier concrete in the Joint.
In case of a high lateral constraint and the 100% quality of the filier concrete the maximum
load was 4.75 times of the admissible loading. The failure mode is of the concrete arch-type,
with crushing of the cracked concrete in the web and in the bending compression zone.
The effect of the filier concrete quality on the frame behavior can be seen from the comparison
of Figures 4(c) and 6. The load level is 2 times of the admissible load. The quality of the
filier concrete described by the compressive strength was 100% in Fig.4(c), 60% in Fig.6(a)
and 10% in Fig.6(b). In case of the lowest quality, Fig.6(b), the filier concrete in the Joint
fails in compression and after that the behavior is fairly ductile with all major reinforcement
yielding. The ultimate load factor, (related to the admissible load) is in this case 3.5. The
reduction of load capacity was also caused by partially releasing the lateral constraint. In the
most unfavorable case, with the elastic Springs modeling the lateral constraint and 1% of filier
concrete quality the ultimate load was almost equal to the admissible load with no marginal
safety.

The results of the numerical analysis were used to support the design of measures for extending
the service life of the frame, which were based on the restoring the füll interaction of girders
and strengthening the space frame. They were also utilized in a reliability analysis. It was a
valuable contribution to the safety and economy of the engineering solutuion.
 J. MARGOLDOVÄ - R. PUKL - V. CERVENKA 169

4. CONCLUSIONS
The Computer program SBETA was succesfully applied to the assesment of the remaining
structural capacity of the turbine foundation frame. The nonlinear finite element analysis
proved to be a rational method for determination of ultimate load capacity of this statically
indetermined structure, whose behavior significantly deviates from the simple design modeis
based on cross-section al analysis. The FE analysis was used to design the economical and
rational repair measures

REFERENCES
[1] BAZANT, Z.P., OH, B.H. - Crack Band Theory for Fracture of Concrete. Mater. Struct.
RILEM, Paris, France, 16, 1983, pp. 155-177
[2] CERVENKA, V., PUKL, R., ELIGEHAUSEN, R. - Computer Simulation of Anchor-
ing Technique in Reinforced Concrete Beams. Computer Aided Analysis and Design of
Concrete Structures, Proceedings 2nd SCI-C International Conference, Eds. N. Bicanic,
H. Mang, Zell am See, Austria, April 1990, Pineridge Press, Swabsea, U.K. 1990, ISBN
0-906674-74-3, pp. 1-20

[3] CERVENKA, V., PUKL, R., ELIGEHAUSEN, R. - FEM Simulation of Concrete Frac¬
ture. ECF 8 Fracture Behavior and Design of Materials and Structures, Preprints of the
8th Biennial European Conference on Fracture, Torino, Italy, October 1990, pp. 728-733

[4] CERVENKA, V., ELIGEHAUSEN, R., PUKL, R. - SBETA Computer Program for Non¬
linear Finite Element Analysis of Reinforced Concrete Structures. IWB Mitteilungen
1990/1, Stuttgart University, ISSN 0932-5921, ISBN 3-9801833-6-X

[5] CERVENKA, V., ELIGEHAUSEN, R., PUKL, R. - Computer Models Of Concrete Struc¬
tures. IABSE Colloquium "Structural Concrete", Stuttgart, Germany, April 1991, ISBN
3-85748-063-7, pp. 311-320

[6] CERVENKA, V., PUKL, R. - Computer Models Of Concrete Structures. Structural

Engineering Int., Vol. 2, No. 2, May 1992, IABSE Zürich, Switzerland,
ISSN 1016-8664, pp. 103-107

[7] CERVENKA, V., PUKL, R., ELIGEHAUSEN, R. - Fracture Analysis of Concrete Plane
Stress Pull-Out Tests. Fracture Processes in Concrete, Rock and Ceramics, Noordwijk,
Netherlands, June 1991, pp. 899-908

[8] SBETA - Nonlinear Finite Element Analysis of Reinforced Concrete Structures, Software
and Documentation, Peekel Instruments, Industrieweg 161, 3044 AS Rotterdam, The
Netherland.
170 REMAINING STRUCTURAL CAPACITY OF POWER PLANT FRAME

0,50
ä : c
o
o
c^
A2
D [
A1 B1
8,40 9,30 8,00 5,001

Fig.l Power Station turbine frame: schematic plan view (dim. in meters).

* ,* I
IT)
CO

A2 B2
1,10

Fig.2 Structural model of a detail with reinforcement, loading

and boundary conditions.

^ ^

IIJJIL^ ^ JJ 11111III M 111111 [

II |<

Fig.3 Finite element mesh of the structural model.

Jt J. MARGOLDOVÄ - R. PUKL - V. CERVENKA 171

I I *
(a)

P=2.7 MN
§

*
(b)

P=5.9 MN

I
(c)

P=7.7 MN

I V
(d)

P=9.5 MN

Fig.4 Sequence of crack patterns.

172 REMAINING STRUCTURAL CAPACITY OF POWER PLANT FRAME

Load [MN3

10. 00

6. 000

0.000 1.000 2.000 3.000 4.000

Displacement [mm]
Fig.5 Load-displacement diagram.

*
(a)

*
(b)

Fig.6 Comparison of damage states for cases of different Joint concrete qualities.
Load factor 2.
 Analytical Models for Strength and Stiffne
Modeles pour l'6valuation de la resistance
Modelle für die Berechnung von Tragfähigk

Filip C. FILIPPOU Fabio F. TAUCER

Associate Professor Junior Researche
University of California University of Cali
Berkeley, CA, USA Berkeley, CA, US

Filip C. Filippou, born 1955, received Fabio F. Taucer, bo

his civil engineering degrees at his civil enginee
TU München and at the University the Universidad Ce
of California, Berkeley. Since 1983 Caracas and
he is on the faculty at the University of California, B
of California, Berkeley with involved in modelin
174 STRENGTH AND STIFFNESS EVALUATION OF CONCRETE STRUCTURES

1. INTRODUCTION
The development of rational measures for strengthening and retrofitting of existing concrete structures
depends on advanced methods of assessing their strength and stiffness. These methods should
be capable of predicting the future behavior of the entire structure based on information about the
original design and the current State of the structure. The latter can be usually approximated from
current measurements of material and structural properties using System identification methods. In
regions of high seismic risk the difficulty of the problem is compounded by the complex loading
history of existing structures, which might have experienced several small and moderate earthquake
excitations in their service life.
The evaluation of the future behavior of existing concrete structures depends on the development of
advanced analytical modeis, which describe the time and load dependent nonlinear behavior of the
structural members. These modeis should satisfy two basic requirements: (a) they should be reliable,
robust and computationally efficient and (b) they should be of variable complexity depending on the
degree of detail required from the analysis: while individual eritieal members of the structure can be
evaluated with sophisticated finite element modeis, the behavior of multistory buildings and multiple
span freeway structures can be described with sufficient aecuraey with member modeis. In fact, the
ability to mix finite element modeis of eritieal regions of the structure with nonlinear or even linear
member modeis of the rest of the structure should be an important consideration in the development
of such modeis.
In the following, a new fiber beam-column finite element for the analysis of reinforced concrete
structures is presented. Contrary to most existing beam finite elements which are based on the
definition of displacement shape functions, the element described herein assumes a constant axial force

and linear bending moment diagrams inside the element, thus assuming force shape functions. A
general overview of the element formulation and of the element nonlinear iteration scheme needed
for the element State determination is first presented, followed by the description of a few numerical
examples in which the element response is compared with experimental results

2. ELEMENT FORMULATION
The beam-column element is shown in the local reference System x,yj. in Fig. 1. The element is
represented without rigid-body modes, thus forces and deformations are measured with respect to the

cord connecting the two end nodes. Forces and displacements are grouped in the following vectors:
Element force vector Q {& Qi ß3 Qa QsY 0)
Element displacement vector q {qx q2 Q3 q* Qs} (2)

Similarly, section forces and deformations can be grouped in the vectors:

Section force vector D(jc) {Ml (x) M2(x) N(x)}T (3)

Section deformation vector d(x) {jCi (*) X2 M e(x)} (4)

The element is composed of a finite number of longitudinal fibers Each cross section is therefore
described by the number of fibers, their area, location and force-deformation relations. Since the
element has been developed for the analysis of reinforced concrete structures, concrete and steel
constitutive modeis have been used [1]. Small kinematics are postulated and plane sections are
 F.C. FILIPPOU - F.F. TAUCER - E. SPACONE 175

assumed to remain plane and normal to the longitudinal axis. Consequently, the effects of shear and
bond-slip are neglected in the present model. The nonlinear nature of the problem depends entirely
on the nonlinear fiber force-deformation relations. The element formulation is based on the assumption
that the axial force is constant and the bending moment diagram is linear inside the element. In
Symbols this translates to a simple relation between section and element forces:

D(jc)=b(jc).Q and AD(jc) =b(x) • AQ (5)

jments and b(x) is defined by:

"(!-) (i) ° °

° - (H °
(6)

0 0 0 0 1

The force field, as defined by

(5), is exact as long as only
*).Xz(x)
nodal forces act on the
element. Loads acting inside the
<iY
^-r element can be easily
N(x).e(x) introduced using the procedure
described in [1].
M,x.xix

<V*
Q4fl4

\ Q*q
(th) fiber
A/nat
The element is formulated
using the flexibility method
rather than the classical stiffness
method, because of the
advantage of defining an
Q..q "exact" force field inside the
element. Calling P-Q the
element unbalanced forces
(difference between applied
and resisting forces P and Q
respectively) and Aq the
FIGURE 1 BEAM ELEMENT FORCES AND DISPLACEMENTS WITHOUT
element deformation increments,
RIGID BODY MODES IN LOCAL REFERENCE SYSTEM:
FIBER DISCRETIZATION OF CROSS SECTIONS the nonlinear system of equations
at the element level is
written:
[F]-1Aq=(P-Q) (7)

In Eq. 7 the element stiffness appears as the inverse of the element flexibility to indicate that the
element is flexibility-based. The element flexibility matrix is determined integrating the section flexi-
bilities according to:

(8)
F jbT(x)-f(x)b(x)-dx
176 STRENGTH AND STIFFNESS EVALUATION OF CONCRETE STRUCTURES

Section flexibility is obtained by inverting the section stiffness. The element is implemented in a stand
alone program organized along the lines of a typical finite element code. Loads are applied on the
structure and the program computes the corresponding displacements. The nonlinear Solution procedure
is organized as follows:
Load increments AP are applied to the structure and a Newton-Raphson scheme is used to compute
the corresponding structure displacement increments. At every Newton-Raphson iteration it is necessary
to determine the element resisting forces corresponding to the updated element displacements.
This is a challenging task when working with a flexibility-based element, because force and not
displacement shape functions must be used. A new scheme has been developed for the proposed
element, based on residual section and element deformations. Given the updated element displacements,
the following Steps are performed:

1) Compute the element linearized force increments using the last computed element tangent stiff¬
ness, and update the element forces;
2) Compute the section force increment using (5);
3) Compute the section deformation increment using the last computed section flexibility;
4) From the new section deformations, using the hypothesis that plane sections remain plane and
normal to the longitudinal axis, compute the new fiber strains;
5) Compute fiber stresses and tangent moduli using the fiber force-deformation relations;
6) Compute the new section tangent stiffness, the section resisting forces and the section unbal-
anced forces, difference between applied and resisting forces;

7) Transform the section unbalanced forces into section residual deformations using the section
flexibility;
8) Integrate the section residual deformations to compute the element residual deformations;
9) Compute the element flexibility using (8);
10) Compute the new element force increments;

Step 10) is needed because the element residual deformations can not be applied to the element
alone, otherwise node compatibility would be violated. Forces based on the new element stiffness are
applied to the element in order to yield element displacements equal and opposite to the element
residual deformations. Correspondingly, force and deformation increments are applied to all sections:
these increments are computed repeating steps 3) through 9) until convergence is achieved. The
element converges when the unbalanced forces at all sections are sufficiently small. Element convergence

implies that the element resisting forces corresponding to the applied displacements have been
computed and the following Newton-Raphson iteration can be performed.
The new element convergence scheme is based on the equilibrium conditions (5). It can be shown
that during the iterations equilibrium and convergence inside the element is respected, and section
force-deformation relations are satisfied, at least within a certain tolerance, when convergence is
reached. More details on the approach and a thorough description of the iteration scheme are
presented in [1].

3. EXAMPLES
A series of comparisons between analytical and experimental results are used to study the element
Performance. Four examples are illustrated in this section: these refer to three reinforced concrete
cantilevers discretized with a single beam-column element. Displacement control techniques have
 F.C. FILIPPOU - F.F. TAUCER - E. SPACONE 177

been used to match experimental and analytically imposed displacements: a very strong linear elastic
spring has been positioned at free end of the cantilever to control the tip displacements.
The first example shows the
30 uniaxial bending of a

20 - i$
' ZU_
»p cantilever beam Rl with a
rectangular cross section
tested in [2]. The Simulation
H H
of the tip displacement
10
response in the strong
direction v is shown in Fig. 2.
o Analytical and experimental
I
Cl 10
results agree well, especially
for displacements up to
<
yielding of the built-in end. At
-20 - this point bond-slip and shear
deformations become
-30 -
important and since the
i_r 1 1

-2 -1 0 element does not include such

Tip Deflection (in) effects, the analytical and
experimental results show
FIGURE 2 TIP LOAD-DISPLACEMENT RESPONSE OF CANTILEVER some discrepancy.
BEAM UNDER CYCLIC UNIAXIAL BENDING
The remaining examples refer
to the bending behavior of a cantilever under a compressive axial load and biaxial or uniaxial bending
moments, which was tested in [3]. Fig. 3 illustrates the uniaxial case in which a constant axial force
and a cyclic force along the weak axis z are applied at the tip of the cantilever. Displacement control
was not used in this example. Numerical and experimental results are very similar and show a stiffer
fiber model behavior, especially for low levels of lateral force at which "pinching" is evident in the
experimental results.
The same cantilever is
studied under biaxial

i
bending conditions. Two
cyclic transverse loads are P,-10kips
applied at the free end of -3J*"S_
the cantilever. Displacement H es: ^
control is used in this -z? 2
example. Fig. 4 shows the
r-r o
tip response in the strong
direction y. The correlation
between analytical and
experimental results is very analysis
experiment
good both for small and
large displacements. When
-1 2 0 8 -0 4 0.4 0.8 1.2
the concrete is fully cracked
at the built-in section, bond-
Tip Displacement z (in)
slip effects appear in the
experimental data, but their
contribution to the tip FIGURE 3 TIP LOAD-DISPLACEMENT RESPONSE OF CANTILEVER
displacements is small. UNDER CONSTANT AXIAL LOAD AND CYCLIC UNIAXIAL
BENDING: NUMERICAL AND EXPERIMENTAL RESULTS
178 STRENGTH AND STIFFNESS EVALUATION OF CONCRETE STRUCTURES

Finally, the same cantilever

^=#-
beam is studied under both
ups
cyclic bending and cyclic axial
47 force. According to the nota-
tion of Fig. 5, the following
load and displacement histories
have been imposed:
Px= -75 ±30kips
py=±0.96m
pz=±0.96w
analysis
Load and deformation
experiment
applied so that
increments are
cycles simultaneous: all are
-08-06-04-02 0 02 04 06 08 10 three quantities reach their
Tip Displacement y (in) maximum and minimum values
at the same time. This example
Figure 4 - Tip Load-displacement response in the strong is particularly important to
direction v of a cantilever beam under constant show the capability of the
axial load and cyclic biaxial bending: proposed element to represent
numerical and experimental results. softening and stiffness degradation
without any computational difficulües. This is due to the fact that force equilibrium is always
maintained along the element. When softening mitiates at the built-in section, the whole beam
unloads respecting the prescribed linear bendmg moment diagrams. All sections unload elastically
except for the built-in section, which softens. Correspondingly, the end curvature increases while
curvatures at all other sections decrease.

8- 4. CONCLUSIONS
6- -
To predict the response of
4- -^-^^g / " / existing reinforced concrete
structures to strong ground
2- motions and to develop better
n
U - and retrofit
~0 strengthening
^^~
CÜ
O
-2- measures for structures in
zones of high seismic risk inte-
-4 - M 5 ±30 kips
grated experimental and
CO analytical studies are very important.

The beam-column fiber

CO h
1—^— element presented in this paper
1 1 1
r 1
-1
-1 -08-06-04-02 0 02 04 06 08 1 is part of an ongoing effort to
Tip Displacement y (in) develop reliable computational
tools of different levels of
FIGURE 5 - TIP LOAD-DISPLACEMENT RESPONSE IN STRONG complexity and, thus, reliability,
DIRECTION y OF CANTILEVER UNDER CYCLIC AXIAL for modeling reinforced
LOAD AND CYCLIC BIAXIAL BENDING concrete structures. Most two-
node reinforced concrete finite
elements are based on the stiffness approach which postulates linear curvatures and constant axial
strain along the element. These deformation distributions do not represent the physical behavior
 F.C. FILIPPOU - F.F. TAUCER - E. SPACONE 179

when sections Start yielding. The proposed finite element is based on the assumption of linear bending
moment diagram and constant axial force along the element. This hypothesis is exact when no
load is applied inside the element. The computational cost for each element is higher when a flexibility
based element is used, because of the iteration scheme necessary to compute the element resisting
forces corresponding to the applied displacements. However, fewer elements are needed to discretize
the structure, thus requiring a smaller number of total degrees of freedom. Further refinements of the
element are needed to include bond-slip and second order effects.

5. REFERENCES

1. TAUCER F.F., SPACONE E. and FILIPPOU F.C, A Fiber Beam-Column Element for Seismic
Response Analysis of Reinforced Concrete Structures. UCB/EERC-91/17, Earthquake
Engineering Research Center, University of California, Berkeley, 1992.
2. MA S.M., BERTERO V.V. and POPOV E.P., Experimental and Analytical Studies on the
Hysteretic Behavior of Reinforced Concrete Rectangular and T-Beams. Report UCB/EERC-
76/2, Earthquake Engineering Research Center, University of California, Berkeley, 1976.
3. LOW S.S. and MOEHLE J.P., Experimental Study of Reinforced Concrete Columns Subjected
to Multi-Axial Cyclic Loading. Report UCB/EERC-87/14, Earthquake Engineering Research
Center, University of California, Berkeley, 1987.
180
 181

SESSION 3

ANALYTICAL EVALUATION OF BRIDGES

182
 183

KEYNOTE SPEAKER
184
 Finite Element Prediction of a Des
Prediction de la Charge de rupture d'un
Vorausberechnung einer Beto
Feldversuch mittels F

Rene" de BORST Cor van der VEEN

Prof. Dr. Dr. Eng.
Delft Univ. of Technology Delft Univ. of Tech
Delft, The Netherlands Delft, The Netherla

i.
m

,f.lri Mi

Ren6 de Borst, born in 1958, Cor van der Veen,

received his Ir.degree in civil received his Ir. degre
engineering in 1982 and his Dr. engineering in 1982 and

degree in 1986, both from Delft in 1990, both from D

University of Technology. He is of Techbology. He
Professor of computational mechanics senior reseacher in
186 FINITE ELEMENT PREDICTION OF A DESTRUCTIVE FIELD TEST OF A BRIDGE

1. INTRODUCTION
After a thirty-year development the finite element method has become a powerful tool for
analysing structural behaviour. By now, deflections and stresses under service load levels can be
predicted within a tolerance that is narrower than the scatter in material properties and the
uncertainty due the boundary conditions, which are often not known precisely. Unfortunately, this
Statement cannot always be carried over to the failure regime. Especially for concrete and masonry
structures there is still a lack of robust computational tools which can provide reliable predictions
of the structural Performance in the failure and the post-failure regime. Predictions of the load-
carrying capacity that exceed the experimental failure load by a factor two are not uncommon, and
at some instances a proper failure load cannot be obtained at all. Publications in which the failure
load of such structures is computed accurately mostly relate to postdictions rather than to
predictions.
In this contribution we shall describe predictive finite element analyses of a three-span, skewed-
slab bridge, which has been tested to failure afterwards by a team of the University of Cincinnati
and Wiss, Janney and Elstner [ 1]. The bridge was built in 1953 and was located on Route 222 in
Clermont County, Ohio. Visual inspection prior to the analyses and the testing revealed that the
concrete had experienced extensive deterioration and that there was corrosion in some bars.

In the non-linear finite element analyses concrete cracking and plastification of reinforcement and
concrete in compression have been taken into aecount. The sensitivity of the model to the various
material parameters has been investigated by means of parametric studies. In this way the effect of
the existing damage on the load-carrying capacity could be quantified. Furthermore, the possible
impact of the assumed boundary conditions on the predicted failure load was assessed.

2. DESCRIPTION OF THE BRIDGE AND THE FIELD TEST

The bridge which has been analysed and tested is a three-span, skewed-slab bridge and is shown in
Figure 1, see also Reference [1], which provides the necessary details on the lay-out of the
reinforcement as well. Inspection prior to the analyses and the testing revealed that severe damage
had oecurred, especially near the sides of the bridge [1], while the driving lanes were in a
reasonable condition. In the areas of visible damage of the concrete the reinforcing bars had
corroded severely. The visual inspection was hampered by the existing asphaltic overlay, which
was removed only shortly before the final destruetive testing.

To obtain a better judgement of the concrete properties cores were drilled at several places. The
concrete test specimens were then subjected to uniaxial compression tests which resulted in values
for the mass density p and for the uniaxial compressive strength fc which ranged from 2450 kg/m
to 2470 kg/m3 and from 49 MPa to 56 MPa respectively. The value for Young's modulus E
appeared to be around 33000 MPa. As will be detailed below, in the analyses the possible effect on
the remaining structural capacity of the observed poor quality of the concrete was modelled by
adopting artificially low values for Young's modulus and for the uniaxial compressive strength.
Properties for the reinforcing steel could be derived from uniaxial tension tests on rebars, see for
instance Figure 2 [1].

Prior to the final destruetion test modal tests were carried out in order to obtain data on the
boundary conditions which would have to be applied in the analyses. In contrast to the final
destruetion test the modal test were condueted with the asphaltic layer still in place, and resulted in
a lowest eigenfrequency of approximately 8.3 Hz [1].
 R. DE BORST - C. VAN DER VEEN - J. BLAAUWENDRAAD 187

7K
¦^ 9.76 m
^ 12.18 m
^ 9.76 m
y.

11.12m centreline

)k

6.42 m
1^ ^

Figure 1 Plan view and side view of the bridge.

800
G[MPa
600-

400

200-

0 0.01 0.02 0.03 e 0.04

Figure 2 Experimentally obtained stress-strain curve of a tension test on a reinforcing bar.

The actual destructive tests were carried by pulling down two concrete blocks of 0.6 m by 1.8 m,
which were placed on the bridge deck in order to distribute the forces exerted by servo-controlled
hydraulic actuators. On each block two of such actuators were placed. Rock anchors were attached
to the actuators to provide the reaction force that was needed to load the bridge. The total load at
which failure occurred was 3.24 MN. In the remainder of this article we shall always refer to the
load that was exerted on one of the blocks only, so that collapse occurs at a load level of 1.62 MN.
188 FINITE ELEMENT PREDICTION OF A DESTRUCTIVE FIELD TEST OF A BRIDGE

3. PREDICTIVE ANALYSES
3.1 Analytical yield line analyses

When the authors started their non-linear finite element analyses a rough estimate of the final
collapse load had been obtained by the investigators at the University of Cincinnati. They had
carried out yield line analyses assuming four different yield line patterns [1]. For the boundary
conditions that have been used in the Delft finite element analyses and for the material data used in
the reference finite element calculation (see below), the collapse load per loading block varied
from 1.5 MNto 2.67 MN [1].
3.2 Discretization and loading configuration
The finite element mesh that was adopted in the analyses which have been carried out with the
DIANA finite element package is shown in Figure 3. It consists of 144 eight-noded degenerated
plate/shell elements with a 2x2 Gauss Integration in the plane and a nine-point Simpson Integration
through the depth. Reinforcement was modelled using an embedded approach, that is the
interpolation functions of the concrete were used also for the reinforcement. The reinforcement
grid has its own Integration stations, which do not have to coincide with the layers of the plate/shell
element. The discretization of Figure 3 was considered sufficiently refined for the expected
bending-type failure. Analyses with different meshes should have been tried to verify this, but,
because of time restrictions - the analyses had to be completed before the actual bridge testing -
this has not been done.

The loading blocks have been modelled as two line loads which were each placed at the edges of
two elements, see Figure 3. Linear dependence relations have been supplied to ensure that all nodes
beneath a line load had the same vertical displacements.

31.70 m
K- ->l
7\

11. 12

^
¦

^ 6.42 m ^,
Figure 3 Finite element model for the Delft FE analysis of the bridge and position of the loads.

3.3 The numerical Solution procedure

All load-deflection curves that will be presented in the remainder of this article correspond to
converged Solutions in which a force norm of 10~2 was
satisfied after 4-5 equilibrium iterations
with a Modified Newton-Raphson scheme, in which the stiffness matrix was set up at the
beginning of each loading step. For the plasticity modeis this matrix was the tangent stiffness
matrix and for the cracking model the secant stiffness matrix was substituted. At the points were
 R. DE BORST - C. VAN DER VEEN - J. BLAAUWENDRAAD 189

the calculations have been terminated further analysis was always possible. Rather large Steps have
been taken, approximately flfty for the upper bound Solution and only five for the lower bound
Solution. The upper bound Solution calculations in which the step size was halved showed that for
this structure even the coarse loading Steps were fine enough.

3.4 Assessment ofthe boundary conditions

The piers have not been modelled in the predictive analyses, because they have no influence on the
final collapse load or the failure pattern. However, there is some influence on the load-deflection
pattern, not only because of the neglected axial stiffness of the piers, but also because the piers act
as rotational Springs on the bridge deck. From a hand calculation it appeared that the maximum
axial shortening of the piers would be approximately 0.1 mm, which is negligible. The rotational
stiffness of the piers was not taken into aecount either. This simplification will be justified below.

A most important issue when modelling an existing structure is the interaction ofthe structure with
the environment. At the abutments as well as at the piers we have the question whether the most
appropriate boundary condition would be a clamped support, a hinged support, or a roller support.
The question of clamped support vs hinged support can be partly resolved by carrying out
eigenvalue analyses and comparing the numerical results with the lowest eigenfrequency that
comes out of the modal test 8.3 Hz). In the finite element analyses with the mesh of Figure 3 the
mass density of the concrete
was taken as p 2370 kg/m3 and Young's modulus E and Poisson's
ratio were assumed as 24800 MPa and 0.2 respectively. The reduced value for Young's modulus
was adopted to model the observed deterioration of the concrete. In the first analyses all supports
were assumed to be hinged. When the influence of the asphaltic concrete cover was neglected an
eigenfrequency of 7.43 Hz was computed, whilst the slightly lower value of 6.76 Hz was found for
the analysis in which the asphaltic concrete cover was included. These values are much closer to
the experimentally determined eigenfrequency than the value of 22.69 Hz that was obtained for the
case with clamped ends and hinged supports at the piers. This indicates that (i) the supports at the
abutments are not clamped and (ii) neglecting the bending stiffness of the piers is reasonable.
However, the issue of hinged vs roller supports cannot be answered by modal analyses and will be
investigated below.
3.5 Model parameters for non-linear finite element analysis
In the non-linear analyses the following data have been used. For the reinforcement an elastic-
plastic model was utilized with a Young's modulus Es 200000 MPa, an initial yield strength o^
345 MPa and a hardening modulus h 7000 MPa, which is in agreement with the experimentally
supplied data. The inelastic behaviour of concrete in tension has been modelled by the multiple
fixed crack model of de Borst and Nauta [2] and Rots [5]. The shear retention factor ß was set
equal to 0.2 in all analyses. For the expected type of bending failure a Variation of ß hardly has any
impact on the results.
To aecount for the stiffness of the concrete between the smeared-out cracks a tension-stiffening
model was adopted with a linear softening branch and an ultimate strain at which the residual load-
carrying capacity is exhausted Eu l/2fsy/Es. The factor V2 has been introduced because previous
experience has shown that this generally leads to a better prediction of the structural behaviour [4]
and because a hand calculation for a rectangular cross section showed that taking Eu =fsyIEs would
result in a moment at which the steel Starts yielding, Msyi that is larger than the moment at which
collapse ultimately occurs (Mu).
The concrete stresses in biaxial compression were limited by a Drucker-Prager yield contour,
190 FINITE ELEMENT PREDICTION OF A DESTRUCTIVE FIELD TEST OF A BRIDGE

30

a[MPa]
20

10-

e«
¦/*

-0.001 0.001 0.002

Figure 4 Uniaxial response of concrete for the reference set of model parameters.

which was fitted such that the pure biaxial compressive strength fbc equals 1.16 times the uniaxial
compressive strength fc. Perfectly plastic behaviour was assumed thereafter, because any
introduction of softening in compression would result in an extreme mesh sensitivity, which cannot
yet be modelled properly. The uniaxial compressive strength fc itself was set equal to 27.5 MPa.
This is a relatively low value, and was adopted to aecount for observed damage in the concrete.
The tensile strength was initially set equal to ft 3.2 MPa, which is the value that had been
suggested by investigators ofthe University of Cincinnati [1], but later the value ft 1.8 MPa has
been used which follows from applying /,=0.75-(l +/c/20), which formula is used in the Dutch
Codes of Practice. In parameter studies it later appeared that the tensile strength affects the load-
deflection curve only in the first stages of cracking. The uniaxial stress-strain curve that results
from the adopted set of reference parameters is shown in Figure 4.
3.6 Numerical results
When carrying out non-linear finite element analyses it is sensible to first concentrate on the most
important causes of the non-linear structural behaviour, cf. the almost pedagogic treatise of Meyer
[3]. For 90% of all reinforced concrete structures cracking and yielding of the reinforcement are
the dominant non-linear phenomena which govern the structural response. Therefore, firstly
analyses were carried out in which concrete plasticity was not taken into aecount. These analyses
were carried out under arc-length control with a novel and very robust method for estimating the
load increment in a step [6]. The results are the upper and lower curves of Figure 5. In these figure
half the total load has been plotted against the deflection of the outermost loading block. The upper
curve was obtained under the assumption that all supports (at the abutments and at the piers) were
hinges, while the lowermost curve was obtained assuming that all supports were rollers except for
one of the abutments.

We observe that this Variation in boundary conditions has a tremendous impact on the structural
response of the bridge. This phenomenon can be explained as follows. In the latter case (the lower-
bound Solution) cracks due to the bending moments penetrate deep into depth of the slab which
causes large horizontal strains in the midplane of the slab. As a consequence an axial elongation of
the midplane occurs. On the other hand, this elongation is entirely prevented in case of hinges at all
supports. This means that additional in-plane forces prestress the slab. These membrane forces
effectively prevent collapse of the bridge, as an almost linearly ascending load-displacement curve
 R. DE BORST - C. VAN DER VEEN - J. BLAAUWENDRAAD 191

4-n
F/2 [MN] hinges at
abutments and piers
3-

2- \Jiinges at abutments
^

1-

0
\ roller supports

1 1
i 1

0.025 0.05 0.075 0.1

w m
[ ]

Figure 5 Influence of boundary conditions on load-bearing capacity.

was computed up to a displacement of 0.2 m, at which point the calculations were stopped. At this
point a large part of the reinforcement was yielding. Because no real collapse load could be
identified at this point, which is far beyond the failure loads predicted by yield line Solutions, the
role of the membrane forces seems unrealistically high for these boundary conditions.
To further illustrate the important role of the membrane forces an additional analysis was
undertaken in which the piers were roller-supported, but where both abutments were modelled as
hinges. The membrane actions that develop are now distributed over all three spans and, as a result,
the load-displacement curve nicely falls between both extremes. At a deflection of 0.2 m
significant yielding of the reinforcement was again observed, but there were no signs of impending
collapse.
The Solutions with hinges at all supports and with hinges at only one abutment can be considered
as upper and lower bound Solutions respectively. Because the precise boundary conditions were
unknown a more accurate prediction of the collapse load could only be obtained by improving the
upper and lower bounds. To this end first the effect of a Variation of the hardening modulus of the
reinforcement was considered. In particular it was thought that the almost linear rising branch of
the analysis with the hinged supports might be caused by the hardening of the reinforcement after
first yielding. Therefore analyses were carried out with the same data, but with an ideally-plastic
behaviour of the reinforcement. Surprisingly, for both types of boundary conditions the differences
were well within 1%.

Next, it was investigated how the type of loading affected the computed load-deflection response.
In the actual test the loading was first carried out under load control and when the collapse load
was approached a switch was made to displacement control. This could not be simulated in the
finite element analyses. As stated most analyses were carried out under arc-length control with
equal loads on both loading blocks. Although the precise form of loading should not influence the
collapse load the deflections can be affected by a different control scheme. Therefore an analysis
was also made under pure displacement control, in which both loading blocks were pushed down
by the same amount in each loading step. In this loading arrangement there is no relative tilting of
the outermost loading block compared to the other block, which results in a somewhat stiffer
response. However, the differences in displacements remained within 10-15% for a given load
level.
192 FINITE ELEMENT PREDICTION OF A DESTRUCTIVE FIELD TEST OF A BRIDGE

6-^

F/2 [ MN ]
fc-*°° ^-
4- -^jT= 40 MPa
\/c 27.5 MPa

2-

0 I 1 1

0.05 0.1 0.15 0.2

w[ m]
Figure 6 Influence of concrete plasticity on failure load for the case that there are hinges assumed
at the abutments and roller supports at all piers.

The most important parameter that influences the upper bound Solution is the compressive strength
fc. Figure 6 shows the effect of varying fc on the load-deflection curves of the upper bound
Solution. The calculation in which the compressive stresses were not limited, gives the stiffest
structural response, but computations with fc 40 MPa and/c 27.5 MPa give a markedly softer
response. In fact, we consider the calculation with fc 27.5 MPa as the best upper bound Solution,
and the structural response should be between this Solution and the lower bound Solution of Figure
5. It is finally remarked that the lower-bound Solution is not affected by adopting a plasticity model
to bound the compressive regime, since the maximum compressive stresses remain well below the
uniaxial compressive strength fc.

4. DISCUSSION OF THE RESULTS AND FAILURE MECHANISM

The authors expected that the lower bound Solution would be closer to the experiment than the
upper bound Solution, since it was believed that the abutment was not sufficiently rigid to sustain
the large horizontal forces without undergoing horizontal displacements. Accordingly, the most
realistic assumption for the conditions at the abutments would be roller supports rather than hinges.
This expectation was confirmed when the testing had been carried out, Figure 7.

Although the numerically predicted lower-bound Solution for the failure load and the
experimentally obtained collapse load agree extremely well, this is not completely the case for the
failure mechanism. From observations on the experimental failure pattern it seems that first a pure
bending type failure has occurred, but that after significant deformations the shear capacity was
exhausted. This point, that is when the capacity to sustain external loads Starts to decrease, is
marked by the onset of the softening branch in Figure 7. This hypothesis is strengthened by the
following observations. Firstly, the used plate/shell elements can only predict accurately bending
type failures. Yet, it predicted the experimental failure load very well. Secondly, not only did our
(lower-bound) numerical Solution, in which membrane effects played no role, match the
experimental failure load, also the yield line Solutions obtained at the University of Cincinnati [1]
fall in the same ränge, indicating that at the peak of the experimental load-deflection curve only
bending effects have played a role of importance. Exhaustion of the shear capacity and subsequent
punching only comes into play after significant yielding of the reinforcement.
 R. DE BORST - C. VAN DER VEEN - J. BLAAUWENDRAAD 193

F/2 [ MN ] numerical lower-bound prediction

1.5-I

experiment

computation with
smaller load Steps
0.5

0 0.025 0.05 0.075 0.1

w[ m]
Figure 7 Numerically obtained lower-bound Solution and the experimental failure load.

In the numerical simulations yielding started at a load level of F12= 1.05 MN at the edge of the
outermost loading block near the abutment. The extent of the area in which the bottom
reinforcement was yielding at a load level of F12= 1.40MN is shown in Figure 8. The maximum
plastic strains at this point were approximately 0.27%. A recalculation, conducted with smaller
load Steps, resulted in a somewhat softer response, but the computed failure load was hardly
affected, Figure 7.

Figure 8 Spread of zone in which the bottom reinforcement is yielding at F12= 1.40 MN.
The lighter shaded area has experienced less plastic flow.

5. CONCLUDING REMARKS
The numerical simulations have shown that the uncertainty in the boundary conditions of the
bridge was more important than the fact that the material parameters could not be determined
exactly. Nevertheless, reasonable predictions for upper and lower bounds of the failure load could
be obtained by a proper combination of sensitivity studies on the influence of the boundary
conditions and the material data. The actual field test resulted in a collapse load that was
marginally above the predicted numerical lower bound Solution.

ACKNOWLEDGEMENTS
Partial financial support from the Commission of the European Communities through the Brite-
Euram programme (Project BE-3275) to the first author is gratefully acknowledged.
194 FINITE ELEMENT PREDICTION OF A DESTRUCTIVE FIELD TEST OF A BRIDGE

REFERENCES
1. AKTAN, A.E., MILLER, R. and SHAHROOZ, B., Destructive field testing of a r/c slab bridge
and associated analytical correlation studies, Research Status Report, University of
Cincinnati, Cincinnati, 1991.
2. BORST, R. de and NAUTA, P., Non-orthogonal cracks in a smeared finite element model,
Engineering Computations 2 ,1985, pp. 35-46.
3. MEYER, C, Analysis of underwater tunner for internal gas explosion, in Computational
Mechanics of Concrete Structures, IABSE Reports 54,1987, pp. 473-486.
4. MlER, J.G.M. van, Examples of non-linear analysis of reinforced concrete structures with
DIANA, Heron 32 No. 3, 1987.
5. ROTS, J.G., Computational modeling of concrete fracture, Dissertation, Delft University of
Technology, Delft, 1988.
6. SCHELLEKENS, J.C.J., FEENSTRA, P.H. and BORST, R. de, A self-adaptive load estimator
based on strain energy, in Computational Plasticity: Fundamentals and Applications
Pineridge Press, Swansea, 1992, pp. 187-198.
 195

SELECTED PAPERS
196
J%

Analytical Modelling for Fatigue Assessm

Modele analytique d'eValuation de la fa
Analytisches Modell zum Ermüdungs

M.S.G. CULLIMORE Peter J. MASON

Visiting Research Fellow Director
University of Bristol Howard Humphrey
Bristol, UK Dorking, UK

f : ;

M. Stuart Cullimore, born 1920, Peter Mason, born

graduated in civil engineering from the Bridge Departme
Bristol University in 1940. Reader Humphreys & Partne
in Structural Engineering until engineers, in 1967
1985. Reserach latterly on fatigue from Imperial Colleg
198 FATIGUE ASSESSMENT OF THE C

1. INTRODUCTION

The Clifton Suspension Bridge

was designed originally by the
eminent Victorian engineer
Isambard Kingdom Brunei. The
bridge was completed in 1864,
after his death, with a number
of important modifications to
his design [1]. The
spectacular setting of the
bridge spanning the Avon Gorge feS
makes it an important tourist
attraction and focus of civic sS
pride (Fig 1). But it is also
an important link in the
traffic System of Bristol
carrying nearly 4 million
vehicles per year, although
there is a gross weight limit
of 40 kN.
F

Structurally the bridge is awrought iron

suspended structure carrying an asphalt s
are three chains on each side of the roadw
shown in Fig 1. They are made up of 175mm
to each end. E
 M.S.G. CULLIMORE - P.J. MASON - J.W. SMITH 199

A number of studies of the structural capacity of the bridge have been

undertaken this Century, some of them resulting in remedial or strengthening
works [1,2]. The collapse of the Point Pleasant Bridge over the Ohio river in
1967, resulting from corrosion-fatigue in an eye bar, prompted an extensive
fatigue appraisal of the Clifton Bridge [3]. It was concluded that there was
an adequate margin of safety against fatigue failure at that time. However,
traffic loading was steadily increasing and there was concern over progressive
deterioration of the riveted joints of the parapet girder and other signs of
wear or damage. It was decided that a global analysis of the structure should
be carried out, using modern analytical methods, so that the effects of a
ränge of load cases could be studied.
2. FINITE ELEMENT MODEL

2.1 Modelling assumptions and analytical procedure

Suspension bridge behaviour under load is geometrically non-linear. This is
because the cable or chain adapts its shape when a concentrated load is
applied at a particular point on the structure. In most Suspension bridges,
as at Clifton, the longitudinal girder provides some stiffening and
effectively distributes these deformations over part of the structure. For
this reason it was decided to use a finite element program which had
geometrically non-linear Solution capabilities.
The finite
element model was designed to represent the effects of vertical
loading the structure. For this purpose a two-dimensional model was
on
considered to be adequate. Dead load and traffic load in both lanes is
symmetrical about the longitudinal axis of the bridge and therefore only one
half of the bridge needed to be modelled, i.e. one set of three chains
supporting the main girder and parapet girder. Eccentric traffic loads, in
the form of single vehicles or traffic in one lane only, were dealt with by
means of a separate torsional analysis.

A schematic diagram of the node and element geometry is shown in Fig 3a. From
the left anchorage to the tower the three chains were represented by a single
chain of beam elements. The final link was connected by a pin Joint to the
saddle elements. The saddle nodes were all effectively constrained to move
horizontally as a single unit, simulating the roller bearing that exists at
the top of each tower. The three chains of the main span were represented by
beam elements of the same length as each eye bar link. Thus the correct
sequence of connection to the suspender rods could be modelled as shown. Each
link of ten, eleven or twelve bars, was modelled by a Single element of
equivalent area. It has been observed that the chain links behave as if
they
are rigidly connected to each other over the main part of the span. The pins
work freely only at the tower saddles. The main girder was modelled by beam
elements pin connected to the vertical rods as shown. There is no vertical or
horizontal restraint to movement of the main girders of the bridge. However,
in order to avoid computational instability, a soft horizontal spring
restraint was connected to the middle node of the deck.
The behaviour of the cross girders and parapet girder were modelled by
suspending longitudinal beam elements from the main girder elements by means
of vertical linkages. The stiffness of the linkage elements was determined
from the stiffness of the cross girder in shear between points of connection
of the main and parapet girders as shown in Fig 3b. In order to avoid the
problem of horizontal instability itwas sufficient to introduce
coupling between main and parapet girders at mid-span.
horizontal
200 FATIGUE ASSESSMENT OF THE CLIFTON SUSPENSION BRIDGE

river
saddle chains
land
chains
suspender
rods

main girder fc
hnkage elements——*:
parapet girder

(a) Schematic diagram of node and element geometry

Parapet girder
T Main girder T

1 >s\AAfewW
)p tp
Stiffness of hnkage element, k P/A
(b) Modelling of cross—girder stiffness

Fig 3 Finite Element Model

2.2 Analysis of eccentric loading

It was mentioned earlier that eccentric loads, such as single vehicles or
traffic in one lane only, could be dealt with by introducing a torsional
component. This is illustrated in Fig 4 where it
can be seen that the loading
can be resolved into Symmetrie and torsional components at the line of the
chains. If the cross girder is rigid, then because of its rotation about the
centre line, it
forces the parapet girder to deflect more than the main girder
in the torsional case. Hence the effective stiffness of the parapet girder,
if it is transposed to the same plane as the main girder and the chains,
becomes:

leffPa Ip(B/b)- (1)

\
I I

V
P P

11 J ^zz
11
—-?
i!
N^
1

£ - "
[-1
y
A.\- t—

Symmetrie component Ps P

Torsional component PA P e/b

Fig 4 Analysis of eccentric loading

 M.S.G. CULLIMORE - P.J. MASON - J.W. SMITH 201

In this way it
would be possible to analyse the problem as two separate load
cases and add the results, provided that
of the chains was linear with changes in live load applied at a point. In
it
could be assumed that deflection

practice it
was possible to combine the Symmetrie and torsional components
into a Single eccentric load case. This was done by evaluating the effective
stiffness of the parapet girder when transposed to the line of the chains with
both Symmetrie and torsional component loads present. The formula is as
follows:

Ieffp Ip [ ](1 + f) ; where f Pa/Ps (2)

B2 + fb2

Further refinements were included which took aecount of cross girder

flexibility in the torsional case, but are outside the scope of this paper.
2.3 Load tests and comparison with analytical model
In order to confirm the analytical model, loading tests were carried out on
the bridge at night. The weight limit on the bridge is 40 kN. This
represents vehicles such as ambulances, loaded vans and pick-up trucks. Two
loading cases were identified as follows:
(a) Dual vehicle Symmetrie loading
The maximum concentrated load occurs when two 40 kN vehicles, travelling in
opposite directions, pass each other on the span. Assuming a load
distribution of 15 kN at front axles and 25 kN at rear axles the load case iE
as shown in Fig 5. Although the front axles are in opposite lanes of the
carriageway, and would produce an anti-Symmetrie torsional component of
loading, it
was assumed that this would be a small localised effect and that
the load could be treated as Symmetrie as shown.

15 kN 25 kN
15 kN 50 kN 15 kN
motion
o
25 kN 15 kN
o
h
b ° K
b °—%
motion b b
m »-^ m

b 3.66 m

Fig 5 Dual vehicle load case (Symmetrie)

(b) Single vehicle eccentric loading

The bridge is torsionally flexible and therefore it was considered important
to study the effects of a Single 40 kN vehicle travelling in one lane of the
carriageway, thereby applying an eccentric loading to the structure. The
eccentricity of the vehicle in the analysis was taken as 1.0 m from the centre
line of the carriageway.
For the loading tests on the bridge, two pick-up trucks were hired and loaded
with boxes of nails to provide the appropriate distribution. Strain gauges
were fixed to the top and bottom flanges of one parapet girder and both main
girders at 1/4 span. The Signals were logged by a Computer data acquisition
system while the vehicles were crossing the span in the loading configurations
described above.
202 FATIGUE ASSESSMENT OF THE CLIFTON SUSPENSION BRIDGE

The results of a typical analysis are shown in Fig 6 and the results of a
single vehicle load test run are shown in Fig 7. Similarity in the shapes of
the curves is evident. The test run results are effectively influence lines
for strain at 1/4 span when the vehicle passes over the span. The analysis
represents the distribution of deflection and bending moments when the vehicle
is stationary at the 1/4 point. But since the wheel base of the vehicle is
very short compared with the span, and there is evidence of linear ity under
live loads, it may be considered that the analysis results approximate to
influence lines. It may also be noted that the bending in the main girder is
sharper than that of the parapet girder directly under the load. This is
because flexibility of the cross girder results in transfer of bending from
the main to parapet girder to be distributed longitudinal ly to some extent. A
further point to note about the results is that the parapet girder bending
moment is of the same order as that of the main girder. is not known if
It aecount
this structural action of the parapet girder was taken into in the
original design.

^rniTTTTTTTTTIIlllllllllirnTTTTTTTmTTmw
¦^m[' jp*~ Main Girder Bending Moment
64.0
\y/
Max BM kNm

^
^TTTTimm rrrmrrr^
Y Parapet Girder Bending Moment
Max BM - 70.6 kNm

MJ/

Fig 6 Finite element analysis of eccentric 40 kN load at 1/4 span

100. -r 100.
t
BO.O BO.O

60.0 BO.O •¦

40.0 40.0 ¦¦
20.0 ¦¦ 20.0
-.000 -.000 -•

-20.0 -20.0
-40.0 •¦ -40.0 ¦•
60.0 ¦• 60.0 ••

-BO.O ¦• BO.O

12.0 24.1 36.1 48.1 60.2 72.2 84.2 96.3 12.0 24.1 3 48.1 60.2 72.2 B4.2
tisne/seconds time/seconda

(a) Main Girder (bottom) (b) Parapet Girder (bottom)

Fig 7 Strains at 1/4 span under action of 40 kN vehicle eccentric load

In order to obtain quantitative comparison between the experimental results
and the analysis it was necessary to convert the observed strains at top and
bottom of the girders to equivalent bending moments. This was done using the
measured sectional properties of the girders and a value of E for wrought iron
of 192 GN/m2. The results are compared in Table 1 below.
#% M.S.G. CULLIMORE - P.J. MASON - J.W. SMITH 203

Table 1 Bending Moments at 1/4 span in kNm (analysis in parentheses)

LOAD POSITION MAIN GIRDER PARAPET GIRDER TOTAL MOMENT

8 ton 1/2 span 18.2 (19.7) 13.1 (24.6) 38.4 (44.3)

symm. 3/8 " 3.3 (4.0) 5.3 (1.8) 11.6 (5.8)
1/4 " -89.1 (-111.7) 64.1 (-86.3) -170.6 (-198.0)
1/8 " -2.1 (-8.5) -7.0 (-10.6) -18.8 (-19.1)
4 ton 1/4 " -56.4 (-64.0) -43.7 (-70.6) -111.2 (-134.6)
ecc.
In Table 1 the experimental moments were evaluated from the strain gauge data
so as to provide a comparison with the analysis. The analytical results are
generally greater than the experimental results. This is probably a result of
the influence of the deck which acts like an additional flange to the main
girder. This was difficult to include in the analytical model although some
allowance was made by modifying the girder section properties to simulate the
shift of the neutral axis.
The 'Total Moment' in Table 1 is the sum of the girder moments in the case of
the analysis. However, the longitudinal forces introduced by the presence of
the deck could be evaluated from the strain gauge data together with the
girder moments. Hence, the 'Total Moment' of the experimental results is
always greater than the direct sum of the girder moments.
3. CONTINUOUS MONITORING OF GIRDER STRAINS UNDER NORMAL TRAFFIC

3.1 Installation and use of "Stress Analvser"

The same strain gauge locations, as used in the vehicle loading tests,
were
monitored continuously under normal traffic for several one week periods. The
equipment for doing this, called a "Stress Analyser" [4], was capable of
amplifying the signal from the gauges, detecting peaks and troughs of the
fluctuating signal, and performing a "rainflow" count in real time.
"Rainflow" counting is an accepted method for interpreting a varying amplitude
signal in terms of an equivalent number of simple cycles of different
amplitudes. The fatigue damaging potential of the signal may then be assessed
by summing the fatigue damage contributions of all the simple cycles.

The data are provided in the form of numbers of cycles of different strain
ranges. The mean strain was not recorded because, although has an effect it
it is generally accepted that, for materials such as wrought iron with many
defects, strain ränge is the dominant factor affeeting fatigue life.
3.2 Prediction of strain ränge cycle count and comparison with observations
The results of the global analysis were used to make a prediction of the
strain ränge cycle count under normal traffic. This was achieved by looking
at the output from the analysis of the bridge under a 40 kN eccentric load as
shown in Fig 6. It has already been said that this figure approximates to an
influence line and therefore the ränge of bending moment at the 1/4 span when
a 40 kN vehicle crosses the bridge may be deduced from the maximum and minimum
of this figure. It was further assumed that the bending moments at this point
were linear with load within the ränge of live loading. Hence, it was
possible to evaluate strain ranges oecurring under the passage of a ränge of
vehicle weights as they cross the bridge.
A Classification count was carried out on the bridge, grouping weekday traffic
into seven weight classes. Cars were relatively easy to classify according to
weight, but estimates had to be made for larger vehicles such as pick-up
204 FATIGUE ASSESSMENT OF THE CLIFTON SUSPENSION BRIDGE

trucks, vans and ambulances. A count was also made of the number of times
vehicles travelling in opposite directions were applying load to a particular
cross girder simultaneously. The count is set out in Table 2 below.
Table 2 Number of loadings of a cross girder bv vehicles of different weight
Vehicle weight 8 10 14 20 25 30 40 (kN)

Left lane 264 633 252 26 22 11

Right lane 343 819 243 41 22 11
Both 21 184 56 10 1

The weights were converted into strains at the top of the main girder and a
table of the number of loading cycles within strain ränge bands was compiled.
Data on the number of vehicle crossings was available from the toll records
and amounted to 72,000 vehicles per week during the period of the study. The
number of cycles for the short count (four hours in total) was then factored
up to give the number of cycles that would occur at the same rate during one
week of normal traffic. The predicted cycle count was compared with the data
obtained using the "Stress Analyser" and is shown in Table 3.
Table 3 Strain ränge cvcle count: Predicted v. Stress Analvser
Number of Loading Cycles
Range Predicted Stress Analyser (avg
(x 10'-6> Short count Seven days of 3 seven day periods)

0 - 10 343 8,332 124,022

10 - 20 1,326 32,210 20,713
20 - 30 695 16,882 12,403
30 - 40 469 11,393 6,989
40 - 50 82 1,992 3,328
50 - 60 25 607 1,356
60 - 70 21 510 519
70 - 80 1 24 174
80 - 90 2 48 59
90 - 100 18
100 - 110 7
110 - 120 3
120 - 130 1

Considering the difficulties of assessing the loads from the Visual

Classification count the correlation is remarkably good. The large number of
cycles occurring in the smallest strain ränge may be the result of small
vibrations and electronic noise. A further comparison can be made by
evaluating the fatigue damage done by each loading cycle. This can be
achieved by assuming a power law for fatigue life with an index of 3, together
with Miner's law of cumulative damage. It is then possible to calculate the
equivalent strain ränge per vehicle, if
applied repetitively, that would yield
the same fatigue damage as the actual variable loads. This quantity (ESRV) is
given by
ESRV (niSim/72,000)1/m (3)
where n^ is the number of cycles of strain ränge S^ and m is the index of the
power law. m=3 is a reliable mean value for fatigue of wrought iron.
 M.S.G. CULLIMORE - P.J. MASON - J.W. SMITH 205

Using the data in Table 3 the following comparison may be made:

ESRV predicted 26.4 x 10-6

ESRV experiment 26.8 x 10
This result confirms that the method of prediction provides a very accurate
measure of fatigue damage.

4. FATIGUE ASSESSMENT OF MAJOR COMPONENTS

4.1 Saddle link

Rotation of the chain links attached to the saddle bearings at the tops of the
towers were found to produce significant variations in principal stress while
vehicles crossed the bridge. In an earlier study this was found to be the
most significant fatigue loading on the bridge [3]. The results of the load
tests carried out on the bridge at that time were found to compare favourably
with the global analysis. Hence the conclusion of the earlier assessment,
that there was sufficient factor of safety against fatigue or fracture, was
confirmed.
4.2 Main girders
The strains observed in the main girder under normal traffic (Table 3) were
converted to stress cycles. These were compared with S-N curves for riveted
girders [5,6]. Assuming traffic totalling 4 million vehicles per year the
fatigue life of the main girder was calculated to be 468 years.
4.3 Parapet girders
For fatigue loading the eritieal location is the spliced Joint in the top
flange of the parapet girder. Strains were obtained using the same procedure
as for the main girders and were converted to stress ranges. The joints have
been progressively deteriorating in recent years and a new friction grip
bolted assembly has been designed as a replacement. Using the current UK code
for fatigue assessment, the life of the Joint was estimated to be 197 years.
REFERENCES

1. MITCHELL-BAKER D and CULLIMORE M S G. Operation and maintenance of the

Clifton Suspension Bridge. Proc Inst Civil Engrs, Part 1, n 84, 291-308,
April 1988.
2. FLINT A R and PUGSLEY A G. Some experiments on Clifton Suspension Bridge.
Correlation between calculated and observed stresses and displacements in
structures. Inst Civil Engrs, Preliminary vol, 124-134, London 1955.
3. CULLIMORE M S G and MASON P J., Fatigue and fracture investigation carried
out on the Clifton Suspension Bridge. Proc Inst Civil Engrs, Part 1, n 84,
309-329, April 1988.
4. WASTLING M A and SMITH J W., An instrument for detecting arbitrary peaks
and troughs of a fluetuating stress signal. Strain, Brit Soc Strain
Measurement, 127-131, August 1987.

5. BRUHWILER E, SMITH I F C and HIRT M A. Fatigue and fracture of riveted

bridge members. J.Struct.Engng., ASCE, vll6, nl, 198-214, 1990.
6. FISHER J W, YEN B T and WANG D., Fatigue strength of riveted bridge
members. J.Struct.Engng., ASCE, vll6, nll, 2968-2981, 1990.
206
J%

Long-term Prediction of Behavi

Prediction du comportement ä lo
Voraussage des langfristigen Verh

Eiichi WATANABE Masahiro KAMEI

Professor Chief Engineer
Kyoto University Osaka Municipial
Kyoto, Japan Osaka, Japan

\
^^^tm^" w^
Eiichi Watanabe, born 1942, B.S., Born 1950, B.S., M
M.S. & Dr. Eng. from Kyoto City University.
University, M.S. & Ph.D. from Iowa
State University, USA.

Osamu NAKADE
Bridge Engineer
208 LONG-TERM PREDICTION OF BEHAVIOUR OF CABLE-STAYED BRIDGES

1. INTRODUCTION
Presented herein is a method to predict the long-term change in the cable forces and cables slip-
out from their sockets in cable-stayed bridges. Based on long-term tension tests on full-scale cables
of 5m length, a very simple analytical model is proposed and an effort is made to determine the visco-
elastic constants of the cables and sockets, taking into aecount the seale effect of the length of the
cables, by extrapolating the results of the measurements carried out in cables with limited length to
actual cables with arbitrary length. In addition, visco-elastic F.E.M. analysis using the experimental
results was carried out to predict the long-term behavior of several existing bridges and these results
were compared to the measured values at the site.
Due to the fact that the erection of bridges is usually completed within a period of one year or one
year and a half at the site, the visco-elastic constants of the cables and sockets were determined
emphasizing first, the initial relatively short period of the erection stages and, secondly, focusing on
the control of the cable forces for the much longer period of service life.

2. LONG-TERM TEST OF PROTOTYPE CABLES

2.1 Experimental Method
When investigating the time-dependent behavior of materials, there are two types of tests,
namely, creep and relaxation tests. The former being carried out under constant loading with
increasing deformation, and the latter under constant deformation with decreasing stress. The type of
test carried out in the present study is a combination of both types [1].

The measurement System is presented in Fig.l, where the load is measured by a load cell and the
relative displacements between the cable and the steel frame, by displacement transducers. To
investigate the visco-elastic characteristics ofthe cables due to the difference in cable strength, two
types of cables (Specimen types 1 and 2) were tested. In addition, four different combinations of
cables and sockets (Specimen types 3 to 6) were tested, in a total of 6 specimens. Table 1 shows the
specimen dimensions and characteristics.
Strain gage
Loadcell |
„H _*
Displacement |
\JF*^~ transducer """-^-ü-fl
~
/ Shim plate

Specimen cable A
v
\
Socket Socket
Strain gage

Fig.l Measurement System

Table 1 Dimensions and Characteristics ofthe Test Specimen

Specimen Name Type 1 Type 2 Type 3 Type 4 Type 5 Type 6
Cable Length (m) 5 9 5 9 5 4 5 4 5 4 5 4
Diameter of Wire (mm) 5 0 5 1 7 0 7 0 7 0 7 0
Number of Wires 127 127 19 19 19 19

Cable Type PWS shghtly PWS shghtly

PWS PWS PWS PWS
twisted twisted
Anchorage Length(cm) 44 0 44 0 16 3 16 3 30 7 30 7

Anchorage Type Zinc-poured Zinc-poured Zinc-poured

Zinc-poured HiAm HiAm
(Incl 2% Cu) (Incl 2% Cu)
Cross Sectional Area (cm2) 24 94 25 94 7 31 7 31 7 31 7 31
Breaking Force (kN) 4400 3910 1137 1137 1137 1137
Initial Cable Force (kN) 1400 1310 380 377 368 371
 E. WATANABE - M. KAMEI - Y. ISHIHARA - 0. NAKADE - L.H. ICHINOSE 209

2.2 Test Results

Fig.2 shows the time Variation of the tensile force in the cables. As it can be observed, the forces
in specimen types 3 and 4 tend to stabilize in a relatively short time (about 20 days), whereas in the
other cables, continue to decrease even after one year's measurement.
1450 390 ¦ 390
r- 1 1

1400 —Typel Type 3 Type 5

--Type 2 380 Type 4 380

i
--Type 6
§ 1350
»—

370.
Q
<; 1300 <
3 1250
J 360 - 360

..«««•JA*..,
1200 —1 L... 1
350
180 360 540 720 900 0 100 200 300 400 100 200 300
TIME (day) TIME (day) TIME (day)

Fig.2 Time Variation of Cable Tensile Force

Although the tested cables and sockets were the same as the ones used in actual bridges, the
cable length of the specimens differed from the actual cables, thus, it was decided herein to consider
the cables and sockets separately. Fig. 3 illustrates the time-dependent strain of each cable type and
Fig.4, the slip-out behavior of the different types of sockets.
1 1
v —i ¦¦ i ¦ "i l T~
<r
1

o Type 3 © Type 5
- - - 80 -
Type4 Type 6

Typel
60 -
es
Type 2
* 40 40 -

- & 20 V 20
r-^™^"'"'""v~ w
U 200 400
1

600
1

800 100
i
200
i
300
1

u 0 i
100
i

200
i

300
TIME (day) TIME (day) TIME (day)

Fig.3 Time-dependent Behavior of Cables

20 1 1 1 1 20 I '1 1 20 1
' 1

^ 16 -
^ 16 - —
Type 3
- -Type 4
- "? 16 - Type 5
- - -Type 6
-
^H 12 i12 ii2
D D
O 08 ¦ em O 08 " O 08 - -
/ CU

^ 04 Typel
---Type2 -
^—
i i "T*—'—
200 400 600 800 1000 100 200 300 100 200 300
TIME (day) TIME (day) TIME (day)

Fig.4 Time-dependent Behavior of Sockets

Specimen types 1 and 2 presented large values for the amount of slip-out from the sockets,
compared to the values of cable creep. Specimen type 3 and 4, 5 and 6 have respectively the same
sockets. In the formers, the forces, as well as the amount of slip-out stabilized in relatively short time
(20 days); whereas in the latters, the amount of slip-out continued to increase.

Cable material of the specimen types 3 and 5, 4 and 6, being respectively the same, presented
similar values for the final creep. However, time Variation between cables of the same type were
different, suggesting the influence of their sockets.
210 LONG-TERM PREDICTION OF BEHAVIOUR OF CABLE-STAYED BRIDGES

3. MODEL TO EVALUATE TIME-DEPENDENT CONSTANTS

3.1 Mechanical Model
Due to its simplicity, the analytical model adopted is the three-element model shown in Fig.5 [2].
The total strain of the model e, can be expressed as the sum of the elastic (ee) and viscous (£v) strain:

e e0 + e9 (1) Ei
For the total stress a, the strain-stress
can be expressed as follows:
relationship Ei ¦A\\-i
HArV- V
(2)
-E-
a=av E2ev + ri£v
(3) Fig.5 Three-element Visco-elastic Model
where Ei, Eiare the elastic coefficients and rj the viscosity coefficient of the three-element model.
Differentiating Eq.l and Eq.2 in relation to time t and introducing them in Eq.3, it yields the
following equation, after some arrangement.

T+Ei±JkG=Ei{e+*ke)
(4)

3.2 Evaluation of Time-dependency

The visco-elastic constants of the model were evaluated according to the three different methods
described bellow.

3.2.1 Method 1

Focusing on the viscous part of the model in Fig.5 (Eq.3) the following approach curve for the
strain due to the viscosity was assumed.

c, e,(l-«-*) (5)
The coefficient X can be obtained through the least Square method. The results of the evaluation
for one of the cables is shown in Fig.6, with the corresponding slip-out from the sockets shown in
Fig.7 and the viscosity constants thus evaluated are presented in Table 2.
This method is effective for cables in which the phenomena of creep and slip-out from the
sockets stabilize in a short time, however, in cases in which the time dependent curves are not steep
and long-term Variation is observed, the curves tend to diverge from the predicted values.

10
<r
1
r r_,

.=i 40- Type 6

O 08 " Type 6 -
- -Predicted
- - -Predicted

« 20-
i°6
O
- -

04
Cu

3 02 ¦/v -
// i i i
100 200 300
TOME (day) TIME (day)

Fig.6 Predicted Time-dependent Fig.7 Predicted Time-dependent

Behavior of Cables (Method 1) Behavior of Sockets (Method 1)
 E. WATANABE - M. KAMEI - Y. ISHIHARA - 0. NAKADE - L.H. ICHINOSE 211

Table 2 Evaluated Visco-elastic Constants (Method 1)

Cable Socket

Specimen l/X El E2 1A *2 il
(day) (GPa) (GPa) (year GPa) (day) (MN/m) (MN/m) (year MN/m)

Typel 2.97 217 6272 51.0 80.0 381 704 154

Type 2 0.26 196 30184 21.5 113.6 543 948 295
type 3 1.168 202 19306 61.8 0.153 210 4192 1.76
Type 4 1.043 201 20690 59.1 1.188 228 2454 7.99
Type 5 25.5 191 20482 1428.4 75.0 785 769 158
Type 6 23.0 198 24304 1533 1 50.3 582 769 106

3.2.2 Method 2
The prediction of the time dependent behavior of bridges during their construction stages
requires more accurate values for the initial steep part of the time Variation curves. Thus, in the
second method, the equations used in Method 1 were applied to a relatively short time interval
corresponding to the average interval of time between the prestress of one of the cables and the
prestress of the cable of the succeeding stage. This method converges for the initial part of the time
Variation curves and leads to reliable values for the initial 40 days. Fig.8 shows one of the curves
evaluated by this method with the respective experimental curve. Table 3 presents the visco-elastic
constants evaluated for specimen types 1,2,5 and 6. " ¦ 1 T 1—

Table 3 Evaluated Visco-elastic Constants for

I" " Type 6 -
Predicted
Sockets (Method 2)
H06
D
l/X Ki K2 O 04 - -

llll
Specimen
(day) (MN/m) (MN/m) (year MN/m) ä
d 02
Type 1 0.921 381 1491 3.76
Type 2 1.043 543 2154 6.16 0 10 20 30 40 50
Type 5 2.48 785 2030 13.8 TIME (day)
Type 6 1.319 582 1910 6.90 Fig.8 Predicted Time-dependent
Behavior of Sockets (Method 2)
3.2.3 Method 3

For the maintenance of the bridge during its service life, a long-term prediction is necessary and
the strain Variation for the time period succeeding the one considered in Method 2, shall be assumed
as it follows.
ev ew(l- ae-»)9 where (6)
X E2 / 7]
Considering Ev as a determined parameter, oc and X can be determined by the least Square
method. Fig.9 illustrates the results of the analysis for one of the specimens and Table 4 shows the
evaluated visco-elastic constants for specimen types 1,2, 5 and 6.
1 1 —,_.,.

