Sie sind auf Seite 1von 205

Forschungsberichte

aus dem

Institut für Werkstofftechnik


Metallische Werkstoffe

der

Herausgeber: Prof. Dr.-Ing. B. Scholtes

Band 9

Enrique Garcia Sobolevski

Residual Stress Analysis of Components with Real Geometries


Using the Incremental Hole-Drilling Technique
and a Differential Evaluation Method
Forschungsberichte aus dem Institut für Werkstofftechnik - Metallische Werkstoffe
der Universität Kassel
Band 9

Herausgeber:
Prof. Dr.-Ing. B. Scholtes
Institut für Werkstofftechnik
Metallische Werkstoffe
Universität Kassel
Sophie-Henschel-Haus
Mönchebergstr. 3
34109 Kassel

Die vorliegende Arbeit wurde vom Fachbereich Maschinenbau der Universität Kassel
als Dissertation zur Erlangung des akademischen Grades eines Doktors der
Ingenieurwissenschaften (Dr.-Ing.) angenommen.

Erster Gutachter: Prof. Dr.-Ing. habil. B. Scholtes


Zweiter Gutachter: Prof. Dr. rer. nat. A. Wanner

Tag der mündlichen Prüfung 20. Juli 2007

Bibliografische Information der Deutschen Nationalbibliothek


Die Deutsche Nationalbibliothek verzeichnet diese Publikation in der Deutschen
Nationalbibliografie; detaillierte bibliografische Daten sind im Internet über
http://dnb.ddb.de abrufbar

Zugl.: Kassel, Univ., Diss. 2007


ISBN 978-3-89958-343-4
URN urn:nbn:de:0002-3438

© 2007, kassel university press GmbH, Kassel


www.upress.uni-kassel.de

Umschlaggestaltung: Melchior von Wallenberg, Nürnberg


Druck und Verarbeitung: Unidruckerei der Universität Kassel
Printed in Germany
Vorwort des Herausgebers

Bei einer zunehmenden Verbreitung elektronischer Medien kommt dem gedruckten


Fachbericht auch weiterhin eine große Bedeutung zu. In der vorliegenden Reihe werden
deshalb wichtige Forschungsarbeiten präsentiert, die am Institut für Werkstofftechnik –
Metallische Werkstoffe der Universität Kassel gewonnen wurden. Das Institut kommt
damit auch – neben der Publikationstätigkeit in Fachzeitschriften – seiner Verpflichtung
nach, über seine Forschungsaktivitäten Rechenschaft abzulegen und die Resultate der
interessierten Öffentlichkeit kenntlich und nutzbar zu machen.

Allen Institutionen, die durch Sach- und Personalmittel die durchgeführten Forschungs-
arbeiten unterstützen, sei an dieser Stelle verbindlich gedankt.

Kassel, im Oktober 2007

Prof. Dr.-Ing. habil. B. Scholtes


„Везде исследуйте всечасно,
Что есть велико и прекрасно,
Чего еще не видел свет.“

Михаил Васильевич Ломоносов

„Überall erforschet ohne Unterlaß


Was herrlich ist und wunderschön,
was die Welt noch nicht geseh’n.“

Michail Vasil’evič Lomonosov

Посвящается моей маме,


Александре Всеволодовне Соболевской-Глёкнер
5

Contents

Symbols and Abbreviations.................................................................. 9


1 Introduction........................................................................................... 13
2 State of Knowledge................................................................................ 14
2.1 Residual Stress........................................................................................ 14
2.1.1 Definition and Origin of Residual Stress................................................ 14
2.1.2 Origin...................................................................................................... 15
2.1.3 Technological Relevance........................................................................ 16
2.1.4 Overview of Residual Stress Measurement Methods............................. 17
2.2 General Aspects of the Hole-Drilling Method........................................ 18
2.2.1 Principle.................................................................................................. 18
2.2.2 Development........................................................................................... 19
2.2.3 Scope, Influence and Restrictions........................................................... 20
2.3 Residual Stress Calculation..................................................................... 21
2.3.1 Average Stress Value over Specimen’s Thickness - Through-Hole
Case 21
2.3.2 Average Value over Hole Depth - Blind-Hole Case............................... 24
2.3.3 Residual Stress Depth Distribution......................................................... 25
2.3.3.1 Differential MPA Method....................................................................... 25
2.3.2.2 Integral Method....................................................................................... 28
2.4 Geometrical Boundary Conditions.......................................................... 30
2.4.1 Prefacing Remarks.................................................................................. 30
2.4.2 Definition of Possible Geometrical Boundary Conditions...................... 31
2.4.3 Overview of Previous Investigations...................................................... 32
2.4.3.1 Thickness................................................................................................. 32
2.4.3.2 Distance between Hole Center and Component Edge............................ 35
2.4.3.3 Radius of Surface Curvature................................................................... 37
2.5 Concluding Remarks on state of Knowledge.......................................... 39
3 Experimental and Numerical Methods of Investigation.................... 41
3.1 Prefacing Remarks.................................................................................. 41
3.2 Experimental Details............................................................................... 42
3.2.1 Material................................................................................................... 42
6

3.2.2 Specimens................................................................................................ 43
3.2.3 Hole-Drilling Device and Calibration Equipment.................................. 48
3.2.4 Experimental Calibration Procedure....................................................... 49
3.3 Numerical Methods and Calibration....................................................... 51
3.3.1 Models and Simulation Procedure.......................................................... 51
3.3.2 Material Model, Load Application and Boundary Conditions of the
FE-Models............................................................................................... 57
3.3.3 Calculation of Strain................................................................................ 58
4 Results of Experimental and Numerical Calibration......................... 61
4.1 Prefacing Remarks.................................................................................. 61
4.2 Reference Specimen and Variation of Hole Diameter -d0..................... 62
4.3 Variation of Specimen Thickness – T..................................................... 64
4.4 Variation of Distance Hole-Edge – D..................................................... 68
4.5 Combination of Parameter T and D........................................................ 71
4.6 Variation of Cylinder Radius – R............................................................ 74
4.6.1 Solid Cylinders........................................................................................ 74
4.6.2 Hollow Cylinders.................................................................................... 77
4.6.3 Comparison of Solid Cylinders and Hollow Cylinders........................... 80
4.7 Specimens with Center Through-Hole -Combined Geometrical
Parameters............................................................................................... 82
4.7.1 Flat Tensile Specimen with a ø10 mm Center Through-Hole................ 82
4.7.2 Hollow Cylindrical Tensile Specimen with a ø6 mm Through-Hole..... 85
5 Discussion of Calibration Results........................................................ 89
5.1 Prefacing Remarks.................................................................................. 89
5.2 Calibration Uncertainty........................................................................... 90
5.2.1 Uncertainty in Numerical Calibration..................................................... 90
5.2.2 Uncertainty in Experimental Calibration................................................ 94
5.2.3 Graphical Representation of Uncertainty in Calibration Results............ 101
5.3 Stress Evaluation on specimens with Non-Reference Geometry............ 105
5.4 Influence of Geometrical Parameters on Results of Hole-Drilling
Measurements.......................................................................................... 110
5.4.1 Influence of Hole Diameter - d0............................................................. 110
5.4.2 Influence of Specimen Thickness - T...................................................... 112
7

5.4.3 Influence of Distance Hole-Edge - D...................................................... 118


5.4.4 Influence of Cylinder Radius - R............................................................ 121
5.5 Possible Reduction of Calculated Stress Differences using Geometry
Specific Calibration Functions................................................................ 129
5.6 Summary of Discussion of Calibration Results...................................... 134
6 Implementation of Geometry Specific Calibration Functions into
Evaluation Program.............................................................................. 136
6.1 Prefacing Remarks.................................................................................. 136
6.2 Modification and Differentiation of Measured Strain Distributions....... 138
6.2.1 Block 1 - Import of Input File................................................................. 138
6.2.2 Block 2 - Parameter Modification........................................................... 139
6.2.3 Block 3 - Strain Modification.................................................................. 139
6.2.4 Block 4 - Strain Smoothing..................................................................... 139
6.2.5 Block 5 - Storage of Modified Input File................................................ 142
6.2.6 Block 6 - Local Polynomial Approximation of Measured Strains.......... 142
6.3 Stress Calculation.................................................................................... 143
6.3.1 General Aspects of Stress Calculation in Prototype Program................. 143
6.3.2 Block 7 - Stress Calculation “MPA Standard”........................................ 143
6.3.3 Block 8 - Stress Calculation “MPA Expanded”...................................... 144
6.3.4 Block 9 - Stress Calculation “MPA Specific”......................................... 146
6.3.5 Block 10 - Import of Calibration Files.................................................... 147
6.3.6 Block 11 - Polynomial Approximation of Calibrations Strains.............. 148
6.3.7 Block 12 - Angle Calculation.................................................................. 148
6.3.8 Block 13 - Visualization of Results......................................................... 149
6.3.9 Block 14 - Export and Storage of Result File......................................... 149
7 Evaluation of the Implementation of Geometry Specific
Calibration Functions into Stress Calculation Program.................. 151
7.1 Prefacing Remarks................................................................................. 151
7.2 General Aspects of Calculation.............................................................. 151
7.3 Influence of Local Polynomial Approximation of Measured Strains... 154
7.4 Influence of Strain Smoothing............................................................... 158
7.5 Influence of Order of Polynomial Approximation of Calibration
Strains...................................................................................................... 162
8

7.6 Influence of Selected Distribution of Depth Increments......................... 166


7.7 Interpolation between Geometry Parameters......................................... 169
7.8 Summary of the Evaluation of the Calculation Program with
Implemented Geometry Specific Calibration Functions........................ 172
8 Conclusion............................................................................................. 173
9 References............................................................................................. 177
10 Zusammenfassung in deutscher Sprache........................................... 183
11 Appendix................................................................................................ 187
11.1 FEM Post Processor Plots....................................................................... 187
11.2 Program Listings..................................................................................... 194
11.2.1 Smoothing using Weighted Average Algorithm..................................... 194
11.2.2 Smoothing using Compensative Cubic Splines...................................... 194
11.2.3 Local Polynomial Approximation of Measured Strains.......................... 196
11.2.4 Calculation of Geometry Calibrations Coefficients via Linear
Interpolation between Single Geometry Specific Coefficients............... 197
Acknowledgments……………………………………...……………... 201
9

Symbols and Abbreviations

Latin Symbols

A(H,h) strain relaxation per unit depth


a1,...,n single values of uncertainty
aij matrix of strain relaxation
ASG strain gage area
ATH calibration constant for finite strain gage areas
BTH calibration constant for finite strain gage areas
D distance between hole-center and free edge or border of component
d0 hole diameter
DE distance between hole-center and edge of component
Den denominator
DH distance between two holes
dm mean rosette diameter
E Young’s modulus
fil compensating factor
H normalized depth from surface
h normalized hole depth (integral method)
Ktn concentration factor according to nominal stress
Kx calibration function in x-direction
Ky calibration function in y-direction
M strain matrix
n number of measurements
nsc number of scale divison
Num numerator
p(h) equi-biaxial plane strain
P(H) equi-biaxial plane stress
R cylinder outside radius
r0 hole radius
r1,2 radial distance from hole center to gage ends
10

RB outside radius of sphere or sphere


RBI inside radius of sphere or sphere
ReS yield strength
RI cylinder inside radius
s standard deviation
S1,2,3 integration constants
s.g. strain gage
T thickness
t distribution according to Student
u uncertainty
us systematic uncertainty
uz random uncertainty
w strain gage width
W component width
z hole depth
Z depth from surface

Greek Symbols

Δεi relived strain


ε0(i) initial strain
ε1,2,3 strain in strain gage direction
εΕ(i) final strain
εx nodal strain value in x-direction
εx,m calculated average strain value in x-direction
φ orientation angle
φ∗ intermediate angle
ν Poisson’s ratio
σ1,2,3 stress in strain gage direction
σc,x calibration stress in x-direction
σc,y calibration stress in y-direction
σI,II,III kinds of residual stress
11

σmax maximal calculated stress


σmax,min principal stresses
σmax,sim maximal stress (FEM)
σnom nominal stress
σRS residual stress
σx-FEM initial stress distribution at measurement point of FEM model
ξ normalized depth (differential method)

Abbreviations of Specimens and Models

The denominations of the used specimens and models are a combination of


abbreviations, which describe the main geometrical specifications:

D Distance between center of measurement point (hole) and specimen’s edge


f flat specimen
fh flat specimen with ø10 mm center through-hole
hc hollow cylindrical specimen
hch hollow cylindrical specimen with ø6 mm through-hole
m FEM model
R cylinder outside radius
ref geometrical reference
sc solid cylindrical specimen
T specimen’s thickness

Examples:

fT1.5D20 flat specimen with a thickness of T = 1.5 mm and a distance hole-edge of


D = 20 mm
hcR3m FEM model* of a hollow cylindrical specimen with an outside radius of
R = 3 mm
hchR6 hollow cylindrical specimen with ø6 mm through-hole and an outside
cylinder radius of R = 6 mm

* The FEM models are additionally indicated using the abbreviation “m” at the end of
the denomination.
13

1 Introduction
As a result of the manufacturing processes, residual stresses are found in nearly all
components. In some cases, the behavior of the component can be affected in a
beneficial or detrimental manner depending on the magnitude and direction of the
residual stresses. Thus, the demand for reliable and economically justifiable analysis
methods is of great significance in industry. One accepted and widely used technique
for measuring residual stresses is the incremental hole-drilling method. The basic hole-
drilling procedure involves drilling a small hole into the surface of a component at the
centre of a special strain gage rosette and measuring the relieved strains. The residual
stress distributions originally present at the hole location are then calculated from the
stepwise established strain distributions against the hole depth. The introduced blind-
hole relaxes the residual stresses only partially and, therefore, the available evaluation
procedures include, as a reference basis, a calibration of the method on a known stress
state under reference geometrical conditions. Because of this, the violation of
recommended geometrical restrictions could have an influence on the calculated
residual stresses, e.g. an overestimation of stress values. In this case, some authors
recommend the determination of geometry specific calibration functions.

This work investigates the geometry influence on the residual stress calculation
according to the differential MPA evaluation algorithm [1, 2] for hole-drilling
measurements. The first aim is to show the individual influence of the specimen
geometry by means of strain and stress deviations related to the results determined with
a reference specimen. The second aim is the implementation of geometry specific
calibration functions into the differential MPA evaluation algorithm in order to show a
possible correction of the calculated stress deviations. For both purposes, experimental
and finite element calibrations are carried out using specimens or models which
systematically violate the geometry parameters thickness, distance from the hole to the
specimen edge and the surface curvature. An uncertainty analysis discusses the quality
of the presented calibrations results. In the last part of the work, a prototype program for
the stress evaluation according to the MPA algorithm is described and evaluated. In
contrast to current software packages, the presented program calculates, additionally,
geometry specific calibration functions for the consideration of the actual component
shape.
14

2 State of Knowledge
2.1 Residual Stress
2.1.1 Definition and Origin of Residual Stress
Residual stresses (or locked-in stresses) are defined as those stresses which exist in a
solid body without the application of external forces or any other sources of load, such
as thermal gradients, gravity, etc. Residual stresses can be found in almost all rigid parts
independent of the material. All residual stress systems in a component are self-
equilibrating, i.e. the resultant forces and the moments that are produced inside the
component by residual stresses must be zero. [3, 4].

There are three kinds of residual stresses that are classified according to the range over
which they can be observed [4, 5, 6]:

• The first kind of residual stress σI is also called macroscopic stress. It is long-
range in nature and extends over at least several grains of the material, and usually
over many more.
• The second kind of residual stress σII is also called structural micro-stress. It
covers the distance of one grain or a part of a grain. It can be found between
different phases and has different physical characteristics. It can also be found
between embedded particles, such as inclusions and the matrix.
• The third kind of residual stress σIII ranges over several atomic distances within
the grain. It is equilibrated over a small part of the grain. This third type of residual
stress is inhomogeneous across the described areas of the material.

The relationship between the three different types of residual stresses is shown
schematically in Fig. 2.1. The total residual stress σRS (magnitude and direction) at a
particular point of a material is the result of the superposition of the locally existent
kinds of stresses.
15

Fig. 2.1: Definitions of residual stresses of I., II., III. kinds [7]

2.1.2 Origin
Main Groups Manufacturing Processes
Elastic-plastic loading bending, torsion, tension, compression
Machining drilling, grinding, milling, planning, turning
Joining adhering, brazing, soldering, welding
Founding -
shot peening, deep rolling ,drawing, forging,
Forming
pressing, spinning
hardening, case hardening, nitriding,
Heat-treating
quenching
cladding, electroplating, galvanizing, plating,
Coating
spraying, depositing
Tab. 2.1: Origin of residual stresses after technological processes [4]

The origin of residual stresses is the consequence of inhomogeneous elastic or plastic


deformation caused by the previous technological treatment of structural parts or
components. This treatment can be either mechanical, thermal or chemical, or a
combination of these treatments. The generated residual stress depends on the geometric
16

conditions of the treated technical part and on the parameters of the treatments and
applied processes. Tab. 2.1 lists possible origins of residual stresses in main groups and
some exemplary treatments or manufacturing processes. The residual stress distribution
caused by manufacturing is usually non-uniform over the component cross-section and
in some cases high stress gradients are characteristic.

2.1.3 Technological Relevance


Residual Stress induced by manufacturing processes can influence the performance of a
component in a positive or negative way. Harmful effects are usually caused by tensile
residual stresses on the surface of a part since they are often the major cause of fatigue
failures, quench cracking and stress corrosion cracking. On the other hand, compressive
residual stresses on the surface induced e.g. by shot peening are usually desirable since
they increase both fatigue strength and resistance to stress corrosion cracking. In general
terms, residual stresses are beneficial when they are parallel to the direction of the
applied external load and of opposite sense, e.g. a compressive residual stress and a
tensile applied load. When considering the possible effects of residual stresses the need
of reliable and cost-effective analysis methods is of great practical interest. Three main
applications for residual stress measurements are mentioned in [8] exemplary for the
automotive industry:

1. Correcting design flaws: Using residual stress analysis, technical failure caused
by improper design can be corrected before it appears during later processing,
storage or in-service of the component.
2. Improving material properties: In many cases compressive residual stress is
intentionally induced to improve the lifetime of parts, e.g. by mechanical surface
treatment. At this stage, residual stress analysis can support the quality control of
the treatment process used.
3. Predicting behavior of working parts: Residual stresses are often considered
for simulation purposes in order to predict fatigue limits and deformation
following different manufacturing operations. The used numerical models
should be validated with proper residual stress measurements.

A recent example in which residual stress analysis plays a central role in a practical
context is shown in a collaborative research project [9] between partners from industry
17

and a research institute. One main objective of the project was the minimization of
distortions of thin walled light metal die casting components that could be generated,
among other things, by manufacturing relevant residual stresses. Within the project the
whole manufacturing process was analyzed and the die casting was simulated with the
objective of optimizing design and individual steps of manufacturing. For this purpose
residual stress measurements in selected structural components were carried out in order
to evaluate processes such as heat treatment and to validate the die casting simulations.

2.1.4 Overview of Residual Stress Measurement Methods


A variety of residual stress measurement techniques exist at the present time for
different analysis objectives, materials, kinds of residual stresses, part geometries, the
cost of the measurement and the prices of the equipment required. The main
characteristic of the methods is the physical principal of the measurement i.e. the
directly measured parameters. For this reason it is common in practice to use a
classification of four main groups [3, 6]:

1. Mechanical Methods: These methods are based on the measurement of strains


caused after the removing of stressed material. With these relieved strains the
stress state before material removing can be calculated using the elastic theory.
The hole-drilling method, the ring core technique, the bending deflection method
and the sectioning method are four of the most extensively used mechanical
methods in industry.
2. Diffraction Methods: The x-ray diffraction method and the neutron diffraction
method rely on elastic lattice strains by detecting the changes in the interplanar
spacing of polycrystalline material based on knowledge of the incident
wavelength λ and the change in the Bragg scattering angle Δθ. The stress state is
calculated using the lattice strains and the elastic theory.
3. Magnetic Methods: The magnetic methods measure either the Barkhausen
noise (analysis of magnetic domain wall motion) or the magnetostriction
(measurement of permeability and magnetic induction) of ferromagnetic
materials. The stress state depends on the interaction between magnetization and
the strain state of these materials.
18

4. Ultrasonic Methods: These techniques are based on the variations in the


velocity of ultrasonic waves, which can be related to the residual stress state of
the material they fly through.
5. Calculation (Modeling) of Residual Stress: Although this is not a
measurement method, the numerical calculation, e.g. using finite elements, is of
particular importance for the analysis of the origin, influence and relaxation of
residual stress on components. The modeling of residual stresses is carried out,
among other things, with the aim of saving costs and time-consuming
measurements.

At all events, according to an industrial survey [10] the hole-drilling method and the x-
ray diffraction method are the most widely used techniques for measuring residual
stresses in practice.

2.2 General Aspects of the Hole-Drilling Method


2.2.1 Principle
The hole-drilling method is a mechanical method for measuring residual stresses and it
is standardized in the ASTM E 837 [11]. Other publications with reference character are
the technical note TN 503 [12] and the NPL guide No. 53 [13]. The principle of the
hole-drilling method is visualized in Fig. 2.2. The basic hole-drilling procedure involves
(1) the drilling of a small hole of diameter d0 into the surface of a stressed material. The
hole can be drilled with mechanical drills and milling cutters (low-speed and high-
speed), with an air-abrasive system or with an electrical discharge machining system.
The stress equilibrium is locally disturbed due to this intervention whereby a new
equilibrium is reached. This change is measured (2) usually radial to the hole with
special strain gage rosettes (dm mean rosette diameter) in the form of relieved strains.
Another possibility is the measurement of the full strain field using optical methods, e.g.
laser interferometry, holography, moiré interferometry and strain mapping. The residual
stresses originally present at the hole location (3) are then calculated from these strain
values using the elastic theory either as an average over the drilling depth from the total
strains upon reaching the final depth, or as a depth distribution from the strain
distribution over the depth as established by incremental drilling.
19

Fig. 2.2: Principle of the hole-drilling method

2.2.2 Development
Many scientists have worked, and continue to research, on the different aspects of the
hole drilling method as e.g. the drilling process itself, the strain measurement and the
evaluation methods. Some important contributions for the hole-drilling method using
electrical strain gage rosettes and conventional drills or mills are mentioned here.

The hole-drilling method was first proposed by Mathar in 1933 [14, 15]. He measured
diametral changes of a drilled through hole over the cross-section of a plate with a
mechanical extensometer and calculated the average stresses using Kirsch’s theoretical
solution [16] for the stress state in an infinite thin plate. In 1950, Soete and
Vancrombrugge used electrical-resistance strain gages instead of the former mirror
extensometers [17] and thus improved the accuracy of the measurements. In 1956,
Kelsey developed an evaluation method to analyze residual stress depth distributions by
incrementally drilling a blind hole [18]. Kelsey’s method was the basis for later
differential algorithms. The hole-drilling method became a systematic and easily
reproducible procedure for measuring residual stresses after the improvements made by
Rendler and Vigness in 1966 [19], who defined among other things the geometry of the
ASTM E837 standard hole drilling rosette. The practicability of high-speed drilling to
provide a stress-free drilling method was shown in the investigations of Flaman in 1980
[20]. Using finite elements calculations, Schajer developed, among other things, two
practical applications for the evaluation of in-depth stress distributions based on an
20

integral approach [21]: the Powerseries Method in 1981 [22] and the Integral Method in
1988 [23, 24]. In 1993 Kockelmann and Schwarz proposed a differential method that
allows the determination of inhomogeneous residual stress depth distributions with
reduced calibration complexity [1, 25].

2.2.3 Scope, Influences and Restrictions


The hole-drilling method is, in comparison to other residual stress measuring
techniques, a common, cheap, fast and popular method which covers the following
scope:

• Type of Measured Stresses: The method determines macro residual stresses.


Most of the in-depth evaluation algorithms provide a solution to determine an
elastic plane stress state. However, to avoid local yielding because of the stress
concentration due to the hole, the maximal magnitude of measured residual
stresses should not exceed 60-70 % of local yield stress ReS [1].
• Material Condition: The method is applicable in general to all groups of
materials. Firstly, the materials should be isotropic and the elastic parameters
should be known. Secondly, the analyzed materials should be machinable, i.e.
the boring of the hole should not prejudice the measured strains.
• Resolution: The local resolution of the method is dependent on the equipment
used. Laterally, the resolution ranges in the area of the produced hole diameter.
The minimal infeed or boring increment is in the region of some µm, whereas
the maximal analyzable depth of the hole does not exceed 0.5 × d 0 .

On the whole, the hole-drilling method is simple; disturbing influences and sources of
error have to be taken into account by eliminating or compensating them. Several
publications deal with problems that may arise when using the hole-drilling technique,
e.g. in [26, 27, 28].
The possible disturbing influences and sources of error are mentioned, investigated and
classified into five main groups in [29]. Tab. 2.2 lists the sources of these errors. The
authors consider the influences of the strain gage technique, the influences of the
drilling technique as well as most of the influences of the stress condition as not being
of importance, if these techniques are correctly applied. Possible adverse effects should
21

be considered for all influences related to the geometrical boundary conditions, the
plasticization near the hole and the calculation errors of stress gradients into depth.
• Change in the residual stress condition when the strain gage
Strain Gage
bonding point is pretreated
Technique
• Temperature response due to heating when drilling
• Production of residual stresses due to drilling
• Deviations from the ideal blind-hole shape
Drilling Technique
• Eccentricity of hole
• Incorrect depth setting and depth measurement
• Influence of component edgings
Boundary Conditions • Influence of neighboring holes
• Influence of component geometry
• Influence of multi-axiality
• Influence of stress gradients
Stress Condition
• Influence of orientation of strain gages
• Plasticization as a result of the notch effect of the hole
Determination of • Scatter of calculation coefficients and calibration curves

Residual Stresses • Error in system of calculation of stress gradients into depth

Tab. 2.2: Disturbing influences and sources of error of the hole-drilling method [29]

In this work, the influences of the component geometry are investigated in detail.
Therefore, the problems regarding the geometrical boundary conditions as well as an
overview of previous investigations are, respectively, mentioned and discussed in
Chapter 2.4.

2.3 Residual Stress Calculation


2.3.1 Average Stress Value over Specimen’s Thickness - Through-Hole Case
The through-hole case is the basic calculation procedure for the hole-drilling method.
This solution is based on the solution of the stress state in an infinite thin plate [16]. An
equation {2.1} calculates average values over the thickness of a thin plate (using
Young’s Modulus E) of the principal stresses σ max and σ min for strain gage rosettes
with finite gage length and gage width.

E (Δε 3 + Δε 1 ) E 2
{2.1} σ max,min = ± (Δε 3 − Δε 2 ) 2 + (Δε 2 − Δε 1° ) 2
4 ATH 4 B TH
22

In Fig. 2.3 the numbering of the single strain gages, as well as the direction and angle
orientation of the principal stresses σ max and σ min , are defined for a strain gage rosette.

Fig. 2.3: Hole- drilling strain gage rosette CEA-XX-062UM-120 with angle definition and direction of
principal stresses

The relieved strains Δε i {2.2} are the difference between the strain readings after the

last drilling step to create the through-hole ε E (i ) and the initial strain readings ε 0( i )

before drilling the hole.

{2.2} Δε i = ε E (i ) − ε 0(i )

Hole-drilling calibration constants for finite area strains ATH {2.3} and B TH {2.4} are
used in equation {2.1} according to the procedure proposed by [30]. An equivalent
procedure using trigonometric relationships is described in [31]:

r0 (1 + ν )S1
2

{2.3} ATH = −
2 w(r1 − r2 )

2
− r0 ⎡ 3 2 ⎤
{2.4} B TH
= ⎢ (1 + ν )( S 2 − r0 S 3 ) + (1 − ν ) S1 ⎥
w(r1 − r2 ) ⎣ 2 ⎦
23

In these equations the geometry of the hole is considered in the hole radius r0 (Fig. 2.4).

The geometry of the strain gages is included in the strain gage width w and the radial
distances from hole center to strain gage ends r1 , r2 respectively (Fig. 2.4).

Fig. 2.4: Schematic definition of geometric parameters for hole and strain gage

The integrations constants S1 , S 2 , S 3 are listed in equations {2.5}, {2.6} and {2.7}:

x=r
⎡ x ⎤ 2
{2.5} S1 = 2 ⎢arctan
⎣ 0.5w ⎥⎦ x = r1

x=r
⎡ ⎤
2
x
{2.6} S 2 = − w⎢ 2
⎣ x + 0.25w ⎥⎦ x = r1
2

x = r2
1 ⎡ x ⎤
{2.7} S 3 = − w⎢ ⎥
(
3 ⎣⎢ x 2 + 0.25w 2 2 ⎦⎥ )
x = r1

The orientation of the principal stresses σ max and σ min can be defined using the angle

ϕ (angle between strain gage 3 and σ max , Fig. 2.3). For this, an intermediate angle ϕ ∗
has to be calculated {2.8}.
24

1 2Δε 2 − Δε 3 − Δε 1 1 Num
{2.8} ϕ ∗ = arctan = arctan
2 Δε 1 − Δε 3 2 Den

The angle ϕ can be designated after verifying the relationships {2.9} to {2.17} between
the numerator Num and denominator Den in {2.8}.

