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.
http://www.e-periodica.ch
IABSE REPORT - RAPPORTS AIPC - IVBH REPORTS
Volume 67
IABSE COLLOQUIUM
COPENHAGEN 1993
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.
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.
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.
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.
Colloquium Committees
Scientific Committee
P. Clausen (Chairman)
E.Y. Andersen
P. Kristensen
L. Lovgren
J.E. Nielsen
J. Skov
0. Olsen
J. Vejlby Thomsen (Secretary)
Sponsored by:
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
G Labbe* J. 443
Lao Y. 223
Gao L. 233 Lin M.S. 149
VIII
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
Sakimoto T. 215
Savard M. 283
Scordelis A.C. 401
Shen D. 223
Shimamura M. 369
Sluszka M. 393
Smith J.W. 197
IX
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
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
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
Session 2:
Analytical Evaluation of Structures
Analyse de la fiabilite des structures
Analytische Beurteilung von Trag werken 101
A.R. KEMP
Evaluation of Reserve Capacity of Frames
Determination de la rEserve de capacitE portante des cadres
Bestimmung der Tragreserve von Rahmen 105
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
J.-M. HOHBERG
Planes of Weakness in Finite Element Analysis
Surfaces de rupture en analyses aux EIEments finis
Versagensflächen in Finite-Element-Berechnungen 157
Session 3:
Analytical Evaluation of bridges
Analyse de la fiabilitE des ponts
Analytische Beurteilung von Brücken 181
Selected Papers 19 5
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
Session 4:
Structural Evaluation by Testing
Contröle par des essais de Charge
Tragwerksbeurteilung durch Versuche 257
A.S. NOWAK
Reliability-Based Evaluation of Existing Bridges
ApprEciation de la fiabilitE des ponts existants
Zuverlässigkeitsuntersuchung bestehender Brücken 275
Session 5:
Case Histories
Etudes de cas
Fallstudien 349
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
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
K. NOORLANDER
Remaining Strength of Bridges in Rotterdam
CapacitE restante des ponts ä Rotterdam
Resttragfähigkeit von Brücken in Rotterdam 409
IABSE Reports
Rapports AIPC
IVBH Reports 451
SESSION 1
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.
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.
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.
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.
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.
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.
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.
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.
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
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 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.
- 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
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.
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:
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.
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:
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
> <
> 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
SELECTED PAPERS
18
19
Max Karlsson has worked with Johannes Morsing Johannesen got Ove Ditlevsen has contributed to
practical applications of structural his Master's Degree at the Technical the development of the theory of
reliability since 1980. He is a University of Denmark in structural reliability since 1961
member of the permanent 1990 and is currently PhD with a large number of publica-
Committee of the Danish Society of Student. He is studying methods of tions dealing with a large variety
Engineers on Actions on Structures reliability analysis of damaged of topics. Ove Ditlevsen is a member
and he is chairman of the structures. of the Joint Committee on
working group of the Danish Structural safety and of CIB, W81:
Society of Engineers on Structures in Actions on Structures.
Glass.
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.
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.
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
are given representing the "realistic" model and the idealized model respectively. It is assum-
ed that for each fixed (xq,Xp,Xß) the equations
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
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.
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
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
Yield line
1
b/4
b/4 <L
b/4
b/4
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.
<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.
RFy3 Uniform:[0.3,0.5] - -
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
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.
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.
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.
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
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)
\ //\
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
^ '
1I 1'
r n
Discussions
n Decisions on
--?
1
'
1
and owners &f jture m anagement
1
J
Risk : 0 significant
not slight 1 2 medium 3 high
Table 1
Significance of moisture State in influencing different durability
processes
G. SOMERVILLE 33
Region 1
Bond anchorage
e^
^ö Flexural compression
Shear
©
7^
Region 3
(cracking, flexural tension)
SU« ngw
K&*
co!2
Ä
bondstreng^ ,€>
6*2
o^
Time
*> Pm4
5 Pm2
~ Pml
Pm3
©
0
® ©
Time
5. CONCLUDING REMARKS
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
2. RILEM, Materials and Structures, Chapman & Hall, 1991, Vol. 24, No. 142,
320 pp.