Type 6 _
Table 4 Evaluated Visco-elastic Constants for Sockets (Method 3) Predicted

Specimen a
1A
(day)
Ki
(MN/m)
K2
(MN/m)
Tl

(year MN/m)
H06
D
O 04
CU
/
Type 1 0.450 351 344 684 658
3 02 «r^
Type 2 0.541 413 491 866 980 00 i i i

0 100 200 300 400

Type 5 0.748 607 505 479 797 TIME (day)
Type 6 0.588 178 406 731 356
Fig.9 Predicted Time-dependent
Behavior of Sockets (Method 3)
212 LONG-TERM PREDICTION OF BEHAVIOUR OF CABLE-STAYED BRIDGES

4. PREDICTION OF TIME-DEPENDENT BEHAVIOR OF A CABLE-STAYED BRIDGE

4.1 Finite Element Formulation

The equilibrium equations for the cable, tower and girder elements after applying Laplace
transform leads to a linear system of equations, whose stiffness matrices Kij are as presented bellow.

K^ ttjvB^mn(s)BfydV
,=i ,=i (7)
where Bim and Bnj are strain matrices and Emn(s) is the elastic modulus corresponding to the Laplace
space:

£^(5)= K + E2 El
s +— -
^ for visco-elastic elements (8-a)
for elastic elements (8-b)
The F.E.M. analysis as above described provides Solutions in the Laplace space which has to be
converted into the real time domain, so as to give the final Solution. Therefore, an inverse
transformation has to be carried out [2].
In case of cable-stayed bridges, the cables, towers and girders have different visco-elastic
constants and therefore, different intervals of convergence, which makes it difficult to perform a
numeric inverse transform considering, simultaneously, all the structural elements visco-elastic.
Thus, the bridge analysis was carried out considering each element, separately, visco-elastic (case
1,..., case n) and the final Solution was assumed to be a linear combination of all the n cases [3]. The
contribution factor for each of the terms of the linear combination is determined by means of the least
Square scheme in the Laplace image space and the use of the Regula-Falsi method.

4.2 Model Bridge

The model bridge, as presented in Fig. 10, is a cable-stayed bridge with a central span of 238.0m
and the side spans supported by a PC rigid frame bridge. The structural analysis was carried out for
half the bridge, considering its structural symmetry and the cables actually used in this bridge were of
type 4 and type 6.
Thus, the following cases were considered for the analysis:
- case 1: only the cables are linearly visco-elastic;
- case 2: only the concrete members of the PC rigid frame bridge are linearly visco-elastic, and
- case 3: final Solution assumed as a linear combination ofthe former cases.
Cable 11
Cable 1
Steel Girder PC Rigid Frame Bridge *. _

©0 OZZ2ZZ
0© ®/ |

0X0

Node 31

119m 238m 119m

Fig. 10 Model Bridge
4.3 Analytical Results
The structural analysis was performed using the experimental values of the specimen types 4 and
6, which, although having the same cable material, had different types of sockets. Fig.l 1 presents the
time Variation ofthe axial force in cable No.l (one ofthe longest cable) and Fig. 12, the time Variation
Jk E. WATANABE - M. KAMEI - Y. ISHIHARA - 0. NAKADE - L.H. ICHINOSE 213

of the axial force in cable No.l 1 (one of the shortest cables). The forces stabilized in a short time,
presenting similar values for both types of cables.

w
OOCCCOOOOOCCD o PJ cxxtPcdoodocö °
U 1000 U2200
O O
(JU
500
¦Type 4
Type 6 1 Type 4
Type 6
O Measured n O Measured

-500 1 1 i I 1 1 1 1

200 400 600 800 200 400 600

TIME (day) TIME (day)

Fig. 11 Variation of the Cable Force Fig.12 Variation of the Cable Force
Cable No.l (Types 4 and 6) Cable No. 11 (Types 4 and 6)

On the other hand, as it can be noticed in Fig. 13, the time-dependent behavior of the nodal
displacements were more remarkable in type 6, whose socket is of a more sensitive type. The
effectiveness of considering the cables and sockets separately in the evaluation of the visco-elastic
parameters is shown in Fig. 14, where the curves for the case in which cable and sockets are
considered separetly provides values closer to that of the data measured in situ.

50
1
^.^
Type 4 ti 40 Socket-Cable Separated
-J m o
Type 6 Socket-Cable Non-separated
O Measured £
30 v_ O Measured
20
"¦>"*
W 10
u O ü
o O <
¦J
o
Cu
00 CO -10
s 1 1 1 1
-20
200 400 600 1000
TIME (day) TIME (day)

Fig. 13 Displacement of Node No.31 Fig. 14 Displacement of Node No.31

(Types 4 and 6) Comparison between (i) Socket-Cable
Separated and (ii) Not Separated

A Simulation was also performed using a ficticious type of cable, assuming elastic constants
similar to that of type 1 (£i=200GPa, £2=4000GPa) and the delay-time (T=^\/Ei) of 50 days, which
correspond to the values of locked coiled rope [1]. Fig. 15 illustrates the time-dependent behavior of
the nodal displacement when using the ficticious cable.

2000

s >5
M 1800
-.
1600 -

X 1400

200 400 600 800 1000 400 600 1000

TIME (day) TIME (day)

Fig. 15 Displacement of Node No.31 Fig. 16 Variation of the Cable Force

(Ficticious Cable) Cable No.ll (Ficticious Cable)
214
 Spatial Behaviour of Plate Girde
Comportement spatial des ponts
de la rgnovation d
Räumliche Wirkung von Plattenbalken

S. NAKAMURA T. YAMAO Y.
Kawasaki Steel Corp. Kumamoto University Ka
Tokyo, Japan Kumamoto, Japan To

•rze *w%

^r""^T""
216 SPATIAL BEHAVIOUR OF PLATE GIRDER BRIDGES UNDER REDECKING

1. INTRODUCTION
In recent years, many reinforced concrete (RC) floor slabs of highway bridges
have been deteriorated or damaged and require rehabilitation, repair and
replacement. This is caused by the increase in traffic volume and the illegal
passing of over-loaded heavy vehicles. Although many studies have been reported
in reference to the repairing or strengthening method of damaged RC floor
slabs, the redecking method by orthotropic steel deck has become of major
interest in the view of the reduction of the dead load and the expected
remaining life of bridge lately[l,2]. The authors have been studied the useful
method of replacing damaged floor slabs with the prefabricated steel deck of
Battledeck Floor Type[3] which is easily manufacturedf4,5]. The bridge has to
be partially open to traffic because a long traffic close of bridges with heavy
traffic is usually not allowed in the work of replacing. Therefore, it
is
important to clarify the spatial behavior as a whole system and local stresses
of plate girder bridges during the redecking.
This paper presents the results of statical loading tests for a large scale
model and the finite element analysis for plate girder bridges. In the
analysis, the RC floor slab is modelled by thin plate elements having six
degrees of freedom for one node and the main girder by equivalent substituted
truss system. The supporting cross beam and the cross frame are modelled by
beam-column elements. Since the shear connector transfers the load primarily by
shear, itis assumed that their flexural and torsional stiffness can be
neglected. Validity and efficiency of the theoretical method are examined by
comparing with the experimental results. Many useful information for redecking
design are obtained from the results of tests and analyses.

2, SUMMARY OF LOADING TEST

2.1 Test model

Although the test model is designed to simulate the replacement of a damaged RC
floor slab with the steel deck, a bridge model with the steel deck in place of
the RC floor slab is used here. This is due to the practical reason that the
thin concrete slab in the model may cause the structural unbalance between the
floor slab and steel girders and it
may be difficult to conduct the casting and
removing Operations of the RC floor slab in the testing frame.
The bridge model is composed of three main girders, the prefabricated steel
deck, supporting cross beams and cross frames. The steel deck used is the
Battledeck Floor Type with welded longitudinal ribs as shown Fig.l and is
connected by high strength bolts to the main girders. In order to simulate the
behavior of the actual bridge under redecking, the steel deck is divided into
two panels along the length. The supporting cross beam is corresponding to the
transverse rib of the orthotropic steel deck and is fastened by high strength
bolts to main girders.
2.2 Test procedure
The objectives of the experimental programme are to investigate the spatial
structural behavior of the girder bridges as a whole system and the local
stress during the redecking. The statical load is applied at the centre of the
span through 300mm x hard rubber plates which are placed on the steel
120mm
deck. The loading conditions and the applied load are summarized in Table 1.
The test procedure corresponding to the actual replacement of the RC floor slab
with the steel deck is selected and consists of a total of seven steps as shown
in Table 2. The modeis of STEP 1 & 2 are composed of three main girders and
i\ S. NAKAMURA - T. YAMAO - Y. KAWAI - T. SAKIMOTO 217

SP^SPS^-
rß>

8S 4™
A00

1000
ioocr

26 <¥
1675 925

ffi
6 (3)250 500
.1
ia-125. 3@ 250 750 5Q
Deck IE

J T.J.
o^ B£>

1000
["
^^ 1000
12 XU
Supporting cross beam

Fig. 1 General layout of the bridge model

Table 2 Test procedure
Table 1 Loading cases
Loading
Max. Conditions
Load Loading Conditions CASEI
(kN) STEP1 }

J
o4- J J o J^L
J J-, J L
CASE III
CASE 1
150
'S
in i
OP
eu
-
__
0o
Li STEP 2
CASEI
l
o CASE III

CASE II 70
htr^-W^^Ti Ti CASEI
Hr lü
o
i ü STEP 3 l
CASEV

CASE III 100

1^LT
I
tf-^rh^^riij Ol
STEP 4
CASEI
l
CASEV

CASE IV 150
r J-
r1 J J J o
Jru-I^l-
nh
-1
o
_c _ L
CASEI
rx] n n STEP 5 l
Lü DT Eü CASEV

J- J J J -Jt_J"'"''L J
O CASEI

ri l
r1 STEP 6
CASEV 70
r dn nh _c
n CASEV
111 iL Li J
CASEI
STEP 7 l
CASEV
218 SPATIAL BEHAVIOUR OF PLATE GIRDER BRIDGES UNDER REDECKING

Plate element

Supporting cross beam Connecting member

Cross frame
Qy

/^
Truss element

Fig. 2 Theoretical model

cross frames without the steel deck, and for the model of STEP 2 supporting
cross beams are installed in addition to cross frames. beams Measurements of
strains in the main girders, steel deck and supporting cross were made by
using strain gages. The deflection of the steel girders and the shear slip
between the deck plate and the main girder were measured using electrical and
cantilever type displacement meters, respectively.

3. SUMMARY OF ANALYSIS

3.1 Theoretical model

The theoretical model used for the analysis of the plate girder bridge which
composed of steel deck, main girders, supporting cross beams and cross frames
is shown in Fig. 2. The steel deck is modelled by triangulär thin plate
elements having six degrees of freedom for one node. The stiffness matrix of
the plate element is derived with the consideration for in-plane flexural
stiffness[6]. The supporting cross beam is modelled by thin-walled beam-column
elements[7] and the cross frame by truss elements. The main girder is modelled
by an equivalent truss system which can störe the same amount of strain energy
with that stored in the plate girder. With reference to Fig. 3, the cross
sectional areas of the substituted truss members are given by the following
equations[8]:
Upper chord member: Acc (hc+ht)hc
Lower chord member: Act
I
(hc+ht)ht
 S. NAKAMURA - T. YAMAO - Y. KAWAI - T. SAKIMOTO 219

N Acc N
6AwG
Vertical member: Av
jlEtanB $\CAd>
Diagonal member: Ad
3AwG
2^Esin9cos28 r
V—\

N
Av

Act
Av

Fig.3 Substituted truss

where I is the moment inertia of the girder cross section about the neutral
axis, ht and hc are the distances from the neutral axis to each outermost
fiber, respectively, E and G are the Young's modulus and the shear modulus, \l
is the shape factor and Aw is the cross sectional area of the web plate. Though
the steel deck is connected by high strength bolts to main girders, the shear
connector is modelled by cantilever elements which permits only the horizontal
shear at the free end of element.
3.2 Computation method
The computation was carried out for the half of the girder bridge considering
the symmetrical condition at the midspan. The deck plate is divided into 10
elements both along the length and the width of the deck. In the modeling, the
cross sectional shape of the theoretical model is used the rectangular
determined according to the condition that the flexural stiffness of the deck
plate section equals those of experimental model. The bridge part of which
steel deck is removed, is modelled in the same way, but the thickness of deck
plates supposed to be removed is assumed to be negligibly small (0.001cm) in
the analysis.

4. RESULTS AND DISCUSSIONS

4.1 Load distribution effect of cross frames and supporting cross beams
The load distribution effects of cross frames and supporting cross beams are
examined at the test Steps 1 & 2 (without the steel deck) for the loading CASE I

100kN
_,____tjQ:

0
-J
1
- .-

GA
-J _j

N/
-j

\ \^x
L

GB
1 L 1

GC
I-.
1

100kN
£>-
L
"t _L-

^
1

2 ^^Nv
P 1Y J^
I
I I
1
I I I I I I I

3 ^----a -50 0 50
Stress(MPa)
CD 4 - O : STEP 1
Q O:STEP 1

5 a:STEP 2 A : STEP 2

Fig. 4 Deflection of main Fig. 5 Stress distribution of main

girders at midspan girders at midspan
220 SPATIAL BEHAVIOUR OF PLATE GIRDER BRIDGES UNDER REDECKING

75kN 75kN 75kN 75kN

O
-LJ-'LJ j
r J -I J JL L
ij
J L J J J J—L J I
__ _

\/ \/
I
s.
n y
y |
v y n y

0
GA GB GC
3 GB GC

1 ^,^j^^*^

^/ E 2
°\

/
~-^&"^^
2 a
/ ^c
/
ect

'S 4

y
*S 4 CL
Q O: STEP 4 O : STEP 4
5 ~ A STEP 5 5 - A : STEP 5
>0
O STEP 6 O : STEP 6
6 - <s 6 ¦-
(a) CASE I (b) CASE IV
Fig. 6 Deflection at midspan for each STEP
75kN 75kN 70kN

J J J J :r"ir."]":r"rri"3"nfT

ImmI l ¦ ' ¦ ¦
I >
¦ ¦ » l
X
-50 0 50 -50 0 50
Stress(MPa) Stress(MPa)
O STEP 4 O:STEP 6
A STEP 7 A STEP 7
(a) CASE I (b) CASE V
Fig. 7 Distribution of normal stresses at midspan

Figs. 4 and 5 show the deflection at midspan and the distribution of normal
stresses on the girder under concentrated load at the midspan, respectively.
It can be seen that 40% of the applied load is distributed to the exterior
girders by using cross frames (STEP 1) and for the model of STEP 2 by using
supporting cross beams in conjunction with cross frames 50 ~ 60% of the applied
load is distributed to the exterior girders. From these test results, we can
recognize that the use of supporting cross beams in the prefabricated steel
deck system can give large effects of the load distribution.

4.2 Behavior during redecking

The deflection at midspan subjected to a lateral load at the midspan for STEP
4~6 are shown in Fig. 6. It
may be found that the deflection for STEP 4 in
CASE IV and STEP 6 in CASE I are larger than those of the other loading cases.
This is due to the reason that the loads are directly applied to the main
girders where part of the deck plate is removed. Fig. 7 shows the distribution
of normal stresses on main girders at midspan for STEP 4,7 in CASE I and STEP 6,
7 in CASE V. By comparing the stress in the low flange of main girders with
(STEP 7) and without (STEP 6) the deck plate, we notice that the stress of main
girder without a part of deck plate is relatively small and is nearly equal
to those of a complete system (STEP 7). The reason of this is that the
moment inertia of the composite section is 2 ~ 3 times larger than that of a
main girder and this section is mainly in Charge of the applied load. From test
results, it
can be noticed that the supporting cross beams play satisfactory
J% S. NAKAMURA - T. YAMAO - Y. JWVAI - T. SAKIMOTO 221

Table Comparison of test and

3
100kN

\ T ^^ ö
analysis for each slip modulus k
(a) Deflection ijnitimm \~
GA GB GC
Experiment
k=9.8
Analysis
98
0.90
0.86
1.16
1.32
1.16
1.31
1.42
1.31
1.48
1.31
1.16
1.31
0.94
0.86
1.16
--i*0 *
-50 0 50
(b) Stress Unit:MPa
Stress(MPa)
GA GB GC
Experiment Analysis
Comp Tens Comp Tens Comp Tens STEP 3 A
Experiment 5.2 16.5 12.2 34.4 4.8 15.8 STEP 4 O
Analysis k=9.8 3.8 16.0 9.2 34.1 3.8 16.0 STEP 6 O
98 6.0 17.0 16.6 34.8 6.0 17.0
Fig. _9 Stress distribution at
midspan
75kN 75kN 70kN

J _J L J J J JN L J JL
H
L
j^tj
\/ _j
yy

0
GA GB \y y

GC GA GB GC

Pa -
„
1
&^
^ß^ £ 2
1 -
AT
&
c O^
•°- 3 _ W^ y
y 2 3
y tf
o 4 y' 4 Experiment Analysis

/' Experiment Analysis m

Q STEP 3 A
STEP3 A
5
y'O 5 - STEP 4 O
'' O
STEP 4
STEP 6 O STEP 6 O
6 6 6 •-
(a) CASE I (b) CASE II
Fig. 8 Comparison of deflection at midspan between theory and experiment

role for the load distribution in transverse direction together with for the
outer girder without the deck.
4.3 Comparison of tests and theory
To determine the slip modulus k of the shear connector prior to model analysis,
the complete modeis (STEP 3) subjected to a lateral load at midspan were
analyzed for two k values, k= 9.8 and 98 (N/m/m), based on the previous studies
[4,5]. The comparisons of tests and the theory for deflection and stress of
main girders are given in Table 3. The slip modulus k= 9.8 (N/m/m) was adopted
in this analysis because it
tends to give overestimated conservative results in
comparison with test results. The deflection at the midspan for STEP 3~6 in
CASE and II are shown in Fig. 8. Though the small differences between the
I

theoretical and experimental results for STEP 6 in CASE is observed,

be noticed that the
can
good agreement exists between the two sets of results.
I
it
Fig. 9 shows the normal stress distributions on three main girders at the
midspan for CASE IM. The theoretical results correspond to the experimental
results quite well. In spite of small differences for some loading cases,
can be seen that the results of the present analysis have fairly good
it
correspondence with the test results as a whole. The validity of the present
222 SPATIAL BEHAVIOUR OF PLATE GIRDER BRIDGES UNDER REDECKING

method can be recognized. This numerical method will be utilized to investigate

the spatial behavior of existing plate girder bridges under redecking.

5. CONCLUSIONS

Spatial behavior of plate girder bridges under redecking is studied

experimentally and theoretically. From this study the following conclusions
can be drawn.
l)The redecking procedure of actual bridges with deteriorated RC slabs is
simulated well by the experiment.
2)It is recognized that the use of supporting cross beams in the prefabricated
steel deck system can give large effects of the load distribution.
3) In spite of small differences for some loading cases, it
can be seen that
the results of the present analysis show good agreement with the test
results as a whole. The validity of the present method can be recognized.
4)The present method will be utilized to investigate the spatial behavior of
existing plate girder bridges under redecking.
The writers express his thanks to Mr. K. Ishii, Kajima Ltd., the former
graduate Student of Kumamoto University, for his valuable assistance in
numerical computations.

REFERENCES

1. R.
WOLCHUK Applications of Orthotropic Decks in Bridge Rehabilitation,
Engineering Journal. 3rd Qtr., 1987.
2. Bridge Design Department Ishikawajima-Tekko Construction Co. Ltd. The Work
of Replacement of Reinforced Concrete Slab of "Shin-Kannon Bridge" for
Chougoku Regional Construction Bureau, Ministry of Construction.
Ishikawajima-Harima Engineering Review, Sep., 1985. (in Japanese)
3. A.I.S.C. Design Manual for Orthotropic Steel Plate Deck Bridges.
4. FUJIWARA M., HARA M., YOSHIDA H. and KAWAI Y., Basic Experiments on Pre-
fabricated Steel Deck for Redecking. Proc. of 43th Annual Conference of
JSCE, Oct., 1988. (in Japanese)
5. MURAKOSHI J., NAKAMURA S., Statical Loading Tests
KAWAI,Y., YOSHIDA H. and
and Fatigue Tests on Prefabricated Steel Deck for Redecking using large
scale modeis. Proc. of 44th Annual Conference of JSCE, Oct., 1989.
(in Japanese)
6. YAMAO T. and SAKIMOTO T., Nonlinear Analysis of Thin-walled Structures by a
Coupled Finite Element Method. Eng./Earthquake Eng., Vol.3, No.2, Oct.,1986.
(in Japanese)
7. YAMAO T., SAKIMOTO T., YUJI S. and KAWAI Y., Analysis of a Plate Girder
Bridge with a Reinforced Concrete Slab. IABSE Symposium Brüssels 1990,
Sep., 1990.
8. YAMAO T., SAKIMOTO T., SHIIHARA K., KAWANO K. and KAWAI
I., An Analysis of
the Behavior of Crane Girders using an Equivalent Substituted Truss. Journal
of Structures and Materials in Civil Engineering, Vol.5, Jan., 1990.
(in Japanese)
 223

Method for Evaluating the Load Carrying Capacity of Existing Bridges

Methode d'evalution de la capacitö portante des ponts
Eine Methode zur Tragfähigkeitsbewertung bestehender Brücken

Souxin WU Yuanchang LAO

Doctoral Student Professor
Southwest Jiaotong Univ. Southwest Jiaotong Univ.
Chengdu, China Chengdu, China

Souxin Wu, born 1966, received
his M.S. degree at Southwest
Jiaotong Univ. He is now involved
in finite element analysis of
concrete structures and evaluation of
existing bridges.
Yanchang Lao, born 1920, received
his Ph.D. at Imperial College
of Science and Technology in U.K.
He is now involved in assessment
of structural capacity.

Dayuan SHEN Yalin PEI

Professor Lecturer
Southwest Jiaotong Univ. Southwest Jiaotong Univ.
Chengdu, China Chengdu, China

Dayuan Shen, born 1932, graduated
from Tangshan Jiaotong Univ.
He is now involved in research on
concrete bridges.
Yalin Pei, born 1958, received her
M.S. degree from Southwest Jiao-
ton Univ. She is now involved in
finte element analysis of concrete
structures.

SUMMARY
On the basis of inspection, testing and analysis for a number of existing bridges, the factors which have
influence on the load carrying capacity of existing reinforced concrete bridges are identified and a finite
element model for evaluating the load capacity is developed. A Simulation method for the evaluation is
proposed, by which the load testing of bridges can be simulated by means of Computers and the
characteristics of the load carrying capacity of existing bridges can be calculated.

RESUME
Sur la base d'inspections, d'essais et d'analyses d'un grand nombre de ponts, les facteurs qui influencent
la capacitä portante des ponts en böton armö ont 6t6 identifiäs et un modele par elöments finis a 6te
döveloppö. Une methode de Simulation pour l'ävalution est pr6sent£e, par laquelle l'essai de Charge du
pont peut etre simule au moyen de l'ordinateur; les caracteYistiques de la capacite portante des ponts
peuvent ötre ainsi egalement calculäes.

ZUSAMMENFASSUNG
Auf Grund der Untersuchungen, Proben und Analysen von zahlreichen bestehenden Brücken werden die
Hauptfaktoren, die Einfluss auf die Tragfähigkeiten von bestehenden Stahlbetonbrücken ausüben,
festgestellt und ein Finite-Elemente- Modell für die Tragfähigkeitsbewertung entwickelt. Mittels dieser
Methode kann die Belastungsprobe der Brücken mit Hilfe von Computern simuliert und dadurch können
die Kennziffern der Tragfähigkeit von Brücken berechnet werden.
224 LOAD CARRYING CAPACITY OF EXISTING BRIDGES

l.INTRODUCTION

With the development of transportation, the load carrying Capacity of a number of existing bridges are
found to be insufficient due to progressing deterioration and increased loads. There are nearly 136,000
highway bridges in China, About 5,000 of these are judged to be functionally obsolete or inadequate
for current requirements, the service age of these bridges are ranged from 30—years to 40—years. In
addition, some of bridges constructed over the last 20 years are considered to be structurally deficient
because of deterioration or distress[l][2].

While replacing all the deficient bridges mentioned above with new bridges is often extremely difficult
and expensive, a moderate increase of structural capacity through rehabilitation and repair is fairly
cheap and easy to obtain. To avoid high costs of rehabilitation and repair, the evaluation of the bridges
must accuratelly reveal the present load carrying capacity and any further changes in the capacity in the
applicable time span. In recent years, many method for evaluating the load carrying capacity of
existing bridges have been developed. these method can be roughly divided into three kind: Knowledge
—based method; computalional method; and load testing method. The knowledge—based method has
the advantage of assessing the damage state of the bridges, but can not give exact index about the load
carrying capacity. Most computational methods are similar to design methods, the hypothesis on which
design methods based are not quite the same as practical behavior of existing bridges, and computational
results may be doubtful. load testing on bridges can directly examine the load capacity and the results
are more reliable than other methods. However, it would be very expensive to test all the deficient
bridges. In order to explore the load carrying capacity of existing bridges, it is necessary to develop an
inexpensive evaluation method which can fully take into aecount the real behavior of existing bridges
and give reliable results about the load capacity.

The objeetive of this paper is to identify the factors which affect the load carrying capacity of the
bridges and develop a Simulation method for the evaluation. Using the method, the load testing of the
bridges can be simulated by means pf Computers and the index about the load carrying capacity can be
caculated. An example of evaluating a T-beam bridge is also presented.

2. FACTORS AFFACT1NG THE LOAD CAPACITY OF R. C. BRIDGES

Over the years of servicing, various forms of deterioration would appear on beams, piers and bases of
bridges and all affect the load carrying capacity. So the load carrying capacity include the capacity of
upper structure which composed of beams and deck and that of lower structure constituted by piers and
bases. Only the upper structural capacity is studied in this paper.

In order to assesss the damage State of old bridges and identify the factors which affect the load carrying
capacity of the bridges, a thorough field survey of old R. C. bridges located in Guangdong province in
China was made and static and dynamic load tests were performed on some of these bridges [3].
Inspection and testing show that the deterioration emerged on beams and deck are main factors which
influence the load carrying capacity. The deficiencies on the attachment such as discharge orifices and
expansion joints results in the damages on beams and deck, then have indirect influence on load
capacity. Various types of deterioration -efflorescence, leakage, cracking and spalling-can contribute to
the reduction of bridge' s load capacity to different degree. Ignoring the deficiencies which have little
influence on the load capacity and only beams and deck are considered, main factors äffecting the R. C.
bridge's load capacity can be identified as shown in Fig. 1.

Among the factors, cracking of concrete is very important to estimate the load capacity. The density,
width, length and pattern of cracks are significant Indexes for the estimation. The factors given in Fig.
1 must be take fully into aecount in the computational model for evaluation of R. C. bridge' s load

carrying capacity.
 S. WU - Y. LAO - D. SHEN - Y. PEI 225

Cracking and spalling Corrosion of Inadequate material

of concrete reinforcement properties

Insufficient load capacity

Fig. 1 Main factors affecting load capacity

3. COMPUTATIONAL MODEL FOR R. C. BRIDGE'S LOAD CAPACITY

3. 1 Mathematical Model

Theoretically, for a given loadings and structure type, a bridge' s load carrying capacity can be
determinated in terms of the parameters such as load components, dimensions, strength of materials,
etc. The evaluation philosophy of an existing bridge must differ from the design philosphy of a new
bridge. In this study, the load carrying capacity index, R is defined as follows:

R gls(0, /(0, x(0, <p(0 ] (1)

where S(t) represents load effect; /(0

represent strength of materials; x(f) represent dimensions; <p(t)
is structural integrity and is equal to 1 — d in which d is damage. All of these parameters are
functions of service time t Assuming that, R is differentiable continuous function of all the
parameters, the change of load capacity, AR can be derived from Eq. (1) as*

Aä (dg/dt) As + (dg/df)Af + (dg/az)A* + Og/d<p)A<P (2)

in which As A/ A% and A<p denote the changes of load effect, strength of materials, dimensions
>

and structural integrity, respectively. Over the years of Performance, the actual load carrying
capacity, R* at the time of evaluating, t is Witten as.

Rt R — AR (3)

The resistance coefficient, k can be defined as:

K= (Rt- G)/S (4)

where G represents dead load effect; S represent the effect of the live loads used for evaluation.

3. 2 Finite Element Model

To reveal the real load carrying capacity of bridges, a rational computational model must be
established. Actual behavior and the factors affecting the load capacity must be fully considered in the
model. The finite element method is appropriate for dealing with non — homogeneous materials,
nonlinear constitutive relationships and complicated boundary conditions. It is easy to dispose
226 LOAD CARRYING CAPACITY OF EXISTING BRIDGES

deterioration or distress which contributes to reduction of the load capacity by using the finite element
method. Thus a finite element model can be developed to caculate the load carrying capacity index, R
The major factors in Fig. 1 can be taken into aecount in the finite element model and constitutive
relationships (Fig. 2)

Fig. 3(a)shows a damaged simply supported T—beam, which has four cracks located at A, B, C,and
D, respectively. Spalling occurs at location E, and result in corrosion of parts of reinforcing bars.
Meanwhile, expansion bearing lose its efficaey. Finite element model of the beam is illustrated in Fig.
3. (b). Details of the model will be described in the following.

Cracking]—»—( FE meshes configuration Numerical techniques

Spalling &.
corrosion
Modified dimensions
iu FE model
Simulation of
load test

Inadequate material properties [ Constitutive modeis

Fig. 2 Management of the factors affecting load capeity

P,(i=l~n) P, P2 P,

t J l L

.Cracks

Ineffective//^7)//
Ü Spalling Cracks

L/2
(a) Deteriorated beam

P,(i= l~n)
1 S,
+&
Y/y
¦^
.^
¦g

7TT7T77

(b) Finite Element Idealization

Fig. 3 Finite element model of deteriorated beam

3. 2. 1 Concrete

The assumption of the plane stress is considered to be reasonable for T—beam subject to loadings in—
web plane. Reduction of concrete section due to spalling is taken into aecount by the modified element
 S. WU - Y. LAO - D. SHEN - Y. PEI 227

thickness, Tc defined as:

Tc KCT (5)

where Kc is modification coefficient, T is original thickness.

Existing cracks are modelled with two techniques* Discrete model and smeared crack model. The
discrete model approaches a Single fully separat!ve crack by disconnecting nodal points (Fig. 4(a)).
Interlock elements are placed across the incompletely separative cracks to simulate aggragate interlock
(Fig. 4(b))

> i >

f ' T '

i 1 i.-.i • i 1

**s
(a) (b)
Fig. 4 Discrete crack model

The smeared crack model represent overall influence of many discrete cracks existing in the domain of
the element. the constitutive eqation is expressed by[8];

Tda{\ VEX 0 Ol fitei]

\dai\ =10 E2 0 I
\de2 I
(6)
LfcrJ LO 0 gJ \dyj

where G„ is reduced shear modulus which reflects the density, width and degree of Separation of the
cracks.

3. 2. 2 Steel reinforcement

Main reinforcing bars are represented by axial force elements with two translational degrees of freesom
defined at each node. Corrosion of rebars is considered by reduction of the cross sectionl area of the
bars. Secondary reinforcement, such as stirrups is assumed to be distributed over concrete elements and
forms composite concrete — steel element. The material stiffness of the element is defined as follows
[8]:

M-M + SM (7)

where [ U ] and [ DI ] are the concrete and reinforcement material stiffness matrices, respectively.

3. 2. 3 Bond between steel reinforcement and concrete

If the influence of bond slip is considered, linkage elements must be set up to model bond behavior, so
the number of nodal points will increase greatly. For the sake of utilizing memory capacity of Computer
228 LOAD CARRYING CAPACITY OF EXISTING BRIDGES

effectively, reinforcement may be assumed to be connected directly to the concrete at the nodal
points.

4. CONSTTTUnVE RELATIONS FOR THE CONSTTTUENT MATERIALS

Material properties have a significant influence on the load capacity. Actual material parameters must
be used in constitutive relations. Biaxial nonlinear constitutive relations and failure theories should be
applied to explore the realistic load capacity of existing bridges.

The concrete constitutive model and failure crireria of Balakrishnan &. Murray[5] are introduced in this
study. The model divides the uniaxial response curve for concrete into five damage regions described as
linear elastic;compressive strain hardening; compressive strain softening;tensile strain softening; and
tensile stiffening regions. When used under biaxial stress conditions, the model is considered
orthotropic after cracking. The effect of biaxial stress conditions on peak strength is represented by a
Variation of the Kupfer — Hilsdorf failure curve in stress Space in which the compressive and tensile
envelopes are separatelly specified[4][5]. In the tension—compression region, tensile stress less than
0. 5 ft does not reduce compressive strength. If o\ >
0. 5ft the ultimate strength may be described
as:

/* 2f c/ü/2 + 2fjfd (8)

/« (-1.5+ /,„//<) l/cl (9)

In compression—compression region, the ultimate compressive strength may be written ass

U (1 + 3. 65)JV(1 + «)2 (10)

in which a o\/<Jt <j\ and <r2 are major and minor principal stresses respectively, fa,and ftuare peak
compressive and tensile stresses respectively. Details of the model are described in reference [5]. The
paramefers in the model such as cylinder compressive strength fc elastic modulus E should be
determined with nondestructive inspection such as sonic pulse velocity measurements. If neccessary,
cores are taken from bridge for compressive and split test.

Reinforcing steel is assumed to be elastic perfctly plastic material. Actual values of the yield strength fr
and elastic modulus Es are used. The strain—hardening region may be considered, if neccessary.

The bond stress—slip relationships of Mirza and Houde is used, expressed as[6]:

x (54 X 102S-25. 7 X lO5^ + 5. 98 X IOW-O. 558 X 10nS4 Vfc/41. 5) (11)

in which r is the bond stress in MPa, S is the slip in cm.

Interlock elements are employed to model the interface shear transfer across the crack by aggragate
interlock and friction. The stiffness of the element is derived from Horde &• Mirza' s shear stress-
displacement relation as[6]:
 S. WU - Y. LAO - D. SHEN - Y. PEI 229

Kg 63. 85(1/C)3/2 Vfc/35,4 (12)

in which A is the area for which one element is responsible, in cm2; f' c in MPa; C is crack width in cm;
kf in N/cm.

5. EVALUATING PROCEDURE

Based on the mathematical model and the finite element model aforementioned, a finite element
program can be developed and implemented into a particular Computer code to simulate the load tests of
R. C. bridges. For T—beam bridges, each beam of the bridge is evaluated as a Single unit. Simulation
results of all beams are synthesized and analyzed to give the resistance factor of the whole bridge. The
evaluating procedure is described in the following.

5. 1 Field Survey and Review of Design Documents

A thorough field survey is needed to obtain the information about deterioration of the bridge, including
crack location and size, corrosion of reinforcement, actual material properties, and as — built
dimensions. The presence and location of reinforcement can be determined though review of design
documents. If the design drawings are not available, the reinforcement location may be determined
using a pachometer which locate steel magnetically. The parameters needed in the computational model
can be defined through field survey and review of design documents.

5. 2 Determination of Loading pattern and Finite Element Meshes

According to loads for evaluation,the most detrimental loading pattern can be decided. The transverse
distribution of loads must be taken into aecount to determine the detrimental loading pattern applied to
each beam evaluated. Generally for simply supported beam, the internal forces such as bending
moments and shear forces at beam ends and mid—span control the position of loads. After defining the
load pattern, finite element meshes are construeted. Data file are prepared and inputed into Computer to
Start Simulation of load test using incremental load procedure described in the next.

5. 3 Incremental Loading

j
For the i th T-beam, the loading factor is defined as vhj in the th loading increment. If failure occurs or
the specified indexes, such as deflection and crack width, are reached at m th loading increment. The
resistance coefficient for the beam is defined as

Ki ^Wj (13)

5. 4 Evaluating the Load Carrying Capacity

If the bridge consist of n of beams and the resistance coefficient for every beam is caculated, the
resistance coefficient of the bridge is given as:

K <f>r* Min(KuK2»',Km) (14)

where <pr 1 — d? q>r and dr are integrity and damage of the transverse diaphragms, respectively; fc(i
230 LOAD CARRYING CAPACITY OF EXISTING BRIDGES

1,2, •••») is the i th beam's resistance coefficient.

If k > 1,
the load carrying capacity of the bridge is enough to meet the need of present traffic.
Otherwise, repairs or rehabilitation must be made to restore or increase the load capacity.

6. EXAMPLE OF BRIDGE EVALUATION

Using the computational model given is this paper,a Computer program RCBM for simulaing the load
tests of R. C. bridges is developed. The proposed approach is illustrated by an example. The load
capacity is evaluated for one beam of a T-beam bridge which is located in chengdu city in China. The
bridge, built in 1961, is a three-span simply supported T-beam bridge. The cross section are composed
of 12 T-beams. The span length is 16. 3m. The midspan cross section of one beam is shown in Fig. 5.
There are 14 main unnotched rebars which have diameter of 32mm. Concrete design compressive
strength is 18. 4MPa. A field survey was conducted. Spalling, corrosion of rebars, several inclined
and vertical cracks were found. Actual concrete compressive strengen is only 8. 3Mpa. One beam was
taken from the bridge for failure test in order to judge whether the load capacity is enough or not. The
load pattern is shown in Fig. 6 [7].

\* 140cm »(
' """
1

B
if"i
S P P p P

<f>32mm JlJ LJ_

-L .[
17cm
13. 0
H
I

-h'H
I 4.0 I

+4-
I
6.9 I

J(m)
1.2 1.2

Fig. 5 Mid—span cross section Fig. 6 Loading pattern

The design load p is 70. 5KN, corresponding to the bending moment of 782. 6KN—M in mid—span.
The program RCBM is employed to simulate the failure test and evaluate the experimental results as
shown in Table 1. The load—mid- pan defletion curves of the beam are shown in Fig. 7. It can be

\
seen, from Table 1 and Fig. 7 that the proposed Simulation method gives a good approximation to
>

failure test of the T—beam, the evaluation resuts are available.

Table 1 Comparision of Simulation results with test results

Max. Load (4P) Mid—span bending The resistance

(KN) Moment (KN—m) coeffcient

Design 282 782.6 __^—-^^

Simulation 540 1497 1.91

Test 600 1665 2.13

 S. WU - Y. LAO - D. SHEN - Y. PEI 231

Load (KN)
960

720

yV
480 Xy
— Simulation

Test
240 *>

20 40
f (mm)
Fig. 7 Load—midspan deflection curve

7. CONCLUSIONS

—An evaluation method for load carrying capacity of existing bridges should distinct from methods in
design of new bridges. A finite element model for evaluating the load capacity is developed, in which
the factors influencing the load capacity are taken into aecount. Nonlinear constitutive relationships and
failure criteria under biaxial stresses are employed to explore the realistic load carrying capacity.

—A Simulation method is proposed to estimated the load capacity of existing bridges. Using the method,
load tests can be simulated in order to given reliable resistance coefficient of the bridge. The method is
effective and inexpensive, since it may replace many load tests. Although the present approach is
developed for R. C. bridges, it can be used in the evaluation of various types of bridges,such as steel
bridges and composite bridges.

REFERENCES

1. "Inspection, Evaluation and Rehabilitation of Old Bridges," Special publication, Scientific &•
Technical Information Institute, Ministry of Transp. and Communications, China, June 1986. (in
Chinese)

2. Lou, Zhanghong, "Evaluation, Rehabilitation and Replacement of Existing Bridges," proc. of the
1987 Symposium, Bridge and Structural Engrg. Division comttee of China Highway and Transp.
Society 1987. (in Chinese)

3. Wu, Souxin," Research on Nolinear Finite Element Method for Evaluating the Load Carrying
Capacity of Existing R. C. Bridges," M. S. thesis, Southwest Jiaotong University, Chengdu,
China, 1990. (in Chinese)

4. Kupfer, K. Hilsdorf, H. K. and Rusch, H. ," Behavior of Concrete Under Biaxial stresses,"
Journal of ACI, No. 8, Aug. 1969.

5. Balakrishnan, S. Murray, D. W. ," Concrete Constitutive Model For NLFE Aualysis of

Structures," Journal of Structurl Engineering, Vol 114, No. 7, July 1988.
232 LOAD CARRYING CAPACITY OF EXISTING BRIDGES

6. Houde, J. Mirza, M. S. " A Finite Element Analysis of Shear Strength of Reinforced Concrete
Beams," ACI «Shear in Reinforced Concrete >SP42, vol. 1, 1974.

7. " Evaluation Report on the Load Carrying Capacity of Chengdu Baihuatan Bridge," Report,
Chongqing Jiaotong Institute Dcember 1988. (in Chinese)

8. "Finte Element Analysis of Reinforced Concrete," State —of —the —Art —Report, ASCE, New
York, 1982.
 Damage Assessment and Remainin
Evaluation des dommages et de
Schadensbewertung- und Restleben

Lubin GAO
Assoc. Professor
Railway Eng. Res. Inst.
Beijing, China
234 DAMAGE ASSESSMENT AND REMAINING LIFE PREDICTION FOR STRUCTURES

1. INTRODUCTION

The safety of railway transportation has been focused serious

attention of railway administration, research and design branches,
not only because railway accidents always induce heavy loss of
the life and property of peoples, but also because the most
important is it
is directly related to traffic schedule of
railway transportation. Of a lot of factors affecting railway
traffic safety, it
is the service state of railway track and
railway structures which plays a significant role to. Firstly,
railway construction, especially railway bridges have a long
service life and difficult rehabilitation characteristics.
Secondly repair expense of railway bridge is very great. Hence
railway maintenance branches has a very important, long-time task
which is bridge service state inspection. Since 1950's, damage
assessment of railway bridges has been investigated and a lot of
experimental data and experience have also been obtained. But due
to the complexity of the problem, it
is still a very important
research project to start deepgoing investigation on damage
assessment and remaining life prediction method for existing
structures. At the end of the paper, a prototype expert system
to diagnose damage state of railway bridges will be briefly
introduced.

2. BRIEF REVIEW ON THE RESEARCH

For a long time, railway bridge inspectors and researchers in

China have been searching for a reasonable and effective method
to assess service state of railway bridges. But up
no well method has been developed, owing to the limit of
tili
present,
experience and research work as well as the complexity of the
problem. Existing approaches may be classified into two kinds:
physical inspection and mechanical measurement, such as
structural outward appearance Observation for cracks and dynamic
response test. Physical means for interior inspection include
ultrasonic method etc. Not all of the existing methods is
effective.
In Northern China, railway has served for a considerablly long
time. Bridge damage is more serious. Hence Harbin Railway
Administration Bureau has paid more attention to the research and
practice of damage state inspection and assessment for bridges.
Since 1979fs, a lot of situ test data for dynamic response and
time-domain properties of railway bridges have been accumulated,
and a great deal of experience for diagnosing damage state of
bridges has been obtained. They applied maximum transver dynamic
displacement of bridge pier and some peculiarities of Vibration
wave of pier-head displacement as identifying rule to sum up into
10 types of damage for bridge pier.

China Academy of Railway Sciences started the investigation and

research work on this topics in last 1970*s. At the first stage
of research, CARS has gathered a lot of situ dynamic test data
/v L. GAO 235

for existing bridges, which have been also analyzed, for example,
the effects of damage on frequencies, modes, time-domain and
frequency-domain peculiarity of bridges. Classifying the test
data of about 80 bridge piers, CARS has suggested a Standard for
damage diagnosing of bridge pier.
The investigation on damage assessment for superstructure of
bridges in China started in middle 1980's. CARS firstly did a lot
of inspection work on riveted and welded steel bridges. Then a
great amount of tests on fracture mechanics and fatigue
properties of situ samples of bridge elements. For these kinds of
bridges, CARS has done about 10-year investigation and has
obtained some preliminary results. In recent years, CARS
does an amount of researches on damage assessment method for
still
reinforced and prestressed concrete railway bridges, as well as
strengthening method for damaged bridge rehabilitation.
According to present research state, the investigation and
research on damage assessment and remaining life prediction
approaches for existing railway bridges must be continued.

3. PREPARE WORK IN DAMAGE ASSESSMENT PROCEDURE

Prepare work for damage assessment of existing structures is

composed of two stages:
* situ inspection and testing
* laboratory testing and analysis
The main objeetive in the first stage is to acquisite practical
structure geometric, physical and mechanical knowledge and
information about the bridge to be assessed. The methods to
inspect the bridge include outward appearance Observation,
physical measurement for cracks etc. as well as mechanical test
to measure free Vibration frequencies, modes and dynamic response
to loading etc.

3.1 Situ Inspection and Testing

Outward appearance Observation

Outward appearnace Observation for railway bridges is one of the
routine duties of bridge maintenance branches. From the
inspection, we can obtain basic knowledge and information about
bridge service state, such as
cracks: length, location, direction, depth etc.
crust of reinforcement: location, extent etc.
deterioration of concrete: location extent etc.
inclination of bridge pier: direction, degree etc.
link of bridge and piers: state, relative displacement etc.
236 DAMAGE ASSESSMENT AND REMAINING LIFE PREDICTION FOR STRUCTURES

foundation: state
Interior defect inspection
On the basis of outward appearnace Observation, interior defect
inspection for existing bridges is mainly to survey interior
damage of bridge structures by means of physical method, for
example using ultrasonic approach to measure crack depth and
interior cracks of structures.
Situ mechanical testing
The aim of situ mechanical testing is to ac qui site basic
mechanical properties of the bridge to be assessed, generally
using dynamic testing to measure free Vibration frequencies,
modes and dynamic response to vehicle loading. In addition, situ
testing work also include obtaining samples for laboratory test
and some basic material property experiment such as concrete
strength test etc.

3.2 Laboratory Testing and Analysis

The in this stage composed of data treatment for the

main tasks
first stage, necessary laboratory test of situ samples in order
to deliever sufficient information for damage assessment and
remaining life prediction of the bridge assessed. Laboratory test
is generally to measure fundamental properties, fracture
mechanics and fatigue properties of samples of bridge structure
and materials.
Time-domain and frequency-domain analysis as well as dynamic
parameter identification are very useful to assess bridge service
state. The second work in this stage is posttreating dynamic
testing data recorded in situ test, including analysing and
identifying physical and modal parameters, time-domain and
frequency-domain peculiarities of the bridge.

4. DAMAGE ASSESSMENT AND REMAINING LIFE PREDICTION

From prepare work, we have obtained some knowledge and
the
information for geometric, physical and mechanical properties of
bridge assessed. The remained work is to assess service state of
the bridge and to predict its remaining life. How to do this
work? The
discussed above,
first damage
problem is how to model and quantif y damage .As
phenomena are very different and complex.
So it
is considerablly difficult modelling damage of structures.
As an example, we investigate effects of damage on dynamic
properties of bridge pier in following so as to find some
variables sensitive to damage.
 L. GAO 237

(1) Frequencies of bridge pier

Dr. Sun [21 investigated frequencies of bridge pier with and
without cracks in 1991. The first four frequencies of a pier is
listed in Table 1, from which we may know the effect of damage on
frequencies of bridge pier.

12
Table 1. Frequencies of a bridge pier (Hz)
Mode 3 4

Normal pier 17.05 18.93 113.95 137.21

Cracked pier 16.39 17.42 110.51

(2) Modal shape of bridge pier

Convex point may be found near cracks in modal shape of cracked
pier.
(3) Modal damping of bridge pier
Chen [3] found that modal damping of damaged bridge pier becomes
greater and greater as damage becomes larger.
(4) Maximum pier-head displacement
With the reduction of pier stiffness, maximum pier-head displacement
under vehicle loading becomes greater.
(5) Time-domain peculiarity
Forced Vibration response wave is smooth, similar to harmonic
wave for good state bridge piers, but
peculiarities for damaged piers.
it
may emerge some Singular

(6) Frequency-domain peculiarity

Power spectrum of damaged pier may be found with some differences
to that of normal pier, such as with a wider band and smaller
value at low frequency, a continuous spectrum and a greater
number of peaks etc.
For superstructure of railway bridge, the effect of damage is
more complex. In general, dynamic response parameters reflect the
synthesized properties of structures. Some local damages have no
evident effect on dynamic parameters. From identification of
system parameters, we can only obtain a global and average damage
description.
Here we classify damages of structures into two categories: local
damage and global damage. The first may also called nonstructural
damage, and the second structural damage. "Nonstructural" and
"structural" do not mean without or with effect on structural
capacity. But nonstructural damage only affects local stress and
deformation of strcutures, and does not greatly affect
macroscopic global displacement of structures. However the second
is directly reflected from global mechanical response, such as
frequencies, modes etc.
238 DAMAGE ASSESSMENT AND REMAINING LIFE PREDICTION FOR STRUCTURES

The first kind of damage includes local cracks, corrosion

reinforcement, exposure of concrete etc. The second is
of
the
worsening of structural materials.

4.1 Damage Tolerance Method

For the first kind of damage, we can use several characteristic

variables to measure
depth
it.
For example, using crack length 1, crack
a and crack open displacement d to describe a local crack
in structures. In general,
Characteristic variables: { Ai / i=l,2,...n }
Assume the damage produce a response quantity J to local stress,
deformation etc. of the structure:
quantity:
Response J

Through fracture analysis and material test, we may obtain a

eritieal value of J as Je. A simple method to define local damage
is
D=J/Jc
D=0 when J=0 structure
undamaged
D=l local failure
when J=Jc

This method for the first kind of damage is similar to the main
idea of Damage Tolerance Method of fracture mechanics.

4.2 Combined Damage Theory-System Identification Method

According
materials, elastic
to damage theory,
modulus
if there are cracks or vacancy
of material will be reduced.
in
The
author [5] obtained:

where % and JE are elastic tensors of damaged and undamaged

materials respectively. D is called damage tensor. ^ is a unit
tensor. In case of isotropy, it can be simplified as
E=(1-D)E
where D is a damage scalar.
Applying identification method, we can obtain elastic
system
modulus distribution of damaged bridges. Therefore damage state
of the bridges may be also identified.
The approach discussed above is called Combined Damage Theory-
iiik L. GAO 239

System Identification Method.

4.3 Remaining Life Prediction

In order to predict remaining life of existing structures, it is
necessary to model damage evoluting property of bridges. In
general,

or
D=f(D, t, F)
SD/ &N=g(D, N, F)
where represents external act. N represents loading cycles.
F

It is Firstly
a very complex work to obtain the damage evoluting equation.
we have to do deep investigation on damage
mechanism. Secondly we must grasp a lot of experiment results for
bridge materials and elements with and without typical damages.
If evoluting equation has been developed, remaining life
damage
of bridges assessed can be predicted from the present damage
state and the future loading spectrum of the bridges.
Nr=Nf-Ns
D(N+dN)-D(N)=g(D(N), N, F)(N+dN-N)
D=Ds when N=Ns
D=l when N=Nf
where Ds represents the present damage of the bridge.

5. A PROTOTYPS EXPERT SYSTEM FOR BRIDGE PIER ASSESSMENT

The expert system, DREPM-DP, is a prototype system developed by

China Academy of Railway Sciences, which is based on an expert
system tool DREPM. In acquisition of knowledges, domain knowledge
and expert knowledge as well as an analyse module are used. The
system has two kinds of knowledge representation method: frame
representation and productive representation. A ask-answer
blackboard is designed, which makes the system has a good humanmach
ine interface.
DREPM-DP has been used to assess damage state of about 20 bridge
piers. The diagnosing results are identical to practical
Situation.
Example: Haoshi River Bridge on Han-Dan Railline
The piers of the bridge is of a circular cross-section, enlarge-
ment foundation on rough sand with gravel layer. Bridge maintenance
branch reported that Pier No.3 rocked greatly when train
passed the bridge. In 1988, Zhengzhou Bridge Inspection Branch
did situ Vibration testing to the bridge. According to the
240 DAMAGE ASSESSMENT AND REMAINING LIFE PREDICTION FOR STRUCTURES

results recorded in the test, DREPM-DP gives such diagnosing

results:
Pier No.2: little worse state, D=0.4
Pier No•3: very worse state, D=0.1
Pier No.4: worse state, D=0.25

6. CLOSE REMARKS

Damage state assessment and remaining life prediction for railway

bridges are closely related to transportation safety. Deepgoing
investigation on how to assess the state and to predict the life
is very significant at present. According to the discussion
above, it has been found following topics must be done futher
studies:
(1) Modelling method for damage of bridges
(2) Damage criterion and damage-servicability relation
(3) Damage evoluting properties

ACKNOWLEDGEMENT: The author would like to express his heartful

thanks to Prof. Qingguo Cheng and Dr. Ning Sun for their
invaluable help and discussion with the author on this paper.

REFERENCES

1. GAO Lubin, CHENG Qingguo, Fatigue and Life Prediction for

High-speed Railway Bridges, IABSE Symposium, Leningrad, 1991
2. SUN Ning, Dynamic Behavior and Damage Diagnosing Expert
System of Bridge Pier, Thesis for Doctorate Degree, China
Academy of Railway Sciences, 1991
3. CHEN Xinzhong, Investigation on Dynamic Diagnosing Method for
Bridge Pier Damage, Thesis for M.S. Degree, China Academy of
Railway Sciences, 1986
4. GAO Lubin, Qingguo, Damage Assessment of Structures by
CHENG
Identification of StructureDynamic Parameters, Proceedings
of EASEC-3, Shanghai, 1991
5. GAO Lubin, A New Isotropie Damage Constitutive Model and Its
Applications, China Railway Science Journal, Vol.10, No.2,
1989
Probabilistic Response of Reinforced and
R6ponse probabiliste de sections de po
Wahrscheinliche Belastungsa
Stahlbetonbrücken

Juan R. CASAS
Dr. Eng.
Technical Univ. of Catalunya
Barcelona, Spain
•">'

Juan A. SOBRINO
Civil Engineer
Technical Univ. of Catalunya
Barcelona, Spain
¦
242 PROBABILISTIC RESPONSE OF BRIDGE CROSS-SECTIONS

1.- INTRODUCTION
In order to obtain an accurate reliability analysis of existing or future structures it is necessary to
use more realistic modeis for materials, geometrical variabilities and structural analysis, taking into
aecount the non-linear behavior [1]. The use of analytical or semi-empirical relations for the
obtention of the bridge section resistance available in design code must be improved.

The Statistical parameters of geometrical variability and the uncertainties in the physical properties
of the involved materials has been normally derived for the available literature data for building [2].
Nevertheless, in bridge construction both the construction considerations stipulated are different
and more accurate techniques are used. Thus, a higher quality construction can be obtained. On the
other hand, the available data are not directly suitable for other countries with different modes of
construction or quality control. So that, more information should be obtained.
Recently, numerical methods to evaluate the structural cross-section response of prestressed and
reinforced girders have been developed to obtain ultimate moment and shear responses and the
moment-curvature relationship [3], and they has been widely used for bridge evaluation and code
calibration [4]. There is a need for the analysis of other concrete bridge typical cross-sections.

2.- GEOMETRICAL VARIABILITY

The parameters involved in the geometrical snapes, alignment, configuration of building

components are subjected to uncertainties due to different sources: form works, placement of
reinforcing or prestressing steel, placing of concrete, assemblage procedures, etc. Geometrical
variations affect both, the seif weight of the structural elements, (dead load) and the cross section
response (effective depth, concrete Covers, etc.) Most of the different literature data available has
been collected in Western Europe, USA and Japan, mainly for building structures [5] [6].

2.1.- Experimental dam and proposed modeis

A large experimental data bank has been collected until now in different bridges recendy built in
Barcelona area (post-tensioned concrete slabs and one box girder bridge). Also, a large data has
been obtained from a group of reinforced and prestressed concrete bridges (slabs and girder
bridges) demolished, for urbanistic reasons because of the 1992 Olympic Games and the
construction of new infrastrueture, in Barcelona.

The parameters collected have been: Deck tickness in slab-girder bridges, geometric definition of
girders, depth of slabs, thickness of top and bottom slabs in box girder bridge, effective depth of
reinforcing, diameter of voids in voided slab, thickness of asphalt.
The measurements of all these variables have been analyzed in order to obtain the Statistical
parameters. A Kolmogorov-Smirnov test has been used to derive the theoretical probabilistic
model. The probability distribution function included in these study were: Normal, Lognormal,
Truncated Lognormal, Gamma and Truncated Gamma. For each of these distributions the
Kolmogorov-Smirnov test provide a rational measurement of the approach. In many cases, all of
these functions fitted well the sample, thus to simplify the rational use of the theoretical modeis the
criteria was to select the Normal or Lognormal probability density distributions. The results are
summarized in Table 1.

3.- MATERIAL UNCERTAINTIES

The available data and modeling for the physical uncertainties involved in the material and
mechanical properties is very large [7] [8] [9]. Anywise, to obtain accurate modeis it is necessary
to define the source and to process the samples that are homogeneous, in order to establish a
suitable probability functions for a well definite random variable to use in further calculations.
 J R CASAS - J A SOBRINO 243

In this paper, the data collected is restncted to materials recently used in concrete bridge
construction in Spain, with a mean quabty control of the materials and high quality control of
construction

Parameter Xnormnal Xmean Xmin Xmax Standard Dev Typeof

X (mm) X nominal Xmean Xmean Xmean distribution

Deck Slab
Girder Thick. 250 0 100 0 79 113 0 07 Normal

Horizontal
dimensions 250 600 099 1003 0 99 1007 0003-0 007 Normal
of Girders

Vertical
dimensions 150 094 1025 0 95 105 0025 Normal
of Girders 600 0 003

Depth of east 300 0 996 0 94 105 0 026 Normal or

in situ Slabs 1800 0 015 Lognormal

Depth of 266 1006 091 107 0 045 Normal

top Reinforc

Depth of 50 141 0 48 1 83 0 27 Normal

bottom Reinf

Thickness
of top slab 250 0 95 1 03 0 89 0 92 106 1 10 0 02 0 07 Normal
in Box Girder

Thickness of 200 1002 0 987 1012 0011

bottom slab Normal
m Box Girder 350^50 0 95 1 1 0 95 0 97 1 03 1 05 0 016 0 025

Diameter of 1200 1400 0 97 0 98 102 0 007 0 008 Normal

voids (Slab)

Thickness 45 80 0 95 114 0 58 0 76 125 153 0 26 0 11 Normal or

of asphalt Lognormal

The measurements of geometrical definition of girder cross seeuons have been classified in vertical and horizontal
dimensions

Table 1 - Geometrical vanabüity

3 1 - Concernmg Concrete

Different samples of compressive strength of concrete have been processed to get the Statistical
parameters and to obtain a good fit Normal and Lognormal PDF provide a rational approach for
modeling this parameter It is recommended to use Normal distribution for high quality concrete
The Statistical data are summanzed in Table 2, for three types of concrete
244 PROBABILISTIC RESPONSE OF BRIDGE CROSS-SECTIONS

Specified Age at fc, mean fcmin fcmax cov Typeof

distribution
fc,k(MPa) test (days) (MPa) fc,mean fc,mean

25 7 22 30 0 71-0 85 1 13 1 19 008 011 Normal

Lognormal

25 28 28 33 0 74-0 82 120 009011 Normal

Lognormal

30 7 34 081 1 19 0 14 Normal
Lognormal

30 28 36 0 70 120 0 11 Normal
Lognormal

35 7 30 35 0 72-0 88 109 1 16 0 07 011 Normal

Lognormal

35 28 40 42 0 77 0 94 1 05 1 16 0 03 0 10 Normal
Lognormal

Table 2 - Compressive concrete strength

3 2 - Concerning prestressing steel

A large data bank has been processed for two different types of Strands, 0 5" and 0 6", and steel
270K (186/167 MPa) The source has been the more important manufacturer of prestressing steel in
Spam, which provided hundreds of quality control tests, conform to ASTM A-416 specifications
All these data corresponds to prestressing steel used in post-tensioning concrete bridges in the last 3
years Analysis of data and results of the Kolmogorov-Smirnov test are summanzed in Table 3 An
example of the sample is shown in Figure 1

Typeof Parameter X mean X nominal Xnun Xmax COV Typeof

Strand X Xmean Xmean distnbuti

Normal
05 Emodul 197 0 190 0 0 96 104 0 018 Lognormal

Normal
05 Ty 0 2% 180 6 166 0 0 92 108 0 028 Lognormal

Normal
05 Tmax 195 5 186 0 0 95 106 0 017 Lognormal
Gamma
Normal
06 Emodul 196 5 190 0 0 95 106 0 019 Lognormal

Tr Gamma
06 Ty,0 2% 247 0 238 0 0 94 108 0 022 Tr Lognor

Lognormal
06 Tmax 2716 266 0 0 96 107 0 018 Normal

E Deformation module (kN/mm2), Tmax Tensile strength (kN) and Ty,0 2%= Yield force (kN)
Tr Gamma= Truncated Gamma Tr Lognorm= Truncated Lognormal

Table 3 - Prestressing Steel (270K) properties

 J.R. CASAS - J.A. SOBRINO 245

3.3 - Concerning reinforcing steel

The available data for reinforcing steel is not as large as in the case of prestressing. Different
quality controls, made in some bridges, provide us the data. The mechanical properties are very
related with the bar diameter in the analysis performed. The results are presented in Table 4
although its can not be significant until the data bank will be more representative.

NORMAL NORMAL
LOGNORMAL LOGNORMAL
TRUNC LOGN TRUNC LOGN
TRUNC GAMMA TRUNC GAMMA
GAMMA GAMMA

N g;
1 ^

iL 2« 24-5 250 255

O^EL
2<>0 265 23Ü 235 240 245 250 255 260 265

YIFLD TENSILE FORCE (kN) YIELD TENSILE FORCE (kN)

Figure 1.- Histogram, PDF and CDF curves, for yield tensile force prestressing steel. (Strand 0.6")

Parameter X nominal X mean Xmia Xmax c.o.v. Type

X Xmean Xmean of
Distribution

Area.real 1.0000 1.003 0.98 1.05 0.002 Normal

Area,nom Lognormal

fy 51.0 58.0 0.91 1.10 0.057 Lognormal

(kN/mrn2)

f max 60.0 67.9 0.98 1.10 0.050 Lognormal

(kN/mm2)

Table 4.- Reinforcing steel parameters (fy yield tensile stress f max= maximum tensile stress)

4.- RESISTANCE MODELS

Accurate resistance reliability modeis to obtain the real response of cross sections must take into
aecount the real strain-stress relationship of the materials involved, and consider the uncertainty in
the geometry and material properties [4].

A numerical procedure has been developed, considering the above mentioned needs, to obtain the
moment, shear and torque response and the moment-curvature relationship of typical cross-sections
246 PROBABILISTIC RESPONSE OF BRIDGE CROSS-SECTIONS

of reinforced and prestressed concrete bridges, conform to CEB Model Code [10]. The model has
been computerized for easy application. In order to predict the probabilistic response and to fit a
theoretical probability distribution a 400 Monte-Carlo simulations, for each case, were performed.
The following assumptions were made:

- Strain-stress curves from CEB Model Code.

- The theoretical PDF used in the simulations are in conformity with data bank collected. The user
can also use the experimental histogram.

- All parameters involved to define the cross-section geometry and strain-stress curves are
considered as a independent random variable, in Statistical sense.

- In each Simulation all the above mentioned parameters are actualized.

Sensitivity analysis is made, in a recently built bridge in Barcelona (Fig. 2), to reveal the effects
on the random response due to selected parameters such as:

- C.O.V of vertical and horizontal magnitudes of geometry, diameter of voids, fc, fct, yield stress
of prestressing, effective steel prestress after losses, depth of prestressing.
The parametric analysis concerning C.O.V. is because of this statistic is directly correlated with
quality of materials and construction and human error, for new bridges, or with the level of
uncertainty in the unknown involved parameters of existing bridges. Thus, the more relevant
parameters in the resistance evaluation can be selected and used to rationalize the test tasks and
inspections.

4.1- Voided $lat>, Margenat Bridge in Barcelona,

This is a simply supported prestressed concrete voided slab, east in situ in 1991, with a span length
of 27.40 m, the cross-section and the placement of prestressing are shown in Figure 2 [11]. The
Moment-curvature relationships are shown in Figure 3, with design values (factored resistance) and
with mean and characteristic values of parameters involved. Numerical results of the response
value analyzed are given in Table 5. Due to the lack of space only the most relevant results of
parametric study are summarized in Figures 4, 5, 6 and 7.

10 00
Rj 3.- MOMENT CURVATURE RELATIONSHI?

200 i 300 3 00 l 2 00
70.00
t—i—i—i—i—r
6000
135
13b I07I07
10 1i
135

Hi
1 7 10 7 ti

tI I
P=2%
?5000
Q
-4000

^::Ql:@-l^ g30.00
1

1 70
/ L 1 70^ I
0
2
20.00 MOMENT (mean param)
FACT0REDM0MENT
3 30 340 3 30 1000 MOMENT(charac param)

000
¦ ¦ ' ' ' L
3x4 prestressing rendons 31 0 0 5 00000002000400060008001000120014
DIMENSIONS IN METERS Curvature (m-1)

Figure 2.- Margenat bridge. Typical cross-section. Figure 3.- Moment-curvature relationship
from decompression of concrete
 J.R. CASAS - J.A. SOBRINO 247

Parameter Xmean Xmean Xmin Xmax c.o.v. Type

X X nominal Xmean Xmean of
Distribution

Ultimate 1.254 *
Bending- 57.45 0.92 1.10 0.034 Normal
Moment 1.067 Lognormal
(MN*m)

Normal
M, crack 1.00 32.48 0.90 1.14 0.046 Lognormal
(MN*m) Gamma

Lognormal
Inertia 1.00 0.829 0.75 1.23 0.089 Tr. Lognor.
section (m4) Gamma

Area 1.00 6.657 0.91 1.08 0.032 Normal

section Lognormal
(m2)

Table 5.- Simulation results. * design value, factored resistance 0c 1.5 and 0s 1.15).
Nominal values conform to CEB Model Code, with characteristic parameters.