{2.9} Num < 0 Den > 0 ⇒ ϕ = ϕ*

{2.10} Num < 0 Den < 0 ⇒ ϕ = ϕ * − 90°

{2.11} Num > 0 Den < 0 ⇒ ϕ = ϕ * − 90°

{2.12} Num > 0 Den > 0 ⇒ ϕ = ϕ * − 180°

{2.13} Num = 0 Den > 0 ⇒ ϕ = 0°

{2.14} Num < 0 Den = 0 ⇒ ϕ = −45°

{2.15} Num = 0 Den < 0 ⇒ ϕ = −90°

{2.16} Num > 0 Den = 0 ⇒ ϕ = −135°

{2.17} Num = 0 Den = 0 ⇒ rotation symmetric state of stress

2.3.2 Average Value over Hole Depth - Blind-Hole Case


In many measurement applications where an average stress value is desirable (e.g. in
quality control tasks) the generation of a through-hole is not feasible because the
component thickness is too big compared with the drilling tool. In addition, it is
impossible to register analyzable strain changes after drilling a hole with a depth z of
approximately the half hole diameter d0. Alternatively, it is possible to work with a
bigger hole diameter or to drill a blind hole in the material. The latter leads to a complex
stress state near the hole because the strains are always partially relieved in these cases.
Therefore, an analytical solution does not exist for the average value in the blind hole
case. Rather, a previous calibration is necessary to determine a relationship between a
known stress state and the relieved strains after drilling the hole. The known stress state
can be applied either experimentally using e.g. a tensile machine or numerically by
calculating a finite element model. The calibration specimen or calibration model
respectively has normally an ideal geometry, i.e. a thick wide plate. In practice,
calibration constants A BH and B BH for the blind hole case are determined to substitute
the calibration constants ATH and B TH in {2.1} [11].
25

2.3.3 Residual Stress Depth Distribution


It is possible to determine a residual stress depth distribution by introducing the hole
incrementally i.e. drilling the hole in consecutive steps and recording strain values for
every depth. There are several methods of evaluation for the calculation of the residual
stress depth distribution, which differ basically in the physical assumption that leads to
the strain relaxation. The most common methods are:

• the differential method [1, 2, 18]


• the integral method [21, 23, 24]
• the powerseries method [22]
• the average stress method [32]

For example, in differential methods it is assumed that the registered strain values at a
specific depth z depend only on the stress values at this depth z. In contrast, integral
methods consider simultaneously the contribution of all stresses at all depths to the
measured strain for a specific depth z. As can also be seen in the blind hole case
(chapter 2.3.2) a previous calibration is needed according to the specific evaluation
method. The actual evaluation algorithms also take into consideration the ideal plate
geometry for the calibration specimen or calibration model respectively.

In the next chapters, the differential MPA method and the integral method are described
in detail.

2.3.3.1 Differential MPA Method


The differential MPA-Method was developed in 1993 by Kockelmann and Schwarz in
the Material Research Laboratory (Materialprüfungsanstalt MPA) at the University of
Stuttgart [1, 2, 25]. It is based on the works of Kelsey [18] and Kockelmann and König
[33]. Fig. 2.5 resumes for the MPA-Method the relationship between the calibration
(upper part) and the stress calculation over the measured strains (lower part). To reduce
the amount of possible combinations the hole depth z is normalized over the hole
diameter d 0 obtaining the non-dimensional depth ξ {2.18}. In the following

explanations, it is assumed that the calibration hole diameter d c ,0 equals the

measurement hole diameter d 0 .


26

z
{2.18} ξ =
d0

When carrying out the calibration for the hole drilling method using the MPA-Method,
a hole of the diameter d c ,0 is drilled incrementally in a specimen subjected to a known

external stress state with the calibration stresses σ c , x (ξ ) and σ c , y (ξ ) . The calibration

stresses in the MPA-Method can be applied either in an experimental or numerical way.


The relaxed calibration strains ε c, x (ξ ) and ε c, y (ξ ) are differentiated in order to calculate

the calibration functions K x (ξ ) {2.19} and K y (ξ ) {2.20}. Note that in the case of a

calibration, only two strain gages, which are positioned in direction of the two
(principal) calibration stresses, are required The calibration is independent of the
material, i.e. the elastic parameters E and ν of the calibration specimen or model are
not necessarily equal to the material of the measured component.
dε c , x (ξ ) dε (ξ )
⋅ σ c , x (ξ ) − c , y ⋅ σ c , y (ξ )
dξ dξ
{2.19} K x (ξ ) =
1 2
E
[
σ c , x (ξ ) − σ c2, y (ξ ) ]

dε c , x (ξ ) dε (ξ )
⋅ σ c , y (ξ ) − c , y ⋅ σ c , x (ξ )
dξ dξ
{2.20} K y (ξ ) =
ν
E
[σ 2
c, x (ξ ) − σ c2, y (ξ ) ]
To calculate the unknown residual stress depth distribution in a component, a hole with
the hole diameter d 0 is also incrementally introduced. In this case the depth increments

of the measurement Δz may be different to the calibration depth increments Δz c . The

derived strain readings dε i (ξ ) / dξ of the three different strain gages as well as the

calibration values K x (ξ ) and K y (ξ ) are used in {2.21} to {2.23} in order to calculate

the stress values σ i (ξ ) in the direction of the respective strain gages.

E ⎡ dε (ξ ) dε (ξ ) ⎤
{2.21} σ 1 (ξ ) = ⋅ ⎢ K x (ξ ) ⋅ 1 + ν ⋅ K y (ξ ) ⋅ 3 ⎥
K (ξ ) − ν K y (ξ ) ⎣
2
x
2 2
dξ dξ ⎦
27

{2.22}

E ⎡ dε (ξ ) ⎛ dε (ξ ) dε 3 (ξ ) dε 2 (ξ ) ⎞⎤
σ 2 (ξ ) = ⋅ ⎢ K x (ξ ) ⋅ 2 + ν ⋅ K y (ξ ) ⋅ ⎜⎜ 1 + − ⎟⎥
K (ξ ) − ν K y (ξ ) ⎣
2
x
2 2
dξ ⎝ dξ dξ dξ ⎟⎠⎦

E ⎡ dε (ξ ) dε (ξ ) ⎤
{2.23} σ 3 (ξ ) = ⋅ ⎢ K x (ξ ) ⋅ 3 + ν ⋅ K y (ξ ) ⋅ 1 ⎥
K (ξ ) − ν K y (ξ ) ⎣
2
x
2 2
dξ dξ ⎦

Fig. 2.5: Differential MPAII Method: Relationship between Calibration and Calculation of Stress

The maximal and minimal stresses are then calculated easily by using the relationships
in Mohr’s circle {2.24}.

σ 1 (ξ ) + σ 3 (ξ ) 1
{2.24} σ max, min (ξ ) = ± ⋅ (σ 1 (ξ ) − σ 2 (ξ ))2 + (σ 3 (ξ ) − σ 2 (ξ ))2
2 2
28

The angle ϕ ( z ) is calculated for each increment similar to {2.8}, with the difference
that the equation uses the stress values in strain gage direction σ1,2,3 instead of strain
values ε1,2,3.

2.3.3.2 Integral Method


The basic formulae of the integral method shown in this chapter are taken from
Schajer’s 1988 work [23, 24] and are based on the previous works of Bijak-Zochowski
[21]. The method is implemented among other things in the commercial residual stress
evaluation programme H-Drill [34]. To give a brief explanation of the integral method,
an equi-biaxal plane stress state P(H ) at a normalized depth from surface H {2.25} is
assumed (using almost the same nomenclature as in [23, 24]). Thus, the global strain
response p (h ) in the actual normalized hole depth h {2.26} would be equal for all
strain gages. In the case of the integral method the normalizing factor is the strain gage
mean radius rm and not the hole diameter d 0 as in the differential MPA-Method.

Z
{2.25} H = with Z: depth from surface
rm

z
{2.26} h = with z: hole-depth
rm

The relationship between the strain p (h ) and the stress P(H ) is given in {2.27}, where

Aˆ ( H , h) is the strain relaxation per unit depth or influence function to integrate.

1 +ν
Aˆ ( H , h) P( H )dH
h
{2.27} p (h) =
E ∫
0
0≤H ≤h

The differential notation of the strain relaxation p (h ) {2.27} is unusable because the
strain values pi are recorded in practice after each drilling step or increment i and not
continuously. Therefore, {2.27} can be approximated by the discrete form in {2.28}
where Pj is the equivalent uniform stress within the j − th hole depth increment and aij

is the strain relaxation due to a unit stress within increment j of a hole of i increments
29

deep. {2.29} is an equivalent matrix notation in which four hole depth increments are
given as examples.

1 +ν j =i
{2.28} pi =
E
∑a
j =1
ij Pj 1≤ j ≤ i ≤ n

⎡ p1 ⎤ ⎡ a11 ⎤ ⎡ P1 ⎤
⎢p ⎥ ⎢a ⎥ ⎢P ⎥
1 +ν ⎢ 21 a 22 ⎥⎢ 2 ⎥
{2.29} ⎢ 2 ⎥ =
⎢ p3 ⎥ E ⎢a31 a 32 a33 ⎥ ⎢ P3 ⎥
⎢ ⎥ ⎢ ⎥⎢ ⎥
⎣ p4 ⎦ ⎣a 41 a 42 a 43 a 44 ⎦ ⎣ P4 ⎦

Fig. 2.6: Integral method: loading steps of calibration and matrix of strain relaxation

The calibration of the integral method is usually done by FEM calculation because it is
almost impossible to apply experimentally the stress Pj at a specific increment j of a
30

hole. The loading steps of the calibration with the stress Pj and the resultant matrix of

strain relaxations aij are exemplarily shown in Fig. 2.6 for the case of four hole-

depth increments i . Here, Pj is applied e.g. as a pressure load on specific depth sector

j of the inner surface of the hole, which is schematically represented for each loading
step with a geometrical two dimensional hole which has been halved. Hence, each value
of aij can be numerically determined and the unknown residual stress distribution can

then be then calculated using the measured strain distributions pi in the inverted form
of {2.28}.

2.4 Geometrical Boundary Conditions


2.4.1 Prefacing Remarks
The geometrical conditions of the component play, in some cases, an important role
when carrying out a hole drilling measurement. One of the first problems that may arise
is the application of the strain gage rosette at the desired measurement point. This
means that the component geometry at the measurement point should allow the bonding
of the strain gage rosette, e.g. the width of the component should be greater than the
dimensions of the rosette. Moreover, the geometrical conditions around the
measurement point should not constrict the drilling tool or device when introducing the
hole.

Apart from these practical problems, different geometrical parameters influence the
evaluation of a hole-drilling measurement. The most important geometrical parameters
are the strain gage mean diameter d m , the hole diameter d 0 and the hole depth z ,
because they directly influence the calibration strain as well as the measured strain. For
example, a set of calibration strains is always valid for a specific strain gage rosette and
a specific hole diameter. The strain gage mean diameter d m is treated as a constant
parameter, because the calibration is normally carried out for one specific strain gage
rosette. In contrast, the hole-diameter can differ between the diameter d c ,0 , which is set

after the calibration, and the diameter d 0 , which is determined after the measurement.
Evaluation algorithms contain a limited set of calibration functions or calibration
constants for specific hole-diameters, which range between the maximum and the
31

minimum possible (measurement) hole diameter d 0 . In this case, an interpolation of the


specific calibration strains, calibration functions or calibration factors respectively is a
common procedure [24, 33]. The same difficulties appear concerning differences
between the calibration hole depth increments Δz c and the hole depth increments of the
measurement Δz .

In addition, the calibration of the hole-drilling method for the different residual stress
evaluation algorithms is normally carried out using, as already mentioned, the geometry
of an ideal thick and wide plate. For this reason, errors in stress calculation could appear
when evaluating the residual stress after a hole-drilling measurement on a real
component e.g. a part with a thin cross section and a curved shape. In some works, a
geometry-specific calibration is suggested in order to minimize the calculation errors (s.
Ch. 2.4.3). This means that in these cases some geometrical values should be treated as
an additional parameter in the evaluation.

Thus, the main focus of this work is the investigation of the influence of the component
geometry on the residual stress calculation and a correction of potential errors. In the
following chapters geometrical parameters are listed and an overview of previous
researches dealing with the geometry influence on the results of the hole-drilling
method is given.

2.4.2 Definition of Possible Geometrical Boundary Conditions


Real components are characterized by a complex shape compared to the shape of the
calibration specimens or calibration models respectively, on which the hole-drilling
evaluation methods are based. Although some hole-drilling measurements are carried
out in simple geometrical parts, e.g. in sheet metal, many measurement points of real
components are characterized by a combination of two or more geometrical conditions.
Tab 2.3 lists geometric parameters that may influence the results of the hole-drilling
measurement, if the defined limit of these parameters is under-run.
32

Symbol Description
T Component Thickness
D Distance between hole center and free edge or border of component
DE Distance between hole center and edge of component
DH Distance between two holes
W Component Width
R Uniaxial curved shape: Outside radius of shell or cylinder
RI Uniaxial curved shape: Inside radius of shell or cylinder
RB Biaxial curved shape: Outside radius of shell or sphere
RBI Biaxial curved shape: Inside radius of shell or sphere
Tab. 2.3: Geometrical parameters which can influence a hole-drilling measurement

2.4.3 Overview of Previous Investigations


The influence of geometry on residual stress analysis has already been investigated by
several researchers, who basically discuss the individual effect of the component
thickness T, the distance between the hole and the component edge D (or the
component width W) and the radius of surface curvature R (s. Fig. 2.7).

2.4.3.1 Thickness
In ASTM 837 [11] a definition of thick and thin specimens is already considered. Thus,
a specimen with a thickness T > 0.4 × d m is considered to be thin whereas T < 1.2 × d m

is considered to be thick. (e.g., thin T > 2.052mm and thick T < 6.156mm for a
062UM-120 strain gage rosette). The intermediate case T = 0.4...1.2 × d m is not within
the scope of the standard. According to ASTM 837, the stresses are assumed to be
uniform throughout the hole depth. Otherwise the result should be considered as non-
standard. A measurement on a thin specimen should be carried out and evaluated using
the procedure for the through-hole case. In the case of drilling a hole in a thick
specimen, the standard provides calibration data for different strain gage rosette types.
33

Fig. 2.7: Selected types of geometrical boundary conditions for the hole-drilling technique

In earlier works Altrichter [35] and Motzfeld [36] pointed out possible sources of errors
when carrying out a hole-drilling measurement. They both described the procedure for
the through-hole case and gave limits of thickness with T > d 0 [35] and T > 0.5 × d 0
[36]. At that time, no standardized strain gage rosettes were used and the dimensions of
e.g. the hole-diameter were generally bigger then in modern hole-drilling
measurements. Motzfeld used e.g. a self-made hole-gage assembly with a hole diameter
of d0 = 6mm and a sheet thickness of T = 12mm respectively [36]. Using such
experimental configuration the total relieved strain is the sum of the strain relaxation
due to the applied stress, the strains due to drilling operation and the strain due to
localized plastic flow as established in [37], where the alignment of the hole-gage
assembly was modified to a recommended range of nondimensional hole diameter of
34

λ = d0/rm = 0.5...1.2, which corresponds approximately to the suggested values in the


later TN-503 [12].

The influence of the specimen thickness on the results of a residual stress vs. depth
calculation using an in-depth evaluation method was shown by Münker in [38]. He
calculated, among other things, the stress difference in percentage which arises after
comparing the hole-drilling results of thin specimens in relation to a reference thick
specimen. Fig 2.8 shows the stress error in percent vs. the specimen thickness taken
from [38]. The calculations were made using FEM models of different thickness which
were loaded with an equibiaxal plane stress condition σII / σI = 1 and an equibiaxal pure
shear stress condition σII / σI = -1. The stress evaluation was carried out according to the
powerseries method. The stress differences are negligible for thick specimens with
T > 2.5mm . For thin specimens the maximal difference is more than 20% for the
models with plane stress condition and more than 10% for the models with pure shear
condition. Münker pointed out that for the investigation of thin components a specific
calibration which considers the component thickness is necessary.

Fig. 2.8: Stress difference against the component thickness T according to [38]

The relationship between the calibration coefficient for the integral in-depth evaluation
method and plate thickness was investigated numerically as well as experimentally by
Aoh and Wei in [39] and [40]. Thus, the calibration coefficients vary with the thickness
35

of the flat specimen so that the accuracy of residual stress analysis can be improved if
appropriate calibration coefficients are chosen to match the component thickness. For
this purpose, a transition range between thin and thick plates is given with
T = 1.34...2.0 × d m ( T = 1.88...2.811 × d 0 ). Considering the dimensions of the strain
gage rosette (062RK-120) used in [40] the transition range of the specimen thickness is
about T = 3.44...5.13mm . In a practical application, Schajer specifies in [34] the
acceptable component thickness as being T > 1.2 × d m for the correct calculation of
residual stresses using the H-Drill software.

In his work in the area of metal forming, Haase [41] underlines the need to adapt the
calibration to the component thickness, if the component is thin with T < 2.5 × d 0 . The
developed evaluation program there is based on the integral method and has
implemented several sets of calibration coefficients for different metal sheet thickness.

Kockelmann and König [33], who developed a parent version of the differential in-depth
evaluation MPA-Method, defines the limit of component thickness for a correct stress
calculation as being T > 3 × d 0 . He pointed out that a measurement on thin structures
could lead to misinterpretations because global stress changes and deformation
additionally influence the measurement results.

2.4.3.2 Distance between Hole Center and Component Edge


The ASTM 837 standard as well as the TN-503 do not contain recommendations
dealing with limits of the distance between the hole and the component edge. Altrichter
[35] and Motzfeld [36] set the limits of the distance hole-edge with D > 15 × d 0 for a
correct calculation of a through-hole measurement.

Münker [38] investigated, analogical to the thickness, the influence of the distance
between the hole and the component edge by varying the model width W = 2 × D (s.
Fig 2.9). A significant edge effect appears only by a width of W < 20mm or a distance
D > 10mm respectively. Considering specimens with D = 5mm , the maximal
difference is c. 7.5% for the models with plane stress condition and c. 12.5% for the
models with pure shear condition. Schajer [34] gives the minimal component width
36

with as W = 3 × d m for the correct use of the H-Drill software. This means a component

width of W = 15.4mm for a 062UM-120 strain gage rosette.

Fig. 2.9: Stress difference against the distance hole-edge D according to [2.35]

Preckel [42] carried out investigations in a 2 mm thick metal sheet plate using different
loading conditions (equibiaxal plane stress condition and equibiaxal shear stress
condition). He defined the minimal distance from the border of the hole to the
component edge as being Da = 5 × d 0 . Distances of Da ≥ 5 × d 0 yield to a maximum
relative error of less then 10% (s. Fig 2.10). Similar observations were made by König
[33].
37

Fig. 2.10: Maximum Relative Error against the distance hole-edge Da according to [33]

2.4.3.3 Radius of Surface Curvature


Preckel calculated in [42] the effect of the surface curvature with a finite element
analysis using a sphere model. According to this, the maximal relative error is c. 15% in
a specimen with surface curvature radius of R B = 40mm . Therefore, for a correct
execution of the hole-drilling method he suggested a surface radius of a biaxial curved
shape with R B ≥ 100mm whereas the other surface radius should in be all cases
RB > 15mm , to provide an appropriate application of the strain gage rosette. Also, in the
case of the surface curvature, Haase [41] underlines the need to implement a geometry-
specific calibration in the evaluation algorithms.

According to Kockelmann and König [33], the correction of the influence of the surface
curvature cannot be given in general and it is possible only in specific cases. To avoid
errors greater than 20% the surface radius should be R ≥ 3 × d 0 with an evaluable depth

of z ≤ R / 4d 0 (e.g. d 0 = 1.8mm correspond to R = 5.4mm and z ≤ 0.75mm ).

In [43], Zhu and Smith developed an analysis for interpreting relaxed strain data from
the incremental hole drilling method and an integral in-depth evaluation for components
38

with curved surfaces. For this purpose, the geometry of an 8mm diameter steel was
explicitly considered in three dimensional FE calculations, which provide the stiffness
coefficients for the evaluation method. The developed analysis was tested for 8mm
diameter hot forged steel round bars. According to the paper, the obtained results
showed a good correlation when compared with x-ray diffraction measurements.

The influence of a complex geometry on the results obtained with the hole-drilling
method was investigated in several works by Montay, Lu et al. Using FE calculations
and an integral evaluation method the hole-drilling method was adapted to the following
three dimensional shapes: (1) a sphere [44], (2) a cylinder [45] and (3) a crankshaft [46].
The calibration coefficients show a dependence on the curvature radius for the spherical
models as well as for the cylindrical models. According to [2.41], the surface of a
sphere can be considered to be plane, if the sphere radius is RB > 17.5 × d 0 . As shown in
Fig. 2.11, the maximal calculated error on numerical calibration coefficients is c. 35%
for a specimen with a sphere radius of R B = 5 × d 0 (with RB = 10mm and d 0 = 2mm ).

Fig. 2.11: Calculated error of calibration coefficients over the hole-depth for different sphere radius R
[44]
39

Fig. 2.11: FE mesh around a hole in the fillet of a crankshaft for the calculation of geometry-specific
calibration coefficients for the hole-drilling method [46]

In case of the investigation with the cylindrical model [45], the authors suggest a
cylinder radius of c. R > 25 × d 0 (with R > 50mm and d 0 = 2mm ) to be considered as
plane, because no significant variation in the calibration coefficients depending on the
cylinder radius was observed. In [46] the hole-drilling method was adapted in the fillet
of a crankshaft model (s. Fig. 2.12) and as a result, geometry-specific calibration
coefficients were calculated in order to solve residual stress evaluation in complex
shapes.

2.5 Concluding Remarks on State of Knowledge


The measurement and analysis of residual stress is of great interest in the field of
material engineering as well as for industrial objectives. Therefore, the hole-drilling
method provides a versatile, easy to handle and economic measurement solution. Like
other measurement techniques, the hole drilling method is influenced by different
parameters which can restrict in special cases the use of the method. One possible origin
of measurement errors is the influence of the geometry of a component with complex
shape. This is due to the fact that most of the evaluation algorithms include calibration
functions or coefficients, which were determined by using the geometry of a simply-
shaped specimen or model. One possible approach to this problem is the definition of
geometrical limits as established by different studies. Tab. 2.4 summarizes the
geometrical boundary conditions for the adequate use of the differential MPA-Method
[2] and is based mainly on the previous investigations of Kockelmann and König [33].
40

Another possibility proposed by different researchers is the implementation of a


geometry-specific calibration on the evaluation algorithms. Examples of such an
implementation on evaluation algorithms based on the integral method have been shown
in different works. A possible correction of the geometry influence using a differential
evaluation method (e.g. MPA-Method) has not yet been investigated.

Boundary Condition Reference Value e.g. d0 = 1.8 mm


Min. Component Width W > 10 - 20 x d0 W > 18 - 36 mm
Min. Component Thickness T > 3 x d0 T > 5.4 mm
Min. Distance Between Hole and Component
D > 5 - 10 x d0 D > 9 - 18 mm
Edge
Min. Distance Between Two Holes DH > 5 x d0 DH > 9 mm
Min. Radius of Surface Curvature R > 3 d0 R > 5.4 mm
Tab. 2.4: Reference values for geometrical boundary conditions according to the MPA-method [2]
41

3 Experimental and Numerical Methods of Investigation


3.1 Prefacing Remarks
The general procedure of the experiments, as well as that of the numerical calculations
in this investigation, is based on the calibration principle of the differential MPA-
Method. This allows the analysis of the influence on the results of a hole-drilling
measurement caused by:
• individual geometrical parameters T, D or R using flat and cylindrical tensile
specimens respectively, or,
• a combination of these geometrical parameters using the same specimens and,
additionally, specimens of complex shape which produce a stress state similar to
that in a real component.

The aim of the experimental calibration is to provide evaluation data in order to


compare and verify the numerical models. For this reason, the specimen’s material, the
specimen’s shape, the used equipment and the calibration process are described in the
first part of this chapter. Ultimately, the objective of the numerical calculations is the
systematic investigation of the geometry influences on the hole-drilling results and the
calculation of geometry specific calibration functions. Hence, the second part of this
chapter specifies the different numerical models used in this investigation as well as the
numerical calibration procedure.

Fig. 3.1: Calibration coordinate system and position of strain gages


42

Fig. 3.1 shows the conventions used for the experimental investigation as well as for the
numerical calibration concerning the coordinate system, the strain gage rosette and the
loading conditions. The point of origin is in both cases, experimentally and numerically,
the hole center at the surface. The experimental calibration is carried out in a uniaxial
testing machine, i.e. the x-axis is defined as the load direction with σc,x as the
corresponding calibration stress. Thus, strain gage no. 3 is positioned parallel to the x-
axis or load direction whereas strain gage no. 1 is parallel to the y-axis. The geometry of
the FEM-models and their loading conditions are based on the experiments. The
numerical calibrations distinguish a stress component σc,x in x-direction and σc,y. in y-
direction. These conventions are effective within the whole work.

3.2 Experimental Details


3.2.1 Material
The specimens were made of fine grain structural steel S 690 QL, which is water
quenched and tempered. The chemical composition as established after an analysis
using sparking spectroscopy is listed in Tab. 3.1.

C Si Mn P Cr Mo Fe
0.180 % 0.599 % 0.943 % 0.00724 % 0.740 % 0.255 % balance
Tab. 3.1: Chemical composition of S690 QL

E [MPa] Rp0,2 [MPa] ReH [MPa] Rm [MPa] Elongation


[%]
Experiment 210000 747 - 833 14
[47] - 690 690 790-940 16
Tab. 3.2: Mechanical proprieties of S690 QL

The hole-drilling method is based on elastic equations and, consequently, yielding and
plastic strains respectively must be avoided during the calibration. However, stress
concentrations, three times greater than the nominal or calibration stress, arise in the
vicinity of the hole. This could lead to erroneous strain measurements, if the local stress
concentration exceeds the yield stress. On the other hand, the experimental calibration
43

requires the highest possible stress values in order to measure evaluable calibration
strains. With a minimum yield stress of ReH = 690 MPa (s. mechanical properties in
Tab. 3.2) the chosen specimen material S 690 QL offers the possibility to apply
maximal calibration stresses of approx. 210 MPa without local yielding near the hole.

Before carrying out the experiments, the specimens were submitted to a stress relief
annealing to avoid detrimental effects due to residual stresses induced by the
manufacturing. This process was carried out at a temperature of approx. 600 °C during
1.5 h in a protective-gas tube furnace of the type “Labotherm R50/300/13”. Finally, the
specimens were cooled down slowly inside the furnace.

3.2.2 Specimens
Four different types of specimens were used:

Type 1: Flat tensile specimens with different thickness (s. Fig. 3.2).

Fig. 3.2: Flat tensile specimen

The flat tensile specimens allow an individual investigation or a combination of the


geometrical parameter T and D by using specimens with different thickness and by
bonding the strain gage rosette with a defined distance between the hole center and the
specimen’s edge.

The following flat tensile specimens with different geometrical parameters were
investigated:
44

• Specimens with a variable thickness of T= 1.5; 2.0; 2.5; 3.0; 4.0; 6 mm and a
constant distance of D = 20 mm.
• Specimens with a variable distance of D = 2.0; 5.0; 10; 20 mm and a constant
thickness of T = 6 mm.
• Thin specimens with T=1.5 mm and a minimal distance of D = 2 mm.

The reference specimen has a thickness of T = 6 mm with a strain gage rosette bonded
on the center of the specimen with a resulting hole-edge distance of D = 20 mm. In this
case, assuming a maximal hole diameter of d0 = 2 mm, the reference specimen satisfies
the criteria for the geometrical boundary conditions as established in Tab. 2.4.

The individual flat tensile specimens are identified by an abbreviation which includes
both parameters T and D. For example, the abbreviation “fT1.5D2” refers to the flat
specimen with a thickness of T = 1.5 mm and a distance of D = 2 mm.

Type 2: Solid and hollow cylindrical tensile specimens with different cylinder
radius (s. Fig. 3.3).

The solid cylindrical specimens allow the investigation of the geometrical parameter R,
the outside radius of a uniaxial curved shape. Three solid cylinders with R = 3; 4; 6 mm
and a constant gage length of l = 40 mm were manufactured. The strain gage rosette was
always bonded in the center of the specimen (s. Fig 3.3 specimen scR6), i.e. the
geometrical parameter D does not fall below the boundary condition.

The hollow cylindrical specimens allow the investigation of the combination of the
geometrical parameter R and T. They are based on the geometry of the solid cylindrical
specimens with the difference that the wall thickness is always constant with
T = 1.25 mm, which is considered to be thin.
45

Fig. 3.3: Solid and hollow cylindrical tensile specimen

The abbreviations of the cylindrical specimens consider shape of the cross-section of


the specimen (“sc” solid cylinder and “hc” hollow cylinder) as well as the outside
radius R.

Type 3: Flat tensile specimens with a ø10 mm center through-hole and different
thickness (s. Fig. 3.4).

With this type of specimen, a stress state similar to that in a simple flat component can
experimentally be simulated. Considering the notch effect, a biaxial stress state before
drilling at the measurement point can be applied.

The geometry of these specimens base on that of the flat specimens of Type 1 and,
additionally, have a through-hole of ø10 mm in the center. The strain gage rosette is
bonded near the edge of the ø10 mm through-hole with D = 2 mm. A thick specimen
with T =6 mm and a thin specimen with T =1.5 mm were manufactured.
46

Fig. 3.4: Flat tensile specimens with a ø10 mm center through-hole and different thickness

In addition, a single strain gage is placed axial to the centerline of the ø10 mm hole (s.
Fig. 3.4) in order to control the load application and correlate the control strain readings
with the subsequent simulation.