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:
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 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.
(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).
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 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
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.
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
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
C.Inter bridge
combinations
11 Cross girders 1.128 8.251 E + 06
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
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
D (N/K) y Sjn 5
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
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,,
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
<
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
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
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
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
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 :
service / monitoring
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.
- 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
:
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.
In the following chapter, the discussion is focused on stage 2 relating to the evaluation of
failure probability.
The calculation of remaining fatigue life is improved by combining the advantages of both
damage aecumulation and crack propagation calculations.
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.
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)
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
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.
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]:
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 %
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
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.
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
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.
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
400 -
\\ N
\
V>
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
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)
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
*"(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
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.
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
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
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.
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.
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].
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.
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)
(MPa) fc MPA)
15-
0-
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:
Considering the local values of the rebound Schmidt hammer index, the parameters of the
concrete compression strength pdf are derived as follows.
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
fy =LN(347; 19710)
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
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
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:
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)
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
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:
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
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].
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]
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.
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
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
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
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
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.
STORAGE
SCOPE
n in ADAPTER
AXLE
DETECTORS
BALANCING
VOLTAGES DVM
SUPPLY
L. VZ&Z' BRIDGE
DEFLECTION
SENSING
DEVICES
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
13 500/2 13 600/2
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
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
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
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.
LOADED
1,000 4,700 12,500 3,900 5,667 8,820
BRIDGE
UNLOADED
BRIDGE - - - 5,560 5,759 8,200
SECOND -0,134 0,170 0,500 1,00 1,118 1,875 0,010 0,072 0,180
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
Impact Reception
serisor sensor Oscilloscope
Object
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]
<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
(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
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
mm m
&
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
<D
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)
Bridge name
^^^^ Occurrence Tolerance j
Judgment Occurrence Tolerance Judgment
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
SESSION 2
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
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 "~^
~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:
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.
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>:
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 '
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:
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].
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:
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
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
The method is described firstly in terms of the unknown end moments and then in terms of the
additional unknown, independent sway-deflections.
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) :
Vm-m+m+m+m
L-*t i ij ik i/ im - M' i
(10)
The only exception to this subdivision of equilibrium and compatibility equations occurs at a fixed
support where all n equations will be compatibility equations.
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.
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.
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:
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).
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.
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
^w
\— 4
r
SS
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.
After the state of the sheetpiling wall is well mapped, the evaluation will
finally Start.
A. VAN DER TOORN 123
¦r
/
codes othe*. msks
M o /V*ot0
/S .t |,>vrK-
<W
sea* opt ~^*
KtSc Ä/ic
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
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**.
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
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.
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)
(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
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)
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
It is readily noticed that equation (7) can be written in the following form
MrM{*}-Mr{F} (9)
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.
The applied loads are written in the following Standard load vector
kN
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
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.
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.
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)
K —1 ^^
(|)C(|-)2 + 4]
(2-4)
in which,t,h and b are the thickness,height and width of masonry wall respectively.
-£-
(1) P/Pu (0
/\u < 3)
< P/Pu < 0. 78,0 < — (2 — 5)
2. 6 0.
/\u
äk XINGGANG ZHOU 135
(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
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.
W ±mXL (3 - 1)
136 STRENGTH EVALUATION OF EXISTING MASONRY STRUCTURE #%
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)
It is noted that:
Jyxidt Jxidxi
xx hxX\/h\
therefore,
» »
J
* 7
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
* *
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
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
(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.
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
Within the limits of assumed model of a brittle body in the case of macroscopic
cracks the following equation is valid [2]:
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
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.
lim
s->o l r ay(*v)yJ «
{Vs
2& ß
'
OC7T
r
Vt (5)
k -
er
1.6 f vr£
s fJC
<7)
h+h^-^
r/Vj et
(8)
Modulus
K
ML t -
m AM NfN
isi ^ r-
Vh
JE
/Q\
V?)