The real case of cross-section herein studied yield a good example to evaluate the main important
parameters involved in its ultimate resistance and serviceability behaviour. It is easy to realize that
the most important parameters are related with geometry and not with those concerning with
strength of material, due to the ductile behaviour. In the same way, the parametric study varying
C.O.V of void diameters and with yield tensile stress of prestressing shows a not important
correlation with the section properties analyzed.

5.- CONCLUSIONS

5.1.- Conclusions concerning geometrical and material variability.

Due to the lack of data available for modeling geometrical and material variability in concrete
bridges an important experimental data has been collected and presented. The Statistical analysis of
data fairly shows that different Statistical parameters has to be considered in the analysis of
reinforced and prestressed concrete bridges, which usually have a more accurate construction than
buildings.

5.2. - Conclusions concerning probabilistic response of concrete bridges

The scarcity of analysis of the most typical cross-sections in concrete bridges has conducted to
develop a numerical procedure to obtain the probabilistic response, in terms of moment, torque and
shear, taking into aecount the non-linear behaviour of materials and the uncertainties in the
parameters involved. A parametric study has been presented as a guideline to determine the
sensitivity of resistance and geometrical properties of the cross-section to different varying C.O.V.
of main parameters. The results show that the most important parameters to be correctly and
accurately evaluated are cross-section geometry and depth of prestressing steel.
 248 PROBABILISTIC RESPONSE OF BRIDGE CROSS-SECTIONS

Fig 4 - Influence of C O V depth prestressing

Fig 5.- Influence of Slab depth
0 12 0 30

COV ofMu COV Are*

0.10 0 25 COV Mcr
Mcr COV Inerüa
0 08 0 20

0 06 015

0 04 0 10

O
Ü 0 02 005 <r »>^ »¦*-—

0 00 0 02 0 04 0 06 0 08 0 10 0 12 0 00 0 02 0 06 0 08 0 10 0 12
C.O.V. depth of prestressing C.O.V. slab depth

Fig 6.- Influence of C O.V of Tx Fig 7.- Infl uence of CO.V. fc

008 i i J i 0 05
i 1 1 1 1
1
—•— COVMo

\\
1

007 • COV Mcr 1* 0 05 ~

006 - < ^ _ 004 <

* + >

0 05 ~
^r*^~ - ' - 004 -
-^* ~

004 - ' - - 0 03
- cov
¦" Mu r
'
1 \ 1 1 1
—a— COV Mcr 1
1 I
0 03 1

0 08 010 0 12 0 00 0 05 010 0 15 0 20 0 25 0 30
COV. Tx C O.V. fc

Note: Tx=prestress after losses, fc= compressive concrete strength.

7.- REFERENCES

1.- KARAGEORGOU, P. AND SKABARDONIS, A.; " Reliability of structural members under
time-dependent loads" Proceedings of the 4th specialty Conference on Probabilistic and structural
reliability, pp. 359-362, January, 1984.
2.-ELLINGWOOD, B. GALAMBOS, T.V., MAcGREGOR, J.G. AND CORNELL CA.;
"Development of a probability based load criterion for American national Standard A58 ". National
Bureau of Standards, NB's Special publication 577, Washintong, D.C., 1980.
3.-TING, S.C.;" The effects of corrosion on the reliability of concrete bridge girders"; Ph D Thesis,
Department of Civil Engineering, University of Michigan, Ann-Arbor, Michigan, 1989.
4.-NOWAK, A.S.;" Calibration of LRFD bridge design code" NCHRP Project 12-33. Department
of Civil Engineering, University of Michigan, Ann-Arbor, Michigan, 1992.
5.-CASCIATI, NEGRI AND RACKWITZ; " Geometrical variability in structural members and
Systems ". Joint Committee and Structural Safety. Working document, January 1991.
6.-MIRZA, S. A., MAcGREGOR, J. G.; " Variations in dimensions of reinforced concrete
members". Journal of the Structural Division ASCE, Vol. 105, ST4, pp. 751-766, April, 1979.
7.-MIRZA, S. A., MAcGREGOR J.G.; " Variability of mechanical properties of reinforcing bars
". Journal ofthe Structural Division ASCE, Vol. 104, ST4, pp. 921-937, May, 1979.
8.-MIRZA, S. A., HATZINIKOLAS, M., MAcGREGOR J.G. ; " Statistical description of strength
of concrete". Journal of the Structural Division ASCE, Vol. 105, ST6,pp. 1021-1037, June, 1979.
9.-VEGH, L.;"Quality control and rehabilitation of concrete structures in Czechoslovaquia". Bull,
ofthe International Assoc. for Shell and Spatial Structures, No 105, Vol. 32, pp. 41-49, April, 1991.
10.- CEB-FIP Model Code for concrete structures, Lausanne, 1990.
IL- CASAS J.R. and APARICIO A.C.; "Project of the bridge in Margenat street". Regional
Government of Catalunya. 1991.
 249

Dynamic Behaviour in Existing Concrete Bridges and Damage Detection

Comportement dynamique des ponts en bäton et dötection des dommages
Dynamisches Verhalten bestehender Betonbrücken zur Schadenerkennung

Ayaho MIYAMOTO Hidenori MORIKAWA Yoshiaki KAJITANI

Prof. Dr. Research Assoc. Dep. Dir.
Kobe Univ. Kobe Univ. Hyogo Prefecture Government
Kobe Japan Kobe Japan Kobe, Japan

A. Miyamoto, born 1949, received
his Doctor of Engineering at Kyoto
University in 1985. His recent
research activities are in the areas
of structural safety assessment on
concrete bridges and establish-
ment of design concept for
concrete structures under impact
loads.
H. Monkawa, born 1959, received
his Master of Engineering degree
from Kobe University in 1984. His
current research activities are in
structural safety and reliability
analysis of concrete structures. He
is now developing a fuzzy expert
system for structural safety
assessment of concrete bridges.
Y. Kajitani, born 1943, received
his Master of Enginenng degree
from Kobe University in 1969. He
is attached to the Road Construction
Division and is now working
on technical management related
to middle span bridges, and also
has great mterest in integrated
management system of steel and
concrete bridges.

SUMMARY
In this paper, the sensivities of the dynamic behaviour of concrete bridges constituted by natural
frequency, mode shape, damping and phase angle in damage detection is studied utilizing an analysis of
the complex eigenproblem considering non-proportional damping and using dynamic loading tests on an
existing bridge in which some specified artificial damage had been included. Also, a method of damage
assessment for concrete bridges integrated from both the concise detection of damage location based on
the difference in the sensitivities of modal parameters and the exact evaluation by localization and
quantification of multiple damage based on the system identification method, is discussed.

r£sum£
Les sensibilitös du comportement dynamique des ponts en böton constituöes par la fröquence naturelle,
la forme modale, l'amortissement et le däphasage sont examinäes en vue de la detection des dommages,
avec le recours a l'analyse de problemes propres complexes, et en tenant compte de l'amortissement non
proportionnel et des essais en Charge dynamiques sur des ponts existants od quelques dommages
artificiels späcifiques ont 6t6 induits. On y discute ggalement une methode d'estimation des dommages
des ponts en bäton int£gr6e a la fois par une detection concise de la localisation du dommage selon la
difference des sensibilitäs des parametres modaux et par une Evaluation exacte par localisation et
quantification des dommages multiples selon la möthode d'identification des sySternes.

ZUSAMMENFASSUNG
In dieser Arbeit wird die Empfindlichkeit im dynamischen Verhalten, ausgedrückt durch natürliche
Frequenz, Modalform, Dämpfung und Phasenwinkel, von Betonbrücken in der Schadenserkennung
untersucht. Dabei wird die Analyse von komplexen Eigenproblemen unter Berücksichtigung der nicht-
proportionalen Dämpfung und dynamischer Belastungstests auf vorhandenen Brücken, bei denen
spezifizierte künstliche Lasten aufgebracht wurden, eingesetzt. Ausserdem wird eine Methode der
Schadensbewertung von Betonbrücken diskutiert, bei der die schnelle Erkennung der Schadenslage und
die genaue Bewertung der Lage und Quantifizierung von mehrfachen integriert sind. Erstere basiert auf
dem Unterschied der Empfindlichkeit modaler Parameter, während für letztere die Systemindenfikation
eingesetzt wird.
250 DYNAMIC BEHAVIOUR IN EXISTING CONCRETE BRIDGES

1. INTRODUCTTON

The need for damage assessment of existing concrete bridges by a combination of visual inspections,
loading tests and analytical studies, has been pointed out with reference to the diagnosis of bridge
serviceability [1]. Since there are a number of factors included in the relationship between the damage
to existing bridges and dynamic behavior, it is necessary to develop an efficient method for damage
detection based on dynamic loading tests [2]. The most important aspect of this problem is to focus on
the dynamic sensitive parameters to the damage, because this has a significant influence on the
aecuraey of assessment.
In this paper, the sensitivities of dynamic behavior constituted by the natural frequency, the mode
shape, the damping constant and the phase angle, for damage detection was studied using an analysis of
the complex eigenproblem considering non-proportional damping and also dynamic loading tests. For
the analytical study, the component mode synthesis (CMS) method, which is one type of coupling
technique for substruetures in the dynamic analysis was applied to the complex eigenproblem for
simplification and an iterative analyses for damage detection was utilized. The system identification
(SI) method was developed based on sequential linear programming (SLP), combined mth the dynamic
sensitivity analysis, to quantify the degree of damage for each member in the whole system.
For the application of this method to existing concrete bridges, parametric analyses for simply
supported RC-T beam bridges in service were executed to evaluate the sensitivities of damage to
dynamic behavior and to construet a concise flow for damage detection. Furthermore, the SI method
was applied to the results from the dynamic loading tests, performed on an existing bridge in
whichsome specified artificial damages were induced. Finally, the concise flow and the SI method were
integrated, to enable an efficient damage assessment by multi-level and multi-aspect approaches.

2.ANALYTICAL METHOD FOR DAMAGE

DETECTION OF EXISTING BRIDGES
2.1 Modeling of bridges Gir der A
For existing concrete bridges, stiffness reduction
of the main girder has been caused due to the li r^
Itasdi
Dam a&e A A
interactive effect of flexural cracks and Gir der B
retrogression in the modulus of elasticity of the
li 5r^
bsm!
Dam ige B iL
concrete, The safety of bridges is strongly Gir, der C
influenced by this process as due to a change of l* *-L/4
"
*-L/2
"
*-3L/4~
A
load distribution and hence a reduction in load 10800
carrying capacity. In this research, the stiffness (a) Analysis model
reduction and change of damping constant were
considered to be the damage factors. The
modeling of the target bridge was carried out by n.r
1
CD 1
5600
™n \
using a lumped mass gridform model of finite
/Asphalt pavement 770
£Oy
beam elements and spring elements for the
elastic restraint of rotation at the supports. Fig. 1
1

//\? \ Cv)

shows an example of the model for an existing / ~

wju.a
RC-T beam bridge, "Nakaibashi" located in 670

Hyogo prefecture in Japan. ,46Q K lbt ü ÜSU! Ibh Ü 6Q

Girder A Girder B Girder C (mm)
2.2 Component mode svnthesis method in the (b) Section of Nakaibahi brid&e
complex eigenproblem
The CMS method deals with the equation for the
whole structure in modal coordinates obtained by Fig. 1 Modeling of existing RC-T beam bridge
synthesis of the boundary modes of substruetures
previously evaluated by modal analysis in the
physical coordinates. When the whole structure is
divided into n pieces of substruetures, the equation of motion for the substrueture i can be expressed by:
Maa Mab IM +
Cab M +
J\-aa K.ab
•IJol (i)
M\a Mbb - IM Cba Cbb I Kba Kbb 18*1
where subscript a, b denote the internal area and the boundary area, respectively.
 A. MIYAMOTO - H. MORIKAWA - Y. KAJITANI 251

By using Guyan reduction of the stiffness matrix, a correlation between the displacement in the internal
area and the boundary area can be expressed by:
ö* --[jüJ1[j&]ö* - r,ö» (2)
After this, an analysis of the complex eigenproblem for the synthesized whole structure, of which the
matrix size is reduced to the degree of freedom for the boundary area of each substructure, is carried
out Displacement at the boundary area b can be evaluated in the form of complex conjugates by linear
combination of each mode as:
_ _
yb - fy
4>* (3), where §b - \$bu ?«. <j>w, §b% $t*, <(>*]: the complex mode matrix,
k denotes the adopted number of modes, yb \§b> ö&/ %a - {?u» &h ?«&}T: the modal coordinates.

On the other hand, carrying out the analysis of the constrained mode for each substructure,
displacement at the internal area a for each substructure can be expressed by:
y™ - \j\$ü>, 4>i«] & (4), where <|>m - [<j>wi, 4>,ai» <!>*& $vq> * tyum, bumü : the complex mode matrix,
T, 0
m denotes the adopted number of modes, & {&& ?w) T{ -
0 Ti

Displacement of the whole structure can be expressed by using the modes at the boundary area and the
internal area, as:
{yis ya)T - X {& %a)T -X (5), where yfl » {v^ v^
Ti
y„)T, g« - {$to ?«,}T,
?ia
^
7*i
0
r-
<|>&

[T^b <|>a

Finally, the equation of motion for the whole structure can be written as:
0 M» 0 Mto -Af» 0 -Mba 0

0 Mob 0 Maa -Mab 0 -Maa 0

XTPX + XTQX 5 -0 (6), where P - .<?-
Mo, Cbb Mba Cba 0 Kbb 0 Kba

M* Cab Maa Caa _ 0 Kab 0 Kaa

By analyzing the eigenproblem for Eq.(6), the modal parameters for the whole structure can be
obtained. Furthermore, substituting the modes evaluated by Eq.(6) into Eq.(5), the modes of the whole
structure in the physical coordinates can be obtained. The degree of freedom for this analysis is the sum
of the adopted number of modes for the
whole structure synthesized by the modes for
the boundary area of all substruetures; 2k,
and the adopted number of modes for the
internal area of all substruetures; 2Lmh (i-1,
«, n: number of substruetures), and it can be [l-l]th mode [l-2]th nxfe [l-3]th mode
seen to be much less than the total degree of
freedom for the whole structure. Through
study of the aecuraey of this method using
the existing bridge model shown as Fig.l, it [2-l]th mode [2-2]th mode [2-3]th mode
was founded that the results of this analysis
were sufficiently accurate for the target
modes of bridge Vibration as shown in Fig.2, Fig.2 Shape of target modes
even if the adopted degree of freedom for the
whole structure was half the total degree of
freedom.

2.3 SI method using sensitivity analysis

The SI method [3] is one type of back analysis method, which can be used to identify system
parameters such as the flexural rigidity corresponding to the degree of damage in the problem, by
minimizing the error between the mechanical behavior as obtained from test and analysis. Here, the
existing concrete bridge was modeled as shown in Fig.l, considering spring elements for the friction
252 DYNAMIC BEHAVIOUR IN EXISTING CONCRETE BRIDGES

restraint of rotation at the supports. In this research, the sensitivity analysis of damage to mechanical
behaviors and the SLP method were integrated. The objeetive function was defined as minimizing the
total squared error between the mechanical behavior obtained from the field test and the analysis, by:

'•"IM + Wi
l IIa min

where/? is the order of normal Vibration, n is the number of measuring points, u, um are the eigenvalues
(7)

obtained from the analysis and the field test, respectively, Z, TP1 are the normalized modes of Vibration,
and W\, Wz are the weights for the eigenvalue and the Vibration modes. Here, it is assumed that
Wi=1.0,W2=l/n.
The sensitivity (derivative) ofthe design variable(for identification, here assign it the rigidity K± *=1~A
/: the number of members) to the objeetive function, can be expressed by:
OL
ÖKt
W± I^L - i\
uf» \tf dK
^
+ y 2W2 [Zj*
ki z£ \Z£ I dK
t\ dZj (8)

where
d\ip _r dK - dZp
-t dK
dK
and K is the stiffness matrix
_ Lp
—
dkt
Z*p
dK
of the whole structure.
* vr-vp
L.j
dK
7
£j\

Following this, identification of the

design variables can be performed by START
applying the SLP method using the
objeetive function and its derivative for
the design variables. Fig. 3 shows the Determination of initial values for design
variables Xi (i-I, 2, •••, n),
flow of the SLP method for the dynamic such as flexural rigidity of raain girders
problem. Firstly, the initial values of the and spring rigidity for rotation at supports
design variables are evaluated by
analysis. Linearization of the objeetive
function is then carried out within the
region of movement limits for the
design variables, and a search for the
minimum point of the objeetive value is Evaluation of modal parameters by analysis
tried using the simplex method. In the X (x Eigen value
event that the change of design variables ym(x Mode
exceeds the movement limits, m number of nodes
reanalysis of the modal parameters and
restart of the search for the minimum
values from updated initial values are
executed by the same procedure
Judgement of yes
iteratively up to the stage at which the k+1 error for
objeetive function is within the
^ STOP
identification
allowable limits.

3.APPLICATION TO AN EXISTING
RC BRIDGE WITH ARTIFICIAL
DAMAGE
Calculation of objeetive function Fk and
3.1 Outline of target bridge derivative of objeetive function VFk
The target existing bridge "Oyasubashi"
is 27 years old and a simply supported 4
Calculation of updated objeetive function
girder RC-T beam bridge with a skew Fk+1 Fk + VFk-SXk
angle of 46 degrees and a span length of I SXk | < S{ m tf öllu t Move linit
14.7m. This bridge seemed to be almost
intact according to the results of Visual
inspections. Fig.4 shows an outline Xk+i Xk + £Xk
section for this bridge.

3.2 Inducement of artificial damage

Fig.3 Flow of SLP (for dynamic problem)
The artificial damage corresponding to
 A. MIYAMOTO - H. MORIKAWA - Y. KAJITANI 253

flexural cracks in concrete in the tensile region C.L.

were induced partially in the girders A and B as
shown in Fig.5. Introduction of cracks was „Asphalt pavement
carried out by core boring along a vertical line
from the bottom surface of concrete in the
tensile region to the neutral axis, as shown in
Fig.6. The core concrete blocks were then
replaced in the holes from which they had been
1500 450
extracted, in order to avoid altering the weight mm

of the girders. Girder A Girder B Girder C Girder D

3.3 Procedure for field test [41 Fig.4 Section oiOyasubashi bridge
Prior to the field test, the target bridge was
modeled as the lumped mass system shown in
Fig.7, and the modal parameters were
evaluated by analysis. From this a T Tjg^e j» m GIrdbr A jmss drcppin
»in&
measurement method was determined focusing 4150 £ 4150 J • Sensor
the antinodes of target Vibration modes, as Girrjer lN

\
shown in Fig.5. Positions of forced vibrations t\4150 JTl39Q*T
by falling mass were arranged to obtain the
Girder N
—u f
various modes of Vibration and the mass (unit mraj\ Girder
dropping was carried out from about 70cm
height for ten times at the same loading point
to cancel white noise and to obtain a stable Flg.5 Outline of dynamic loading test
average value. The modal analysis [5] was then
applied to the acceleration data to identify the
modal parameters.

3.4 Identification of damage parameters Slab

The stiffness reduction of main girders was Haunch
identified by the SI method based on the [1-
Artificial
l]th eigenvalue and eigenvector obtained from Web damage by
the field tests, using the following procedure: concrete lOOmm
(l)For the target bridge before inducement of core
boring
the artificial damage, the system parameters
constituted by the stiffness of main girders 4150mm(GirderA), 4390mm(Girder B)
and cross beam, and the spring coefficient of
rotation at the supports, were identified. Fig.6 Inducement of artificial damage
(2)After inducement of the artificial damage,
the stiffness of the main girder in Sie
damaged region and the spring coefficient of
rotation at supports were identified under
the condition that the girder stiffness except »
Girder A
for damaged region was fixed at the value Girder B
identified in (1).
Girder
Table. 3 shows the results of the above
procedure. According to the results, the change Girder ^Ä^ 4 6

^-t
of natural frequency due to inducement of r.
&xJ
Cross
cross beam
beam
^^
artificial damage is about 8% i.e. relatively
great The results of the SI method can be seen Fig.7 Analysis model for Oyasubashi bridge
to show that the spring rigidity of rotation at
the supports changed sharply due to
inducement of the damage, although this bridge has simple support conditions in its design. In such
cases, the SI method considering the spring rigidity of rotation at the supports as the design variables is
effective. Identified girder stiffness before inducement of the damages was equivalent to the theoretical
value considering the stiffness of concrete in the tensile region, and it agreed with the results of visual
inspections and material tests of concrete cores extracted from the target bridge. Also the evaluated
degree of damage was then relatively great and qualitatively matched the theory.
254 DYNAMIC BEHAVIOUR IN EXISTING CONCRETE BRIDGES

Table 1 Results of damage detection for Oyasubashi bridge

Girder stiffnessCx 1012Kg«cm2) Spring rigidity of rotation at supports (x109kg-cra/rad) Ist

Girder Girder frequency

A B C D A B C D (HZ)

Theoretical value 7 198 6 755 6 755 7 255

-
6 346 310 310 100 100 0 900 0 900 0 690 0 690 12 69
Identified value 8 242 7 701 7 160 1 1 1 1

without damage

0203 0 0122 0 0100 0 0089 0 0075 0 0073 0 0046 0 0044 11 71

Identified value 5 496* 5 671* 0

with damage

* Value at damaged region

4. EFFECTIVENESS OF MODAL PARAMETERS IN DAMAGE DETECTION

4.1 Modeling of hridge with damage
The target bridge is shown in Fig.l, and this was modeled considering partial damage. For the damage
condition, the change of damping constant in the region of 0-30% in the region of A, B were assumed.

4.2 Damping characteristics

Fig.8 shows the relationship between the change of damping
constant in the damaged area and that for the whole bridge
system. For the [l-l]1*1 mode of Vibration, the sensitivities of
damage A and B are the same and have linearity, independently 20 O Damage A
of the change of damping constant in the non-damaged area. On A Damage B
the other hand, according to the results for the [1-2]111 mode, • Damage A, B

damage B in girder B which is the node of Vibration has no 515

influence on the whole bridge system. Accordingly, localization $
Damping constant
of the damage can be carried out using the damping at non-damaged
15%
characteristics for both the [l-l]01 and [1-2]111 modes. 10 region
Damping constant
4.3 Phase angle at non-damaged
region 1%

Fig.9 a)~c) shows the relationship between the change of »2 5

damping constant in the damaged area and the difference in
phase angle for each girder for the [l-l]111 mode of Vibration.
Here, the difference in phase angle was evaluated on the basis 10 20 30
of the phase angle at the midpoint of girder B. The difference in Damping constant at
phase angle in the damaged girder is greatest, in particular, the damaged region (%)
influence of damage A is significant. The sensitivities of these (a) [l-l]th mode

parameters have linearity independently of the change of

damping constant in the non-damaged area. On the other hand,
Fig.9 d) shows the result for the [2-1]111 mode and it can be seen 10
ODamage A
that the sensitivity of damage A is great and the difference in CO =3
phase angle between girder A and C is about 0.5rad i.e. largest. +-> U

Accordingly, localization of the damage can be carried out by

using these characteristics, similar to the above-mentioned
damping characteristics. -H35

10 20 30
4.4 Flow of damage detection based on modal parameters Damping constant at
Fig. 10 shows the flow of damage detection by localization and damaged region (%)
quantification based on the sensitivities of modal parameters to (b) [l-2]th mode
damages. Firstly, a brief evaluation for the location of damage
can be carried out using all or a portion of modal parameters
constituted by natural frequencies /i_i, /i_2, fi-2, damping Fig.8 Damping characteristics
constants £i-i, Ci-2, and phase differences yi-1» \|/2-i. At the first
 A. MIYAMOTO - H. MORIKAWA - Y. KAJITANI 255

O.OB 0.05

—-A
«^Jtr-—— O—

—-o— _____ A— —-o

_-W4of Girder.4
Vm—ml/1 of Girder/1
äJ-0.05 - £-0.05
^ a—-A3L/4 of Girder/J
O—ol/4 of Girder^
«« A—A3L/4 of Girderi?
-0.10
2 • #1/4 of Girderf
11-O.lOh» «1/2 of Girderf
a A3/4 of Girderf
10 20 30 0 10 20 30
constant at
Damping constant at
Damping
region (%)
damaged region (%)
damaged
(a) [1-1]th mode with damage A, (b) [l-l]th mode with damage A,
Damping constant at non- Damping constant at non-
damaged region 1% damaged region 15%

0.05 - 0.3 - •—•1/4 of Girder,4

a—A3L/4 of Girder^
• #1/4 of Girder^
| 0.2 - a A3L/4 of Girderr A

1/4 of Girder/I
"
1 of Girder/1
^-o difference
A •
-0.05 74 of Girder^
O—OL/4 of Girderi?
• «)
A3L/4 of Girder£
#1/4 of GirderC N^IO 20
Damping
30
¦1/2 of GirderC
Phase

-0.10
A3L/4 of Girderf
constant at
^s. damaged
0 10 20 30 v ^a region
Damping constant at -0.2
damaged region (%)
(c) [l-l]th mode with damage B
-0.3 -

(d) [2-1]th mode with damage A

Fig.9 Characteristics of phase angle

step, the existence of damage can be detected by searching whether the parameters (f\-\ and^i), &-i,
indicate large values. After this, at the second step, localization of the damages i.e. the distinction
\|/1_1
between damage the inside girder or the outside girder, can be carried out by searching whether
parameters fi-2, Cl-2> \|*2-l indicate large values. Though the above evaluation can be carried out
independently for each modal parameter, the final decision for damage detection should be carried out
comprehensively by comparison among the results from all parameters. Furthermore, the exact
evaluation by localization and quantification of multiple damages can be carried out by the SI method.
As the above mentioned procedure, the concise flow and the SI method can be integrated, to enable an
efficient damage assessment by the multi-level and multi-aspect approaches.
5. CONCLUSIONS

The main conclusions obtained from this study can be summarized as follows:
(l)For the simplification of analysis of the complex eigenproblem considering non-proportional
damping, the CMS method was applied and its suitability was demonstrated.
(2)The SI method based on dynamic sensitivity analysis and the SLP method has been studied and
applied to the results of dynamic loading tests performed on an existing concrete bridge in which
some specified artificial damage was induced.
(3)The sensitivities of dynamic behaviors to damage were evaluated by analysis and the results could be
seen to show that the [2-1]111 and [2-2]1*1 natural frequencies, the [l-l]*
[1-2]01 and [2A]^ damping
constants, and the [l-l]1*1 and [2-1]1*1 phase differences had high sensitivity.
256 DYNAMIC BEHAVIOUR IN EXISTING CONCRETE BRIDGES

START

J___.
Dynamic loading test
4r

Modal analysis

S or UK

ÜK

Damage in inside girder Damage in outside girder

first
I
Damage assessment at level
SiSmall

I
•> Application of SI method L:Large
UK : Unknown

Damage assessment at second level

Fig. 10 Flow of damage detection

(4)By using these modal parameters, a concise flow of the damage detection without any complex
analysis was constructed. Furthermore, this concise flow and the SI method were integrated, to allow
efficient damage assessment by multi-level and multi-aspect approaches.

REFERENCES

1. Natke, H. G. and Yao, J. T. P. (ed.), Structural Safety Evaluation Based on System Identification
approaches, Proceedings of the Workshop at Lambrecht/Pfalz, Friedr. Vieweg & Sohn, 1988.
2. Nowak, A. S. (ed.), Bridge Evaluation, Repair and Rehabilitation, Proceedings of the NATO
Advanced Research Workshop on Bridge Evaluation, Repair and Rehabilitation, 1990.
3. Douglas, B. M. and Reid, W. H., Dynamic Tests and System Identification of Bridges, Proceedings
of ASCE, Vol. 108, No.STIO, 1982.
4. Morikawa, H. and Miyamoto, A., Structural Safety Evaluation and Remaining Life Prediction of
Existing Concrete Bridges, Proceedings of the 17th Conference on Our World in Concrete &
Structures, 1992.
5. Ewins, D. J., Modal Testing-Theory and Practice, Research Studies Press, 1984.
 257

SESSION 4

STRUCTURAL EVALUATION BY TESTING

258
 259

KEYNOTE SPEAKER
260
 Analytical and Observed Respo
Comportement analytique et obse
Berechnetes und beobachtetes Ve

Baidar BAKHT
Principal Research Engineer
Ministry of Transp. of Ontario
Downsview, ON, Canada

Leslie G. JAEGER
Associate
Vaughan Eng. Associates Ltd
Halifax, NS, Canada
?**£¦¦ -¦»
262 ANALYTICAL AND OBSERVED RESPONSES OF STEEL GIRDER BRIDGES

1. INTRODUCTION

Very significant advances in analytical techniques during the past three decades have led to the
development of extremely powerful and versatile methods of structural analysis that are generally
computer-based. Especially within the linear elastic ränge, these methods are known to predict
accurate and consistent sets of results. Even methods developed from different fundamental
considerations predict virtually the same set of results, thereby lending credence to each other's
aecuraey. Because of their rigor and pereeived capability to predict the actual behaviour of
structures, these advanced methods of analysis are used extensively for both the design and strength
evaluation of bridges.

Predictions from rigorous analyses are usually accepted with confidence as representing the actual
behaviour of bridges, and reliance on the ability of analysis to predict the actual behaviour of a
bridge increases with the rigor of the method.

Jointly between them, the authors have spent more than 60 years on research related to different
aspects of bridge engineering; much of this research has been condueted in the analytical aspects of
bridge engineering, e.g. [1], [2], [3], [4] and [5]; they also had the opportunity of being involved in
a large number of field tests on short and medium span bridges, e.g. [6], [7], [8], [9] and [10].
Through such involvement, many comparisons have been made between the observed responses of
bridges and those given by advanced methods analysis. It has frequently been found that significant
discrepancies existed between the predicted and observed responses, even when the loading was
within the linear elastic ränge of the structure.

As might have been foreseen, the discrepancies between the analytical and measured responses were
subsequently found to be due not to inadequacies of the methods of analysis, but rather to the
presence of behavioral factors which could not be included in the mathematical modelling because
of difficulties in their quantification.
Without for a moment denying their usefulness, especially in the design office, the authors have
come to believe that even highly rigorous methods of analysis cannot be relied upon
unquestioningly to predict the actual response of a bridge. In support of this somewhat
provocative assertion, results from tests on five bridges with steel girders are presented. It is
emphasized that the limitation of the examples to bridges with steel girders is due only to
considerations of space availability for the paper. Similar examples, underlying the difficulties in
realistic analysis, are also available for other structures. It may also be noted that results are
discussed herein of only those tests in which the authors had a direct involvement. For this
reason, the references cited are only those contributed by the authors.

2. BRIDGE WITH TIMBER DECKING

The first example presented is that of the rolled steel girder Lord's bridge with nail-laminated
timber decking in which the wood laminates are laid transversely. As described in [11], the bridge
is 6.25 m wide and has a single span that is apparently simply-supported. The girders are 10.2 m
long with a bearing length of 0.53 m at each end, and rest directly on timber crib abutments.
There are no mechanical devices to transfer interface shear between the girders and the timber
decking although there are 100 x 200 mm nailing Strips bolted to the top flanges of the girders; the
decking is nailed to these Strips. The Lord's bridge was tested with a test vehicle under several load
levels and different longitudinal and transverse positions. Even up to the highest load level, the
girders responded in a linear elastic manner. For two of the load cases, the longitudinal position of
the vehicle was the same but the eccentric transverse positions were the mirror images of each
other.
 B. BAKHT - L.G. JAEGER 263

o
"¦S2.0
qj
*aä
¦o
2.4 ttti **»«
f nrr
measured (girder 1 on left)
*>* measured (girder 8 on left)
— Semi-contmuum analysis
Note: discrete distribution factors for
girders are joined by continuous
lines to facilitate readability

|1.6
¦o
"E1.2
o
o
§0.8
o
^0.4
"co

0.0
transverse girder Position
Figure 1/ Distribution factors for mid-span deflections
in the Lord's bridge

For these two load cases at the highest test load level, the distribution factors for mid-span
deflections are plotted in Fig. 1 by viewing the cross-section of the bridge from two different ends
so that the two transverse distribution profiles overlap each other for easy comparison. It is noted
that the distribution factor for deflection is the ratio of the actual and average girder deflections at
the transverse section under consideration.

If the geometrically-symmetrical bridge was also symmetrical with respect to its structural
response, the distribution factors for the two mirror-image load cases, noted above, would have led
to transverse distribution profiles that lie exactly on top of each other. As can be seen in Fig. 1,
the two profiles are fairly close to each other but not exactly the same, thus pointing out that the
two transverse halves of the bridge do not respond in exactly similar manner to corresponding
loads. The two sets of distribution factors obtained from measured deflections, are also compared
in Fig. 1 with those obtained from deflections given by the semicontinuum method of analysis
[4]. It can be seen that the analytical values of the non-dimensionalized deflections are not any
more different from the two sets of observed values than the latter are from each other. This
confirms that for the bridge under consideration, the semi-continuum method used for analysis is
able to predict the pattern of transverse distribution of loads fairly accurately.

The same aecuraey of prediction, however, cannot be claimed in the case of the absolute values of
girder deflections. This is because of uncertainty in quantifying the parameters discussed below.

As noted earlier, the girders for the Lord's bridge are 10.2 m long with an unusually long bearing
length of 0.53 m at each end. It is customary to assume that the nominal point-support for a
girder lies midway along the bearing length, in which case the nominal span of each girder would
be 9.67 m. It can be demonstrated, however, that for the case under consideration, the vertical
pressure under the supported length of a girder, should have its peak away from the midway point
and towards the free edge of the abutment. Determination of the exact location of this peak
requires detailed knowledge of the modulus of subgrade reaction of the timber crib abutment.
Clearly, this factor is not easily quantifiable thus making the task of determining the effective span
very difficult. It can be appreciated readily that the clear span of the girder, being 9.14 m, is the
lower-bound of the effective span of the girder.
264 ANALYTICAL AND OBSERVED RESPONSES OF STEEL GIRDER BRIDGES

25
—analytical, non-composite, L 9.67 m
—analytical. composite, L 9.14 m
«««measured

20

|15

"S510-

0L

transverse girder Position

Figure 2/ Girder deflections at mid-span in

the Lord's bridge

The transverse modulus of elasticity of wood, which is operative in the longitudinal direction of
the bridge, is extremely small compared to the longitudinal modulus. Even if the transverse
laminated deck were made composite with the girders, the contribution of the deck to the strength
and stiffness of the composite section would usually be expected to be so small as to be
negligible. Consequently, no attempt is usually made to provide shear connectors in such bridges.
There are some holding down devices, however, to connect the deck to the girders through the
nailing Strips; these devices, by transferring some interface shear, do make the girders partially
composite with the nailing Strips and the decking. From measured girder strains, it was discovered
that despite the absence of shear connectors, the decking and the nailing Strips of the Lord's bridge
were partially composite with the girders. The degree of composite action was found to vary from
girder to girder, and clearly was not quantifiable.

The Lord's bridge was analyzed using two different sets of idealizations. In one idealization, the
girders were assumed to be non-composite and with a simply-supported span of 9.67 m. In the
other idealization, füll composite action was assumed between the girders and the timber
components, being the nailing Strips and the decking; the girders were assumed to have the lower -
bound span of 9.14 m. As can be seen in Fig. 2, the measured deflections for the same load case
for which the distribution factors are plotted in Fig. 1, are bracketed entirely with very large
margins by the analytical results corresponding to the two idealizations. It is tempting to believe
that the actual condition of the bridge lies somewhere between the two sets of conditions assumed
in these idealizations and consequently, errors in analysis are related only to the uncertainties of
span length and degree of composite action. However, there is at least one other complicating
factor, namely bearing restraint, which was not accounted for in these idealizations and which can
have a significant influence on the bridge response; this factor is discussed below.

Observed bottom strains of the girders near the two abutments were generally found to be
compressive, indicating the presence of significant bearing restraint forces which varied almost
randomly between the girders. It was found that there was no consistent pattern in the bottom
flange strains at the mid-span, these being smaller or larger than the corresponding top flange
strains. This Observation points towards the random, and hence deterministically unquantifiable,
nature of both the bearing restraint and the degree of composite action. Because of the presence of
 B. BAKHT - L.G. JAEGER 265

these factors and the difficulty in the estimation of the effective span, the analysis for the bridge
under consideration cannot be expected to replicate the actual behaviour of the bridge.

3. TWO-GIRDER BRIDGE

The Adair bridge is a single span, single lane structure with a clear span of 12.8 m, as shown in
Fig. 3. As is also shown in this figure, the bridge comprises a concrete deck slab supported by
two outer steel girders and five inner steel stringers, with the latter spanning between the
abutments but also supported within the span by two transverse floor beams that frame into the
two girders. A proof test on this bridge is described in [12].

Mid-span strains in the top and bottom Hanges of the two girders due to two load cases are plotted
in Fig. 4 against the longitudinal position of the test vehicle. It can be seen in this figure that the
strains in the top flanges are always much higher than the corresponding strains in the bottom
flanges. This Observation confirms the presence of fairly large bearing restraint forces. Large
compressive strains in the bottom flanges of the girders near their supports also confirmed the
presence of significant bearing restraint which again cannot be practically quantified for inclusion
in the mathematical model for analysis.

Much larger magnitudes of strains in the top flanges of the girders also point to the lack of
composite action between the girders and the deck slab, this bridge not having any mechanical
shear connection with the girders. Because of the lack of composite action, the top flanges of the
girders getting little relief from bearing restraint at the bottom flanges, govern the load carrying
capacity of the girders.

It is interesting to note that, unlike the case in the Lord's bridge and other bridges discussed later,
bearing restraint does not provide any significant reserve of strength in the Adair bridge.

The uncertain nature of the composite action in slab-on-girder bridges without mechanical shear
connection is underlined by the Observation that, in the same Adair bridge, the inner stringers are
able to develop füll composite action with the deck slab despite the lack of mechanical shear
connectors.

Because of the composite action, the stringers had become considerably stiffer thus relieving the
non-composite girders of a much greater share of the applied loading than would have been the case
if they were also non-composite. It can be appreciated that analysis cannot be very effective
without the knowledge of the degree of composite action in the various beams; such knowledge is
practically impossible to obtain without a test.

west -floor beam 1

^-floor beam 2east 4.88 m
TT _150mm (min.)

pnjL,
Unorth south
12.80 m n
13.71 m i¦ j
_l
^.76^1076^076^76^^76^1076^
0.46 m cross-section
elevation

Figure 3/ Details of the Adair bridge

266 ANALYTICAL AND OBSERVED RESPONSES OF STEEL GIRDER BRIDGES

Ear .*. »SSI ssST u. tSa

reference axle

strain gauge at top flange /.

strain gauge at bottom flange

400

300
north girder
\
$h* 'ift: ¦&&

$&^ fe*M?<$l&\
E 200 <^v
M' >» south girder

P
¦s 100
compressive strains in the top flange
tensile strains in the bottom flange

Figure 4/ Girder strains at

mid-span in the
3 4 5 6 7
Adair bridge
longitudinal Position of axle no. 2

4. ULTIMATE LOAD TEST ON A BRIDGE

An ultimate load test on a single span, right, i.e. skewless, slab-on-girder bridge, called the Stoney
Creek bridge, is described in [10]. The bridge, which had a clear span of 13.26 m, was loaded to
failure in 1978 by loading it with concrete blocks piled in six layers. Girder strains at the midspan
were recorded after each layer of blocks had been placed on the bridge.

To check the validity of the recorded data, the mid-span moments taken by the girders and the
associated portions of the deck slab, computed from measured strains, were compared with the total
applied moments. It is recalled that in a right, simply-supported bridge, the total moment across
any transverse section is obtained by simple beam analysis, and is statically determinate. When it
was found that the moments computed from measured strains were up to 30% smaller than the
applied moments, the aecuraey of the measured data was initially questioned. An example of the
comparison of moments thus computed from measured strains and average applied moments is
presented in Fig. 5 for load due to one layer of concrete blocks. It is noted that the girder strains
under this loading were well within the limit of computed elastic strains.

The initial computations of moments from measured strains were made by assuming that the
girders were free from any horizontal restraint at the bearings. The bearing restraint forces were not
initially entertained as possible cause for the moment discrepancies mentioned above. This was
because bearing restraint forces of the magnitude needed to reduce the applied moments by up to
30%, were believed to be unlikely to develop in practice.
 B. BAKHT - L.G. JAEGER 267

Load layer No.6

*&&} Load layer No. 1

¦^ffW^^i.

111111
Load layer No.5

Load layer No.4

Girder No. 1
Load layer No.3

50 Load layer No.2

X
Average of applied s Load layer No.1
beam moments q'
E 100
rl I ] L 1 1 J
Girder No. 1 2 3 4
2.5
Load layer No.;
Average of computed
moments Load layer No. 1
« 50
E 2.0
Moments computed from
/
only measured strains at mid-span
1

Girder No.
Load layer
1
0
Figure 5/ Girder moments in the Stoney Creek
bridge computed from observed data
by ignoring bearing restraint
i 0.5
Load layer o. 6

b 0.0
3 4
Girder Number

Figure 6/Girder moments in the Stoney Creek

bridge due to load at different levels

Subsequent tests, some of which are discussed herein, confirmed the presence of significant bearing
restraint forces in similar slab-on-girder bridges in which girders rest upon steel bearing plates.
The presence of these forces invalidates the assumption of simple supports and the compution of
moments obtained from measured strains on the basis of no external forces. In light of the
knowledge gained from the other tests, the data from the test on the Stoney Creek bridge were re-
analyzed about ten years after the test by back-calculating the bearing restraint forces that may have
occurred. From these revised computations, it was found that the bearing restraint reduced the
applied moment by up to 18%, rather than the 30% ränge that had been wrongly deduced by
previous calculations.

Distribution factors for mid-span moments taken by the girders and the associated portions of the
deck slab are plotted in Fig. 6 for loads at different levels. It is interesting to note that the
transverse distribution pattern of the bridge does not change very significantly as the load
approaches the ultimate sixth layer. As the failure of the bridge approaches, the load gets
redistributed only slightly among the most heavily loaded girders. The girder most remote from
applied loading, receiving little load at early stages of loading, continues to receive low levels of
load even when the total load approaches the failure load of the bridge.

An important outcome of the test was the Observation that in the absence of mechanical shear
connection, the composite action between a girder and the deck slab, that may exist at low levels
of load, breaks down completely as the load approaches the failure load for the girder.
268 ANALYTICAL AND OBSERVED RESPONSES OF STEEL GIRDER BRIDGES

5. A NON-COMPOSITE SLAB-ON-GIRDER BRIDGE

The unquantifiable and random nature of the bearing restraint forces, and of the degree of composite
action in the absence of mechanical shear connection, is illustrated by the results obtained from a
test on the Belle River bridge [13]. The Belle River bridge is also a slab-on-girder bridge with steel
girders and an apparently non-composite concrete deck slab. The nominal span of the bridge is
16.3 m and the width 9.1 m.

As indicated earlier, the transverse load distribution analysis of slab-on-girder bridges without
mechanical shear connectors is made difficult, to the point of becoming impossible, by the
uncertain degree of the composite action. One is tempted to believe that the actual load
distribution pattern of such bridges could be bracketed by two sets of analyses: one corresponding
to füll composite action and the other to no composite action at all, with the former analysis
always leading to safe-side estimates of the maximum load effects in the girders. In reality, a
deterministic analysis, no matter how advanced, might fail completely to predict safely such
maximum load effects. This assertion is illustrated below with the help of the results from the
test on the Belle River bridge.

Transverse profiles of the distribution factors for mid-span girder moments in the bridge under
consideration, are plotted in Fig. 7 for a transversely symmetrical load case. One of these profiles
corresponds to moments computed from observed girder strains both at the mid-span and near the
abutments, with the latter providing information regarding the bearing restraint forces. The other
two transverse profiles are obtained from the results of the semicontinuum method of analysis [4]
for the two bounds of the composite action. It is noted that no attempt was made to model the
bearing restraint in these analyses.

It can be seen in Fig. 7 that the pattern of transverse distribution of actual moments is similar, but
only in a general way, to the two analytical patterns. It is also quite random. Unlike the
analytical patterns, the actual pattern is far from being symmetrical. In fact, the actual distribution
factor for maximum girder moments is about 10% larger than the corresponding analytical factor
for the fully composite bridge. It can be appreciated that the occurrence of the very high
distribution factor and significant departure from symmetry are probably caused by the middle
girder becoming accidentally much stiffer through composite action by bond than the adjoining
girders. In light of the results plotted in Fig. 7, there can be little doubt that, for the kind of

CD (0

Ö
OT

E
2.5

2.0
I I I I
Moments computed
from observed data III
Semicontinuum analysis
for composite girders
Semicontinuum analysis for
non-composite girders 400
I Im
I I I II
1.5 300
O <d
200
xa o 1.0
100
0.5
fc -100
0.0 -200 —
Transverse position

Figure 7/ Distribution factors for midspan Figure 8/ Bearing restraint forces in the
moments in the Belle River bridge Belle River bridge
 B. BAKHT - L.G. JAEGER 269

bridge under consideration, even the most rigorous deterministic analysis is at best only a fairly
close approximation.

Bearing restraint forces in the girders of the Belle River bridge were computed from observed girder
strains near the abutments. From these bearing restraint forces and approximately-calculated girder
reactions at the supports, it was concluded that the effective coefficient of friction varied between
0.66 and 0.95; the former limit relates to loading by single vehicles and the latter to two side-by -
side vehicles. Such effective coefficients of friction may be on the high side but are not
uncommon in bridges in which the girders rest directly on highly rusted steel bearing plates.

Bearing restraint forces computed from measured girder strains are plotted in Fig. 8 for the same
load case for which the distribution factors for mid-span girder moments are plotted in Fig. 7. The
bearing restraint forces are shown as positive when they tend to push the abutment away from the
girders.

It can be seen in Fig. 8 that the bearing restraint forces, in all the girders except one, are positive.
At the location of the left hand outer girder, the bearing restraint force was found to be not only
negative but also fairly large in magnitude. It is postulated that this unusual response is the result
of a relatively soft pocket in the backfill behind the abutment in the vicinity of the left hand outer
girder.

In light of the uncertainties discussed above, it can be seen that for the kind of bridge under
consideration, no deterministic analysis can be expected to predict the actual behaviour of the
bridge.

6. A NEW MEDIUM SPAN COMPOSITE BRIDGE

The examples presented so far in the paper are of relatively short span bridges in which there are no
mechanical shear connectors and in which girders rest either directly on the abutment or on fairly
rusted steel bearing plates. In such bridges, there may be difficulties in assessing the degree of
composite action and the magnitude of bearing restraint forces. Further, because of the spans
being themselves short, even small errors in the estimation of the effective span can have a
relatively large influence on the computed responses of the bridge. Consequently, one might
conclude that the difficulties in predicting the realistic response of a bridge are limited to only the
kinds of bridge discussed earlier. It is shown in the following that errors in predicting bridge
behaviour can also extend to medium span bridges in which mechanical shear connectors ensure
virtually füll composite action and in which the girders are supported by elastomeric bearings
which apparently permit free longitudinal moment of the girders.
14.62 m

r
203 mi
mm 2C03 mm
-i
25 mm
203 mm
JE610 mm
£= 246 mm

^,37m><2.97m 2.97 m 2.97 m ><237n\ ^37 m

Cross-Section
Figure 9/ Cross-section of the North Muskoka River bridge
270 ANALYTICAL AND OBSERVED RESPONSES OF STEEL GIRDER BRIDGES

The cross-section of the single-span North Muskoka River bridge is shown in Fig. 9. This bridge
comprises five steel girders and a composite deck slab; its span and width are 45.7 m and 14.6 m,
respectively. Both ends of every girder rest on laminated elastomeric bearings each measuring 560
x 335 mm in plan and 64 mm in thickness. The design shear rate for each bearing is about 30
kN/mm.

A dynamic test showed the North Muskoka river bridge to be about 20% stiffer flexurally than
could be rationalized by even a very detailed analysis in which all those components of the bridge
were taken into aecount which could conceiva-bly enhance the flexural rigidity of the bridge. To
determine the cause for the apparent discrepancy, a diagnostic static test was condueted
subsequently. For this latter test, all the girders were instrumented with strain measuring devices
to measure longitudinal strains at three transverse sections of the bridge, one section being near the
mid-span and each of the other two near each abutment [8].

If the elastomeric bearings had permitted a free longitudinal movement of the girders, then under
live loads the strains in the bottom flanges near the bearings would have been tensile and very
small. It was found that this was not the case. The test loads induced fairly large compressive
strains in the bottom flanges near the elastomeric bearings. Bearing restraint forces computed
approximately from observed strains are plotted in Fig. 10 for different load cases. It is interesting
to note that under transversely symmetrical loads, the corresponding bearing restraint forces were
not exactly the mirror image of each other, as should have been the case for an ideally symmetrical
structure. Bearing restraint forces as high as 175 kN, which can be seen in Fig. 10, are
considerably larger than a funetioning elastomeric bearing would be expected to develop.
Nevertheless, such large forces were really present despite the fact that the bearings were apparently
in excellent and funetioning condition.

A further proof of the presence of large bearing restraint forces inthe North Muskoka river bridge
was provided by comparisons of applied moments obtained from considerations of simple supports
with those computed from girder strains. Figure 11 shows the comparison of mid-span girder
moments computed from measured strains with those obtained by the familiär grillage analogy
method. The bearing restraint forces were not aecounted for in this analysis. It can be readily
concluded from this figure that the total moment sustained by all the girders is noticeably less than
the corresponding applied moment obtained on the basis of simple supports; this confirms that the
applied moments were reduced by the effect of bearing restraint.

It was found that at the time of the test the bearing restraint in the North Muskoka river bridge
reduced the mid-span deflections due to test loads by about 12%. This reduction is considerably
smaller than the 20% reduction observed in the previous test on the same bridge. The first test
was condueted on a relatively cool day in October and the second on a very hot day in June. It is
hypothesized that the elastomeric bearings had become stiffer in the cold temperature when the first
test was condueted thereby generating higher restraint forces which consequently caused the bridge
to become effectively stiffer than it was at the time of the second test.

Results of tests on the North Muskoka river bridge demonstrate the significant influence of the
restraining effects of elastomeric bearing which may change with load level and temperature. To
be able to analyze bridges with these bearings more accurately, it is essential to include their
effective shear stiffness in the mathematical model.
 B. BAKHT - L.G. JAEGER 271

0* Line 4
Q_QJ
f ^ Line 1

[i n une5
j
11111
150

100 100

50 .*^s
'S 50 "
y.i
fc*. *V*is,&'
Transverse girder position Transverse girder position
Figure 10/ Bearing restraint forces in the North Muskoka River bridge

Line 4 f~> f-J

liii
Linel f—fr n

1
Line 4

£
z
3000

Linel
2000
I from
1000 :=> Computed
measured strains

Grillage analysis
(Without accounting for
bearing restraint)
Transverse Position
Figure 11/ Mid-span girder moments in the North Muskoka River bridge

7. CONCLUSIONS

The purpose of this paper is not to discredit the rigorous methods of analysis, but to note that
there are certain unquantifiable behavioral aspects in bridges which cannot be accounted for
realistically in a mathematical model. Because of this difficulty, the predictions of even highly
rigorous and very accurate methods of analysis may not reflect reality. This contention has been
illustrated with the help of results obtained from tests on five short or medium span bridges with
steel girders. It is suggested that in some cases, a realistic evaluation of the load carrying capacity
of an existing bridge can be condueted only through a field test.
272 ANALYTICAL AND OBSERVED RESPONSES OF STEEL GIRDER BRIDGES

REFERENCES

1. Hendry, A.W. and Jaeger, L.G. The Analysis of Grid Frameworks and Related Structures.
Prentice Hall, Englewood Cliffs, N.J., USA, 1958.

2. Jaeger, L.G. Elementary Theory of Elastic Plates. Pergamon, Oxford, England, 1964.
3. Bakht B. and Jaeger, L.G. Bridge Analysis Simplified. McGraw Hill, New York, USA,
1985.

4. Jaeger, L.G. and Bakht, B. Bridge Analysis by Microcomputer. McGraw Hill, New York,
USA, 1989.

5. Cheung, M.S., Li, W. and Jaeger, L.G. Finite Strip Analysis of Bridges. Book in
preparation.

6. Bakht, B. and Csagoly, P.F. Diagnostic Testing of a Bridge. ASCE Journal of the
Structural Division, 106 (7), July, 1990.

7. Bakht, B. and Jaeger, L.G. Behaviour and Evaluation of Pin-connected Steel Truss Bridges.

Canadian Journal of Civil Engineering, 14(3), 1987.

8. Bakht, B. and Jaeger, L.G. Observed Behaviour of a New Medium Span Slab-on-Girder
Bridge. Journal ofthe Institution of Engineers (India), January, 1990.

9. Bakht, B. and Jaeger, L.G. Bridge Testing - A Surprise Every Time. ASCE Journal of
Structural Engineering, 116(5), May 1990.

10. Bakht, B. and Jaeger, L.G. Ultimate Load Test on a Slab-on-Girder Bridge. ASCE Journal of

Structural Engineering, 118(6), June, 1992.

11. Bakht, B. and Mufti, A.A. Behaviour of a Steel Girder Bridge with Timber Decking.
Structures Research Report SRR-92-02. Ministry of Transportation of Ontario, Canada,
March, 1992.

12. Bakht, B. and Mufti, A.A. Evaluation by Testing of a Bridge with Girders, Floor Beams and
Stringers. Structures Research Report SRR-91-05, Ministry of Transportation of Ontario,
Canada, April, 1992.
13. Bakht, B. Testing of an Old Short Span Slab-on-Girder Bridge. Structures Research Report
SRR-88-01, Ministry of Transportation of Ontario, Canada, June, 1988.
 273

SELECTED PAPERS
274
 Reliability-Based Evaluatio
Appr6ciation de la fiabilitö
Zuverlässigkeitsuntersuchun

Andrzej S. NOWAK
Professor
Univ. of Michigan
Ann Arbor, MI, USA

_,,/
276 RELIABILITY-BASED EVALUATION OF EXISTING BRIDGES

1. INTRODUCTION

Bridge evaluation is an increasingly important topic in the effort to deal with

the deteriorating infrastructure. There is a need for accurate and inexpensive
methods to determine the actual strength of the bridge and the actual load
spectrum. The major factors that have contributed to the present Situation are:
the age, inadequate maintenance, increasing load spectra and environmental
contamination [3,7], The deficient bridges are posted, repaired or replaced.
The disposition of bridges involves clear economical and safety implications. To
avoid high costs of replacement or repair, the evaluation must accurately reveal
the present load carrying capacity of the structure and predict loads.
The major parameters which affect the structural Performance are loads
and resistance (load carrying capacity). Bridge loads include dead load (own
weight of the structural and non structural components), live load (weight of
trucks), dynamic load (dynamic effects of moving trucks), environmental loads
(wind, earthquake, temperature) and special forces (collisions, emergency
braking). Resistance is determined by material properties, dimensions and
geometry, and it depends on the method of analysis. Loads and resistance are
random in nature. Their Variation and uncertainty involved in the analysis can
be expressed by Statistical parameters. Knowledge of loads, their magnitude
and frequency of occurrence, can be gained through surveys, field observations,
measurements and Statistical analysis. In this study, bridge load and resistance
modeis are reviewed and a procedure is formulated for evaluation existing
bridges. Structural Performance is measured in terms of the reliability index.

2. BRIDGE LOAD MODEL

The basic load combination for highway bridges is a simultaneous occurrence of

dead load, live load and dynamic load. The load modeis are developed using the
available Statistical data, surveys and other observations. Load components are
treated as random variables. Their Variation is described by two parameters: X
ratio of mean-to-nominal and V coefficient of Variation. Existing bridges are
evaluated for various periods of time, e.g. 1, 5 or 75 years. Therefore, the
extreme values of loads are extrapolated from the available data base. Nominal
values of load components are calculated according to the current AASHTO [1].
Dead load, D, is the gravity load due to the seif weight of the structural and non
structural elements permanently connected to the bridge. Because of different
degrees of Variation, it is convenient to consider four components of D, as
shown in Table 1.
Table 2-1. Statistical Parameters of Dead Load

Component X V
Factory-made members 1.03 0.08
Cast-in-place members 1.05 0.10
Asphalt (mean thickness) 90mm* 0.25
Miscellaneous 1.03-1.05 0.08-0.10
 A.S. NOWAK 277

The need for a reliable truck weight data has been recognized by many bridge
authorities. In this study, the load spectra are determined on the basis of truck
survey data, truck counts and weigh-in-motion measurements. The data
includes truck weights, axle spacings and axle loads. Multiple truck occurrence
(more than one truck on the bridge simultaneously) is determined by special
truck counts. Bridge Performance is affected by moments and shears rather
than gross vehicle weights. Therefore, the surveyed trucks were run over the
influence lines to determine the moments and shears.
The development of live load model for highway bridges is described by Nowak
and Hong [6] and Nowak [5]. The expected maximum live load moments and
shears are evaluated for various time periods. Life time is 75 years for newly
designed bridges, but 1 to 5 years for evaluation of existing structures. The
measured trucks represent a Statistical sample of the total number of trucks
which cross a bridge in 1, 5 or 75 years. Therefore, calculation of the maximum
moment (shear) for longer periods involves extrapolation of the obtained
results.
The maximum effect is calculated by Simulation of the actual traffic. For
multiple occurrence, various truck configurations are considered: in lane and
side-by-side. The analysis indicates that a lane load is governed by a Single
truck up to about 30-36m span. For longer spans two fully correlated trucks
govern. For two lanes, the live load is governed by two fully correlated trucks
(side-by-side), each being about 85% of the mean maximum 75 year truck. The
actual values of the mean maximum moments and shears for various time
periods also depend on traffic volume.
The Statistical parameters of live load moment for various spans and for time
periods 1, 5 and 75 years are presented in Table 2. The nominal value is
calculated as a design moment specified by AASHTO [1], as shown in Fig. 1.
Table 2 Statistical Parameters of Live Load Moment

Time Period
1
year 5 years 75 years
Span (m) X V X V X V

3 1.37 0.15 1.46 0.15 1.65 0.14

12 1.58 0.13 1.64 0.12 1.74 0.11
36 1.90 0.135 1.97 0.12 2.08 0.11
60 1.78 0.14 1.85 0.125 1.96 0.11
278 RELIABILITY-BASED EVALUATION OF EXISTING BRIDGES

(a) Standard HS20 Truck

I 142 kN I 142 kN
35 kN
\ 4.25 m J 4.25 m J
(b) HS20 Lane Loading

80 kN (for moment)
116 kN (for shear)
9.35 kN/m

Fig. 1 AASHTO Standard Truck and Lane Loading [1].

Traditionally, the dynamic load is considered as an equivalent static load. Many

codes, including AASHTO [1], specify the dynamic load as the function of span
length only. However, it has been observed that the dynamic load depends on
dynamic properties of the vehicle, dynamic properties of the bridge, and
pavement roughness. A procedure was developed by Hwang and Nowak [2] to
model the dynamic behavior of girder bridges including the three factors
(pavement roughness, bridge and vehicle). The procedure was used for
extensive simulations. The major parameters considered include the degree of
road roughness, truck type and weight, speed, and structural type of bridge.
The results indicate that the absolute value of the dynamic load component is
almost constant (measured in terms of deflection). Static deflection increases
with truck weight and therefore the dynamic load, as a percent of static live
load, decreases for heavier trucks.
Simulations were carried out for the cases of one truck and two trucks side-by-
side. The results indicate that, on average, dynamic loads do not exceed 15% of
live load for a Single truck and 10% for two trucks. The coefficient of Variation
of dynamic load is about 0.80.

3. RESISTANCE MODELS
The capacity of a bridge depends on the resistance of its components and
connections. The component resistance, R, is determined mostly by material
strength and dimensions. R is a random variable. The causes of uncertainty can
be put into three categories:

- material; strength of material, modulus of elasticity, cracking stress, and

chemical compositum.
- fabrication; geometry, dimensions, and section modulus.
- analysis; approximate method of analysis, idealized stress and strain
distribution model.
The resulting Variation of resistance has been modeled by tests, observations of
existing structures and by engineering judgment. The information is available
 A.S. NOWAK 279

for the basic structural materials and components. However, bridge members
are often made of several materials (composite members) which require special
methods of analysis. Verification of the analytical model may be very expensive
because of the large size of bridge members. Therefore, the resistance modeis
are developed using the available material test data and by numerical
simulations.
In this study, R is considered as a product of the nominal resistance, Rn and
three parameters: strength of material, M, fabrication (dimensions) factor, F,
and analysis (professional) factor, P,
R Rn M F P (1)
The mean value of R, hir, is
niR Rn niM mF mp (2)
and the coefficient of Variation, Vr, is,

VR= (VM2+VF2+Vp2)1/2 (3)

where, ihm, hif, and mp are the means of M, F, sind P, and Vm, Vf, and Vp are
the coefficients of Variation of M, F, and P, respectively.
The Statistical parameters are developed for steel girders, reinforced concrete
T-beams, and prestressed concrete AASHTO-type girders. The results are
presented in Table 3.
Table 3 Statistical Parameters of Resistance

Type of Structure FM P R
X V X V X V
Non-composite steel girders
Moment 1.095 0.075 1.02 0.06 1.12 0.10
Shear 1.12 0.08 1.02 0.07 1.14 0.105
Composite steel girders
Moment 1.07 0.08 1.05 0.06 1.12 0.10
Shear 1.12 0.08 1.02 0.07 1.14 0.105
Reinforced concrete
Moment 1.12 0.12 1.02 0.06 1.14 0.13
Shear 1.13 0.12 1.075 0.10 1.20 0.155
Prestressed concrete
Moment 1.04 0.045 1.01 0.06 1.05 0.075
Shear 1.07 0.10 1.075 0.10 1.15 0.14
4. RELIABILITY ANALYSIS

The available reliability methods are presented in several publications e.g. [4,
8]. In this study the reliability analysis is performed using Rackwitz and Fiessler
procedure. The reliability index, ß, is defined as a function of Pf,
280 RELIABILITY-BASED EVALUATION OF EXISTING BRIDGES .#%

ß - O-I(Pf) (4)

where <D_1 inverse Standard normal distribution function.

There are various procedures available for calculation of ß. These procedures

vary with regard to aecuraey, required input data and Computing costs. The
simplest case involves a linear limit State function. If both R and Q are
independent (in the Statistical sense), normal random variables, then the
reliability index is,
ß (mR - mQ)/(aR2 + aQ2)i/2 (5)

where mR mean of R, mg mean of Q, or Standard deviation of R and oq

Standard deviation of Q.

If both R and Q are lognormal random variables, then b can be approximated by

ß In (mR/mQ)/(VR2 + Vq2)1/2 (6)

where Vr coefficient of Variation of R and Vq coefficient of Variation of Q. A

different formula is needed for larger coefficients of Variation.
If the parameters R an Q are not both normal or lognormal, then the formulas
give only an approximate value of ß. In such a case, the reliability index can be
calculated using Rackwitz and Fiessler procedure, sampling techniques or by
Monte Carlo simulations. Rackwitz and Fiessler [4, 8] developed an iterative
procedure based on normal approximations to non-normal distributions at the
so called design point. The design point is the point of maximum probability
on the failure boundary (limit state function). The procedure has been
programmed and calculations are carried out by the Computer.

5. RELIABILITY INDICES FOR SELECTED BRIDGES

The calculations are performed for a selected set of girder bridges. The
selection was based on material, span, number of girders and girder spacing.
For the selected bridges, moments and shears are calculated due to dead load
components, live load and dynamic load. Nominal (design) values are calculated
using AASHTO [1], The mean maximum values of live load are obtained using
the Statistical parameters given in Table 1 and 2. Resistance is calculated in
terms of the moment carrying capacity. For each case, two values of the
nominal resistance are considered: Ractual. the actual as-built load carrying
capacity and Rmin. the minimum required resistance which satisfies the
AASHTO [1]. In general, Ractual is larger than Rmin- The basic design
requirement is expressed in terms of moments as follows [1],
1.3 D + 2.17 (L+ I) < (|>R (7)
#L A.S. NOWAK 281

where D, L and I are moments due to dead load, live load and impact, R is the
moment carrying capacity, and is the resistance factor,
<|> 1.00 for steel and
<|>

prestressed concrete girders, and 0.90 for reinforced concrete T-beams. The
/
ratio of Ractual Rmin is an indication of overdesign and it is about 1.5 for steel
girders and about 1.1 for prestressed concrete girders.
The selected bridges do not cover a füll ränge of spans and other parameters.
Therefore, additional bridges are designed as a part of this study. The analysis
is focused on girder bridges with spans from 9 to 60 m. Five girder spacings
are considered: 1.2, 1.8, 2.4, 3.0 and 3.6 m. Typical cross sections are assumed.
In all considered cases, the actual resistance, Ractual. is made equal exactly to
Rmin (the sections are neither overdesigned nor underdesigned).

The reliability indices are calculated for girder bridges described by the
representative load components and resistance. The results are presented in
Fig. 2-4 for steel girders, reinforced concrete T-beams and prestressed
concrete girders.