The specimens are identified with an abbreviation which includes the thickness (T) of
the specimen and the distance to the edge of the ø10 mm hole (D). For example
“fhT1.5D2” refers to the flat specimen with center hole, a thickness of T = 1.5 mm and
a distance D = 2 mm.

Type 4: Hollow cylindrical tensile specimens with a ø6 mm center through-hole (s.


Fig. 3.5).

Similar to the Type 3 specimens a through-hole of ø6 mm is introduced in the center of


a hollow cylindrical specimen and over its whole cross-section in order to
experimentally simulate a stress state which can be found in a component.

This specimen bases on the geometry of the hollow cylindrical specimen “hcR6” and
provides, on the other hand, a combination of the three geometrical parameters, which
are investigated in this work: a wall thickness of T = 1.25 mm, a distance to the edge of
the ø6 mm through-hole of D = 2 mm (when bonding the strain gage rosette
accordingly) and a curvature radius of R = 6 mm.
47

Fig. 3.5: Hollow cylindrical tensile specimen with a ø6 mm center through-hole

Analog to the Type 3 specimens, a single strain gage is placed axial to the centerline of
the ø6 mm hole (s. Fig. 3.5) in order to control the load application and correlate the
control strain readings with the subsequent simulation.

The abbreviation for this type of specimen is “hchR6” and means a hollow cylindrical
specimen with a center hole and an outside radius of R = 6 mm.
48

3.2.3 Hole Drilling Device and Calibration Equipment

Fig. 3.6: Experimental calibration set-up

The hole in the measurement point on the specimen was introduced using the high-
speed drilling device “RS-200 Milling Guide” of the company “Measurements Group”.
The high-speed turbine generates, according to the manufacture’s specifications, a
rotation of the cutting tool of up to 300.0000 r.p.m., which ensures that almost no
plastic deformation in the vicinity of the hole occurs due to the drilling process. Special
milling cutters with titanium-nitrite coating were used for drilling. Their diameter is
ø1.6 mm and leads to a typical hole diameter of approx. d0 = 1.8 mm. The relieved
strains were measured using the strain gage rosettes of type CEA-06-062-UM (also
“Measurements Group”). The strain gages were connected in a (Wheatstone) quarter
bridge and were completed to a full bridge by using an additional temperature
compensation strain gage and the amplifier “DMCPlus” of “HBM”. During calibration
the values of the different strain gages were displayed on-line using data acquisition
software programmed with “Testpoint” of the company “Keithley”. After each depth-
49

increment, the strain values were recorded. The final calibration data was written to an
ascii- file.

The experimental calibration set-up is shown in Fig 3.6. The tensile specimens were
loaded using a universal testing machine of the type Zwick Z100 which has a 100 kN
load cell. For the specimens without a center through-hole the calibration stress was
calculated with σc = F / A (F: applied uniaxial force; A: specimen’s cross-section area).
In the case of the specimens of Type 3 and 4, the applied load was set using a control
strain gage. The hole-drilling device was jointed using fixing screws to the lower
clamping jaw of the testing machine in order to place the milling cutter perpendicular to
the clamped specimen. For this purpose, a special mount was designed. The mount
allows height and lateral adjustment in order to arrange the hole-drilling device and,
especially, the milling cutter over the measurement point.

3.2.4 Experimental Calibration Procedure


Before starting with the calibration the strain gage rosettes were applied at the desired
place on the specimen according to the recommendations in [48, 49]. The steps of the
calibration procedure are (s. Fig 3.7):

(1) Engage specimen to clamping jaws by loading and unloading with a small stress
magnitude.
(2) Zero-balance the strain gage circuits.
(3) Apply a load σc,x,min (50 MPa) and measure strain (before drilling).
(4) Apply a load σc,x,max (210 MPa) and measure strain (before drilling).
(5) Unload specimen. Arrange the hole-drilling device over the measurement point
and establish zero depth.
(6) Apply a load σc,x,min (50 MPa) and measure strain (z = 0) in order to verify step
3.
(7) Apply a load σc,x,max (210 MPa) and measure strain (z = 0) in order to verify step
4.
(8) Unload specimen. Introduce first drilling increment (e.g. z = 0.02 mm).
(9) Apply a load σc,x,min (50 MPa) and measure strain at first increment (e.g.
z = 0.02).
50

(10) Apply a load σc,x,max (210 MPa) and measure strain at first increment (e.g.
z = 0.02).

… Repeat steps 8 to 10 using sufficient depth increments until reaching the


maximum hole depth.
(n-1) Apply a load σc,x,max (210 MPa) and measure strain at maximum hole depth.
(n) Unload specimen and measure the hole diameter.

Fig. 3.7: Experimental calibration procedure

The relieved calibration strain Δε cΔ,σi c ( z ) , which is related to the calibration stress

Δσ c {3.1}, is calculated after the calibration procedure using {3.2}.

{3.1} Δσ c = σ max − σ min here:

Δσ c = 210MPa − 50 MPa = 160MPa

{3.2} Δε CΔσ,i c ( z ) = [ε cσ,imax ( z ) − ε cσ,imin ( z )] − [ε cσ,imax (0) − ε cσ,imin (0)]

This procedure for the determination of the calibration strain Δε c ,i ( z ) is proposed in

different works [38, 41] and should ensure that the calibration stress is only dependent
on the applied uniaxial calibration load. Using stress differences, other sources of load
are omitted from the calculation. In general, undesirable sources of load could be the
51

prestress caused by some small inexactness when fixing the specimen to the clamping
jaws in the form of bending or torsion or residual stresses in the specimen caused by
manufacturing.

After finishing the calibration, the hole diameter is measured using the microscope
assembly which is provided with the “RS-200 milling guide”.

3.3 Numerical Methods and Calibration


3.3.1 Models and Simulation Procedure
The geometry of all three-dimensional finite elements models was generated using the
pre-processor “MSC.Patran”. The simulation itself was carried out by means of the
finite element code ABAQUS Standard [50]. For this purpose, a PC was used with a
“Pentium 4” processor of 3.4 GHz and a RAM of 4.0 GB. In all cases, the drilling of the
hole was simulated by applying the birth and death option to the elements forming the
hole volume. All models are based on the geometry and the strain gage rosette
arrangement of the specimens of the experimental calibration (s. Ch 3.2.2). Thus, four
types of models were generated:

Type 1: 3D-models of flat tensile specimens with different thickness (s. Fig. 3.8).

The models of the flat tensile specimen allow, similar to the experimental calibration, an
individual investigation or a combination of the geometrical parameter T, D and, in
addition, the consideration of different hole diameters d0. For this, a prototype model
was first generated. The models with different geometrical parameters T, D and d0 were
modelled by deleting and rearranging the elements of the prototype model.

The following geometrical parameters were considered for the models of the flat
specimens:

• Thickness: T = 1; 1.5; 2; 3; 6 mm
• Distance hole-edge: D = 2; 3.5; 5; 10; 20 mm
• Hole diameter d0 = 1.7; 1.75; 1.8; 1.85; 1.9 mm
52

The combination of all geometrical parameters leads to 125 models of the flat tensile
specimens. The number of elements varies between 26,460 for the models with
T = 1 mm and 41,580 for the models with T = 6 mm. The elements are three-
dimensional solid elements of the type C3D8 (8-node linear brick). Only some elements
in the hole volume are of the type C3D6 (6-node linear triangular prism). Due to
symmetry, only the half geometry of the specimens was modelled, i.e. the left side of
the models is regarded as the symmetric plane. The main dimensions of the models as
well as the approximate location of the strain gages are shown in Fig 3.8.

Fig. 3.8: 3D-model of a flat tensile specimen

The model fT6D20m with T = 6 mm and D = 20 mm is defined as the reference model,


because it satisfies the criteria for the boundary conditions (s. Tab 2.4). The reference
hole diameter is specified with d0 = 1.8 mm for this investigation.
53

The abbreviation of the models is the same as the specimen with the addition of an “m”
for “model” at the end. For example, “fT1.5D2m” refers to the model of the flat
specimen with T = 1.5 mm and D = 2 mm.

Type 2: 3D-models of solid and hollow cylindrical tensile specimens with different
cylinder radius (s. Fig. 3.9 and Fig. 3.10).
In the case of the solid cylinder, three different models were generated which consider
three different cylinder radius R = 3, 4, 6 mm. The hollow cylinders were generated out
of them by deleting the corresponding elements and creating a tube profile. All hollow
cylindrical models have a thickness of T = 1.25 mm. Fig 3.9 shows exemplarily the
shape and the main dimensions of the models of the hollow cylindrical tensile
specimens. For all cylindrical models a hole diameter of d0 = 1.8 mm was modelled.

Fig. 3.9: 3D-models of hollow cylindrical tensile specimen whit different cylinder radius R
54

Fig. 3.10: Detailed views of the hole region in the hcR6m model

Only a quarter of the cylinder needs to be modeled due to symmetry. Fig. 3.10 shows
exemplarily for the hcR6m model two detailed views of the hole region. In this picture,
there is schematically illustrated the hole volume, both symmetric planes and the
approximate location of the strain gages in x- and y-direction.

The majority of the elements used are of the Type C3D8. In some regions of the models
C3D6 elements were needed in order to create a congruent mesh of the cylinder
curvature. In Tab. 3.1 the number of elements for each cylindrical model is listed.

Models scR6m hcR6m scR4m hcR4m scR3m hcR3m


No. Elements 70,746 46,410 47,760 33,348 33,988 26,284
Fig. 3.3: Total number of elements for the different models of the cylindrical tensile specimens.

The abbreviation for the different cylindrical models is the same as for the specimens
with the addition of an “m” at the end.
55

Type 3: 3D-models of flat tensile specimens with a ø10 mm center through-hole


and different thickness (s. Fig. 3.11).

The models of this type of flat tensile specimen utilize the notch effect near the ø10 mm
center through-hole, which causes a non-uniform bi-axial stress state. In a real
component it is assumed that approximately such a stress state exists.

Two specimens were modelled, a “thin” one with T = 1.5 mm and a “thick” one with
T = 6 mm. The measurement point is modelled with a distance of D = 2 mm to the edge
of the ø10 mm through-hole. The stress magnitudes before drilling at the hole volume,
i.e. the resulting calibration stresses σc,x and σc,y, can be averaged over the depth after
each numerical calibration. Fig. 3.11 shows the main dimensions of these Type 3
specimens, the location of the measurement point (with a variable hole diameter of
d0 = 1.7; 1.8, 1.9 mm) and the approximate location of the strain gages. Additionally,
the models of Type 3 are longer than the models of Type 1 in order to get strain
readings in the region of the control strain gage. The number of elements varies between
47,120 for the models with T = 1.5 mm and 76,570 for the models with T = 6 mm. With
the exception of some C3D6 elements in the hole volume, the elements are three-
dimensional solid elements of the type C3D8. Because of the symmetry, only half of the
geometry of the specimens was modelled.

Two abbreviations based on the experiments are used for the two different models:
“fhT1.5D2m” for the “thin” model and “fhT6D2m” for the thick model.
56

Fig. 3.11: 3D-model of a flat tensile specimen with a ø10 mm center through-hole

Type 4: 3D-model of the hollow cylindrical tensile specimens with a ø6 mm center


through-hole (s. Fig. 3.5).

Similar to the models of Type 3, this model of a cylindrical tensile specimen utilizes the
notch effect near the ø6 mm center through-hole, which causes a bi-axial stress state
with a different depth distribution at the measurement point. In this case, it is also
assumed that the stress state is similar to that in a real component. This model combines
all three geometrical parameters used in this investigation and bases on the geometry of
the hcR6m model with an identical mesh of the hole region or measurement point (s.
Fig. 3.10).

Fig. 3.12 shows the shape of the Type 4 model with the main dimensions, the location
of the measurement point, the approximate location of the strain gages of the hole
drilling rosette, the approximate location of the control strain gage and the symmetric
plane. Similar to the previous examples, only half of the geometry of the specimens was
modelled, due to symmetry. The stress magnitudes at the hole volume before drilling,
i.e. the resulting calibration stresses σc,x and σc,y, can be averaged over each drilling
depth increment after each numerical calibration. The model contains a total of 92,596
elements; most of them are of the type C3D8. Nevertheless, some elements of the type
57

C3D6 were required in order to mesh the curvature of the cylinder near the
measurement point.

The abbreviation for this model is “hchR6m” and is based on the previous abbreviation
of the experimental specimen with the addition of an “m” for “model”.

Fig. 3.12: 3D-model of the hollow cylindrical tensile specimen with an ø6 mm center through-hole

3.3.2 Material Model, Load Application and Boundary Conditions of the FE-
Models
The evaluation methods of the hole-drilling method are based on the linear-elastic
theory. For this reason, the used material model is in all calculations, unless otherwise
noted, linear-elastic with a Young Modulus of E = 210000 MPa and a Poisson’s Ratio
of ν = 0.285.

The loading and the boundary conditions of the models are shown schematically in
Fig. 3.13. The displacement of all nodes in z-direction is always free for all four model
58

types. The calibration stresses for the models of Type 1 and 2 were applied using
preloaded elements with the ABAQUS function “initial conditions, type = stress”. Thus,
to satisfy the symmetry conditions and to ensure a uniform stress state in x-direction
(and / or sometimes in y-direction) before simulating the drilling process, all nodes of
the faces perpendicular to the xy-plane were fixed, i.e. they were assigned with zero
displacement for the whole simulation (s. loose bearing symbols in Fig. 3.13).

Fig. 3.13: Boundary conditions of the models and load application

For the models with a center through-hole (Type 3 and 4) an external load was applied
to the face in front of the symmetric plane. This external load generates, due to the
notch effect, a stress concentration at the measurement point. The stress concentration
can thus be separated in resulting calibration stresses σc,x and σc,y, which are averaged
over the hole volume. The magnitude of the external loading was set comparing the
control strain gage of the experimental calibration of the specimens of these types. Only
the nodes on the face of the symmetric plane were fixed in order to satisfy the symmetry
conditions.

3.3.3 Calculation of Strain


The strain field near the hole is inhomogeneous due to the stress concentration. This
means that the measured strains of the hole-drilling rosette are an average value of the
59

strain field over the gage area ASG. Therefore, the strain evaluation of the FEM
simulations should also consider the inhomogeneity of the strain field at the surface of
the model by calculating average values over the location and dimension of the
corresponding strain gage areas ASG.

The main procedure of the strain calculation for one depth increment is illustrated in
Fig. 3.14. On the left side, a model of a flat specimen with the hole area is shown.
Assuming a calibration stress in x-direction σc,x, the limits of a strain field are defined
considering the location of the virtual strain gage in x-direction (1). The dimensions of
the defined strain field are slightly bigger than those from the strain gage. The single
node strain values within the strain field are saved in a matrix in the form of
r r r r r
M = [x , y, ε x ] (2). Here, x and y are the xy-locations of the single surface nodes
r
within the strain field and ε x the corresponding node strains in x-direction. After this,

the coefficients of a polynomial surface function ε x = f ( x, y ) are calculated using

Gaussian elimination (3). The average strain value in x-direction ε x,m is finally

calculated by numerically evaluating a double integral of the surface function ε x


according to {3.3}. In this equation, the radial distances from hole center to strain gage
ends r1 , r2 and the strain gage width w are the endpoints of the interval (s. Fig. 2.4).
r 2 0.5 w
1
{3.3} ε x,m =
ASG ∫ ∫ ε ( x, y )
r1 − 0.5 w
x

This calculation has to be repeated for each depth increment as well as for each
calibration strain gage (εx and εy). In this work, the calculation of the average strain was
automated using the numerical computing environment “Matlab” created by
“MathWorks”. To simplify matters, a small algorithm was programmed in “Matlab”, in
order to sort the output data taken from the ABAQUS result files of the type *.dat and
to create the strain field matrices M for each calibration strain gage and each depth
increment.

It must be pointed out for the cylindrical models that the tangential strain εy was
calculated using a cylinder coordinate system within the models. In this case, the y-
60

coordinate of the cylindrical models is specified as the segment of the circle which
arises from the cylinder circumference.

Fig. 3.14: Calculation of average strain over the gage area ASG using the strain field near the hole

The calibration strains εx(z) and εy(z) are the principal strains because they are oriented
parallel to the principal stresses (s. Fig 3.1). Hence, according to the relationships in
Mohr’s circle, the strain values of strain gage no. 2, which is oriented ± 45° between
strain gage no. 1 and no. 3, are calculated according to {3.4}.

εx +εy
{3.4} ε2 =
2

This calculation is done in order to give an indication of the strain distribution in this
direction in the results diagrams of Chap. 4. The calculated strain values of strain gage
no. 2 are not taken for the subsequent calibration.
61

4 Results of Experimental and Numerical Calibration


4.1 Prefacing Remarks
This chapter presents the results of the experimental calibration as well as selected
results of the numerical calibration. The following individual geometrical parameters
were varied:

• d0 hole diameter
• T specimen thickness (specimens of Type 1)
• D distance between hole and component edge (specimens of Type 1)
• R cylinder radius of solid cylinder (specimens of Type 2).

Additionally, the following combinations were taken into account in the calibrations:

• Thickness T and distance hole-edge D (specimens of Type 1)


• Cylinder Radius R and thickness T (specimens of Type 2)
• Thickness T and distance to center through-hole D (specimens of Type 3)
• Cylinder Radius R, thickness T and distance to center through-hole D (Type 4
specimens)

Two types of diagrams are mostly used to illustrate the obtained calibration results. The
first type of diagram compares the experimental calibration with the numerical
calibration. Three strain distributions Δε measured (or calculated) from the strain gages
of the hole-drilling rosette (strain gage no. 1, 2 and 3) are plotted over the hole depth z.
The strain values are plotted until a depth of z = 1 mm. This depth exceeds a little the
maximal analyzable depth of 0.5 × d 0 (with d0 = 1.8 mm) as mentioned in Chap. 2.23.
The investigated specimens and the experimental hole diameter d0,exp as well as the
simulated hole diameter d0,sim are indicated in the respective diagrams. The applied load
is distinguished in a calibration tensile stress σc,x in the case of the specimens with a
defined cross section (flat and cylindrical specimens of Type 1 and Type 2) and in an
external applied load σx in the case of the specimens with a center through-hole (flat
and cylindrical specimens of Type 3 and Type 4, respectively). In addition to these first
type diagrams, a strain comparison table is included after each diagram. The tables list
62

the experimental and numerical absolute strain values as well as the difference of these
values at selected hole depths.

The second type of diagram shows the influence of geometrical parameters on the
absolute values of the strain distributions of strain gage no. 1 and strain gage no. 3,
respectively. The diagrams of this second type, which always include the strain
distribution of the reference specimen, summarize the previous numerical results into
one figure. Therewith, a first qualitative comparison of the geometrical influences on
the hole-drilling results is established. With the exception of the diagrams in which the
influence of the specimen thickness is demonstrated, the strain values are also plotted to
a depth of z = 1 mm. A discussion of the influence of geometrical parameters is carried
out in the following Chap. 5.

4.2 Reference Specimen and Variation of Hole Diameter - d0


The reference specimen is the one in which the geometrical boundary conditions for the
hole-drilling method are not violated according to the differential stress calculation
method MPA (s. Tab. 2.4). In addition, a hole diameter of d0 = 1.8 mm is defined as the
reference hole diameter. Fig. 4.1 shows the calibration results of the reference specimen
fT6D20. It is a typical hole-drilling calibration result for a tensile-loaded calibration
specimen or model. Strain gage no. 3 is in the load direction of the tensile load; strain
gage no. 1 is transverse to the load direction. Hence, the absolute strain values of strain
gage no. 3 are the highest with negative algebraic sign whereas the absolute strain
values of strain gage no. 1 are the lowest with positive algebraic sign.
63

100

sg 1
50

0
Δε [μm/m]
-50
sg 2
-100 Experiment ·
___
Simulation
-150
fT6D20 (Type 1)
σc,x = 160 MPa
d0,exp = 1.80 mm
sg 3
-200
d0,sim = 1.80 mm

-250
0.0 0.2 0.4 0.6 0.8 1.0 1.2
z [mm]

Fig. 4.1: Comparison of experimental and numerical calibration strain depth distributions for
reference specimen fT6D20 (T =6 mm, D = 20 mm, R = ∞)

Typical comparative strain values for reference steel specimens with a calibration stress
of σc,x = 160 MPa at a hole depth of 0.2 mm, 0.5 mm and 1 mm are marked in the
diagram of Fig. 4.1 with a dashed line box. The experimental and numerical values as
well as the difference between them are listed in Tab. 4.1.

z ε1 [μm/m] ε3 [μm/m]
[mm] experiment simulation difference experiment simulation difference
0.2 7 7 0 -36 -36 0
0.5 23 23 0 -104 -108 4
1.0 58 53 5 -199 -195 -4
Tab .4.1: Comparison of experimental and numerical calibration strain values in selected hole
depths for reference specimen fT6D20 (T =6 mm, D = 20 mm, R = ∞)

Fig. 4.2 and Fig. 4.3 show the influence of different dimensions of the hole diameter on
the values of the strain depth distributions for strain gage no. 3 and strain gage no. 1,
respectively. The results of these diagrams were calculated using the reference flat
model with five different hole diameters. Both diagrams show that an enlargement of
the hole diameter leads to a magnification of the relieved strain values.
64

0.0

Simulation of fT6D20 specimen


with variable hole diameter d0
-50.0 Strain gage no. 3
σc,x = 160 MPa

-100.0
Δε [μm/m]

-150.0
d0 = 1.70 mm
d0 = 1.75 mm

-200.0 d0 = 1.80 mm
d0 = 1.85 mm
d0 = 1.90 mm
Reference
-250.0
0.00 0.20 0.40 0.60 0.80 1.00 1.20
z [mm]

Fig. 4.2: Influence of hole diameter d0 on numerical calibration strain depth distribution (strain
gage no. 3) for reference model fT6D20m
60 d0 = 1.90 mm
Simulation of fT6D20 specimen d0 = 1.85 mm
with variable hole diameter d0
50 d0 = 1.80 mm
Strain gage no. 1
d0 = 1.75 mm
σc,x = 160 MPa
d0 = 1.70 mm
40
Δε [μm/m]

30

20

10

Reference
0
0 0.2 0.4 0.6 0.8 1 1.2
z [mm]

Fig. 4.3: Influence of hole diameter d0 on numerical calibration strain depth distribution (strain
gage no. 1) for reference model fT6D20m

4.3 Variation of Specimen Thickness - T


The results for the experimental and numerical calibration for the flat specimens with a
thickness T of 3 mm, 2 mm and 1.5 mm and D =20 mm = const. are shown in Fig. 4.4,
Fig. 4.5 and Fig. 4.6, respectively. The numerically calculated strain distributions are in
accordance with the experimentally measured depth distributions of the relieved strains,
as can be seen in Tab. 4.2 (fT3D20), Tab. 4.3 (fT2D20) and Tab. 4.4 (fT1.5D20).
65

100

sg 1
50

0
Δε [μm/m]
-50
Experiment • sg 2
-100 Simulation ___

fT3D20 (Type 1)
-150
σc,x = 160 MPa
d0,exp = 1.79 mm sg 3
-200
d0,sim = 1.80 mm

-250
0.0 0.2 0.4 0.6 0.8 1.0 1.2
z [mm]

Fig. 4.4: Comparison of experimental and numerical calibration strain depth distributions for
specimen fT3D20 (T =3 mm, D = 20 mm, R = ∞)

z ε1 [μm/m] ε3 [μm/m]
[mm] experiment simulation difference experiment simulation difference
0.2 6 7 -1 -25 -36 11
0.5 23 24 -1 -92 -107 15
1.0 55 54 1 -186 -194 8
Tab. 4.2: Comparison of experimental and numerical calibration strain values in selected hole
depths for specimen fT3D20 (T =3 mm, D = 20 mm, R = ∞)
100

sg 1
50

0
Δε [μm/m]

-50
Experiment • sg 2
-100 Simulation ___

-150
fT2D20 (Type 1)
σc,x = 160 MPa
d0, exp = 1.78 mm
-200
d0,sim = 1.80 mm sg 3

-250
0.0 0.2 0.4 0.6 0.8 1.0 1.2
z [mm]

Fig. 4.5: Comparison of experimental and numerical calibration strain depth distributions for
specimen fT2D20 (T =2 mm, D = 20 mm, R = ∞)
66

z ε1 [μm/m] ε3 [μm/m]
[mm] experiment simulation difference experiment simulation difference
0.2 7 6 1 -37 -38 1
0.5 23 21 2 -113 -115 2
1.0 55 51 4.0 -210 -208 -2
Tab. 4.3: Comparison of experimental and numerical calibration strain values in selected hole
depths for specimen fT2D20 (T =2 mm, D = 20 mm, R = ∞)

100

50
sg 1

0
Δε [μm/m]

-50

Experiment • sg 2
-100
Simulation ___

-150 fT1.5D20 (Type 1)


σc,x = 160 MPa
-200 d0,exp = 1.78 mm
d0,sim = 1.80 mm sg 3
-250
0.0 0.2 0.4 0.6 0.8 1.0 1.2
z [mm]

Fig.4.6: Comparison of experimental and numerical calibration strain depth distributions for
specimen fT1.5D20 (T =1.5 mm, D = 20 mm, R = ∞)

z ε1 [μm/m] ε3 [μm/m]
[mm] experiment simulation difference experiment simulation difference
0.2 5 5 0 -38 -44 6
0.5 20 17 3 -123 -129 6
1.0 55 51 4 -219 -224 5
Tab. 4.4: Comparison of experimental and numerical calibration strain values in selected hole
depths for specimen fT1.5D20 (T =1.5 mm, D = 20 mm, R = ∞)

The influence of the geometrical parameter thickness T on the calibration strains is


demonstrated in Fig. 4.7 (sg. no. 3) and Fig. 4.8 (sg. no. 1). In these diagrams based on
numerically determined data, a strain distribution of the calibration using a model with a
thickness of T = 0.55xd0 = 1 mm is additionally included. The diagrams show strain
distributions to a depth of z = 1.5 mm, in order to show the change of the slope a few
drilling increments before the through-hole is introduced in the models with T = 1.5 mm
67

and T = 1mm. It can be seen that the strain distributions recorded with both strain gages
for the calibration with the model fT3D20m with T = 1.67xd0 = 3 mm are almost equal
to those from the reference model. Beginning with a thickness of T = 1.11xd0 = 2 mm, a
reduction of the thickness leads to a change in the absolute values of the strain
distributions compared to the reference. In the case of strain gage no. 3, the absolute
strain values become higher with decreasing thickness. All strain distributions range in
the same level. In the transverse case of strain gage no. 1, the absolute strain values
decrease slightly with decreasing thickness. A few depth increments before the blind-
hole is transformed into a through-hole the slopes of the strain distributions of the
models with the thickness T = 0.55xd0 = 1 mm and T = 0.83xd0 = 1.5 mm change and
the values converge roughly to that of the reference calibration.
0
Simulation of Type 1 specimens
with variable thickness
-50
Strain gage no. 3
σc,x = 160 MPa
d0, sim = 1.8 mm
-100
Δε [μm/m]

T = 1.11d0 = 2 mm
-150
T = 1.67d0 = 3 mm

Reference
T = 0.83d0 = 1.5 mm
T = 3.33d0 = 6 mm
-200

T = 0.55d0 = 1 mm

-250
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6
z [mm]

Fig.4.7: Influence of geometrical parameter T on numerical calibration strain depth distribution


(strain gage no. 3)
68

120
Simulation of Type 1 specimens
with variable thickness
100 Strain gage no. 1
σc,x = 160 MPa
d0, sim = 1.8 mm
80
Δε [μm/m]

60
Reference
T = 3.33d0 = 6 mm
40
T = 1.67d0 = 3 mm
T = 1.11d0 = 2 mm
20
T = 0.83d0 = 1.5 mm
T = 0.55d0 = 1 mm
0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6
z [mm]

Fig.4.8: Influence of geometrical parameter T on numerical calibration strain depth distribution


(strain gage no. 1)

4.4 Variation of Distance Hole-Edge - D


The results for the experimental and numerical calibration for the flat specimens with a
distance hole-edge of 10 mm, 5 mm and 2 mm and T = 6 mm = const are shown in
Fig. 4.9, Fig. 4.10 and Fig. 4.11, respectively. Similar to the results for the variation of
the thickness, the numerically calculated strain distributions in this calibration case
reproduce well the experimentally measured depth distributions of relieved strain
(s. Tab. 4.5, Tab. 4.6 and Tab. 4.7 for specimen fT6D10, fT6D5 and fT6D2,
respectively).
100

sg 1
50

0
Δε [μm/m]

-50
sg 2
Experiment •
-100 Simulation ___

-150
fT6D10 (Type 1)
σc,x = 160 MPa
d0,exp = 1.90 mm sg 3
-200
d0,sim = 1.90 mm

-250
0.0 0.2 0.4 0.6 0.8 1.0 1.2
z [mm]

Fig.4.9: Comparison of experimental and numerical calibration strain depth distributions for
specimen fT6D10 (T =6 mm, D = 10 mm, R = ∞)
69

z ε1 [μm/m] ε3 [μm/m]
[mm] experiment simulation difference experiment simulation difference
0.2 6 8 -2 -32 -41 9
0.5 24 25 -1 -115 -121 6
1.0 59 57 2 -216 -215 -1
Tab. 4.5: Comparison of experimental and numerical calibration strain values in selected hole
depths for specimen fT6D10 (T =6 mm, D = 10 mm, R = ∞)
100

sg 1
50

0
Δε [μm/m]