*> t Wj
-
A1 ; ' 2
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.
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.
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)
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
REFERENCES
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
(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
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
k+p
G.K. CHAN - M.S. LIN - T. BALENDRA 153
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
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.
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
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)
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),
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:
fr
M
/ fr
|
plane stress
h es
45°- % 90 +Oc
r^ (ft.H-fcl
plane strain
(o,-t
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].
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]:
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+
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').
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
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.
1152 4mm
l3M2Jn(io/9
3048mm
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
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
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.
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
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.
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
[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
^ ^
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
Load [MN3
10. 00
6. 000
*
(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
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)
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:
"(!-) (i) ° °
° - (H °
(6)
0 0 0 0 1
<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
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
^=#-
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.
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
KEYNOTE SPEAKER
184
Finite Element Prediction of a Des
Prediction de la Charge de rupture d'un
Vorausberechnung einer Beto
Feldversuch mittels F
i.
m
,f.lri Mi
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.
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^ ^
800
G[MPa
600-
400
200-
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.
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.
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«
¦/*
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
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
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.
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
experiment
computation with
smaller load Steps
0.5
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%
f : ;
1. INTRODUCTION
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
Parapet girder
T Main girder T
1 >s\AAfewW
)p tp
Stiffness of hnkage element, k P/A
(b) Modelling of cross—girder stiffness
\
I I
V
P P
11 J ^zz
11
—-?
i!
N^
1
£ - "
[-1
y
A.\- t—
Symmetrie component Ps P
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:
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
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/
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
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)
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)
\
^^^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.
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.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
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)
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)
^ 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)
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
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)
3.2.1 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_,
« 20-
i°6
O
- -
04
Cu
3 02 ¦/v -
// i i i
100 200 300
TOME (day) TIME (day)
Cable Socket
Specimen l/X El E2 1A *2 il
(day) (GPa) (GPa) (year GPa) (day) (MN/m) (MN/m) (year MN/m)
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—
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
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.
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
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
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)
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
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.
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
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 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
Cross frame
Qy
/^
Truss element
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
N Acc N
6AwG
Vertical member: Av
jlEtanB $\CAd>
Diagonal member: Ad
3AwG
2^Esin9cos28 r
V—\
N
Av
Act
Av
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
\/ \/
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.
\ 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
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
5. CONCLUSIONS
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
Souxin Wu, born 1966, received Yanchang Lao, born 1920, received
his M.S. degree at Southwest his Ph.D. at Imperial College
Jiaotong Univ. He is now involved of Science and Technology in U.K.
in finite element analysis of He is now involved in assessment
concrete structures and evaluation of of structural capacity.
existing bridges.
Dayuan Shen, born 1932, graduated Yalin Pei, born 1958, received her
from Tangshan Jiaotong Univ. M.S. degree from Southwest Jiao-
He is now involved in research on ton Univ. She is now involved in
concrete bridges. 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.
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
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:
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)
where G represents dead load effect; S represent the effect of the live loads used for evaluation.
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.
Spalling &.
corrosion
Modified dimensions
iu FE model
Simulation of
load test
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
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
Tc KCT (5)
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];
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.
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.
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:
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]:
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
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.
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.
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)
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:
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
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.
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
-h'H
I 4.0 I
+4-
I
6.9 I
J(m)
1.2 1.2
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
>
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.
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
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.
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.
12
Table 1. Frequencies of a bridge pier (Hz)
Mode 3 4
This method for the first kind of damage is similar to the main
idea of Damage Tolerance Method of fracture mechanics.
According
materials, elastic
to damage theory,
modulus
if there are cracks or vacancy
of material will be reduced.
in
The
author [5] obtained:
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.
6. CLOSE REMARKS
REFERENCES
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.
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.