3.6
¦s
^
¦

1 •s 3.0^^
I
S — Z.4^>*^^ ¦
s 1.83^^
s 1.2 s girder spac ng (m) ¦

2 ¦ ¦

u
0 30 Span(m) 60
Fig. 2 Reliability Indices for Steel Girders.
5
¦s 3.6 ¦

s 3.0f^^ ¦

s 2.4g^^
£ s= 1.8^^^
i
¦

2
s 1.2 s girder spac ing (m) ¦

2 ¦ ¦

u
0 30 Span(m) 60
Fig. 3 Reliability Indices for Reinforced Concrete T-beams.

s 3.6 ^jf^
3.0jj^>*
-3L_

r
s
S 2.4*^*^" ¦

3 2 S= l-S^^ ¦
s 1.2 s girder spac ng (m)
s ¦ ¦

0 30 Span (in) 60
Fig. 4 Reliability Indices for Prestressed Concrete Girders.
282 RELIABILITY-BASED EVALUATION OF EXISTING BRIDGES

6. CONCLUSIONS

The reliability index can be used as an objeetive measure of structural

Performance of an existing bridge. The calculation requires the knowledge of
site-speeifie load and resistance parameters. Load parameters can be
determined by truck surveys. Resistance depends on the degree of
deterioration (e.g. corrosion). The Statistical parameters for a general case are
presented in Tables 1-3.
Reliability indices are calculated for typical girder bridges, designed using
AASHTO [1]. The results show a considerable degree of Variation depending on
girder spacing, material and span length. This is a clear indication that the
current design provides a higher safety reserve for larger girder spacings and
spans about 20 m.
ACKNOWLEDGMENTS
The presented research has been carried out in conjunetion with the NCHRP
Project 12-33. The author thanks Professor Palle Thoft-Christensen and Dr.
John M. Kulicki for fruitful discussions on the presented load modeis and
evaluation procedures. A support from the NATO Scientific Affairs Division is
gratefully acknowledged.

REFERENCES
1. AASHTO, "Standard Specifications for Highway Bridges" American
Association of State Highway and Transportation Officials, Washington, D.C.,
1989.
2. HWANG, E.S. and NOWAK, A.S., "Simulation of Dynamic Load for Bridges",
ASCE Journal of Structural Eng., Vol. 117, No. 5, May 1991, pp. 1413-1434.

3. KONIG, G. and NOWAK, A.S., ed., "Bridge Rehabilitation", Ernst & Sohn,
Berlin, Germany, 1992.
4. MELCHERS, R., "Structural Reliability, Analysis and Prediction", Ellis
Horwood Limited, Chichester, 1987.
5. NOWAK, A.S., "Calibration of LRFD Bridge Design Code", Report, Department
of Civil and Env. Eng., University of Michigan, Ann Arbor, Mi, May 1992.
6. NOWAK, A.S. and HONG, Y.K., "Bridge Live Load Models", ASCE Journal of
Structural Engineering, Vol. 117, No. 9, September 1991, pp. 2757-2767.
7. NOWAK, A.S., ed., "Bridge Evaluation, Repair and Rehabilitation", Kluwer
Academic Publishers, Vol. 187, Dordrecht, Netherlands, 1990.
8. THOFT-CHRISTENSEN, P. and BAKER, M.J., "Structural Reliability Theory
and Its Applications", Springer-Verlag, New York 1982.
J%.

Load Testing and Numerical M

Essais de Charge et modeles num
Belastungsversuche und ihre numerisch

Martin TALBOT Constantin HALC

Civil Engineer Civil Engineer
Ministry of Transportation Ministry of Trans
Quebec, PQ, Canada Quebec, PQ, Can

Constantin Halchini
for the mobile
at the Central Labo
at the Mini
Transportation.

Martin Talbot, born in 1959, got

his Bachelor, Master and Ph.D.
degrees in civil engineering at
284 LOAD TESTING AND NUMERICAL MODELLING OF QUEBEC BRIDGES

1. INTRODUCTION
Aging of structures and public budget cuts require more than ever before the exploration of the
remaining structural capacity. This is especially true, in many countries, in the field ot bridge
structures since a major number of these bridges have been build many decades ago. Since then,
usual deterioration and fatigue as well as important increase of traffic and loading intensity have been
observed. Many aspects concerning the remaining bridge capacity are still unknown. For this
reason, the Quebec government has initiated an extensive program of experimental in situ
measurements of bridges showing real or potential structural problems.

The paper presents the experimental results as well as the calibration of a related detailed
tridimensional finite element model for recent specific bridge cases. The experimental part has been
performed here by an efficient mobile laboratory for which strain gauges and accelerometers are
available. A special processing of strain data makes it possible to obtain directly the stress resultants
(axial force, moments) instead of local stresses. These values are then directly comparable to a
simple or more complex finite element model to be calibrated. This calibrated model is then used as
the most realistic model for the more complex or more intensive loading cases to be checked
according to actual design codes [13]. Both static effects as well as dynamic behavior can be taken
into aecount. interesting results for a first bridge have already shown the scientific as well as the
economic benefit of such an approach [1]. New results for two other deck slab steel truss bridges are
now available and will be presented in this paper.
The first objeetive of the paper is to expose the methodology of the experimental approach as well as
the calibration of the finite element model for the two bridges. It also shows how much valuable
information can be readily available in such a combined approach. Finally, the paper also shows how
Standard evaluations of existing bridges may sometimes be inadequate in predicting the exact
remaining capacity. In fact, in situ measurements and refined modelling can yield in a cost reduction
of expected repairs. It also improves the engineer confidence in the making of repair decisions.
2. STANDARD BRIDGE EVALUATION
The Standard evaluation practice of old bridges is usually based on methods which are similar to those
used for new bridges. Standard linear two dimension displacement modeis (rigidity method) are
normally selected. A distributed factor is used to consider the transverse non symmetrical distribution
of the live load. The evaluation is based mainly on the results of an inspection of the bridge which can
show structural problems such as rust, cracks and fatigue in members. The assumptions of this
analysis are mainly the same as in the original design. One of the main differences lies however in
the intensity of the live loads to apply. In fact, truck weights have severely increased for the past
decades. That is why, a few years ago, a Standard truck specifically developed for the Quebec traffic,
the QS660 truck, has been adopted for design or analysis of new and old bridges. This truck has a
total weight of 660 kN with a 60 kN, 240 kN, 200 kN, 160 kN distribution from front to end of the
truck and with related distances of 4 m, 6 m and 6m respectively and a width of 1.8 m. For instance,
for old bridges, like the two bridges presented in this paper, the design live loads were two to three
times less than those of the QS660 truck.

For this reason, these live loads are now often eritieal for the structure. Moreover, the live loads
must naturally be increased to keep into aecount the dynamic effects. This dynamic allowance factor
can be related to many factors as shown in the literature [10]. In the Canadian Bridge Code [13],
however, the rate of this factor is only related here to the first flexural frequency. A maximum value
of 40% increase will be applied for a frequency lying between 2.5 and 4.5 Hz. Otherwise, this value
will be at least 20%. For steel truss bridges, this factor is often at its maximum value.
If the evaluation shows an acceptable resistance to the loadings in the Norm, it is considered as being
over. If the bridge shows too much weakness, it might be replaced, repaired or limited to lighter
vehicles. Sometimes, however, the results of the analysis lies somewhere in a "grey zone". This
means that the analyst has a good presumption that the Standard evaluation is probably too
conservative due to approximations such as the two dimension modelling or as neglecting the rigidity
of some components. Two more check steps are then available as described in the next two sections.
 M. TALBOT - C. HALCHINI - M. SAVARD 285

3. EXPERIMENTAL EVALUATION
The Ministry of transportation of Quebec acquired, in 1990, a mobile laboratory specifically dedicated
to the testing of bridges. Actually, this mobile laboratory has two independent data acquisition
Systems. One for static tests that includes 60 Channels for strain gauges and 20 for other sensors
such as LVDT or full-bridge gauges. The dynamic test System embodies 12 Channels for
accelerometers and 8 Channels for strain gauges. As many as 8 pressure tubes can be used to evaluate
the total time of the forced vibrations and the averaged speed of the vehicles over the bridge. Up to
50000 samples per second can be recorded from each instruments. The laboratory also includes a
micro Computer, a color plotter and two printers. A 5000 watts diesel generator is the main power
supply of the vehicle. Test trucks used to load and excite the bridge are Standard 10 wheel trucks and
weight around 260 kN (250 kN for Armagh and 265 kN for Sainte-Marie). Because of the Canadian
weather, all tests are performed from late spring to mid fall. It is impossible to give here all the
details of the experimental results so that it has been decided to present specific results for each of the
two bridges. The Armagh bridge has been selected to present the static approach while the the Sainte-
Marie bridge will show the dynamic results (see the corresponding subsections 6.1 and 6.2).
4. FINITE ELEMENT EVALUATION
Once the in-situ tests have been performed and the experimental values postprocessed, a more refined
numerical model is build. A commercial program GIFTS (CASA) has been used here on a IBM PC
386. Models are kept moderately large with about 1000 degrees of freedom. The model is kept
linear and truss, beam, spring and quadrilateral shell elements are used. A diagonal mass matrix has
been selected for Computing the eigenmodes via the subspace iteration algorithm. The basic
assumption here is that experimental results make the best information available about the structure in
spite of normal experimental inaccuracies. Since a first finite element model never fits exactly
experimental values, this one has to be calibrated by an iterative process in order to get closer to the
experimental data. Once the model has been locally calibrated (local forces in specific member) or
more globally calibrated (via dynamic modal values), this one is believed to be adequate in all its
tested and non tested members.

There is no Standard recipe or Standard Steps for a successful calibration. The choice of the
components to be calibrated depends on many factors like the kind of structures (steel or concrete, for
instance) and the results of the in-situ inspection (actual State of some components). For the kind of
bridges which are studied in this paper two main factors were selected. The longitudinal restraints of
the supports have been modified to take into aecount the partial restraint of the theoretically free
supports. Spring elements have been tested to model this partial restraint. Some artificial beam
elements have also been used to link the deck to the trusses in order to modulate the partieipation of
this deck. This appeared to be the most important step. To a lower extent, we also checked the
influence of the Young modulus of concrete. A slight dynamic calibration of the mass of the deck
(pavement thickness) has also been tested. Finally, once the model has been calibrated, we assume a
linear behavior of the structure.Loadings according to the Canadian Code are then performed with
strictly localized QS660 trucks and maxima efforts are extracted. These efforts are considered to be
the "real" efforts that would appear in the structure in its actual state.

5. DESCRIPTION OF THE TWO BRIDGES

gtl The Armasfr Bringe
The Armagh bridge is a deck slab steel truss bridge which is located on the small river near Saint-
Cajetan d'Armagh (Fig. 1). It was build in 1931 and is 51.2m long and 7.3m large. This average size
structure has three spans which are 11.9m, 27.4m, 11.9m long with the central one with steel
trusses. The thickness of the concrete slab is 0.178m and is 0.065m for the pavement and no specific
composite action is present (in theory only). For this structure, steel and concrete Young modulus are
200000 MPa and 250 Mpa. The total number of elements for the Standard 2-dimensional truss model
is 33. Practically all structural components have been discretized for the 3-dimensional model. These
include trusses, stringers, floor beams, the bracings and the concrete slab (Fig. 3). The total number
of elements of all kinds is 415 while the total number of degrees of freedom is 976.
286 LOAD TESTING AND NUMERICAL MODELLING OF QUEBEC BRIDGES

5.2 The Sainte-Marie Bridge

The Sainte-Marie bridge is also a deck slab steel truss bridge which is located on the rather large
"Chaudiere" river in the Sainte-Marie Town (Fig. 2). It is an important bridge in the area. It was
build in 1918 and is 130m long and 6.6m large. This structure has two 65m identical spans. Only
one steel truss span has been studied. The thickness of the concrete slab is 0.178m and is 0.065m for
the pavement and no specific composite action is present in theory. The total number of elements for
the Standard 2-dimensional truss model is 54. Practically all components have been discretized for
the 3-dimensional model that is trusses, stringers, floor beams, the bracings and the slab (Fig. 4).
The total number of elements is 523 while the total number of degrees of freedom is 1246.

6. RESULTS
<?tl The Armagh Bringe
As mentioned previously, the Armagh bridge has been specifically selected here to present the static
results. Six steel members have been instrumented with a total of 40 strain gauges. These members
are those identified with numbers on Fig. 1. Each member had at least 3 or 4 gauges in order to
determine all internal efforts developed in the member (axial force N; bending moments Mx and My;
warping moment B). Considering that the recorded deformation is the sum of the deformation
produced by each effort, we have [15]:
N Mx My R
yi l
EI« ' (i)
1

EA EIX EIy
where E is the Young modulus, A is the cross-section area, Ix and ly are the moment of inertia with
respect to the x and y axis, xi, yi and wi are the x, y and sectorial coordinates of the strain gauge.
Having a similar equation for each strain gauge i of a member, this sets up a System of equations easy
to resolve for the four unknown N, Mx, My and Mq).

The tests took place on August 1991, during which 8 load paths from A to H (from one sidewalk to
the other) have been followed along the bridge. For each path, 12 positions of the 27 tonne 2 axle
truck have been occupied (see Fig 1). The truck positions were selected as a function of the second
row of wheels being placed directly over each vertical member of the truss. Since a simultaneous
calibration of all tested members at any truck position was not possible in practice, the direction of the
calibration was dictated by the values of a very few selected members for selected truck positions.
Once these values had a satisfactory convergence towards the experimental results, graphics of the
axial force in all members as a function of truck positions were plotted. Two examples of these
results are shown in Fig. 5 and 6. The Fig. 5 shows the best results while the Fig. 6 shows the worst
one which are actually rather good too. The precision of all other results lie between these two cases.
Moreover, some dynamic data have been obtained from a few gauges and one accelerometer. The
FFT gave us the first flexural frequency which value was 4.10 Hz. As a final check, we computed
this value for the calibrated model and obtained 3.83 Hz which is quite satisfactory.

The calibrated model was then ready for its maximum loading according to the Canadian Code. The
maximum forces induced by two QS660 trucks (one at 100% and one at 70%) were computed. A
significant reduction of 55% to 170% in the load effects was observed compared to the Standard
analysis for the horizontal top and bottom chords, reducing by almost the half the number of these
members to be replaced. Diagonal members improved slightly but appeared to remain weak and will
have to be replaced. The vertical members at both extremities of the steel truss appeared to be more
loaded in the refined model (by 34%). This shows an interesting example in which the Standard
approach is here less conservative and then less secure than its more refined counterpart.

6.2 The Sainte-Marie Bridge

The dynamic tests of the Ste-Marie bridge were performed in April 1992. For these tests, 8 low
frequency accelerometers have been used, 3 on both sides of the deck for vertical movements
measured and 2 for horizontal movements at the abutment and at mid-span, as shown in Fig. 2. One
pressure tube was placed at each end of the span. A total of 10 runs have been made at different
speeds along the center of the roadway or on the right lane side. The maximum speed was limited to
40 km/h because of the very steep hill at the entry of the bridge.
 M. TALBOT - C. HALCHINI - M. SAVARD 287

12 3

l
4 8 9 10 11 12

^
I
I

r
I

**/ \*6

Fig. 1 Armagh Bridge

2 4 6 0 10 12 14 16 10 20 22 24 28 28

Fig. 2 Sainte-Marie Bridge

Fig. 3 Numerical Armagh Fig. 4 Numerical Sainte-Marie

Bridge model Bridge model

50
3-D Finite element 285 —O— 3-D Finite element

Experimental 235 —D— Experimental

¦
>s.

185

i
73
¦3 - -
4ä
'-50- 85- *
vi *¦ *

35 - -
y -}• < >

Q^O^Jrr
-100- -15 1
1 1 1 1 1
i

12 4 8 12
Longitudinal position Longitudinal position
Fig. 5 Axial force in member # 3 Fig. 6 Axial force in member # 6
288 LOAD TESTING AND NUMERICAL MODELLING OF QUEBEC BRIDGES

In order to evaluate the frequencies of Vibration and the corresponding mode shapes included in the
response of a bridge under traffic loading, the response of one accelerometer placed away from
eventual nodal points is considered as the excitation in the modal analysis. The five other
accelerometers measured the Outputs. This approach is preferred since the excitation force, which is
the Variation of the load intensity with time, is not measured.

The computation of the FFT of these 6 Signals displays how the energy is distributed in the frequency
domain. High peaks usually correspond to the bridge frequencies of Vibration. However, during the
forced Vibration, the frequencies caused by the bridge-vehicle interaction may be included in the
spectra [2]. To avoid this potential confusion, the coherence between the excitation signal and the
Outputs was also computed. The coherence indicates the degree of linearity between the input and the
output [10], [6], [7]. Since the amplitude and phase of all points follow a fixed relationship when the
bridge vibrates in one of its modes, the coherence should be over 0,9 at mode frequencies of the
bridge, and low at other frequencies such as the frequencies of the bridge-vehicle interaction [12].
One may also compute the FFT of each signal for each runs and take the average FFT in the
subsequent calculations. This may eliminate noise and some irrelevant peaks. To identify
experimental mode shapes, the amplitude of the imaginary part of the transfer function FRF and the
phase of the same function give the relative displacements and phases of other accelerometers relative
to the reference accelerometer at all frequencies [10]. The procedure described above has been
followed for the 8 dynamic runs and the results for accelerometer 2 are presented in Fig. 7. The
experimental operating mode shapes are also presented in Fig. 8. One should note a lack of symmetry
of three mode shapes, probably due to the short duration of each free Vibration records.
The static calibration technique was exactly the same as for the Armagh bridge. Then, a final tuning
of model was imposed in order to get closer to the first flexural frequency. A slight modification of
the mass of the slab (for pavement) improved the values of the four eigenvalues. These numerical
eigenmodes are shown in Fig. 8. A good correlation is observed between the experimental and the
numerical values. The model showed however that the experience failed to give some higher flexural
modes with lower values than the fourth one but this was due to the small number of accelerometers.

7. CONCLUSIONS
7.1 Summarv of the results
In this paper, we presented practical results for two old bridges. The paper showed how experimental
results from a mobile laboratory can help to build a well calibrated numerical model. Once calibrated,
this model can then give realistic structural values in any part of the structures. The postprocessing of
the experimental dynamic data can give the eigenvalues as well as the eigenmodes which can be
compared to the numerical model. These values show how Standard evaluations are sometimes
inadequate. They are often too conservative and, in this case, the combined experimental-numerical
approach can save money. They are sometimes (more rarely) not enough conservative and in this
case are less secure. Consequently, this combined experimental-numerical approach is certainly going
to become a must for the analysis of complex or doubtful aging structures. Final decisions could then
be made with a high degree of confidence, allowing engineers to make the proper economic and
safety decisions.

7.2 Future considerations

It should be noted that since complete tests for a bridge may request between one to two weeks, the
testing program is usually limited to about 10 bridges per season. This kind of program should
remain a current practice for the next few years. After this, rapid testing (one or two days per bridge)
based only on dynamical excitations might be sufficient for typical or simple bridges. Calibration of
the finite element method would then be based only on the dynamic tests. This would avoid the very
time consuming strain gauges set up and would partially reduce the weather limitations of these
gauges. It would also require the systematic use of the modal synthesis techniques. For the moment,
our limited number of accelerometers could still be a problem for more complex structures. For some
evaluations based on precise investigations of stress components such as the axial force of certain
members, strain gauges will remain an essential tool.
 M. TALBOT - C. HALCHINI - M. SAVARD 289

+ 1.60
*ec 2

-1.60
—wJ||y
pjb^—^—-—^-
1.0 4 3.5 +8.0 + 12.5 + 17.0
llmt (t)

Fig. 7 Accelerations of accelerometer 2

X
2.73 Hz ; Flexural 2.73 Hz ; Flexural

--«cc3 8 0

occ5 5 0

accZ -1 0
- - acd -3 5

4.66 Hz ; Torsional 4.60 Hz ; Torsional

«CC4 0
0....
"C^C^^ ^KXXXX^,
5.90 Hz ; Flexural 5.73 Hz ; Flexural

^
•cc3 13 -- -
^SZ

9.76 Hz ; Torsional 10.12 Hz ; Torsional

Experimental modes Numerical modes

Fig. 8 Experimental and numerical Vibration modes

290 LOAD TESTING AND NUMERICAL MODELLING OF QUEBEC BRIDGES

On a more general basis, many aspects concerning old bridge structures are still to be clarified,
however. This is why an important scientific collaboration has been settled in order to improve our
knowledge of some complex experimental aspects of different kinds of old bridge structures as well
as some specific bridge components. This has already led to many interesting results such as the
experimental determination ofthe dynamic amplification factor [6], [7],[10], the ultimate capacity of
noncomposite concrete slab on steel girder bridge [8] or the in-situ study of a prestressed bridge [11].
This collaboration has also led, much in the same way, to many improvements and developments of
the "domestic" finite element tools to be used soon for bridge modelling. Fundamental algorithmic
aspects for dynamic analysis [3] and new thick/thin shell elements for concrete structures [5],[14]
have been developed as well as non linear analysis of noncomposite effects in slab [9]. A finite
element program written in C (called CLE) is under development for direct bridge-structure
interactions [2]. All these developments will be necessary to yield in a better understanding ofthe
behavior of the increasing number of old bridge structures.

ACKNOWLEDGMENTS
The authors would like to thank, Mr. G. Richard, Chief of the Structure Division, Mr. C. Tremblay,
Director of the Expertise Division and Mr. A. Ares, Chief of the Central Laboratory for their valuable
supports. We also thank Mr. G. Ouellet and F. Brisson who were in Charge of the technical aspects.

REFERENCES
1. MASSICOTTE, B., HALCHINI, C, LABBE, J., FAFARD,M. " Load Carrying Capacity of the
Massawippi River Bridge", Proceedings of the Annual Conference of the Canadian Society for
Civil Engineering, Vancouver, Volume 3, pp. 235-244, May 1991.
2. SAVARD, M. 1992. Etüde par elements finis du facteur d'amplification dynamique des charges
dans les ponts sollicites par des vehicules routiers. PhD Thesis, Laval University, Quebec.
3. TALBOT, M., " Analyse dynamique lineaire et non lineaire des structures elancees et son
application aux structures immergees", Ph.D. Thesis, Laval University, Canada, 1990.
4. DHATT, G., TOUZOT, G., "The Finite Element Displayed", Wiley and Sons,1986
5. BATOZ,J.L.,DHATT,G.,Modelisation des structures par elements finis,Vol.3,Ed.Hermes, 1991
6. PROULX, J., PAULTRE, P., "Dynamic Testing of a Composite Arch Bridge", Proc. Ann.
Conf. ofthe Can. Soc. for Civ. Eng., Vancouver, Volume 2, pp.21-30, May 1991.
7. PROULX, J., HEBERT, D., PAULTRE, P., Second Series of Dynamic Testing of the Milnikek
Bridge, Proc. Ann. Conf. ofthe Can. Soc. for Civ. Eng, Quebec, Vol 1, pp.21-30, May 1992.
8. DIONNE, G., BEAULIEU, D., PICARD, A., FAFARD, M., "Capacite ultime des ponts en
acier avec dalle de beton non participante: etude experimentale", Proc. Ann. Conf. of the Can.
Soc. for Civ. Eng, Quebec, Volume 1, pp.223-232, May 1992.
9. LIN, J., FAFARD, M., BEAULIEU, D., "Ultimate Capacity of Noncomposite Concrete Slab on
Steel Girder Bridges: numerical Study", Proc. Ann. Conf. of the Can. Soc. for Civ. Eng,
Quebec, Volume 1, pp.233-242, May 1992.
10.PAULTRE, P., PROULX, J., THIBODEAU, L. Dynamic Testing of Milnikek Bridge, Research
Report SMS-91/04, University of Sherbrooke, Sherbrooke, Quebec, 1991 (In French).
11.MASSICOTTE, B., PICARD, A., TREMBLAY, S., "Instrumentation du^pont de Grand-Mere
lors de son renforcement", Proc. Ann. Conf. ofthe Can. Soc. for Civ. Eng, Quebec, Volume 1,
pp.373-382, May 1992.
12.BILLING, J. R. 1982. Dynamic Test of Bridges in Ontario, 1980: Data Capture, Test
Procedures and Data Processing. Report SRR-82-02, Research and Development Branch,
Ministry of Transportation and Communications, Ontario, Canada.
13. CANADIAN STANDARD ASSOCIATION, 1988. "Design of highway bridges", CAN/CSA-
S6-M88. Rexdale, Ontario.
14.TALBOT, M., DHATT, G., HENCHI, K., "Un nouvel element de coque applicable ä l'analyse
3-D de structures de pont", Proc. Ann. Conf. of the Can. Soc. for Civ. Eng, Quebec, Vol. 4,
pp.385-394, May 1992.
15.HALCHINI, C. 1989. Essais extensometriques d'un pont acier-bois: les poutres pnncipales.
Report RTQ-89-18, Etudes et recherches en transports, Ministere des transports, Quebec.
 291

Redundancy and the Reserve and Residual Strength of Frames

Redondance, reserve de resistance et capacite restante de cadres
Redundanz, Reserve- und Restfestigkeit von Rahmen

Colin J. BILLINGTON Helen M. BOLT J. Keith WARD

Director Supervising Engineer Associate
Billington Osborne-Moss Billington Osborne-Moss Billington Osborne-Moss
Maidenhead, UK Maidenhead, UK Maidenhead, UK

Director and co-founder of BO-
MEL, Colin Billington has been
responsible at senior management
levels for establishing and develo-
ping R&D organisations including
the Steel Construction Institute.
Since her academic training at
Imperial College, Helen Bolt has
been involved with development
of the reserve and residual
strength technology for offshore
jacket structures. She was the
Keith Ward's expertise is in the
analysis of complex problems in
offshore engineering. He led the
development of the nonlinear
Programme described in this
paper, which is now used to determine

His contributions to structural
engineering technology and practice
are widely recognised, with
more than 80 publications to his
credit.
engineer on site for the experimental
work described in this paper.
As Supervising Engineer she now
takes overall responsibility for
R&D activity.
the reserve and residual
strength of jacket structures. As
Associate he leads the company's
analysis and Software development
activities.

SUMMARY
For undamaged structures the conservatism in conventional design practices ensures that a substantial
reserve strength exists beyond the design event. Similarly, the redundancy of the structure gives a
residual strength enabling it to sustain load even in a damaged condition. Reserve and residual strength,
redundancy and collapse mechanisms are important considerations in the design and reassessment of
offshore jacket structures. Large scale collapse tests of frames were undertaken and these are described
in the paper with a discussion of the findings and analysis predictions.

R£SUM£
Pour les structures intactes, la pratique prudente des methodes d'etude traditionnelles assure la presence
d'une reserve de resistance au-dela de la resistance nominale. De möme, la redondance de la structure
offre une capacite restante qui lui permet de register a une Charge möme en cas de dommages. La reserve
de resistance et capacite restante, la redondance et les mäcanismes de rupture sont des considerations
importantes dans l'etude et la re-evaluation des structures des chemises des plate-formes en mer. Des
essais de rupture de cadres ont ete effectues sur une grande echelle; ils sont decrits dans l'expose, avec
une discussion des conclusions et des recommandations.

ZUSAMMENFASSUNG
Bei unbeschädigten Bauwerken wird durch den Konservatismus der konventionalen Entwurfspraktiken eine
bedeutende Reservefestigkeit über den Auslegungsfall hinaus gewährleistet. Auf ähnliche Weise wird
durch die Redundanz der Konstruktion eine Restfestigkeit erreicht, wodurch gewährleistet wird, dass sie
auch in einem beschädigten Zustand die Last tragen kann. Reserve- und Restfestigkeit, Redundanz und
Einsturz-mechanismen sind für die Entwurfsgestaltung und Neubewertung von Offshoreplattformen von
grosser Bedeutung. An Rahmen wurden Kollapsgrossversuche ausgeführt, die in dem Schriftstück mit
einer Diskussion der Ergebnisse und Analyse voraussagen beschrieben werden.
292 REDUNDANCY AND THE RESERVE AND RESIDUAL STRENGTH OF FRAMES

1. INTRODUCTION

Traditionally offshore steel jackets are designed on a component by component

basis. Each member and Joint is checked against the strengths given in design
codes, which themselves have been established from the database of isolated Joint
and tubulär beam-column test results.
In practice a structure's ability to resist loads in excess of the design load
(the Reserve strength') depends not only on the conservatism in the design of
individual members and joints, but also on the Performance of these components
within the frame. Thus the ratios of the collapse load to the design load derived
from the eritieal component, may differ even between structures which have been
designed to the same code. Influences of the overall system response on the
ultimate capacity of frames, are not generally aecounted for in the design of
offshore jacket structures since no specific guidance is given within the codes.
However, these are being recognised increasingly as important factors in the
selection of new platform configurations and in the reassessment of existing
installations.
Beyond the ultimate state, be it brought about by accidental damage or over-
loading, there is an additional requirement for the structure to remain intact,
redistributing the loads safely without catastrophic collapse. The ability of a
structure to sustain damage in this way, its Residual strength7« is quantified
by the ratio of the collapse loads for the damaged and intact structures and
depends largely on the structural redundancy within the system. Again, in
traditional engineering practice, residual capacity is not considered explicitly
in design. However, experiences offshore of changing operational requirements and
instances of damage, have emphasised the importance of building redundancy into
jacket structures.
If reserve and beresidual strengths either cannot be be
quantified justifiably or are
considered to inadequate, the consequences can costly in terms of repairs
or strengthening measures which may in fact be unnecessary or may be carried out
at greater risk in inclement weather conditions. With a better understanding of
the issues and more prudent designs in the future, it may be that fewer offshore
modifications will be required. More detailed description of the sources of
reserve and residual strength and early experimental investigations are given in
the companion paper presented by the authors at the Inspection, Repair and
Maintenance Conference in Aberdeen in 1988 [1].
The ability to understand and quantify the influences on the reserve and residual
strength of frames, their redundancy and collapse mechanisms is now an increasing
demand of the offshore industry. This is required both for the reassessment of
existing structures and for the selection of configurations for new structures.
In recognition of both the technical and economic benefits to be gained for the
offshore industry, the Joint Industry Tubulär Frames Project was established in
1987 to address some of the many questions then arising and to develop a
calibrated technique for reserve and residual strength calculations. Sponsored
by nine offshore Operators and the UK Department of Energy, the Project marks a
significant advance in the application of the reserve and residual strength
technology to offshore jacket structures.
Through experiments the project has given unparalleled examples of the ultimate
response of frames related to both member and Joint failures and has illustrated
the important role of redundancy. The frames were the largest ever to have been
pushed to collapse in a controlled manner. Coupled with the development of a new
nonlinear program, SAFJAC, the Project gave the Participating Organisations a
significant advancement in their ability to understand and predict the behaviour
of both planned and existing installations.
 CJ. BILLINGTON - H.M. BOLT - J.K. WARD 293

2. THE JOINT INDUSTRY TUBULÄR FRAMES PROJECT

Although early studies had emphasised the role of reserve and residual strength
in determining the ultimate capacity of frames, they did not supply sufficient
information or a calibrated numerical tool to assess offshore jacket structures
of current concern in the North Sea. The Tubulär Frames Project was therefore
established [2]. The first
phase of this Joint industry project commenced in
November 1987 and was completed in January 1990. The confidentiality period for
the Phase I work expires in 1993 and the objectives, scope and findings from the
work are described in these sections. In the following section Phase
commenced in June 1990, is introduced.
which II,

2.1 Obiectives of the Frames Proiect

The overall objectives of the project may be summarised as follows:
to establish the effects of non-linear joint/member behaviour on frame
behaviour and collapse mechanisms.
to quantify the reserve and residual strength of frames (global safety
margins) and to investigate redundancy and load shedding characteristics.
to investigate the collapse Performance of members and joints within frames
and to develop procedures for the exploitation of available component data.
to investigate residual strength and load shedding behaviour of a frame which
includes a 'cracked' Joint (Phase I).
to develop a non-linear numerical procedure for the collapse analysis of
frames.

2.2 Experimental Scope - Phase I

Four tubulär frame tests were condueted to investigate the influence of
different modes of failure on reserve and residual strength in Phase I. The
general arrangement of the frames and the test set-up are shown in Figure 1.

.FRAME SUPPORT (TYP)

UPPER JACK ASSEMBLY

TUBULÄR FRAME HINGE UNIT

ly

UPPER TIE BEAM

HINGE UNIT

STRONGBACK

STRONGBACK

RIG SUPPORT (TYP)

Fig. 1 Frame test general arrangement

294 REDUNDANCY AND THE RESERVE AND RESIDUAL STRENGTH OF FRAMES

The frames constitute the largest specimens ever to be tested to collapse in a

controlled manner. General features of the test frames are summarised below:
the four frames resembled prototypes as closely as possible, representing a
scale factor of greater than one-third with respect to typical southern North
Sea jacket structures.
size effects were avoided by adopting minimum tubulär diameters and
thicknesses of 168mm and 4.5mm, respectively.
non-dimensional parameters reflected current offshore practice.
Standard offshore fabrication procedures were adopted.
the influence of boundary conditions was minimised by pinning the frame legs
at the base.
The basic configuration of the two-bay, X-braced test frames, 15m high and 6m
wide, is shown in Figure 2.

Two-bay X-braced frames were selected

for the following reasons:-
X-braced configurations are a
LOAD populär choice for offshore
substruetures.
168 OD X-braced frame behaviour provided
a stringent test for calibration
of the Computer program.
Early generation X-braced jackets
usually did not have Joint cans,
LOAD CELLS
and this Situation of practical
concern was replicated by one
test enabling Joint failure and
168 OD 68 OD load shedding characteristics
within the frame to be studied.
No large scale X-braced frames of
geometries typical of offshore
LOAD CELLS
construction had previously been
tested.
Two-bay frame tests were selected
to enable load shedding between
bay panels, important in the
understanding of frame behaviour,
and redundancy to be studied. X-
braced panels exhibit a
significant residual strength
168 OD 168 OD after failure of one member, and
the presence of a second bay
enables this form of behaviour to
be fully and realistically
captured withoutinterference
from end restraints.

168 OD

Fig.2 X-braced test frame configuration

 4, CJ. BILLINGTON - H.M.

The following features of the four frames

Frames I and III
were dominated by c
frames were nominally identical except
two bays was omitted in Frame
be assessed.
III, en

Frames I and II
were nominally identic
can for the top bay DT Joint in Frame I
precipitated failure.
Frame IV failure was initiated by prop
at the eritieal DT Joint of a specimen
The frames were fabricated in Aberdeen an
self-reacting rig, seen in the diagram in
296 REDUNDANCY AND THE RESERVE AND RESIDUAL STRENGTH OF FRAMES

The analytical workscope began with a review of published data and information to
define load-deflection characteristics of members and joints against which the
existing finite element programs were calibrated. They were then used to develop
new parametric relationships for the specification of Joint stiffness curves in
terms of both peak loads and associated displacements.
The numerical activities within the Project then concentrated on the development
of a Software package with new finite elements encompassing automatic sub-division
to accommodate plasticityand with a non-linear representation of joints [3].
This route was chosento complement commonly available numerical methods which are
finite element based, some of which require several beam elements to monitor the
P-delta effects associated with large deflections. Further, the new Software
contrasts with the phenomenological modeis available which, whilst computational ly
efficient, may be considered to require greater expertise from the analyst.
An elastic quartic element was developed to model each member in a frame. This
was programmed to sub-divide automatically as plasticity occurred, introducing at
its ends either plastic hinge or a new cubic element devised to monitor the
a
spread of plasticity. The facility to model non-linear Joint characteristics was
introduced in the program with a piece-wise linear representation. The new
elements were calibrated in isolation and against frame test results from both the
open literature and the experimental Programme. In Phase II the elements and
automatic subdivision facilities were extended to give füll two and three-
dimensional capability.

2.4 Findings - Phase I

The findings from the experimental Programme have important ramifications for the
repair and maintenance of existing installations as well as for the design of new
structures. In addition to the wealth of information gathered from the associated
tests, the frames themselves demonstrated substantial reserves of strength.
The findings can be summarised as follows:
Significantly larger than expected reserve and residual strengths were
recorded compared with individual component responses.
Frames dominated by Joint failure exhibited greater reserve strengths than
those in which member failure occurred first.
The relationship obtained in respect of interaction between Joint loading-
unloading characteristics and overall frame system response was unexpected for
both intact and fatigue cracked eritieal joints.
For the first time, experimental and numerical evidence had been generated which
seemed to indicate unusual and unexpected frame action effects for tubulär joints.
These findings potentially impact on all aspects of tubulär Joint design practices
(both static and fatigue), from isolated Joint testing procedures, (which may not
be adequately capturing frame effects [4]), to Joint data interpretation, failure
definition, capacity and Joint detailing practices. In recognition of these
unexpected and significant findings, a Phase II project was proposed as an
extension to Phase I, to address these issues within the context of
reserve/residual strength calculations. This is described in the next section.
The role of redundancy in the X bracing was demonstrated by the ultimate frame
response. As the compression Joint in the top bay began to yield, so the
compression load path softened. A greater proportion of the applied load was
therefore distributed via the alternative tension diagonal enabling the structure
as a whole to sustain increasing loads. This response may be contrasted with a
K braced structure where the lack of redundancy through the Joint ensures that
failure of one component constitutes failure of a panel.
 CJ. BILLINGTON - H.M. BOLT - J.K. WARD 297

With regard to member failures, the frame tests enabled reserve and residual
capacities to be quantified. Furthermore, comparison of results from tests I and
III supplied much needed information about the role of horizontal bracing in the
ultimate response of X-braced jacket structures. These two frames were tested
specifically to illustrate the role of redundancy on the system reliability.
Frames I and III were nominally identical but for the absence of the midheight
horizontal (see Figure 2) in the latter case. In elastic design this member
carries no load and with trends towards lighter, liftable jackets designers are
being encouraged to omit these redundant members. The tests showed that although
initial failure in both frames was by buckling of the compression brace in the
upper bay, the post-peak response was severely compromised in the absence of the
horizontal.
In this second case a rapid succession of failures was initiated with a residual
frame capacity below the original design load. As the compression brace buckled,
so a greater proportion of load was transmitted via the tension diagonal. At the
midheight level the only path for the load was the lower bay compression diagonal
which soon buckled. The redundancy afforded by the horizontal in the first
instance however, had ensured a more even redistribution of load without
initiating further component failures. At the midheight level the load from the
top bay tension diagonal divided between the horizontal and lower bay compression
diagonal.
The midheight horizontal constituted just 2.5% of the structural weight yet the
alternative load paths that the redundancy afforded assured a factor of 1.3 on
residual capacity. In terms of safety, the redundancy had a significant
contribution.
Excellent agreement with the frame test results was achieved by SAFJAC analyses
and this has ensured that the program may be used with confidence for the analysis
of offshore jacket structures. Indeed, Participating Organisations are already
performing 2-D and 3-D pushover analyses using SAFJAC as part of their
reassessment of existing installations for re-certification.
In addition to conclusions regarding reassessment procedures, Phase I led to
recommendations for the use of the reserve strength technology in the evaluation
of new designs. The aim is to ensure that minimum operator-specified reserve
strength factors are achieved, thereby enabling structural redundancy to be fully
and safely exploited even though this requirement is not yet specifically
stipulated in design codes. This approach should lead to efficient and versatile
structures for which the need for in-service modifications and/or repairs is much
reduced.
Other recommendations focused on the need for future work to examine the issues
surrounding Joint failure and frame mounted Joint capacities, and for K
configurations to be addressed. It was also acknowledged that SAFJAC should be
developed and extended to enhance its capabilities. The importance of the
findings from Phase I of the Frames Project gave impetus to the commencement of
Phase II along the lines noted above.

3. FRAMES PROJECT PHASE II

3.1 Overview
A Phase II of the Frames Project was developed by the Participants and BOMEL, with
the following objectives:
To establish further the Performance of joints in X-braced and K-braced frames
and to investigate the effects of Joint failures on the Performance of these
frames up to and beyond the ultimate limit state.
To establish levels of reserve strength in X-braced and K-braced frames.
298 REDUNDANCY AND THE RESERVE AND RESIDUAL STRENGTH OF FRAMES

To examine the effect of load reversals on the ultimate response

characteristics of joints.
To undertake lack-of-fit stress measurements.
To enhance, calibrate and apply the non-linear numerical procedure SAFJAC for
the collapse analysis of frames.
Work commenced in June 1990 and was recently completed. Of the frame tests, two
were double-bay X-braced as in Phase I, (Figure 2), and four were of single bay
K-braced configurations as shown in Figure 4.

168 OD

168 OD

LOAD
CELLS
168 OD 168 OD

168 OD

168 OD

Fig, 4 K-braced test frame configuration

3.2 Experimental Programme - Phase II
Taking aecount of the findings from Phase I and recent numerical investigations
into the ultimate response of frame mounted joints, six frames were tested within
Phase II follows:
as
Frame V is nominally identical to the two-bay X-braced Frame II from Phase I
and was tested up to the point of Joint failure to verify the ultimate
capacity of the Joint in the frame.
Frame VI reuses the Frame V structure but with a replacement Joint, carrying
compression in the through member rather than the braces. The test continued
through Joint failure up to the maximum displacement available in the rig,
thereby investigating post-peak response.
Frame VII, the first K-braced frame, is detailed for first failure of the
simple gap K-joint. Frames VIII and X are variations of Frame VII with
different diameter and gap ratios, ß and f. For all the K-braced frames the
tests continued such that both reserve and residual characteristics of the
Joint and frame were revealed.
Frame IX, complements the Frame VII test being nominally identical but for the
speeification of a significant overlap at the K-joint.
Companion tests of nominally identical isolated joints are an important part
of the Phase II Programme enabling the role of constraints within the frame
to be quantified [4].
 CJ. BILLINGTON - H.M. BOLT - J.K. WARD 299

These results, coupled with detailed support monitoring and residual 'locked-in'
stress measurements, have significantly advanced the understanding of the ultimate
response of frames and joints.

4. CONCLUSIONS
From the description of the Joint Industry Tubulär Frames Project presented in
this paper, it
can be seen that the tests are providing new and important
information about the reserve and residual strength of structures. A number of
significant and unexpected findings have been noted. For the first time, frame
tests have been carried out where Joint failure precedes member failure, a
scenario in line with current design practices which dictate an equal likelihood
of Joint failure as of member failure. Frame behaviour has been observed which
impacts on all aspects of tubulär Joint and frame design practices. In
recognition of these unexpected and unanswered findings, a second phase of work
was recently undertaken to develop the technology further. The parallel
development of a calibrated numerical tool, SAFJAC, ensures that the findings can
be directly applied to both planned and existing offshore jacket structures.

An increasing awareness of the need to quantify reserve and residual strength is

being raised for a variety of reasons:
Re-assessment of existing installation is being required more often due to:-
more onerous loading, as environmental conditions are reviewed,
additional topside loading to extend facilities or meet new safety
criteria,
requirements to extend the platform life beyond its design value to
exploit remaining hydrocarbon reserves,
deterioration of capacity through damage or corrosion.
The intention of re-analysis is to assess the fitness for purpose of the
structure under the modified load/resistance regime and to ascertain whether
strengthening measures are essential for safe Operation. In many instances,
if
the combined reserve strength resources could be demonstrated, the need for
costly strengthening measure may be removed.
Cost is also driving the trend towards lightweight structures with few primary
members and a commensurate reduction in reserve and residual strength. The
implications for such structures subjected to design storm loading or
accidental loading can only realistically be assessed through pushover
analyses.
The advent of limit State codes requires a thorough understanding of the
ultimate strength of structures in both the intact and damaged conditions.
Hard evidence is being generated by the frames tests.
The data and information resulting from the Frames Project and the availability
of a calibrated and substantiated non-linear numerical procedure will allow the
safe and economic application of steel jacket structures. It will enable the
proper exploitation of redundancy inherent in jacket structures whilst maintaining
desired safety levels. In addition, it
will provide an important tool in the
reassessment of intact or damaged existing structures, leading to the
identification of global safety levels and the development of inspection,
maintenance and repair (or strengthening) procedures in a more rational, cost-
effective and complete manner than has been possible in the past.
Analytical phases of the work are continuing and a 3D structural collapse test is
planned to investigate load redistribution between structural planes and the role
of redundancy.
300 REDUNDANCY AND THE RESERVE AND RESIDUAL STRENGTH OF FRAMES

5. ACKNOWLEDGEMENT

Phase I of the project was carried out by the authors under the auspices of the
Steel Construction Institute. The Software (SAFJAC), which was developed by
personnel from Imperial College working under the direction of the authors, is
confidential to the Participants.

6. REFERENCES

1. LALANI, M AND BOLT, HM, The role of reserve and residual strength in
determining the repair and maintenance needs of steel jackets. IRM '88
Conference, Aberdeen, 1988.
2. BILLINGTON OSBORNE-MOSS ENGINEERING LIMITED, Joint industry funded research
Programme. Analytical and experimental investigation of the behaviour of
tubulär frames. Phase II.
Invitation to participate. Document Nfl BOMEL
138/89, September 1989 and Document Nft BOMEL 166/90, January 1990.

3. LALANI, M AND SHUTTLEWORTH, EP, The ultimate limit state of offshore platforms
using reserve and residual strength principles. Paper OTC 6309, Offshore
Technology Conference, Texas 1990.
4. CONNELLY, LM AND ZETTLEMOYER, N, Frame behaviour effects on tubulär Joint
capacity. IIW Conference, Finland 1989.
 301

Stress Histogramme and Fatigue Life Evaluation of Highway Bridges

Historgrammes des contraintes et Evaluation de la duröe
de vie ä la fatigue des ponts-routes
Spannungshistogramme und Ermüdungslebensdauer von Autobahnbrücken

Minoru FUJIWARA Kazuhiro NISHIKAWA

Director Head, Bridge Div.
PWRI PWRI
Tsukuba, Japan Tsukuba, Japan

Minoru Fujiwara, born 1942, received Kazuhiro Nishikawa, born 1953,

his M.Eng. at Nagoya University
in 1967. He has been engaged
in bridge engineering works and
road administration since 1967.
received his M.Eng. at Tokyo Institute
of Technology in 1978. He
has been engaged in bridge
engineering works and road
administration since 1978.

Jun MURAKOSHI Takahiro HIROSE

Res. Eng. Struct. Eng.
PWRI Pacific Consultants K.K.
Tsukuba, Japan Tokyo, Japan

Jun Murakoshi, born 1963, received
his M Eng. at Tokyo Institute
of Technology in 1987. He has
been engaged in bridge engineering
works since 1987.
Takahiro Hirose, born 1964, received
B.S. at Kobe University in
1988. He has been engaged in
maintenance and management
works of civil structures since
1988.

SUMMARY
In the case of field measurement of stress histogramme acting on steel bridge members under
traffic, a measuring device to measure stress ranges and their frequencies which is called
"Histogramme recorder", has been used. This paper presents the outline of the histogramme
recorder and its applications to evaluating the condition of steel highway bridges in service.

RESUME
Un dispositif appelä enregistreur d'histogrammes a äte~ däveloppe* pour saisir Involution des
contraintes, tant en amplitude qu'en fräquence, se produisant dans les gläments porteurs de
ponts mgtalliques sous Charge mobile. L'auteur präsente cet appareil ainsi que son application
pour Evaluation de l'6tat des ponts-routes mötalliques.

ZUSAMMENFASSUNG
Zur Aufnahme der Spannungsgeschichte an Bauteilen von Stahlbrücken unter Verkehr wurde ein
sogenannter Histogramm-Rekorder entwickelt, der die Spannungsamplituden und ihre
Auftretenshäufigkeit registriert. Der Beitrag stellt das Gerät und seine Anwendung bei der
Zustandsbewertung von Autobahnbrücken vor.
302 STRESS HISTOGRAMME AND FATIGUE LIFE EVALUATION OF HIGHWAY BRIDGES

1. INTRODUCTION

Total number of highway bridges defined that their lengths are 2 m or over is more
than 650,000 in Japan and the number is increasing still more. Maintenance
technology of existing bridges including inspection, diagnosis, repair and
strengthening methods is most essential to keep them in service for long period.
Safety of bridges is generally evaluated based on various informations such as
damage conditions, traffic conditions, structural characteristics obtained through
inspection or more detailed survey. Stress histogram measurement is one of direct
and effective means to evaluate structural behavior of bridges or their individual
components. In the past, it had been a time -consuming work for preparing and
analysing obtained data when using former measurement devices. Now we have used a
much simpler measurement device, which is called "Histogram recorder", in order to
measure stress ranges and their frequencies acting on bridge members under traffic
for about 10 years. It can obtain stress histogram from measured stress
automatically in field on time.
This paper presents the outline of histogram recorder and its applications for
checking load carrying capacity and durability for fatigue of steel highway bridges
in service.

2. OUTLINE 0F FIELD STRESS HISTOGRAM MEASUREMENT BY HISTOGRAM RECORDER

2.1 Outline of histogram recorder

The histogram recorder used is an device which consists of strain meter, amplifier,
A / D Converter, processor and memories to analyze stress ranges and their
frequencies of histogram by digital process of analog data obtained from strain
gages or other sensors. The histogram recorder we use has 8 Channels to input and
8 data memories. Appearance of devices used for stress measurement is shown in Fig.
1 and Photo 1. The histogram recorder is small in size (200mm(W)X75mm(H)Xl20mm(D))
and light in weight (about 2kg) Histogram analyser in the figure is used only
for initial setting and Operation of histogram recorder, collecting of obtained data.
It contains CRT, printer, and micro floppy disk drive. Obtained data can be
checked in field on time. System disks and data disks are used for analytical

Strain gage AC/DC source

fnl
OOOO
Offline

g OGOO
MA
mn
am
OOOOO
Bridge box l Histogram recorder
Histogram analyser

FlS- 1 System of measuring devices

 M. FUJIWARA - K. NISHIKAWA -

Histogram recorder

[Histogram analyser

Photo 1 Appearance of measuring devices

Number of cycles
10' 105

60 (MPa)
304 STRESS HISTOGRAMME AND FATIGUE LIFE EVALUATION OF HIGHWAY BRIDGES

various effects such as traffic condition and impact by the irregurality of road
surface and truck loading, etc.. These data provide us much information with
respect to load carrying capacity or degree of fatigue damage of bridges in use.
Fig. 2 shows typical stress ränge histograms measured on a bottom flange of span
center of a plate girder bridge during 24-hour period. (a) and (b) are obtained
applying Peak-valley method and Rain -flow method to process measured stress
respectively. In Peak-valley method, number of peak and valley appeared is
recorded. The maximum stress obtained by Peak-valley method can be utilized to
evaluate the load carrying capacity. Rain-flow method processing produces stress
ränge histogram. Stress ränge histogram directly represents the accumulated
fatigue damage of a bridge member during the measuring period. Remaining life for
the fatigue damage can be estimated by applying the modified Miner's cumulative
damage rule for such stress ränge histogram.

3. APPLICATION OF STRESS HISTOGRAM MEASUREMENT

3.1 Evaluation of load carrying capacity of an existing bridge

Almost half of highway bridges in service defined that their lengths are over 15m
or over in Japan are designed by the design live load (DLL) of 14 ton truck or
lighter because Japanese DLL was smaller before 1956. In trunk highway, most of
such bridges have already been replaced or strengthened. However, a great number
of bridges are still
in use. The load carrying capacities of these bridges are
generally evaluated for the present DLL. If
the present DLL is simply applied to
evaluate these bridges, the calculated stress naturally exceed the allowable stress.
But with respect to most of these bridges, it
can be considered a rare case that
heavy trucks comparable to the present DLL come one after another on a bridge and
make lines on it. Therefore, live load for evaluation of these bridges in use can
be reduced less than the present DLL load according to the actual traffic
conditions on the bridges. To clarify the live load, research based on Computer
Simulation of traffic passing through bridge had been condueted in various research
institutions in Japan. This paper proposes a more simple and practical evaluation
method for administrators based on measured stress under traffic. As mentioned
before, measured stress contains the various effects with respect to the structral
characteristics and the traffic condition. Therfore, it
would enable more
rational evaluation of the load carrying capacity of these bridges.
Table 1 shows comparison between the measured maximum
stresses during 24 hours and
calueulated stresses at the span center of main girders of a plate girder bridge by
present DLL. Fig. 3 shows the cross section of the bridge, which was designed by

Table 1 Comparison between measured maximum stress and calueulated stress

Main Measured maximum Calculated stress G a- G D O"
a- G D G a — G D

girder stress Omax(MPa) by DLL a,, (MPa) (MPa) G L G max

G2 45.0 77.4 77.2 1.0 1.7
G3 57.0 92.3 65.6 0.7 1.2
 M. FUJIWARA - K. NISHIKAWA -

12 ton truck. As ono of limit states,

consider condition that residual deformatio
does not remain in bridge components, i.e.
in case of main girder, it does not yield
For example, with respect to a main girde
assuming that the acting stress mos
consists of stress components due to de
and live load, the value of a — a n)
required to be larger than measured maximu
stress due to actual live load as follows.
o-
«- „)/ ff„.,>l.
CT 0

where, o allowable stress for the

steel grade of main girder
n calculated stress for the
CT

dead load is better

öm„„=the measured maximum stress
As shownin table 1, the values of a .- a»
the present design live load, are approxi
girder G3, which mean the safety factors of
other hand, the values of a - a a) I a „«»
respectively, which are both larger than 1.0
are less than calculated stresses for CT

that this bridge can be kept in use to prese

306 STRESS HISTOGRAMME AND FATIGUE LIFE EVALUATION OF HIGHWAY BRIDGES

lead to serious problem quickly.

a 105
However, it
can results in loss 200
of durability of the bridge Before stiffening
itself. The causes of damage 1982.12
are not clear enough so far. cd 84,OOOvehcles/day
But at least, it can be said SO

Q)
150
that these cracks were caused by bO
CS
cd
secondary stresses due to
relative deflection of main
101 105
girders and deflection of RC
floor slabs by wheel load of 100
heavy traffic. The After stiffening
characteristics of damaged Er_ 1986.7
bridges and the effectiveness of 1_ 97,OOOvehcles/day
repair and strengthening method 50
have been investigated by
directly monitoring stress
histogram at the top ends of
vertical stiffners where cracks
0
possibly occur.
Fig.4 Stress ränge histograms measured before
Fig. 4 the results of
shows
and after the stiffening
stress ränge histogram measured
during 24 hours without traffic disturbance before and after stiffening of RC floor
slab with additional stringers between main girders. The stiffening is one of the
strengthening methods for RC floor slab. From this figure, the distribution of
stress histogram can be seen shifted to much lower stress ränge after the stiffening.
The fatigue life after the stiffening is estimated about 80 times as long as before.
It is confirmed that the stiffening is effective for not only stiffening the RC
floor slab but also reducing the local stresses at the cross-beam connection.
This measurement could be carried out without disturbing traffic, therfore is it
effective as one of the direct and easy approach for the investigation of fatigue
damaged bridges.

3.2.2 Investigation of the characteristics of fatigue damaged bridges

With respect to thementioned in Paragraph 3.2.1, to clarify
fatigue cracks
structural chracteristics of damaged bridges, the relationship between structural
parameter and measured local stress at cross-beam connection was investigated for
each different bridge.
The equivalent stress ränge aeq(the root-mean-cube stress ränge), which will cause
the same fatigue damage as the measured stress histogram for equal number of the
cycles, derived from the modified Miner's cumulative damage rule is defined by the
following equation:
aeq=(S ai3- ni/N)1 3

where, g i stress ränge

ni number of stress ränge g i
N number of heavy traffic passing through the bridge during the
 M. FUJIWARA - K. NISHIKAWA - J. MURAKOSHI - T. HIROSE 307

Strain gage [
i) )>))) )y yjy
•O:
: Damaged bridges
Non-damaged bridges
y<

o
Cross beam
20
Measured point
iBefore the stiffening
15

i2/t3
-o u-# s=3
10
-o
-^
o
—o
ö~"0 "
l2/t3 based on the present
design code
-O-
\
After the stiffening
j i 1
1—i—i-J
0.5 1 5

The local stress amplitude parameter A

Fig. 5 Relationship between l2/t3 and local stress
amplitude parameter A for each bridge.

measuring period(=2ni)
Suffix i is counted from maximum stress side.
ränge As one of structural
Parameters of damaged bridges, a local stress amplitude parameter A is defined as
follows:
A G eq/ er eq'
where, g eq and are equivalent stress ranges obtained from measured stresses
a eq'
at the cross-beam connection and bottom flange at the span center of main girder
respectively. a eq is divided by g eq' to reduce influence of the traffic
condition of each bridge.
Fig. 5 shows relationship between l2/ t
and the local stress amplitude parameter
3

A for each bridge. l2/t3 is the same order as a rotational angle of RC floor
slab on main girder web where l spacing of main girders, t =thickness of RC floor
slab. The local stress amplitude parameter A in each bridge is shown by solid
line for the scatter ränge of A at different cross-beam connections. Black dots
means fatigue damaged bridges at cross-beam connection. Also indicated in this
figure, is the comparison before and after the stiffening of RC floor slab with
addtional stringers mentioned in paragraph 3.2.1. By such direct approach, it is
confirmed that occurrence of fatigue crack is influenced by the structural
parameter l2/t3 to a considerable degree.
308 STRESS HISTOGRAMME AND FATIGUE LIFE EVALUATION OF HIGHWAY BRIDGES

The Japanese design code of floor slab had been revised several times for the
RC

purpose of increase of its stiffness from the durability point of view. As shown
in the figure, The values of l2/t3 for the RC floor slab designed by the present
design code are approximately 3.8 to 5.8. These values are less than those of the
fatigue damaged bridges. This indicates that the fatigue cracks are unlikly to
occur at bridges designed by the present design code.

4.CONCLUSIONS
This paper presents the outline of histogram recorder and its applications. We

are also trying to apply the histogram recorder for long time period measurement in
order to monitor condition of bridge members, especially for the change of traffic
condition and propagation of fatigue cracks. Field stress measurement had been
condueted on more than 60 plate girders from fiscal 1986 as a 3-year-project in PWRI.
Obtained data will contribute significantly not only to preparing both durability
and load-carrying-capacity evaluation metohds of existing bridges but also to
introducing fatigue design to future bridge design code.

REFERENCES

1.Bridge Division et al. Research on Duarability Evaluation and Improvement

:

Technology of Existing Highway Bridges, PWRI Technical Report, Vol.2420, pp. 57-98,
1986.11
2.FUJIWARA M., et al. : Durability Evaluation for Fatigue of Existing
NISHIKAWA K.
Plate Girder Bridges, 17th Proceedings of Japan Road Conference, Bridge Section,
pp.796-797, 1987
3.FUJIWARA M., MURAKOSHI J., KASHIMA H. : A Report the Result of Stress
on
Measurement of Steel Highway Bridges, Journal of Structural Engineering, Vol.37A,
pp.1181-1188, 1991.3
 309

Field Testing and Safety Evaluation of Concrete Bridges

Essai sur site et Evaluation de la s6cunt6 des ponts en bEton
Feldtest und Sicherheitsbegutachtung von Betonbrücken

Ayaho MIYAMOTO Hidenori MORIKAWA Kazuyoshi TAKEUCHI

Prof Dr Research Assoc Chief Eng
Kobe Univ Kobe Univ Hyogo Prefect Gov
Kobe Japan Kobe, Japan Kobe, Japan

A Miyamoto, born 1949, received
his Doctor of Engineering Degree
from Kyoto Univ in 1985 His
recent research activities are in
the areas of structural safety as
sessment on concrete bridges and
estabhshment of design concept
for concrete structures under
impact loads
H Monkawa, born 1959, received
his Master of Engineering Degree
from Kobe Univ in 1984 His cur
rent research activities are in
structural safety and reliability
analysis of concrete structures He
is now developing a fuzzy expert
system for structural safety as
sessment of concrete bridges
K Takeuchi, born 1948, received
his Bachelor of Engineering Degree
from Osaka Univ in 1972 He is
attached to the Road Construction
Division and is now involved in
technical investigation on repair,
rehabilitation and replacement
techniques of existing bridges

Munehiro ISHIDA M Ishida, born 1966, received his

Civil Eng Master of Engineering Degree from
Nippon Steel Corporation Kobe Univ in 1992 He is attac
hed to the Construction & Ar
Tokyo, Japan
chitectural Materials Development
& Engineering Service Division
which is now developing new
materials related to long span
bridges
SUMMARY
This paper descnbes a method for safety evaluation of concrete bridges and its venfication based upon
field tests performed either under static loading by test trucks for application of the System identification
method or under dynamic loading by falling mass for application to the modal analysis The safety factors
for flexural and shear failures were evaluated from these test results Then, they were venfied through
the ultimate load test carried out in situ on the reinforced concrete main girders isolated by cutting off
from the bridge system Finally, the remaining life of the bridge was predicted by applying the fuzzy set
theory which deals with the subjective Information of bridge engineers

R£SUM£
Ce document dement une mäthode d'eValuation de la s6cunt6 des ponts en bäton et sa v6nfication d'apres
des essais sur place r6ahs6s soit sous Charge statique avec camions pour application de la möthode
d'identification de Systeme, soit sous Charge dynamique avec chutes de masses pour application de
l'analyse modale Les facteurs de säcuntö pour la flexion et la torsion ont 6t6 eValuäs a partir des rösultats
de ces essais Ensuite, ils ont 6t6 ve>ifi6s au cours d'un essai de Charge limite mene' sur le terrain sur les
poutres pnncipales isolöes de l'ensemble du pont Et a la fin, la duröe de vie restante du pont a 6t6 prövue
en appliquant la thäone de l'ensemble flou qui tient compte des informations subjeetives des Ingenieurs
des ponts

ZUSAMMENFASSUNG
Diese Arbeit beschreibt eine Methode zur Sicherheitsbewertung von Betonbrücken und eine dazugehörende
Verifikation, basierend auf Feldtests, die entweder unter statischer Belastung durch Testlastwagen
zur Anwendung der Systemidentifikationsmethode oder unter dynamischer Belastung durch fallende
Massen für Anwendung der Modellanalyse durchgeführt wird Die Sicherheitsfaktoren für Biege- und
Bruchschaden werden auf der Grundlage dieser Testergebnissen bewertet Danach werden sie durch einen
endgültigen Belastungstest vor Ort auf Haupttragern aus Stahlbeton, die durch Abschneiden vom
Bruckensystem isoliert sind, verifiziert Schliesslich wird die Restlebensdauer der Brücke durch Anwendung
der Fuzzy Set Theorie vorausgesagt, die subjektive Information von Bruckeningenieuren verwerndet
310 FIELD TESTING AND SAFETY EVALUATION OF CONCRETE BRIDGES

1. INTRODUCTION
It is an important problem for maintenance and rehabilitation of existing bridges to develop a method of
safety evaluation such as remaining life and load carrying capacity of bridges. This paper describes a
method of safety evaluation of concrete bridges in service and the verification of the evaluated results
by field tests. The field tests to evaluate structural safety were performed as static loading test by test
trucks for application to the system identification(SI) method, and also as dynamic loading test under
forced Vibration caused by falling mass for application to the modal analysis.

Both of the safety factors for flexural and shear failures and the change of dynamic behavior were
evaluated ftom the field tests. These results were verified through the ultimate load test carried out in
field on the reinforced concrete main girders isolated by cutting off from the bridge system. Finally, the
remaining life of the bridge was predicted by application of fuzzy set theory which deal with the
subjective information of bridge engineers. The fuzzy mapping which was determined based on
questionnaire results performed on more than 20 experts was introduced for remaining life prediction of
the existing bridges. A few concrete bridges on which field data have been collected are analyzed to
demonstrate the applicability of this method. Through the application to the cracked reinforced concrete
bridge girders, reasonable results were obtained by field tests.

2. FIELD TEST FOR STRUCTURAL SAFETY EVALUATION

2.1 Flow of Structural Safety Evaluation
Fig. 1 shows a general flow of safety evaluation and its verification for concrete bridges based on field
tests. As shown in the figure, there are two types of field tests, one is non-destructive test to identify the
system parameters which consist of the geometrical moment of inertia of both main girder and cross
beam and Young's modulus of concrete, while the other is destructive tests such as ultimate load test
and material tests to verify the evaluated results[l]. The non-destructive tests in situ are performed as
static and dynamic loading tests, and deflection and acceleration are measured as mechanical behaviors.
The stiffness(flexural rigidity) of each girder is identified by applying the SI method to the mechanical
behaviors[2, 3], and then the section forces such as bending moment and shear force for the modeling
of load variables are evaluated by structural analysis for the design load using the identified system
parameters for each girder. Furthermore, the material tests are carried out for modeling of resistance
variables linked with the Statistical data of results of ultimate load tests which have been previously
carried out on other bridges. Then, the structural safety for bending and shear failure are evaluated by
calculating the safety factor, y, safety index, ß and the probability of failure, Pf. Finally, the ultimate
load test is condueted in order to verify these evaluated results with the material tests.

2.2 Bridge Descriotion

The bridges for which tests were performed are six national highway bridges as shown in Table 1.
These bridges consist of five RC-T simply supported beam bridges and a RC-T continuous beam
bridge, and all of the bridges are located in Hyogo prefecture, Japan. It is noted that the five simply
supported bridges have almost the same bridge characteristics and history of service conditions except

Non Destructive Measurement of Evaluation of System

Test in Situ Mechanical Behavior Parameters for
• Static Load Test • Deflection Each Girder
• Dynamic Load Test • Acceleration

Application of System Identification Method

Evaluation of Safety for
Bending and Shear Failure
by Indices as:
Ultimate Load Test
in Situ
• Safety Factor
• Safety Index ß
j
Material Tests ' Probability of Failure Pf

D- Field Test Data Analysis and Evaluation

Fig. 1 Flow of safety evaluation and verification for concrete bridges

,0J\^ A. MIYAMOTO - H. MORIKAWA

for age, number of girders, existence of cross members

the region of about 40 to 60 years and the number of
is about 10m. Widening using an additional girder wa
Fig. 2 illustrates the general view of Maenobashi bridg

2.3 Details of Field Tests

2.3.1 Non-Destructive Tests
Static loading was applied to the deck by a three-axled
weight(about 20tf) being accurately measured prior to
a way as to cause a realistic severe loading condition
forced Vibration by falling mass(300kgf) were arrang
shown also in Fig. 3. Typical loading procedure for
dropping was carried out from about 70cm height, for
white noise and to obtain a stable average value. Defle
static and dynamic loading tests were measured
l/1000mm aecuraey and 10 mm capacity and ultra-sm
constant frequency response of up to 700Hz and 20G
deflection meters and the acceleration sensors are show
acceleration data to identify the modal parameters such

2.3.2 Ultimate Load Test and Material Tests

Test apparatus for the ultimate load test was specially
load to the reinforced concrete main girders which wer
system. Fig. 5 shows the details of the loading syste
applied at the midpoint of the span via a lOOtf capac
Table 1 Outline of tested

Bridge Name Sakurabashi Bridge Mae

Total Length 21. 84m

312 FIELD TESTING AND SAFETY EVAL

re: I

=;-',:

Fig. 4 Details of

embedded in the loading system. The loading schem

up to failure. The relationship between applied load
X-Y recorder. On the other hand, a dynamic load tes
carried out on each main girder after unloading in the
dynamic behavior up to failure. The acceleration res
girder. On the material tests, specimens of concrete c
were tested to evaluate mechanical properties such a
of carbonization.
3. SAFETY EVALUATION AND ITS VERIFICA
 A. MIYAMOTO - H. MORIKAWA - K. TAKEUCHI - M. ISHIDA 313

rigidity of the main girder and the spring coefficient of rotation of the support, are assumed, and modal
parameters such as eigen values and Vibration mode are evaluated by analysis. Next, linearization of the
objeetive function is carried out within the region of movement limits for design variables, using Eq.
(1), and a search for the minimum point of the objeetive value is tried using the simplex method. In the
event that the change of design variables exceeds the movement limits, reanalysis of modal parameters
and restart of the search for Sie minimum values from updated initial values are executed in the same
procedure iteratively up to the stage in which the objeetive function is within the region of allowable
limit.
3.2 Evaluation of Load Carrying Capacity
For material tests, the specimens of concrete core and reinforcing bar extracted from the main girders of
the six existing bridges were tested to evaluate the deterioration of material. It was found that the
compressive strength and modulus of elasticity of concrete were evaluated to be very low compared
with the design value, and the degree of carbonization of concrete was very great However, the
effective section area of reinforcing bars providing tensile strength maintained their initial value,
although surface corrosion was detected. Since the carbonization of concrete is considered to be the
index of deterioration which includes the effects of quality of material and construction Standards,
traffic loading condition, permeability of slab and web concrete, aeid environment, and so on, the
relationship between the depth of carbonization and the compressive strength of concrete, as shown in
Fig. 7, is suitable for the evaluation of the load carrying capacity considering the effects of these
factors.