-50
sg 2
Experiment •
-100 Simulation ___

-150 fT6D5 (Type 1)


σc,x = 160 MPa
d0,exp = 1.85 mm sg 3
-200
d0,sim = 1.85 mm

-250
0.0 0.2 0.4 0.6 0.8 1.0 1.2
z [mm]

Fig.4.10: Comparison of experimental and numerical calibration strain depth distributions for
specimen fT6D5 (T =6 mm, D = 5 mm, R = ∞)

z ε1 [μm/m] ε3 [μm/m]
[mm] experiment simulation difference experiment simulation difference
0.2 5 7 -2 -32 -39 7
0.5 23 24 -1 -108 -114 6
1.0 57 55 2 -204 -204 0
Tab. 4.6: Comparison of experimental and numerical calibration strain values in selected hole
depths for specimen fT6D5 (T =6 mm, D = 5 mm, R = ∞)
70

100

sg 1
50

0
Δε [μm/m]

-50
sg 2
Experiment •
-100 Simulation ___

-150 fT6D2 (Type 1)


σc,x = 160 MPa
d0,exp = 1.90 mm sg 3
-200
d0,sim = 1.90 mm

-250
0.0 0.2 0.4 0.6 0.8 1.0 1.2
z [mm]

Fig.4.11: Comparison of experimental and numerical calibration strain depth distributions for
specimen fT6D2 (T =6 mm, D = 2 mm, R = ∞)

z ε1 [μm/m] ε3 [μm/m]
[mm] experiment simulation difference experiment simulation difference
0.2 5 8 -3 -31 -40 9
0.5 21 26 -5 -109 -118 9
1.0 56 59 -3 -213 -207 -6
Tab. 4.7: Comparison of experimental and numerical calibration strain values in selected hole
depths for specimen fT6D2 (T =6 mm, D = 2 mm, R = ∞)

In contrast to the variation of thickness, the variation of the distance hole-edge D


appears to have only little influence on the calibration strains as can be seen in Fig. 4.12
for strain gage no. 3 and in Fig 4.13 for strain gage no. 1. In both diagrams, only the
distribution for the model with D = 2 mm shows a slight deviation from the reference
distribution, i.e. the absolute strain values of the calibration with model fT6D2m
decrease for strain gage no. 3 and increase for strain gage no. 1. The distributions for
model fT6D10m with D = 10 mm and model fT6D5m with D = 5 mm are superimposed
on those from the reference model. All calibration curves show here a similar trend in
the strain distribution.
71

0.0
Simulation of Type 1 specimens
with variable distance hole-edge
Strain gage no. 3
-50.0 σc,x = 160 MPa
d0, sim = 1.8 mm
Δε [μm/m]

-100.0

Reference
D = 11.11d0 = 20 mm
-150.0
D = 1.1d0 = 2 mm

D = 1.94 … 5.56d0 = 3.5 … 10 mm

-200.0
0 0.2 0.4 0.6 0.8 1 1.2
z [mm]

Fig.4.12: Influence of geometrical parameter D on numerical calibration strain depth distribution


(strain gage no. 3)
60.0
Simulation of Type 1 specimens D = 1.1d0 = 2 mm
with variable distance hole-edge
Strain gage no. 1
σc,x = 160 MPa
40.0 d0, sim = 1.8 mm
Δε [μm/m]

D = 1.94 … 5.56d0 = 3.5 … 10 mm

20.0

Reference
D = 11.11d0 = 20 mm

0.0
0 0.2 0.4 0.6 0.8 1 1.2
z [mm]

Fig.4.13: Influence of geometrical parameter D on numerical calibration strain depth distribution


(strain gage no. 1)

4.5 Combination of Parameter T and D


Fig. 4.14 compares the experimental and numerical results of the calibration for
specimen fT1.5D2, which is considered to be thin with T = 1.5 mm and at the same
time, which has a small distance hole-edge with D = 2 mm.
72

100

sg 1
50

0
Δε [μm/m]

-50
sg 2
-100 Experiment ·
___
Simulation
-150
fT1.5D2 (Type 1)
σc,x = 160 MPa sg 3
-200 d0,exp = 1.90 mm
d0,sim = 1.90 mm

-250
0.0 0.2 0.4 0.6 0.8 1.0 1.2
z [mm]

Fig.4.14: Comparison of experimental and numerical calibration strain depth distributions for
specimen fT1.5D2 (T =1.5 mm, D = 2 mm, R = ∞)

Both experimental and numerical calibration strain distributions show accordance,


although the absolute strain differences between experiment and simulation for strain
gage no. 3 are higher than the previous comparisons in Chap. 4.3 and Chap. 4.4. The
absolute values as well as the difference are listed for selected hole depths in Tab. 4.8.

z ε1 [μm/m] ε3 [μm/m]
[mm] experiment simulation difference experiment simulation difference
0.2 5 4 1 -46 -41 -5
0.5 17 15 2 -137 -122 -15
1.0 54 50 4 -232 -209 -23
Tab. 4.8: Comparison of experimental and numerical calibration strain values in selected hole
depths for specimen fT1.5D2 (T =1.5 mm, D = 2 mm, R = ∞)

The diagrams in Fig. 4.15 (strain gage no. 3) and Fig. 4.16 (strain gage no. 1) display
the numerical results for previous calibrations of flat models with different
combinations of parameters T and D. The results presented in the diagrams are
calculated from the reference model fT6D20m, the “thin” model fT1.5D20m, the model
with a “small” distance D fT6D2m and the “combined” flat model fT1.5D2m. The
strain distributions are plotted to a hole depth of z = 1.5 mm. In this comparison, the
geometrical parameter thickness has a major influence on the distribution of the
calibration strain. In both diagrams the magnitudes of the relieved strains for the “thin”
73

models differ from those for the reference model with higher absolute values. Also the
trajectories of the distributions of both “thin” models are more similar than those of the
“thick” models.
0
Simulation of Type 1 specimens
Strain gage no. 3
-50 σc,x = 160 MPa
d0, sim = 1.8 mm

-100
Δε [μm/m]

fT6D2
-150
Reference
fT6D20

-200
fT1.5D2

fT1.5D20
-250
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6
z [mm]

Fig.4.15: Comparison of numerical calibration strain depth distributions (strain gage no. 3) for
flat specimens with different combinations of geometrical parameters T and D

120.0
Simulation of Type 1 specimens fT1.5D2
Strain gage no. 1
100.0
σc,x = 160 MPa fT1.5D20
d0, sim = 1.8 mm
80.0
Δε [μm/m]

fT6D2
60.0
Reference
fT6D20
40.0

20.0

0.0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6
z [mm]

Fig.4.16: Comparison of numerical calibration strain depth distributions (strain gage no. 1) for
flat specimens with different combinations of geometrical parameters T and D

When comparing the two “thin” models fTD1.5D20m and fT1.5D2m, a reduction of
parameter D from D = 20 mm to D = 2 mm causes also (s. Chap. 4.4) a decrease of the
absolute strain values as registered with strain gage no. 3. For strain gage no. 1, the
74

calibration with the “thin” model with D = 2 mm causes initially smaller absolute strain
values and after a hole depth of z ≈ 1.12 mm they became slightly bigger.

4.6 Variation of Cylinder Radius - R


4.6.1 Solid Cylinder
The results for the experimental and numerical calibration for the solid cylinder
specimens with a cylinder radius of 6 mm, 4 mm and 3 mm are shown in Fig. 4.17,
Fig. 4.18 and Fig. 4.19, respectively. The numerically calculated strain distributions in
this calibration case reproduce sufficiently the experimentally measured depth
distributions of relieved strain as shown in Tab. 4.9, Tab. 4.10 and Tab. 4.11 for the
mentioned specimens. The experimental results at the first drilling increments scatter
until a hole depth of z ≈ 0.1 mm.
100

sg 1
50

0
Δε [μm/m]

-50

Experiment • sg 2
-100 Simulation ___

-150 scR6 (Type 2)


σc,x = 160 MPa
d0,exp = 1.83 mm
-200
d0,sim = 1.80 mm sg 3
-250
0.0 0.2 0.4 0.6 0.8 1.0 1.2
z [mm]

Fig.4.17: Comparison of experimental and numerical calibration strain depth distributions for
specimen scR6 (R = 6 mm)

z ε1 [μm/m] ε3 [μm/m]
[mm] experiment simulation difference experiment simulation difference
0.2 7 8 -1 -26 -37 11
0.5 27 29 -2 -104 -115 11
1.0 71 67 4 -218 -210 -8
Tab. 4.9: Comparison of experimental and numerical calibration strain values in selected hole
depths for specimen scR6 (R = 6 mm)
75

100

sg 1
50

0
Δε [μm/m]
-50

Experiment • sg 2
-100 Simulation ___

-150 scR4 (Type 2)


σc,x = 160 MPa
d0,exp = 1.82 mm
-200
d0,sim = 1.80 mm sg 3
-250
0.0 0.2 0.4 0.6 0.8 1.0 1.2
z [mm]

Fig.4.18: Comparison of experimental and numerical calibration strain depth distributions for
specimen scR4 (R = 4 mm)

z ε1 [μm/m] ε3 [μm/m]
[mm] experiment simulation difference experiment simulation difference
0.2 7 7 0 -17 -36 19
0.5 25 27 -2 -91 -114 23
1.0 70 66 4 -207 -211 4
Tab. 4.10: Comparison of experimental and numerical calibration strain values in selected hole
depths for specimen scR4 (R = 4 mm)

100

sg 1
50

0
Δε [μm/m]

-50

Experiment • sg 2
-100 Simulation ___

-150 scR3 (Type 2)


σc,x = 160 MPa
d0,exp = 1.80
-200
d0,sim = 1.80 mm sg 3
-250
0.0 0.2 0.4 0.6 0.8 1.0 1.2
z [mm]

Fig.4.19: Comparison of experimental and numerical calibration strain depth distributions for
specimen scR3 (R = 3 mm)
76

z ε1 [μm/m] ε3 [μm/m]
[mm] experiment simulation difference experiment simulation difference
0.2 12 5 7 -30 -32 2
0.5 33 20 13 -109 -109 0
1.0 70 54 16 -210 -208 -2
Tab. 4.11: Comparison of experimental and numerical calibration strain values in selected hole
depths for specimen scR3 (R = 3 mm)

0.00
Simulation of Type 2 specimens
with variable solid cylinder radius
Strain gage no. 3
-50.00
σc,x = 160 MPa
d0, sim = 1.8 mm

-100.00
Δε [μm/m]

-150.00
Reference
R=∞

-200.00

R = 1.67 … 3.33 d0 = 3 … 6 mm

-250.00
0 0.2 0.4 0.6 0.8 1 1.2
z [mm]

Fig.4.20: Influence of geometrical parameter R on numerical calibration strain depth distribution


(strain gage no. 3)
80.00
Simulation of Type 2 specimens
70.00 with variable solid cylinder radius
Strain gage no. 1
60.00 σc,x = 160 MPa
d0, sim = 1.8 mm
50.00
Δε [μm/m]

Reference
R = 2.22 … 3.33 d0 = 4 … 6 mm R=∞
40.00

30.00
R = 1.67d0 = 3 mm
20.00

10.00

0.00
0 0.2 0.4 0.6 0.8 1 1.2
z [mm]

Fig.4.21: Influence of geometrical parameter R on numerical calibration strain depth distribution


(strain gage no. 1)
77

The influence of the geometrical parameter R for the solid cylinder on the calibration
strains is demonstrated in Fig. 4.20 (strain gage no. 3) and Fig. 4.21 (strain gage no. 1).
It can be seen that the strain distributions in load direction (strain gage no. 3) obtained
from calibration with the different solid cylinders are very close to each other. The
absolute strain values are higher compared to the results of the reference model. In the
case of the strains recorded with strain gage no. 1, the distributions for the models
scR6m and scR4m, respectively, are close together and have higher absolute strain
values compared to the distributions of the reference specimen. The distribution for
strain gage no. 1 from the calibration with the model scR3m has lower absolute strain
values than the results of the models with R =4 mm and R = 6 mm.

4.6.2 Hollow Cylinders


Fig. 4.22, Fig. 4.23 and Fig. 4.24 compare the experimental and numerical results of the
calibration for the hollow cylindrical specimens hcR6, hcR4 and hcR3, respectively. In
addition to the change of cylinder diameter, these specimens consider also a wall
thickness of T = 1.25 mm = const. On the whole, the numerical results match
adequately to the experimental results (s. Tab. 4.12, Tab. 4.13 and Tab. 4.14 for
absolute strain values and differences between experiment and simulations of
calibrations using specimen hcR6, hcR4 and hcR3, respectively).

100

sg 1
50

-50
Δε [μm/m]

-100 Experiment • sg 2
Simulation ___
-150
hcR6 (Type 2)
-200 σc,x = 160 MPa
d0,exp = 1.86 mm
-250 d0,sim = 1.80 mm sg 3
-300
0.0 0.2 0.4 0.6 0.8 1.0 1.2
z [mm]

Fig.4.22: Comparison of experimental and numerical calibration strain depth distributions for
specimen hcR6 (T = 1.25 mm, R = 6 mm)
78

z ε1 [μm/m] ε3 [μm/m]
[mm] experiment simulation difference experiment simulation difference
0.2 7 78 -1.0 -49 -47 -2
0.5 22 25 -3 -161 -144 -17
1.0 66 67 -1 -264 -238 -26
Tab. 4.12: Comparison of experimental and numerical calibration strain values in selected hole
depths for specimen hcR6 (T = 1.25 mm, R = 6 mm)

100

sg 1
50

0
Δε [μm/m]

-50

Experiment • sg 2
-100
Simulation ___

-150 hcR4 (Type 2)


σc,x = 160 MPa
-200 d0,exp = 1.84 mm
d0,sim = 1.80 mm sg 3
-250
0.0 0.2 0.4 0.6 0.8 1.0 1.2
z [mm]

Fig.4.23: Comparison of experimental and numerical calibration strain depth distributions for
specimen hcR4 (T = 1.25 mm, R = 4 mm)

z ε1 [μm/m] ε3 [μm/m]
[mm] experiment simulation difference experiment simulation difference
0.2 7 5 2 -32 -44 12
0.5 18 19 -1 -125 -138 13
1.0 54 56 -2 -238 -237 -1
Tab. 4.13: Comparison of experimental and numerical calibration strain values in selected hole
depths for specimen hcR4 (T = 1.25 mm, R = 4 mm)
79

100

sg 1
50

0
Δε [μm/m]
-50

Experiment • sg 2
-100 Simulation ___

-150 hcR3 (Type 2)


σc,x = 160 MPa
d0,exp = 1.86 mm
-200
d0,sim = 1.80 mm sg 3
-250
0.0 0.2 0.4 0.6 0.8 1.0 1.2
z [mm]

Fig.4.24: Comparison of experimental and numerical calibration strain depth distributions for
specimen hcR3 (T = 1.25 mm, R = 3 mm)

z ε1 [μm/m] ε3 [μm/m]
[mm] experiment simulation difference experiment simulation difference
0.2 3 3 0 -39 -38 1
0.5 14 12 2 -126 -130 4
1.0 51 41 10 -228 -234 6
Tab. 4.14: Comparison of experimental and numerical calibration strain values in selected hole
depths for specimen hcR3 (T = 1.25 mm, R = 3 mm)

0
Simulation of Type 2 specimens
with variable hollow cylinder radius
-50
Strain gage no. 3
σc,x = 160 MPa
d0,sim = 1.8 mm
-100
Δε [μm/m]

-150
Reference
R=∞
R = 1.67d0 = 3 mm
-200
R = 2.22d0 = 4 mm

R = 3.33d0 = 6 mm
-250
0 0.2 0.4 0.6 0.8 1 1.2
z [mm]

Fig.4.25: Influence of geometrical parameter R and T on numerical calibration strain depth


distribution (strain gage no. 3)
80

80
Simulation of Type 2 specimens R = 3.33d0 = 6 mm
70 with variable hollow cylinder radius
Strain gage no. 1
60 σc,x = 160 MPa R = 2.22d0 = 4 mm
d0, sim = 1.8 mm
50
Δε [μm/m]

Reference
40 R=∞
R = 1.67d0 = 3 mm
30

20

10

0
0 0.2 0.4 0.6 0.8 1 1.2
z [mm]

Fig.4.26: Influence of geometrical parameter R and T on numerical calibration strain depth


distribution (strain gage no. 1)

The diagram in Fig. 4.25 compares (for strain gage no. 3) the strain distributions of the
reference model with those of the different hollow cylindrical models. It can be seen
that the distributions of the different hollow cylinder calibration are sufficiently close to
each other and of higher absolute strain values than the reference.

In the case of strain gage no. 1 (Fig. 4.26), the numerical strain distributions vary with
varying hollow cylinder radius. The strain curve of the calibration with a hollow
cylinder radius of R = 6 mm is above the reference curve. The strain curves of the
hollow cylinder calibration fall with decreasing hollow cylinder radius.

4.6.3 Comparison of Solid Cylinders and Hollow Cylinders


The following results summarize the previous results of Chap. 4.6.1 and Chap. 4.6.2.
This chapter is included in order to compare the calibration results of the solid cylinders
with the calibrations results of the hollow cylinders, in which a specimen’s wall
thickness of T = 1.25 mm is additionally considered. The diagram in Fig. 4.27 shows
distributions registered with strain gage no. 3. The solid curves represent the calibration
strain of the solid cylinders whereas the dashed lines represent those of the hollow
cylinders. The dotted curve is the strain distribution of the reference model.
81

0
Simulation of Type 2 specimens
comparison sc vs. hc
-50
Strain gage no. 3
σc,x = 160 MPa
d0,sim = 1.8 mm
-100
Δε [μm/m]

-150
Reference
R=∞

-200
hcR3 scR3
scR4
hcR4 scR6
hcR6
-250
0 0.2 0.4 0.6 0.8 1 1.2
z [mm]

Fig.4.27: Comparison of numerical calibration strain depth distributions between solid cylinders
and hollow cylinders (strain gage no. 3)

One can see that the change of the cross section from a solid cylinder to a hollow
cylinder causes an increase of the absolute strain values for the strain gage no. 3, which
is parallel to the load direction.
70
Simulation of Type 2 specimens scR6
comparison sc vs. hc scR4
60
Strain gage no. 1 hcR6
σc,x = 160 MPa scR3
50 d0, sim = 1.8 mm
hcR4
Δε [μm/m]

40
hcR3
Reference
30

20

10

0
0 0.2 0.4 0.6 0.8 1 1.2
z [mm]

Fig.4.28: Comparison of numerical calibration strain depth distributions between solid cylinders
and hollow cylinders (strain gage no. 1)

Fig. 4.28 shows the perpendicular case for strain gage no. 3 and has the same
representation of curves as in the previous figure. It is visible that the strain distributions
of the calibrations with the hollow cylinders are lower (for hcR6m slightly lower than
82

scR6m) and they disperse clearly from each other compared with the distributions with
the solid cylinders.

4.7 Specimens with Center Through-Hole - Combined Geometrical


Parameters
This chapter presents the calibration results of the specimens of Type 3 and Type 4. The
procedure of experimental calibration is described in Chap. 3.2.3 and Chap. 3.2.4. Due
to the stress concentration caused by the center through-hole, which is present in both
specimen types, plastic deformation may occur near the through-hole. In order to take
into account the plastic deformations, the numerical calibration of the specimens of
Type 3 and Type 4 uses an elastic-plastic material model with isotropic hardening
instead of the linear-elastic material model of the previous examples. The procedure of
loading in these numerical calibrations is similar to the experimental one, i.e. the
simulation considers the different load steps cycles of 50 MPa, 210 MPa and 0 MPa.
These loads are applied as external uniform loads (s. Fig. 3.13). Hence, the relieved
strains of numerical calibrations are calculated using {3.2}.

4.7.1 Flat Tensile Specimens with a ø10 mm center through-hole


The result of the experimental calibration as well as that of the numerical calibration of
the specimens of Type 3 is shown in Fig. 4.29 and Fig. 4.30 for the specimen fhT6D2
and fhT1.5D2, respectively. The experimental and numerical strain distributions are in
accordance to each other. Tab. 4.15 and Tab. 4.16 listed the absolute strain values of
experiment and simulation. The differences between both calibration methods are high,
especially for strain gage no. 3, compared to the flat specimens without center through-
hole. The resulting calibration loads were calculated from the model as an average value
over the hole volume. Thus, the calibration stresses are σc,x = 251.5 MPa,
σc,y =61.4 MPa for model fhT6D2m (s. Fig. 11.1 and Fig. 11.2 in appendix Chap. 11)
and σc,x = 261.9 MPa, σc,y =60.2 MPa for model fhT1.5D2m (s. Fig. 11.3 and Fig. 11.4
in appendix Chap. 11).
83

50

sg 1
0

-50

-100
Δε [μm/m]

-150 Experiment • sg 2
Simulation ___
-200
fhT6D2 (Type 3)
-250 σx = 160 MPa
d0,exp = 1.83 mm
-300 d0,sim = 1.80 mm sg 3
-350
0.0 0.2 0.4 0.6 0.8 1.0 1.2
z [mm]

Fig.4.29: Comparison of experimental and numerical calibration strain depth distributions for
specimen fhT6D2 (T =6 mm, D = 2 mm, R = ∞)

z ε1 [μm/m] ε3 [μm/m]
[mm] experiment simulation difference experiment simulation difference
0.2 -3 -1 -2 -28 -49 21
0.5 -8 -2 -6 -159 -157 -2
1.0 4 13 -9 -323 -288 -35
Tab. 4.15: Comparison of experimental and numerical calibration strain values in selected hole
depths for specimen fhT6D2 (T = 6 mm, D = 2 mm, R = ∞)

50

0
sg 1
-50

-100
Δε [μm/m]

-150
Experiment •
Simulation ___ sg 2
-200

-250 fhT1.5D2 (Type 3)


σx = 160 MPa
-300 d0,exp = 1.87 mm
d0,sim = 1.90 mm
-350
sg 3
-400
0.0 0.2 0.4 0.6 0.8 1.0 1.2
z [mm]

Fig.4.30: Comparison of experimental and numerical calibration strain depth distributions for
specimen fhT1.5D2 (T =1.5 mm, D = 2 mm, R = ∞)
84

z ε1 [μm/m] ε3 [μm/m]
[mm] experiment simulation difference experiment simulation difference
0.2 -9 -9 0 -66 -60 -6
0.5 -23 -21 -2 -212 -190 -22
1.0 0 1 -1 -371 -335 -36
Tab. 4.16: Comparison of experimental and numerical calibration strain values in selected hole
depths for specimen fhT1.5D2 (T = 1.5 mm, D = 2 mm, R = ∞)

The strain distributions of the flat specimens with a center through-hole are plotted in
one diagram together with the strain distributions of the flat reference specimen in order
to see the influence of the introduced ø10 mm hole on the calibration results. Fig. 4.31
compares the numerical strain distribution calculated with strain gage no. 3 and
Fig. 4.32 the distributions with strain gage no. 1. Compared to the reference calibration,
the results of the calibration with the specimens of Type 3 differ in the strain magnitude
and in the trajectory of the curves. The absolute strain values of the specimens with the
center through-hole are clearly higher than the reference for strain gage no. 3 and
clearly lower for strain gage no. 1.
0
Simulation of Type 3 specimens
-50 with T = 1.5 mm and T = 6 mm
Strain gage no. 3
-100 σx = 160 MPa
d0,sim = 1.8 mm
-150
Δε [μm/m]

Reference
fT6D20
-200

-250
fhT6D2

-300
fhT1.5D2
-350

-400
0.00 0.20 0.40 0.60 0.80 1.00 1.20 1.40 1.60
z [mm]

Fig.4.31: Comparison of numerical calibration strain depth distributions for flat models with
ø10 mm center through-hole and different thickness T (strain gage no. 3)
85

110
Simulation of Type 3 specimens
90 with T = 1.5 mm and T = 6 mm
Strain gage no. 1 Reference
σx = 160 MPa fT6D20
70
d0,sim = 1.8 mm
Δε [μm/m]
50

30

fhT6D2
10

fhT1.5D2
-10

-30
0.00 0.20 0.40 0.60 0.80 1.00 1.20 1.40 1.60
z [mm]

Fig.4.32: Comparison of numerical calibration strain depth distributions for flat models with
ø10 mm center through-hole and different thickness T (strain gage no. 1)

The distributions of the calibration strains with the “thin” flat model with hole
fhT1.5D2m run similar to the strain distributions of the “thin” flat model without hole
fT1.5D20m compared to its “thick” model counterparts fhT6D2m and fTD6D20m. First
of all, the strain values of the “thin” models are higher for strain gage no. 3 and lower
for strain gage no. 1. Secondly, the slopes of the calibration curves of the “thin” models
change with increasing depth to a greater extent than the calibration curves of the
“thick” models.

4.7.2 Hollow Cylindrical Tensile Specimens with a ø6 mm Center Through-Hole


The hollow cylindrical tensile specimens with an ø6 mm center through-hole combine
all three geometrical parameters which are considered in this investigation about the
geometrical influence on the results of a hole-drilling measurement. In Fig. 4.33, the
experimental strain distributions of the calibration on specimen hchR6 are confronted
with the numerical ones. The numerical strain distribution for strain gage no. 1
reproduces enough the experimental distributions. In case of strain gage no. 3 the
trajectory of the experimental distribution is qualitatively reproduced by the simulation
with clearly lower absolute strain values. The absolute strain values as well as the
difference between the experiment and simulation are listed in Tab. 4.17. One can see
that the calculated differences in this case are the highest of all calibrations shown in
these results chapter (Chap. 4).
86

100
sg 1

-100
Δε [μm/m]

sg 2
Experiment •
-200
Simulation ___

hchR6 (Type 4)
σx = 160 MPa
-300
d0,exp = 1.76 mm
sg 3
d0,sim = 1.80 mm

-400
0.0 0.2 0.4 0.6 0.8 1.0 1.2
z [mm]

Fig.4.33: Comparison of experimental and numerical calibration strain depth distributions for
specimen hchR6 (T =1.25 mm, D = 2 mm, R = 6 mm)

z ε1 [μm/m] ε3 [μm/m]
[mm] experiment simulation difference experiment simulation difference
0.2 8 7 1 -38 -62 24
0.5 17 19 -2 -187 -186 23
1.0 72 62 10 -356 -308 -48
Tab. 4.17: Comparison of experimental and numerical calibration strain values in selected hole
depths for specimen hchR6 (T = 1.25 mm, D = 2 mm, R = 6 mm)

350
hchR6 (Type 4) σc,x
σx = 160 MPa
300
d0 = 1.8 mm

250

200
σ [MPa]

150

100

σc,y
50

0
0 0.2 0.4 0.6 0.8 1 1.2
z [mm]

Fig.4.34: Stress distribution versus depth within hole volume before introducing the hole for
numerical calibration of model hchR6m
87

The resulting calibration stresses σc,x and σc,y over the hole volume before drilling are
plotted in Fig. 4.34. It can be see that the calibration stresses in the hole area are non-
uniform in depth (s. also Fig. 11.5 and Fig. 11.6 in appendix Chap. 11).

0.00
Simulation of Type 4 specimens
with T = 1.25 mm, D = 2 mm and R = 6 mm
-50.00 Strain gage no. 3
σx = 160 MPa
d0, sim = 1.8 mm
-100.00
Δε [μm/m]

-150.00
Reference
fT6D20
-200.00
hcR6

-250.00

-300.00
hchR6

-350.00
0 0.2 0.4 0.6 0.8 1 1.2
z [mm]

Fig.4.35: Comparison of numerical calibration strain depth distributions for the hollow
cylindrical model with ø6 mm center through-hole with reference distributions (strain
gage no. 3)
70.0
Simulation of Type 4 specimens hcR6 hchR6
60.0 with T = 1.25 mm, D = 2 mm and R = 6 mm
Strain gage no. 1
50.0 σx = 160 MPa
d0, sim = 1.8 mm
40.0
Reference
Δε [μm/m]

fT6D20
30.0

20.0

10.0

0.0

-10.0
0 0.2 0.4 0.6 0.8 1 1.2
z [mm]

Fig.4.36: Comparison of numerical calibration strain depth distributions for the hollow
cylindrical model with ø6 mm center through-hole with reference distributions (strain
gage no. 1)

Fig. 4.35 (sg. no. 3) and Fig. 4.36 (sg. no. 1) illustrate the difference of the calibration
using the simulation of the hollow cylinder with the ø6 mm center through-hole
88

hchR6m compared to the reference calibration with the flat model and, additionally, the
calibration for the hollow cylinder without the center through-hole hcR6m.

The calibration strain distributions of the hchR6m model are different from the
reference and the hollow cylinder model hcR6m. In the case of strain gage no. 3, the
absolute strain values are clearly higher than the reference and the hcR6m distributions.
For strain gage no. 1, the absolute strain values are lower than the two comparative
distributions.
89

5 Discussion of Calibration Results


5.1 Prefacing Remarks
In this chapter the results of the experimental calibration as well as the numerical
calibration for the hole-drilling method are discussed by pointing out two main topics.
The first topic deals with the question of whether the numerical calibration reproduces
sufficiently the experimental strain distributions. The second topic concerns the
influence of the component geometry on the results of a hole-drilling measurement:
Questions that arise are e.g. in which cases the single geometrical limits listed in [2]
(Tab. 2.4 of Chap.2.5) are sufficient or may be reduced and in which cases a geometry
specific calculation could be an alternative. For the purpose of the discussion of these
topics, this chapter is divided into four parts:

The first part deals with the uncertainty of the calibration results shown in Chap. 4. For
this reason the uncertainty of the numerical calibration as well as the uncertainty of the
experimental calibration are discussed and visualized for some selected strain
distributions.