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
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
Thickness
of top slab 250 0 95 1 03 0 89 0 92 106 1 10 0 02 0 07 Normal
in Box Girder
The measurements of geometrical definition of girder cross seeuons have been classified in vertical and horizontal
dimensions
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
30 7 34 081 1 19 0 14 Normal
Lognormal
30 28 36 0 70 120 0 11 Normal
Lognormal
35 28 40 42 0 77 0 94 1 05 1 16 0 03 0 10 Normal
Lognormal
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
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
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 ^
Figure 1.- Histogram, PDF and CDF curves, for yield tensile force prestressing steel. (Strand 0.6")
Table 4.- Reinforcing steel parameters (fy yield tensile stress f max= maximum tensile stress)
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:
- 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.
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.
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
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
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
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
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
\\
1
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
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
A. Miyamoto, born 1949, received H. Monkawa, born 1959, received Y. Kajitani, born 1943, received
his Doctor of Engineering at Kyoto his Master of Engineering degree his Master of Enginenng degree
University in 1985. His recent from Kobe University in 1984. His from Kobe University in 1969. He
research activities are in the areas current research activities are in is attached to the Road Construction
of structural safety assessment on structural safety and reliability Division and is now working
concrete bridges and establish- analysis of concrete structures. He on technical management related
ment of design concept for is now developing a fuzzy expert to middle span bridges, and also
concrete structures under impact system for structural safety has great mterest in integrated
loads. assessment of concrete bridges. 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.
//\? \ Cv)
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
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.
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\
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.
3.3 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.
^-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
A B C D A B C D (HZ)
without damage
with damage
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—
_-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%
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 -
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
first
I
Damage assessment at level
SiSmall
I
•> Application of SI method L:Large
UK : Unknown
(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
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.
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
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.
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
reference axle
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
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
¦^ffW^^i.
111111
Load layer No.5
Girder No. 1
Load layer No.3
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
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
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.
r
203 mi
mm 2C03 mm
-i
25 mm
203 mm
JE610 mm
£= 246 mm
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
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.
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
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
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
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
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:
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,
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)
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
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%.
Constantin Halchini
for the mobile
at the Central Labo
at the Mini
Transportation.
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.
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.
12 3
l
4 8 9 10 11 12
^
I
I
r
I
**/ \*6
2 4 6 0 10 12 14 16 10 20 22 24 28 28
50
3-D Finite element 285 —O— 3-D Finite element
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.
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)
X
2.73 Hz ; Flexural 2.73 Hz ; Flexural
--«cc3 8 0
occ5 5 0
accZ -1 0
- - acd -3 5
«CC4 0
0....
"C^C^^ ^KXXXX^,
5.90 Hz ; Flexural 5.73 Hz ; Flexural
^
•cc3 13 -- -
^SZ
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
Director and co-founder of BO- Since her academic training at Keith Ward's expertise is in the
MEL, Colin Billington has been Imperial College, Helen Bolt has analysis of complex problems in
responsible at senior management been involved with development offshore engineering. He led the
levels for establishing and develo- of the reserve and residual development of the nonlinear
ping R&D organisations including strength technology for offshore Programme described in this
the Steel Construction Institute. jacket structures. She was the paper, which is now used to determine
His contributions to structural engineer on site for the experimental the reserve and residual
engineering technology and practice work described in this paper. strength of jacket structures. As
are widely recognised, with As Supervising Engineer she now Associate he leads the company's
more than 80 publications to his takes overall responsibility for analysis and Software development
credit. R&D activity. 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
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,
ly
HINGE UNIT
STRONGBACK
STRONGBACK
168 OD
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.
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.
168 OD
168 OD
LOAD
CELLS
168 OD 168 OD
168 OD
168 OD
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.
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
Jun Murakoshi, born 1963, received Takahiro Hirose, born 1964, received
his M Eng. at Tokyo Institute B.S. at Kobe University in
of Technology in 1987. He has 1988. He has been engaged in
been engaged in bridge engineering maintenance and management
works since 1987. 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.