Ultimate load tests of main girders were carried out on five existing bridges, and results of these tests
are summarized in Table 2, compared with calculated values based on the evaluation equations defined
in the specification, and using compressive strength obtained from material tests. In this case, the load
carrying capacity was defined to be bending moment at yield point of reinforcing bar for bending
failure and shearing force at yield point of stirrup for shear failure. The results show that the calculated
values give good agreement with the results of field tests, except for the value of bending failure on
Taitabashi bridge which was affected by the bond failure between the reinforcing bars and concrete, and
the value of shear failure on Sakurabashi Bridge, including the effect of scatter in the section area of the
compression zone.

3.3 Modeling of Resistance and Load Variables

In the safety evaluation method, the probability model for load effect and load carrying capacity should
be construeted considering the scatter of data and error in evaluation. The probability model for load
carrying capacity can be evaluated by considering the correction coefficient corresponding to the ratio
of the value obtained from the ultimate load test to that from analysis. The correction coefficient a is
assumed to be a random variable characterized by the normal distribution, N (\Xa9 oa) and then, the load
carrying capacity for bending and shear failure can be expressed by[4]:

N(pMut • QMut N(iiaMucai, oaMucai) ,N(iiSut vs* ^(M^aca/, oaSucai) (3)

where, MUcal and Sucal are calculation results of the load carrying capacity for bending failure and
shear failure, respectively, and Mut and Sut are results of ultimate load tests for bending failure and
shear failure, respectively. The coefficient, N (Ua, oa) can be determined through comparison of
estimated and measured values for load carrying capacity, as shown in Table 2 and Table 3.

On the other hand, for the evaluation of the probability model for load effect, the mean value, u^and
Standard deviation, os of the section force, S for total weight of the three-axled truck, W -N (pw, o\y)
can be expressed by the following equation using the influence surface[4]:

us
as
- l*w (0.1
ow (0.1
I
I
(ifc + Hä)f + 0.4
(*U + 1k)f + °-4
lI (tu + n*)*) \
(Hl + Hk)ä) J
(4)

Here, the distribution of the total weight of three-axled truck was assumed to be as given in the results
of research by Hanshin Expressway Public Corporation [5].

In this way, the safety factor, y, safety index, p and probability of failure, Pf can be evaluated by the
314 FIELD TESTING AND SAFETY EVALUATION OF CONCRETE BRIDGES

START
Table 2 Comparison of estimated and measured
values for load carrying capacity
Determination of initial values for design
variables Xi (i l,2, •••,n). (a) Case of bending failure (tf«m)
such as flexural rigidity of main girders
and spring rigidity for rotation at supports
Bridge Name Span No Girder Mue.l Mut

k 0 Maenobashi 1 C 73.28 78.22

2 B 73 28 74.15
2 C 73 28 69 82
Cvaluation of modal parameters by analysis
A(x.) : Eigen value Taitabashi 1 B 78 51 34.46
ym(x,) : Mode 2 B 79.24 41 30
m number of node 3 B 79 77 56 23

Oyasubashi 3 B 318 4 388 78

Judgement of Yes
k k+1 error for $ STOP b) Case of she ar failure (tf)
"^ identification
Bridge Name Span No Girder Sue.. Sut

Sakurabashi 1 C 16 43 12.49
1 D 16 43 9 92

Calculation of objeetive function Fk and Nakaibashi 1 B 18 20 22.29

derivative of objeetive function VFk 2 B 25.85 25 59
3 B 26.63 26 59

Calculation of updated objeetive function Oyasubashi 1 B 57 91 62 49

Fk + Fk + VFk-SXk
1
C 56.20 64.09
| <5Xk |
^ di.mit. äi.m.t • Move limit 1

75.24
2 B 53.61
2 C 54 88 62.10

- Xk + 1
Xk + SXk

Fig. 6 Flow of sequential linear programming (for dynamic problem)

350 Mean value *-Sakurabashi bridge
ÜT ju/c - 300 0 exp (- 0 175C) o Menobashi bridge
^^.rn 300
Standard deviation
x Taitabashi bridge
— Nakaibashi bridge

^ 250
<r/c 55 0exp(-0 131C) • Oyasubashi bridge
Aokibashi bridge
.er.
cn 200 Standard deviation
©
</>
05 1h()
>
CO
-o
^
CO
(D 1UU
Q. XX-
E
Ü 50
O A
AUA
B 10 12 14

Depth of carbonization, C (cm)

and depth of carbonization for concrete
Fig. 7 Relationship between compressive sttength
load carrying capacity
Table 3 Ratio of measured value to estimated value for
Mu,/Mu M,./Mu«,i
(ex Taitabashi)

Mean Value 0 845 1 063 1 034

294 0 115 0 239

Standard Deviation 0
 A. MIYAMOTO - H. MORIKAWA - K. TAKEUCHI - M. ISHIDA 315

following equations:

Y-(|iÄ-l.l-|tia)/|ifi, P=s(üÄ-Ms)/Voi+oi, Pf P((R-S)<0) (5)

3.4 Evaluation Results and Verification

The proposed method for structural safety evaluation based on field tests was applied to the existing
bridges as shown in Table 1. The flexural rigidity identified by applying the SI method to results of
field tests, and the safety evaluation based on safety factor, y, safety index, ß and probability of failure,
Pf are summarized in Table 4. The reduction of the flexural rigidity of the oldest bridge, Nakaibashi
bridge, was found to be remarkable compared with the design value, that is, the detected degree of
damage for this bridge is very large. From the comparison between the safety evaluation for bending
failure and shear failure, it can be judged that the failure mode for Aokibashi bridge would be of the
bending type and that for Sakurabashi and Nakaibashi bridges would be of the shear type. In particular,
since the probability of failure for Nakaibashi bridge is extremely large compared with the Standard for
the initial condition, of which the ränge is about IQ-3 to 10-5, the safety condition of this bridge seems
to be eritieal from the bridge maintenance viewpoint.

The comparative study between the above mentioned evaluation and the results of the ultimate load test
or material tests, can be considered to be available for the verification of suitability and Utility of the
proposed method. Firstly, for the girder B of Nakaibashi bridge, of which the degree of damage was
estimated to be the largest of all tested bridges, both results of Table 4 and the ultimate load test can be
seen to show that the failure mode of this girder would be of the shear type. Similarly, for the girder of
Taitabashi bridge which has the second highest grade of damage of all tested bridges succeeding
Nakaibashi bridge, both results of Table 4 and the ultimate load test suggest that the failure mode
would be of the bending type. Since remarkable damage was not detected in the evaluation for inside
girder of Oyasubashi bridge, the possibility and verification of bending failure or shear failure cannot be
distinguished exactly. Accordingly, the prediction of the failure mode at the stage of heavy damage can
be considered to be appropriate. In addition, the results of the SI method and safety evaluation shown in
Table 4 can be found to be adjusted by the results of material tests, where compressive strength and
modulus of elasticity of concrete for Nakaibashi bridge are lower than half the design values. For
Aokibashi bridge, compressive strength and modulus of elasticity of concrete were about half the
design values and the order of the resulting values of the three girders A, B and C was A > B > C.
Therefore, the results of the SI method shown in Table 4 can be seen to be reasonable.

Table 4 Summary of structural safety evaluation results for existing concrete bridges

Tlexural Rigidity S a f e t y E v a 1 u a t i o n

(x 104 tf-m) Bending Tailure Shear railure

Bridge Name Girder
Design Lstimated
Value Value r M /?M P ,M 7 s ß* P fS

A 7 66 6 14 3 88 2 173 1
483x 10 2
2 53 1 934 2 631x 10 3

B 7 Ol 2 88 8 22 2 789 2 626x 10 3
4 31 2 543 5 451x 10 3

Sakurabashi C 7 Ol 1 68 16 12 3 174 7 502x 10 4

8 94 3 318 4 494x 10 4

D 7 66 4 73 5 64 2 422 7 685x 10 3
4 39 2 525 5 734x 10 3

E 13 22 7 80 3 75 2 097 1
793x 10 2
2 68 2 280 9 872x 10 3

A 8 38 5 00 4 25 2 424 7 665x 10 3
3 64 2 594 4 740x 10 3

Taitabashi B 10 03 6 23 5 34 2 491 6 363x 10 3

5 05 2 806 2 508x 10 3

C 13 25 6 63 3 68 2 259 1
194x 10 2
3 37 2 483 6 506x 10 8

A 25 90 7 91 4 80 2 380 8 664x 10 a 1 80 1 326 9 238x 10 2

Nakaibashi B 17 21 9 11 5 49 2 414 7 886x 10 3 1 97 1 383 8 332x 10 z

C 25 90 8 93 4 31 2 265 1
176x 10 2 1 67 1 185 1
181x 10 »

A 101 04 110 25 7 68 2 567 5 103x 10 3

7 02 2 899 1
846x 10 8
Oyasubashi B 88 32 92 27 16 32 2 877 1
998x 10 3
11 55 3 013 8 355x 10 "
C 88 32 86 41 16 02 2 880 1
980x 10 3
14 00 3 050 1
128x 10 3
D 101 04 89 68 8 41 2 638 4 158x 10 3
7 96 2 895 1
272x 10 8

A 9 98 5 74 4 23 2 367 8 916x 10 3
4 58 2 797 2 561x 10 3

Aokibashi B 10 87 4 48 5 86 2 560 5 225x 10 3

6 12 2 866 2 058x 10 3

C 18 50 3 21 5 29 2 578 4 973x 10 3
5 91 3 005 1
342x 10 3
316 FIELD TESTING AND SAFETY EVALUATION OF CONCRETE BRIDGES

4. REMAINING LIFE PREDICTION

Based upon the above mentioned evaluated results, it is of interest to predict the remaining life of the
bridges. However, the practical procedure for remaining life prediction involves an uncertainty scheme
such as engineering judgement based on technical knowledge with professional experience. Therefore,
the fiizzy set theory was introduced in a process of remaining life prediction for dealing with the
subjective uncertainty.

The relationship between the integrity(soundness), S and the remaining life, t is assumed based on
questionnaire results for the serviceability of actual bridges(Maenobashi bridge, Taitabashi bridge and
Nakaibashi bridge). The questionnaires were performed on more than 20 experts for each bridge. The
remaining life, t can be assumed as in the following equation[6]:

t 3.03 x lO'1^0-98 (6)

Since the deviation of the data from the equation is very large, the treatment of the subjective
uncertainties included in this evaluation should be considered. In order to represent these uncertainty,
the "kerner of log-normal distribution is introduced for membership function of fuzzy set theory [6].

u,(,,^fexp^I|-L|__Aj where, \s lnfe] ln[(3a03xlO-1)1S°s>8] (7)

Next, the membership function for the remaining life, t is related to the membership function for the
integrity, S by Fuzzy synthesis with aid of fuzzy relation, R[6\. Relative prediction of remaining life for
the bridges is performed by integrating the forecast of respective remaining life, 7} based upon the
corresponding membership function obtained from evaluated results for each judgement factor.

(a) Modliied membcislup function bcioic combination

Y^tc hs [t Numenzed reliability of
hl \r judgement factor h
M7Vi
hi\t) \ 1 0

\
My« h-t
/72 0 9

\/
/
,/\
\
\\\ '
/?3=0 8
^ta h\ v Assigned membership
function(M F for degree
of bar corrosion
M/t? hi {r Assigned M F for degree
\
*

of stiffness
*
V/c /i*[r) Assigned M F for degree
V of safety factor

50 90
Remaining life, f (years)

(b) Membership lunctiou aftci combination

1 0
M7^\ h\ tu ia ic h\ (t

50 90
Remaining life, t (years)

Fig. 8 Prediction of remaining life

 A. MIYAMOTO - H. MORIKAWA - K. TAKEUCHI - M. ISHIDA 317

As an example, Fig. 8 shows the result of calculating the remaining life for Taitabashi bridge using the
"degree of bar corrosion", "stiffness of girder" and "safety factor" evaluated by field inspections and
tests as a measure of the integrity. From this figure, the remaining life of the bridge is predicted to be
about 35 years corresponding to the peak value of the membership function after combination.
Furthermore, the predicted values for Sakurabashi bridge, Maenobashi bridge and Nakaibashi bridge
are roughly about 15 years, 18 years and 18 years, respectively.

5. CONCLUSIONS
The major part of bridge diagnosis which is the kernel of the systematization for bridge maintenance is
to develop a method of safety evaluation on items such as remaining life and load carrying capacity.
The main conclusions obtained in this study can be summarized as follows:
(1) A few examples of the field testing and the measuring procedures, the evaluation methods of both
the safety factors, including the safety index and the probability of failure, for flexural and shear failure
and the verification concept of the evaluated results by field tests were performed on six reinforced
concrete bridges.

(2) The prediction of remaining life of concrete bridges in service, which is a major problem faced in
the process of bridge maintenance, can be quantitatively evaluated based on the evaluated results for the
judgement factors. The relation among the factors are obtained through the fuzzy relations in the
questionnaire results performed on experts.

REFERENCES
1. Miyamoto, A. and Fujii, M., Safety Evaluation and Verification by Field Testing of Concrete
Bridges. Proc.of ASCE Structures Congress '89,1989.
2. Miyamoto, A., Morikawa H. and Yamamoto M., Sensitivity Analysis of Dynamic Behavior in
Existing Concrete Bridges and Its Application to Damage Detection. Proc. of JSCE, No.442, 1992.
3. Natke, H.G. and Yao, J. T. P.(ed.), Structural Safety Evaluation Based on System Identification
Approaches. Proc. of the Workshop at Lambrecht / Pfalz, Friedr. Vieweg & Sohn, 1988.
4. Morikawa, H. and Miyamoto, A., Structural Safety Evaluation and Remaining Life Prediction of
Existing Concrete Bridges. Proc. of the 17th Conference on Our World in Concrete & Structures, 1992.
5. Committee of Design Load in Hanshin Expressway Public Corporation, Research for Design load of
Hanshin Expressway. Hanshin Expressway Public Corporation, 1986.
6. Miyamoto, A., Morikawa, H. and Fujii, M., Development of Expert System for Serviceability Rating
of Concrete Bridges and Its Practical Application. Proc. of the 16th Conference on Our World in
Concrete & Structures, 1991.
318
i\
 Interpretation of 200 Load
Interpretation de 200 essais de
Auswertung von 200 Belastungsv

Munzer HASSAN Olivier BÜRDET

PH.D. candidate Dr. Eng.
Swiss Fed. Inst, of Techn. Swiss Fed. Inst,
Lausanne, Switzerland Lausanne, Switze

jt^^S.

SUMMARY
i
320 INTERPRETATION OF 200 LOAD

1. INTRODUCTION
For the past 20 years, the Institute of Reinforced and
with full-scale load testing of bridges (Fig. 1). This
The Swiss Codes recommend a load test for any new
The objeetive of a load test is to determine and quanti
of load tests are acceptance tests, aimed at examin
consequence put it into service or not. The decis
concordance between the measured and calculated de
and on the affinity between measured and calculate
correlation between an unsatisfactory behaviour d
behaviour of the bridge, characterized by a non-stab
abnormal behaviour of the bridge under a load tes
frequent inspections and early maintenance work.
the main results of these two hundred tests have bee
conclusions on the real behaviour of a bridge under l
this article.

,:-;
;•- :

\
 M. HASSAN - 0. BÜRDET - R. FAVRE 321

taken during construction. The influence on the calculated deflections of the method used for
determining the modulus of elasticity is shown in Fig. 2, taking the example of the viaduct of
Coudray.

14 <>1 m

cqa d^D
hn
J
120

2 30|;
1 09 I 0 59 i
1 11
74 ^Z 174
2 07 2 07 0 95 1 11 0 51 1

097 2 36 2 66 0 80 0 97 0 43
0 82 3 16
2 36
0 82
-400
3 61 mm
5 40
32 00 32 00 m 32 00

Ec=38 GPa from elastic modulus-compressivc strength relationship (SIA),

Ec=32 GPa from cores, Ec=28 GPa from ultrasonic pulse velocity
Fig. 2: Calculated deflections for viaduct of Coudray based on three methods for determining the modulus of elasticity

22 Effective moment pf inertia

section is easily determined as it
At first sight, it would seem that the moment of inertia of the cross
depends only on geometry. However, the effective inertia depends on several parameters such as
reinforcement, parapets, asphalt deck surface, cracks and micro-cracks due to construction or other
causes.
In the past, these parameters have been roughly taken into aecount in the evaluation of the results of
load tests by adjusting the modulus of elasticity of the concrete. To cover the partieipation of such
secondary elements, the value of the modulus of elasticity was taken as 40 GPa in most cases. A
better Solution is to base the calculated deflections on the real modulus of elasticity of the concrete and
on a better estimate of the effective inertia of the cross section.

3. DATABASE
The differences between calculated and measured deflections during load testing led IBAP to establish
a computerized database. The database contains the main characteristics of each bridge along with the
results of the static and dynamic load tests. The objeetive of this database is a Statistical study of
bridge behaviour as observed during load tests. The influence on the behaviour of the type of
structure, cross-section, parapet and other factors can be discovered and quantified by sorting and
filtering the results.
Table 1 shows the structural system and type of cross-section of all bridges contained in the database.
One hundred and sixty five of the tested bridges are post-tensioned concrete structures.

Structural system Cross-section

Type Number Type Number

beam 168 box-girder of constant depth 55

rigid portal frame 21 box-girder of variable depth 25
arch 5 slab beam 28
cable stayed 3 open (beams connected by concrete deck) 89

Table 1: Structural system and type of cross-section of all bridges contained in the database
322 INTERPRETATION OF 200 LOAD TESTS OF SWISS BRIDGES

2J—Selection of bridge tvpe and Statistical parameters

In order to obtain a homogeneous set of results, the analysis focused on intermediate spans of post-
tensioned bridges with the structural system being a continuous beam. Because it was anticipated that
the influence of massive R.C. parapets and of an asphalt layer would be important, bridges for which
no information was available on the presence of these items at the time of testing were also rejected.
Finally, bridges for which the ratio of measured to calculated deflections were inferior to 0.65 and
superior to 1.30 were rejected because of doubts on the reliability ofthe engineer's calculations.
Sixty six bridges satisfied these criteria. In this article, "Rm" is used for the average ratio of measured
to calculated deflections. At mid-span of the loaded span, Rm 0.94 and a 0.16 (Standard
deviation) for the sample of 66 bridges; taking Ec 40 GPa for the calculated deflections. The effect
of the R.C. parapets and the asphalt layer is neglected for the calculated deflections. While the
difference of the average ratio to one can be explained by the participation to the inertia of secondary
elements, and by the value of the modulus of elasticity, the dispersion is large. In order to understand
and reduce this dispersion, the following three sub-samples were examined:

- bridges with RC parapets (22 bridges), Rm 0.82, o 0.095;

- bridges with a box-girder of variable depth (10 bridges), Rm 1.03, G 0.11;

- bridges with an asphalt layer (33 bridges), Rm 0.93, <J 0.15.

The smaller dispersions of these three sub-samples shows the influence of certain characteristics of
the bridge on its behaviour, as it clearly appears in Fig. 3.

¦ 33 Bridges tested with an asphalt

layer Rm-0 93
" 10 Box girders of variable depth
Rm-1 03

22 Bridges with RC parapets

Rm-0 82

RATIO OF MEASURED TO CALCULATED DEFLECTIONS

Fig. 3: Normal probability distribution curve for the ratio of measured and calculated deflections for: bridges tested
with an asphalt layer; the box-girder bridges of variable depth; all the bridges with massive R.C. parapets
(Ec 40 GPa)

3.2 Analysis of data

The results of the load tests in the sample of 66 bridges were successively corrected using the
reciprocal of the Rm determined for each factor contributing to the moment of inertia. The increase in
stiffness due to the presence of normal and prestressed reinforcement was estimated to 6%.
After these corrections, the results of the calculations led to an average ratio of measured to calculated
deflections of 1.19, with a modulus of elasticity of concrete equal to 40 GPa. In order to obtain
Rm 1.0 for this sample, the modulus of elasticity of the concrete should be taken to 33.5 GPa,
which is much closer to actually measured values. Fig. 4 shows the histogram of the ratio of
measured to calculated deflections and the Gauss distribution after correction of the influence of the
parapets, of the asphalt layer, of the reinforcement and of the cross-section. Table 2 summanzes the
participation to inertia of asphalt layer, R.C. parapets and reinforcement.
 M. HASSAN - 0. BÜRDET - R. FAVRE 323

The asphalt surface increases the effective inertia of the superstructure. This participation depends
strongly on the surrounding temperature and on the cross-section of the bridge. Fig. 5 shows the
normal probability distribution for the ratio of measured to calculated deflections for bridges tested
with and without asphalt.

Element Asphalt layer R.C. parapets Reinforcement

Participation to inertia 6% 24% 6% (estimated)

Table 2: Participation to inertia of asphalt layer, R.C. parapets and reinforcement

10 --

I Real distribution
b e - ' Gauss distribution

2 -

0 76 0 84 0 92 1 00 1 08 116 1 24 1
32
RATIO OF MEASURED TO CALCULATED DEFLECTIONS

Fig. 4: Histogram of distribution of ratio of measured to calculated deflections for 66 bridges after correction
(Ec 33.5 GPa)

Bridges tested without asphalt

layer Rm=»1

Bridges tested with asphalt

layer Rm=0 94

0 6 0 8 1
12 14
RATIO OF MEASURED TO CALCULATED DEFLECTIONS

Fig. 5: Normal probability distribution curve for the ratio of measured to calculated deflections for bndges tested with
and without asphalt (Ec 33.5 GPa)
The parapets increase the effective inertia of the superstructure. The amount of this participation
depends on the type of parapets, on the connection between the parapets and the superstructure and
pn the cross-section of the bridge. Fig. 6 shows the normal probability distribution for the ratio of
measured to calculated deflections for bridges tested with and without reinforced concrete parapets. It
should be noted that the effect of parapets is neglected for the calculated deflections.
324 INTERPRETATION OF 200 LOAD TESTS OF SWISS BRIDGES

- Bridges tested without R C.

parapets Rm-1
"
Bridges tested with RC
parapets Rm-0 76

0 6 0 8 1 1.2
RATIO OF MEASURED TO CALCULATED DEFLECTIONS

Fig. 6: Normal probability distribution curve for the ratio of measured to calculated deflections for bridges tested with
and without R.C. parapets (Ec 33.5 GPa)

Box-girder bridges systematically exhibit an effective modulus of elasticity lower than average,
especially box-girders of variable depth. On the other hand, slab bridges and bridges of an open
cross-section systematically show an effective modulus of elasticity greater than the average modulus.
Table 3 shows the effective modulus of elasticity for the bridges with various cross-sections.

Effective modulus
Type of cross-section of elasticity [GPa]
Box-girder of constant depth 32

Box-girder of variable depth 31

Slab beam 37

Open (beams connected by concrete deck) 35

Table 3: Effective modulus of elasticity for the bridgesof different types of cross section
A possible explanation for these variations lies in the fact that the method of construction is strongly
dependent on the type of cross-section. Fig.7 shows the normal probability distribution for the ratio
of measured to calculated deflections for box-girder bridges with a constant depth constructed on
fixed scaffolding, for box-girder bridges with a variable depth constructed by the balanced cantilever
method and for box-girder bridges with a constant depth constructed by incremental launching. In the
same manner slab beams and open cross-section bridges are shown in the same figure.

- Box girder constructed on fixed

scaffolding Rm-1 04
" Box girder con by balanced
cantilever method Rm-1 07

" Box girder constructed by

incremental launching Rm-1 32

- Slab beams Rm-0 91

¦ Open cross section

Rm-0 96
brigdes

0 6 0 8 1 12 14
RATIO OF MEASURED TO CALCULATED DEFLECTIONS

Fig. 7: Normal probability distnbuüon curve for the ratio of measured to calculated deflections for bridges of different
type of cross-sections and different construction methods (Ec 33.5 GPa)
 M. HASSAN - 0. BÜRDET - R. FAVRE 325

13. Affinitv between the measured and calculated deflections

The ratio between the measured and calculated deflections in the loaded spans is systematically larger
than those in the adjacent spans. If we make the calculated deflections correspond to those measured
in the loaded spans at mid-span (Rm 1), Rm becomes 0.84 in the adjacent spans (see Fig. 8). This
systematic difference in stiffness between adjacent spans could be explained by cracking of the loaded
span and/or a lack of continuity on supports. Another factor can be a reduction of the modulus of
elasticity in the loaded span due to the level of stress in the concrete.

¦ Loaded span Rm-1

¦ Adjacent span Rm-0 84

0 6 0 8 1 12 14
RATIO OF MEASURED TO CALCULATED DEFLECTIONS

Fig. 8: Normal probability distribution curve for the ratio of measured to calculated deflections for the loaded central
span and for the adjacent span

4. ULTIMATE LOAD FROM LOAD TEST RESULTS

The load-deflection curve for a prestressed concrete bridge can be approached by a tri-linear
relationship as shown in figure 9, in which we have assumed that 80% of the permanent load is
balanced by the effect of prestressing, a permanent load of 20 kN/m2 and a live load of 5 kN/m2.
Considering these different values and assuming a global safety factor of 1.7, the ultimate design load
is 1.7 (20+5) 42.5 kN/m2. Because the load test is an acceptance test the upper limit of loading is
generally about 5 kN/m2, which normally should not lead to cracking of the bridge. It seems that the
extrapolation of the ultimate load based on the results of a single load test is senseless. However, a
good correlation between measured and calculated deflections indicates a satisfactory structural
behaviour, provided the deflections are reversibles and the requirements of ductility are fulfilled.

LOAD [KN/m2]
IMPORTANT DEFLECTION
AND MOBILISATION OF
ii THE UNDETERMINATE
RESERVE
ULTIMATE LOAD 1«
LOAD CORRESPONDING 42.5
TO FIRST YIELDING

CRACKED CONCRETE

CRACKING LOAD qi -•
LOAD TEST

PERMANENT LOAD REAL BEHAVIOUR

UNCRACKED CONCRETE
BALANCED LOAD
SIMPLDFIED BEHAVIOUR AND QUASI LINEAR
BEHAVIOUR

u
measured dcflecüoo
200 DEFLECTION [ mm ]

Fig. 9: Load-deflection curve of a prestressed concrete bndge (order of magnitude)

326 INTERPRETATION OF 200 LOAD TESTS OF SWISS BRIDGES

5. CONCLUSIONS
The load test is a real chance and a challenge to observe and interpret the actual behaviour of a
structure under load on a 1:1 scale. A frequent correlation between an unsatisfactory behaviour during
a load test and the long-term behaviour of a bridge regarding a non-stabilization of cracking and of
sagging has been observed.
The contribution to the rigidity of the parapets, the asphalt layer and the reinforcement has to be taken
into aecount for calculated deflections. The modulus of elasticity has to be experimentally determined
and not only estimated according to the compressive strength of the concrete.
Although our research concerning 200 load tests is not completely finished, some important
preliminary conclusions can be drawn:
Box-girder bridges exhibit an effective modulus of elasticity lower than the average. On the other
hand slab bridges and bridges of open cross-section show an effective modulus of elasticity greater
than the average modulus.
The method of construction influences the stiffness of the bridge; box-girder bridges with a constant
depth constructed by incremental launching are greatly less rigid than those constructed on fixed
scaffolding. This can be explained by micro-cracking during the execution.
Even if the calculated deflections are corrected to correspond to those measured at mid-span of the
loaded span, the affinity between the measured and calculated deflected shapes is not perfect. A
systematic difference in stiffness between loaded spans and adjacent spans is observed. This can be
explained by cracking of the loaded span and/or a lack of continuity over the supports.

6. REFERENCES
[ 1] Norme SIA 160 (Swiss Standard). Actions sur les structures porteuses (Actions on load-bearing
structures). Soci&e" suisse des ing&iieurs et des architectes. Zürich, Switzerland, 1989.
[ 2] Recommandation SIA 169 (Swiss Standard). Recommandation pour la maintenance des ouvrages de
g&iie civil (Recommendation for the maintenance of Civil Engineering structures). SociSte' suisse des
ing£nieurs et des architectes. Zürich, Switzerland, 1987.
[ 3] RILEM 20-TBS-2 General recommendation for statical loading tests of load-bearing concrete structures
in situ. Drafted by D. SZOKE. Materials and Structures, Vol. 16, N°96, pp. 405-419,1983.
[ 4] RILEM 20-TBS-3 Testing bridges in situ. Drafted by T. JA VOR. Materials and Structures, Vol. N° 96,
pp. 420-431, 1983.
[ 5] LADNER M., In situ Load Testing of Concrete Bridges in Switzerland. ACI, SP-88-8, pp. 59-70,
1985.
[ 6] SUTER R., "Ponts de Saint-Maurice: essais de Charge statiques et dynamiques" (The St. Maurice
Bridges: Static and dymanic load testing) Chantiers (Switzerland) vol. 6,1988.
[ 7] FAVRE R., and SUTER R., "Lauscultation des ouvrages d'art: conclusions des inspections et des
6preuves de Charge en Suisse". International Conference on Inspection, Maintenance and Repair of Road
and Railway Bridges, Brussels-Paris, 1981.
[ 8] FAVRE R. and MARKEY I., "Long-term Monitoring of Bridge Deformation". NATO Research
Workshop, Bridge Evaluation, Repair and Rehabilitation. NATO ASI series E: vol. 187, Baltimore,
USA, 1990.
[ 9] FAVRE R., HASSAN M., and MARKEY I., "Bridge Behaviour Drown from Load Tests". Third
International Workshop on Bridge Rehabilitation, Darmstadt, Germany, June 1992.
[10] MARTINEZ Y CABRERA F., and POZZO E., "Experimental control of deformability at short-term
loadings in testing large-span prestressed structures". Materials and Structures, Vol. 25, N° 148,
pp. 231-238, May 1992.
[11] MOSES F., LEBET J.-P., and BEZ R., "Use of Testing for Bridge Evaluation". Third International
Workshop on Bridge Rehabilitation, Darmstadt, Germany, June 1992.
[12] KIRKPATRICK J., LONG A.E., STEVENSON W.M.C. and THOMPSON A., "Load Tests on
Prestressed Concrete Multi-Girder Bridge Decks. ACI, SP-88-2, 1985.
[13] BAKHT B., "The Role of Bridge Testing in Evaluation", pp. 209-218 Bridge Evaluation, Repair and
Rehabilitation. Ed. Nowak and Absi. University of Michigan USA, 1987.
aA

Remaining Carrying Capacity o

Capacite restante des pon
Resttragfähigkeit proviso

Pavol JUHÄS
Prof. Dr.
Technical University
Koäice, Slovakia

Emilia JUHÄSOVÄ
Dr. Eng.
Inst, of Constr. and Archit. SAS,
Bratislava, Slovakia
328 REMAINING CARRYING CAPACITY OF TEMPORARY STEEL BRIDGES

1. INTRODUCTION
Nearly in every country we can meet special steel elements which
are intended to be used as components of temporary bridge structures
during the reconstruction of existing bridges or for special
transport purposes. After some short time of service they
are disassembled and stored for another utilization. After
certain time periods
it is necessary to decide what is thei^ remaining
carrying capacity and whether they can be used again with
sufficient degree of reliability. Similar problems can arise in
the case of such steel elements which were produced with some
degree of defects as notehes, small cracks, geometrical imperfec-
tions, etc. In our paper we deal with the results obtcHned from
the tests and analyses of assembled elements of czecho slovak
temporary steel bridges. Two asembly bridge Systems were tested.
In the first system one from the tested elements was previously
used in the temporary bridge during the reconstruction of the
tram bridge for the period of two years. Others of them did not
fully answer to conditions of safety due to initial production
defects. The second system was later developed with an intention
to replace the first system. Due to results of our tests its
details and a production technology were modified and improved.

2. PROPERTIES OF LOADING IN ACTUAL AND LABORATORY CONDITIONS

Dynamic fatigue loading of bridges is caused in large degree by
variable combined static-dynamic actions from different vehicles
moving through the bridge. Some dynamic loading can be caused by
natural effects as are strong winds and earthquakes. Then the
dynamic loading process of bridges is random stationary or non-sta-
tionary. As a base for the analysis of dynamic loading properties
we can use time records of accelerations, velocities, deflections,
strains, etc. Dynamic measurements on structures and in laboratory
need a wide assortment of different devices such as pickups,
amplifiers, recorders, Computers with A/D and D/A convertors,
appropriate softwares, etc.
A random stochastic process as a continuous random function of a
non-random variable t€I( -oo ,oo) may be written as a set X(t),
where X(t) x-j(t), x 2 (t), x (t)
.j in which
x.j(t) (i=1,2, are realizations of random process. Having
realizations x.j(t) we can calculate for each of them all necessary
characteristics as distribution functions, probability den-
sities, correlation functions, power spectral densities, transfer
functions, coherence functions, etc., due to the purpose of
analysis [5], [6]. Frequency distribution in power spectral density
is influenced by properties of the source loading, e.g. from the
motion of vehicles, and by dynamic properties of the bridge structure
alone. In Fig. we can see^acceleration records in the
1

midspan of tram bridge points 10, 10 and chosen power spectral

densities. When passing from acceleration to velocity or deflection
spectrum, the peaks in upper frequency region are falling down
as can be seen in Fig. 1-A. This record was obtained from the
motion of the sole heavy truck through the bridge. The fluent
motion of heavy vehicles can cause the vibrations which are nearly
harmonic with narrow frequency band [7]. When using piesoelect-
ric measuring system with primary acceleration measurements it is
convinient to apply modified integration according to algorithms
described in [6]. The obtained frequency distribution is important
for the fatigue analysis when following the expected number
 P. JUHÄS - E. JUHÄSOVÄ 329

0 w
© V)
© X
-a
0)

-o

03

*- >

ZH V VZ

103
c
"O

2 51

rvj co <u
X 2 Q.
(0 0)
*»-*.

o ac o h- 09 £_
T3
£_

35
<D o
2 o
O 0)
Q. c_
m
|
i i i
il! i^tp^ T3 >s
C
03
¦M
•r-
O
CO O
o T3 _j
c_ <D
O >
"
M- CO
X
© © ©
«•
•r- >
h- >
CO
*H S£6E •
ZH S 6€ 0> s
CJ) CO
"O T3
•r— t_
Z £_ o
1 s > 1
er o -O o
0)
9 ^E er
o E £_
2 03
18Z
UJ
E £_
c
X +¦» o
H- >fr •r-
Z
>*
¦ J
8 n 03 +-»
03
00 6ZZ H- C_
t O o>
»- x03 _J
Z 5 C o>
o e zh C9u3 o ü
Q- > 9*91-< 6 91 ü
•1—
¦M 03
03
8E*El< C_ E
-O O
SZ U
V Ol •r- L.
88'6<- 9Z'6 > M-

zu -=^<
£9 5
i ¦ 6 V
>szz <

P 88 k
CO
l 1 1 1
L.J, i J I
L0
I I I I I I I 1 1
I I I I I I I
330 REMAINING CARRYING CAPACITY OF TEMPORARY STEEL BRIDGES

of cycles and frequency components with larger amplitudes. Dynamic

measurements give usually informations only about dynamic component
of deformation. Then for laboratory fatigue tests we use the
combination of measured and calculated values. In laboratory
conditions we can apply actual loading processes directly through
the Simulation of a chosen loading, or we can apply one level or
multi-level harmonic loading with the prescribed spectral distribution
[1], [2], [8], [9] [11].
3. LABORATORY TESTS AND THEIR RESULTS
Experimental verification of the fatigue strength and the life-time
of steel bridges is not usually available on full-scale structures.
Technical and time factors are decisive. The technical and
spatial facilities are also limiting factors for the fatigue tests
of full-scale bridge structures in special Laboratories. Therefore,
the fatigue strength and life-time of steel bridges is often
investigated on their modeis. The transfer of conclusions and results
from model tests into actual structures can be problematic. This
is, andhowever, very problematic in the cases of determination of total
remaining fatigue life-time of already exploited bridge
structures with certain constructional influences and production-
-technological defects or cracks.
In Czecho-Slovakia the special assembly bridge Systems for temporary
repeated exploitation have been developed and produced. Due
to originality of technical Solution, valid Standards and former
scientific knowledge have not enabled to determine their total and
remainig fatigue life-time. The connecting bolts and areas of
individual assembly elements have been especially problematic from
the view of fatigue strength and failure. Therefore, the experimental
verification of these phenomena was realized at our Institute
t1],[2],riOD.
Our laboratory enables to realize static-dynamic tests of structures
their elements with constructional length up to 18 m. It
and
was not enough for the test of the whole bridge system, therefore,
only individual assembly elements were tested in special configuration
(Fig. 2). The aim of this was to simulate the design loading
of investigated connecting bolts and areas near connections.
In the
system was
firstrealized
stage the static
succesive and repeated loading of
two forces until P
by one or P where P is
a theoretical ultimate static load, the effect of which corres*
ponds to the effect of design loading on füll scale bridge structures.
Deflections v and strains £ were measured in chosen points
of flanges, webs and stiffeners. Then fatigue tests were realized.
with loading P(t) n( 27Tf t), P =(P
° 1 si min )/2,
P^+Pi ° max
+P f 5 Hz,
P1=(P
1 v
max
-P )/2
min'
The fatigue tests continued in stages with interuption after every
50 000 or 100 000 cycles, when the Systems were controlled both
visually and through the Computer. Controlled values were deflections
andstrains for load levels P=0; Pm-jn and Pmav. The differences
in strains indicated beginning of cracks and their position.
After appearance of cracks their succesive increasing was
recorded with investigation of their influence on the system
behaviour. The tests continued until the failure of Systems.
From system 1
there were tested five assembly elements. Every
element was tested twice in cantilever composition. It means
 P. JUHÄS - E. JUHÄSOVÄ 331

following the behaviour of both its ends in position a or b in

Table 1. The element, previously used in the tram bridge, had
already visible fatigue cracks in the place of connections. The
others were chosen from Stocks. From fatigue tests six different
types of cracks were ascertained - A up to F, as can be seen in
Fig. 2. The decisive cracks and total numbers of cycles N for
individual elements and tests are in Table 1.
Table 1. The results of fatigue tests of system 1.
Indication MSO MS1 MS2 MS3 MS4

cracks C A,E D F B
a
tion
N 1 150 000 1 160 000 1 170 000 585 000 1 010 000
cracks A D F F F
Posi
b
N 680 000 1 820 000 750 000 260 000 565 000
For system previously four elements were tested with inter-
2
changing of their ends. All these tests finished with fatigue
cracks in places of tension connections of types A and B from
Fig. 3. On the base of obtained results the modification of
technology was recommended. Then next four tests with another four
elements were realized. Here arose new types of fatigue cracks
C, D and E, which were decisive for the failure of the whole System
(Fig. 3). Total numbers of cycles N and decisive cracks
are in Table 2.
Table 2. The results of fatigue tests of system 2.
Indi cati on MS1 MS2 MS3 MS4

cracks - - C C

I - 850 000
N
- 2 223 000 1

cracks - -
Element

C D

N II 3 380 000 3 690 000 - -

The knowledge and results of presented laboratory tests were
used for the determination of the total and remaining fatigue
life-time of verified temporary steel bridge Systems. Owing to
the small number of tests and their character, the obtained re-
could not be, however, sufficient and competent theoreti-
results
cal accesses. It means to apply them together with the Standards
data in the fatigue analysis.
FATIGUE STRENGTH AND LIFE-TIME
The tandards for the judgement of fatigue strength
antecedent s
and life-time of
ridge structures have considered a harmonic
b
loading with const ant amplitude. Also experimental fatigue verifi-
cation has been mo stly realized with this type of loading. But,
as we have mention ed previously, actual steel bridges are subjec-
ted to variable-am plitude loading.
Much activity has already been concentrated on the development of
a fatigue design m ethod for service variable-amplitude loading.
The result of it i s a definition of the effective constant-ampl i-
tude stress, which allows to use the constant-amplitude relation-
ships for the vari able-amplitude load conditions. Palmgren-Mi-
332 REMAINING CARRYING CAPACITY OF TEMPORARY STEEL BRIDGES

STEEL A
15 422.5 - FLANGES 30075100

tt
11 523.1 - WEB 3^-2 MSO/I
L*1
T

^2
^
SUPPORTING PART 3?? TESTED PART

375, A 5250
JJ& uP965 >¦
P
5410
6000 t-«t 6000

320

^Ä^1
# C

Hi
A-

A-
B

fcn

Fig. 2. Static and geometric scheme of test system 1. Positions

of fatigue cracks.

P P
crTLh ' ^

550
© | U550
®

^t-
2 700
ik_

^ -«
3000
^ ¦¦i
«!o |Y^t
6 600
1--
3000
x
»- ¦^
2700

"V <to
—1t- ^a
STEEL
15 422.5
_rfe
Ä B

y C

/v <io::;5:: A JA
B
400

Fig. 3. Static and geometric scheme of test system 2. Positions

of fatigue cracks.
 P. JUHÄS - E. JUHÄSOVÄ 333

ner's linear sumation is usually used for accounting the cumu-

lation of fatigue damage caused by variable-amplitude loading.
It is generally known that Palmgren-Miner's cumulative rule is
not sufficiently accurate. In our Institute the extensive
experimental program has been realized for verification of some
cumulative assumptions and fatigue design methods [3],[4],[8],[91.
The result of this program is a new conception for the determination
of the fatigue life-time.
In the case of constant harmonic loading the fatigue life-time
can be determined from experimental or Standard fatigue curve
concerning the most exposed place of structure. The stochastic
stationary loading can be expressed in simplified wayA using
loading spectra with different double stress amplitude ö*.j and
answering number of cycles n^. Considering constant harmonic
loading withAdmax we can state -from responsible fatigue curve the
minimal number of cycles Nmin. It means that according to the
prevailing influence of variable amplitude loading the total number
of cycles should be higher than Nm-jn.
N

We have used our obtained results like a basis for the new
determination of fatigue life-time using Nmi-n, Admax and aggressivity
of loading r :

.£ T^T A6max
A6_.
i n
i (1)
n(D
oc,1 - 1
(D
is i \* is
where 6 double stress amplitude, n./1x
max
maximum the num-
ber of cycles for individual loading levels and n(<\\ "is total
number of cycles in one block loading. * '
For constant harmonic loading r 1. Coefficient or^ 1
for
A6.j >A6q; and o^ <. 1
for Atf.j<A6o- 6q is defined as the
A
limit fatigue stress. Due to obtained test results when A6.|< L6q
answers. very well an- 0.8. Therefore we recommend for the
it
determination of total number of cycles N:
N
¦ IN
NA
A " Nmin/F
N
min
I~ ' (2)
where
k (1 + r2) (3.25 - r (3)
and the total life-time T.
TA " Vn(1) ¦ (4)
When calculating the life-time of the lower chord of the railway
steeUbridge [2J we can compare the results of different fatigue
conceptions T8], T9j, where
- authors: T. 10.506 years,
- Palmgren-Miner: TpM 9.721 years,
- SVUM: T$M 4.347 years,
- CSN 73 6205: Tg$N 7.130 years.
It can be seen that authors' fatigue life-time is larger than
that which is supposed due to valid Standard demands. Moreover,
from the analysis and comparison of
theoretical life-time, described
results
by number
it follows that the
of cycles N^ is more
dose to experimental results N than the theoretical ones NpjV|
334 REMAINING CARRYING CAPACITY OF TEMPORARY STEEL BRIDGES

or Ngjvj determined according hypothesis Pa Img ren-Mi ner and SVUM.

5. CONCLUSION
The total and remaini ng fatigue life-time is very relevant for the
reliable exploitation of assembled temporary steel bridges. The
fatigue of these structures depends on many factors as are in
particular actual service loading process, material characteristics,
geometrical parameters, constructional Solution, production-tech-
nological effects and defects. All these factors are, of course,
random variables. Therefore, the exact judgement of fatigue
strength and the determination of the life-time for such structures
is problematic also nowadays. In cases of original pretentious
steel bridges, as those presented in the paper, the experimental
verification is inevitable. The appropriate Simulation of actual
conditions and characteristics is, however, very important for
the reliable results. Our proposed approach to the total and
remaining fatigue life-time determination concerns variable multi-
-level loading. For general cases it can be determined using the
value of minimum life-time in the case of maximum loading together
with the application of aggression effect of the whole loading
spectrum. The fatigue life-time is then appropriately increased.
The new conception is accounting for the influence of general
loading process. It is more exacting than the application of
usual linear cumulation hypotheses.
REFERENCES
1. JUHÄS P. et al., Static-Dynamic Tests of Elements for Tempora¬
ry Steel Bridges. [Expert Report]. ICA SAS Bratislava, May 1987.
2. JUHÄS P. et al., Fatigue Tests of Elements for Temporary Steel
Bridges. [Expert Report]. ICA SAS Bratislava, Feb 1989.
3. JUHÄS P., New Cumulative Conception for Calculation of Fatigue
Life-Time. In: Proc. 10th Congr. Material Testing, Budapest 1991.
4. JUHÄS, P., Fatigue Life-Time of Steel Structures. Stav. Cas. 39,
No. 11/1991.
5. JUHÄSOVÄ E., Diagnostics of Dynamic Properties of Structures.
[Reseaech Report]. ICA Bratislava, Nov 1987.
SAS

6. JUHÄSOVÄ E., Seismic Effects on Structures. Elsevier, 1991.

7. E., Induced Vibrations from the Transport on Brati¬
JUHÄSOVÄ
slava In: Traffic Effects on Structures and
SNP Bridge.
Tatras, April 1991.
Environment, The Low
8. JUHÄS P. and JUHÄSOVÄ E., Fatigue Life-Time and Mechanism of
Failure of Thin-Walled Hybrid Steel Girders. In: Proc. Intern.
Symp. on Fatigue and Fracture in Steel and Concrete Structures.
SERC CSIR India, Madras, Dec 1991.
9. JUHÄSOVÄ E. and JUHÄS and the Fatigue Life-
P., Dynamic Loading
-Time of Bridges. In: Proc. of Conf. Theory and Construction
of Bridges, SvF VSDS Zilina, Donovaly, Low Tatras, June 1992.
10. JUHÄS P. and KRIVÄCEK J.: Experimental Verification of the
Fatigue Life-Time of Steel Bridges. In: Proc. of Conf. Steel
Bridges '91, CSOK and SUDOP Hradec Krälove, Nov 1991.
11. SMITH I.C.F., CASTIGLIONI CA. and KEATING P.B., Analysis of
Fatigue Recommendations Considering New Data. IABSE Proceedings,
P-137/89.
 Load Tests on Acute-Ang
Essai de Charge sur un p
Belastungsversuch an einer spitz

Bjarn Lykke JENSEN

Dr. Eng.
Axel Nielsen - Carl Bro as
Odense, Denmark

Soren Grubbe NIELSEN

Civil Eng.
Axel Nielsen - Carl Bro as
Odense, Denmark
336 LOAD TESTS ON ACUTE-ANGLED MOTORWAY BRIDGE

1. INTRODUCTION

The paper describes a load test performed on a six year old motorway bridge.
Originally the bridge was constructed as a part of a large ferry terminal project
at the Great Belt which separates the eastern and western part of Denmark. The
terminal was a part of a project in which both railway and roadtraffic were to be
shipped from the same terminal instead of from two separate terminals several
kilometers apart, as was the case until then. However, as a decision was made to
establish a fixed link across the Great Belt, the terminal plans were given up
during the construction of the bridge, and the bridge therefore never served its
real purpose.
As itturned out impossible to re-use the bridge as a part of the on-shore
constructions for the Great Belt crossing, the bridge was demolished just after the
load testing had ended.
The bridge was constructed using high quality materials, and thus, being practically
undeteriorated, it
offered a unique opportunity to perform a variety of
tests. These could e.g. through a demonstration of the real life behaviour indicate
possible needs for alterations in the current design practice and the basic
assumptions concerning the concrete properties.
Only tests in the serviceability limit State with live loads were considered, due
to limitations on time and budgets.

2. DESCRIPTION OF THE BRIDGE

The bridge consisted of a 2-span post-tensioned plate, simply supported on retaining

walls and columns. The angle of intersection was only about 18 degrees,
resulting in free edges spanning up to 40 metres. Prior to the load testing, the
concrete properties were measured by testing drilled-out cores. The average
strength was determined to be 40 MPa and the modulus of elasticity approx. 39.500
MPa.
The bridge is shown in fig 1.

10800

^Tj ^31

A-A
Fig 1 The bridge
 B.L. JENSEN - S

3. DESCRIPTION OF THE TESTS

The tests were carried out over two fü

weather was dry and sunny, with temperatur
two heavy vehicles from the Danish Road D
Army, respectively. The load characteristi
table 1.

ax

The Danish Road Directorate

Loaded trailer
The Danish Road Directorate
Volvo truck
The Royal Danish Army
Tank on tank-transporter
The Royal Danish Army
MAN Truck

*) With tank in normal driving position

Table 1 Load characteristics of the heavy

The vehicle from the Danish Road Directo

338 LOAD TESTS ON ACUTE-ANGLED MOTORWAY BRIDGE

By placing the tank directly rear axles of the tank transporter, a load
above the
on these two axles of about tons was obtained. Further, a relatively concentrated
63
load of more than 100 tons could be obtained by placing the two vehicles
in back-to-back in-line configuration.
a
The axle loads were measured with an aecuraey of + 50 kg and the vehicles could
be placed on the deck within an aecuraey of + 0.5 m.
13 tests were carried out with different combinations of the loads: 5 tests with
the DRD-truck alone, 6 tests with both vehicles positioned in-line next to the
edge beam and 2 tests in a side-by-side configuration. Furthermore, 6 reference
tests were performed without loading. These reference tests were performed mor-
ning, noon and evening, and were performed for calibration purposes, in order to
be able to compensate for temperature effects.

TANK WITH TRANSPORTER

*?*
^</

TRAILER FROM DANISH ROAD DIRECTORATE

Fig 3 Typical set-up of the heavy vehicles (Test no 18).

The following measurements were carried out:

(The position of the measuring points are given in fig 3)

- Strains at 19 points at the lower side of the deck. 3 components were

measured at each point. (points 1 - 19)

- Deflections of the edge beam (points 21 - 34), and at the points 15 20

- Strains at points at the upper side of the edge beam.

14 1 component (the
longitudinal one) was measured at each point. (points 21 - 34)

- Strains of unloaded parts of the construction (for reference purposes).

The strains were measured by DEMEC mechanical strain-gauges with an aeurraey of
10"5. The deflections were measured using a GEODIMETER theodolite.
Besides for calibration purposes, the unloaded tests served for the detection and
quantification of temperature effects. During testing the Upper side of the
bridge-deck was subject to a temperature rise of about 10°, which represented
potential temperature induced stains of the same order of magnitude as the maximum
load-induced strains. The temperature rise underneath the deck, on the other
hand, was only about 2°. By comparing morning and evening reference tests,
could be concluded that the bridge deck was subject to a slight compression in
it
the x-direction, and a corresponding slight expansion in the y-direction. The
magnitude of these strains was only 10 - 15 x 10"6, which is considerably smaller
 B.L. JENSEN - S.G. NIELSEN 339

than the maximun bending strains, thus leading to the conclusion, that the
rise leads to an increase in the internal stress level, rather than to
temperature
stains. Consequently, considering the procedure of calibrating against unloaded
tests, scattered througout the day, the temperature effects were neglected.
The reproduceability of the strain measurements was determined
by means of a
special test bar. Fluctuations within ± 5 x 10"6 were observed. The aecuraey of
the deflection-measurements was measured to + 0.75 mm.
Caption (strains)
10-6 10-6
120-j
100-
— ex Calculated 120

100
80- _ e xy
60- ex Measured
60-
40 40-
o e xy

^F—-i
-20- 20-
-40-
-60- 60-
-80-
-100- -100
-120-
-120
10 [4
point nc point no
10-6 mm
120

100

50-

-^
\>
10-

12-

100 18-
120 -20 '
26 28 30
point no point no
10-6 mm
lZO-i 2-i
100
0-
-2-
-4-
40-
-6
20-
^y^ -8
-10-
-20
-40 -12-
-14-

-16-
100 -18-
-20-
15 16 17 18 19 20
point no point no
Fig 4: Test Results. Measured (markers) and calculated (lines) results. The x-
axis indicates the number of the measuring point (see fig. 3).
(a - d: strains, e and f: deflections).
340 LOAD TESTS ON ACUTE-ANGLED MOTORWAY BRIDGE

4. RESULTS

The maximum edge beam deflection was measured to approx. 10 mm with the load con-
sisting of the the steel plate loaded DRD-truck only, and to approx. 17 mm when
both vehicles were present.
The corresponding strains were in the ränge of -50 x 10"6 to 60 x 10~6 and
-60 x 10~6 to 130 x 10"6, respectively.
The longitudinal strains measured at the Upper side of the edge beam were in the
ränge of -100 x 10"6 to 80 x 10"6.
No cracks were detected, even though the loads by far exeeded the level of the
serviceability limit state, and actually were close to the ULS design loads.
The results obtained from a selected test are shown in fig 4. Note that all
strains are 'true' strains, measured in three directions: ex parallel to the
retaining walls (- the x-direction), ey shifted an angle of -90 degrees with respect
to the x-direction and e^ shifted -45 degrees.
The test results have been compared to theoretical results provided through an
FEM-analysis, in which the bridge was represented by 300 triangulär plate
elements in each of the unloaded quadrants of the bridge, and by 600 elements in the
loaded section. The plate edges, i.e. the edge beam and the load distributing
beam at the bearing line, were modelled by additional beam elements. The curved
shape of the plate at the free edge was modelled using wedge shaped plate
elements

5. CONCLUSIONS

Fig 4 shows an exellent agreement between the measured and the calculated
results, thus indicating that the behaviour of bridges, even with very acute
intersection angles, may be successfully predicted by the use of Standard FEM
modelling. The good agreement also confirms, that the bridge was fully intact,
without sign of cracking.
Finally, the tests demonstrate that the mechanical strain gauge technique, which
was used, is a well suited tool, and provides a sufficient degree of aecuraey,
for the kind of load testing described in this paper.

6. AKNOWLEDGEMENTS

The was sponsored by The Danish Road Directorate and The Danish State
testing
Railways. Valuable contributions to the planning of the project was offered by
The Danish Engineering Academy, The Technical University of Denmark and Great
Belt Als.

REFERENCES

1. Load testing of motorway bridge at Knudshoved. Report prepared for the

Danish Road Directorate and the Danish State Railway. October 1992 (in
Danish).
 341

Field Testing of Existing Bridge on Remaining Load Capacity

Essais de Charge et capacite restante d'un pont existant
Feldversuch zur Resttragfähigkeit einer bestehenden Brücke

Hitoshi FURUTA
Nobuo NISHIMURA
Prof. of Civil Eng.
Prof. of Civil Eng.
Osaka University Kyoto University
Kyoto, Japan
Osaka, Japan

Nobuo Nishimura, born 1941, Hitoshi Furuta, born 1948, received

received his Dr. Eng. degree in
civil engineering from Osaka
University. He has held pro-
fessorship at Osaka University
from 1965.
his Dr. Eng. degree in civil
engineering from Kyoto University.
He has held prof essorship at Kyoto
University from 1976.

Atushi NANJYO Yoshihoko MINE

Civil Engineer Civil Engineer
Hanshin Expr. Public Corp. Yokogawa-Bridge Co.
Osaka, Japan Osaka, Japan

Atushi Nanjyo, born 1952, received
his M. Eng. degree in civil
engineering from Kyoto University.
For about forteen years he has
been engaged in designing or
repairing bridges.
Yoshihiko Mine, born 1947, received
his Civil Engineering degree at
Kyusyu University. For about
twenty years he has been engaged
in bridge design.

SUMMARY
This study is devoted to defining the load sharing of secondary members, such as lateral bracings, sway
bracings and concrete walls compared with the loading test results of an existing steel bridge.
Furthermore, it is examined how to evaluate the effects of secondary members of the structural load
capacity of the existing steel bridge in the modelling for structural analysis.

r£sum£
Cette etude a pout but de definir la repartition de la Charge du trafic sur les membres secondaires, tels
que contreventements, entretoises et murs en beton, par comparaison avec les resultats des essais de
Charge d'un pont metallique existant. Elle montre egalement comment estimer les effets des membres
secondaires sur la resistance du pont existant dans le cadre de l'analyse de la structure.

ZUSAMMENFASSUNG
Diese Studie ist der Ermittlung der Mitwirkung von sekundären Bauelementen wie seitlichen Verstrebungen,
Querversteifungen und Betonwänden unter Verkehrsbelastung gewidmet. Dazu werden Vergleiche
mit den Ergebnissen von Belastungsversuchen an einer bestehenden Stahlbrücke gezogen. Weiters wird
untersucht, wie die Auswirkung sekundärer Bauelemente auf die strukturelle Belastungskapazität der
bestehenden Brücke bei der analytischen Modellierung berücksichtigt werden kann.
342 FIELD TESTSTING OF EXISTING BRIDGE ON REMAINING LOAD CAPACITY

1. INTRODUCTION

In recent years, some of the composite plate-girder bridges on urban

highways in Japan have suffered fatigue damages which have occuered on the
connections between main girders and secondary members such as lateral
bracings and sway bracings0. Local corrosion have also occuered in the
secondary members. Furthermore, cracks and vehicular collision have caused
damages to the concrete wall parapets.

Lateral bracings and sway bracings are provided to ensure the lateral
stability of bridges while their erection work is in progress or after
the construction has been completed. Wall parapets are constructed to assure
the safety for vehicular falls from bridge floor. These damages, therefore, do
not reduce the reliability of steel bridges in direct relation to vehicular
traffic loads.
On the other hand, the secondary members form a three-dimensional structure
with the main girders, and contribute to load distribution50 3). They are
considered effective for increasing the structural load carring capacity of
the existing bridges. Failure to conduct periodical inspections and to perform
adequate repair or reinforcement of these secondary members may help
accelerate the damages, resulting in decreasing load carring capacity.
Through the field loading tests for an existing bridge, it was clarified that
an individual secondary members can share the traffic load. Furthermore, it was

examined how to evaluate the effects of secondary members on the structural

load carring capacity of the exsisting bridge by modelling the structural
analysis.

2.TESTED BRIDGE OVERVIEW

2. I Structure
steel bridge on the Hanshin expressway in the Umeda district in Osaka was
A

chosen as a testing object. The bridge is 28.73m in effective span and 7.5m in
total width.It is a simply-supported composite plate-girder bridge composed of
three main girders, five pairs Elevation and piain view
29 330
of intermediate sway bracings
placed at the distance of 8.5
approximatelly 4, 8m, two pairs
of end sway bracings each
MOV
Cross section U075 255
7500
S5QQ 250 URP4 URP5
*
r
Asphalt pavement
Concrete slab 8
^
r
11

7
i£JÄL>
G2 G3

*
G1

\m 265Q ¦ 2650 illOC 26 730

Fig. 1 General view of plate girder bridge

 N. NISHIMURA - H. FURUTA - A. NANJYO - Y. MINE 343

located at either girder end, and lateral bracings provided in a horizontal

plane near the bottom flangein the exterior bays. The concrete deck slab is
170mm thick, which is overlaid with 75mm thick asphalt pavement. Concrete wall
parapets of 250mm in thicknessand 1000mm in height are located alongside
the edge of the concrete slab. Figure illustrates the brige structure.1

2. 2 Historical background
This bridge has been served for vehicular traffic for approximately twenty-
four years since its completion in 1965. The urban redevelopment of this
district necessitated the traffic route to be altered, and this bridge to be
demolished. Before this bridge was pulled down,loading tests and other various
investigations were performed.
This bridge has not received many vehicular loads, because was a ramp which it
provided vehicular access to the expressway. Since was opened, the it
replacement of expansion joints was major maintenance work. Little
deterioration has occurred in this bridge, and this bridge has been maintained
in a sound condition.

3. TEST PROGRA»

3. 1
Structural Systems for loading tests
Figure 2
il lustrates
the structura 1 Systems for which loading tests were
carried out. A t first, loading test wa s performed to examine its load carrying
capacity after 24 years of service. Th is is called the loading test of system
1. Next, loadi ng tests evaluating the load share of secondary members such
as lateral bra cings, sway bracings and concrete wall parapets were carried out.
Those tests a re called the loading te st of system 2, 3 and 4, respectively.
In each system lateral bracings, s way bracings and wall parapets were
removed in a serial manner, and ide ntical loading tests were excuted. By
comparing the test results of each sys tem, it is possible to clarify the
extent of the load sharing at seconda ry members.
/Lateral braclnqs
3. Loading
y\ 25 y\ y\ X X
2
System 1
XX
Loading trucks were used as a means
of providing the required load.
XX \y \y \y V X
.Swav bracinqs
><

Figure 3 shows the dimensions of the

Unit( mm)
System 2
/
7 700 .Main girders
r1500

E a-
3250

^
c3
3
1300. 1650

C
C
3
u
System 3
z
s

^
CN
2010

_£ 3 CU
I .1900
- System 4 —-Wall parapets
j I.. E 3- 13 c3
50kN 150kN Concrete slab

Fig.3 Loading truck Fig. 2 Structural Systems for loading tests

344 FIELD TESTSTING OF EXISTING BRIDGE ON REMAINING LOAD CAPACITY

loading truck. Figure 4 shows examples Unit (mm)

of loading truck positions. Tests are 3900, 8100 .5000 8100 r3900.

arranged so that loading is placed dfo <k jBät <** <h

ffi
3
<|

symmetrically on the travelling lane and 1 r

in the passing lane, and also one lane 28A730

225 2000 1020 2000 1221

only as an extreme case.

ttt
T case LI
3.3 Measurement items
Deflections and strains in main girders
G3
are measured with regard to the following G1 G2

items: 3900,650
• Deflections of three main girders in
the center of the effective span.
• Strains in the bridge-axis direction of 28730
the bridge at both top and bottom 275Q

franges of main girders in the center m —xcase S1

of the effective span. Pv caseS2jj

4. ANALYSIS

Analysis
OF THE TESTED BRIDGE

condueted by using the four

was
G1

Fig.4 Loading truck positions

IT
G2 G3

modeis. Table shows the structural modeis

1

used for the analysis. Table 2 shows the

material properties of steel and concrete, Table 1
Structural modeis
used for the analysis. Models Gl, G2 and G3 Structural Lateral fall Sway
Model Bracings parapets bracings
are used for the grid-girder analysis. Gl N N C

Figure 5 illustrates the frame model used G2

G3 C
N C
C
C

for grid-girder analysis. Model Gl is a G4 C c C

structural model generally applied to the Note C

N
Considered
Not Considered
design of steel plate-girder bridges. In this
model, the effective width of the slab
complies with in the Japan
the Provision Table 2 Material properties
Specification for Highway Bridges4). Material
Steel
Young's aodulus
2 06x 10'UPa
Poisson's ratio
0 3
Model G2 includes the assumptions employed in Concrete 2 94x 104MPa 0 167

model Gl as well as adding the bending-

rigidity of concrete wall parapets to the
exterior girders. Model 3 includes G2' s
assumption, and in this model the torsional
rigidity increased due to the installation of Main girders
lateral bracings between the main girders is
added to the rigidity of main girders. Model F
is a model for the three-dimensional finite- Sway bracings
eleraent analysis.The mesh division of model F
is shown in Fig. 6. For concrete deck slab,
concrete wall parapets and webs of main
girders quadrilateral plate elements taking Fig. 5 Frame model
for grid-girder analysis
 N. NISHIMURA - H. FURUTA - A. NANJYO - Y. MINE 345

with m-plane and out-of-plane rigidities are used. Top and bottom flanges
of the main girders are divided into beam elements. The components of lateral
baracings, sway bracings, are also divided into beam elements.