The second part of this discussion tends towards introducing the problematic of the
stress calculation using commercially available evaluation software. The input data for
this evaluation is selected from calibration results determined by using specimens with
non-reference geometry and the purpose is to show in general the possible stress
evaluation error if the boundary conditions are violated.

The third part deals with the individual contribution of the investigated geometry
parameters on the obtained results. This is done by calculating the strain and stress
deviations or differences between distributions established with non-reference geometry
calibrations and those established with the reference calibration. As a result, a possible
recommendation is analyzed concerning the minimization of the geometry limit in the
case of only one geometrical boundary condition being violated.

The last part consists of demonstrating the possible reduction of the stress calculation
error. For this purpose, hole-drilling measurements are simulated using models of
90

selected non-reference specimens and evaluating them with a simplified MPA algorithm
in which the geometry specific calibration functions are included.

5.2 Calibration Uncertainty


5.2.1 Uncertainty in Numerical Calibration
The quality of a finite element model depends mainly on the type of elements used, the
used finite element formulation, the mesh refinement, the boundary conditions and the
material model. In the special case of the hole-drilling method the quality of the
corresponding model itself can be proved as a first approximation comparing selected
results with already existing theoretical solutions. In the best case, the numerical results
should converge to the theoretical ones.

160
Analytical solution _______ strain gage position
Simulation fT1.5D20 • ▲ σrad
120

80
σx [MPa]

40

σtan
0
0.5 x d0

-40
0 1 2 3 4 5
distance x from hole center [mm]

Fig. 5.1: Stress distribution near a through-hole: Comparison of results between a numerical
calibration using model fT1.5D20 and analytical solution according to Kirsch
91

In case of the of the flat specimen models of Type 1 Kirsch’s analytical solution [16] for
the stress state near a through-hole in a thin and infinitely wide plate may be used for
the comparison. This procedure was used for finding an appropriate meshing of the
models of Type 1. Models with different number of elements, which have a small
thickness and a sufficiently wide broadness, were generated in order to obtain the stress
values near the simulated through-hole. The results of the simulations were compared to
each other in order to select one model which reproduces the analytical solution
adequately and has an acceptable calculation time. The result of the selected model,
which was the base of the used flat specimen models in this investigation, is shown in
Fig. 5.1. The upper part of Fig. 5.1 shows schematically the loading conditions of the
flat model, the mesh near the hole and the transition area between the hole mesh and the
rectangular mesh of the rest of the model with the approximate location of the strain
gage. The lower diagram shows the stress distribution radial σrad and tangential σtan to
hole against the x-axis, which is parallel to the axis of strain gage no. 3 as can be seen in
the picture above the diagram. The stress values are the average values extrapolated to
the nodes of the elements along the x-axis. Clear deviations of max. 30 %-46 % are
visible at the transition area near the hole (x = 0.9 … 1.26 mm) for the radial stress
distribution σrad. It should be noticed that these high percental deviations arise due to the
fact that the absolute stress values are very small and each small difference between
them leads to a relatively high percental deviation. In the strain gage location
(x = 1.78 … 3.35 mm) the differences are about 0.3-2.0 %, which can be considered as
a good approximation of Kirsch’s solution. A similar observation is done with the
distribution of the tangential stress σtan. The reason for the high stress deviation values
near the hole could be the low mesh refinement at the transition area between the hole
mesh and the rectangular mesh. The meshing of this transition area was necessary in
order to design the further rectangular mesh which allows the generation of the models
with different distance hole-edge D. This transition between the two different meshing
shapes could be too sharp, i.e. the few elements at this area do not sufficiently
reproduce Kirsch’s solution. Other sources for possible inhomogeneous stress
distributions near the hole were not detected with the finite element post-processor.

The flat specimens of Type 3 were generated using a mesh design similar to that of the
Type 1 flat specimens. Therefore, the numerical uncertainty should be in the range as
92

the one from the models of Type 1. A quantitative statement about the quality of these
models can be made by calculating the stress concentration factors at the edge of the
ø10 mm center through-hole of these flat models and comparing them with theoretical
values e.g. in [51]. The theoretical value (Ktg = 3.23) differs 8.8 % from that obtained
with the fhT6D2m model and 2.9 % from that obtained with the fhT1.5D2m model.

3.25
theoretical value

3.2

3.15

fhT1.5D2m
Ktg [-]

3.1

3.05

fhT6D2m
2.95
0 1 2 3 4 5 6
Thickness [mm]

Fig. 5.2: Stress concentration factor at the edge of a ø10 mm through-hole as a function of the
specimen thickness (according to [51])

Fig. 5.2 shows the distribution of the stress concentration factor Ktg against the
thickness for specimens with a ø10 mm center through-hole and different thickness. The
value of Ktg increases with decreasing thickness. The theoretical value, similar to
Kirsch’s solution, does not consider the specimen thickness and is plotted at T = 0 mm.
Nevertheless, the Ktg values established with the finite element simulations converge
with decreasing thickness to the theoretical value. This thickness dependency of the
stress concentration factor is described in [51]. Thus, a stress variation in the thickness
direction of a plate exists, as thick plates show lower maximum tangential stresses at the
surface and higher tangential stresses at the midplane than the values determined with
thin sheets. This observation can also be made in this investigation (s. also Fig. 11.1 and
Fig. 11.3 in appendix chapter Chap. 11). The surface stress at the edge of the ø10 mm
center through-hole is 476 MPa for the thick model fhT6D2m and 502 MPA for the thin
model fhT1.5D2m. In contrast, the stress at the midplane is 526 MPa and 514 MPa for
93

models fhT6D2m and fhT1.5D2m, respectively. For this reason, it can be assumed, that
the solutions established with the models of Type 3 reproduce the theoretical solution.

In the case of the cylindrical specimens of Type 2 the models were generated only once,
i.e. there was no selection of models with different number of elements. The quality of
each cylindrical model is verified by comparing the stress concentration factor near the
hole with the stress concentration factor taken from literature. Tab. 5.1 compares the
maximal stress σmax,sim at the edge of the hole (hole-drilling measurement point)
established with the hollow cylinder models hcR6m, hcR4m and hcR3m with the
maximal stress σmax calculated using the stress concentration factor Ktn according to
[51] and gives the deviation in percent. The geometrical parameters which are
considered for the determination of Ktn are also listed in Tab. 5.1 and shown in Fig. 5.3.

σmax,sim σmax = Ktn x σnom Deviation Ktn σnom 2R 2RI d0


Model
[MPa] [MPa] [%] [-] [MPa] [mm] [mm] [mm]
hcR6m 453 464 2.4 2.90 160 12 9.5 1.8
hcR4m 465 472 1.5 2.95 160 8 5.5 1.8
hcR3m 506 480 -5.1 3.00 160 6 3.5 1.8
Tab. 5.1: Comparison of numerically established maximal stress σmax,sim and calculated maximal
stress σmax according to [51]

Fig. 5.3: Location of maximal stress σmax near a transverse circular hole in a hollow cylinder
under axial tension

The deviations concerning the maximal stress at the edge of the hole (hole-drilling
measurement point) between the cylindrical models used in this investigation and the
literature solution are minimum 1.5 % and maximum 5.1 %. These deviations are
possibly caused by the mesh design near the hole. In Fig. 11.7, Fig. 11.8 and Fig. 11.9
(appendix chapter Chap.11), the initial stress distribution in x-direction near the hole is
plotted for model hcR6m, hcR4m and hcR3m, respectively. Although the initial stress is
set with 160 MPa by preloading all elements of the model, small stress differences in
94

the range of 0.3 % to 1 % are observed near the hole. This stress inhomogeneity near the
hole may cause, in terms of error propagation, the listed deviations in Tab. 5.1 after
introducing the through-hole into the modeled specimens.

The same comparison could not be applied in case of the cylindrical model with an
ø6 mm center through-hole hchR6m (specimen of Type 4). In this case, no theoretical
solution for the calculation of a stress distribution near the edge of the hole-drilling
measurement point was found in literature. A solution for the calculation of a stress
concentration factor at the edge of the ø6 mm center through-hole, similar to the
calculation summarized in Tab. 5.1, was not exactly possible because the stress
concentration factors according to [51] are plotted up to the ratio d0/2R = 0.4.
Considering the “big” center through-hole with d0 = 6 mm, the limit is exceeded with
the ratio d0/2R = 0.5. A extrapolation of the concentration factor according to the
diagram in the literature [51] determines a value, which is 10 % lower than the
calculated value of σmax,sim/σnom= 3.9. Nevertheless, it is assumed that the hchR6m
model has a similar numerical uncertainty as the hcR6m due to the fact that the model
hchR6m of Type 4 was generated from the hcR6m model of Type 2. Both have the
same mesh within the hole region (hole-drilling measurement point).

5.2.2 Uncertainty in Experimental Calibration


As in all measurement applications, the strain measurement and consequently the
residual stress calculation using the hole-drilling method is always associated with
errors and uncertainty. In Chap. 2.2.3 (Tab. 2.2) the influences and sources of error of
the whole residual stress evaluation process are listed. Some of these influences and
sources of errors are of importance only for the final steps of the residual stress analysis,
i.e. the stress calculation after the introducing of the hole (e.g. by drilling) and the
measuring of the relieved strains. To these influences one can include the selection of
the evaluation method and indirectly the geometrical boundary conditions. This is due
to the fact that the geometrical boundary conditions were specified according to the
evaluation method. Every evaluation method, e.g. the differential MPA method,
includes a specified amount of calibration functions or factors, which were determined
using models or specimens with a simple geometry (e.g. a flat plate). Another
influences or source of error can be related to the selection and use of the measurement
95

hardware. These are e.g. the strain gage technique or the drilling technique. Most of
these influences can be minimized or mostly eliminated if the experimental techniques
are correctly applied. At all events, some sources of errors remain and should be
considered in an uncertainty analysis.

In general, the measurement uncertainty u is defined as the interval between the correct
mean value of a series of measurements x and the upper and lower limit [52]. It
considers two components: the random component u z for the random deviations and
the systematic deviations u s for the unknown systematic deviations. The measurement
uncertainty can be calculated by the linear addition of both components or, by
geometrical summation if both components are approximately of equal size {5.1}.

{5.1} u = u z + u s or u = u z2 + u s2

The random component u z can be calculated for a series of measurements under


repetitive conditions with unknown standard deviation using {5.2}, in which the
variable t is a distribution according to Student, n is the number of measurements and
s is the empirical standard deviation.

t
{5.2} u z = s
n

{5.3} u s = ± a12 + a 22 ... + a n2

The systematic uncertainty u s , which exists in all measuring devices, can generally only
be determined using reliable data from manufacturers or using appropriate experimental
experience. A possible way to estimate the systematic uncertainty u s is by calculating

the geometrical sum of all single values ai {5.3}.

In the case of the analysis of uncertainty for the experimental calibration of the hole-
drilling method two points are to be taken into consideration. First of all, it must be
considered that with the available equipment only one measurement can be performed
for each calibration specimen. For example, each milling cutter has a certain
96

dimensional tolerance and introduces for this reason a hole with a specific diameter d0.
An additional measurement requires another measuring point, which may have a
different hole diameter d0. Thus, it is impossible to calculate the random uncertainty u z
in terms of a statistical analysis because it is impossible to repeat the same calibration
again. For this reason, the random uncertainty is quantified with zero in this
investigation. The second component, the standard uncertainty u s of the experimental
calibration, may be approximated by defining and quantifying the single uncertainties
and subsequent calculation of u s according to {5.3}. For this work it is especially
important that the uncertainty analysis should be applied on the experimental calibration
strain distributions in order to quantify the uncertainty and to compare, from this point
of view, the experiments with the simulations.

Although the experimental calibrations in this investigation were carried out as


accurately as possible, factors affecting the magnitude of relieved strains still remain.
The following factors, which may be not exhaustive, were listed according to literature
[33, 38, 53], and the experience made during the work:

Strain measurement technique: The sources of measurement errors caused by the


strain gage technique can be found in the strain gages themselves, the measuring
amplifier, the wires, etc. According to [54], the accuracy requirements for static stress
analysis for CEA gage series are “moderate”, which means, in general, for the whole
measuring chain that the accuracy is in the range of 2 % to 5 %. However, it should be
noted that the resolution of the strain gages is approximately 1 µm/m. Hence, the
uncertainty becomes greater, especially at small hole depths and small strain values. For
the subsequent calculation, the strain gage uncertainty is quantified for all examples
given in this discussion with a fixed value of 3.0 %, which is less than the average in the
range of 2 % to 5 %.

hole diameter d0: The dimension of the introduced hole diameter affects directly the
magnitude of the relieved strains. As it can be seen in Fig. 5.11 and Fig. 5.12 in a
subsequent chapter Chap. 5.3.2, the absolute strain values increase with increasing hole
diameter and vice versa. A change of 2.7 % in the dimension of the hole diameter (e.g.
1.80 mm to 1.85 mm) leads to a change of the strain values of approximately 5 %. A
97

change of 5.5 % in the dimension of the hole diameter (e.g. 1.80 mm to 1.90 mm) may
even lead to a change of the strain values of approximately 10 %. This is of particular
importance when comparing the experimental calibrations with the numerical
calibrations in this work. In many cases, the experimentally introduced hole diameter
does not coincide with the designed model hole diameter (s. results chapter Chap. 5).

In the last statements it is assumed that the hole diameter is determined exactly. In
practice, the hole diameter was determined using the microscope of the RS-200 after
each experimental calibration. The hole diameter was measured by reading the scale
division nsc in the microscope and by calculating it according to {5.4} The factors in
this formula are the specific factors of the used microscope. They convert the scale
division into the metric system.

n sc × 0.158 × 25.4
{5.3} d 0 = in mm
102

This means that an error in the range of nsc = 1 is possible when measuring the hole
diameter. For example, a hole diameter of nsc = 46 is approx. 1.80 mm and a hole
diameter of nsc = 47 means approx. 1.85 mm. As already mentioned, a change of
0.05 mm of the hole diameter leads to a change of approx. 5 % in the relieved strain
values. For the comparison of the experimental and numerical calibration it should be
noticed that the previously described effect of a smaller or bigger experimental hole
diameter on the strain values could be compensated by the measurement error of the
determination of the hole diameter. For the subsequent calculations, the uncertainty of
the hole diameter is quantified for each subsequent example individually depending on
the measured hole diameter after the specific calibration.

Establishing zero depth: The determination of z = 0 mm is in many cases difficult.


The reason is that the strain gage insulating rosette backing as well as the adhesive layer
must be removed before the drilling process starts on the testing material. This rosette
backing/adhesive layer is in the range of 0.1 mm. When working with the RS-200 hole-
drilling device, the user has to remove this polyimide/adhesive layer by making several
attempts, i.e. drilling a very small increment, controlling the change in the drilling noise
and inspecting wheter the material surface is visible when using the microscope. This
98

process of “listening and looking” [55] can lead to errors, especially if the surface is
rough or the surface curvature is high. In practice, the effect of an erroneous zero depth
is a shifting of the strain distributions along the depth axis z [38]. The uncertainty of the
zero depth determination is quantified in this special case with 0.02 mm because this
value is the smallest depth increment, which can be set with the depth setting
micrometer of the RS-200 hole-drilling device, which was used for the calibration
experiments.

Setting up of hole depth increments Δz: After establishing zero depth with the depth
setting micrometer, the subsequent depth increments Δz are also adjusted to an accuracy
of 0.02 mm. This error can possibly compensate the previous error of the zero depth
determination so that it is not considered explicitly in the subsequent calculation.

Eccentricity of the hole: Eccentricities of the holes as a result of an incorrect


adjustment will lead to a magnification or minimization of the calibration strain values,
similar to the effect of the change of the hole diameter or the hole depth. The
uncertainty related to an eccentricity less than 0.02 mm is quantified in [38] with less
than 2 % r. However, this uncertainty value is related to the stress evaluation and not to
the relieved strain of the calibration. This error is not considered in the subsequent
calibration.

Perpendicularity of the hole axis relevant to the surface: According to [53], the
uncertainty concerning the perpendicularity of the hole axis relevant to the surface is
negligible for plane surfaces and unknown for bent surfaces. The experience in this
work shows that in the case of the flat specimens, the correct perpendicular positioning
of the milling cutter was set without difficulties. In the case of the cylindrical
specimens, the perpendicular positioning was controlled particularly during the process
of the alignment of the milling cutter over the hole-drilling rosette. This process may
affect the subsequent drilling and leads to uncertainties, which were not investigated in
this work.

Irregularities in the hole shape: In contrast to the models, the experimentally


introduced hole during the calibration process shows deviations from the ideal blind-
hole shape. One reason is that the used milling cutters have a 45° bevel, which creates
99

consequently a bevel at the hole bottom. In order to quantify the uncertainty caused by
the irregularities in the hole shape, a model with a 45° bevel at the bottom of each
drilling increment was additionally generated. This additional model is, with exception
of the hole region, identical to the reference model fT6D20. The difference between the
strain distribution of the model with bevel Δεbevel and the strain distribution of the
reference model Δεreference (with ideal blind-hole shape) is plotted in the diagram of
Fig. 5.4 for strain gage no. 3. An axial load of σx = 160 MPa was applied on both
models. At a hole depth of 0.02 mm the difference is 0.38 µm/m which is approx. 15 %.
The maximal absolute difference is 4.3 μm/m which means 3.9 % at a hole depth of
0.5 mm. At a hole depth of 1 mm the difference is 3.1 µm/m which is approx. 1.6 %. In
the case of strain gage no. 1, the maximal difference does not exceed 1 μm/m. In order
to provide a uniform value for the subsequent calculation, the uncertainty is set with a
value of 2 %.

4
Δεbevel - Δεreference [μm/m]

σx = 160 MPa
2
strain gage no. 3

0
0 0.2 0.4 0.6 0.8 1 1.2
z [mm]

Fig. 5.4: Strain difference between a simulation with a model which considers a 45°hole bottom
bevel at each hole increment and the reference model with an ideal blind hole shape.

Magnitude of calibration stress - plastification: To avoid plastic strains as a


consequence of stress concentrations, the residual stresses in general and, specifically,
the calibration stresses should not exceed approx. 60 % of the local yield stress as
mentioned in Chap. 2.2.3. This request is fulfilled for most experimental calibrations,
which use the flat specimens of Type 1 and the cylindrical specimens of Type 2. In the
case of the specimens of Type 3 (flat specimens with ø10 mm center through-hole) and
100

especially of Type 4 (hollow cylindrical specimens with ø6 mm center through-hole),


the notch effect is intensified due to the geometrical condition of the hole (measurement
point) close to the center through-hole. This leads to plastification and, as a result, the
deformations near the hole (measurement point) are inhomogeneous and not elastic. An
example for this statement is shown in Fig. 11.10 (appendix Chap. 11). This figure
shows the equivalent stress according to Mises, which is symbolized by a color plot
over the hchR6m model. The actual hole depth is z = 1 mm and the model is externally
loaded with 210 MPa, which is the maximum load in the experimental calibration. In
the hole vicinity (measurement point), it can be seen, that in all grey colored sections of
the model the equivalent stresses exceed 690 MPa. This value is, according to the
manufacturer information, the minimal yield stress of the material used in this
investigation. The observation is especially noticeable between the ø6 mm center
through-hole and the measurement point. The resulting error was not investigated and
for this reason, this effect is not quantified for the subsequent calculation of the
measurement uncertainty. The plastification effect is ignored in this work in order to
separate only the geometry effect during determination of the calibration functions. This
means that the subsequent calculations using the MPA algorithm assumes linear-elastic
material behavior and, for this reason, all FEM calibrations are elastic. Nevertheless, it
should be mentioned that plastification problems can be significantly and several
investigation are found in literature concerning this matter, e.g. [33, 56, 57].

Magnitude of calibration stress due to exact positioning of strain gage rosette near
edges: The strain gage rosette is bonded manually to the measurement point of the
specimen and hence, it might be placed imprecisely. If the distance to specimen edges is
sufficiently large and the stress distribution is mostly uniform, this uncertainty does not
emerge. The situation is different for other specimens e.g. the flat specimens with the
ø10 mm center through-hole and the hollow cylindrical specimen with the ø6 mm
through-hole. Due to the stress concentration caused by the through-hole, the stress
distribution is inhomogeneous in this region. The figures Fig. 10.1 to Fig. 10.4 (model
fhT6D2m and fhT1.5D2m) and Fig. 10.5-Fig.10.6 (model hchR6m) show the stress
variation with increasing distance D. Therefore, if the strain gage rosette is not placed
exactly, the experimental calibration stress values may differ from those of the
simulation. In the case of the hchR6 specimen for example, the hole center
(measurement point) should be 2 mm away from the edge of the ø6 mm through-hole.
101

This distance is exactly modeled in the simulation but, in the case of the experiments, an
uncertainty is existent. This uncertainty was not investigated in this work and is not
considered for the subsequent calculation.

5.2.3 Graphical Representation of Uncertainty in Calibration Results


In this chapter four calibrations are selected in order to compare the quality of the
numerical results with the experimental results from the perspective of the experimental
uncertainty. Importance was attached to choosing calibrations which consider the
different geometrical parameters for this graphical representation of uncertainty. The
selected calibrations were carried out using the following specimen:

• fT6D20 - “reference flat specimen” (s. Fig. 4.1)


• fT1.5D20 - “thin flat specimen” (s. Fig. 4.6)
• hcR3 - “thin and curved specimen” (s. Fig. 4.24)
• hchR6 - “thin, curved and near to edge specimen (s. Fig. 4.33)

The random uncertainty is set to zero as explained in the previous chapter. The
systematic uncertainty is determined considering four single values which are discussed
in the previous chapter Chap. 5.22. The first three values are the uncertainty from the
strain gage technique astrain gage, the uncertainty from the hole diameter adiameter and the
uncertainty of the hole shape ashape. These three values calculate the error bars parallel to
the strain axis Δε according to {5.4}. The values of the strain gage technique
uncertainty and the hole shape uncertainty are set for all four calculations with astrain
gage = 3 % and ashape = 2 %. The hole diameter uncertainty adiameter is set individually for
each calculation. The fourth value is the uncertainty of the determination of the hole
depth with ahole depth = 0.02 mm and is plotted as an error bar parallel to the depth axis z
{5.5}.

{5.4} u s ,ε = a straingage
2
+ a diameter
2
+ a shape
2
in [%]

{5.5} u s , z = a holedepth = 0.02mm


102

Fig. 5.5 shows the calibration results of the reference specimen fT6D20 (same diagram
as Fig. 4.1) with the addition that the experimental uncertainty is symbolized with error
bars. The calculated total systematic experimental uncertainty of the strain values (error
bar in Δε-direction) is for this calibration 6.16 %. In this case the numerical hole
diameter of the simulation d0,sim coincides with the experimental one d0,exp. Therefore,
adiameter considers for this example the measurement error when determining the
diameter with the microscope and is set to 5 %. For this example it can be held that the
numerical strain distribution of strain gage no. 3 runs exactly within the range of the
error bars. The strain distribution of strain gage no. 1 is, in deeper increments, slightly
beneath the tolerance range.

The same observation has been made in case of the calibration of the “thin flat
specimen” fT1.5D20. The calculated total systematic uncertainty here is also 6.16 %
with a hole diameter uncertainty adiameter = 5 %, which considers only the diameter
measurement error (s. Fig. 5.6).

Another example is the calibration result of the “thin and curved” specimen hcR3,
which is plotted in Fig. 5.7 with the corresponding error bars. The hole diameter
uncertainty of adiameter = 10 % considers, in this case, the difference between the
experimental hole diameter d0,exp and the hole diameter of the model d0,sim. This leads to
a calculated total strain uncertainty of 10.63 %. Similar to the two previous examples,
the numerical strain distribution of strain gage no. 3 runs within the tolerance range and
the distribution of strain gage no. 1 runs, for deeper increments, beneath the tolerance
range.
103

100

sg 1
50

0
Δε [μm/m]
-50
sg 2
-100 Experiment ·
___
Simulation
-150
fT6D20 (Type 1)
σc,x = 160 MPa
sg 3
-200 d0,exp = 1.80 mm
d0,sim = 1.80 mm

-250
0.0 0.2 0.4 0.6 0.8 1.0 1.2
z [mm]

Fig. 5.5: Comparison of experimental and numerical calibration with reference specimen
fT6D20. Uncertainty values: astrain gage = 3 %, adiameter = 5 %, ashape = 2 %,
ahole depth = 0.02 mm

100

sg 1
50

0
Δε [μm/m]

-50

Experiment • sg 2
-100
Simulation ___

-150 fT1.5D20 (Type 1)


σc,x = 160 MPa
-200 d0,exp = 1.78 mm
d0,sim = 1.80 mm sg 3
-250
0.0 0.2 0.4 0.6 0.8 1.0 1.2
z [mm]

Fig. 5.6: Comparison of experimental and numerical calibration with specimen fT1.5D20.
Uncertainty values: astrain gage = 3 %, adiameter = 5 %, ashape = 2 %, ahole depth = 0.02 mm

The last calibration example in Fig. 5.8 shows numerical and experimental strain
distributions which differ at the first sight and also when the described uncertainty
analysis is applied. This calibration was carried out using the hchR6 specimen, which
highly violates the geometrical conditions by being “thin”, “curved” and having the
measurement point “close to the edge”. The hole diameter uncertainty of adiameter = 5 %
considers in this case the difference between the experimental hole diameter d0,exp and
104

the hole diameter of the model d0,sim. The total strain uncertainty is then 6.16 %. The
numerical strain distribution for strain gage no. 1 mostly coincide with the experimental
tolerance range whereas the numerical strain distribution for strain gage no. 3 coincide
only at hole depths between 0.4 mm and 0.7 mm within the experimental tolerance
range.

100

50
sg 1
0

-50
Δε [μm/m]

sg 2
-100 Experiment •
Simulation ___
-150
hcR3 (Type 2)
-200 σc,x = 160 MPa
d0,exp = 1.86 mm sg 3
-250 d0,sim = 1.80 mm
0.0 0.2 0.4 0.6 0.8 1.0 1.2
-300
z [mm]

Fig. 5.7: Comparison of experimental and numerical calibration with specimen hcR3.
Uncertainty values: astrain gage = 3 %, adiameter = 10 %, ashape = 2 %, ahole depth = 0.02 mm
100
sg 1

-100
Δε [μm/m]

sg 2
Experiment •
-200
Simulation ___

hchR6 (Type 4)
σx = 160 MPa
-300
d0,exp = 1.76 mm
sg 3
d0,sim = 1.80 mm

-400
0.0 0.2 0.4 0.6 0.8 1.0 1.2
z [mm]

Fig. 5.8: Comparison of experimental and numerical calibration with specimen hchR6.
Uncertainty values: astrain gage = 3 %, adiameter = 5 %, ashape = 2 %, ahole depth = 0.02 mm

With the exception of the last example determined with the hchR6 specimen, the
uncertainty analysis in this chapter shows that the numerical calibration almost
105

reproduces the experimental calibration within the calculated tolerance range. It must be
pointed out that several experimental uncertainty sources, which were discussed in the
previous chapter, are not considered in the uncertainty calculation. Especially in the
case of the calibrations with the hchR6 specimen, uncertainties concerning
plastification, the zero-depth determination, the perpendicularity of the hole axis
relevant to the surface and the exact positioning of the strain gage rosette, may cause
additional errors and, thus, contribute to a higher total strain uncertainty.

Taking the results of this chapter into consideration, it can be said that a calibration is
always associated with uncertainties, as many influencing factors affect the results. In
the case of a calibration carried out with finite elements, the accuracy of the results
depends on influencing factors like the selected type of elements, the refinement of the
model, the design of the mesh, etc. The quality of the hole-drilling calibration model
can be analyzed by comparing some specific results with existent solutions. Once an
acceptable accuracy is achieved with a finite element calibration, the resulting
calibration strain distributions run almost consistently and smoothly. In contrast to a
numerical calibration, the experimental distributions may run within a tolerance range
and are not always smooth due to the different sources of uncertainties described in this
chapter. This is problematic inasmuch as the determined calibration strain distributions
are used for the approximation of the calibration functions Kx and Ky of the MPA
algorithm. A scattering or a shift of the strain distribution leads to errors in the
determination of the calibration functions and consequently in the stress calculation. For
this reason, an experimental calibration of the hole drilling method according to the
MPA algorithm should be regarded as complementary to the numerical calibration in
order to have comparative values and to gain experience. The proper determination of
the calibration functions should be carried out using the results of the numerical
calibration.

5.3 Stress Evaluation on Specimens with Non-Reference Geometry

The calculation of stress depth distributions after a hole-drilling measurement is


possible using different evaluation methods. Some of them are already implemented in
commercial software. On the one hand, the evaluation algorithms of these programs
106

differ in the physical assumption of the strain relaxation after the removal of stressed
material. On the other hand, geometrical boundary conditions are defined for each
evaluation program, because the determination of these algorithms bases on calibrations
using specimens or models which have an almost ideal shape, e.g. a thick and wide flat
plate. The application of the hole-drilling method on components, which violate the
specific boundary conditions, leads to stress errors, i.e. the calculated stress distribution
value differs from the real one. In order to confirm this statement, four numerical
calibration results (same examples as in Chap. 5.2.3) are selected as input strain
distributions for the MPA software. Fig. 5.9 shows the stress distribution in x-direction
σx calculated with the MPA software for the models fT6D20m (reference), fT1.5D20m
and hcR3m. The uniform stress in x-direction within the hole volume is σc,x-
FEM = 160 MPa. The fourth example is shown in Fig. 5.10, which compares the
calculated (with MPA software) stress distribution σx for the model hchR6m with the
initial stress distribution in x-direction at the measurement point σc,x-FEM.

240

220
hcR3m
200
fT1.5D20m
σx [MPa]

180

σc,x-FEM
160

fT6D20m
140

120

100
0 0.2 0.4 0.6 0.8 1
z [mm]

Fig. 5.9: Comparison of calculated residual stress σx with MPA software (differential Method)
[2] and stress distribution σc,x-FEM at measurement point before drilling. Input strain
distribution calculated from calibration simulations using models fT620m (reference),
fT1.5D20m and hcR3m.

All four examples were smoothed using compensative spline functions with a strain
smoothing factor of fil = 0.5 (s. Chap.6.2.4). In addition to the diagrams in Fig. 5.9 and
Fig. 5.10, stress differences are listed for selected depth values in Tab. 5.2. The
calculation of these percental differences is related to the uniform stress distribution
107

σc,x-FEM = 160 MPa for the models fT6D20m, fT1.5D20m and hcR3m. In the case of
the model hchR6m, the related initial stress value depends on the hole depth.

350

σc,x-FEM
300

hchR6m
250
σx [MPa]

200

150

100

50
0 0.2 0.4 0.6 0.8 1
z [mm]

Fig. 5.10: Comparison of calculated residual stress σx with MPA software (differential Method)
[2] and stress distribution σc,x-FEM at the measurement point before drilling. Input
strain distribution calculated from calibration simulation using model hchR6m

Stress deviation Δσx related to calibration stress σc,x-FEM [%]


depth [mm] fT6D20m
hchR6m hcR3m fT1.5D20m
(reference)
0.0 34.9 5.6 16.9 -5
0.2 31.3 21.3 19.4 -2.5
0.5 16.6 31.3 20.0 0.6
1.0 -82.5 -22.5 -11.3 -3.1
Tab. 5.2: Stress deviation Δσx related to calibration stress σc,x-FEM for evaluated simulations
fT6D20m, fT1.5D20m, hcR3m, hchR6m. All calculations were carried out with the
differential MPA method

The four exemplarily calculated stress distributions show considerable differences to the
initial stress distributions. The calculated result of the ft6D20m reference model shows
a characteristic oscillating distribution for the evaluation of uniform stress using the
MPA software. This oscillating distribution, which has its origin in the calculation
formalism of the MPA software, depends, among other things, in the shape of the
calibration functions Kx and Ky and the application of the previous smoothing. The
topics concerning the execution of the evaluation formalism are discussed in the
108

following chapters (Chap. 6 and Chap 7). The reference model fT6D20 does not violate
the geometrical boundary conditions and hence, the calculated differences related to the
initial stress distribution σc,x-FEM = 160 MPa are small compared to the other
calculations with a maximal difference of 5 % at the selected hole depths. At the same
time, this result represents the current status of the stress evaluation using the
differential MPA method. The other calculation results show clearly the fact that a
violation of the boundary conditions specified for the MPA method leads to errors in the
stress calculation. In the case of the three non-reference models, all calculations show
higher deviations than the reference. The maximal deviation values are calculated for
the hchR6m model, which is “thin”, “curved” and whose measurement point is “close to
the edge” of the ø6 mm hole. The calculated distribution of this model also shows a
different trend compared to the initial stress distribution. The stress differences as well
as the distribution of the calculated result of the “curved” and “thin” model hcR3m are
between the results of the hchR6m model and the fT1.5D20m model. The “thin” and
“flat” fT1.5D20m model shows a max. 20 % stress deviation at the selected values and
the stress results run up to a hole depth of approx. 0.5 mm approx. 35 MPa parallel over
the distribution of the reference specimen.

240

220
hcR3m
200

fT1.5D20m
σx [MPa]

180

160
σc,x-FEM
fT6D20m
140

120

100
0 0.2 0.4 0.6 0.8 1
z [mm]

Fig. 5.11: Comparison of calculated residual stress σx with H-Drill software (integral Method)
[34] and stress distribution σc,x-FEM at the measurement point before drilling. Input
strain distribution calculated from calibration simulations using models fT620m
(reference), fT1.5D20m and hcR3m.
109

To finish this initial analysis, the same strain distributions of the models fT6D20m,
fT1.5D20m, hcR3m and hchR6m are used as exemplary input data for the integral
method implemented in the HDrill software [34]. The diagrams in Fig. 5.11 (fT6D20m,
fT1.5D20m and hcR3m) and in Fig. 5.12 (hchR6m) show that a stress calculation on
models with non-reference geometry with the integral method also leads to stress
differences.

350

σc,x-FEM
300
hchR6m
250
σx [MPa]

200

150

100

50

0
0 0.2 0.4 0.6 0.8 1
z [mm]

Fig. 5.12: Comparison of calculated residual stress σx with H-Drill software (integral Method)
[34] and stress distribution σc,x-FEM at the measurement point before drilling. Input
strain distribution calculated from calibration simulations using model hchR6m.

It should be noticed that the integral method requires a different depth increment
distribution from the one used in these examples. A depth increment distribution with
small depth increments leads to large errors according to [28, 58]. This problem is
related to the matrix character of the calculation, in which a high spatial resolution
causes a poorly conditioned calibration matrix and, consequently, leads to higher stress
errors. For this reason, the stress values for a hole depth smaller than 0.15-0.17 mm are
not plotted in Fig. 5.11 because some of them tend to extreme oscillation. Nevertheless,
except the result of the reference model fT6D20m, which runs close to the initial
uniform stress distribution of σc,x-FEM = 160 MPa, the other results show deviations
from the initial stress distribution and are comparable, at first sight, to the results
obtained with the MPA software. A further discussion on these results is not continued,
as the integral method is not a subject of investigation in this work.
110

It can be concluded that the calculation of a residual stress depth distribution using
current evaluation methods on components which violate the specific geometrical
boundary conditions leads to considerable errors. The individual geometrical influences
are discussed in the following chapters.

5.4 Influence of Geometrical Parameters on Results of Hole-Drilling


Measurements
5.4.1 Influence of Hole Diameter - d0
This chapter is included in order to underline the importance of the hole diameter for the
evaluation algorithms of the hole-drilling method. When carrying out a hole-drilling
measurement, the experimentally introduced hole diameter may differ depending on the
elastic properties of the material and the used milling cutter. The different current
evaluation programs of the hole drilling method (e.g. MPA, H-Drill) consider this fact
by including a set of diameter specific calibration functions (or diameter specific
factors) within the possible range of diameters that may be introduced for the particular
strain gage rosette. If the dimension of the introduced hole diameter is between two
calibration hole diameters, the calibration function (or the calibration factor) is
calculated using linear interpolation between the two sets of diameter specific
calibration functions.
0

z = 0.2 mm
-50

-100
Δε [μm/m]

z = 0.5 mm

-150

-200
Simulation of fT6D20 specimen z = 1.0 mm
influence of hole diameter d0
-250 at selected hole depths
Strain gage no. 3
-300
1.65 1.7 1.75 1.8 1.85 1.9 1.95
d0 [mm]

Fig. 5.13: Relieved Strain (strain gage no. 3) versus hole diameter at selected hole depths
(reference model fT6D20m)
111

depth strain gage no. 3: strain deviation [%]


[mm] d0 = 1.70 mm d0 = 1.75 mm d0 = 1.80 mm d0 = 1.85 mm d0 = 1.90 mm
0.2 -9.9 -4.3 0.0 7.7 14.0
0.5 -10.2 -4.9 0.0 6.4 12.3
1.0 -9.8 -4.9 0.0 5.4 10.6
Tab. 5.3: Strain deviation (strain gage no. 3) related to relieved strain of reference model
fT6D20m with reference hole diameter of d0 = 1.8 mm

The influence of the hole diameter at selected hole depths on the relieved strain is
shown in Fig. 5.13 and Fig. 5.14 for strain gage no. 3 and strain gage no. 1, respectively.
These results are calculated from the reference model fT6D20m (s. in Chap. 4.4 Fig. 4.2
and Fig. 4.3). The linear dependency of the relieved strain Δε on the hole diameter d0
can be clearly seen in both diagrams where the strain values increase with increasing
hole diameter.

In addition to the diagrams, Tab. 5.3 and Tab. 5.4 give the strain difference related to
the reference hole diameter d0 = 1.8 mm at the same hole depths. These differences are
the basis for the calculation of the hole diameter uncertainty adiameter in the previously
discussed chapter Chap. 5.23.

60
z = 1.0 mm

50

Simulation of fT6D20 specimen


40
influence of hole diameter d0
Δε [μm/m]

at selected hole depths


30
Strain gage no. 1
z = 0.5 mm

20

10

z = 0.2 mm

0
1.65 1.7 1.75 1.8 1.85 1.9 1.95
d0 [mm]

Fig. 5.14: Relieved Strain (strain gage no. 1) versus hole diameter at selected hole depths
(reference model fT6D20m)
112

depth strain gage no. 1: strain deviation [%]


[mm] d0 = 1.70 mm d0 = 1.75 mm d0 = 1.80 mm d0 = 1.85 mm d0 = 1.90 mm
0.2 -8.9 -4.1 0.0 5.4 10.3
0.5 -8.4 -4.0 0.0 4.9 9.4
1.0 -8.1 -4.0 0.0 4.3 8.4
Tab. 5.4: Stain deviation (strain gage no. 1) related to relieved strain of reference model
fT6D20m with reference hole diameter of d0 = 1.8 mm

An evaluation using the diameter specific strain distributions as input data for the MPA
software calculates almost equal stress distributions. The maximal average difference of
the stress calculation related to the reference calculated stress distribution is only 2.3 %.
This can be interpreted as a calculation uncertainty.

The hole diameter influences linearly the relieved strain. An implementation of (new)
calibration functions or calibration factors into stress evaluation algorithms should
always consider the possible range of the effectively introduced hole diameter by the
experimental measurement.

5.4.2 Influence of Specimen Thickness - T


The influence of the specimen thickness based on the previous numerical calibration
results (s. in Chap. 4.3 Fig. 4.7 and Fig. 4.8) is discussed in this chapter. In all
examples, the models with different thickness T were loaded with a uniform tensile
stress and had a constant hole diameter of d0 = 1.8mm.

The first question is whether the established boundary condition of T = 3.33 x d0 [2] is
adequate or may be minimized. For this purpose, the relieved strain Δε is plotted versus
the specimen thickness T at selected hole depths in Fig 5.15 (strain gage no. 3) and
Fig. 5.16 (strain gage no. 1). It can be seen that the strain distributions over the
specimen thickness are almost constant between T = 3.33 x d0 = 6 mm and
T = 1.67 x d0 = 3 mm. This observation is confirmed by the strain deviations listed in
Tab. 5.5 (strain gage no. 3) and Tab. 5.6 (strain gage no. 1). The strain deviations
between the results of T = 1.67 x d0 = 3 mm and the reference are small with values
between 0.1-3.9 %. Thus, in this special case (uniaxial stress and violation of single
boundary condition thickness) the boundary condition may be minimized.
113

The second question concerns the developing of the strain distributions if the thickness
of the specimen exceeds the limit and consequently the strain values change. In the
diagrams of Fig 5.15 and Fig. 5.16 it can be seen that for a decreasing thickness
between T = 1.67 x d0 = 3 mm and T = 0.55 x d0 = 1 mm the strain values change non-
linearly without a consistent trend at the selected hole depths. When the specimen is
perforated (at T = 1.0 mm and z = 1.0 mm), the absolute strain values reach a maximum
which is clearly visible for strain gage no. 1 (transverse to load direction) with a
difference of 90.1 %. To complete the discussion about the influence of the specimen
thickness, the distribution of the strain differences over the hole depth are plotted in
Fig. 5.17 and Fig. 5.18 for different thicknesses. The strain difference ΔεDIFF.(T,z) is
calculated according to {5.6}.

{5.6} Λε DIFF . (T , z ) = Λε (T , z ) − Λε ( REF , z )

The strain differences increase with increasing hole depth and with decreasing specimen
thickness. First of all, it can be observed that the strain difference of the distribution
with T = 1.67 x d0 = 3 mm runs close to zero for both strain gages. The distribution with
T = 1.11 x d0 = 2 mm shows a visible difference to the reference distribution for strain
gage no. 3 and small difference for strain gage no. 1. Especially in the case of the thin
specimens with T = 1.5 mm and T = 1.0 mm the differences as well as the slope change
of the distribution are high. The maximal strain difference is -60 μm/m at
z = 0.7 mm = 0.7 x T for T = 1.0 mm and -30 μm/m at z = 0.9 mm = 0.6 x T for
T = 1.5 mm. After introducing the through-hole, i.e. when the specimen is completely
perforated, the absolute values of the strain differences decrease and increase for strain
gage no. 3 (parallel to load direction) and strain gage no. 1 (transverse to load
direction), respectively. These facts can be related to the loss of stiffness in the hole
vicinity of the almost introduced or completely perforated through-hole, i.e. the effect of
the blind hole, which causes a partial release of strains, is reduced gradually as the
specimen is almost perforated. Finally, an elastic spring-back occurs after the
perforation of the specimen. This causes finally a total relaxation of stress near the hole.
114

z = 0.2 mm

-50

z = 0.5 mm
-100
Δε(T)

Influence of thickness T
at selected hole depths
-150
Strain gage no. 3

-200
z = 1.0 mm

-250
0 1 2 3 4 5 6
T [mm]

Fig. 5.15: Relieved strain (strain gage no. 3) versus thickness at selected hole depths

depth strain gage no. 3: strain deviation [%]


[mm] T = 1 mm T = 1.5 mm T = 2 mm T = 3 mm T = 6 mm
0.2 61.1 21.2 6.2 -1.5 0.0
0.5 49.8 19.4 6.6 -0.7 0.0
1.0 20.0 15.3 6.7 -0.1 0.0
Tab. 5.5: Models with different thickness: strain deviation (strain gage no. 3) related to relieved
strain of reference model fT6D20m with reference thickness of T = 3.33xd0 = 6 mm

110
Influence of thickness T
100
at selected hole depths
90 Strain gage no. 1
80

70
Δε [μm/m]

60
z = 1.0 mm
50

40

30 z = 0.5 mm
20

10
z = 0.2 mm
0
0 1 2 3 4 5 6
T [mm]

Fig. 5.16: Relieved strain (strain gage no. 1) versus thickness at selected hole depths
115

depth strain gage no. 1: strain deviation [%]


[mm] T = 1 mm T = 1.5 mm T = 2 mm T = 3 mm T = 6 mm
0.2 -92.7 -29.7 -6.3 3.9 0.0
0.5 -50.4 25.2 -7.9 1.7 0.0
1.0 90.1 -3.6 -3.0 1.2 0.0
Tab. 5.6: Models with different thickness: strain deviation (strain gage no. 1) related to relieved
strain of reference model fT6D20m with reference thickness of T = 3.33xd0 = 6 mm

10

0
T = 1.67d0 = 3 mm
Δε(T)−Δε(Ref.) [μm/m]

-10
T = 1.11d0 = 2 mm

-20

-30
T = 0.83d0 = 1.5 mm

-40
Influence of thickness
strain difference
-50
T = 0.55d0 = 1 mm Strain gage no. 3

-60
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6
z [mm]

Fig. 5.17: Influence of thickness: strain difference versus hole depth: strain difference of strain
gage no. 3 is related to strain distribution of reference model Δε(Ref.)

50
Influence of thickness
40 strain difference
Strain gage no. 1 T = 0.55d0 = 1 mm
T = 0.83d0 = 1.5 mm
Δε(T)−Δε(Ref.) [μm/m]

30

20

10
T = 1.11d0 = 2 mm

T = 1.67d0 = 3 mm
-10

-20
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6
z [mm]

Fig. 5.18: Influence of thickness: strain difference versus hole depth: strain difference of strain
gage no. 1 is related to strain distribution of reference model Δε(Ref.)
116

To conclude, the last question is about the thickness influence on the stress calculation
using the MPA software. For this purpose, the thickness specific calibration strain
distributions are used as input data and evaluated into stress distribution with the MPA
software. After this, the stress difference is calculated from the stress values σx(T) and
σy(T) and the stress values of the reference specimen according to {5.7}.

{5.7} σ DIFF . (T , z ) = σ (T , z ) − σ (REF ., z )

The input strain distributions for all examples consider values up to z = 1 mm, because
the terminating condition of the MPA software evaluates up to a hole depth of
ξ = z/d0 = 0.6 mm. In this special case with a constant hole diameter of d0 = 1.8 mm, the
software terminates the evaluation at a hole depth of 1.08 mm. The results are plotted in
Fig. 5.19 for the stress differences in x-direction and Fig. 5.20 for the stress differences
in y-direction. In addition, Tab. 5.7 and Tab. 5.8 list the stress deviations (or
differences) for the corresponding stress direction. Note that the values in Tab. 5.8 show
the stress differences in transverse direction σy and are listed in MPa. This difference in
the units is because some of the calculated σy(Ref.)-values are zero or close to zero.
Hence, a calculation of a deviation in percent is impossible or leads to very high values.

150

Influence of thickness:
T = 0.55d0 = 1 mm
100 stress difference σx
using MPA-Method
σx(T)−σx(Ref.) [MPa]

T = 0.83d0 = 1.5 mm
50

T = 1.11d0 = 2 mm
0

T = 1.67d0 = 3 mm
-50

-100

-150
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
z [mm]

Fig. 5.19: Stress difference σx(T) versus hole depth: Stress difference σx(T) is related to calculated
stress distribution of reference specimen σx(Ref.) using MPA software
117

Stress deviation Δσx related to calculated reference stress σx(Ref.)


depth [mm] [%]
T = 1 mm T = 1.5 mm T = 2 mm T = 3 mm
0.0 67.1 23.0 6.6 2.0
0.2 62.2 22.4 7.1 0.6
0.5 28.6 19.3 8.1 0.6
1.0 -225.2 -8.4 6.5 1.3
Tab. 5.7: Stress deviation Δσx related to calculated stress distribution (MPA software) of
reference model σx(Ref.) for evaluated simulations with different specimen thickness

50
T = 0.55d0 = 1 mm
40 Influence of thickness:
stress difference σy
30 using MPA-Method
T = 0.83d0 = 1.5 mm
σy(T)−σy(Ref.) [MPa]

20

10 T = 1.11d = 2 mm
0

0
T = 1.67d0 = 3 mm
-10

-20

-30

-40

-50
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
z [mm]

Fig. 5.20: Stress difference σy(T) versus hole depth: Stress difference σy(T) is related to calculated
stress distribution of reference specimen σy(Ref.) using MPA software

Stress difference Δσy related to calculated reference stress σy(Ref.)


depth
[MPa]
[mm]
T = 1 mm T = 1.5 mm T = 2 mm T = 3 mm
0.0 40 12 2 -2
0.2 44 16 5 -1
0.5 1 13 6 0
1.0 -478 -32 -1 0
Tab. 5.8: Stress difference Δσy related to calculated stress distribution (MPA software) of
reference model σy(Ref.) for evaluated simulations with different specimen thickness
118

The differences for the calculated stress of the specimen with T = 1.67 x d0 = 3 mm are
in both directions negligible as can be seen in the diagrams as well as in the deviation
values in the tables. With decreasing thickness the calculated stress values increase and
this is reflected on the calculated deviations. For the specimen with
T = 1.11 x d0 = 2 mm, the difference for σx is in average 7 % and for σy 3.5 MPa. In
case of the thin specimens, the stress differences are considerable. For the specimen
with T = 0.83 x d0 = 1.5 mm the differences for σx are for the selected hole depths on
average approx. 18 % and 18 MPa for σy. The highest differences are determined for the
thinnest specimen with T = 0.55 x d0 = 1 mm and especially at depth increments higher
than z =0.6 mm the results are totally incorrect. The calculated stress difference is at the
depths of z = 0, 0.02, 0.5 mm in average approx. Δσx = 53 %. But at a hole depth of
z = 1 mm the difference is very high with Δσx = 225.2 % and Δσy = -478 MPa.

Conclusion: The boundary condition “thickness” is set in [2] with T = 3 x d0. The
results in this chapter, which were all determined using uniformly loaded finite elements
models under violation of only the single boundary condition T, show that the geometry
limit may be minimized up to a thickness of T = 1.66 x d0. After this thickness, the
relieved strains values increase with decreasing thickness. The trajectory of the strain
distributions after a calibration on thin specimens depend upon whether the introduced
hole almost perforates or completely perforates the specimen cross section. A stress
calculation with the MPA software [2.44] leads consequently to higher stress values
compared to a calculation using the strain distributions determined with the reference
specimen. The stress differences are on average over the hole depth 18 % for
T = 0.83 x d0 = 1.5 mm and about 53 % (225% at z = 1 mm) for T = 0.55 x d0 = 1 mm.

5.4.3 Influence of Distance Hole-Edge - D


The results of Chap. 4.3 (Fig. 4.7 and Fig. 4.8) are taken as a basis for the discussion of
the influence of the distance from the hole center (measurement point) to the free edge
of the specimen D. The boundary condition is set with D = 5 .. 10 x d0 according to the
MPA specification [2]. Similar to the discussion of the previous chapter, the question to
be answered about the distance influence is whether this specific limit may be reduced
and what does it mean for a stress calculation with the MPA software. Therefore, the
relieved strains at selected hole depths are plotted against the distance hole-edge in
119

Fig. 5.21 and Fig. 5.22 for strain gage no. 3 and strain gage no. 1, respectively. In both
diagrams, the strain distributions of both strain gages at all selected hole depths run
almost constantly in the range between D = 1.94 … 11.1 x d0 = 3.5 … 20 mm. Only at
D = 1.11 x d0 = 2 mm and z = 1, a maximal strain difference related to the reference is
calculated with approx. 4 % for strain gage no. 3 and approx. 2 % for strain gage no. 1.

These relatively small strain differences lead to small differences of the calculated stress
with the MPA software, as can be seen in Fig. 23. The stress differences σx(D,z) are
calculated analogically as in {5.7}, with the difference that the stress distributions were
dependet on the distance D. In addition, the stress deviations in per cent at selected hole
depths are listed in Tab. 5.9.

0
z = 0.2 mm
-20

-40

-60

-80
Δε [μm/m]

z = 0.5 mm
-100

-120
Influence of Distance Hole-Edge
-140 at selected hole depths
-160 strain gage no. 3
z = 1.0 mm
-180

-200
0 2 4 6 8 10 12 14 16 18 20
Distance Hole-Edge [mm]

Fig. 5.21: Relieved strain (strain gage no. 3) versus distance hole-edge at selected hole depths
120

60
z = 1.0 mm

50

Influence of Distance Hole-Edge


40 at selected hole depths
strain gage no. 1
Δε [μm/m]

z = 0.5 mm
30

20

z = 0.2 mm
10

0
0 2 4 6 8 10 12 14 16 18 20
Distance Hole-Edge [mm]

Fig. 5.22: Relieved strain (strain gage no. 1) versus distance hole-edge at selected hole depths

50

40 Influence of distance hole-edge:


stress difference σx
30 using MPA-Method
σx(D)−σx(Ref.) [MPa]

20
D = 1.94 … 5.56 x d0 = 3.5 … 10 mm
10

-10

-20
D = 1.11 x d0 = 2 mm
-30

-40

-50
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
z [mm]

Fig. 5.23: Stress difference σx(D) versus hole depth: Stress difference σx(D) is related to
calculated stress distribution of reference specimen σx(Ref.) using MPA software
Stress deviation Δσx related to calculated reference stress σx(Ref.)
depth [mm] [%]
D = 2 mm D = 3.5 mm D = 5 mm D = 10 mm
0.0 -3.9 1.3 0.7 0.0
0.2 -3.2 1.3 0.6 0.0
0.5 -3.7 1.2 0.6 0.0
1.0 -9.7 2.6 1.3 0.0
Tab. 5.9: Stress deviation Δσx related to calculated stress distribution (MPA software) of
reference model σx(Ref.) for evaluated simulations with different distance hole-edge
121

The maximal stress difference in x-direction is calculated for D = 1.11 x d0 = 2 mm with


approx. -5 % averaged over the hole depth and -9.7 % at z = 1 mm. The stress
differences in y-direction are even smaller and, therefore, a graphical representation is
not worthwhile.

Within the scope of an additional investigation [59], FEM calculations were carried out
using the models of the flat specimens with different distance hole-edge D. The models
were loaded with a pure shear stress field of σc,x =160 MPa and σc,y = -160 MPa. For
this loading case, the calculated stress deviation was in average 5 % for a model with
D = 2 mm and T = 6 mm, which is similar to the deviations presented in this work for
the uniaxial stress state. The stress deviation increases to a value of 20 % if the same
model becomes “thin” with a thickness of T = 1.5 mm. In this work, the position of the
CEA-06-062-UM strain gage rosette concerning the free edge of the specimen was
always identical (s. also Fig. 3.1, Fig. 3.2 and Fig. 3.8). Thus, strain no. 3 is parallel to
the free edge and strain gage no. 1 is perpendicular to the specimen edge with the
following order: “specimens edge → hole → strain gage no. 1”. This order allows a
positioning of the strain gage very near to the specimens edge with a minimum distance
hole edge of D = 2 mm. If the order of the positioning of strain gage no. 1 is changed,
e.g. “specimens edge → strain gage no. 1 → hole”, then the hole can not be introduced
so closely to the specimen’s edge because of the length of the strain gage itself and the
length of the soldering lug. The minimum distance hole-edge is then D = 9 mm. In such
a positioning case, different FEM calculation with different loading conditions (uniaxial
in x-direction, uniaxial in y-direction, biaxial plane stress, biaxial pure shear stress)
show no significant differences of the relieved strains compared to the results with
models with reference geometry.

Conclusion: In contrast to the specimen thickness, the distance from the center of the
measurement point (hole center) to the specimen’s free edge does not influence the hole
drilling results significantly. The maximum stress difference of approx. 5 % on average
related to the reference is found for the results of the model with D = 1.11 x d0 = 2 mm.
Thus, the boundary condition for this special case, which is set with min. D = 5 x d0 [2]
may be minimized up to D = 1.94 x d0.
122

5.4.4 Influence of Cylinder Radius - R


This chapter is based on the numerical results determined on cylindrical models (s.
Chap. 4.6) and is discussed similarly as in the previous examples. The boundary
condition for the geometrical parameter radius of surface curvature is set with
R = 3 x d0 according to [2]. A possible minimization of this boundary condition can be
analyzed using the results in Fig. 5.24 and Fig. 5.25, which plot the relieved strain
against the cylinder radius at selected hole depths for strain gage no. 3 and strain gage
no. 1, respectively. These diagrams also distinguish between the relieved strain of the
solid cylinders and the ones with the hollow cylinders.

z = 0.2 mm sc
-50 hc

Ref.
-100
Δε [μm/m]

z = 0.5 mm sc

hc
-150 Influence of Cylinder Radius
at selected hole depths
strain gage no. 3
-200
sc
z = 1.0 mm
hc
-250 / /
2 3 4 5 6 ∞7
Cylinder Radius [mm]

Fig. 5.24: Relieved strain (strain gage no. 3) versus cylinder radius at selected hole depths

strain gage no. 3: strain deviation [%]


depth
solid cylinder hollow cylinder
[mm]
R = 3 mm R = 4 mm R = 6 mm R = 3 mm R = 4 mm R = 6 mm
0.2 -10.8 -0.8 1.6 8.0 23.0 31.5
0.5 1.1 5.1 6.1 20.2 27.7 32.9
1.0 7.1 8.4 8.1 20.4 20.0 22.4
Tab. 5.10: Cylindrical models: strain deviation (strain gage no. 3) related to relieved strain of
reference flat model fT6D20m
123

70
sc
60
hc Influence of Cylinder Radius
50 at selected hole depths
z = 1.0 mm
Δε [μm/m] strain gage no. 1
40

Ref.
30
sc
20
z = 0.5 mm hc
10
sc
z = 0.2 mm
hc
0 / /
2 3 4 5 6 ∞7
Cylinder Radius [mm]

Fig. 5.25: Relieved strain (strain gage no. 1) versus cylinder radius at selected hole depths

strain gage no. 1: strain deviation [%]


depth
solid cylinder hollow cylinder
[mm]
R = 3 mm R = 4 mm R = 6 mm R = 3 mm R = 4 mm R = 6 mm
0.2 -34.2 -1.6 11.2 -61.2 -21.3 10.5
0.5 -12.9 16.6 25.2 -50.6 -16.9 8.4
1.0 2.3 24.8 27.2 -22.4 6.4 26.2
Tab. 5.11: Cylindrical models: strain deviation (strain gage no. 1) related to relieved strain of
reference flat model fT6D20m

Another illustration of the radius influence is plotted in the diagrams of Fig. 5.26 (strain
gage no. 3) and Fig. 5.27 (strain gage no. 1), which show for all cylindrical models the
distributions of the strain differences related to the reference strain distribution. The
strain differences are calculated similarly to {5.6} using R instead of T as a variable. In
addition to the diagrams, Tab. 5.10 (strain gage no. 3) and Tab. 5.11 (strain gage no. 1)
list the strain deviation related to the reference strain of model fT6D20m (in %).