Fig. 6 Mesh division of Model F

5. RESULTS OF LOADING TEST

5. 1 Load-carrying capacity in its existing condition

The measurement results of cross-
caSfiU
sectional stress distributions in
the main girders at the center • • Measured
of the effective span are shown ; Calculated

in Fig. 7. The mesured values are

caseLI t^-iLr
smaller than the calculated values
in design of this bridge, This
indicates that the bridge which
has served for twenty-four years -20 20 40 60-20 0 20 40 60 80-20 0 20 40 60
completely the specified
meets caseL2l~l
Stress(MPa)

requirements. Differences between

the measured values and calculated
values are presumably resulted
from the fact that effects of the -20 0 20 40 60-20 0 20 40 60 80-20 0 40 60 80
Stress(MPa)
secondary members on load sharing
were disregarded in design. Fig. 7 Stress distributions
in main girders
5. 2 Effects of secondary members on load distribution
Figure 8 shows the results of the loading tests condueted for the each
structural system.
Based on the test results of Systems 1 and 2, effects of lateral bracings
the
on load distribution are described. In the symmetrical loading test (case SI)
346 FIELD TESTSTING OF EXISTING BRIDGE ON REMAINING LOAD CAPACITY

the measurement result of system

se 51
1
is almost similar to that of [UK!§ -fl
system 2. In the unsymmetrical
loading test (case S2) the
lateral distribution of system Or*
G2 S3
c0# & S3

^
1
differs from that of system
2 In the structural system 2 « 2
the deflection of the exterior
z
Nv
girder(Gl) mcreases. Stress in N^
the bottom flange of girder Gl ^P
"
increases by about 10 percent ri
1
The elimmation lateral
of
bracings in system affects

^
xJX**1 r*i?y'

^
2

the lateral distribution. This

20±
indicates that considering the
effects of the lateral bracings O—o; System 1 a- --a ; System 2

on load sharing would be O- -o: System 3 • • ; System 4

necessary to accurately evaluate Fig. 8 Deflections and stresses

load carrying capacity of in each system
existing bridges.
For the comparison of the test results of system 2 with that of system 3, the
effects of mtermediate sway bracings on the load sharing are studied. In the
results of system 3 in which mtermediate sway barcmgs are elimmated,
deflection of interior girder(G2) increases in both symmetrical and eccentric
loading tests, and stresses in the bottom flange at the interior girder
increase by about 30 percent in comparison with the result in system 2. Hence,
the effects of mtermediate sway bracings on lateral distoribution are
significant. Based on the test result of system 4, the extent of the effect
of concrete wall parapets on load sharing is studied. The rate of the increase
in both deflection of main girders and stress in bottom flanges of the girders
without concrete wall parapets are noticeable. This proves that the concrete
wall parapets contribute greatly to the overrall stiffness of the bridge.

6. ANALYSIS RESULTS

Figure 9 shows the comparison between the calculat ed values and measured
values of the deflections of the main girders in the structural system 1.
Models Gl,G2 and G3 represent the results of the grid-girder analysis, and
model F represen ts results of the three-dimensional f mite element analysis.
Model Gl is gene rally used in the design of plate gi rder bridges With this
model, the calcu lated values greatly differ from the measured values. The
calculated value s for the deflection are approximat ely two times as much as
the measured val ues. The effects of concrete wall pa rapets are considered in
model 2 In th e symmetrical loading test(case LI) the calculated values
derived from the analysis using model 2 are much clos er to the measured values.
In the unsymmetr lcal loading test (caseL2), the calcu lated values differ from
the measured val ues.
 N. NISHIMURA - H. FURUTA - A. NANJYO - Y. MINE 347

In model the torsional rigidity increased due to the installation of

G3,
lateral bracings is considered. The calculated values at model G3 agree well
with the measured values, proving that the use of model G3 would enable the
accurate evaluation of the actual behavior of any existing girder bridge.
By using the model F, the better understanding of the structural behavior
of the bridge could be achieved, but such an analysis requires a considerable
set-up time and more expensible computational cost than model G3. Thus, the
analysis with the use of model G3 is more appropriate and more economical
method in the evaluation of the load carrying capacity of the existing bridges
with the effects of secondary members.
caseS2 caseSI
L L
ä= J\
J PJe 3
G3
•POr

^ ^
E
JE
2-
.o
U
V
Q
^4
Q
/
• Measured ;ModelG1 ;ModelG2
; Model G3 ; Model F
Fjjl _9 Comparisons between measured values
and calculated ones

7. CONCLUSIONS

From loading tests using an existing bridge, it can be seen that lateral
bracings and intermediate sway the load
bracings effectively act on
distribution, reducing the stress generated in the main girders, and concrete
wall parapets significantly contribute to the overrall stiffness of the bridge,
reducing the stress generated in the main girders. Therefore, the analysis
which also take of the effects of secondary members is required for the better
understanding of the structural behavior of steel bridges. Both grid-girder
analysis three-dimensional finite element analysis may be available
and as
methods to evaluate accurate load carrying capacity of existing bridges.

REFERENCES

1. A.NISHIMURA, K. HORIKAWA,H.TAKAYAMA,M.KUNIHIRO, H. HAYASHI, Repair of Fatigue-

Cracked Plate Girders in Highway Bridges, IABSE SYMPOSIUM LISBON 1989.
2. S. Analysis of Curved I-Girder Bridge with Lateral Bracing,
OSHITA,
Proceedings, JSSC, No. 324, 1982.8.(in Japanese)
3. AISC,Steel Bridges,The Best of Current Practice, Plate-Girder & Roiled
Beam Bridges
4. JAPN ROAD ASSOCIATION,Specification for Highway Bridges, 1989.(in Japanese)
348
 349

SESSION 5

CASE HISTORIES
350
 351

SELECTED PAPERS
352
 Testing and Modelling to Assess the
Exp6rimentation et modölisation pour eVa
Modellierung und Versuche zur Einschätzun

Charles ABDUNUR
Dr. Eng.
Lab. Central Ponts et Chaussees
Paris, France
354 TESTING AND MODELLING TO ASSESS THE CAPACITY OF BRIDGES

l.INTRODUCTION
The initial State and evolution of the stress profile is the major parameter of structural reserves in
prestressed concrete bridges. A reduced prestressing force may jeopardize their safety. In aging and
often damaged structures, the remaining prestress is very difficult to compute with a sufficient
aecuraey, owing to uncertainties about initial frictional losses, time-dependent hydric and visco-elastic
properties of concrete, bond redistribution, random steel corrosion failure and concrete cracks.
With the continuing development of this type of construction, several theoretical modeis were
successively proposed to predict prestress losses. These modeis need Validation or amendment in the
light of experiments involving direct stress measurement. In return, modelling may guide the
experiment and better interpret its results.
To evaluate the remaining capacity of structures, the present methods of field investigation do require
If
external test loading and indirectly estimate the corresponding stress. transversal concrete cracks
already exist, their adequate intrumentation would provide a better control of test loads. On the other
hand, in the absence of cracks and with no prior knowledge of the actual absolute stresses, test
loading may become arbitrary and destructive, somehow beating its own purpose. Hence it was
necessary to develop new parallel assessment methods that can do without test loading, measure
directly the existing stress and apply in both the presence and absence of individual cracks.

2. ASSESSMENT METHODS REQUIRING EXTERNAL TEST LOADS.

For continuous prestressed structures with flexural cracks, experimental computer-aided methods
were developed to assess the actual mechanical State and behaviour under dead and live loads.

2.1 Flexural Reserves under Dead Load

For a cracked section, under a convenient gradually increasing test load, figure 1 shows the tendon
over-tension Variation A oa, with the applied moment AM, [1]. This curve, reflecting a reinforced
concrete behaviour model, is a reliable criterion ofthe crack opening. Let X be the point on the curve
representing the State of the crack under the dead load of the tested structure ; let I be the crack
opening point. The load testing objeetif will then be to locate the position of X with respect to I, and
determine the bending moment reserves or deficit in the section. If the crack is completely closed
under dead load, X will lie on the straight portion around A ; the applied moment experimentally
necessary to open it (reaching I) gives the remaining flexural capacity. If the crack is already open, X
will lie on the curved portion. In this case, a general way is first to trace, for the same test load, the
theoretical strain-moment curve ofthe tendon ea /(M), similar to figure 1, using the above-
mentioned reinfored concrete model. Where the experimental curve eoineides with a portion ofthe
theoretical one, the origin ofthe former will be the point X and its relative position with respect to I
will give the flexural deficit.

A8a,Aoa
crack

closed open ^ FigJ.

Theoretical and experimental
X=B tendon strain and stress Variation
X=A with the applied bending moment.

AM M
theoritical reference experimental

2.2 Prediction of Flexural Stresses under Live Loads

In statically indeterminate structures, transversal cracks may be assimilated to a series of elastic or
plastic hinges, alternating with sound beam segments and jointhy setting up a new system in
equilibrium [2]. The hinge residual flexural stiffness is a multi-parameter function, subject to major
 C. ABDUNUR 355

(a) b El
[El] [i !!„<*> [El]

M,(
^ h h )M
£
-x ß 1
H B2 x
Fig.2 (a) Moments Mand stiffnesses Elin cracked sections H and sound segments B
(b) Redistribution of curvature 6' and stiffness EI
assumptions. An experimental evaluation was therefore adopted as follows : Under given test loads,
the curvature redistribution is accurately measured throughout the spans. With the flexural stiffnesses
ofthe sound segments usually known, a chain application ofthe classical beam equation leads to the
bending moments and, chiefly, to the required actual stiffness functions ofthe hinges (Fig 2 These
are introduced into a structural analysis program and the new statical system ofthe bridge is defined.
The actual flexural stresses can thus be predicted throughout the structure, under any given live loads.

3. THE CASE HISTORY

In the 1950's, France witnessed a construction wave of prestressed simply supported concrete girder
bridges, known by some as the "first generation". Confident of prior tests and careful workmanship,
designers pushed the still improving materials to the verge of their Performance limits.
Three decades later, hardly any concrete cracks were observed. However, the unfavourable effects of
time remained almost unknown on prestress loss, but clearly materialized in severe steel corrosion
with local failure of wires.
With a completely uncertaint stress profile and in the still lucky absence of concrete cracks, structural
assessment through extemal load testing connot be recommended for the afore-mentioned reasons.

4. NEW ASSESSMENT METHODS WITHOUT EXTERNAL LOADING

In a fundamentally different approach, we developed two new parallel methods of field investigation.
One can directly measure the actual stress in concrete, the other in steel. Both are now successfully
applied on site. No test loads are needed. Initial
& Compressive
Stress (Co)
4.1 Direct Stress Measurement in Concrete ;
The Release Method a Stress release
It is a local and partial release of stress, followed by a (stageA)
controlled pressure compensation [3]. In pratice (figure 3),
a displacement reference field is first set up on the concrete
M
surface ; a tiny slot, 4 mm wide, is then cut in a plane
normal to the desired stress direction ; finally, a special
very thin flat jack is introduced into the slot and used to (b)
restore the initial displacement field. The amount of
cancelling pressure gives the absolute compressive stress (c) %
normal to the slot. In the same way, with the same Applied pressure
aecuraey, tensile stresses are obtained by a corollary. The (stage B)
stress profile is traced by repeating the Operation at closely
successive depths of the same slot, then by treating the Re-estabüshment
data numerically. The depth operating ränge is 80 mm. In Id) (stage C~C0);P~CT0

spite of imposing a minute working scale, counter Fig. 3

measures kept the error within 0.3 MPa. Measurement is Stages A, B, C of direct stress
"direct" in the sense that the same physical quantity is measurement in concrete by the
involved (pressure for stress) and that none ofthe material release method
356 TESTING AND MODELLING TO ASSESS THE CAPACITY OF BRIDGES

elastic properties are needed. These are even determined in the process. Material eigenstresses can
now be correctly isolated and subtracted from the measured absolute stress, giving access to the
mechanical normal stress acting on the section. Besides the miniaturization imperative, a post-
operation specific remedial technique also restores the initial mechanical and esthetic State of the
medium. The release method can thus be classified as non-destructive.

Pinciple of stress measurement

in steel by the crossbow method 7 ////// A'o £NW ZZ
Wl NN
W
///</,

4.2 Stress Measurement in the tendons: tm^ »«

The Crossbow Method
The effort necessary to deflect a tight rope is obviously proportional to the axial tensile stress. Figure
4 outlines the application of this fact to the assessment of tendons after carefully Clearing the adjacent
concrete cover, duct and groute over a 60 cm length [4]. A controlled perpendicular force P, coupled
with a displacement sensor, deflects successive prestressing wires throught a distance f limited to 4
mm ; the tensile force F in the steel is deduced by the formula :
P - 2 (F + k) (W) + K (d)3
where k and K are given constants. In practice, the parasitie effects of friction, flexural stiffness and
overstretching necessitate prior calibration tests on simulating modeis in the laboratory.

4.3 Other remarks

The release method remains mechanically more comprehensive. Both methods, however, are
complementary. They enable now to reason directly in terms of stress thus providing an immediate
acces to the applied forces and moments and offering a straightforward comparison with the material
strength, a main criterion of structural safety.

5.APPLICATION TO THE CONSIDERED CASE

The two new assessment methods were jointly used to evaluate the remaining prestress in a large
number of the simply supported, post-tensioned concrete bridges described in chapter 3 and situated
in the north eastem part of France. These structures are 22 to 38 years old, but of identical design and
construction : same Standard T-sections and spans, same materials and prestressing system.

5.1 Procedure
Concrete stresses at midspan were measured on the neutral axis to deduce directly the actual
prestressing force. Steel stresses were evaluated near this section at convient points. Before and
during the campaign, thorough inspection was carried out to detect steel corrosion. Affected bridges
were left for a separate consideration. On the 12 remaining identical bridges still free from corrosion,
the stress values obtained by both methods were quite consistent and fully agreed. They were selected
for the following analysis on the mechanical loss of prestress.
5.2 Experimental Data Synthesis
In a homogeneous group of N bridges, solely differing in age, the N respective stresses measured at
a given date may be assumed equivalent to the measurements that could have been taken on a single
typical bridge at N consecutive dates of its lifetime. In our case, the 12 corresponding stress values
can thus form a synthesis curve representing an experimental time-dependent loss of prestress.
 C. ABDUNUR 357

percentage
remaining
prestress ^a^
100

instantaneous
loss
90
Ä
time-dependent
loss
80-

IP1 p 4%) (b)

Experimental points (a)

BPEL p 8%) (b)

20 40 60 (years)
age

Fig. 5
a) Directly measured remaining prestress at mid-span, on 12 identical bridges of different age and
with unknown steel relaxation p and friction coefficient ji.
b) Corresponding envelop theoretical curves obtained from two predictive modeis with different basic
assumptions and p values ; they help to extrapolate the experimental results in time and determine p
for the steel actually used ; experiment may in return improve the model assumptions.

In figure 5, these points are converted to forces over the section and expressed as a percentage ofthe
original anchorage prestress force FQ found in the records of each bridge. They lie between two
theoretical envelop curves, stemming from constitutive modeis discussed below, and guiding the
extrapolation ofthe present measurements towards the future (or the past).

6.THEORETICAL MODELS ON PRESTRESS LOSS

Three main frameworks of prestress loss estimation were proposed at sucessive periods of
construction development in France : pre-code recommendations, IP1 model, BPEL code.

6.1 Instantaneous losses

The effects of friction, elastic shortening and anchorage loss are all well known. The main uncertainty
concerns the values ofthe friction and wobble coefficients, u and 4>, in the exponential formula which
all modeis use for frictional losses.

6.2 Time-dependent losses

On prestress losses due to concrete shrinkage and creep or steel relaxation, the three modeis diverge
as follows.
* The pre-code of thel950's, construction period ofthe bridges now under scrutiny, gives a lump
sum estimation of 12 to 15 % prestress loss in all, after the instantaneous effects [5]. It assumes
proper overtensioning to reduce steel relaxation and overcome friction and anchorage losses.
* In the IP1 model ofthe 1960's, based on the concept of allowable stresses, a group of constitutive
equations accounts for the following sources of loss [6]:
- Concrete shrinkage, depending on the mix and thickness, the atmospheric humidity and a time-
dependent strain coefficient.
358 TESTING AND MODELLING TO ASSESS THE CAPACITY OF BRIDGES

- Concrete creep, as a function of the initial elastic strain, atmospheric humidity, time-dependent
strain coefficient and other constitutive factors.
- Steel relaxation, depending on the initial steel stress o i, strength Rg and percentage stress loss p at
1000 and 3000 hours in a Standard relaxation test.
The main assumption in IP1 is the simple straightforward superposition of these three.
* The B.P.E.L., a more elaborate model ofthe 1980's, is based on the service and ultimate State
limits [7]. Compared with the IP1, there is hardly any difference as far as concrete shrinkage. For
concrete creep, in a new BPEL Version : The effect of atmospheric humidity is more accurately
formulated in the basic law ; an additional but different law was introduced for "creep recovery" ;
a new "equivalent time" method was developed to replace the superposition principle.
Concerning steel relaxation, the same IP1 parameters were kept in a slightly different formula, less
optimistic for medium relaxation steel used in the bridges under consideration.
BPEL's basic assumption is the partial interaction between concrete creep and steel relaxation.

7.MODEL-AIDED INTERPRETATION OF EXPERIMENTAL RESULTS

For both the IP1 and BPEL predictive modeis, detailed Computer programs were elaborated [8].
Given the material properties, they calculate all the prestress loss compoments in an exact step-by-
step Simulation of the bridge construction phases and loading history. For the assessed bridges, the
properties related to the calculation of concrete shrinkage and creep were either given or deduced from
the design and construction records. However, the percentage steel relaxation at 1000 hours, p, did
not exist as such in the early 1950's ; it was actually very high but masked by short-time tests.
Another uncertainty was p, the prestress friction coefficient. In a parametric study using both modeis,
p was varied from 4 to 8% (extreme value), and p from 0.16 to 0.23. The analysis showed that p had
an effect several times greater than p on the total prestress loss. Hence, we thought of fixing p at
0.23, corresponding statistically to the type of ducts used. Left with one influential unknown material
parameter p, it was not possible to have a clear-cut comparison between experimental and theoretical
results but rather a Joint contribution : the release method giving the present actual prestress force, the
two modeis guiding its extrapolation to the future under a p Variation ränge suggested by previous
data synthesis and living memory. In this context, the adopted Interpretation strategy is illustrated in
figure 5 for the 12 assessed bridges, then explained in table 1 by an example of a typical 30-year-old
bridge. In the example, Interpretation proceeds as follows :
1) From the anchorage prestress force FG given in construction records, calculate by each model the
remaining mid-span force F at infinity (°° 60 years), using two envelop values of p and the adopted
constant value of p. (Ml, M2 and M4, M5 in table 1).
2) Deduce the present force F (30 years), assuming a continuons quasi-asymptotic function of time.
3) Compare the present experimental value of force F with the two theoretical ones of each model.
4) By identification or interpolation, find in each model a particular theoretical function that agrees
with the present measurement (5500 KN at 30 years). The functions obtained nearly coincide and can
predict the prestress losses still expected.
Thus, at 30 years, the prestress ofthe bridge, reflecting its structural capacity, has already actually
suffered a 32 % total loss, expected to attain 38 %. Of these, 27 % are time-dependent contrasting
with the 15 % estimated in the design. The two particular functions, satisfying the 30-year stress
measurement and giving the same losses, correspond to p 8% for IP1 and only 6.27 % for BPEL.
Despite the lessening effect of its partial interaction between concrete creep and steel relaxation, BPEL
seems then to grow passimistic for p values exceeding the 2 % of our present low relaxation steel.
Finally we notice practically stable values of instantaneous losses, 10-11 %, appearing in the
corresponding differences between the last two columns of table 1.

8. UP-TO-DATE ACTIONS ON THE STRUCTURE

With its remaining capacity evaluated, the structure must face external loads not necessarily equivalent
to those its was designed for. The Eurocode will thus be proposing two partial load Systems :
- A double-axle concentrated load model (indivisible tandem system), each axle having a weight Qk
 C. ABDUNUR 359

Means of Evaluation Prestress Steel Concrete Time-dept Total

Force Stress Stress % Loss % loss
modelling ^^^^ F(KN) öa(MPa) ob (MPa) 1 - F/Fi 1 - F/F0

^^^^^ Experiment 30 y oo 30 y oo 30 y oo 30 y oo 30 y oo

Pre-code: M 0
approximation 6430 5900 994 912 9.0 8.3 8 15 20 27
(design)
IP1 model: M 1

5500 5040 850 779 7.7 7.1 21 28 32 38

p - 8 % ; ju 0.23

IP1 model :M2 5800 5400 897 835 8.1 7.6 17 22 28 33

p 4 % ; u 0.23

IP1 model: M 3
5600 5090 866 787 7.8 7.1 19 27 31 37
p 8% ; ju 0.16
BPEL code : M 4
5340 4760 826 736 7.5 6.7 23 32 34 41
p 8 % ; u 0.23

BPEL code : M 5
5710 5260 884 814 8.0 7.4 19 24 29 35
p 4 % ; u 0.23
EXPERIMENT
(Direct Stress 5500 — 850 — 7.7 — 21 — 32 —
measurement)

Identification 5500 5040 850 779 7.7 1A 21 28 32 38

with M 1

Interpolation
M4-M5 5500 4980 850 769 1.1 7.0 21 28 32 38
p - 6.27 %

Conclusion : The bridge may still lose 10 % of its present prestress.

Table 1 : Example of an experimental model-aided assessment of present and future prestress losses
in a bridge, 30 years old at the testing date. Underlined quantities are experimental values
extrapolated to infinity using these modeis (p is the steel relaxation at 1000 hours, p the
friction coefficient, °° assumed at 60 years. F is the present and future prestress force at
mid-span ; Fi and F0 are the initial forces respectively at mid-span and the anchorage).

equal to 300, 200, 100, KN on lane 1, 2, 3, respectively.

- A uniformly distributed load (UKL system), having a weight density q^ per Square metre, applied in
the unfavourable areas ; cfc is assumed 9, 2.5, 2.5, 2.5 KN/m2on lane 1, 2, 3, elsewhere.

10. RANDOM EFFECTS OF CORROSION

On the second group of bridges, the experimental results are far less consistent. The impact of steel
corrosion failure, on the remaining prestress in nearby sections, varies from negligible to
considerable, depending on the ability ofthe broken wires to develop local friction anchorages in both
senses (fig. 6). This random factor is still difficult to model. Direct stress measurement, by the
release method in particular, remains the only unchallenged means of assessment for the time being.
360 TESTING AND MODELLING TO ASSESS THE CAPACITY OF BRIDGES

percentage
remaining A^
prestress
100 Fo

instantaneous
loss
90
o apparently sound
+ corroded
80-

70
08
* +
60 —t— —P-
20 40 60 (years)
age
Fig-6
Effect of steel corrosion on the remaining prestress of otherwise identical bridges.
Both groups seem to have undergone similar time -dependent losses.

10. CONCLUSION
The present actual structural reserves of a prestressed concrete bridge can now be evaluated through
direct stress measurements by the release method. A model-aided extrapolation of these stresses gives
the time-dependent losses still expected and, hence, the future residual capacity. Proposed load
modeis, such as "Eurocode 1", will be contributing to an up-to-date estimation ofthe total required
capacity of a structure. The difference between the required and remaining reserves leads obviously
to the needed additional external prestress. At a later stage, stress can again be measured to check the
effectiveness ofthe Operation. The release method on one hand, two predictive modeis on the other,
can thus jointly contribute to an Optimum strengthening of aging bridges, which inadequate
prestressing has rendered structurally deficient or obsolete.

ACKNOWLEDGEMENT
The author expresses his thanks to Y. Jaffre and P. Paillusseau (S.E.T.R.A) for their help in the
computations regarding the application ofthe IP1 and BPEL modeis.

REFERENCES

1. CHATELAIN, J. and GODART, B., "Evaluation de l'etat mecanique reel de ponts en beton
precontraint". IABSE Symposium, Helsinki, 1988.
2. ABDUNUR, C, "Structural assessment of a bridge with transversal cracks", First International
Conference or Bridge Management, GUILDFORD, UK, 1990.
3. ABDUNUR, C, "Mesure de contraintes sur ouvrages d'art par une methode de liberation
miniaturisee", Bulletin des Liaison des LPC, n° 138, Paris, 1985.
4. TONNOIR, B., "Mesure de la tension residuelle dans les armatures de precontrainte ä l'aide de
Farbalette", journee portes ouvertes au L.R.P.C de Lille, 1991.
5. DUMAS, F., "L'evolution de la precontrainte du beton arme au cours de la reconstruction des
O.A. du Nord et du Pas-de-Calais", revue "Travaux", p. 588-576, octobre 1951.
6. "Instructions Provisoires sur Pemploi du beton precontraint", Paris, 1965.
7. "Regles techniques de conception et de calcul des ouvrages et constructions en beton precontraint
suivant la methode des etats-limites, Paris, 1983.
8. Dossier "VIPP", S.E.T.R.A., France, 1967.
jp4

Reliability Aalysis of
Analyse de la fiabilit.6
Zuverlässigkeitsanalyse fü

R. RANGANATHAN
Prof. of Civil Eng.
Indian Inst, of Technology
Bombay, India

VL

Zubin MISTRY
Struct. Engineer
Gherzi Eastern Ltd
Bombay, India
362 RELIABILITY ANALYSIS OF A HIGHWAY BRIDGE

1. INTRODUCTION
The development of reliability analysis and reliability based design criteria has
led to the rational evaluation of structural safety. One area of reliability
analysis which is receiving considerable attention from engineers is System
reliability. The major source of hidden safety reserve is in the system behaviour
of structures. A bridge is a system of interacting members with redundancies and
load sharing. However, there is a different degree of redundancy and load sharing
in different types of bridges. The difference between member and system
reliabilities increases with increase in the level of redundancy. This paper uses
a probability based approach, developed by Nowak and Zhou [1], considering a bridge
as a system of interconnected elements. System approach has been used to evaluate
the reliability of a continuous prestressed concrete bridge which is under
construction across Thana Creek in Bombay, India. For the evaluation of the
reliability of a bridge,be different limit states of flexure, shear, fatigue,
deflection, etc. are to considered. However, only the limit states of collapse
in flexure and shear are considered and the evaluation of structural reliability
is presented for the same using Monte Carlo technique. Field data on the strengths
of materials and traffic load have been collected and statistically analysed.
Results of the same have been used in the reliability analysis.

2. DETAILS OF THE BRIDGE

The new Thana Creek Bridge is a prestressed concrete box girder bridge of total
length 1.837 km. It
consists of six spans viz. 205.46 m, four spans of 321 m each
and 347.63 m. It
is a Class AA bridge [2] with three lanes on each of the two
parallel continuous girders. From the structural analysis point of view, the two
girders may be considered to act independently of each other. Each span consists
of a variable depth box girder - a Square parabola from either support upto the
midspan. The girder is supported on steel rocker/roller bearings on each pier. A
longitudinal section of a typical intermediate unit and cross section are shown in
Fig. 1. Details of eritieal sections are shown in Table 1.

Section Area of Moment Number of Slope of Area of Centroid

No. section of prestressing duet stirrups distance
inertia cable duets in <rads> from top
<m2> <m4> deck soffit <mm2/m> <m>

1 7.4660 13.511 2 6 0.0000 4400.83 1234.21

2 9.3258 36.725 36 6 0.1000 4997.73 2040.96
3 14.470 113.11 58 0 0.1315 6544.93 3832.94
4 9.3258 36.725 40 4 0.1000 4581.47 2040.96
5 7.2300 13.262 2 38 0.0000 4400.83 1215.74
6 9.3258 36.725 40 4 0.1000 4581.47 2040.96
7 14.470 113.11 64 0 0.1315 6544.93 3832.94

Table 1 Section properties for the Thana creek bridge

3. STATISTICS OF STRENGTH AND LOAD VARIABLES

Concrete mix M 40 (cube strength 40 N/mm2) and 12.7 mm diameter, 1/7 ungalvanised
stress relieved Strands for prestressing the bridge and Torsteel (yield strength
415 N/mm2) for vertical stirrups have been used in the bridge. Field data have been
collected on cube strength of concrete, f^, ultimate tensile strength of Strands,
fp, and yield strength of Torsteel, fsy, and statistically analysed. It is found that
fw, fp and f,y follow normal distributions with mean value and Standard deviation
equal to 47.28 and 4.28, 1925.8 and 24.5, and 468.9 and 34.2 N/mm2 respectively
 R. RANGANATHAN - Z. MISTRY 363

r-3-5 r-5-0 r-5'0 r-3-5 70

© ®

n ® ®
®
©
®
®<g) n
16-5
!2lij u
*
350
^
»4« 350 -I 350 i 350

53-5 107 0 1070 53-5

321-0
All dimensions in m
(a) Typical mtermediate unit

i-i 12300
Sections h tw tb
325 1,5,9,13 3500 300 225
*-tw
3,7,11 7000 350 1000

2,4, 6, 8 5000 363 335

tb 10,12
6500
All dimensions in mm

(b) Details of cross sections

Fia.1 Details of new Thana Creek Bridge

[3].
The live load data has been collected for the old Thana Creek Bridge as the bridge
under reliability study is still
under construction. The last traffic survey was
condueted in 1989. The collected data has been statistically analysed. Based on the
analysis of the actual loads, a new Standard truck has been fixed with a mean axle
load of 92 kN per axle (total load 184 kN) with a wheel base of 1.2 m. Considering
the uncertainties due to several factors, the coefficient of Variation of the load
is found to be 30 per cent [3]. The Standard truck obtained above represents the
arbitrary point-in-time varying load, Qapl, as obtained from the results of the
traffic survey. For checking the reliability of the bridge at ultimate limit
states, statistics of the lifetime maximum live load have to be used. Considering
the life of the bridge as 50 years, and assuming that (i) Q,^ follows Type I
extremal (largest) distribution [4], (ii) the live load data represents a constant
traffic distribution throughout the year and (iii) all annual maximum loads are
identically distributed and independent, the statistics of the lifetime maximum
total load, Qm, for the Standard truck have been obtained. It is found that Qm
follows Type I extremal (largest) distribution with parameters u= 327.5 kN and a
364 RELIABILITY ANALYSIS OF A HIGHWAY BRIDGE

* 0.232. The mean value of Qm is [3].

352.4 kN
The dead load,D, is assumed to be normally distributed with mean to nominal ratio
equal to 1.05 and coefficient of Variation five per cent.

4. RELIABILITY ANALYSIS
Having determined the statistics of the basic random variables, both load and
strength, the reliability analysis of the bridge is now considered. The criteria
selected are (i) the limit state of collapse in flexure and (ii) the limit state
of collapse in shear.
4.1 Limit state of collapse in flexure
The given bridge is a redundant structure. In the analysis of such structures, the
failure of a section is assumed to take place when the plastic moment capacity of
the section is reached. The failure is called the formation of a plastic hinge at
the section. In the case of a redundant structure, a collapse mechanism forms only
when a sufficient number of plastic hinges have developed.

4.1.1 Plastic rotation capacity and plastic moment

The failure of a structure through the formation of a collapse mechanism requires
the plastic hinges to have a large rotational capacity. In the case of prestressed
concrete structures, any eritieal region, hinged earlier, may fail due to
insufficient rotation capacity before a collapse mechanism is formed. This failure
is called a rotation failure mode. The plastic rotation capacity, 6p, of a section
is given by
8p d{eu - ecy)/xu (1)

where d is the effective depth, eu is the ultimate strain in concrete and e^ is the
strain in concrete at the extreme fibre at the start of yield in steel. The depth
of the neutral axis at failure, xu, can be determined from the equilibrium
condition at failure. As per IS:456-1978 [5], the limiting values eu= 0.0035 and
e^* 0.002 are used. Thus, knowing the geometry of the section and the depth of the
neutral axis at failure, the ultimate rotation capacity of the section can be
found.
The ultimate resisting moment of a section is taken as the plastic moment capacity,
Mp, of the section. This is given by [5]
Mp B fpuÄp (d - 0.42xu) (2)

where f^ is the ultimate tensile stress in the tendons at failure and Ap is the
area of prestressing tendons. Using equilibrium equations, strain compatibility
conditions and the stress-strain relationships for steel and concrete, a given
section can be analysed and f^ and xu can be determined. Knowing f^ and xu, Mp can
be calculated.In Eq. 2, the model parameter, B, introduced to take care of
uncertainty in the prediction equation, is assumed to be normally distributed with
mean value 1.01 and Standard deviation equal to 0.0465 [6].

4.1.2 Determination of reliability

The reliability analysis of the bridge as a system Starts with the determination
of eritieal sections in the bridge. The Standard truck is placed at a eritieal
section. Using the Monte Carlo technique, a set of values is generated for the
basic random variables viz. load and strength variables (f^, fp, Qm, B and D). All
other parameters are considered to be deterministic. Using the generated values,
 R. RANGANATHAN - Z. MISTRY 365

plastic moments and plastic rotation capacities of eritieal sections are calculated
as explained in the previous section.
For a given position of the wheel load, the structure is analysed using the
stiffness method of linear elastic and pieeewise linear elastic-plastic analysis
[7]. The Steps involved are as follows:
i) Develop the load vector {L}, and the stiffness matrix, [K], for the structure.
ii) Invert the stiffness matrix and obtain the displacement vector {D> as
iD) [IC]'1 [L] (3)

iii) Determine the member end forces and check

locations where plastic hinges have formed.
for excessive rotation at all
iv) Find the load factor, X, the ratio of the plastic moment capacity to the actual
moment, at each eritieal location. Select Xnin and increase the load by multiplying
the initial load by Xnin.
v) Introduce a plastic hinge at the node for which X X^ and modify [K] To do
this, add a row and a column to [K], the elements of which are the stiffnesses
corresponding to the rotational degree of freedom at the hinge point. Add the
plastic moment capacity of the section and the hinge rotation to the load and
displacement vectors respectively. Check for the formation of a mechanism.
vi) Repeat Steps (ii) to (v) until the check in step (iii) fails or a mechanism is
formed. A mechanism is said to have formed when the determinant of the matrix,
|[K]| < 0.
vii) The ultimate load factor, X^ will be the sum of the minimum values of X in
successive cycles. The above procedure results in a pieeewise linear
moment-curvature relationship. Hence, during the analysis, if at any stage it is
found that the rotation capacity of a section is exceeded, then a direct
interpolation will suffice to obtain X and hence A.^.
The safety margin, Z, is calculated as
„ _ resistance _ ultimate load _ * 4*
action actual load ult

Hence a value for Z is obtained. The Monte Carlo technique is repeated and a number
of values are generated for Z. As the elastic-plastic analysis is to be carried out
each time, the number of samples to be generated is restricted to only 200 to get
estimates of mean and Standard deviation of Z.
For the generated samples, mean value, Z, and coefficient of Variation,5 z, of Z
are calculated. For calculating reliability, the lognormal format has been adopted.
Hence for the lognormal variate Z, the parameters are calculated as follows:

olnZ ^ln(ö| + 1) (5>

Z 2 exp(-| o\nZ) (6)

The probability of failure, pf, and the reliability index, ß, are given by

P/ *[ ln(i)/alnz] (7)

ß -fc-Mp^) (8)

This gives the probability of failure of the bridge for a given load position.
There are four eritieal load positions (Table 2). For each eritieal load position,
the above procedure is repeated and pf and ß are calculated. The results are given
in Table 2. Since the values of pf «
1, the probability of failure of the system,
pf8, has been taken as
366 RELIABILITY ANALYSIS OF A HIGHWAY BRIDGE

Pf. £"< (9)

assuming the failure modes to be independent. pfl is the value of pf for i* load
position and N is the number of eritieal load positions. The value of pft is also
given in Table 2.

Load position Z °z Pf ß
Section No.
14 4.6643 0.8716 1.018 x 1016 8.2191
4 4.6418 0.8581 6.159 x 1017 8.2827
5 4.3353 0.8174 4.538 x 1015 7.7548
6 4.8220 0.9152 6.537 x 1017 8.2691
For system 4.77 x 1015 7.75

Table 2 Results of analysis for limit State of collapse in flexure

4.2 Limit state of collapse in shear

4.2.1 Ultimate Shear Strength
The analysis for the limit state of collapse in shear differs from the elastic-
plastic analysis for flexure in one major respect. There is no successive formation
of hinges as was considered in the previous limit state (See. 4.1). This is because
once a section fails in shear, the entire unit is considered to have failed. The
ultimate shear strength of a prestressed concrete girder is considered to be the
lesser of its strengths in the cracked and uncracked states. It is calculated using
the Indian Roads Congress code [8]. The equations in the code are used after
removing partial safety factors. While estimating the ultimate shear strength of
a section, a model error parameter, B, is attached to the prediction equation and
this is assumed to be normally distributed with a mean value of 1.09 and a Standard
deviation of 0.14 [6]. The analysis of a girder at the limit state of collapse in
shear is simpler than that at the limit state of collapse in flexure.

4.2.2 Determination of reliability

The Standard truck wheel load is fixed at a eritieal section. Using the Monte Carlo
technique, a set of values is generated for the basic random variables viz. load
and strength variables (fw, Qm, B, f8y and D). Using the generated values, the
ultimate shear strength of each eritieal section is calculated. For a given load
position, elastic analysis of the bridge is carried out using stiffness matrix
method and values of actual shear forces at eritieal sections are determined. The
load factor, X, which is the ratio of ultimate shear strength to actual shear, is
calculated at each eritieal section. The lowest value, X^, is selected and this
gives the load factor for the ultimate shear capacity of the weakest section of the
girder to be reached. The safety margin is calculated as

z lesi3tcLnce _ «
(10)
min
action
This Xmin is nothing but a value for the safety margin Z. Similar ly, using the Monte
Carlo technique, the-process is repeated and a number of values of Z are generated.
Mean and Standard deviation of Z are calculated. Values of pf and ß are calculated
using Eqs. 7 and 8.
There are seven eritieal load positions (Table 3). For each eritieal load position,
the above procedure is repeated and probability of failure calculated. The
probability of failure of the system is calculated using Eq. 9. The results of the
ßk R. RANGANATHAN - Z. MISTRY 367

reliability analysis are given in Table 3.

Load position Z °z Pf ß Failure at

Section No. section
1 2.9256 0.4420 7.60 x 10" 7.0707 1, 6 or 8
2 2.9203 0.4410 8.33 x lO'" 7.0618 2 or 6
3 2.9257 0.4419 7.59 x lO"" 7.0728 6 or 8
4 2.9237 0.4424 9.00 x 10" 7.0553 4
5 2.9296 0.4437 8.33 x 10" 7.0621 6 or 8
6 2.9095 0.4393 9.61 x 10" 7.0383 6 or 10
7 2.9256 0.4418 7.57 x 10" 7.0739 6 or 8

For system 5.81 x 1010 6.14

Table 3 Results of analysis for limit State of collapse in shear

5. DISCUSSION AND CONCLUSION

With the available data, a fairly accurate Standard truck has been developed. The
mean value of the lifetime maximum live (truck) load has been found to be 352.4 kN.
It is observed that this value is very close to the Standard wheeled vehicle of 400
kN (200 kN per axle) specified by Indian Roads Congress [2] for Class AA bridges.
During the reliability analysis at limit state of collapse in flexure, it
has been
observed that the failure in each case is in the rotation failure mode. Failures
take place before the last hinge is formed to cause a complete failure mechanism.
A study of the results of the reliability analysis at limit State of collapse in
flexure given in Table 2, reveals a slight decrease in the reliability index when
the load is over the eritieal section 5 (Fig. 1). This is due to the special method
of analysis and design used for the bridge. The bridge is being constructed by the
cantilever method. As a result, during construction, the dead load due to self-
weight is carried by cantilever action. Even after the fixation of the continuity
cables at midspan, only the live load and the super-imposed dead load are carried
by continuous beam action. Only after a period of 10 to 12 years, due to the
relaxation of steel and the creep of concrete, the dead load will be carried by
continuous beam action. As a result, there is an excess of strength in flexure at
the supports. When the load is over the eritieal section 5, the moment at midspan,
the weakest section of the girder, is more than that for the other load positions.
Hence the reliability index for this position is lower than that for the rest. The
reliability index for the limitbe State of collapse in flexure, for the1015)
girder unit
taken as a system, is seen to 7.75 (corresponding to pf 4.77 x as shown
in Table 2. This is much higher than the acceptable value of 2.5 to 3, normally
used abroad. Thus, as expected, there appears to be an excess reserve of strength
in the girder unit.
A study of Table 3, giving values of reliability index at limit state of collapse
in shear, does not reveal any Sharp variations in the reliability index as was
observed in the case of flexure. This is because the failure in each case is
predominantly at one of the 5 m deep sections, close to the interior supports, viz.
2, 4, 6, 8 or 10 (Fig. 1). The reliability index for the limit State of collapse
in shear, for the girder unit taken as a system, is found to be 6.137
(corresponding to pf 5.81 x 10'10) This is higher than the generally accepted
value of 2.5 to 3.0. The reason for this is the same as that for the limit state
of collapse in flexure, viz. the use of the cantilever method of construction for
the bridge.
Within the limitations of the present study, it
can be concluded that the bridge
has high reliability (reliability index greater than 6; probability of failure less
than 10"9) at limit states of collapse in flexure and shear. Reliability of the
bridge in fatigue is an important aspect. Currently the state of knowledge in
368 RELIABILITY ANALYSIS OF A HIGHWAY BRIDGE

fatigue evaluation of concrete structures is not adequate to carry out reliability

studies.

7. ACKNOWLEDGEMENTS
The authors gratefully acknowledge the assistance given by CIDCO, New Bombay,
India, in providing the traffic survey data for the old Thana Creek Bridge and the
Maharashtra Public Works Department, Bombay, in furnishing data on the test results
of concrete cubes and steel Strands. Thanks are also due to STUP Consultants
Limited, Bombay, for providing the structural details of the bridge.

REFERENCES

1. NOWAK, A. S. and ZHOU, J. H. Reliability Models for Bridge Analysis.

Report No. FHWA/OH-85/005, Department of Civil Engineering, Michigan,
March 1985.
2. IRC:6-1966-SECTION II,
Standard Specifications and Code of Practice for Road
Bridges - Loads and Stresses. Indian Roads Congress, New Delhi, 1967.
3. MISTRY, Z. A. Reliability Based Design of Bridges. M.Tech. Thesis,
Department of Civil Engineering, Indian Institute of Technology, January
1992.
4. BENJAMIN, J. R. and CORNELL, C. A. Probability, Statistics and Decision for
Civil Engineers. McGraw Hill, New York, 1970, pp. 273-274.
5. IS:1343-1980, Code of Practice for Prestressed Concrete, Bureau of Indian
Standards, New Delhi, 1981.
6. ELLINGWOOD, B. R. GALOMBOS, T. V., McGREGOR, J. G. and CORNELL, C. A.
Development of a Probability Based Load Criterion for American National
Standard A58. NBS Special Publication 577, U.S. Department of Commerce,
Washington, D.C., June 1980.
7. MAJID, K. I., Nonlinear Structures. Butterworths, London, 1972.
8. IRC:18-1985, Design Criteria for Prestressed Concrete Road Bridges (Post
-tensioned Concrete). Indian Roads Congress, New Delhi, 1986.
 Remaining Structural Capacitie
Evaluation de la capacit.6 restant
Verbleibende Tragwerksfestigkeit v

Makoto SHIMAMURA Masao SUGIDATE

Chief Researcher Constr. Supervisio
East Japan Railway Co. East Japan Railwa
Tokyo, Japan Tokyo, Japan

ifMh,i,

Makoto ABE
Researcher
Railway Techn. Res. Institute
Tokyo, Japan
370 REMAINING STRUCTURAL CAPACITIES OF STEEL RAILWAY BRIDGES

1. INTRODUCTION

Maintenance and management of steel railway bridges in Japan is faced with two
Problems. One is a fatigue problem as exemplified in Tokaido Shinkansen. The
other is that the average age of steel railways is higher than 60 years-the
design age. Both problems are based on fatigue. Evaluating remaining life is
necessary. This remaining life evaluation was not generalized sufficiently.
Recently, however, remaining life evaluation procedures and Standards also have
been developed. In addition, development of Systems to process such evaluation
has been made. This report is an outline of a fatigue damage evaluation system
developed by Railway Technical Institute and BMC Corp.

2. DEFINITION OF SERVICE LIFE

Remaining life evaluation isin relation to the following two points.

made
(1)"Structural life" the end of its service life as
when a main member reaches
a result of repetition of nominal stress.
(2)"The time of occurrence of fatigue crack" when members suffer fatigue damage
due to local stress.
There are several definitions of "service life" which determines a period of time
when a steel railway bridge can survive. According to one theory, service life
expires when a bridge has suffered a fatal damage, economically and physically
to such an extent that justifies the judgment that its structural strength and
functions have been lost. Under this theory, service life means a physical
service life considering the economic aspects of
a bridge. Remaining service
life is estimated
by this method.
The "time of occurrence of fatigue crack" due to local stress is when fatigue
damage has impaired the function of a particular member.

3. CALCULATING REMAINING LIFE

The Fig.l illustrates an idea about formulation.

sevice expected life service
start present in design stop
&

term
passing remaining life >
(years) endurance years
f >
fatigue DPT D.T
damage k ¦ »1 <
*

DaT : (annual cumulative fatigue damage anticipated thereafter) XTr

Dpt : up to a given time and cumulative fatigue damage
Fig.l Life vs. passing years of a steel railway bridge
This analysis needs following data.
(1) History of live load
(2) Relation of live load with active stress
(3) Fatigue strength of joints in the precise target position.
These will generally require large data and practically made the work
 M. SHIMAMURA - M. SUGIDATE - T. HORIGUSHI - M. ABE - A. KOSHIBA 371

troublesome. Therefore, the following items were taken into consideration for
generalization.
(1) Regarding live load, axle load distributions were investigated for all
railroad line territories and statistically processed to obtain a central value.
(2) For the relation of live load with a stress, an actual stress ratio (measured
stress divided by design stress) was examined and rearranged. Fig.2 shows stress
ratios (a) used for railway bridges at present.
(3) A fatigue test was condueted for long-life areas by service stress to
determine evaluation strength.
This is a specimen. Live load was determined from the result of long-term (from
about three months to one year here) measurement of real bridges.
The remaining life evaluation formula is as Fig.3.

0.9
0 main girder
0.8 floor
+ system
0.75 J

0.7 8
CO

H- 0.65 ¦
« 0 - 8
o
1
(0 0.6
CO
CO
• o 1

>i o
<^
p

• •
• o o
O
CO
0..4
I 1
o

•
¦
°.
0

10 20 30 40 50 60 70 80 90 100
influence line (m)
Fig.2 Research result of stress ratio

A 0 amax (i)
I
CC
Tr N0( 1-Dpt )/ i 1
{n ,c<,(i)
Atffo
3 }

A 0 amax (i) maximum stress ränge exerted by trains operated in future.

n aeq (i) one year-equivalent number of cycles of maximum stress ranges
exerted by passing trains.
A Ö (i) respectively stress ränge and its cycle, as obtaind by analizing
n (i) the frequencies of variable stresses exercised by trains.
nn level of stress ränge for frequency analysis of one passing train
Ny number of trains passed in a year; if this is not a variable,
N=365Xn ad
n ad number of trains passed in a day.

Fig.3 Remaining life evaluation formula

372 REMAINING STRUCTURAL CAPACITIES OF STEEL RAILWAY BRIDGES

4. BMC (Bridge Maintenance Consulting) SYSTEM

The results as mentioned above were programmed and rearranged into a fatigue
diagnosis system capable of automatic processing. We call it FATIDAC system.
This FATIDAC system is a part of the bridge maintenance and management system
called BMC system. BMC system consists of :
(1) Bridge database
(2) Bridge diagnosis system
(3) Supporting system (expert system)
FATIDAC system forms a part of the bridge diagnosis system. The system has
functions as a measuring instrument and executes measurement, evaluation analysis
and outputting of a diagnosis report on the spot in a series of process. Data is
registered on the database of the system and outputted in the form of a
tabulation when necessary. This system is used for bridge diagnosis of decaying
bridges abroad. BMC system performs also the following.
(1) Live load calculation
(2) Bearing capacity calculation
(3) Calculating Vibration in relation to running Performance
(4) Bridge stability calculation
Select the bridge Joint (portion)
EXAMPLE to be checked in a given bridge

Items for processing are as

Arbitrarily sample a Joint to be
shown in the fig.4. The first measured; measure the nominal or
example relates to prediction of local stress; or draw a stress
occurrence of a local fatigue wave from by Simulation
damage.
(1) Stress in the precise target
position in measured. Concerning Caluculate the maximum stress
fatigue to be measured, a ränge by using ß or through
Statistical processing and grasp
position to which a gage is the tendency of all the bridge
fitted is shown in the girders (monitering)
standardized manual.
(2) Sampling is made several Where or:working stress ratio
#:coefficient to corrowded
times, which is necessary for YES
the measured value
stress frequency analysis.
(generally for Statistical NG
purpose)
(3) Live load history is Analysis the degree of fatigue History of dynamic
inputted on the basis of the damage in terms of cumulative loading size of
result. fatigue or fracture mechanics initial defect
(4) The cumulative fatigue
damage ratio calculated.
(5) D value as a result of this dge the degree
calculation is
judged in cumuiau ve fat gue
accordance with the category of damage (D)
judgment. A requires some
B will be confirmed
measure and Other than ^p
on a priority basis by
inspection in the future.(Fig.5) Take the measure as prescribed
(6) Finally an examination in the diagnosis
report is outputted. In this
i
Finish

Fig.4 Flow chart of fatigue damage cheking

J% M. SHIMAMURA - M. SUGIDATE - T. HORIGUSHI - M. ABE - A. KOSHIBA 373

case, in accordance with the results of judgement early, a specific remedy is

suggested. A suggests that a measure should be taken immediately in accordance
with S-13 in the manual. B indicates that
item in the
should be inputted as a priority
next general inspection. (Fig.6) This system is being
it
inspection
employed widely for diagnosis of steel railway bridges by all Japanese Railway
Companies, including Tokaido Shinkansen and private railway companies. Timely
diagnosis ensures quantitative and objeetive bridge diagnosis.

JRM£#»1»KK«fc 5«£ffi¥«50E# (1) - tt BMC Fatidac SySTEM

« JS £ B**** «i»* 1 3
(a®*^ WDG610-101
#»«#»1** (D) ?1 3t E it 1* S ^ © 5 "ft

« « € £ tfi K & ±T7?^y » » - 7

* BMC- 1

D * 1 0 AI flH gi| tö i£ © * HÜ
If (35
* -»
J R»SJ«JI <«) 5i a a 19 9 2 6 4
1 0 > D ü 0 8 A2
(2) **gtE0rfZ*t-rS!¥<I}*jlt
0 8 > D 2 0 5

*
» « rt s If 1 <o X 9 « JR «
Ü ^ B S « J« 1 ^
Mft«**F*r
2007JE] ft tt » ff flt
0 5 >D* 02
Na
* S tt¥ BÄH tt flt m n # * # »{* k ICttt*
« * C t**©*itBp/r^
(MPa) (MPa)

toiiQJmfft 1
£*rn)^t7f 6 5 3 2 1 2 7 2 A.
D < 0 2 S d # ä K> © * £ 2 £thT75^> 1 0 0 3 3 9 0 7 5 B

3 3Elh±7j^»t^ 6 5 3 2 1 0 6 1 B
tt) 1 DcTMäli, ffi?ijic»g LA.HWrJ»\ fc L< ««f±*Jcaffl-^4«*Ji,
4

5
<*B3flM«rtl 0 0«

Fig.5 Category of cummulative

fatigue damage ratio (3) 4i£IMUfMt

Na ia *
1
A,
2 B
LTTSl».
3 B AB© r£»t«j Tli Ä*l«J&B±l,T:iIftlU *rwi8
LTTSt».

Fig.6 Example of examination report

REFERENCES

1. ABE,M. et al., Development )f a System for Estimating the Remaining Life of

a Steel Railway Bridge. The Second Japan-Korea Joint Seminar on Steel Bridges,
Tokyo, Japan, July 1992.

2. HORIGUCHI,T. et al.,Development of Fatigue-Damage Rating System for Steel

Bridges.
Proc. of the 4th Inst. Conf. on Fatigue and Fatigue Thresholds, July 1990 in
Honolulu.
374
 Remaining Service Life of a
Duröe de vie restante d
Restnutzungsdauer einer gen

Eugene BRUHWILER
Dr. Eng.
Swiss Federal Railways
Zürich, Switzerland
r^T^ *Wi
.-.

Peter KUNZ
Research Associate
Swiss Fed. Inst, of Technology
Lausanne, Switzerland
376 REMAINING SERVICE LIFE OF A RIVETED RAILWAY BRIDGE

1 INTRODUCTION

Riveted bridges were built until the 1950s. There are thousands of riveted bridges around
the world which are still in service. Most old bridges have been restricted to reduced
permissible loads or repaired; eritieal elements have been reinforced or even replaced.
Some bridges are subjected to more severe loading than was envisaged during their
design due to increased traffic and higher axle loads. Economically, it is not possible nor
justified to replace all bridges when they reach their "design lives". Often there is a
considerable reserve and a remaining service life may be justified provided that inspection
guidelines are followed.

This paper deals with the investigation of the reliability and remaining service life of a 118
year-old bridge on a branch railway line. Standard engineering methods revealed various
bridge members to be fatigue eritieal. Since fracture of a chord member could result in
collapse of the bridge, a severe load restriction was imposed on the bridge. This load
restriction hampered the freight service on the line resulting in substantial financial loss.
As a consequence, a thorough investigation of the structural reliability was condueted
applying, the method for evaluating the fatigue safety as developed in [1] and summarized
in a companion paper [2].

2 DESCRIPTION OF THE BRIDGE

The investigated bridge crosses the Rhine river in Northern Switzerland near Lake
Constance. It was built in 1875 and comprises riveted wrought iron members. The straight
through truss bridge is a continuous girder bridge over four spans with a total length of
254 m (Fig. 1). The two main girders are of a truss-type construction with parallel chords
and cross-wise diagonals. The Single track is not ballasted; the timber sleepers are fixed
directly to the stringers.

The bridge structure has been strengthened several times to allow for higher traffic loads.
At the end of the last Century, all major elements of the bridge were reinforced. In 1936,
the stringers were strengthened for a second time, and in 1964 several truss nodes of the
main girders were reinforced using high-strength bolts. Apart from some local corrosion,
the present State of the bridge structure is satisfactory.

From 1875 until 1968, a mixed freight and passenger traffic of about 25 trains per day was
crossing the bridge. Since 1969, this branch line has been exclusively in service for a daily
traffic of 15 freight trains. To date, a total of 850*000 trains have crossed the bridge which
is a moderate traffic compared to main lines with a daily traffic of about 120 trains.

3 LOAD CARRYING CAPACITY

A structural analysis is performed first to determine the load carrying capacity of the
bridge. The action effect S is calculated using actions and load factors according to the
Code SIA 160 [3] with the load model corresponding to UIC 702 [4]. The resistance R for
wrought iron is chosen based on [5, 6, 7] which include test results for wrought iron
 E. BRÜHWILER - P. KUNZ 377

specimens from bridges built in Switzerland in the last Century. The load carrying capacity
is expressed by

load carrying capacity ratio

R resistance
YR resistance factor (=1.20)
S action effect

For jii > 1, the investigated element fulfills the structural safety required by SIA Code. For
H < 1, the load carrying capacity in terms of admissible rail traffic load is established. The
structural analysis reveals that the main girders are the governing elements regarding
load carrying capacity of the bridge. The minimum value for ji, equal to 0.80, is obtained
for the top chord members over the piers.

4 FATIGUE SAFETY

4.1 Proceeding in stages

The fatigue safety is verified using a method described in [1], proceeding in stages with
increasing level of sophistication:
1. Simplified deterministic method (chap. 4.2)
2. Simplified probabilistic method (chap. 4.3)
3. Inspection intervals (chap. 4.4)
4. Detailed probabilistic method (chap. 4.5)

The aim of the first stage is to identify the fatigue eritieal members of the structure. The
probability of failure pf due to fatigue is calculated in stages 2 and 4. The probability of
crack detection p(cd) during inspection is evaluated in stage 3, and subsequently linked to
Pl to obtain the probability of rupture prupt:

Prupt=Pf[l-P(cd)] (2)

Prupt probability of rupture

Pf probability of failure
P(cd) probability of crack detection in a construction detail

The probability of rupture can also be expressed by means of the reliability index ßrupt
according to the Standard normal distribution. Finally the reliability of a structural element
is compared to the target value ßt:

ßrupt >ßt (3)

ßrupt reliability index with respect to rupture

ßt target reliability index
378 REMAINING SERVICE LIFE OF A RIVETED RAILWAY BRIDGE

The target value ßt for fatigue safety is adopted based on proved values assumed in the
design of structures where a reliability index of 4.7 is often used; this corresponds to a
probability of rupture of about 10-6 per year. Assuming a design life of 100 years, pt drops
to 10"4 or ßt 3.7 over the entire service life of the structure. The probability of rupture
due to fatigue should not be larger than structural safety. Consequently, a target value ßt
3.7 is considered in this study for the total collapse of the structure resulting from fatigue
failure of the detail being assessed.

4.2 Simplified deterministic method

The fatigue safety of all bridge members is checked by the following ratio:

Ao^Mat (4)

with Aae aS| a «N

• • Ao(0 Qfat)
• (5)

Vfat fatigue safety ratio

Aac fatigue resistance to 2-106 cycles 67 N/mm2) [5, 7, 8]
Aae fatigue load effect refered at 2-106 cycles
Yfat fatigue resistance factor
asiA correction factor from the SIA Code 161
ocfsj correction factor taking into aecount the number of trains in the past [1 ]
Aa(a>-Qfat): stress difference due to the fatigue load defined by UIC (International Union
of Railways) multiplied by the dynamic factor

The resistance factor Yfat is adopted to obtain a reliability index ßt of 3.7 [1]. For redundant
details causing local failure, Yfat 1.20 is chosen, and for elements leading to total
collapse of the structure, Yfat 1-34 is considered.

Based on this deterministic method, the bridge members are compared, and a ranked list
identifying fatigue eritieal bridge details is established. Details with Vfat < 1 require further
investigation and inspection. Fatigue safety is verified if Vfat > 1.

The lowest Vfat-values are calculated for truss elements of the main girders. This is an
uncommon result because frequently the stringers are the most fatigue eritieal elements of
old bridges with non-ballasted track. This result is explained by the various strengthening
measures; stringers and cross girders were reinforced more than the main girders.

Due to the continuity of the bridge over four spans, the influence line for the chord
members (Fig. 4) shows both high compressive and tensile stresses at the zero-moment
locations. The fatigue action is thus greater than the action effect that was considered for
structural safety during design and strengthening. Consequently, the simplified
deterministic method identifies the chord members at zero-moment locations between the
nodes 7 and 8, 12 and 13 as well as 19 and 20 as the most fatigue eritieal members of the
main girders (Fig. 1).
J% E. BRÜHWILER - P. KUNZ 379

The minimum value for vfat 0.69) considering 850'000 trains is calculated for chord
member 18/19. This is considerably smaller than 1, and a more thorough investigation
based on probabilistic concepts is therefore performed.

4.3 Simplified probabilistic method

The reliability index for fatigue failure of a bridge member is evaluated as follows:

R
_mR-ms(Nfut)
ßt-—rj—y~
VSR + SS
(6)

mR mean of the fatigue strength logAac + 2-sr)

ms(Nfut): mean of the fatigue load effect as a function of the number of future trains
Nfut
Sr Standard deviation of the fatigue strength 0.11)
Ss Standard deviation of the fatigue load effect 0.04)

The background and the assumptions as well as the results and the correction factors
needed to calculate ms are presented and discussed in [1, 2].

The reliability index ßf as a function of the number of future trains Nfut is calculated. The
probability of crack detection due to inspections is not considered, thus p(cd) 0; and
from equation (2) follows: ßrupt ßf. The result is shown in Figure 2 with Nfut represented
on a log-scale.

The reliability index is now compared to the target value ßt 3.7. The calculated ß-value
of chord members 18/19 is smaller than ßt; a more detailed analysis is necessary. There
are two ways to influence the calculated reliability index:
(1) to optimise the inspection intervals, considering the probability of crack detection,
and
(2) to refine the assumptions and calculate Pf using a detailed probabilistic method.

4.4 Inspection intervals

By a fracture mechanics analysis, the eritieal crack length of the most likely crack in chord
member 18/19 is calculated using the R6-method [9]. Assuming fracture toughness values
K|C of 1500 N/mm3'2 and 2000 N/mm3'2, eritieal crack sizes ac of 70 and 100 mm
respectively are calculated. Subsequently, the probability of detecting a crack shorter than
the eritieal crack is assessed. For this, the relation between the probability of crack
detection and visible crack length must be known for steel bridges. Due to the lack of
relevant information, an estimate based on literature regarding offshore structures is
adopted [1].

Figure 3 shows the probability of crack detection p(cd) as a function of the number of
trains between two inspections Np jnsp. Inspection intervals can be fixed; for example for
380 REMAINING SERVICE LIFE OF A RIVETED RAILWAY BRIDGE

p(cd) 0.95, the chord member must be inspected after every 17'000 trains. Putting p(cd)
equal to 0.95 in equation (2) a value of 3.3 is obtained for ßmpt for the present state, which
is still smaller than the target value (Fig. 2).

4.5 Detailed probabilistic method

For the detailed probabilistic analysis, refined rail traffic modeis for the past and future are
needed. The rail traffic model in the past is based on Statistical data provided by the Swiss
Federal Railways regarding the number of trains and towed load. The rail traffic model for
the future considers the expected freight traffic on this line.

The calculated reliability index as a function of the number of future trains is shown in
Figure 2. ß is equal to 3.4 for Nfut 150'000. Fixing the inspection intervals so as to give a
95% probability of crack detection, a reliability index ßrupt of 4.15 is obtained using
equation (2), which is greater than the target value ßt of 3.7. Fatigue safety could also be
verified for even more trains; yet the bridge owner is primarily interested in the reliability of
the bridge for the next 25 years.

Figure 2 shows a striking increase in ß from the simplified to the detailed probabilistic
method. This is explained by the significantly smaller towed load and length of the trains
on this branch line when compared to the traffic load modeis used in the simplified
method. For main lines with usually higher loads and longer trains, there is a better
agreement between service and traffic model loads, and consequently, the difference in ß
between the simplified and the detailed probabilistic method is smaller.

5 FIELD TESTING

The main objeetive of the field testing is to verify the structural modeis. Top and bottom
chord members between the nodes 18 and 19 (Fig. 1) have been equipped with strain
gauges, and the strain history due to an engine of six axle loads equal to 177 kN has been
recorded. There is a small difference between the measured and calculated influence
lines for the bottom chord member 18/19 (Fig. 4); such good agreement is frequently
observed for truss girders.

6 BRIDGE INSPECTION

The field investigation included a thorough Visual bridge inspection. Cracks have been
detected at details with Sharp corners located at the reinforced parts of the stringers.
These cracks may not affect the structural integrity because crack propagation is likely to
stop. Such "undangerous" cracks are usually not revealed by analysis, which emphasizes
the importance of bridge inspection and study of construction details.

"Dangerous" fatigue cracks affecting the structural integrity (such as those investigated in
the foregoing sections) could not be detected. This is in agreement with the theoretical
 E. BRÜHWILER - P. KUNZ 381

investigation predicting that - at present - any fatigue crack is likely to be shorter than the
crack that could be detected by Visual inspection.

7 CONCLUSIONS

1. Comprehensive assessment of the remaining service life of railway bridges includes

theoretical studies of the structural reliability, bridge inspection and field testing.
2. A method for the evaluation of fatigue safety which proceeds by stages of
deterministic and probabilistic calculations of increasing sophistication has been
demonstrated. Probabilistic methods enable the consideration of the scatter of the
parameters that influence both the fatigue strength and damage aecumulation.
3. Based on fracture mechanics and probabilistic concepts, inspection intervals are
assessed and considered in the evaluation of structural reliability.
4. By means of strain measurements on fatigue eritieal members, the load carrying
behaviour of the structure is evaluated and a more accurate (and usually lower) value
for the stress ränge is determined, which is the most dominant parameter influencing
the fatigue safety.

As a consequence of this study, the severe load restriction imposed on the bridge was
removed for the next 170*000 trains. In addition to the main inspection every 5 years, a
supplementary inspection of the fatigue eritieal chord members of the main girders is now
condueted at yearly intervals. This thorough evaluation is cost effective because it now
allows for practically unrestricted rail traffic for the next 25 years.

ACKNOWLEDGEMENTS
The Measurements and Instrumentation group of the Swiss Federal Railways is thanked for the carefully
prepared and condueted bridge measurements.

REFERENCES
[1] KUNZ P., Probabilistic method for the evaluation of fatigue safety of existing steel bridges. Doctoral
Thesis Nr. 1023, ICOM - Steel Structures, Swiss Federal Institute of Technology Lausanne,
Switzerland, 1992. (in German)
[2] KUNZ P. and HIRT M.A., Remaining fatigue life of steel railway bridges. Proceedings, IABSE -
Colloquium "Remaining Structural Capacity", Copenhagen, Denmark, March 17-19,1993.
[3] SIA Code 160: Actions on structures. Swiss Society of Engineers and Architects. 1989.
[4] UIC 702 V. Load model for the analysis of structures of international lines. International Union of
Railways, 1974.
[5] UIC 779-1 E. Recommendation for the evaluation of the load carrying capacity of existing steel bridges.
International Union of Railways, 1986.
[6] KULAK G., FISHER J.W., STRUIK J.H., A Guide to Design Criteria for Bolted and Riveted Joints.
Second Edition. New York: John Wiley & Sons, 1987.
[7] BRÜHWILER E., SMITH I.F., HIRT M.A., Fatigue and fracture of riveted bridge elements. Journal of
Structural Divisbn, ASCE, Vol. 116, No. 1, January 1990, pp. 198-214.
[8] ECCS Publication Nr. 43. Recommendations for the Fatigue Design of Steel Structures. European
Convention for Constructural Steelwork, Brüssels, 1985.
[9] HENSEN W., Concepts for the evaluation of the remaining service life of old steel bridges. Doctoral
Thesis, Technical University RWTH Aachen, 1992. (in German)
382 REMAINING SERVICE LIFE OF A RIVETED RAILWAY BRIDGE Jk
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21

IE SE
'vyJ-

57.15 m 70.00 m

Figure 1 View of the investigated truss bridge (one half)

5.5
Detail edr riet hc d:
5.0
4.5+-
/ with
• withoi jt
ii IS 3€!Ction

4.0
^-._!_.
3.5
3.0
2.5 __
X
iti
^
s iir plifibd met TUJ
| ^ "^
ins pe r»ti
CXI Oti
VA
Vi iFini itl
I
2.0
104 105 1
150'000
number of future trains N tut

Figure 2 Reliability index for chord member 18/19 as a function of the number of future
trains for simplified and detailed probabilistic approach
 E. BRÜHWILER - P. KUNZ 383

p(cd) [%]
100 T
IC

98 "
*N 1500 N/mm372
N 2000 N/mm372
96 -
95
94 ¦
\\
92 ¦¦
\\
\
90
10'000
V
20'000
\ 30'000
N,
p.insp

Figure 3 Probability of crack detection in chord member 18/19 as a function of inspection

intervals

; bridge
18/19
strain e [u,m/m]
150

100

50 calculated
moasured

50

-100
50 100 150 200 250 300
distance from the bridge abutment in [m]

Figure 4 Comparison between calculated and measured influence lines for bottom chord
member 18/19 due to the passage of the engine. (The stress values are
evaluated from the strain readingsusing an elastic modulus of 200'000 MPa.)
384
 Test Results from Suspension C
Resultats d'essais sur des cäbles de S
Erkenntnisse aus der Prüfung baupra

Wolfgang HARRE
Dr. Eng.
Otto-Graf-Institut
Stuttgart, Germany

¦ ¦' i0*
386 TEST RESULTS FROM SUSPENSION CABLES PRELOADED IN PRACTICE

1. INTRODUCTION

Despite the spread use of cables as tension members in bridge structures since
many decades, the knowledge concerning the fatigue behavior of these cables is
still very poor. Neither a sufficiently supported Ao-N-curve (Wöhler-diagram) is
known, nor the substantial proof of a fatigue strength limit for such cables
really exists. Completely unknown is the effect of variable amplitude loading,
i.e. the increase of fatigue strength respectively fatigue life when the fatigue
loading changes from constant amplitudes to variable amplitudes, the more
realistic loading. Testing with variable amplitudes related to the practice will
shift the WÖHLER-diagramm (Ao-N-curve for constant amplitude loading) towards a
higher fatigue strength equivalent a longer fatigue life. The amount of increase
in fatigue life depends on the amplitude spectrum.
In consequence of the lack of comprehensive basic knowledges concerning their
fatigue behavior, the cables for every cable-stayed bridge going under
construction have to pass a Standard test program according to formerly DIN 1073
respectively now DIN 18809, which has no relationship at all to practice and
which represents a very unsatisfactory proceeding since many decades! There are
mainly two reasons for the lack of basic knowledges: a) in the earlier past no
suitable testing equipment was available to run tests under random load
conditions as well as with comparatively high forces (~1 to 10 MN), and b) to create
a complete WÖHLER-diagram or Ao-N-curve requires a great number of specimens,
i.e. the enormous financial expense even only for the specimens has prevented
until now the realization of any proposed test program.
Nowadays the Situation has somewhat changed: Suitable testing equipments are de-
velopped and available. However the fabrication costs of the specimens remain
stillfabricated
extraordinarily high, so that extensive and systematical investigations on
new test specimens of cableswill quasi out of reach now and in the
be
near future. On the other hand, there are meanwhile several cable-stayed bridges
in service, which will be or should be overhauled partly or completely soon. In
the course of such bridge "renovations" the cables will often be replaced. In-
stead of scräpping them, the dismantled cables could be used on for experimental
investigations as practically preloaded test specimens almost free of charge!
First of such investigations would be to set up a sufficiently covered
aim
WÖHLER-diagram of preloaded cables in the sense of a lowerbound-curve for the
Classification of the fatigue behavior of cables in general, as well as to obtain
founded informations regarding the damage progress and the remaining fatigue life
of preloaded cables.
The recently accomplished investigation program, carried out at the Otto-Graf-
Institut, on dismantled cables of the NORDERELBE-bridge near Hamburg and of the
FEHMARNSUND-bridge to Denmark can be considered as a first step in the above
described direction.
The gained results-partly contradictory-reveal the considerable deficit in
knowledge at present and the actual inability to evaluate the fatigue strength,
the influence of damaging factors and the remaining fatigue life of cables. In
this report, the most important results of the condueted investigations up to now
are presented in a compact form and conclusions are pointed out.

2. FEHMARNSUNDBRIDGE

2.1 Statement of the problem

In 1980 the Otto-Graf-Institut was commissioned to investigate a cable dismantled
of the Fehmarnsundbridge. During an inspection and examination of the bridge it
 W. HAR

was found out that some of the 80 Suspensi

The protection against corrosion came part
the outer layer were opened and there were
a decision base for the neccessary sanitat
subjected to detailed material testing and
report, only the most important mechanica
presented. More detailed informations can
2.2 Testing program
The following testing program was arranged
2.2.1 Cable specimen A

Constant amplitude fatigue test with Ao

afterwards a constant amplitude fatigue te
0,20-106 cycles; afterwards a statical rup

2.2.2 Cables specimen B

Constant amplitude fatigue test with Ao

a constant amplitude test with Ao

The tests were condueted in a hydraulic 5-M

:-:^':-:::rr. ::¦-¦¦;.:¦-¦

I«—PS»
388 TEST RESULTS FROM SUSPENSION CABLES PRELOADED IN PRACTICE

69,11m

25B6m <s>

9' 2' 1 1 2

£=248,40m

D9' (specimen) D9

Fig. 2 Fehmarnsundbridge-survey

The cable, about 14,20 m long, <j> 81 mm, was substantially divided into 3 different
segments.
The upper part of the dismantled cable with a length of about 3,60 m was used for
material- and corrosion tests. After cutting off the upper anchorage and extru-
sion of the anchor socket, the latter one was applied as anchorage to the upper
end of cable specimen B.

Cable specimen A was made of the middle part (length 5,40 m) of the dismantled
cable by attaching special test anchor sockets on both ends.
From the lower part of the dismantled cable (length 5,20 m) with the remaining
original anchorage at the lower end, cable specimen B was made by attaching as
mentioned above the anchor socket of the initially upper cable part to the upper
end of the test specimen.

The casting of the sockets was done by Thyssen Draht AG, Gelsenkirchen. Casting
metal was ZnA16Cul (Zamak 610). Test specimens A and B had a free cable length of
4,50 m each.

3. NORDERELBEBRIDGE

3.1 Statement of the problem

When the highwaybridge over the river Norderelbe near Hamburg (built in 1962) was
subjected to a comprehensive sanitation in 1985, the Otto-Graf-Institut was com-
missioned to investigate the dismantled cables 72 mm. On the base of the test
(f>

results obtained from Fehmarnsundbridge [1], the intention was to achieve further
contributions and clarifying informations in the field of longtime fatigue
behavior and remaining fatigue life of cables and cable-stayed bridges. Detailed
descriptions are given in [2].
3.2 Testing program
The following testing program was planned and condueted.
 W. HARRE 389

3.2.1 Cable specimen C

The specimen was taken from the free length of the dismantled cable. Constant
amplitude fatigue test with Ao 150 N/mm2 and with simultaneously applied
transverse pressure of 19 kN/cm. The same test was made in 1962 with a specimen
of the new cables when the bridge was under construction. It was expected from
this test repetition to find out, whether the fatigue strength of the cables is
reduced by the service over a period of more than 25 years and yes, how much if
it is reduced.
3.2.2 Cable specimen D

The specimen was taken from the end area of the cable with an original anchorage
kept on; constant amplitude fatigue test with Ao 150 N/mm2, no transverse
pressure.
3.2.3 Cable specimen E

Constant amplitude test with an original cable-saddle, containing two cables

turned around, one upon the other, Ao 150 N/mm2. The aim of this test was to
study the behavior of the cables under realistic transverse pressure conditions
existing in such saddle-or turn around structures with several layers of cables.
All until the provided limit number of cycles NG
the tests were designed to run
2,0«106 is reached respectively until the recorded wire cracks increase progres-
sively. At the end of the tests, a
controlled opening of the cables and anchorages was planned in order to register
the wire cracks and analyse the damage.
In the following report, only the most essential mechanical-technological results
are presented. Further details and informations can be taken from [2].
The tests with the specimens C and D were carried out in the 5-MN testing machine
in the same way like the tests with the specimens A and B of Fehmarnsundbridge
[1]. As to speciment E a special testing equipment had to be developped (see
figures 3 and 4), a detailed description of it is given in [2].
cable-saddle

m un upper oable\
in
=i
anchor anchor
sockets sockets
5.5°

Ff
25MN
HUf lower
cable \
M m m
H
T> u
£ ü, tWz, fl
5,00 m
Fig. 3 Schematic diagram of test set-up
390 TEST RESULTS FROM SUSPENSION C

k I

Fig. 4 Test se

3.3 Fabrication of the cable specimens C,

Fig. 6 shows the different points of the N

the specimens were drawn from. The cable b
 W. HARRE 391

the other of about 6 m length respectively. From the initial ten cables (two
layers by five) going over the saddle in the bridge, only the middle one of each
layer was used for the fatigue test. At the four ends of the two cables ancho-
rages taken from the bridge were attached.

4. RESULTS AND CONCLUSIONS

The investigations performed on cables dismantled after more than 25 years in

service from the Fehmarnsundbridge and the Norderelbebridge have yielded the
following results and findings:
a) After more than 25 yeais in service, the cables itself showed a remaining
fatigue life of more than 2,0 • 106 cycles. This can be stated for the cables
"in relatively good condition" of the Norderelbebridge as well as for the cables
"considerably corrosion" of the Fehmarnsundbridge. Consequently
damaged by
the tests have proved
relevant DIN specifications.
a still Even
sufficient
fatigue strength according to the
after
fatigue loading of more than 25 years,
no significant loss of carrying capacity of the cables could be found, i.e. an
increased risk of damage or failure did not yet exist for both bridges as far
cables are concerned.
as the
b) Analysing the wire cracks observed during the investigations showed numerous
proofs and confirmations for a dominating influence of fretting corrosion
among all the known corrosive damaging phenomenas like tension crack corrosion,
pitting corrosion etc.
Beside of the pure mechanical induced fatigue damaging process, the mecha-
nical-chemical combined damaging by fretting corrosion exerts the most
decisive influence on the reduction of lifetime of cable structures. The
remarkable difference between the fatigue strength limit of the Single wires
(Ao 400 N/mm2) and the fatigue strength limit of the complete cables
(Ao 150 - 200 N/mm2) is mainly referred to the influence of fretting
corrosion.
c) Points of turn around the multi layered cable bundles (cable-saddles, turn
around anchorages etc) which are usually characterized by high local transverse
pressures, had always been considered therefore as very problematical
details of cable structures related to the fatigue strength. On the base of
the presented test results [1, 2],
are less eritieal as formerly assumed.
it
can be pointed out that these details
They can be classified as at least not
more "delicate" than the anchorages of the cables.
d) The acting stress rate Ao (stress amplitude o0 - ou) has an essential
influence on fatigue life of the cables. Ao is the dominating parameter
governing the fatigue life, whereas the level of the maximum stress o0 is of
subordinated influence as long as it is kept under the yield stress level.
On of the urgent aims of the investigations on preloaded cables is to improve the
understanding and fundamental knowledge of the fatigue behavior of cables by
means of the elaboration of a sufficiently supported Ao-N-curve (WÖHLER-diagram).
The set up of such a diagram however implies the derivation and the definition of
a damage - and failing criterion corresponding to the actual cable behavior.

The failing of cables submitted to fatigue loading is fundamentally different

from that of the usual steel structures: Depending upon structural boundary
conditions and the grade of noteh effect, damaging and destruetion in usual steel
structures proeeeds more or less steadily after a first crack occurred and almost
the whole fatigue life of the structure is related to the period before the first
crack happened, i.e. the crack initiation phase representes usually a life
period
multiple of the period of crack propagation.
2
LJ
r < CO_lLUi/) a_X -LI "1< _Jir 2i
—1w^ z i. r 3.*<J _Ju - w i
O
o
CO
z
CO
Q_
2
C/J
1—
%
LU
Z)
TT
—\
to
H
cc
C/5
H
QC
J)
_/
Ll_
LU
LU
(N
Ol
i
ro
__
1
3 r—14-1 o im chcd 4-1¦HbO3 P0) cn P 3 •H4-1X u fl XrH CU XIT) > CU CO CU
TJ
XI
CO
-H
CO
X
Ol CO
•H
4-1
u
4-1
M
C
cd
c
>
4-1
4-1
4-1
o
et)
CO
4-1
OJ
O
to
O
u
g
cd
c
CO
O
O
CO
CJ oE O.cdkl ro P-H> tu >%rH o o c 3 ¦H XO 3
a.
4-> jesrt
o3 u
i-H 4-1 > HC bOX C >^ <TJ
J4
CO
X
J3
j3
M-1
¦H
4-1
¦H
e
«-i
O
O
PO
td
P
cd
3
P
p
CJ
U
P
Cll
p
0)
Oj
3
c
CO
o
CO
HflC
r-l POJ P 4-1oPcd cmfd •H •HG 60 O L4-I 4JXOJ O efl xi-tOJ H 0J
X
X
X
CO
rH
-H
PCO
4-1
4->
•H
•H
3
E
P
CJ
o
4-1
tu
3
o
c
cd
OJ
P
G
O
p
O
O
C
(U
C
O
CJ
HHCfl •Hi-l-H BO
c o l-l -rlPtu P -HO C •
SC
PO XIOl 3 tu td co O P dl
rH
X
J4
1
X
M-i
•rl
tu CO
-H
Cn
td
P
-H
3
3
-H
P
l-l
o
OJ
CO
CU
rl
cd
tll
CO
3
O
4h
o
tu
M
CO
P
c bOrHOJ S -H tu o|x ro TJ bsto » -H• tu • PX0) m-Hl-l P P Ü X
CO CO CO
-H
O
4-1
oo
m
rH
X
x
-H
P
p
3
P
co
O.
O
C
O
Cd
CO
CO
O
OJ
o
>
LO
O
OJ
Ij
Ol
XOl -o¦Hg Ol C -Ho C O HJ cd g ¦Ho P o CJ P cd CJ >: 4HO rl HG pcd c CJ Ol CO
rH
X
X
-H
1—1
CO
p
>.
•H
3
•H
MJ
c
o
td
cd
CO
C
LO
rd
CO
CJ
rfl
G
>
c
o
Ol
Ol
Ij
o
O CM pXOJ CJ o c P•HC 3 ro rH
MH P WS 0) -HO rd T3H C bO rd
H 3
X
OJ CO
TJ
IH CO
P
rH
rH
-ö
rH
3
P
o
P
p
o
c
3
¦rl
Ol
0)
rl
3
3
Ij
CJ
co
Ol
OJ
cd
CO
Ol
Ol
c
O PxOl P HG
•
3 cj CJ cu tu TJ•HC bo O e tu td CT)
&p OJ cd J*i E cd>> O CJ O 3 p co3cd>>tuc003
73
-ü
CM
X
X
CO
¦H
rH
-O
cd
•H
G
3
cfl
o
CO
O
p
tu
p
a
o
P
CU
E
3
Cd
OJ
P
OJ
G
CJ
O
CO
P (U cfl J4 — 3 -rlPxO G P CJ rd 3 10 -hc bO -hE 1 tu -rl ptu >,X
-rt G rH td
o er.
IH
TJ
-H
rH
rH
-tj
CO
p
X
-H
rH
rH
P
P
-H
O
G
(U
E
(U
cO
CJ
O
P^i-I
P
CO
tu
>
P
P
o
cd
CO
CU
CU
>
-rt tu P coOJ O P-HO G co P cu cd XrH pO o M P e p OJ
CU CO
Mi
X
rH
X
H
X
CU
p
ch
-H
rH
M-1
P
4-1
CJ
Ol
D.
CU
co
tu
tu
P
3
CJ
tu
O
P
cd
CO
CU
O
Og rH cd CHPtu P td 3 p PXtu 3 t= Xtu P
•
tu TJ PP OJ co C4H
TJ
CO CU CO
-H
X
-H
CH
-H
P
-H
CO
td
P
<U
cj
G
G
P
bom
G
O.X
>
OJ
M
td
o
CO
P
o
o
CJ
P
CO
tu
O
G
bC
CJ
O CP Ps:OJ > -Hp tu cd PO XP o tu G h> (U
J»J
5-s td Q.P O bOP Ol CO CO
o
CO
ro
X
CN
rH
i—i
P
ch
rH
4J
¦H
CH
o
3
>>
O
cd
3
G
G
»
rd
>>
cu
(U
to
CJ
G Xcu o Xco tu P> OJ T3 3 X-HCJ x
O
cj
3s 3 rHtd P-hO G CJ 3 P> tu
«'Hfl CHtd -H
,ü
X
X
¦o
-H
-H
ch
P
cd
P
CJ
3
cj
CO
P
OJ
tu
P
P
tu
P
P
bO
tu
to
tu
tu
O
G
Cd
CJ
CO
x
0)
O rHrH Cd o. Ol O t-i X
P II cd tu • rH H X tu CH P bO
3 CO CU
r-i
X
X
-H
<-> CO
P
P
rH
rH
G
P
cd
1
o
bO
P.X
p.
CO
P
O
O
CO
G
3
tu
G
tu
G
rd
>^CH
O
O
co
P
Ol
O
rd
P (U Cd rH PO 3 bOX P 3 o.P P tu 3 -HPX cd J*-HG o 4H
-o
to
-n
X
X
rH
rH
P
cd
to
OJ
P
OJ
cd
N
cd
CJ
P
CJ
G
Ol
cd
o
co
cfl
tu
O
3
CJ
P
tu
¦H
P
¦rl
3
G
tu
CJ
O
>
CO
Cj
o
0)
p
•
G ¦OOJ
3rO a pXtu •HTJCH
p rH¦HT3H OJ cd
¦O¦hcoO3 lo tu T3 cd P E tu co X>*
¦HTJ CO
PI
T3
«I
CU
•HO
TD
•H
X
¦O
X
X
-O
H
¦H
bO
G
CO
G
G
CO
cd
G
E
G
CO
O
Ol
QJ
OJ
P
O
tn,
LO
tu
CU
XJCdos -Hrd M4J p cdE ^-*xO P cj cd XrHtu to cd CNCO ä-s cdE cd bOOJ
[3
S
"Ö
<
-H
^--
3
p
>
0)
p
P
p
3
CO
CO
o
G
1
P
O
D
LU
O
X
CU
tu
LO
(j
O.
1
1
o p PXOl T3 -HbOG O co co tu
«H
H oj Cd -Hto -Ho «H MH MH CH
ji
co CU CO
rH
cd OJ
-H
-H
^S
cn
cn
cd
C
CO
>
tu
O
CO
P
O
CU
P
cd
3
tu
>
tu
p
o
x
1
3-HpX PXOl P co3 Pto o PCd -HGOJ T3 -HG PXCU a.P tu
•
•
-
CU
T3
rH
X
•-!
41
P
-H
T3
1
P
M
3
3
>*
Ol
CO
O
P
CO
OJ
co
cd
rd
E
CO
o
O
CtI
tu
O
O
o
G
td
tu
^
1
Jjitu4-1 X o MH cd CNcn
•
T3 E cd bO0) o3 P> OJ o CH o o3 P tu
o CO CO
cm
-H
CN
4-1
T3
rH
-H
-H
l
i—'
4-1
4-1
'
4->
ta
LO
P
e
LO
O
cd
bO
CO
CU
>
G
LO
C
1
CU
CU
lo
-
-
tu T-J ch¦HbO3 p 0) vn
'
tu tu
1—1
-o
¦o
!—|
CO
rH
4J
rH
>
>
•H
CJ
CJ
o
tu
tu
rd
CU
4->
O
CU
a,
CO
>>
OJ
3
M
Oh
to
nj
Lj
LO
CO
UD
~ •
1
_
J
tä
-s._
E
Ol
n
i/i
cfl
di
"c
-*.,
-C.
>o
£
7
-°
i4
Safe Capacity of Suspe
S§curit6 de la capacitö portante d
Sichere Traglast von H

Peter SLUSZKA
Senior Vice President
Steinman
New York, NY, USA

KS=»"

Martin KENDALL
V.
A
Project Manager
Steinman
New York, NY, USA
394 SAFE CAPACITY OF SUSPENSION BRIDGE CABLES

1. INTRODUCTION

1.1 In-depth inspections have been undertaken recently on a number of older

Suspension bridges with the primary purpose of inspecting the condition of the
wires of the main Suspension cables. This was in order to determine if
any
deterioration has occurred, if
so how much, and to compute the remaining
structural capacity of the overall cable.
1.2 Steinman has undertaken such in-depth inspections on the Bear Mountain
Bridge across the Hudson River approximately 40 miles north of New York City, the
Suspension span of the Triboro Bridge across Hell's Gate in New York City, and
the Golden Gate Bridge in San Francisco. In addition, Steinman has been involved
in a Consulting capacity for the in-depth inspection of the Mid-Hudson Bridge
across the Hudson River at Poughkeepsie.
1.3 All four of these bridges were completed more than 55 years ago. The Bear
Mountain Bridge was completed in 1924, the Triboro Bridge was completed in 1936,
the Golden Gate Bridge was opened to traffic in 1937, and the Mid-Hudson Bridge
was completed in 1936. As such, they represent a good sample of the type of
medium to long span bridges being constructed in the United States between the
two World Wars.

1.4 In-depth inspection techniques for viewing and establishing the condition
of Suspension bridge main cable wires were developed by Steinman in the early
1980's initially for inspecting the main cables of the Brooklyn Bridge, in New
York City. These techniques were successfully utilized on the special
investigation of the Williamsburg Bridge, also in New York City in 1988/89 when
it was demonstrated that the condition and capacity of the cables was fully
adequate for the design loads imposed.

1.5 The four bridges which are the subject of this paper represent a more
commercial and standardized practice of the construction of Suspension bridges
than either the Brooklyn Bridge or the Williamsburg Bridge. All four bridges
have main cables that were constructed using the aerial spinning techniques for
cable wire developed by the John A. Roebling Sons Company. All four bridges have
galvanized bridge wire of approximately 4.95mm in diameter with a minimum
ultimate tensile strength varying from 1500 N/mm to 1600 N/mm The wire for
the Bear Mountain Bridge and the Golden Gate Bridge came from John A. Roebling
Sons Company, and the wire for the Triboro Bridge and the Mid-Hudson Bridge was
supplied by American Steel & Wire Company.
1.6 The size of the spans and cables, with the exception of the Golden Gate
Bridge are all roughly comparable. The make-up, size of cable, and spans are
listed in Table 1 and the bridges are illustrated in Figure 1.
2. IN-DEPTH INSPECTION - A CASE STUDY

2.1 A description of the inspection of the Bear Mountain Bridge is provided to

illustrate the means of inspecting the cables.
2.2 The Bear Mountain Bridge was constructed in 1923/4 as the first fixed
crossing of the Hudson River south of Albany, which is approximately 100 miles
north of the bridge. At the time of construction
in the world, 497.2m.
it
had the longest clear span
 M. SLUSZKA - M.H. KENDALL 395

Bear Mid-Hudson Triborough Golden

Mountain

Year Completed 1924 1936 1936 1937

Main Span (m) 497.4 457.2 420.6 1280.2
Cable Diameter 464 425 527 924
Number of Wires 7292 6080 9176 27572
per Cable
Minimum Ultimate 1500 1500 1600 1550
Tensile Strength
(N/mm2)
Residual Diameter 1680 1830 1470 1830
of Wire Curvature
TABLE 1
SUMMARY OF CABLE DATA
(all dimensions in millimeters unless noted)
2.3 The two towers of the bridge are 106.7m high. The clearance from mean
high water level to the underside of the stiffening truss at the towers is 42.1m,
rising to 46.6m at mid-span. The designed span to sag ratio is 8.16 under dead
load. The bridge has straight backstays with truss supported end spans. The main
stiffening truss is 9.14m deep, supported by two part 57mm diameter wire rope
suspenders connected to cantilever extensions of beams supporting the bottom
chord. The cables are at 18.70m centers, outside the face of the stiffening
truss, and dip down at mid span below the top chord of the stiffening truss.
2.4 The construction of the bridge incorporated certain advances in technology
which have subsequently permitted the adoption of unlimited size of cables on
other bridges. Previous to the construction of the Bear Mountain Bridge
been the practice to spin the Strands and serve them with round wire to hold
had it
together the parallel wire. When the first seven Strands had been completed,
they were formed into a circle, the serving wires to the Strands removed, the
whole formed into a homogeneous group and re-served with round wire. The next
set of Strands were then spun, and formed to a layer around the central core.
Again the serving wire to each Strand and the core was removed, and the whole re-
served. The process continued until the füll cable was completed, and then the
whole was compacted. This was an extremely lengthy and time-consuming process.

2.5 With the Bear Mountain Bridge, the use of flat steel wire serving bands was
adopted for the Strands. This permitted the spinning of each Strand and its
placement in the overall preliminary hexagon shape, without the tedious forming
procedure. When all Strands were completed, the flat bands were all cut, the
bands to all external Strands were removed and the compaction undertaken.

2.6 In-depth inspection of the cable was undertaken at four points along the
cable. These were the backstay just outside the anchorage, the tower top at one
panel point into the main span from the saddle, at mid span, and at the one-third
point of the main span. The backstay and mid span location were selected as
those locations most likely to reveal any water that may have entered the cable.
The tower top location is that point of maximum cable load, and the one-third
point is that location subject to the most flexure under live load. All
locations selected were on the north cable except for the one third point
location.
2.7 Inspection of the cable is undertaken by removing the protective wrapping
wire and driving wedges into the cable in order to provide a Visual check of
aoaina aoarae aoaina v i,

Miuvn 9
IP«4J jn6

NivxNfiow Honouoarax NosanH-aiw

TTTTrl

0S>£ j
j

nvaa
aoaiHfl
aisdaaxHonod
SOVUOHONV
sNssnb
J.SV3 aiivo
BDDBI^^
9*
8**12 822 2*0821
Naaioo

* 2*2.9* sai
L6* 9*02*

30VU0H0NV 8*V12
9*822
AS3M 0*£>£
ÜNV7SI
OMV1HÖIH
OOSIONVUd

S<dUVM

NVS
 M. SLUSZKA - M.H. KENDALL 397

interior wire conditions. Wedges were driven into the cable at four radial
positions around the cable, except at the mid span location where four additional
grooves were opened.
2.8 At the three main span locations, suspender ropes and associated cables
were removed in order to provide Observation of conditions under the cable bands.
Purpose made jacking equipment was designed and supplied in order to free the
suspender rope, and subsequently re-tension the new ropes after installation.

2.9 After füll inspection of the cable surfaces

and wedged grooves, the wedges
were withdrawn and the cable recompacted using a specially designed and supplied
compactor. After completion of compaction, the cable bands were replaced on a
new layer of red lead paste applied over the cable wires, and the new suspender
ropes installed and tensioned.
2.10 cable was then wrapped with soft galvanized wire by means of a purpose
The
built ply wrapping machine, also laid over a fresh layer of red lead paste
two
applied to the main cable. A minimum tension in the wire of 665N was specified.
2.11 Füll visual inspection of the exposed wires was undertaken using
photographs and a standardized form of recording corrosion levels. All broken
wires were located on position Charts showing both the cross section, and
longitudinal locations. Broken wires were labelled and sampled and the samples
sent to Carleton Laboratory at Columbia University for detailed Visual and
microscopic inspection, and physical testing.
2.12 Wherever possible broken wires were respliced using a technique developed
for Brooklyn Bridge, incorporating the use of repair wires joined to bridge wire
with hydraulically crimped ferrules and the repair wires themselves joined and
tensioned with threaded ferrules.
3. TESTING OF SAMPLES

3.1 Visual inspection of all wire samples was undertaken, with special effort
being given to the broken ends of the wires. Microscopic examination of all
broken ends was undertaken and two specimens were viewed under the Scanning
Electron Microscope. Five wires were tested for hydrogen embrittlement with two
samples showing elevated hydrogen levels.

3.2 Longitudinal sections of two wire samples were prepared and viewed under
the microscope in order to provide further information on the failure patterns.

3.3 Standard tensile tests were on a number of wire samples, with

made
specimens approximately 450mm long. Additional tensile tests were performed on
long sections of wire up to 5.2m in length in order to compare the possible
effect of the straightening of the wire from the coil set. Four fatigue tests
were undertaken on short wire samples and finally additional tensile tests were
performed on the broken ends of certain of the long samples which had failed at
low ultimate tensile stresses.

4. DISCUSSION OF RESULTS FROM BEAR MOUNTAIN BRIDGE

4.1 The primary cause of the observed wire breakages appears to be stress
corrosion, with all fracture surfaces observed exhibiting the characteristic
fracture profiles. Representative longitudinal sections of wires also revealed
the typical secondary cracks running into the wire.
398 SAFE CAPACITY OF SUSPENSION BRIDGE CABLES

4.2 A significant number (approximately 40%) of the wire breaks were observed
to be immediately adjacent to the Strand Straps that had been left inside the
cable. In addition, the large majority (approximately 90%) of all wire breaks
are located in the 300 mm diameter central zone of the cable, thus being
topographically situated in the zone where the flat wire serving bands were cut
but not removed.
4.3 These banding Straps have almost certainly initiated zinc depletion of the
galvanized wires as the Straps were made of low carbon steel, with no protection.
The indication of a possibility of hydrogen embrittlement of the wire breaks
tends to support the above-described zinc-depletion effect, and in addition would
render the wire itself susceptible to corrosion.
4.4 The definitive reasons for the occurrence of the stress corrosion cannot be
readily defined. However, the initial crack planes on all fracture surfaces were
located on the inside of the residual curvature of the wire. This seems to
indicate that initial occurrence, and subsequent propagation of the crack could
be related to the initial manufacture of the wire and its subsequent decoiling.

4.5 Thestrength testing of the wires showed that the wire generally had the
same capacity as when it was originally manufactured, except where stress
corrosion has commenced. All tensile tests undertaken showed an average ultimate
tensile stress of 1400N/mm except where stress corrosion crack planes were
located. Retests of four (4) wires taken from the broken ends of two (2) tensile
tests where the failure stresses were low, exhibited revised failure stresses
averaging 1375N/mm indicating that the balance of the wire did not have pre-
existing cracks comparable to those at which the initial breaks occurred.
4.6 Itthewas noted faces
that in all cases the in-place wire breaks showed corrosion
across entire of the fractures. This indicated that the majority if
not all of the breaks occurred a significant time past, and that the stress
corrosion attack has been retarded in the recent past. This is in agreement with
reported maintenance Operations where an increased attention to maintenance and
painting of the cables has been in effect for approximately the last 10 years.
5. SUMMARY OF RESULTS FROM OTHER BRIDGES

5.1 The inspection of the Golden Gate Bridge showed that the wires generally
were in excellent condition. One location was unwrapped and wedged open at six
(6) radial positions for inspection, and other than the exterior wires at the low
points of the cable where some medium surface corrosion was observed, the
interior wires of the cable only exhibit light zinc oxidation with spots of minor
ferrous corrosion. Observation of the samples removed for testing showed that
these minor corrosion locations were generally on the inside of the residual
curvature of the wire. It is to be noted that the wire wrapping system of the
cables was very well applied during original construction, and has been
maintained during the life of the bridge. The cable was dry when inspected, but
this may have been related to the lengthy drought that California has just
experienced.
5.2 Theinspection of the Triborough Bridge showed that the wires are suffering
from some significant corrosion to the surface layers of wires. Four (4)
locations were unwrapped for inspection, with one location being wedged open to a
depth of 250mm and the other three only to a depth of 50mm. Fractures have
occurred in wires at the surface layers, apparently due to hydrogen assisted
stress corrosion cracking. These wires were invariably heavily corroded with
visible section loss. Interior wires observed exhibit corrosion of the zinc
coating with localized areas of ferric corrosion similar to those observed at
 M. SLUSZKA - M.H. KENDALL 399

Bear Mountain. Observation of the samples removed for testing showed that these
corrosion locations, again, were generally located on the inside of the residual
curvature of the wire. The wire wrapping system of these cables was not well
applied with gaps between the wires evident, and red lead paint was used as a
seal layer between the wrapping and main cable wires, instead of red lead paste
used on Bear Mountain and Golden Gate. The combination of these two last factors
has permitted water to enter the cable and initiate corrosion of the wires.
5.3 The inspection of the Mid Hudson Bridge showed that the wires are suffering
from active stress corrosion attack. Forty eight (48) locations have been
unwrapped and wedged open füll depth for inspection, and a large number of
fractured wires have been discovered. In addition there is a wide distribution
of local corrosion locations showing pitting and craters, with severe local loss
of zinc and local corrosion attack of the steel wire. All fracture surfaces
observed were typical of stress corrosion, and all fractures had the initial
normal fracture surface on the inside of the residual curvature of the wire. In
addition, in a number of samples, when the zinc had been removed from the wire,
transverse cracks were observed on the surface of the steel wire at approximately
3mm intervals, and all were located on the inside of the residual wire curvature.
Cracking occurred at local areas exhibiting a hard, black corrosion product, even
where little loss of section had occurred.
6. DEFECTS EVIDENT IN BRIDGE WIRE

6.1 Three of these bridges are exhibiting common defects and deterioration of
the bridge wire making up their main cables. These defects originate with the
failure of the overall protective wrapping wire system to the outside of the main
cable wires, resulting from a combination of failure of the external paint
system, inadequately installed or maintained wrapping wire, and incorrect
material specification of the protective paste system to the outside of the main
cable wires.
6.2 After water has entered the main cable system, depletion of the zinc
galvanizing to the main bridge wires can occur, with the water acting as the
electrolyte. In the case of the Bear Mountain Bridge this has been aggravated by
the flat wire serving bands remaining in the cable.

6.3 The bridge wire appears to be susceptible to stress corrosion of the main
steel wire itself. This can only occur after depletion of the zinc, but after
the zinc is lost, local corrosion of the steel wire seems to occur in
concentrated local positions, with resultant pitting and cratering.
6.4 The stress corrosion cracking appears to be induced to initiate its action
on the inside of the residual curvature of the wire. All three of the Bear
Mountain, Triborough and Mid Hudson Bridges have the initial crack planes of the
wire fractures originating on the inside of this curve. The wire from the Golden
Gate Bridge, exhibited relatively little
ferrous corrosion, moderate zinc
oxidation, but no cracking.
6.5 It is considered possible that the stress corrosion may be directly related
to the initial curvature of the wire. The approximate measured diameters of the
curvature are recorded in Table 1. The simple calculation of the equivalent
decoiling stress from the minimum radius of 1830mm on the Mid- Hudson Bridge
gives an indicative extreme fiber stress of 345N/mm Early research into stress
corrosion cracking indicated that the threshold for crack initiation is around
270N/mm Typically, dead load cable stress is somewhat below this on these
older bridges. However, the secondary "decoiling stress" appears to be
sufficient to exceed this threshold.
400 SAFE CAPACITY OF SUSPENSION BRIDGE CABLES

7. PROTECTION OF CABLES AND REMAINING CAPACITY

7.1 It is evident from the condition of the wires in the four bridges inspected
that the corrosion pattern is only initiated after failure of the protective
wrapping system to the main cables and the entrance of water into the cables. It
is also evident that even if the corrosion attack has commenced, it is possible
to slow, or even halt the attack by increased maintenance of the protective
wrapping system, thereby stopping the entry of water into the cable.
7.2 The primary recommendation therefore, is to undertake some means of
wrapping to the cables which will totally seal the cables from any entry of
water. This can be achieved by means of wrapping the cables with a neoprene
sheet wrap, comprising 150mm wide uncured neoprene tape wrapped around the cable
with a 50% overlap. The curing of the tape is effected by its exposure to air,
and the ends are sealed to the cable bands. The whole is coated with a three-
coat liquid chlorosulfonated Polyethylene paint system. Overall, this results in
a totally impervious flexible sheath to the cable.

7.3 This system can be applied either directly to the surface of the main
bridge wires or over the original wire wrapping system. It is considered
preferable to leave the original wire wrapping system in place if
of
possible, as
this will maintain the compaction of the cable wires. Removal the wrapping
wire can result in loss of compaction up to 5% on the diameter of the cable, thus
providing an increased voids ratio of the whole cable. The cable should
therefore be at least strapped at sufficient intervals to maintain sufficient
compaction. Under no circumstance is
without
it
recommended to use neoprene wrap alone,
a positive compacting provision.
7.4 Calculation of remaining cable capacity can be considered as a reduction in
the Factor of Safety to failure under load. Such reduced factors of safety have
been calculated for the Bear Mountain Bridge and Mid Hudson Bridge. Insufficient
data was available for this to be undertaken for the Triborough Bridge, but the
Factor of Safety is not considered to be at or near any level to cause concern.
not considered necessary to make any estimate of reduction of Factor of
It was for
Safety the Golden Gate Bridge, due to the good condition of the wires.

7.5 Conservative assumptions were made in projecting the available figures into
calculations for remaining cable capacity for both the Bear Mountain and Mid
Hudson Bridges. These were:
7.5.1 The percentage of broken wires as a proportion of all wires actually
observed at any location is considered to extend through the complete cable cross
section.
7.5.2 The strength of unbroken wires is taken as an average strength of the
laboratory test values for that location. It is to be noted that all such test
values were measured on previously fractured wire samples.
7.5.3 All breaks observed at any one location are effective in conjunction,
ignoring any transfer of load by friction and clamping.
7.6 The Factor of Safety on the Bear Mountain Bridge was calculated as having
been reduced from 3.74 to 3.25, and on Mid Hudson Bridge from 3.90 to 3.20. In
both cases, this still represents an ample reserve of capacity, and presents no
cause for alarm. However, the recommended measures referred to above for sealing
the cables have been put into effect and the cable wires will continue to be
monitored on a regulär basis to ensure that the corrosion attack is halted.
 m

Nonlinear Analysis of an Ex
Analyse non linöaire d'une coq
Nichtlineare Berechnung einer beste

Pere ROCA Alex. C. SCORDEL

Assoc. Prof. Prof. Em.
Univ. Politecn. de Catalunya University of Calif
Barcelona, Spain Berkeley, CA, USA

; ¦**:

m*. a

A
402 NONLINEAR ANALYSIS OF AN EXISTING PRESTRESSED SHELL

1.- INTRODUCTION

Although powerful numerical tools for structural analysis have been developed during the
past twenty years to simulate the nonlinear behavior of concrete structures, very few estudies
have been effectivelly carried out on their use for the nonlinear analysis of large span shell
existing structures. However, a knowledge ofthe complete structural response of shell concrete
structures through their elastic, cracking, inelastic and ultimate ranges, is usually difficult to
achieve by means other than a nonlinear analysis which takes into aecount both the geometric
second order effects and the nonlinear aspects of the true behavior of the materials, including
concrete cracking and crushing, yielding of steel, and the time-dependent effects of concrete
creep and shrinkage.
The present paper describes the nonlinear finite element analysis of an existing prestressed
concrete shell structure: the Ponce Coliseum roof, constructed in Puerto Rico in 1971.
Through this analysis, the different causes of nonlinear behavior are characterized and their
partial influence on the service and ultimate behavior is measured. The simulated behavior
as obtained by the analysis is discussed in relation to the resulting reliability of the structure
under instantaneous vertical loading.
The original structural design of the Project was carried out during the period 1968 to 1970
by a Joint Venture of T.Y. Lin International, San Francisco, California, and R. Watson,
Engineer, and Sanchez, Davila and Suarez, Engineers of San Juan, Puerto Rico.

2.- DESCRIPTION OF THE SHELL

Ponce Coliseum consists of a 82.6 m span prestressed HP shell roof built in Ponce, Puerto
Rico, in 1971 for the 1974 Pan American basketball chanpionship games. The complete roof
is made up of four similar 10 cm thick saddle type shells connected to interior and cantilever
edge beams to form a structure supported by four piers at the low points located at the
centers of the four exterior panels. The high points of the shell, which rise 12.2 m above the
low points, are at the four corner tips and at the center of the shell. The clear spans between
opposite support piers in the two directions are 82.6 m and 69.2 m (Fig. 1). The cantilever
edge beams are supported only at the piers and have a constant width of 76 cm throughout
their entire length. These depth decreases gradually from the abutment to the corner tips,
as may be seen in Fig. 2.

Reinforcing for the shell consists of reinforcing bars in the top and bottom surfaces throughout
the shell with additional bar reinforcement added in the zones adjacent to the beams and
near the tips. Prestressing is provided as well by a two-way set of tendons parallel to the
straight line generators. The cantilever edge beams contain both normal reinforcing steel and
prestressing steel, while the interior beams contain only normal reinforcing steel and are not
prestressed.
The final original design used for the structure in 1970 was based on a detailed Computer
analysis using a linear elastic finite element model which coupled a system of one dimensional
beam type elements with another system of two dimensional triangulär plane stress finite
elements. This analysis showed the high degree of interaction that occurs between the shell
and the beams when subjected to their dead load, so that the resulting stress states can
not be completely understood using membrane theory, and coupled states of axial force and
bending moments appear both in the beams and the shell.
 P. ROCA - A.C. SCORDELIS - C. MOLINS 403

"[Pier
— __ 1..l_
zz^r==—
1
1
i

35.4

Pier Center Pier

1 i

6.1
12.2
- <

DIAGONAL ELEVATION VIEW

U Pier i2
I

L 42

Fig. 1. General dimensions of the structure

k -
H#9T
*5 STIRRUP j c^*?6 v—— 0*96 J^ 4*9F>
TIE5 AT 6"| 7" 7" 6" &'
t
6" 6* £»"
^ -
j I I I
i

M 1* + + ^U^
(5.Z hv)
(46 cm)
r,«300'vlvPR>CSTR.E55)
F, 5O0
- 53 m
(135 cm)
- 9* in ™# y
Ce38 cm)
3?
Fig. 2. 42 m cantilever edge beam

Prestressing in the edge beam was used to partially balance its dead load bending moments
and deflections while increasing its axial forces, thus resulting a set of stresses found acceptable
for design. In the shell, the remaining tensile forces after the introduction of the prestressing
were resisted by providing normal steel reinforcement. Also, to minimize cracks, prestressing
shell tendons were used in combination with reinforcement. For bending moments in the shell
adjacent to the edge beam and the tips, additional reinforcement was used.
Additional detailed information regarding concrete dimensions, reinforcement and
prestressing, as well as the original desing and construction of the Ponce Coliseum may
be found in [1-4].

NUMERICAL MODEL USED IN ANALYSIS

Chan's [5-7] previous numerical model for the nonlinear analysis of reinforced concrete shells,
extended by Roca [7-8] to include both beam and shell internal prestressing tendons, has been
adopted for the present study. This model includes both shell and beam elements to aecount
for shell Systems with edge beams, internal ribs or supports. The nine node Lagrangian
404 NONLINEAR ANALYSIS OF AN EXISTING PRESTRESSED SHELL

isoparametric element shown in Fig. 3 was adopted to model two dimensional curved shells,
while beams are simulated by adding two straight one dimensional beam elements to the
side of a curved shell element. Beam elements are prismatic but may have an arbitrary
cross-section made up of discrete numbers of concrete and reinforcing steel filaments.

Shell

beam element
a+ the reference
OXIS

Fig. 3. Adopted L7 shell and straight beam elements

The shell element is regarded as a multilayered system where each layer is assumed to be under
a biaxial state of stresses. The stresses and the state of the materials vary independently at
each layer to aecount for the mechanical changes of the materials throughout the loading
process. An hypoelastic biaxial concrete model is adopted together with a biaxial strength
envelope [5] to reproduce the compressive behavior of concrete and crushing. In tension,
concrete is assumed to behave as a linear elastic- perfect brittle material, aecounting for
tension stiffening in the reinforcement. Smeared cracking is initiated once the tensile stregth
is reached. A second crack is permitted to appear normal to the first. Reinforcement strength
is included as a set of additional layers of uniaxial behavior characterized by an equivalent
thickness. A bilinear diagram is used to model the elasto-plastic behavior of reinforcing steel.
Shell prestressing tendons are individually defined as arbitrary spatial curves contained in the
shell thickness. The geometric treatment of tendons is based upon a method where analytical
parametric expressions are used for the definition of the mid surface of the shell as well as the
tendon curves. Thus, an authomatic calculation of their geometric properties, as well as the
further updating of their whole geometry and forces in the nonlinear geometric analysis, are
possible. Prestressing in beams is treated consistently with the adopted beam straight finite
element, by dividing the tendon into a number of straight segments, each of which spans
a single beam element and is assumed to have a constant force. A multilinear stress-strain
curve is adopted to model the stress-strain relationship for prestressing steel. In addition,
the usual empirical formulae are used for stress relaxation and friction properties.
Nonlinear geometric effects are caused by the consideration of finite movements and finite
rotations. First, the compatibility equations are used with their quadratic terms to obtain the
strains from the displacement field. Furthermore, the geometry of the structure is continuosly
updated by adding the displacement increments to the current nodal coordinates, according to
the Updated Lagrangian Description. This requires an interactive procedure until convergence
is obtained meaning that the equilibrium condition has been finally satisfied on the deformed
geometry of the structure.
J% P. ROCA - A.C. SCORDELIS - C. MOLINS 405

4.- DISCRETIZATION OF THE STRUCTURE

Advantage is taken of the structural symmetry, so that only one of the four connected HP
quadrants is numerically modelled. Two meshes with a different degree of refinement were
defined: Firts, a coarse mesh having 15 shell elements and 24 beam elements, and, second,
a fine mesh with 55 shell elements and 52 beam elements. Although satisfactory results
were obtained using the coarse mesh for linear elastic analyses, it was necessary to use the
fine mesh to accurately reproduce the strongly local effects and local type of failure finally
observed in the ultimate behavior. The results that are presented below were thus obtained
using the fine mesh (Fig. 4).
3
f

V '
j M '
j t
M i 6
I II) | ii | ¦> 1 t» | « Aj„,

Fig. 4. Fine mesh including 50 shell and 52 beam elements

An automatic procedure for the generation of the shell finite element mesh and the courses
of the tendons has been used which needs as input data the parametric description of the
geometry of the choosen HP quadrant and the axial curves of the tendons included in it.
All mesh reinforcement keeps the direction of the HP generatrices parallel to the X,Y global
axes. Untensioned reinforcement in the shell is defined by a set of layers of equivalent thickness
and provided at each integration point in the mesh.
All shell tendons are defined with a cross steel area of Ap^ 3.13 cm2:, and stressed
to 1158iVy mm2. Their distribution is designed to fit the expected average compression
throughout the shell membrane.
Each straight beam element was defined with its cross sectional dimensions, flexural
reinforcement and torsional reinforcement, which are assumed to be constant throughout
the element. Each flexural reinforcement pattern consists of a set of discrete reinforcing bars
defined by their position in the cross section and their individual amount of steel. Torsional
patterns are determined by the core dimensions in x' and z' local directions, the area and
spacing of hoop reinforcing bars, and the top and bottom longitudinal perimeter steel.
The four beam tendons included in the discretized quadrant of the structure produce a
total prestressing force of 3550 kN and 2440 kN respectively for the 42.10 m and the 35.35
m edge beams. In both cases, the minimum cover between tendon axis and beam top
surface is about 11.5 cm. Based on the available design data, the following mechanical
properties were considered in the analysis: Concrete: Eq 26320 N/mm2, fc 28 N/mm2,
f[ 3.0 N/mm2, ec 0.002, v 0.15 and 7C 24 kN/m3. Reinforcing steel:
Es 210000 N/mm2, Esh 0 N/mm2, fsy 270 N/mm2, and esu 0.10. A prestressing
406 NONLINEAR ANALYSIS OF AN EXISTING PRESTRESSED SHELL

steel with fpu 1680 N/mm2 and fpy 0.7 fpu- is considered. A penta-linear stress-strain
diagram has been generated based on Standard shape fitting these properties.

5.- PERFORMED ANALYSES AND RESULTS

5.1- Linear Elastic Analysis

The study in [9] (1991) by Molins of several loading cases in a linear elastic analysis showed
a very good agreement with the similar results first obtained by Lo and Scordelis [1] (1969),
thus illustrating the suitability of the geometric modelling adopted for the present study. The
discussion of the linear elastic analysis, together with more details relative to the nonlinear
analyses which are presented below, may be found in [9].

5.2.- Nonlinear Geometric Analysis (NLG) with Prestressing

In this case, the design live load Qq (1.437 kN/m2) is increased indefinitely until a maximum
load is obtained for the structure. The materials are considered to be linear elastic, so that
no limiting tensile or compressive strengths are defined for the concrete or the steel, while
the nonlinear effects caused by finite displacements are considered.
A failure was obtained for a vertical load of 1.0DL + 32.3Qq> followed by a descending post-
failure branch (Fig. 5). The maximum deflection at failure is reached in the external X—edge
beam at a distance of 11.0 m from the top (Fig. 6). The obtained geometric instabilty seems
related to an excessive deflection of the X—edge beam, which tends to work like a simply
supported beam rather than a cantilever, having a support at the tip of the Y—edge beam.
Thus, the geometric failure is partly related to the geometric asymmetry of the structure.

5.3.- Nonlinear Geometric Analysis (NGL) with no Prestressing

A similar type of analysis was performed without any prestressing to study this effect. No
significant change was observed during the loading process for live load, as illustrated by Fig.
5. As known, prestressing should not affect the global nonlinear geometric behavior of the
structure since the set of equivalent forces caused by prestressing consitute a self-balanced
system.

5.4.- Nonlinear Material Analysis (NLM) with Prestressing

In this case, the limiting material strengths defined in

See. 4 are considered, together with the
nonlinear constitutive equations included in the numerical method of analysis. Displacements
and rotations are assumed to be small.
An ultimate load of 1.0DL + 4.23Qq ls obtained (Fig. 7). The maximum vertical deflection
at failure appears at the tip of the HP quadrant and reaches 22.1 cm. Almost no post-
failure branch is obtained, so that a rather brittle behavior is detected. An inspection of the
analytically obtained state of the materials shows that the failure is mainly due to exceeding
the tensile strength at certain membrane zones near the HP tip which are under-reinforced
while submitted to a state close to pure tension.

5.5.- Nonlinear Geometric and Material Analysis (NLGM) with Prestressing

The coupling of both types of nonlinearity, material and geometric, caused a reduction in the
 P. ROCA - A.C. SCORDELIS - C. MOLINS 407

oo,

^*:
»¦

f a"*" i
1

//
ft rf' t

7 ff
i
j/—Ik
* r
{

if
4 ff —j—i-
i.
ANLG
x--x--Pr es t r no d e A
r O--0-- - - -j - n
o d e C
// ¦-n o pre tr node
s A
n o d e B
n o d e C

Fig. 5 Load vs. vertical deflection for Fig. 6 Vertical deflection in shell at
NLG analysis, (a) with prestressing ultimate load for NLG analysis
(b) without any prestressing. with prestressing.

Ns>^

ES
i. cO>

¦+ ANLM node A
-°ANLM node
¦o ANLM node C
- ANLGM node A
o ANLGM node
-• ANLGM node C

V Disp (m)

Fig. 7 Load vs. vertical deflection for Fig. 8 Vertical deflection at ultimate
NLM and NLGM analyses load for NLGM analysis.
ultimate load of the structure with respect to the analysis in which only material nonlinearities
were considered. The obtained ultimate load is now 1.0DL + 3.S7Qq, showing a drop in the
maximum live load Q of 8.5% from that for the nonlinear material alone analysis (Fig. 7). A
maximum deflection of 22.1 cm for the ultimate load, appears at the tip of HP quadrant. As
Fig. 8 shows, the deformed shape of the roof is qualitatively similar to that one obtained by
the nonlinear material analysis. The influence of the nonlinear geometric effects may be seen
in the slight reduction of the ultimate load and in the ductility acquired in the post-failure
behavior. However, failure is still due to excesive cracking of concrete near the tip of the
shell. Considering the live load capacity of 3.87Qq found in this nonlinear analysis, in which
Qq was the design live load of lA37kN/m2, it can be said that a design level of safety for the
408 NONLINEAR ANALYSIS OF AN EXISTING PRESTRESSED SHELL

structure exists, and the adequacy of the original design process is shown for the structure
subjected to instantaneous vertical loading. Shell and edge beam prestressing affect in a
strong way the Performance of the structure not only during service conditions, but also at
the ultimate ränge.

6.- CONCLUSIONS

A numerical model for the nonlinear geometric and material analysis of prestressed concrete
shells has been used to study an existing long span prestresed concrete shell structure: the
Ponce Coliseum, in Puerto Rico. Using an accurate discretization which included both shell
and beam finite elements, several different analyses were carried out under vertical loads to
study the service and ultimate behavior of the structure, together with its failure mechanism.
In an analysis considering both the geometric and the material nonlinear effects, the designed
safety of the structure under short-term vertical loading was demonstrated.
Some aspects not yet included in the analysis, such as the influence of the actual construction
sequence, as well as the time dependent effects of creep and shrinkage of the concrete and
the long-term losses in the prestress are planned for future studies.

REFERENCES
1 Lo, k.S., Scordelis A.C., Design of a 271 ft. Span Prestressed HP Shell, Proceedings ofthe
IASS Congress, held at Madrid, October 1969.
2 Scordelis, A.C., Lin, T.Y., Ponce Coliseum Shell Roof in Puerto Rico, Proceedings of the
IASS Congress, held at Calgary, July 1972.
3 Yang, Y. C, Record Concrete Shell Roof Featured in Coliseum, Civil Engineering
Magazine— ASCE, June 1972.
4 Lin, T. Y., Kulka F., Lo, K., Giant Prestressed HP Shell for Ponce Coliseum, Journal of
the Prestressed Concrete Institute, Vol. 18, No. 5, September-October 1973.
5 Chan, E.C., Nonlinear Geometric, Material and Time Dependent Analysis of Reinforced
Concrete Shells with Edge Beams, Division of Structural Engineering and Structural
Mechanics, University of California, Berkeley, UC-SESM Report 82-2, December 1982.
6 Scordelis, A.C., Chan, E.C., Nonlinear Analysis of Reinforced Concrete Shells, Proceeding
of Symposium on Computer Applications in Concrete Technology held at American
Concrete Institute Convention in San Francisco, March 1986, ACI Special Publications,
SP-98 1987.
7 Scordelis, A. C, Nonlinear Material, Geometric and Time Dependent Analysis of
Reinforced and Prestressed Concrete Shells, Bulletin of the International Association for
Shell and Spatial Structures, Vol. 31-1,1,2, Nos. 102,103, April-August 1990.
8 Roca, P., Un modelo de analisis no lineal para el estudio del comportamiento
de estructuras laminares de hormigon pretensado, Departament d'Enginyeria de la
Construccio, Universität Politecnica de Catalunya, Barcelona ,September 1988.
9 Mohns, C, Estudi del comportament en servei i fins a ruptura d'una lamina pretensada
real: el Coliseo de Ponce, Universität Politecnica de Catalunya, Barcelona ,1991.
J%

Remaining Strength of B
Capacitö restante des
Resttragfähigkeit von B

Kors NOORLANDER
Civil Engineer
Public Works Rotterdam
Rotterdam, The Netherlands

r
410 REMAINING STENGTH OF BRIDGES IN ROTTERDAM

1. INTRODUCTION
The Public Works Service of the City of Rotterdam is responsible for the maintenance of more
than 700 bridges with a total value far exceeding 109 Ecu. Lifetime varies from new to over a
Century. The spans vary from some meters to over 270 m. Nearly every type of construction is
present.
It is quite clear that, with such a diversity of structures, it can be expected that a considerable
number is not confirming to present day design Standards. Remaining strength of these structures
is of vital importance in decisions concerning maintenance or replacement.
In this paper selection criteria, remaining strength assessment and some selected cases will be
dealt with.

2 SELECTION OF STRUCTURES

2.1 Selection criteria

Selection of structures can not be done on the base of an elaborate structural analysis. The
intention of the selection is to decide which structures have a high priority. Because it is out of
question to base the selection on a thorough analysis good engineering judgement is the only
tool available. But of course there are some criteria which can be helpfiil in this respect: age of
the structure, damage reports, maintenance reports, absence of drawings and/ or calculations,
changed loading conditions, Settlements, problems with similar structures.
Age is, of course, the most important criterium, but damage and maintenance-reports are also
quite useful in this respect.

3 ASSESSMENT PROCEDURE

3.1 InfroductiQn
The procedure for the assessment and, as pointed out in the introduction, the loading and finally
the risk involved is in nearly all cases the same.

3.2 Procedure
After the decision to evaluate a certain structure the following procedure is applied:
a Retrieval of archival material
This involves the material in the technical and administrative archives, historical archives
of the municipality, old engineering handbooks, historical descriptions of the City,
whatever may be useful.
In a number of cases drawings as well as design calculation are absent, because the
original ones were destroyed during the war.

b Review of drawings and design calculations

In this stage the reliability of drawings and calculations is established. Drawings are
compared with the real structure; calculations are checked, especially with respect to
modelling and loading.

c Drawings
 K. NOORLANDER 411

Depending on the Situation drawings should be updated, or, if necessary, completely

renewed. This should be based on a recent survey. Under no circumstances it is allowed
that any serious work is done before this step is completed.

d Updating design calculations

If the reliability of the existing calculation is not enough it should be updated, or even
completely renewed. This can be a quite laborious Operation. After this phase it is
possible to make conclusions on the static strength of the structure and to perform the
fatigue analysis.

e Fatigue life assessment.

Based on the updated calculations and combined with loading spectra derived from recent
traffic counts, and, in the case of movable bridges, bridge Operation records, the life of
the structure is determined.

f Test loading
In a lot of cases the structural system leaves a lot of questions with respect to the
reliability of the modelling. In these cases test loads can be condueted.

g Inspection
After static and fatigue analysis detailed directions for the inspection of the structure are
given. These can involve searching for cracks in members or welds, corrosion, loose
bolts or rivets, specific locations and so on.

h Final conclusions
Based on the finding of the foregoing steps the real state of the structure is determined
In this conclusion static strength and fatigue life are the main items.

i Recommendations.
Depending on the findings actions are recommended; these can be:
Strengthening of the structure
Intensive periodical inspection of eritieal parts
Partly or complete renewal

4 CRITERIA

Relating remaining strength of a structure to a fixed value in a Standard is, to my opinion, not
giving any answer to the question how safe or dangerous it is, because strength as well as
loading are both stochastic in nature. Risk is the only criterium which has any value for taking
decisions. The level of risk that is acceptable depends on a lot of circumstances.
Most important point is the question how serious the collapse really is. Is it a complete collapse
of the structure or is it merely regarded by the public as a minor nuisance comparable to a
pothole in a road-surface.
So minor damage is accepted by us in a lot of cases provided it is detected within a reasonable
time.
Another important point in the possibility of inspection. If it is possible to detect damage just in
time we can aeeept a greater risk.

It is quite clear that in the case of a total collapse of the structure higher levels of safety are
412 REMAINING STENGTH OF BRIDGES IN ROTTERDAM

required.
In all these cases a be in our judgement. After all if we are
lot of commonsense should
predicting problems there is a chance of a near miss. And the smaller safety-margin the greater
the chance.
That is the reason that we use recent observations and measured intensity of the actual traffic on
the bridge. BS 5400:prt 10 is extremely useful because, in this respect, the level of safety can
be varied.

5 CASES

5.1 Bridge with hot-rolled beams and oversized cut-outs.

This fixed bridge which is of a very simple structural concept of parallel INP380 0.79 m'
c.t.c, with a span of 5.83 m' covered with wooden shelfs.

^^'ZyW^

INP380

Y INP38Q 7k§-
V
780
H

figure 1 Oversized cut-out near support

5.1.1 Selection
The structure was selected because the drawing gave rise to questions with respect to details.

5.1.2 Existing documents

Historical research revealed that the bridge, originally, was a movable bridge. Between 1940
and 1945 it was converted to a fixed structure using the original beams. On the only available
drawing a cut-out near one of the abutments showed up. No dimensions were available. In the
stress-calculations no reference was made to this detail.

5.1.3 Strength assessment

The analysis of the bridge was preceded by a measurement of the cut out. At the same time the
corner of the cut-out was inspected by dye-penetrant; no cracks were found.
On behalf of the existing design calculation it could be concluded that the bridge could be rated
as Class 45 which is regarded as quite sufficient in the occurring Situation.
A detailed analysis revealed that the cut-out gave a reduction of the Classification to 22; if the
stress-concentration around the corner was taken into aecount the Classification was further
reduced to 15; equivalent to a wheel-load of 12.5 kN. The collapse load, with plasticising
crossection occurred at a wheel-load of 69 kN (without impact) that explained why there was no
collapse.
 K. NOORL

'•'
* *
,,:;:
e
jf

¦¦ &
fiel '~'Wßm
-¦* 1 -isfl»

-¦ "---:y

" 'r

Photo 1

Recently we analyzed the detail by finite elements.

former calculations by hand.
414 REMAINING STENGTH OF B

V)

Photo 2 constructio

Compared to die loadings specified in die Standard

th
4 K. NOORL

•=!
416 REMAINING STENGTH OF B

Photo 4 detail of m

6. CONCLUSIONS

Experience in the past 25 years has learned us that

 417

Probabilistic Re-Analysis of Existing Offshore Structures

Contröle probabilistique de plates-formes marines existantes
Probabilistische Überprüfung existierender Offshore-Konstruktionen

Marcello BERANGER Dimitris DIAMANTIDIS Gian Maria MANFREDINI

Nuclear Engineer Professor Director of Operations
Snamprogetti S.p.A. Fachhochschule Regensburg D'Appolonia S.p.A.
Milan, Italy Regensburg, Germany Genoaf Italy

Marcello Beranger, born 1962 in
Italy, got his Laurea in Nuclear
Engineering at the Polytechnic of
Milan, Italy Afterjoining Snamprogetti
in 1988, he has been involved
in several studies on structural
reliability analysis of components
for the Offshore, Petroleum and
Chemical Industries. Presently he
is a Project Safety Leader of the
Dimitris Diamantidis, born 1957 in
Athens, Greece, got his Diploma
and Doctor degrees in Civil Engineering
at the Technical University of
Munich, Germany. Since then he
has been senior research engineer
with Det Norske Veritas in Oslo,
Norway, and Project Manager with
D'Appolonia S.p A. in Genoa,
Italy. From September 1992 he is
Gian Maria Manfredini, born 1943
in Italy, holds a Laurea in Mechanical
Engineering from Polytechnic
of Milan, Italy, and a M.S. in Civil
Engineering from Michigan State
University. He has been involved
in research, analysis and design
related to steel and concrete
structures for the Nuclear,
Offshore and Transportation Industries.

Safety and Reliability department Professor of Structural Engineering He is currently Director of

of Snamprogetti S.p.A. at the Fachhochschule Regensburg Operations at D'Appolonia S.p.A.

SUMMARY
When the necessity for re-qualification of an offshore structure arises, the basic problem encountered is
that decisions may not be supported by code criteria. Current code criteria are not flexible in terms of
additional information aquired after the design stage. One has to tum to methods based on reliability
analysis supported by the information gained from field experience, maintenance and monitoring. These
methods are briefly discussed here and three case studies are presented to illustrate their applicability to
various situations.

r£sum£
Lorsque se präsente la näcessrte de reclassement d'une plate-forme marine, le Probleme principal est que
les decisions ne peuvent s'appuyer sur les criteres de projet existants. Les criteres de projet courants ne
se prötent guere ä l'assimilation des informations supplämentaires acquises apres la phase d'ötude d'un
projet. II faut par consöquent adopter des möthodes basöes sur des analyses de fiabilite en exploitant les
donnäes obtenues directement sur le terrain, dans le domaine de la maintenance et du contröle. Ces
möthodes sont brievement expliquöes et trois cas sont präsentes pour illustrer leur faculte d'application
a diffgrentes situations.

ZUSAMMENFASSUNG
Das Hauptproblem beim nachträglichen Nachweis existierender Offshore-Konstruktionen besteht darin,
dass Entscheidungen nicht von Normenkriterien getragen werden können. Aktuelle Normenkriterien sind
nicht flexibel genug, um zusätzliche Erkenntnisse nach der Bemessungsphase zu berücksichtigen. Mann
muss zu Methoden der Sicherheitstheorie zurückgreifen, unterstützt von Informationen, die während der
Betriebs- und Überwachungsphase der Konstruktion gewonnen wurden. Diese Methoden werden hier kurz
diskutiert und drei Falle werden beschrieben, um ihre Anwendung in verschiedenen Situationen
dazustellen.
418 PROBABILISTIC RE-ANALYSIS OF EXISTING OFFSHORE STRUCTURES

1. INTRODUCTION

Offshore exploitation is nowadays a mature field Several platforms have been in

Operation for over twenty years, often exceeding their original design lifetime and
experiencing extreme environmental actions, damages and possible reduced maintenance
and repair due to periods of depressed international offshore market Re- qualification of
these structures for extended lifetime is a major issue in the offshore industry Offshore
Operators must balance their decisions between required levels of safety and available
budgets

This has resulted in the rationalization of the decision process and of the re-analysis
procedures necessary to re-qualify existing offshore structures Efficient and
sophisticated techniques have been developed in the last decade for that purpose which
incorporate modern structural and reliability analysis methods

This contribution discusses experience gained from the implementation of reliability

methods and appropriate criteria for the re-qualification of existing offshore structures
The difference between prior uncertainty modeling and reliability updating on the basis of
available information is emphasized The methodological approaches are illustrated in
case studies dealing with re-qualification of platforms in various offshore locations

2. OVERALL RISK DECISION APPROACH TO OFFSHORE REQUALIFICATION

Various aspects have to be taken into aecount in the decision process for offshore
platform re-evaluation, and consequently several parties might be involved, each of them
contributing with, or requiring, different types of information The final decision is
gradually reached by pooling all these contributions into one These contributions are
Coming from

• Design

• Field experience

• Re-qualification analysis

• Economical analysis

The above listed contributions lead to the collection of information that are of very
diverse nature, e g in terms of type of data and category of persons/deeiders \*ho
provide them In particular, both numerical and non numerical information have to be
used

The general practice in re-evaluating offshore structures is that of relying on a "rational

approach" The safety and economy implied by certain decisions are evaluated by means
of both statistieal/probabilistie considerations (risk analysis), and economical analysis
(cost/benefit analysis)
 M. BERANGER - D. DIAMANTIDIS - G.M. MANFREDINI 419

3. USE OF RELIABILITY METHODS

3.1 Prior uncertainty analysis (design stage)

The largest uncertainties exist at the design stage. Not only is the loading environment
partially known but also the specific quality of manufacture and construction is under
control to a limited extent. Furthermore the actual load-effect relationships involve some
systematic but also random variability. All these uncertainties can be modeled by random
variables. In order to provide safety criteria appropriate limit states "g" have to be
defined; failure occurs when g < 0.