First of all, it can be noticed that the difference between the results taken with the solid
cylinder with R = 3.33 x d0 = 6 mm and the reference are approx. 6-8 % (max. 16 µm/m
at z = 1 mm). This means that the boundary condition may not be minimized. After
R = 6 mm (s. Fig. 5.25), the strain distributions in this case run almost constantly with
decreasing cylinder radius. Furthermore, the addition of the parameter thickness, i.e. in
124

the case of the hollow cylinders, causes a magnification (approx. 10 % to 20 %) of the


absolute strain values registered with strain gage no. 3. In the diagram of Fig. 5.26 the
strain difference distributions for the solid cylinders with R = 6 mm and R = 4 mm run
almost one upon the other and reach a maximal strain difference at z = 1 mm of -
16 µm/m. The strain difference distribution of the solid cylinder with R = 3 mm runs
approx. 2-5 µm parallel over the other two solid cylinder strain difference distributions.
The strain difference distributions of the hollow cylinder compared to those of the solid
cylinder behave almost similarly to the distributions of the thin flat specimens compared
to thick flat specimens. Here, the strain differences increase with increasing hole depth
up to a depth of z = 0.8 mm = 0.64 x T (with T = 1.25 mm). The maximal strain
difference at this hole depth is -47 µm/m for the hollow cylinder with R = 6 mm. After
this depth, the slope of the curves decreases and this thickness effect can be associated
to the nearly perforated cross section of the hollow cylinder. The differences are at
maximum 5-10 µm/m between the single distributions of the hollow cylinders
(Fig. 5.26).

The strains registered with strain gage no. 1 (transverse to the load direction) show
significant strain deviations compared with the reference strain values. Already in the
case of the solid cylinder with a radius of R = 3.33 x d0 = 6 mm the strain deviations at
the selected hole depths are at maximum 26 %. A decrease of the cylinder radius causes
a decrease of the absolute strain values of strain gage no. 1. The deviation is on average
approx. -45 % for R = 1.67 xd0= 3 mm at the selected hole values. This curvature effect
can be attributed to the position of strain gage no. 1 related to the influenced material
section near the hole, which becomes smaller with decreasing radius. In Fig. 5.28,
schematic sectional views of the cylinder specimens are shown. The position of the
strain gage (s.g.) is symbolized with a gray line and the material section directly near
the 1.8 mm hole (with z = 1 mm) is symbolized with a hatched area for all three
examples. It can be seen that the length of the strain gage for the cylinder with
R = 6 mm is completely over the cross-section material area which is near to the hole
and about one half for the specimen with R = 3 mm. This assumption is verified by the
plotted results over the sectional view of the cylinder models in Fig. 11.11 (scR6m),
Fig. 11.12 (scR4m) and Fig. 11.13 (scR3m) in the appendix chapter Chap. 11. In these
three examples the plotted strain is tangential to the cylinder surface with equal color
scale. It can be seen that, in case of the models scR6m and scR4m, the strain gage is
125

almost completely over the orange marked surfaces. These orange surfaces symbolize
strain values between 37.5-145.8 µm/m. In contrast, the strain gage of model scR3m is
not completely over this orange marked area. A strain calculation, which is carried out
by integrating the strain field over the strain gage dimensions, leads to smaller strain
values in the case of model scR3m. Nevertheless, if a measurement should be carried
out on a high curved surface, e.g. R = 3 mm, it is possible to use a smaller strain gage
rosette. For example, a strain gage rosette, which allows the generation of a hole of
max. d0 = 1 mm, reduces the described curvature effect. Moreover, in this case
(R = 3 mm), the boundary condition “surface curvature” according to the MPA
algorithm is then satisfied. The main disadvantage of a measurement using a smaller
hole diameter is that the values of the relived strains are always smaller than the ones
measured with a bigger hole diameter and consequently the measurement uncertainty
has then a higher influence on the results.

10
scR3
scR4
0
Δε(R)-Δε(Ref.) [μm/m]

scR6
-10

-20

-30
Influence of cylinder radius hcR6 hcR4 hcR3
-40 strain difference
strain gage no. 3

-50
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
z [mm]

Fig. 5.26: Influence of cylinder radius: strain difference versus hole depth: strain difference of
strain gage no. 3 is related to strain distribution of reference model Δε(Ref.)
126

20

15
scR6
10
Δε(R)-Δε(Ref.) [μm/m]

scR4 hcR6
5

0
scR3
hcR4
-5

-10 hcR3
Influence of cylinder radius
-15 strain difference
strain gage no. 1
-20
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
z [mm]

Fig. 5.27: Influence of cylinder radius: strain difference versus hole depth: strain difference of
strain gage no. 1 is related to strain distribution of reference model Δε(Ref.)

Fig. 5.28: Schematic sectional view of cylindrical models and position of strain gage (strain gage
no. 1) transverse to cylinder longitudinal axis

100
Influence of cylinder radius:
stress difference σx
75
using MPA-Method
hcR6
σx(R)−σx(Ref.) [MPa]

50
hcR4 hcR3

25

scR6
scR4
0

scR3
-25

-50
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
z [mm]

Fig. 5.29: Stress difference σx(R) versus hole depth: stress difference σx(R) is related to calculated
stress distribution of reference specimen σx(Ref.) using MPA software
127

Stress deviation Δσx related to calculated reference stress σx(Ref.) [%]


depth
solid cylinder hollow cylinder
[mm]
R = 3 mm R = 4 mm R = 6 mm R = 3 mm R = 4 mm R = 6 mm
0.0 -11.8 -7.2 -2.6 11.2 15.8 27.0
0.2 -3.8 7.1 7.1 24.3 34.0 39.1
0.5 2.5 8.1 8.1 30.4 29.8 29.8
1.0 -5.8 14.8 12.9 -20.0 -18.7 -38.7
Tab. 5.12: Stress deviation Δσx related to calculated stress distribution (MPA software) of
reference model σx(Ref.) for evaluated simulations with different cylinder radius R

The main question is concerns the calculated stress differences related to the reference
stress distribution using the MPA software. For this purpose, the calculated stress
differences are plotted in Fig. 5.29 and Fig. 5.30 for σx and σy, respectively. The stress
differences are calculated similarly as in {5.7} with the difference that the stress
distributions are dependent on R instead of T. In addition, the stress deviations (or
differences) at selected hole depths are listed in Tab. 5.12 for σx (in %) and in Tab. 5.13
for σy (in MPa). The different units are chosen due to the fact that for σy some reference
stress values are very small or zero with the result that, in the first case, the calculation
leads to very high percental values, or, in the second case, the calculation is not possible
because of the division with zero.

The stress differences in x-direction as well as in y-direction are on the whole not very
high for the solid cylinders with R = 6 mm and R = 4 mm. The maximal stress deviation
in x-direction is approx.15 % (or less than 25 MPa) at z = 1 mm but on average over the
hole depth the stress deviations are about 8%. Also in y-direction, the calculated stress
differences are almost constant against the hole depth with a maximal stress difference
less than 12 MPa. In contrast, the results taken with the scR3m have, on the whole,
lower stress differences compared to the other two results with the solid cylinders. The
stress deviation in x-direction for the solid cylinder with R = 3 mm is only at z = 0 mm -
11.8 % but on average over the hole depth, the difference values are between -6 % to
2 %. Significant stress differences are calculated in y-direction with the results of the
scR3m model: the absolute values of the stress differences increase from -11 MPa at
z = 0 mm up to -100 MPa at z = 1 mm.
128

50
hcR3
25
hcR4

hcR6 scR6
0
σy(R)−σy(Ref.) [MPa]

scR4
-25
scR3

-50

-75
Influence of cylinder radius:
stress difference σy
-100
using MPA-Method

-125
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
z [mm]

Fig. 5.30: Stress difference σy(R) versus hole depth: stress difference σy(R) is related to calculated
stress distribution of reference specimen σy(Ref.) using MPA software

Stress difference Δσy related to calculated reference stress σy(Ref.) [MPa]


depth
solid cylinder hollow cylinder
[mm]
R = 3 mm R = 4 mm R = 6 mm R = 3 mm R = 4 mm R = 6 mm
0.0 -11 2 0 38 10 5
0.2 -22 -3 -3 41 16 12
0.5 -50 -10 -7 22 13 10
1.0 -100 -6 -2 -113 -43 -63
Tab. 5.13: Stress difference Δσy related to calculated stress distribution (MPA software) of
reference model σy(Ref.) for evaluated simulations with different cylinder radius R

In the case of the hollow cylinders, the stress difference in x-direction for the hcR6m
model starts with a higher value of 27 % at z = 0 mm and has a maximum at z = 0.2-
0.3 mm of approx. 40 %. After this maximum the stress difference shows a decrease up
to -38.7 % at z = 1 mm. The distributions of the hollow cylinders with R = 4 mm and
R = 3 mm of stress differences in x-direction at z = 0-0.3 mm run below the distribution
of R = 6 mm. At z = 4 mm the stress difference curves in x-direction of the hollow
cylinders are very close to each other and they diverge at z = 1 mm. In y-direction the
distribution of the stress differences of the hollow cylinders with R = 6 mm and
R = 4 mm shows a similar behavior of the curves with a small increase up to approx. at
z = 0.3 mm and a decrease up to -63 MPa at z = 1 mm. The stress evaluation in y-
direction on the hollow cylinder with R = 3 mm calculates a difference of approx.
129

40 MPa between z = 0 mm and z = 0.5 mm. After this depth, the values fall up to a
stress difference of -113 MPa at z = 1 mm.

Conclusion: The boundary condition “radius of surface curvature” is set in [2] with
T = 3 x d0. The results in this chapter were determined using uniformly loaded
cylindrical finite elements model, which means that the models have a uniaxial curved
shape instead of a biaxial curved shape (e.g. a shell or a sphere). They show that the
boundary condition could not be minimized and should be even displaced to a higher
value. The maximal stress deviations in x-direction are 12 % for the solid cylinder
(39 % for the hollow cylinder) have already been calculated for the results with
R = 3.33 x d0, which already satisfies the set boundary condition. In y-direction, which
is transverse to the load direction, the maximum stress differences are significantly high
for the results with R = 1.67 x d0. The difference values are approx. 100 MPa for the
solid cylinder as well as for the hollow cylinder. The loss of cross section in the form of
the hollow cylinders causes a similar thickness effect as in the case of the flat
specimens.

5.5 Possible Reduction of Calculated Stress Differences using


Geometry Specific Calibration Functions

The previous chapters show that a stress evaluation on specimens with non-reference
geometry leads in some cases to significant stress differences when using the
differential MPA software for the hole-drilling method. This is due to the fact that the
MPA software includes calibration functions which were determined using reference
(flat) specimens. The possibility of the reduction of the calculated stress differences
using geometry specific calibration instead of a reference calibration is discussed in
general in this chapter. For this purpose hole drilling measurements are simulated using
modified models of the already described models fT1.5D20m, hcR3m and hchR6m.
The modification of the models only concerns the material model and the magnitude of
the initial load, in order to distinguish between the simulated measurements and the
calibration. Additionally, the material independent character of the calibration is thus
demonstrated, i.e. one calibration can be used for measuring all kind of materials. The
exemplarily simulated measurements were carried out on models which have typical
130

elastic parameters of an aluminum alloy (E = 70000 MPa and ν = 0.3). The applied
uniform stress, which equals in the case of models ft1.5D20m and hcR3m the simulated
residual stresses, is uniform over the hole depth with σx = 100 MPa. In case of model
hchR6m, the stress distribution in x-direction σx(z) varies with the hole depth. The
strain distributions of these simulated measurements are taken as input data for the
evaluation of stress using the MPA algorithm, which is defined in Chap. 2.3.3.1 with the
equations {2.18} to {2.23}. The calculation of the calibration functions Kx and Ky is
done (a) by using the strain distributions of the reference specimen fT6D20m and (b) by
using the strain distributions determined by the specific geometry of the simulated
measurement. In both these calibration cases, the elastic constants as well as the loading
magnitude are the ones described in Chap. 3.3.2 (E = 210000 MPa, ν = 0.285 and
σc,x = 160 MPa).

The selected calculation procedure in this chapter is a simplified version of the


algorithm used in the MPA software. Some influencing factors are different or simply
omitted. The main differences are:

Order of polynomial approximation: In [2], the calibration strains are approximated


by a 4th polynomial order. This polynomial order is identified there as sufficiently
precise. In this chapter, the calibration strains, which are used for the derivation
dε c ,i (ξ ) / dξ in the calibration functions {2.19} and {2.20}, are similarly calculated by a

closed 4th order polynomial approximation using the strain distributions of the
corresponding calibration model (reference model fT6D20m or one of the three
geometry specific models). Because of the different calibration models, the calculated
calibrations functions of the examples are inherently different compared to the
calibration functions used in the MPA software.

Approximation of simulated measured strain: The stress calculation using equations


{2.21} to {2.23} requires the approximation of the measured strain distribution. For
these simplified examples, the measured strain is calculated, similar to the calibration
functions, using a closed 4th order polynomial approximation, i.e. all simulated strain
values of the measurement εi(ξ) are used for the approximation. In contrast, the MPA
software uses a local polynomial approximation, which means that 3 adjacent strain
131

values of the measurement are approximated by a 2nd polynomial order [33]. This
procedure has the advantage that the local trajectory of the strain distributions is
described more precisely than in the case of a closed polynomial approximation.

Selection of depth increments and maximal depth: The selected depth increments as
well as the selected maximal depth, which are used for the polynomial approximation,
influence directly the calculated coefficients of the polynomial. This means that
different depth increments as well as a different maximal depth lead to different
calibration coefficients. In this chapter, all polynomial approximations (calibration
strains and simulated measurement strains) are calculated up to a maximal depth of
z = 1 mm. The used depth increments are:

• 0.0, 0.02, 0.04, 0.06, 0.08, 0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9, 1.0 for
fT6D20m, fT1.5D20m and hchR6m
• 0.0, 0.04, 0.08, 0.14, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9, 1.0 for hcR3m.

Strain smoothing: The authors of the MPA algorithm suggest an initial smoothing of
the measured strain distribution in order to perform the subsequent strain differentiation
with an appropriate accuracy [33]. This smoothing has an influence on the resulting
stress distributions. The examples in this chapter were calculated without a smoothing
of the measured strains.

Calibration hole diameter vs. measurement hole diameter: The MPA software
provides a set of calibration functions related to a specific hole diameter. In cases in
which the measured hole diameter diverge from one calibration hole diameter, the
calibration function is calculated by linear interpolation between two sets of calibration
functions. This calculation step is omitted in these examples because the calibration
hole diameter is assumed to be equal to the measurement hole diameter with
dc,0 = d0 = 1.8 mm.

The next figures display the results for each simulated measurement, which are two
stress distributions in x-direction. One was evaluated with the reference calibration
(dotted lines) and one was evaluated with the geometry specific calibration (continuous
lines). These calculated distributions are compared to the initial stress distribution
132

within the hole volume (dot and dash line) before starting the drilling process. Fig. 5.31
shows the calculated stress distributions calculated for the models fT1.5D20m (black
lines) and hcR3m (gray lines) as well as the stress deviations (in per cent) related to the
initial stress value of 100 MPa.

Fig. 5.31: Calculated stress distribution and stress difference for models fT1.5D20m and hcR3m
using a simplified MPA algorithm based on calibration functions which were
determined with (a) reference model and (b) geometry specific model

The stress calculation based on the reference calibration lead in both cases to
considerable stress differences. In the case of model fT1.5D20m, the stress differences
are almost constant over the hole depth with values between 24 % and 30 %. The stress
deviation for the model hcR3m starts at z = 0 mm with approx. 8 %, reaches a
maximum of approx. 48 % at z = 0.8 mm and falls at z = 1 mm to a stress difference
value of approx. 35 %. The stress evaluation using the geometry specific calibration
minimizes the stress difference for a hole depth range of z = 0 - 0.6 mm. In the case of
model fT1.5D20m, the stress difference is between z = 0 mm and z = 0.6 mm less than
6 %. The stress results of model hcR3m start at z = 0 mm with a stress difference of
approx. 10 %, falls to a stress difference of approx. 0 % at z = 0.1 mm and increase
slightly to a difference value less than 6 % at z = 0.6 mm. After z = 0.6 mm, both stress
difference distributions increase highly up to values of approx. 100 % and 46 % for the
models fT1.5D20m and hcR3m, respectively.
133

450

400

350

300
hchR6m (spec. calib.)
σx [MPa]
250
hchR6m (ref. calib.)
200

150
σ x (z)-FEM
100

50

0
0 0.2 0.4 0.6 0.8 1
z [mm]

Fig. 5.32: Calculated stress distribution for model hchR6m using a simplified MPA algorithm
based on calibration functions which were determined with (a) reference model and (b)
geometry specific model
150

100

hchR6m (ref. calib.)


50
Diff σx [%]

hchR6m (spec. calib.)


0

-50

-100
0 0.2 0.4 0.6 0.8 1
z [mm]

Fig. 5.33: Calculated stress difference for model hchR6m using a simplified MPA algorithm based
on calibration functions which were determined with (a) reference model and (b)
geometry specific model

In principle, the results obtained with model hchR6m are similar to the previous results,
as can been seen in Fig. 5.32 and Fig. 5.33 for the absolute stress distribution and the
distribution of the stress deviations, respectively. The initial stress distribution within
the hole volume starts with a stress value of 90 MPa at z = 0 mm and increases almost
linearly with increasing hole depth to a value of approx. 200 MPa at z = 1 mm (dot and
dash line in Fig. 5.32). The stress deviations calculated with the reference calibration
start at z = 0 mm with a stress difference of approx. 8 %, reach a maximum of 57 % at
134

z = 0.2 mm and fall up to a stress difference of approx. -105 % at z = 1 mm. In contrast,


the stress differences with the geometry specific evaluation are less than 5 % between
z = 0 mm and z = 0.5 mm. The distribution of the stress deviation shows a remarkable
S-shaped oscillation between z = 0.7 mm and z = 1 mm with a maximal value of
approx. 134 % at z = 0.8 mm and a minimum of -95 % at z = 0.9 mm.

Conclusion: The results of the simplified version of the MPA algorithm in this chapter
show that the stress deviation using a geometry specific calibration is significantly
minimized for hole depths up to 0.5 mm. At deeper hole increments, the stress
distributions show relevant discrepancies to the initial stress distributions.

5.6 Summary of Discussion of Calibration Results


The discussion of the calibration results covered two main topics which were the
uncertainty of the calibration and the influences of the geometrical parameters on the
stress evaluation using the hole drilling method and the differential MPA algorithm.

The uncertainty of the numerical calibration was discussed comparing selected results
of the developed models with results taken from literature. In the case of the
experimental results, affecting factors of the experimental accuracy were listed in order
to analyze the accordance between the numerical and the experimental calibration. The
numerical and experimental results showed a fairly good accordance for the selected
examples although not all described experimental uncertainty sources were taken into
account for the comparison.

In introducing examples of the second part of the discussion, it was demonstrated that
the stress evaluation with actual stress calculation software on specimens with non-
reference geometry leads to stress differences related to the reference evaluation.

The influence of the geometry parameters thickness T, distance hole-edge D and


cylinder radius R was analyzed concerning the evaluation of stress and the possible
stress differences. The results of the calibrations of these uniaxially loaded specimens
showed that the boundary condition thickness and distance hole-edge may be minimized
135

and redefined to T = 1.66 x d0 and D = 1.94 x d0, respectively. In the case of the
cylinder radius, the boundary condition may not be minimized.

The last subject of this chapter considered the reduction of stress differences using
geometry specific calibration functions. The exemplary calculations with a simplified
differential MPA algorithm showed that a reduction of the stress difference is, in
principle possible, although the calculated distribution diverges at deeper hole
increments from the initial distribution.

The next chapters describe an implementation of geometry specific calibration functions


into evaluation software using the MPA algorithm (Chap. 6). A parametric study of the
previously described influencing factors on the results using the MPA algorithm is then
carried out, in order to analyze a possible optimization of the stress calculation using
geometry specific evaluation (Chap. 7).
136

6 Implementation of Geometry Specific Calibration


Functions into Evaluation Program
6.1 Prefacing Remarks
The previous chapter shows, among other things, that the stress evaluation using
specific hole-drilling calculation software on specimens with non-reference geometry
leads to deviations from the initial stress state. Moreover, it was demonstrated that these
stress deviations could be partially minimized, especially for near surface depth
increments, by using a simplified differential MPA method together with geometry
specific calibration functions. The exemplarily used evaluation software packages in
Chap. 5.3 are valid for specific geometrical boundary conditions. Thus, there is no
explicit consideration of the component’s shape when evaluating a part which violates
the geometrical boundary conditions.

This chapter describes the most important functional blocks of the implementation of
geometry specific calibration functions into a hole-drilling evaluation program
according to the differential MPA algorithm. For this purpose, a prototype program was
developed using the numerical computing environment “Matlab” (MathWorks). The
calculation process can be divided in two main parts which are, firstly, the modification
and differentiation of the strain depth distributions and, secondly, the stress calculation.
In the presented prototype program, the stress calculation can be optionally carried out
by using three different sets of calibration functions:

• MPA standard: Calibration functions for a reference geometry according to [2]


were determined for a specific strain gage rosette and different hole diameters
d0. They are implemented as fixed values within the program.
• MPA expanded: Calibration functions calculated in this work using the models
of the flat tensile specimens. They consider the hole diameter d0, the component
thickness T and the distance hole-edge D. The calibration functions for the
CEA-XX-062-UM rosette are also implemented as fixed values within the
program.
• MPA specific: On-line calculation of specific calibration functions from
imported calibration strain distributions which are, possibly, determined using
non-reference geometries.
137

Fig. 6.1: Block diagram for the calculation of residual stress using the hole-drilling method and
the differential MPA evaluation method
138

The input and output as well as the main functions of each block are described in the
following subchapters. In order to provide an overview of the main functional blocks of
the program, a rough sketch of the calculation process is shown in the diagram of
Fig. 6.1. The aim of this chapter is to describe the background of the stress calculation
for the subsequent parametric study in Chap. 7 and is not an exhaustive software
description.

6.2 Modification and Differentiation of Measured Strain


Distributions
6.2.1 Block 1 - Import of Input File
Input: input file
Output: parameters of measurement and material; component geometry;
measured strain depth distributions.
The first block of the program imports the input file for the stress calculation. This file
is the result of the hole-drilling measurement and contains the measurement parameters
and the registered strains at each drilled hole depth increment (s. Fig. 6.2).

%Measurement fT1.5D20 Al %
(0= Blind Hole; 1= Trough Hole)
0
(Strain Gage Rosette)
4
(Hole Diameter)
1.8
(Young-Modulus)
70000
(Poisson Ratio)
0.33
(Thickness)
1.5
(Distance Hole-Edge)
20
*************************************
z e1 e2 e3
0 0.0000 0.0000 0.0000
0.02 0.8426 -2.4529 -5.7483
0.04 1.6873 -5.2704 -12.228
0.06 2.5460 -8.3994 -19.344
0.08 3.4259 -11.807 -27.041
0.1 4.3329 -15.458 -35.249
...
0.9 80.308 -159.63 -399.56
1 93.886 -164.38 -422.66

Fig. 6.2: Input File: geometry parameters, material parameters and measured strain depth
distribution
139

The first line between the percent signs of the input file is a commentary. The
subsequent lines specify the shape of the hole (blind-hole vs. through hole), the used
strain gage rosette, the hole diameter d0, the material parameters (E, ν) and the
geometrical parameters T and D. The last lines after the asterisks list the drilled depth
increments and the relieved strain of each strain gage.

6.2.2 Block 2 - Parameter Modification


Input: parameters of measurement and material; component geometry
Output: modified parameters of measurement, material and geometry (optional)
The second block of the program shows the measurement parameters and allows
optionally their modification within a special interactive program window (s. Fig. 6.3).
This is useful in cases in which some parameters were wrongly stored in the input file
or a new input file with different parameters should be created for subsequent
calculations.

Fig. 6.3: Program window: modification of parameters of measurement.


140

6.2.3 Block 3 - Strain Modification


Input: measured strain depth distributions
Output: modified measured strain depth distributions (optional)
The third block of the program shows the measured strain depth distribution and allows
optionally their modification within the interactive table on the left side of the program
window in Fig. 6.4. The modification in this case means a manual change of the strain
values or the omission of entire lines of the strain input.

6.2.4 Block 4 - Strain Smoothing


Input: measured strain depth distributions
Output: smoothed strain depth distributions

Fig. 6.4: Program window: modification of strain depth distributions and strain smoothing.

Measured strain depth distributions scatter due to the random uncertainty of the
measuring technique. This fact should be especially considered if the strain distributions
are differentiated in order to calculate residual stress depth distributions. In some cases,
the scatter leads to numerical instability. For this reason, the objective of the smoothing
algorithm is the minimization or omission of the statistical scatter. Outliers due to
141

systematic errors are not count as part of the scatter and should be eliminated before the
actual calculation. In this work, two smoothing algorithms, according to [33], are
implemented in the prototype program: a weighted average algorithm and an algorithm
using compensative cubic spline functions.

The first method, the weighted average algorithm, minimizes the scatter, preserves the
local information of the distribution, ensures the continuity of the connection values, is
numerically stable and it is simply to describe in detail. The smoothed strain value
ε ( z i ) is the weighted average {6.1} of three adjacent values ε ( z i −1 ) , ε ( z i ) , ε ( z i +1 ) and
the intermediate averages ε i12 , ε i 23 {6.2}. The weight factors are set in [2.31] with
fixed values {6.3}.

{6.1} ε ( z i ) = ci −1ε ( z i −1 ) + ci12 ε i12 + ci ε ( z ) + ci 23ε i 23 + ci +1ε ( z i +1 )

{6.2} ε i12 = 0.5 × [ε (z i −1 ) + ε (z i )] and ε i 23 = 0.5 × [ε (z i ) + ε (z i +1 )]

{6.3} ci −1 = 0.1 ; c12 = 0.2 ; ci = 0.4 ; c 23 = 0.2 ; ci +1 = 0.1

The first value ε ( z1 ) as well as the final value ε ( z n ) are calculated considering only the
next adjacent value {6.4} and the weight factors in {6.5} according to [2.31].

{6.4} ε ( z1 ) = c1ε ( z1 ) + c 2 ε (z 2 ) and ε (z n ) = c n ε (z n ) + c n −1ε (z n −1 )

{6.5} c1 = 0.95 ; c 2 = 0.05 ; c n = 0.9 ; c n −1 = 0.1

The second smoothing algorithm taken from [33] is described in detail in [60]. It bases
on compensative cubic spline functions. In this calculation, the measured strain
distribution is smoothed piecewise by cubic polynomial functions. The second
derivations of the splines are set equal to zero at the endpoints of the interval. The
advantage of this method is that it uses a compensating factor (smoothing factor fil)
which can be manually set and allows the online adjustment of the smoothed strain
distributions. As a matter of course, the consequences of the setting of the compensative
142

parameter should be verified by the user, as a large fil factor linearizes the trajectory of
the strain distribution [33].

The right part of the program window of Fig. 6.4 shows exemplarily a diagram with
measured strains (continuous lines) and the smoothed values (asterisks). The smoothing
method can be selected beneath this diagram. A listing of both smoothing procedures
can be found in the appendix Chap. 10.

6.2.5 Block 5 - Storage of Modified Input File


Input: modified measurement parameters and modified strain distributions
Output: new input file
This optional function is necessary if the measured strains and the measurement
parameters were modified and this information should be stored. A file can then be
created by writing a new input file or overwriting the old input file.

6.2.6 Block 6 - Local Polynomial Approximation of Measured Strains


Input: measured and smoothed strain distributions, hole diameter d0
Output: differentiated strain values dεi(ξ)/dξ
In order to provide differentiable functions εi(ξ) (with ξ = z/d0) the measured and
smoothed strain distributions can be approximated with polynomials.

One possibility is the close approximation of the strain distribution, i.e. all measured
points of the curve are simultaneously used for the determination of the coefficients. On
the one hand, this has the advantage that only a small number of coefficients have to be
calculated. On the other hand, the disadvantage of this procedure is the loss of local
information due to the increase of the standard deviation. In addition, an overshoot may
occur, e.g. in cases where the number of measurement points is small or the measured
strains have an alternating distribution against the normalized hole depth.

In order to avoid these disadvantages, an alternative procedure is proposed in [33].


Thus, the measured strain distributions are locally approximated using a quadratic
ansatz {6.6}. For three adjacent points (εi-1, εi and εi+1) a system of equations {6.7} can
be set up. Thus, the polynomial coefficients a0, a1 and a2 can then be determined by
143

means of least squares. The differentiated strain values dεi(ξ)/dξ, which are used in the
stress equations {2.21} to {2.23}, are calculated using the derivation of the {6.8}.

{6.6} ε i = a 0 + a1ξ + a 2ξ 2

ε i −1 = a 0i + a1iξ i −1 + α 2iξ i2−1

{6.7} ε i = a 0i + a1iξ i + α 2iξ i2

ε i +1 = a 0i + a1iξ i +1 + α 2iξ i2+1

dε i (ξ )
{6.8} = a1 + 2a 2ξ i

This procedure, which is implemented in the prototype program, is applied for all points
within the strain distribution with the exception of the first and the last point. A program
listing is included in the appendix Chap. 10.