Safety is assured by requiring that the limit state will be reached with a small probability,
which is dependent from the Joint probability density function ofthe stochastic variables
defined in the problem.

Numerical methods for reliability calculation have been developed during the past decade
and the first order reliability methods FORM have been recognized as very accurate and
efficient [1-3]. The probability of failure by using FORM is estimated by:

Pf**(-ß) (i)

where 0(.) is the Standard normal distribution function and ß is the reliability index.

3.2 Updating through additional informations (re-design stage)

Additional informations gained during lifetime of the structure can be quantified and
implemented in order to update original or codified safety levels. Two basic cases are
briefly described next, updating of limit states and updating of random variables.

In many cases, inspection results can be interpreted as artificial limit states and can be
applied directly in the updating procedure ofthe originally estimated failure probabilities.
The updated failure probabilities Ff are then evaluated as conditional probabilities. The
following two fundamental cases are classified:

P'f=P[g0(X)<0 lgl(X) 0] (2a)

P'f=P[g0(X)<0 lgl(X)>0] (2b)

where go(X) is the original limit State function and g,(X) the artificial limit State function
formulated from the inspection results. The vector (X) includes all basic variables in gj
and go. Examples of equation (2a) are measurements carried out under proof - loading,
observations of crack lengths, and of deformations (or Settlements). A typical example
for equation (2b) is the survival ofthe structure (or ofthe structure components) under
high loads. Updated failure probabilities are derived within a first-order reliability method
framework [3]. The influence of a possible lifetime extension can be taken into aecount
420 PROBABILISTIC RE-ANALYSIS OF EXISTING OFFSHORE STRUCTURES

by modifying the distribution function of the time dependent basic variables in the
original limit State function go(X)
In some cases only specific influencing parameters such as material strength or
geometrical dimensions are inspected After the inspection, updated distributions are
given to the influencing random variables resulting to a decrease or an increase of the
reliability index.

3 3 Compatibility of safety requirements with reliability analysis techniques

Safety requirements for structural design and re-design cannot be established on the basis
of pure science A rational approach for the requirements must be a synthesis of
1 basic understanding of structural behavior,
2 basic knowledge of current practice concerning workmanship, equipment, and
construction methods etc,

3 experience gained from the behavior of existing structures,

4 economical, social, political and legal considerations,

5 reliability theory,

The classical codified safety requirements, for example the partial safety factor format,
cannot be applied in the evaluation of the reliability of existing structures since they are
not flexible in terms of additional information The probabilistic reliability theory
constitutes a rational tool for updating additional information and comparing the actual
reliability level with the assumed reliability level (inherent in present codes)

4. CASE STUDIES

4 1 Mechanical impact re-assessment of North Sea platform

Re-qualification of an existing platform located in the British sector ofthe North Sea was
required both for production Systems reliability and structural integrity The platform is a
"K"- bracing steel jacket type, located at a water depth of about 142 m The aim ofthe
study was to extend the platform lifetime and to update the overall safety level

Structural safety assessment was performed taking into aecount

• updated environmental load,

• re-analysis of ship traffic in the vicinity ofthe platform,

• structural re-analysis ofthe jacket,

• jacket inspection
 M. BERANGER - D. DIAMANTIDIS - G.M. MANFREDINI 421

Due to over-stressed members found after the jacket structural re-analysis, the main
emphasis was paid to assess the platform damage resistance after a ship collision The
scope was also to verify the British authority requirements which State that the structure
should be capable of withstanding an impact from a 2500 tons vessel with a velocity of
0 5 m/s Assuming that all the kinetic energy is absorbed by the installation and that an
added mass factor of 0 4 must be applied an impact energy of 0 44 MJ is obtained
Norwegian authorities requirements were also considered, which specify a design impact
from a supply vessel of 5000 tons at a velocity of 2 0 m/s and an added mass factor of
0 4 for broad side collisions and 0 1 for bow collisions This results in impact energies of
14 MJ and 11 MJ respectively

The probability of failure is evaluated as follows

i c m

where Pc is the probability of platform-ship collision and Pm is the member probability of

failure conditioned to ship impact

The failure probability analysis associated to passing vessels was simplified by the fact
that Pm was considered equal to unity A ship traffic analysis in shipping lanes near the
platform location was performed, considering also updating for future ship traffic
development

In case of a visiting vessel (supply vessel) platform collision, the limit State can be
defined as

g E - E

where Em is the energy absorption capability of the impacted member and Es is the ship
impact energy The probability of failure has been calculated using FORM by introducing
the following random parameters

- ship mass and impact velocity,

- strength and geometric characteristics ofthe impacted member

The results have shown that the Norwegian authorities requirements are not fulfilled, the
British authorities requirements are fulfilled and the annual probability of failure
considering revised data on distributions of impact masses and impact velocities found in
the literature is in the order of 103

The reliability analysis has indicated that the structure cannot sustain an impact from a
vessel since stress redistribution is not possible This is due to the high number of
members over-stressed and to the type of joints

In these platforms the ability ofa "K"-bracings panel to transmit shear is lost if the
compression brace buckles From the study it was concluded that appropriate provisions
to avoid vessel impact or to increase the impact load absorption capability must be
considered
422 PROBABILISTIC RE-ANALYSIS OF EXISTING OFFSHORE STRUCTURES

4 2 Fatigue re-analysis of steel gravity platforms in the Gulf of Guinea

For offshore structures the fatigue limit State is governing the structural dimensions of
several member and particularly of jomt connections Therefore, efforts of inspection
campaigns for re-qualification purposes are aimed to assess the structural integnty of
such joints, and to detect possible cracks and corrosion damage However, due to the
considerable costs, only a limited number of joints can be inspected

The inspection results together with the design data representative of the platform have
been first implemented in a Software system that integrates the data base with procedures
of analysis required in a re-qualification process [8] Procedures include Statistical
analysis, damage aecumulation analysis, crack growth analysis

The Statistical analysis package was required to correlate damage joint characteristics in
order to define probability distnbutions of representative deterioration parameters for not
inspected joints Where applicable, such an approach has been integrated in three Steps

• creation ofthe sample population and definition of explanatory variable such as water
depth, element type, joint geometry, matenal, design fatigue life, etc

• definition of Statistical model for the detenoration parameters (i e pitting depth,

probability of crack presence, crack depth, actual thickness) by using analytical
methods of variance,

• computation of response for not inspected joints

The damage aecumulation package has allowed to compute, for each joint, the
probability of failure due to fatigue in the hypothesis of linear damage aecumulation
(Miner's rule)

A crack growth analysis was performed, to determine failure probabilities by means of a

fracture mechanics approach The model based on the Pans and Erdogan [10] law was
established for a semi-elliptical crack in an infinite plate according to current practice
[9,11] The stochastic model applied to fatigue crack growth aecounts for uncertainties
in loading, initial defects, material parameters and in computation of stress intensity
factors The crack growth model has been combined with FORM to compute failure
probabilities

Figure 1 shows the decrease of the reliability index with time for a eritieal joint The
three curves represent

• pnor reliability index without considenng inspection results,

• postenor reliability index "conditioned" by inspection results, l e

- joint ltself
has been inspected and no crack, or a crack of given
depth has been found,

- other joints of similar characteristics have been inspected and found

without crack, or with a crack of given depth, thus resulting in an
updating ofthe considered not inspected joint
 M. BERANGER - D. DIAMANTIDIS - G.M. MANFREDINI 423

Figure 2 illustrates results obtained for another eritieal joint of the same platform and
emphasizes the time dependency ofthe following parameters

• prior reliability index without information through inspection,

• posterior reliability index corresponding to conditional failure probability index

"given" that no crack has been observed,

• ränge of acceptable safety level (based on the considerations mentioned before),

• inspection effort to achieve acceptable safety level for a desired additional lifetime

The results have been generally used to judge the Inspection Repair Maintenance (IRM)
program versus extended life of service The benefits associated with the use of the
illustrated procedure include

• improved safety related to the knowledge of the risk associated to the extension of
the platform life,

• sound evaluation of the relative importance of detected defects on the structural

safety,

• optimization of future survey by Screening out not eritieal joints

4 3 Re-assessment of seismic loading for a platform in the Adriatic Sea

This example deals with the re-assessment of seismic loading for a platform offshore
Ancona in the Adriatic Sea This platform has been designed in the late 70's according to
the following criteria

• design lifetime of 20 years,

• probabilistic evaluation of ground acceleration based on the attenuation law of
McGuire [12] which is based on strong motion recordings from Western United
States,

• definition of two different peak ground acceleration (PGA) levels [13], Strength
Level Earthquake (SLE) associated to a return period of 100 years, corresponding to
an acceleration of 0 11 g, Rare Intense Earthquake (RIE) associated to a return
period of 1000 years, corresponding to an acceleration of 0 2 g

Figure 3 illustrates peak ground accelerations versus return periods for design and re-
design

After 10 years the safety of

the platform has been addressed together with the
consideration of a total lifetime of 30 years Therefore, the criteria used in the design
stage have been reviewed, evidencing in particular that recent developments allowed to
aecount for
424 PROBABILISTIC RE-ANALYSIS OF EXISTING OFFSHORE STRUCTURES

B Posterior Nl)

ß Postenor (I)

CO
GC <s>

ß Prior

1990 1995 2000 2005 2010

YEPR

Fig 1
Reliability index vs time for inspected (I) and not inspected (Nl) joints

Ponnuol IE 4
p Posterior
x
UJ
o <yX*

^U^^^fir^ ;.jjV^g---«*.wÄ:t^,-=c^* »!ijr~~-»^gLv.A~ LS/f'1 ~. 1

CE*

-s
¦#* I *
yissr.^r

y- ^-^t*^t£r~=Ö
Ponnuol =3E-4
co
a.

tu
or

B Prior

'
^990 1992 199U 1996 1998
'
2000
'
2002 2004
'
2006
'
2008
'
2010
YEOR
Fig 2 Reliability index vs time and recommended inspection program
 M. BERANGER - D. DIAMANTIDIS - G.M. MANFREDINI 425

• revised earthquake catalogues;

• specific attenuation relationship for the platform site [14];

• reliability based definition of the return periods at which SLE and RIE should be
associated to be in accordance with current safety philosophy in the offshore industry
[15]; SLE and RIE have been associated to 200 and 2000 years return periods
respectively.

Although the new return periods for the representative PGA's are more conservative than
those assumed in design, the application of the new attenuation law and the more
accurate characterization of seismicity allowed to obtain lower values for the peak
ground acceleration In particular, the SLE was associated to 0 09 g, while the RIE to
0 18 g, introducing a 10% - 15% reduction with respect to design values

The new results expressed in terms of PGA versus return period are also included in
Figure 3 demonstrating the lower values obtained through re-analysis The method in
Ref [16] has been then used and the results have demonstrated that earthquake is not
endangering platform safety, if compared to the original design assumptions

REQUALIFICATION DESIGN REQUALIFICATION

UtÖHjN ÜLfc LfcVfcL \ SLE LEVEL —^ ^— RIE LEVEL - RIE LEVEL

/
1

250
''
y
//
MCGUIRE, (1974)
*
o
-£ 200
SABETTA, PUGLIESE, (1987)
,* '
y'
E
o^ e
*
<
Ü ,* ,'
CL
,*'
O 150 .*'
H
£
UJ
^~''
^*" /
_J
"*
,-''' „
LU

8 10°
<
Q _«.-•"'
2 _ „-
—

3
O
__«.'"¦
cc ^Ä
ü 50 —
^—-^"^
^
s
CL

l i I
Q
10 20 50 100 200 500 1,000 2,000 5,000 10,(
RETLJRN PERIOD (years)

Fig 3 Peak Ground Accelerations obtained adopting design and re-qualification

attenuation laws
426 PROBABILISTIC RE-ANALYSIS OF EXISTING OFFSHORE STRUCTURES

5. REFERENCES

1 HASOFER, A M, LIND, N.C An exact and invariant first-order reliability format,

J Eng Mech Div ASCE, 1974, 111-121
2 HOHENBICHLER, M, RACKWITZ, R, Non-normal dependent vectors in
structural reliability, J Eng Mech Div ASCE, 1981, 1127-1138
3 MADSEN, H O KRENK, S LIND, N C Methods of structural safety, Prentice-
Hall, Englewood Cliffs, New Jersey, 1986
4 CEB-FIP, Model code, Paris, London, Berlin, 1978
5 DET NORSKE VERITAS (DnV), Rules for the design, construction and
inspection of offshore structures, Hßrvik, Norway, 1981
6 DIAMANTIDIS, D, Reliability aspects for design and re-design of offshore
structures, Proc 5th OMAE-ASME Conf, Tokyo, Japan, 1986, 29-36
7 RACKWITZ, R, Implementation of probabilistic safety concepts in design and
organizational codes, Proc 3rd Int Conf Struct Safety and Reliability, Trondheim,
Norway, Elsevier, Amsterdam, 1981, 593-614
8 VANZINI, R, ROSSETTO, P, CONZ, L, FERRO, G, RIGHETTI, G, Re
qualification of offshore platforms on the basis of inspection results and
probabilistic analysis, Off Techn Conf (OTC), Houston, 1989
9 DIAMANTIDIS, D RIGHETTI, G, FERRO, G, ZUCCARELLI, F Reliability
based re-qualification criteria for existing jacket platforms, Off Mech and Arctic
Eng (OMAE), Stavanger, Norway, 1991
10 PARIS, P ERDOGAN, F A cntical analysis of crack propagation laws, Journal
of Basic Engineering, Trans ASME, vol 85,1963,528-534
11 MADSEN, H O, Probabilistic and deterministic modeis for predicting damage
aecumulation due to time varying loading, DIALOG 5-82, Danish Engineering
Academy, Lyngby, Denmark
12 McGUIRE, R K,Seismic structural response risk analysis incorporating peak
response regressions on earthquake magnitude and distance, MIT Dept of Civil
Engineering, Research Report R74-51, 1974
13 American Petroleum Institute (API), Recommended practice for planning,
designing and construeting fixed offshore platforms, API RP2A, 19th Ed Dallas,
1989
14 SABETTA, F PUGLIESE, A, Attenuation of peak horizontal acceleration and
velocity from Italian strong-motion records, Bulletin of the Seismological Society
of America, vol 77, pp 1491-1513,1987
15 MOSES, F Development of preliminary load and resistance design document for
fixed offshore platforms, Final Report, API-PRAC Project No 85-22, American
Petroleum Institute, Dallas
16 BEA, R G, Reliability criteria for new and existing platforms, Offshore Technology
Conference, Texas, 1990
 Traffic Load Estimation by Long
Estimation des charges due au
de contraintes
Abschätzung von Verkehrslasten mit

Holger WAUBKE
Civil Eng. '¦MF
Techn. Univ. of Munich
Munich, Germany

J
Werner BAUMGARTNER
Dr. Eng.
Techn. Univ. of Munich
Munich, Germany
-
¦HfilMl
428 TRAFFIC LOAD ESTIMATION BY LONG-TERM STRAIN MEASUREMENTS

1. Introduction
In autumn 1991 the bridge Fischerdorf in Germany was opened. The Bridge crosses the
danube near Deggendorf and is connecting the motorway A12 Munich-Deggendorf to three
roads on the other side ofthe danube. To get a control ofthe theoretical calculations ofthe
complex construction as shown in figures 1 and 2 measurements with static loads were
projected. To get more information about the structural behaviour it was decided to use the
installations for additional dynamic long term measurements.

2. Bridge Structure
The bridge has two main girders which cross-section is shown in figure 2. Transverse beams
connect both main beams to rods which are connected to one middle arch. The cross-section of
the main beams are a steel construction with a concrete deckslab. The construction bears the
loads in a combination of torsion and bending ofthe main girders and the arch.

18.90m

1_ s\^
r!
^ j^=^-
i

13.10 7-10.90 13.10

102.50

1 _F _£ .D M

ABC
München
Deggendorf

flg. 1 Bridge Fischerdorf

¦ti
3350 5600 3200 3200 5600 3350

PrnT!^w^-''"" ..fi<gSgl -ZZZZZZZZZZ2

S>
1910
Y
Ä 1
5600 3500 3500 5600

flg. 2 Bridge Fischerdorf cross-section

 H. WAUBKE - W. BAUMGÄRTNER 429

3. Measurements

3.1. Measurement Equipment

Fourty points were selected at which the strains were measured. Four of them were at the arch
(point G1-G4) near the first rod. Another group of four points were installed at the first rod.
(Points H1-H4). The transverse beams have three groups of strain gages (D1-D4, E1-E4, Fl-
F4). The other measurement points (A1-A4, B1-B4, C1-C4, M1-M4, N1-N4) were at different
locations of one ofthe main griders, at the webs near the deckslab and the bottomslab.
At each measurement point one active strain gage and one for compensation were connected
to one Whetstone's half bridge to compensate temperature effects. The gages for
compensation were fixed to little steel plates which were seperated from the steel construction
with a special paste, with the consequnece that the plate has the same temperature as the main
construction without having its stresses. To get longer stability of the strain gages some of
them are capsuled. The long time measurement will also improve the stability of this different
type of strain gages. Eight ofthe fourty measurement points can be measured at the same time.
An amplifier HBM-MGA is used as a signal conditioner. The calculation of collectives or
spectra are done by different personal Computers with build in analog-digital-boards of type
Data Translation 2801A and Analogie MS-DAS 16.

3.2. Stress Collectives

Two personal Computers were used to evaluate the stress collectives permanently.The program
was developed at the Lehrstuhl für Baumechanik [1]. The hystereses counting method
(rainflow method) is used to classify the stresses into a matrix with 64 rows and columns. The
calculation is done without interrupting the measurement. Also hypothesis for life time
estimations are included, in this way nonlinear hypothesis can be calculated incrementally. A
calculation of nonlinear hypothesis afterwards is impossible, because ofthe unknown time
history ofthe stress-cycles. Additional information about life time estimation is given in [2]. At
the moment the results ofthe collectives at different points covering one week measurements
can be presented. The tree lines in figure3 show the collective after 2 hours, 20 hours and 160
hours.

AI

10 N/mm2

Ustzyklen H (lig. lustellunff) lttöfoe

fig-3 Stresscollectives of measurement point AI after 2 hours, 20 hours, 160 hours

430 TRAFFIC LOAD ESTIMATION BY LONG-TERM STRAIN MEASUREMENTS

The collectives after 160 hours at points G2 and H4 ofthe construction is shown in figure 4

G2 H4

20 N/mm2

20 N/mm 10 N/mm2

Lastiykln N U»f. laKttlltwg) Ustzyklc« M

(Uf. liKtelUnj)

fig.4 Stress collectives of measurement locations G2 and H4 after 160 hours

The stress cycles at the highest levels are not the result of the traffic loads but due to the
temperature gradient between day and night. Therefore about one cycle per day can be
calculated from the collectives. Strain-cycles due to high level stresses are measured at the
arch. The point AI near the road way shows much more stress-cycles at low level stresses than
the rod and the arch, for which the local stresses are not important.

3.3. Time Series

After the short time measurements with static loads were done in autumn 1991, some short
time measurements with dynamic loads followed. Figures 5 and 6 show the dynamic response
of the first rod (location Hl) to a transporter with a weigt of 91 tones. Figure 5 belongs to a
speed of 30 km/h. The influence of crossing wooden plank placed on the driving lane in the
middle ofthe bridge can be recogniced in figure 6. It shows the possible influence of dynamic
effects on fatique life.

Zeit reihe Uea Datei:*J3-log Kanal: 8 Zeitatatand [sec]:B,4088E-B2

-.308OE+81 §.88001+01 lin. M 4. 1.1984 23:53:50

ifr*fry*

8.808E+08 8,480E*8i 8.888Et81 8.128E+B2 8.160E+82 8.280E+82

fig. 5 Short time measurements at point Hl of a transporter without a plank.

 H. WAUBKE - W. BAUMGARTNER 431

Zeitreihe Heg Datei:*d4-losr Kanal: 8 Zeitabstand [sec]:B,4B08E-82

-.3BÖ0E+01 8.80801*81 lin. to 5. 1.1984 8:14:36

B.808E+80 8,4001+ei 0.888E+81 B.120E+B2 8,1601+82 6.2B0E+82

fig. 6 Short time measurements at point Hl of a transporter with a plank.

Some time series were recorded for about twenty minutes. Two clips of one ofthe records
(figure 7) are displayed in figure 8 to show the different response of point N4 due to crossing
vehicel In the first clip a static response ofthe bridge is visible while in the second clip a
dynamic reaction ofthe bridge is shown. The influence on fatique life ofthe second vehicle is
much greater than that ofthe first one. Therefore strain measurements as described in chapter
3.2 are necessary to calculate a more realistic life time estimation.

Zeitreihe Heg »atei:«e4cal Kanal: 3 Zeitaistand [sec]:B.2888E-81

-.4580E+81 8.18581+82 lin. fc) 23. 4.1992 12:19:51

UM u »*M u 4*

8.898E+88
n
8.262E*B3
1

8.5241+83
1

8.786E+83 8.185E+M
r 8.131E+84

flg. 7 Time record at point N4

ItitNi)* fctei:i«4c*l Iml: feitiisUnd [$«];!.2W8E-81 Xeitrei)» tatfi:ne4cal bui: leitafefenJ [$«]:!. 2 8»E-il
tej
" ' ""
Urs 3 3
-.45881*81 8.lB5tt*82 lin. lo 23. 4.1)92 12:19:51 -.38881*81 8.38881*81 lin. *- •••—««
k 23. 4.1932 12:19:51

m^^
§.?4BI*82 8.7881+82 •J88E*82 8.1281*82 1.1481*82 8, «81+82 1.9(81*83 8.9611*83 8.97(1*83 8.9I4E*83 8.9921*83 8.188E*M

fig. 8 Two clips ofthe time record with showing different reactions ofthe bridge
432 TRAFFIC LOAD ESTIMATION BY LONG-TERM STRAIN MEASUREMENTS

The measured static response of the bridge due to the crossing vehicel in the first clip can be
compared to a calculated influence line ofthe moment, which is porportional to the measured
strains, because the vehicel is driving with a constant velocity.

^ I • 1

©
N |

I 1 > ¦—H

© 1

fig. 9 Static influence line of deflection and moment at point N4 calculated with finite elements
method.
The frequency spectrum ofthe second figure has two significant frequencies which
correspondends to two neighboured eigenfrequencies of the bridge.

3.4.Averaged Frequency Spectrum

Using a group of time series the spectrum can be averaged using the amplitudes ofthe spectra
with no respect to the phases. Averaging ofthe complex amplitudes is impossible because a
measurement without triggering results in random phases. Averaging amplitudes smooths the
noise but does not eleminate the noise. The following figures show the frequency spectra with
traffic load after about 1400 averages. The effect of smoothing the noise for random signals
gives the possibility to determine the eigenfrequencies without using determined loads ore
impacts. Averaging the traffic load spectrum has similar the effect as using a banded white
noise excitation.

m fuer n4a Kanal 0 Frewenzakstand ü/secl :6.1221E-01

*l.E-2 AmpI lin. Fnq. lin. diskrete Darstellung Di 8. 1.1980 1:26:28
8.60

8.50

8.48 h

0.30

0.20

0.10

0.00
~I T
0.00 1.00 2.00 3.00 4.00 5.00

fig. 10 Averaged frequency spetrum for point N4

 H. WAUBKE - W. BAUMGÄRTNER 433

3.5.Standard Deviation of banded Frequency Spectrum

A possible way to control changements in traffic during time is to control the change in a
banded frequency spectrum between averaged groups. Therefore in a first step the frequency
spectra are averaged. In this case 10 spectra are chosen for one group. In a second step groups
are averaged with a calculation ofthe Standard deviation. In this example 140 averages of
groups are done. The deviation is ofthe same order as the average. This is the result ofthe
differences between day and night and at the weekend. To reduce this effect a special time
control is introduced into the program, which measures the groups always at the same time and
the same day. Long time Observation with this time control might get information about long
time changes in the traffic load.

FFT fuer gsnn4a Kanal 0 Frequenzafctand [1/secl :8.1221E-01

[dB! Mittel wert N= 18 l/48-0ktavband Di 8. 1.1980 1:26:28

0.508 1

fig. 11 Averaged banded frequency Spektrum at point N4

i i

5 iii
FFT fuer gsnn4a Kanal 0 Freguenzakstand [1/secl :8.1221E-B1
[dB! StandardaJbv. N= 10 1/48-Oktavtaid »i 8. 1.1980 1:26:28

0.500 1 5

fig. 12 Standard deviation ofthe frequency spectrum at point N4

434 TRAFFIC LOAD ESTIMATION BY LONG-TERM STRAIN MEASUREMENTS

4.Traffic Load Estimation

The direct calculation ofthe traffic loads based on the measured stresses is difficult. All driving
lanes have different influence lines. Also the speed ofthe vehicels are variable, therefore only
stochastic calculations are possible. Another problem is the dynamic response ofthe bridge to
harmonic loads as shown in chapter 3.3. This effect is dicussed in [3] there the response of a
beam to a moving harmonic load is displayed.

MOV1NG HARMONIC FORCE

V (l/2,t)/v0

v (l/2,i)/v0

v (U2,Ü/v„

fig. 13 Deflection of a massless beam to a moving harmonic load with different speeds [3]
Changements in the traffic load will produce changing stresses at the measured points. Using
long time measurements it is possible to control changing in traffic loads easily by Controlling
the development of numbers of stresscycles and their distribution. As an example an increase of
heavy vehicles will result in an increase of the high level stress cycles. As a second possibilty a
measurement ofthe resonant frequencies can control the development of vehicels with bad
dampers, which are predistined to get in resonance with the bridge structure. This vehicels can
reduce the fatique life of a bridge rapidly.

5.Conclusions
The results ofthe long term measurements can be used for life time estimation ofthe bridge.
Using transfer function a life time estimation of arbitrary points ofthe bridge can be calculated.
An exact estimation ofthe traffic load is difficult because ofthe dynamic response ofthe
bridge, but the measurements can control the development and the composition of the traffic
loads. The only way to calculate the traffic load is to use Statistical functions.
Because ofthe difficulties to develop a realistic load model it is better to measure the stresses
of bridges at dufferent points. The results of this points can be used to calculate different points
with respect to the dynamic eigenfunctions, the distribution in the frequency spectrum and the
stress collectives.

REFERENCES
1. WAUBKE, H.: Kontinuierliche Erfassung von Schwingungsbeanspruchungen mit paraleller
Ermüdungsbewertung unter Berücksichtigung von nichtlinearen Schadensakkumulationshypothesen,
Diplomarbeit Nr, 29, Lehrstuhl für Baumechanik der TU München, 1988
2. BAUMGÄRTNER, W., WAUBKE, H.: Service life estimation of bridges. To be presented
at: 2nd Int. Conf. on Bridge Management, University of Surrey, UK,1993
3. FRYBA, L: Vibration of solids and structures under moving loads, Noordhof International
Publishing, Groningen, 1972
 Monitoring of Trial Loiadings of Bri
Surveillance des charges d'öpreuve de
Probelastungen von Brückenbauwerken

Reinhard WOLFF
Dr. Eng.
SICOM GmbH
Cologne, Germany

Hans-Joachim MIESSELER
Civil Eng.
SICOM GmbH
Cologne, Germany
436 MONITORING OF TRIAL LOADINGS USING OPTICAL FIBER SENSORS

1. INTRODUCTION

For the monitoring of buildings with optical fiber sensors in principle two different types of
application can be distinguished. On the one hand the monitoring of concrete structures, e.g.
monitoring of crack formation and their further development and on the other hand monitoring of
prestressing tendons comprising fiber composite materials. For the monitoring of crack
formations in concrete structures the optical fiber sensors especially designed for this case of
application are directly embedded in new buildings at eritieal positions of the load bearing
structure known from statics. Their function is to provide information about eventually required
maintenance works at a very early stage so that the costs for repair works can be reduced at a
minimum. In already existant buildings - especially already damaged buildings - these optical
fiber sensors are installed at a later date in the structural elements which shall be monitored in
order to monitor cracks and their further development or the repair works which already have
been realized with regard to the load bearing capacity of the structural element. In case of fiber
composite prestressing tendons the optical fiber sensors are directly embedded in the fiber
composite bar in order to indicate its intaetness.

2. SENSOR TECHNOLOGY

In order to monitor buildings in a useful way it is necessary to develop appropriate sensors which
are able to guarantee reliable measured values during a long period of time. Nowadays the
intelligent processing of the high quantity of measured values is no problem due to the available
and efficient personal Computers.
The sensor technology is - regarding to the monitoring of buildings - only in the beginning of its
development. The application of strain gauges on the prestressing steel or the measurement of the
prestressing forces with the aid of load cells is not possible in case of prestressing with post-bond.
Moreover it is not a durable Solution in case of prestressing without bond.
Only the application of fiber composite materials facilitates a permanent control of the
prestressing element over its entire length due to the Integration of copper wire sensors or optical
fiber sensors. Even the monitoring of each individual bar is possible. The sensors are already
integrated into the tendon during its fabrication. The sensors indicate the integrity of the tendon
or they locate the damage.
Besides the monitoring of the prestressing elements the Observation of the stress/strain behaviour
of the concrete in the zone subject to tensile forces is very important. Therefor the tensile zone
above the piers and the spans are monitored permanently with integrated optical fiber sensors
with a measurement exaetitude of ± 0,2 mm.

2.1 The crack detection sensor

In the case of this type of sensor several optical fiber sensors or one optical fiber sensor are
stranded with each other with one or several steel wires. If this sensor - which is connected
normally to the structural element which shall be monitored in a distance of 1 m through a
bearing material - is exposed to tensile load, the sensor effect is created by a loss of light intensity
as a result of micro-bending. This light attenuation allows an integral measurement of the changes
of the length of the sensor with a long term aecuraey of ±_ 0,2 mm. The localization of these
extensions changes (e.g. cracks) is carried out with an Optical Time Domaine Reflector (OTDR)
by reflection measurements and a local dissolution of jh 0,5 m.
 R. WOLFF - H.-J. MIESSELER 437

wire Strand protective sheath

optical fiber

Figure 1:

2.2 The crack width sensor

The crack width sensor is a loop bended optical fiber sensor which is fixed at two bars. These
bars are directly connected to the structure on the right and on the left of the joints or cracks
which shall be monitored. The sensor monitors and measures permanently changes in the joints or
crack width with a measuring sensitivity of 0,02 mm and a measurement ränge of 0,1 -10 mm.

optical fiber sensor transmitter structural cracks

\\\\\\V\\\\^VN\^\^XN\\\^>^^^\\ix\\\\Vl\N\'N\\\^
Q> CEED
X
\ f w >
0 \\^
^^\^\\K\\^V^^\\\\\K\\^^^^\^^\r^^\\\¦-^^^^.\f-\^
\ roroivor \
recording ränge for crack width monitoring load transfer rail

Figure 2:

2.3 The reflector sensor

A light signal is passed through the optical fiber sensor. The reflectors (Fig. 3) reflect part of the
light while non-reflected light travels to the next reflector. In this way up to 30 measuring points
can be monitored with a single sensor. Velocity measurements provide data on the deformations.
An other design is the parallel arrangement of several optical fiber sensors (Fig. 4). In this case a
light signal is passed one after the other through each optical fiber sensor with the aid of a
multiplexer. With velocity measurements the length of the respective optical fiber sensor is
measured and then compared with the sensor length of the parallel arrangement.
438 MONITORING OF TRIAL LOADINGS USING OPTICAL FIBER SENSORS

1. Analysis electronic
2. Bore hole
3. Glass fibre composite pipe
4. n optical fibre sensors
5. Reflective end of sensor
^ 6. Synthetic resin Airi7.
7. Basic length
as required
8. Semi-permeable
reflector

Figure 3: Figure 4:

3. TRIAL LOADINGS

3.1 Marienfelde Bridge, Berlin

The design of the Berlin-Marienfelde pedestrian bridge, a two span TT beam bridge, is an
example of the first fully monitored bridge structure. In addition to glass fiber prestressing
tendons (partial prestressing as external prestressing) with integrated sensors, a whole series of
optical fiber sensors have been embedded directly inside the concrete or been retro-applied.
During a trial loading carried out in November 1989 using 250 concrete slabs (weight per slab 1
t), this was able to be monitored with the aid of sensors. By using particularly highly sensitive
sensors, evidence was even able to be recorded on the dynamic bahaviour of this bridge.
#L R. WOLFF - H.-J. MIESSELER 439

optical fiber sensor for

monitoring of crack width
51.31
27.61 22.98

Vs

optical fiber sensor for

monitoring of crack width
longitudinal section
optical fiber sensor for
optical fiber sensor monitoring of cracks
for monitoring pl cracks and crack width
4.80

optical fiber sensor for

monitoring of crack width

7 glassfiber tendons
optical fit with optical fiber and
monitoring of cracks copper wire sensors
cross section

Figure 5: Measurement layout

SLAB 20 SLAB 21 SLAB 22
LAYER 4 LAYER 4 LAYER 4

\ 1

"
^
w
1
\jnr

0**
<

Time

REACTION OF THE CRACK - WIDTH SENSORS ON REMOVAL OF

THE CONCRETE SLABS FROM THE CENTRE OF THE LONG SPAN

Figure 6: Monitoring of the test load by means of the optical fiber sensors embedded in the
concrete

3.2 Schiessbergstrasse Bridge. Leverkusen

This three span, solid concrete slab bridge (bridge Classification 60/30), is designed with limited
prestressing comprising 27 glass fiber prestressing tendons (working load 600 kN) and post-
bonding. Three glass fiber bars per tendon are provided with sensors and there are to be four
additional optical fiber sensors integrated directly into the concrete on the upper and four on the
lower side of the slab. The trial loading of the Schiessbergstrasse Bridge which has been realized
on 3Ist of March 1992 showed in an impressive way the efficiency of the optical fiber sensors
embedded in the concrete construction. By means of the loading of the bridge with a truck of a
total load of 27,5 t it was possible to measure additional extensions of the bridge in the /x-range.
440 MONITORING OF TRIAL LOADINGS USING OPTICAL FIBER SENSORS

4 optical sensors for

monitoring of cracks

53.00

4 optical fiber sensors for each prestressing

monitoring of cracks tendon integrated with
copper wire sensors
longitudinal section
9.00 chemical sensors
4 optica! fiber sensors for (corrosion cell)
monitoring of cracks
each prestressing
5^ tendon integrated with
optical fiber sensors
optica! fiber sensors for chemicai sensors
monitoring of cracks (corrosion cell)
cross section

Figure 7: Sensor layout

Longitudinal section
mm

£P La

Computer control
Cross section-

A - A B - B

0.030
0,020
0,020
0.010
0,000
0.000
0.010
0.020

Figure 8: Trial loading

 R. WOLFF - H.-J. MIESSELER 441

3.3 Nötsch Bridge. Kärnten. Austria

The Nötsch Bridge is the first bridge in Austria with glass fiber composite prestressing tendons,
and is designed with limited prestressing comprising 41 glass fiber prestressing tendons (working
load 600 kN) and post-bonding. Similar to the Bridge Schiessbergstrasse the suitability of the
sensors has been proved by a trial loading. Two trucks with a total load of 2 x 22 t did produce
additional extensions in the middle field of approx. 50 /xm.
4 optical fiber sensors for
monitoring of cracks
44.00

4 optical fiber sensors for each prestressing

monitoring of cracks tendon integrated with
optical fiber sensors
longitudinal section
• 11.00
4 optical fiber sensors chemical sensors
for (corrosion cell)
monitoring of cracks
each prestressing
Sj?
tendon integrated with
4 optical fiber sensors for
optical fiber sensors
monitoring of cracks chemical sensors
(corrosion cell)
cross section
Figure 9: Sensor layout
extension in //m
70
60
50
sensor section 2
40
30
V
20
sjensor pier 1
/
10
0.0
-10
X5 measuring -
-20 time / see.
-30

pier 1 pier 2
Il | |' ' Ml I' lJM]M_..l

"H section 1 section 2 section 3

J
y i

Figure 10: Trial loading

r 13.00 —* 18.00
44.00
13.00
442 MONITORING OF TRIAL LOADINGS USING OPTICAL FIBER SENSORS

4. CONCLUSIONS

The current constantly increasing requirements and demands on heavy building structures make
the question of the utilised materials1 durability and the useful life of such structures in general to
be of ever greater and wider interest. Compared with essentially new construction, investment for
the preservation of building structures Claims a not inconsiderable proportion of the funds
available.
The possibility of integrating sensors, on the one hand directly into the bar materials during
production, and on the other the embedding of such sensors in the concrete, opens completely
new horizons for the permanent monitoring of building structures.

5. REFERENCES

1. Miesseler, H.-J.; Lessing, R.: Experiences in monitoring of load bearing structures with
optical fiber sensors. IABSE-Congress, Lisbon 1989.
2. Miesseler, H.-J.; Preis, L.: High performace glass fiber composite bars as reinforcement in
concrete and foundation structures. Bauen mit Kunststoffen, Issue 2/1988, Pages 4-14.

3. Miesseler, H.-J.; Wolff, R.: Experience with the monitoring of structures using optical fiber
sensors. FIP Xlth International Congress on Prestressed Concrete, Hamburg, 4-9 June 1990.

4. Wolff, R.; Miesseler, H.-J.: Application and experience with intelligent prestressing Systems
based on fiber composite materials. FIP Xlth International Congress on Prestressed Concrete,
Hamburg, 4-9 June 1990.
5. Wolff, R.; Miesseler, H.-J.: New Materials for Prestressing and Monitoring Heavy Structures.
Concrete International, 9.1989, Vol. 11, No. 9.
 443

Evaluation of the Capacity of an Existing Steel Truss Bridge

Evaluation de la capacitö d'un pont existant ä treillis mötallique
Ermittlung der Traglast einer bestehenden Stahlfachwerkbrücke

Bruno MASSICOTTE Constantin HALCHINI Jocelyn LABB£

Assist. Prof. in Civil Eng. Princ. Bridge Testing Eng. Bridge Engineer
£cole Polytechnique Quebec Min. of Transport Quebec Min. of Transport
Montreal, PQ, Canada Montreal, PQ, Canada Montreal, PQ, Canada

Bruno Massicotte was born in
1959. He received his Ph.D. at the
University of Alberta. He worked
for two years for the Quebec
Ministry of Transportation Bridge
Department. In 1990 he joined the
£cole Polytechnique where he
does research in bridge engineering.
Constantin Halchini graduated in
1953 in Bucarest as bridge engineer.
He worked for 26 years in
Romano for the Railway Institute
of Research as bridge testing
engineer. He joined recently the
Quebec Ministry of Transportation
where he is the leader of the bridge
testing Programme.
Jocelyn Labbö was born in 1961.
He received his B.Sc. from Laval
University in 1984. He worked for
several Consulting companies
before he joined the Quebec Ministry
of Transportation in 1989 as
bridge engineer.

SUMMARY
The paper presents the evaluation of the load capacity of a top deck steel truss bridge. Conventional
analyses, load testing and extensive finite element analyses were used to evaluate the actual capacity.
It is shown that conventional analyses are overly conservative compared to the actual bridge behaviour.
A load testing Programme, coupled with finite element analyses, led to important savings.

r£sum£
Cet article döcrit l'övaluation d'un pont a trellis mötallique avec dalle supeYieure. Des analyses
conventionelles, des essais de chargement et des analyses par eläments finis ont 6te* utilisös afin d'ävaluer
la capacite" röelle. On dömontre que les analyses conventionelles sont trop conservatrices. Les essais de
chargement couplös avec des analyses par elöments finis ont permis des 6conomies appräciables.

ZUSAMMENFASSUNG
Der Beitrag berichtet von der Ermittlung der Traglast einer Stahlfachwerkbrücke mit oben liegender
Fahrbahn. Dabei wurden herkömmliche Berechnungen, Belastungsversuche und umfangreiche Finite-
Elemente-Berechnungen eingesetzt. Wie sich zeigt, sind die herkömmlichen Berechnungsverfahren zu
konservativ im Vergleich zum tatsächlichen Verhalten der Brücken. Die Kombination von Probebelastungen
und Finite-Elemente-Berechnungen ermöglichten bedeutende Einsparungen.
444 EVALUATION OF THE CAPACITY OF AN EXISTING STEEL TRUSS BRIDGE

1. INTRODUCTION

Evaluation of existing bridges is a growing concern for bridge engineers.

Several options are possible to evaluate the carrying capacity of an existing
bridge: conventional or sophisticated analyses and bridge testing.
One can useconventional analyses coupled with design codes modified
specifically for bridge evaluation. Prequently this lead to unrealistic
evaluations of bridge carrying capacity. However conventional methods have
several advantages. They are simple to use and many bridges can be analyzed
rapidly. They are thus essential to classify bridges in categories to determine
the prior ity of action and the type of Intervention.
The degree of complexity of analyses can vary from simple static or empirical
distribution factors to complex finite element. In an other perspective, load
test can be performed. The tests can be set up to measure load effects at
service load level in various instrumented members, they can be proof tests in
which the load is added to the neighborhood of the ultimate capacity or at
onset of nonlinear behavior, or they can be performed up to failure.
The combination of a properly carried load test and refined analyses approach
an ideal Situation. This type of action was undertaken for the Massawippi River
bridge.
This paper presents the evaluation of the load carrying capacity of a deck slab
steel truss bridge carried using a conventional approach, a load testing
program and sophisticated analyses. The aim of this paper is to illustrate the
Steps involved in the strength evaluation procedure of an existing bridge. The
emphasis is directed toward rational strength evaluation involving, when
possible, refined analyses coupled with load testing.

2. BRIDGE DESCRIPTION

The Massawippi River bridge, located on Highway 108, south of Sherbrooke

Quebec, was built in 1937 and is owned by the Ministere des Transports du
Quebec (MTQ). This 183 m long bridge has eight simply supported spans: six with
four reinforced concrete beams, and two with two 32.3 m steel trusses at a skew
angle of about 53 degrees (Fig. 1) The top deck above the trusses is made of a
8.7 m wide and 220 mm thick reinforced concrete slab with a 125 mm asphalt
cover. The deck also includes stringers and floor beams as shown in Fig. 2.
Although the concrete slab was completely replaced some years ago,
mechanically connected to the steel beams so no composite action can
it was not

theoretically be developed. Despite an important number of years in service,

the bridge steel and concrete spans are stillin good condition and only a
limited number of elements require replacement or strengthening. Until
recently, the bridge sustained an intense daily traffic
of cars and trucks of
any configuration and weight since the bridge had no posted weight limits. In
an effort to identify all deficient bridges, the Bridge Department at the MTQ
evaluates all substandard bridges, starting with those potentially eritieal
located on highways. Considering its age and its location, the bridge
evaluation became recently a priority. A conventional strength evaluation was
then performed.
/4 B. MASSICOTTE - C. HALCHINI - J. LABB6 445

CURRENT

^\
1 2 3 4 5 6 7 71 8 9 10 11
N^ 1
i

" — - X I

TRUSSESs
—\^
\ - *¦
y' 1
f
\ \ N
FLOOR BEAM-
v— STRINGER

PLAN VIEW

TOP CHORD

H I

VERTICAL -J

ELEVATION VIEW
/ G
/

/\
\Z DIAGONALS
9

BOTTOM CHORD

Fig. 1 Geometry of the bridge

3064
600

_1800_ I
1800 1800 600

C D and E A and B

<j a ^

X JL A JL JL X I

DOWNSTREAM UPSTREAM

h
TRANSVERSAL VIEW

Fig. 2 Deck system

446 EVALUATION OF THE CAPACITY OF AN EXISTING STEEL TRUSS BRIDGE

3f. CONVENTIONAL STRENGTH EVALUATION

The MTQBridge Department evaluated the load carrying capacity of the

Massawippi River bridge according to the latest provisions of the Canadian
Bridge Code S6-M88 [1] The assumptions used for the analysis were the same as
in the original design. The dead and live load distribution between trusses
were obtained following a conventional approach since the bridge structural
system is satically determinate. The two trusses were assumed to behave
independently, ignoring the transverse bracing in the analysis. Moreover, since
no indication was available, no contribution of the deck, from either floor
beams, stringers or the concrete slab could be included for the evaluation of
the trusses. Also, the supports were assumed free to move horizontally at one
end of the trusses.

The live load shearing between both trusses was considered following a
conventional approach, with floor beams being simply supported on the top
chords of the two trusses. The distribution factor, for the calculation of the
most eritieal live load effects on a given truss according to clause 12 of S6-
M88[l],is equal to 1.16 times the loading model. The bridge calculated first
natural frequency of approximately 3.8 hertz produces a corresponding dynamic
load allowance factor of 40%. The live load rating factors obtained from the
evaluation are listed in Table 1 for the steel trusses and the concrete beams.
The steel spans exhibited the lowest LLRF giving posting limits of 25, 31 and
36 tonnes for two-axle, semi-trailer and train-trailer trucks respectively.

Table 1: Live load rating factors (LLRF)

Members 2 axles-truck Semi-trailler Train-trailler

Tension chord 0.87 0.61 0.54
Compression chord 0.89 0.64 0.54
Tension diagonal: 1.03 0.72 0.64
Compression diag. 0.88 0.85 0.85

Posting (tonnes) 25 31 36

Considering the amount of vehicles traveling on the bridge daily, its location
and its economical importance, it
was decided to increase the bridge carrying
capacity to legal load limits. However, due to the high cost of strengthening
the steel trusses, it
was decided to postpone the Intervention to give
engineers the time to explore other alternatives.

4. LOAD TESTING PROGRAM

In 1990, the MTQ acquired a mobile laboratory dedicated to bridge testing. This
mobile unit has two data acquisition Systems for static and dynamic load tests.
The Massawippi River bridge test was the first duty for the mobile laboratory
team and its new acquisition.
The Instrumentation of the bridge lasted 15 days during which 49 strain gages
were installed in a section of the upstream truss. Members instrumented within
the truss were: 1 bottom chord, 2 top chords, 1 vertical member and 2 diagonals
(Fig. 1) One member of a vertical bracing between the two trusses, 1 floor
Jl B. MASSICOTTE - C. HALCHINI - J. LABBE 447

beam and 2 stringers were also instrumented. Finally, 1 strain gage was
installed on the concrete slab soffit, parallel to a strain gage on a stringer,
to give some indication on the composite action between the concrete slab and
the steel beam.
The test itself took placed on September 26, 1990, and lasted 3 hours, during
which the bridge was closed intermittently. The loading vehicle was a 43 tonne
5 axle semi-trailer truck, loaded with gravel, carefully weighted and measured.
Five load paths along the bridge, named A to E (Fig. 2) and identified with
marking lines on the road, were used to measure load effects on truss members.
For all
members the axial force and bending moments were calculated from the
strain measurements. For most of the truss members, the axial load was
predominant. However even small bending moments measured justified the
utilisation of 4 strain gages per members. Without then the measurements could
have lead to unrealistic results with significant errors. The floor beams and
stringers exhibited a certain degree of composite action with the concrete
slab.
A comparison of load effects predicted by the conventional analysis and the
corresponding measured values, clearly indicated a significant discrepancy
between the behavior observed in the test and the one assumed in the analysis.
The diminution of live load effects measured in the test, compared to the
values obtained using the same assumptions as in the evaluation, were up to 67%
for the tension and compression chords whereas it
was 39% for the diagonals.
This indicates that the bridge capacity obtained initially could be modified
using different assumptions for the behavior.
Although field measurements represent the actual bridge behavior, test results
were available only for 6 of the truss 25 main members. This amount of
information is not sufficient to indicate clearly the reasons for such
discrepancy and safely allow for any significant increase in the bridge load
carrying capacity. It was therefore appropriate to proceed with sophisticated
analyses using the finite element approach to study more deeply the bridge
behavior.

5. FINITE ELEMENT ANALYSIS

Linear finite element analyses were carried out to model correctly the bridge.
A 3-dimensional model was created in which steel trusses, floor beams,
stringers, all bracing and the concrete slab were carefully discretized. To
model adequately the bridge behavior, two factors were used for calibration:
the longitudinal restraint of the truss supports and the participation of the
deck system, including floor beams, stringers and the concrete slab, in the
carrying mechanism.
The first parameter only accounts for the horizontal restraining action of free
supports at one end of the trusses. Although they should theoretically allow
free longitudinal movements of the bridge, the truss supports were rusted and
suspected to be frozen. Analyses involved only two cases for the movement: free
or fixed. To model the effects of the second parameter, two independent meshes
were used to discretize the bridge. A first mesh describes the two trusses and
the bracing system joining them (horizontal and vertical), and a second mesh
for the deck system: the concrete slab, the floor beams and the stringers (Fig.
1 Vertical connection members between the two meshes were used to simulate
various degrees of participation of the deck system in the load carrying
mechanism.
448 EVALUATION OF THE CAPACITY OF AN EXISTING STEEL TRUSS BRIDGE

Calibration of the two parameters was achieved with only one loading case. From
the load effects in the bottom chord, it became rapidly obvious that the mobile
supports were actually frozen horizontally. The second step in the calibration
process involved only the stiffness of the connection member which was then
modified in a trial and error process until satisfactory agreement between test
results and analysis were obtained for this load case. The model was then
considered adhequate to represent satisfactorily the bridge behavior and all
load cases of paths A, B and C were analyzed. Some of the results are presented
in Figs. 3 and 4 for the load path B, together with results of the conventional
analysis used initially for the bridge evaluation and test measurements.
A close examination of the results indicates a significant reduction in load
effects in top and bottom chords due to some composite action between the the
trusses and the deck system. This interaction, although partial, increases the
effective inertia of the two trusses which reduces the live load effects on
horizontal chords. The skewness of the bridge deck induces an important load
transfer between the two trusses through the vertical bracing. Frozen supports
have also some influence on the load carrying mechanism making the bridge
working like an arch.
These effects, important for the truss top and bottom chords, are however less
significant for web members. For the diagonals and the vertical members, a less
pronounced live load reduction is observed since vertical shear forces are
mainly carried by the trusses. The modification in the load carrying mechanism
can be observed by comparing reactions from dead load effects at truss
supports. In Table 2 reactions calculated using the various approaches are
presented. These results indicate that the dead load distribution is notably
affected by the skew angle and the transversal bracing. However, for most of
the truss members, the results indicate that the composite action has a more
important effect than the frozen supports.
Table 2: Dead load reactions at supports

Support Conventional 2D Conventional 3D Finite element

Free Fixed Free Fixed Free Fixed

Other-most 661.5 661.5 507.4 579.9 436.1 523.1

Inner-most 697.0 697.0 853.3 827.0 922.4 863.8

EVALUATION OF THE NEW LOAD CARRYING CAPACITY

The confidence gained in the comparison of the finite element model with test
results, allows one to proceed further and use the finite element model to
calculate the live load effects in the truss members for posting. However, the
evaluation of the actual carrying capacity using the finite element model must
consider assumptions on the conservative side which can differ from those used
for the comparison with test data. The contribution of the deck and the
skewness of the bridge have a clear effect on the bridge behavior and thus on
load effects supported by the bridge main members. These beneficial aspects of
the bridge behavior are assumed to remain the same at service load level.
However, the role played by frozen supports in the apparent gain in carrying
capacity must be reconsidered.
M, B. MASSICOTTE - C. HALCHINI - J. LABBE 449

700

Test
600
2D free
2D fixed
500
Finite element
400
P
300
<3

I 200

100

-100
B3 B4 B5 B6 B7 B8 B9 B10 B11

Vehicle position

Fig. 3 Axial forces in the bottom chord

-100

£, -200

0) Test

-o
(0
-300 2D free

I -400
2D fixed

Finite element

-500

-600
B3 B4 B5 B6 B7 B8 B9 B10 B11

Vehicle position

Fig. 4 Axial forces in the top chord

450 EVALUATION OF THE CAPACITY OF AN EXISTING STEEL TRUSS BRIDGE

This assumption, although based on careful measurements and sensitive analyses,

confront engineers with a dilemma: should the analysis to determine the actual
carrying capacity be done with frozen supports or not? On the other hand, is it
reasonable to repair the frozen supports to allow free longitudinal movements
since the bridge apparently behaves correctly in its current Situation? The
answer are not simple and this matter requires special attention to be assess
correctly.
For weight limit posting, both free and frozen support conditions at one end of
the bridge were considered. Since it is difficult to determine accurately the
bridge history, the dead load effects retained were those obtained from the
conventional analysis or calculated using the finite element model with free or
fixed supports, whichever produces the worst effects. For live load effects,
the values obtained with the finite element model with free supports at one end
were used.

An interesting fact observed in the analysis is that the most eritieal member
governing posting is the vertical strut a-A (Fig. 2) located at the inner-most
corner which carries higher compression forces due to load transfer between the
two trusses through the vertical bracing. The weakness of this member was not
identify in the conventional analysis. This mean that classical analysis
although usually on the safe side, may sometimes be unconservative.

7. STRENGTHENING

Although posting limits did not apparently increase very much after all these
efforts, the top and bottom chords, the weakest members previously, were not
eritieal any more and only web members remain eritieal. Thus only the
strengthening of a few members is now required, bringing down the costs by more
than $200 000. This economy almost justifies by itself the acquisition of the
mobile laboratory. The strengthening of the bridge was done in the summer of
1992.

8. CONCLUSION

The for strengthening a bridge and the cost involved justified the
need
utilization of a field testing campaign combined with refined analyses to
increase the understanding of the load carrying mechanism of the bridge. This
effort resulted in a slight increase in the actual safe carrying capacity of
the bridge and in a significant reduction of members requiring strengthening,
leading to important savings. The success of the project is an excellent
example of the benefit obtained when advantages of two complernentary approaches
are efficiently combined.

REFERENCES

1. CANADIAN STANDARD ASSOCIATION, 1988. "Design of highway bridges". CAN/CSA-

S6-M88. Rexdale, Ontario.

ACKNOWLEDGMENTS

The authors would like to thanks the Ministere des Transports du Quebec
Officials who believed in their dreams of performing bridge testing. Without
their strong supports such work could not have been possible.
 451

IABSE REPORTS RAPPORTS AIPC IVBH REPORTS

Volume 1 Structural Precast Concrete

Constructions präfabnquäes en b6ton
Vorfabrizierte Betonbauten

Symposium - New York 1968 On Weanng Surfaces for Steel Bridge Decks of
Lightweight Construction
Sur les revetements pour les platelages
m£talhques I6gers
Über Fahrbahnbelage von stählernen Leichtfahrbahnen

Volume 2 -
Final Report Rapport final - Schlussbericht
Symposium - London 1969 On Concepts of Safety of Structures and
Methods of Design
Sur les notions de s£cunt£ des structures et
m6thodes d'älaboration des projets
Über neue Aspekte der Tragwerksicherheit und
ihre Berücksichtigung in der Bemessung

Volume 3 Preliminary Pub f. - Pub f. preliminaire -

Vorbericht

Volume 4 Final Report - Rapport final - Schlussbericht

Symposium - Madrid 1970 Design of Concrete Structures for Creep,
Shrinkage and Temperature Changes
L'influence du fluage et du retrait, l'effet des
changements de temperature sur les
constructions en bäton
Der Einfluss des Kriechens, Schwindens und der
Temperaturanderungen in Stahlbeton-
konstruktionen

Volume 5 Preliminary Pub f.

Vorbericht
- Pub/, präliminaire -
Volume 6 Final Report - Rapport final - Schlussbericht
Volume 7 Report on an Enquiry Regarding Corrosion of the
Reinforcement in Reinforced and Prestressed
Concrete Structures
Rapport d'enquete sur la corrosion des armatu-
res des ouvrages en bäton arm§ et en bäton
precontraint
Bericht über eine Umfrage betreffend Korrosion
der Bewehrung in Stahl- und Spannbetonbauten
452

Symposium - Praha 1971 On Mass-produced Steel Structures

Fabncation en sehe des constructions
metalliques
In Serien gefertigte Stahlbauten

Volume 8 Introd. Report-Rapportintrod.

-Einführungsbericht

Volume 9 Prelim. Report - Rapport präfim. - Vorbericht

Volume 10 Final Report - Rapport final - Schlussbericht

Söminar - London 1971 Design of Plate and Box Girders for Ultimate
Strength
Etüde ä la resistance post-critique des poutres ä
äme pleine et des poutres-caissons
Bemessung von Blech- und Kastenträgern im
überkritischen Tragbereich

Volume 11 Proceedings - Rapport - Bericht

Symposium - Lisboa 1973 Resistance and Ultimate Deformability of
Structures Acted on by Well Defined Repeated
Loads
La resistance des structures et leur deformabili-
te ultime sous l'action de charges definies, räpe-
tees ou periodiquement variables.
Tragfähigkeit und Grenzverformung von
Bauwerken unter Einwirkung gegebener
wiederholter Belastungen

Volume 12 Introd. Report- Rapport introd.

Einführungsbericht
-
Volume 13 Prelim. Report - Rapport prä/im. - Vorbericht
Volume 14 Final Report - Rapport final - Schlussbericht
Symposium - Quebec 1974 Design and Safety of Reinforced Concrete
Compression Members
Dimensionnement et securite des pieces
comprimes en beton arme
Bemessung und Sicherheit von Stahlbeton-
Druckgliedern

Volume 15 Introd. Report- Rapport introd.

Einführungsbericht
-
Volume 16 Prelim. Report - Rapport prälim. - Vorbericht
 453

Volume 17 Final Report - Rapport final - Schlussbericht

Volume 18 Three Contributions on Safety of Structures
Trois contributions sur la secunte des structures
Drei Beitrage zur Sicherheit von Tragwerken

Seminar - Bergamo 1974 Concrete Structures subjected to Tnaxial

Stresses
Structures en beton soumises ä des contraintes
tnaxiales
Tnaxial beanspruchte Betontragwerke

Volume 19 Proceedings - Rapport - Bericht

out of print - epuise - vergriffen

Symposium - Dresden 1975 Steel and Composite Structures for User Needs
Structures en acier et mixtes congues pour
l'usage
Nutzungsgerechtes Bauen im Stahl- und
Stahlverbundbau

Volume 20 Introd. Report- Rapport introd.

Einführungsbericht
-
Volume 21 Prelim. Report- Rapport prelim. - Vorbericht
out of print - epuise - vergriffen

Volume 22 Final Report - Rapport final - Schlussbericht

out of print - epuise - vergriffen

Seminar - Paris 1972 Column Strength

Flambement des barres compnmees en acier
Knicken von Stahl-Druckgliedem

Volume 23 Proceedings
out of print - epuise - vergriffen
Volume 24 Pnnciples for Establishmg Tables of Optimum
Solutions for Bridges with Preflexed Steel
Beams
Pnncipes d'etablissement de tables de Solutions
optimalisees de ponts en poutres metalliques
prefiechies
Grundsatze für die Aufstellung der Tabellen der
optimalen Losungen für Brücken mit
vorgebogenen Stahlträgern
454

Symposium - München 1977 Problems Associated with Design and

Construction in Developing Countries
Conception, projet et execution de constructions
dans les regions en developpement
Planen und Bauen in Entwicklungsgebieten

Volume 25 Introd. Report- Rapport introd.

Einführungsbericht
-
Volume 26 Final Report - Rapport final - Schlussbericht

Symposium - Moscow 1978 Main Trends in the Development of Steel

Structures and Modern Methods of their
Fabrication
Tendances principales dans le developpement
des structures en acier et methodes modernes
de fabrication
Wichtigste Entwicklungstendenzen im Stahlbau
und moderne Herstellungsverfahren von
Stahlkonstruktionen

Volume 27 Introd. Report- Rapport introd. - Ein¬

führungsbericht

Colloquium - Copenhagen 1979 Plasticity in Reinforced Concrete

Volume 28 Introd. Report- Rapport introd. - Ein

out of print - epuise - vergriffen führungsbericht

Volume 29 Final Report - Rapport final - Schlussbericht

Colloquium - Bergamo, May 1978 Constructions in Seismic Zones

Construction en zones sismiques
Bauten in Erdbebengebieten

Volume 30 Proceedings
out of print - epuise - vergriffen

Colloquium - Bergamo, August 1978 Interface between Computing and Design in

Structural Engineering
Relations entre le calcul ä l'ordinateur et le projet
de constructions de g6nie civil
Die Beziehungen zwischen elektronischer
Berechnung und dem Entwurf von Tragwerken
im konstruktiven Ingenieurbau

Volume 31 Proceedings
Jk 455

Symposium - Zürich 1979 Bridges

Ponts
Brücken

Volume 32 -
Symposium Report Rapport du Symposium
-Symposiumsbericht

Colloquium - Delft 1981 Advanced Mechanics of Reinforced Concrete

Volume 33 Introd. Report- Rapport introd. -

out of print - epuise - vergriffen Einführungsbericht

Volume 34 Final Report - Rapport final - Schlussbericht

out of print - epuise - vergriffen

Volume 35 JCSS General Principles

out of print - epuise - vergriffen General Principles on Quality Assurance for
Structures
General Principles on Reliability for Structural
Design

Symposium - London 1981 The Selection of Structural Form

Systeme et forme des structures
System und Form von Tragwerken

Volume 36 Symposium Report-Rapport du Symposium

out of print - epuise - vergriffen -Symposiumsbericht

Colloquium - Lausanne 1982 Fatigue of Steel and Concrete Structures

Fatigue des structures en acier et en beton
Ermüdungsverhalten von Stahl- und
Betonbauten

Volume 37 Proceedings
out of print - epuise - vergriffen

Symposium - Washington, DC 1982 Maintenance, Repair and Rehabilitation of

Bridges
Entretien, reparation et modification des ponts
Unterhaltung, Instandsetzung und Sanierung
von Brücken

Volume 38 Introd. Report- Rapport introd. -

out of print - epuise - vergriffen Einführungsbericht

Volume 39 - -
Final Report Rapport final Schlussbericht
out of print - epuise - vergriffen
456 ßk.

Workshop - Bergamo 1982 Informatics in Structural Engineering

Volume 40 Proceedings

Colloquium - Copenhagen 1983 Ship Collision

Problems with Bridges and Offshore Structures

Volume 41 Introd. Report- Rapport introd.

Einführungsbericht
-
Volume 42 Prelim. Report- Rapportprä/im. - Vorbericht
Volume 43 -
Final Report Rapport final - Schlussbericht

Workshop - Tokyo 1982 Health and Safety in Construction

Volume 44 Proceedings

Symposium - Venezia 1983 Strengthening of Building Structures- Diagnosis

and Therapy
Renforcement des structures de bätiment-
Diagnostic et therapie
-
Verstärkung von Bauwerken Diagnose und
Behandlung

Volume 45 Introd. Report-Rapport introd. -

out of print - epuise - vergriffen Einführungsbericht

Volume 46 Final Report - Rapport final - Schlussbericht

out of print - epuise - vergriffen

Workshop-Rigi 1983 Quality Assurance within the Building Process

Volume 47 Proceedings

Symposium - Luxembourg 1985 Steel in Buildings

L'acier dans le bätiment
Stahl im Hochbau

Volume 48 Symposium Report-Rapport du Symposium

-Symposiumsbericht

Colloquium - Stockholm 1986 Thin-Walled Metal Structures in Buildings

Volume 49 Proceedings
 457

Symposium - Tokyo 1986 Safety and Quality Assurance of Civil Engineering

Structures
Secunte et assurance de la qualite des structures
de gerne civil
Sicherheit und Qualitätssicherung von
Ingenieurbauwerken

Volume 50 Introd. Report- Rapport introd.

Einführungsbericht
-
Volume 51 Prelim. Report- Rapport prelim. - Vorbericht
out of print - epuise - vergriffen
Volume 52 Final Report - Rapport final - Schlussbericht
Workshop - Zürich 1986 Organization of the Design Process
Organisation du Processus de projet
Organisation des Planungsprozesses

Volume 53 Proceedings

Colloquium - Delft 1987 Computational Mechanics of Concrete

-
Structures Advances and Applications
Mecanique des structures en beton arme
Progräs et applications dans les modeles
-
analytiques
Rechnerische Mechanik von Konstruktionen aus
-
Stahlbeton Fortschritte und Anwendungen

Volume 54 -
Report Rapport - Bericht
Symposium - Paris 1987 Concrete Structures for the Future
Structures en beton de l'avenir
Betonbauwerke für die Zukunft

Volume 55 -
Report Rapport - Bericht
Colloquium - Bergamo 1987 Monitoring of Large Structures and Assessment
of their Safety

Volume 56 -
Report Rapport - Bericht
Symposium - Lisbon 1989 Durability of Structures
Durabilite des constructions
Dauerhaftigkeit von Bauwerken

Volumes 57/1 and 57/2 -

Report Rapport - Bericht
Colloquium - Bergamo 1989 Expert Systems in Civil Engineering

Volume 58 -
Report Rapport - Bericht
458

Workshop - Lausanne 1990 Remaining Fatigue Life of Steel Structures

Volume 59 Report - Rapport - Bericht

Symposium - Brüssels 1990 Mixed Structures, including New Materials
Construction mixte acier, beton et nouveaux
materiaux
Misch- und Verbundbauweisen, einschliesslich
neuer Baustoffe

Volume 60 Report - Rapport - Bericht

Short Course - Brüssels 1990 Composite Steel-Concrete Construction and
Eurocode 4

Volume 61 Notes

Colloquium - Stuttgart 1991 Structural Concrete

Volume 62 Report

Colloquium - Nyborg, DK 1991 The Interaction between Major Engineering

Structures and the Marine Environment

Volume 63 Report

Symposium - Leningrad 1991 Bridges: Interaction between Construction

Technology and Design

Volume 64 Report

Conference - Davos 1992 Structural Eurocodes

Volume 65 Report