6.3 Stress Calculation


6.3.1 General Aspect of Stress Calculation in Prototype Program
The blocks “MPA Standard”, “MPA Expanded” and “MPA Specific” calculate the
stress distributions in the direction of the single strain gages (σ1, σ2, σ3) and the
distributions of the principal stresses (σmax, σmin) according to the equations {2.21} to
{2.24} (s. Chap. 2.3.1). The major difference of each block is the used sets of
calibration functions. The blocks “MPA Expanded” and “MPA Specific” were
developed during this work in order to provide two different calculation procedures
which consider the component geometry.

6.3.2 Block 7 - Stress Calculation “MPA Standard”


Input: differentiated strain values dεi(ξ)/dξ; parameters of measurement and
material
Output: σ1(z), σ2(z), σ3(z), σmax(z), σmin(z) (s. {2.21} to {2.24})
144

This block no. 7 (“MPA Standard”) uses the calibration functions of the original MPA
algorithm. They are calculated according to {2.19} and {2.20} (s. Chap. 2.3.1). The 4th
order polynomial coefficients of the calibration strain functions εc,x {6.9} and εc,y
{6.10} were determined for different rosettes types using uniaxially loaded finite
element models with σc,x = 100 MPa, σc,y = 0 MPa, E = 200000 MPa and ν = 0.3 [2].

{6.9} ε c , x = a 0 + a1ξ + a 2ξ 2 + a 3ξ 3 + a 4ξ 4

{6.10} ε c , y = b0 + b1ξ + b2ξ 2 + b3ξ 3 + b4ξ 4

The sets of calibration coefficients, which are included as fixed values within the
program, consider five different hole diameters by means of the ratio dn {6.11}.

dm
{6.11} d n =
d0

The implemented set of calibration strain functions covers hole diameters between
1.50…1.97 mm. In case of the hole-drilling rosette CEA-XX-062UM-120, the mean
rosette diameter is dm = 5.13 mm. If the measured hole diameter (e.g. d0,v) is between
two set hole diameters (e.g. d0,u and d0,w), the specific calibration coefficient av is
calculated using linear interpolation {6.11} between the fixed calibration coefficients au
and aw {6.12}.

a w − au
{6.12} a v = au + (d n,v − d n,u ) with dn,u < dn,v < dn,w
d n , w − d n ,u

6.3.3 Block 8 - Stress Calculation “MPA Expanded”


Input: differentiated strain values dεi(ξ)/dξ, parameters of measurement and
material, component geometry parameters T and D
Output: σ1(z), σ2(z), σ3(z), σmax(z), σmin(z) (s. {2.21} to {2.24})
This block calculates the stress distributions identically as in the previous “MPA
Standard” block. In contrast to the “Standard” block, the “Expanded” block signifies
145

that the calibration coefficients of the 4th order polynomials considers not only the hole
diameter d0 but also the component thickness T and the distance hole-edge D for the
hole-drilling rosette CEA-XX-062UM-120:

• Thickness: T = 1; 1.5; 2; 3; 6 mm
• Distance hole-edge: D = 2; 3.5; 5; 10; 20 mm
• Hole diameter d0 = 1.7; 1.75; 1.8; 1.85; 1.9 mm

The calibration coefficient sets are included as fixed values within the program and
were determined in this work using the models of the flat tensile specimens of Type 1 as
described in Chap. 3.3.1. It might happen that the geometrical parameters of the
measurement (T, D, d0) do not exactly coincide with the fixed geometrical parameters
of the calibration coefficient sets (Tc, Dc, dc,0). In this case, the actual calibration
coefficients are determined by linear interpolation between the values of the calibration
coefficients. This process is visualized with a schematic diagram in Fig. 6.5. Thus, two
set of calibration coefficients C1(Tc, Dc, dc,0,u) and C2(Tc, Dc, dc,0,w) span two surfaces in
a three dimensional diagram. The actual calculation coefficient aj(T ,D , d0) is then
interpolated using the single values of C1 and C2.

Fig. 6.5: Schematic illustration of determination of actual calibration coefficient aj using


interpolation in block “MPA-Expanded”
146

The advantage of this calculation procedure with fixed values is that the user only has to
include the geometry parameters T, D and d0 into the input file or in the parameter
modification block (Block 2) in order to consider the shape of the component. The
disadvantage is that the considered component geometry is limited by the sets of
determined specific calibration functions. As a start, a three dimensional interpolation
aj = f(T,D,d0) was chosen but in principle the computing environment MATLAB allows
moreover an interpolation with more than three variables, e.g. if the geometry parameter
curvature should be additionally considered. An exemplarily program listing of this
interpolation procedure is included in the appendix Chap. 10.

6.3.4 Block 9 - Stress Calculation “MPA Specific”


Input: differentiated strain values dεi(ξ)/dξ, parameters of measurement and
material, parameters of specific calibration, calibration strain
distributions, calibration stress distributions
Output: σ1(z), σ2(z), σ3(z), σmax(z), σmin(z) (s. {2.21} to {2.24})
The stress calculation in this block is identical to blocks 7 and 8. The calibration
functions are calculated from the specific calibration files, which are previously
imported and approximated in block 10 and block 11, respectively. In this case,
“specific” means that the calibration was carried out using a component (or specimen)
with a shape, which is identical or similar to the shape of the measured component. The
geometry of the component can be complex and does not comply with the geometrical
boundary conditions. Thus, the resulting calibration strains inherently consider the
specific geometry situation near the measurement point.

One possible application of this calculation procedure is the use for recurring
measurements on complex shaped components. In such a case, the calibration files are
determined once using, e.g. finite element models of the component, in which a hole-
drilling measurement point is modelled. The models should only consider the possible
dimensional range of the introduced hole diameters, so that the approximated functions
could be interpolated for the actual hole diameter of the measurement.
147

6.3.5 Block 10 - Import of Calibration Files


Input: calibration files
Output: parameters of calibration material; calibration strain depth distributions
and calibration stress depth distributions, hole diameters d0.

This block is activated if the stress calculation should be carried out using specific
calibration functions. It imports one or more (max. five) calibration files in order to
consider different hole diameters. An exemplarily calibration file is shown in Fig. 6.6.
The calibration contains information about the used strain gage rosette, the calibration
hole diameter and the elastic parameters of the calibration material. The data columns
after the asterisk are the hole depth, the calibration strains εx and εy, and the calibration
stress σx and σy. The calibration material parameters are passed to block 9 in order to
calculate the calibrations functions Kx and Ky.

%reference calibration fT6D20 %


(Strain Gage Rosette)
4
(Hole Diameter)
1.8
(Young's Modulus)
210000
(Poisson´s Ratio)
0.285
(Mean Diameter Strain Gage Rosette)
5.13
*************************************
depth ey e2 ex sx sy
0.00 0.0000 0.0000 0.0000 160 0
0.02 0.4192 -1.319 -3.057 160 0
0.04 0.8413 -2.831 -6.504 160 0
0.06 1.2731 -4.508 -10.29 160 0
0.08 1.7190 -6.332 -14.38 160 0
0.10 2.1822 -8.283 -18.74 160 0
...
0.90 43.642 -84.19 -212.0 160 0
1.00 51.006 -86.64 -224.3 160 0

Fig. 6.6: Calibration File: calibration parameters, calibration strain depth distributions,
calibration stress depth distribution
148

6.3.6 Block 11 - Polynomial Approximation of Calibration Strains


Input: calibration strain distributions, calibration stress distributions, hole
diameters of calibration d0,c
Output: εc,x(ξ), εc,y(ξ), σc,x(ξ), σc,y(ξ)

The imported calibrations strain distributions, as well as the calibration stress


distributions, are approximated in this block with polynomials using ξ = z/d0 and a close
approximation. The polynomial order can be manually set within a program window
and is, by default, of 4th and 1st order for the calibration strain distributions and
calibration stress distribution, respectively. The approximated calibration coefficients of
the functions εc,x(ξ), εc,y(ξ), σc,x(ξ), σc,y(ξ) as well as the calibration parameters are
passed to block 9 in order to calculate the calibrations functions Kx and Ky.

6.3.7 Block 12 - Angle Calculation


Input: σ1(z), σ2(z), σ3(z)
Output: φ(z)
The orientation angle φ between strain gage no. 3 and the maximal stress is determined
by calculating an intermediate angle φ* {6.13} and by verifying the relationships {2.9}
to {2.17} (s. Chap. 2.3.1).

1 2 Δσ 2 − Δ σ 3 − Δσ 1 1 Num
{6.13} ϕ ∗ = arctan = arctan
2 Δσ 1 − Δσ 3 2 Den

In {6.13}, calculated stress values in the strain gage direction are used instead of
measured strain values {2.8} for the determination of φ. This means that, although the
measured strain distribution of the input file is constant, the value of the orientation
angle may differ as the stress values are dependent on the previous calculation
procedure (strain smoothing, strain approximation, etc.).
149

6.3.8 Block 13 - Visualization of Results


Input: σmax(z), σmin(z), σ2(z), σ3(z), φ(z)
Output: diagrams of stress distributions and orientation angle φ(z)

This block is added in order to verify the stress results of the calculation before
exporting the data into an output file. In Fig. 6.7, the resulting stress depth distributions
as well as the distribution of the orientation angle against the hole depth of a hole
drilling measurement are exemplarily shown within the diagrams of the program
window.

6.3.9 Block 14 - Export and Storage of Result File


Input: σmax(z), σmin(z), σ2(z), σ3(z), φ(z)
Output: result file
The calculated stress depth distributions as well as the distribution of the orientation
angle against the hole depth can be stored in a result file which is exemplarily shown in
Fig. 6.8.

Fig. 6.7: Program window: plot of calculated stress depth distributions and orientation angle
150

STRESS CALCULATION ---- BLIND HOLE


------------------------------------
Data Titel : Measurement fT1.5D20 Al

Rosette Type : 062 UM


Hole Diameter : 1.80
Young's Modulus : 70000
Poisson's Ration : 0.33

Depth SigMax SigMin Alpha Sig3 Sig1


[mm] [MPa] [Mpa] [°] [Mpa] [Mpa]
0.00 116.75 3.642 -0.144 116.75 3.642
0.02 113.16 5.149 -0.128 113.16 5.149
0.04 110.95 6.544 -0.125 110.95 6.545
0.06 110.57 7.937 -0.141 110.57 7.938
0.08 111.51 9.209 -0.147 111.51 9.210
0.10 113.03 10.24 -0.100 113.03 10.24
0.20 120.89 13.86 -179.7 120.89 13.86
0.30 126.25 16.07 -179.8 126.25 16.07
0.40 126.82 16.02 -0.191 126.82 16.02
0.50 122.97 13.23 -0.216 122.97 13.23
0.60 116.03 8.904 -179.9 116.03 8.904
0.70 107.78 4.043 -179.9 107.78 4.043
0.80 99.557 -2.09 -0.119 99.557 -2.09
0.90 94.134 -9.58 -0.215 94.132 -9.58
1.00 90.911 -19.3 -0.455 90.904 -19.3

Fig. 6.8: Result File: stress depth distributions and distribution of orientation angle against hole
depth
151

7 Evaluation of the Implementation of Geometry Specific


Calibration Files into Residual Stress Calculation Program
7.1 Prefacing Remarks
In this chapter, the implementation of geometry specific calibration files into the
differential MPA algorithm for the calculation of residual stress distributions after a
hole drilling measurement is evaluated. For this purpose, exemplary calculations are
carried out using the previously described prototype program in Chap. 6 in order to
discuss the influencing factors of the algorithm on the stress calculation and to evaluate
the two alternative calculation blocks which consider the component geometry.

The first part of this chapter deals with the stress calculation using the MPA algorithm
and illustrates especially the reason for the oscillating form of the calculated stress
distributions (s. also Chap. 5.3).

The next parts discuss influencing factors of the calculation algorithm on the stress
evaluation by means of a parametric study. Therefore, the second part shows the
influence of the local approximation on the calculated stress results. The influence of
the initial strain smoothing is discussed in the third part. In the fourth part, the order of
the polynomial approximation of the calibration strain distribution is discussed whereas
the effect of the selection of the distribution of the depth increments is shown in the
fifth part.

The last part of this chapter compares results using the three different stress calculation
options of the prototype program. These options are the stress calculation blocks MPA
Standard, MPA Expanded and MPA Specific.

7.2 General Aspects of Calculation


Chap. 5.3 shows, among other things, that small differences compared to the initial
stresses are calculated, even if the geometrical boundary conditions are observed. This
is the case for the calculation using the reference specimen. Moreover, especially in
cases in which the initial stress depth distribution is uniform over the depth, the
calculated stress distribution oscillates around the actual initial stress depth distribution.
152

The origin for the oscillating shape of the calculated stress distribution is found in the
calculation itself. In order to visualize this problem exemplarily, the calibration strain
distribution determined with the reference model fT6D20m is selected in order to
calculate the stress distribution and compare it with the initial stress of σx = 160 MPA.
The stress in x-direction is parallel to the direction of strain gage no. 3 and, therefore,
the stress is calculated using {7.1} (s. in Chap. 2.3.3.1 equation {2.23}).

E ⎡ dε (ξ ) dε (ξ ) ⎤
{7.1} σ 3 (ξ ) = ⋅ ⎢ K x (ξ ) ⋅ 3 + ν ⋅ K y (ξ ) ⋅ 1 ⎥
K (ξ ) − ν K y (ξ ) ⎣
2
x
2 2
dξ dξ ⎦

Young’s modulus (E = 210000 MPa) and Poisson’s ratio (ν = 0.285) are constant values
whereas the differentiated strain distributions dε1(ξ)/dξ and dε3(ξ)/dξ as well as the
calibration functions Kx(ξ) and Ky(ξ) are variable against the depth. The distribution of
the differentiated strain and the distribution of the calibration functions for this example
are shown in Fig. 7.1 in diagram part a) and b), respectively. The strains were
differentiated after carrying out a local polynomial approximation and were not initially
smoothed. The calibration functions were calculated using the MPA Standard block by
interpolating the calibration functions, as the hole diameter (d0 = 1.8 mm) of the
selected reference model is between the nominal hole diameters of two sets of
calibration functions. The calculated stress in x-direction is plotted in diagram part c).
Additionally, values of the differentiated strains, the calibration functions and the
calculated stress at selected depths are listed in Tab. 7.1. It can be seen that the
distribution of the differentiated strains and the distribution of the calibration functions
show an alternate shape. This is the main reason for the resulting alternate shape of the
calculated stress distributions. The calculated stress distribution runs thus around the
value of the initial stress.
depth ξ [-] dε3(ξ)/dξ [-] dε1(ξ)/dξ [-] Kx(ξ) [-] Ky(ξ) [-] σx [MPa]
[mm]
0.0 0.000 -291 x10-6 55 x10-6 -0.39 -0.17 152
0.2 0.111 -393 x10-6 78 x10-6 -0.53 -0.36 156
0.5 0.277 -416 x10-6 111 x10-6 -0.54 -0.48 161
1.0 0.555 -219 x10-6 96 x10-6 -0.29 -0.40 155
Tab. 7.1: Stress calculation using MPA Standard block and reference model fT6D20m:
differentiated strains, calibration functions, calculated stress σx at selected hole depths
153

200
a) differentiated strain dε1/dξ
100

dεi/dξ [μm/m]
-100
z = 0.0 mm z = 0.2 mm z = 0.5 mm z = 1.0 mm
-200

-300

dε3/dξ
-400

-500
0 0.1 0.2 0.3 0.4 0.5 0.6

ξ [−]

-0.1
b) calibration functions

-0.2

-0.3
Ky
Ki [-]

-0.4

Kx
-0.5

z = 0.0 mm z = 0.2 mm z = 0.5 mm z = 1.0 mm


-0.6
0 0.1 0.2 0.3 0.4 0.5 0.6

ξ [−]

170
c) calculated stress

160
σx

150

z = 0.0 mm z = 0.2 mm z = 0.5 mm z = 1.0 mm

140
0 0.1 0.2 0.3 0.4 0.5 0.6

ξ [−]

Fig. 7.1: Stress calculation using MPA Standard block and reference model fT6D20m: a)
Distribution of differentiated strains, b) Distribution of calibration functions, c)
Calculated stress σx
154

For this calculation example using the reference model, the average stress deviation
over the hole depth is approximately 6 % with a maximal stress deviation of
approximately 10 %. A similar observation is also made by the authors of the MPA
method [2]. In conclusion, it should be taken into account that the stress calculation
according to the differential MPA evaluation method implies a priori a level of
uncertainty.

7.3 Influence of local polynomial approximation of Measured Strains

The polynomial coefficients of the measured strain distributions are determined within
the prototype program using a local approximation instead of a close approximation as
was carried out with the simplified MPA algorithm in Chap. 5.5. This current chapter
compares the resulting stress distributions using both approximation procedures. For
this purpose, the exemplarily strain distributions in this chapter are the same as in Chap.
5.5. These are the simulated measurements using the “thin” model fT1.5D20, the “thin”
and “curved” model hcR3m and the “thin” and “curved” model hchR6m with the
measurement point “near the edge” of the ø6 mm center through-hole. The elastic
constants of the models are E = 70000 MPa and ν = 0.285. The models are loaded
externally with a stress of σx = 100 MPa. In case of the models fT1.5D20m and hcR3m
the external load value equals the initial uniform stress at the measurement point. In the
case of the hchR6m model, the initial stress distribution at the measurement point
increases with increasing hole depth. The used approximation procedure of the
measured strains is the only difference between the calculations. The other influencing
factors are constant: the measurement strains were not initially smoothed; the
calibrations functions are of 4th polynomial order and the selected distribution of depth
increments is the same as in the simplified algorithm. The stress calculation for all three
examples was carried out using the MPA Specific stress calculation block. In the case of
this specific model of a flat specimen, a calculation with the MPA Expanded block
leads to the same results.

These examples, i.e. the FEM models and the selection of the influencing factors are
also used in the next chapters, Chap. 7.4, Chap. 7.5 and Chap. 7.6. The only difference
is the variation of the respective influencing factor.
155

200
fT1.5D6m close pol.
approx.
180

160
σx
140
local pol.
approx.
120

100

σx-FEM = 100 MPa


80
0 0.2 0.4 0.6 0.8 1 1.2
z [mm]

Fig. 7.2: Stress distributions using MPA Specific block and model fT1.5D20m considering local
or close polynomial approximation of the measured strain distributions.

Fig. 7.2 shows calculated stress distributions in x-direction for the simulated
measurement with model fT1.5D20m. The black dashed line is the distribution
calculated with local approximation and the grey continuous line is the one calculated
with close approximation. It can be clearly seen that the stress deviations over the hole
depth related to the initial stress distribution σx-FEM is, in this specific case of the thin
flat specimen, less when using the local polynomial approximation of the measured
strains instead of the close polynomial approximation. Especially from a depth of
z = 0.6 mm, the calculation using local approximation is more accurate. Tab. 7.2 lists
the stress deviations at selected depths related to the initial stresses for the examples
discussed in this chapter.
Stress Deviation related to initial stress [%]
Depth fT6D20m hcR3m hchR6m
[mm] local close local close local close
approx. approx. approx. approx. approx. approx.
0.0 -2 0 -60 -11 -54 0
0.2 -1 1 -17 0 0 1
0.5 1 3 0 3 1 5
1.0 0 99 0 46 -23 -56
Tab. 7.2: Stress deviation using MPA Specific block and models fT1.5D20m, hcR3m and hchR6m.
Consideration of local or close polynomial approximation of measured strain
distributions
156

Fig. 7.3 and Fig. 7.4 show the calculated stress distributions for the models hcR3m and
hchR6m, respectively. These calculations were carried out similarly to the previously
described calculation of the fT1.5D20m model. With the exception of the initial depth
increments z ≤ 0.2 mm for model hcR3m, the calculated stress distributions using local
polynomial approximation show less stress deviations than the distributions calculated
with the close approximation. In the case of the model hchR6m, the stress values
calculated with the local approximation show less deviation for depth increments
between of z = 0.1 mm and z = 0.8 mm. Both calculated distributions alternate for the
final depth increments from z > 0.8 mm and have completely erroneous stress values.

160
hcR3m close pol.
approx.
140
local pol.
approx.
120
σx

100

σx-FEM = 100 MPa


80

60
blind hole
with d0
40
0 0.2 0.4 0.6 0.8 1 1.2
z [mm]

Fig. 7.3: Stress distributions using MPA Specific block and model hcR3m considering local or
close polynomial approximation of the measured strain distributions.

The deviations at the first depth increments using the local polynomial approximation
on the curved models hcR3m and hchR6m are caused because the local approximation
uses only three strain values in order to determine the coefficients of the strain
distribution of the measurement. At the first depth increments, the hole is not generated
to the full extent of the hole diameter due to the curvature of the cylindrical specimens.
Hence, at these initial depth increments of the simulation, less material is “removed”.
This causes lower strain values and, thus, the local approximation at the initial
increments calculates strain functions with lower slopes. Consequently, the stress values
at these first depth increments are also lower then the initial values. A close
157

approximation averages this initial strain information because all strain values of the
depth distribution contribute to the determination of the strain function.

450
hchR6m local pol.
400 approx.
350 close pol.
approx.
300

250
σx

200

150

100
blind hole σx-FEM
50 with d0

0
0.00 0.20 0.40 0.60 0.80 1.00 1.20
z [mm]

Fig. 7.4: Stress distributions using MPA Specific block and model hchR6m considering local or
close polynomial approximation of the measured strain distributions

The high stress deviations between z = 0.8 mm and z = 1.0 mm of model hchR6m can
be related to the principle of the differential calculation algorithm itself, which does not
consider the partially relaxed stresses above the hole bottom. Deviations at the final
depth increments are found in all three examples but in the case of the hchR6m model,
the stress state is more complex than the other two examples: it has a biaxial initial
stress state which increases against the hole depth.

In conclusion, it can be said that the stress results using a local polynomial
approximation of the measured strains show for the three presented examples in general
less deviation from the initial stress distribution than a calculation using the close
approximation. Deviations related to the local polynomial approximation appear at the
initial depth increments for the curved models. In these cases, the stress values
calculated at depth increments, in which the hole does not have an ideal blind hole
shape with d0, should not be taken into account. For all three examples and for the two
approximation procedures, stress deviations related to the calculation principle appear at
the final depth increments.
158

7.4 Influence of Strain Smoothing

The initial strain smoothing before the differentiation of the measured strain influences
the stress results. To show this, the first example in this chapter uses the thin flat model
fT1.5D20m which has an initial stress of σx = 100 MPa (E = 70000 MPa and ν = 0.33).
Only the smoothing parameters are varied, the other influencing factors on the stress
calculation are kept constant (local polynomial approximation, 4th order polynomials,
equal distribution of depth increments as in the previous examples). First of all, Fig. 7.5
shows distributions of strain ε3 against the normalized depth using different smoothing
procedures for the first three depth increments (z = 0 mm, 0.02 mm, 0.04 mm). These
three values calculate directly the stress values at z = 0 mm, and z = 0.02 mm and
contribute to the calculation of the following stress value, because of the local
polynomial approximation. The diagram shows the distribution of the non-smoothed
strain, the strain smoothed using the weighted average algorithm and strain distributions
using compensative cubic splines functions with a smoothing factor of fil = 0.2, 0.5 and
1.0.

5 fil = 1.0
fil = 0.5
fil = 0.2
0

no smooth
ε3 [μ/m]

-5
weight.

-10

z = 0.0 mm z = 0.02 mm z = 0.04 mm


-15
0 0.011 0.022

ξ [-]

Fig. 7.5: Strain ε3 against normalized depth using different smoothing parameters

It can be seen that the smoothing procedures changes the initial strain values. The non-
smoothed strain of strain gage no. 3 starts exactly at zero and decreases with depth. The
strain distribution smoothed with the weighted average starts slightly beneath the non-
smoothed distribution whereas the distributions smoothed using compensative cubic
159

splines start above the non-smoothed distribution. In the case of the smoothing using
compensative splines functions, the initial strain value increases with the increasing
smoothing factor fil. In the diagram of Fig. 7.5, the different smoothed distributions run
almost together at ξ = 0.022 but the change of the initial strain values causes a change in
the slopes of the strain distributions. This means that, in this case of higher slopes of the
strain distribution, the differentiated strains are determined also higher and consequently
the calculated stress values increase. This is shown in Fig. 7.6 diagram a), in which the
calculated stress in x-direction is plotted between z = 0.0mm and z = 0.04mm for this
first example of model fT1.5D20m. The non-smoothed stress distribution starts at
σx(0) = 98 MPa. The stress distribution calculated with the highest smoothing factor fil
starts with the highest stress value of σx, fil = 0.1(0) = 152 MPa. With decreasing
smoothing factor, the initial stress values also decrease. The surface stress value
calculated using the weighted average algorithm is 6 % lower than the non-smoothed
surface stress value.

The previous diagram a) is expanded in Fig. 7.6 in diagram b) up to a hole depth of


z = 1.0 mm in order to show the whole resulting stress distributions using the different
smoothing procedures. The stress distributions, which were smoothed with
compensative splines functions, run almost equal to the non-smoothed stress
distribution between z = 0.2 mm and z = 0.8 mm. These smoothed distributions disperse
at depth increments of z > 0.8 mm. The highest stress deviation at the final depth
increments are found on the distribution with the highest smoothing factor of fil = 1.0.

The stress distribution, which was smoothed with the weighted average algorithm,
shows almost a good accordance with the non-smoothed stress distribution with the
exception of the values between z = 0.08 mm and z = 0.2 mm. At these depth
increments the calculated stress values increase up to σx,weight(0.1) = 173 MPa. The
reason for this stress deviation is that the depth distribution of the simulated
measurement changes at z = 0.1 mm and has larger depth increments Δz = 0.1 mm
instead of Δz = 0.02 mm. Consequently, this change increases the relieved strain values.
For example, the smoothed strain value at z = 0.1 mm is averaged in the calculation
using the non-smoothed values ε(0.08) = -27 μm/m, ε(0.1) = -35 μm/m and the
significantly higher strain value of ε(0.2) = -82 μm/m. Finally, the strain differentiation
160

also leads also to higher values. The weighted average algorithm only considers the
strain values without any information of the actual depth and should be applied in a
combination of measured strains with uniform distribution of depth increments.

160
a)

fil = 1.0
140

fil = 0.5
σx [MPa]

120
fil = 0.2

no smooth
100
weight.

80
0 0.02 0.04
z [mm]

180
b)

160

fil = 1.0

140
σx [MPa]

weight.
fil = 0.5

120
fil = 0.2

100

no smooth
80
0 0.2 0.4 0.6 0.8 1
z [mm]

Fig. 7.6: Calculated Stress distribution using different smoothing parameters: a) depth z = 0-
0.04 mm; b) depth z = 0-1.0 mm

The previous example shows the influence of the smoothing parameters on the stress
calculation using the results of the fT1.5D20m model. This FEM result is inherently
smooth and, consequently, less stress deviation is calculated using the non-smoothed
strain distributions. The next example shows the importance of smoothing in practice.
Fig. 7.7 shows the stress results taken from a hole-drilling measurement on a tempered
161

42CrMo4 steel specimen, whose surface was shot-peened. The input strain distributions
file for the stress evaluation was the same for all three calculated stress distributions.
The stress distribution with the black continuous line was calculated without smoothing.
The distributions with the grey continuous line and the black dashed line were
calculated using different smoothing factors fil = 0.1 and fil = 0.5, respectively. In order
to have comparative values, a stress distribution on the same shot-peened surface was
determined using X-ray diffraction and additionally included in Fig. 7.7 with a thin grey
continuous line with circles. The effect of shot-peening related to the introduced
residual stresses on metallic surfaces has been investigated in several works [61]. Thus,
compressive residual stresses are commonly expected at the surface. Furthermore, a
maximum of the compressive residual stress is often found beneath the surface of the
peened material. The effective range of shot-peening on the surface of 42CrMo4
decreases at approximately z = 0.1-0.2 mm [61]. It can be seen that a stress evaluation
without the initial smoothing of the measured strains leads to a significantly lower stress
value at the surface (σx,drill.(0) = -47 MPa) than the one measured with X-ray diffraction
(σx,diff.(0) = -666 MPa). Moreover, the non-smoothed stress distribution shows between
z = 0.04 mm and z = 0.1 mm an oscillating trajectory whereas the distribution
determined with X-ray diffraction has a maximum at z = 0.06 mm of σx,diff.(0.06) = -
727 MPa. The smoothed stress distribution with fil = 0.1 starts at σx,drill.(0) = -497 MPa
and the calculated distribution has a characteristic shape with a maximum of
σx,drill.(0.08) = -657 MPa at z = 0.08 mm. In case of the calculated stresses with fil = 0.5,
the distribution apparently shows a linear decreasing distribution starting at the surface
with σx,drill.(0) = -670 MPa and showing no compressive stress maximum.
162

no
smooth
-200
σmin [MPa]

-400
fil = 0.1

fil = 0.5
-600

X-Ray
-800
0.0 0.1 0.2 0.3 0.4
z [mm]

Fig. 7.7: Stress distribution of a shot-peened 42CrMo4 steel specimen: Comparison of a hole
drilling measurement with stress evaluation using different smoothing factors and an X-
ray diffraction measurement.

These two examples should underline, on the one hand, the effects of the initial strain
smoothing after carrying out a real hole-drilling measurement. On the other hand, the
residual stress distribution is not known before the measurement. Thus, the amount of
smoothing to be used is difficult to define. In these cases, good engineering experience
combined with knowledge of the origin and consequences of residual stresses is of
especial importance.

7.5 Influence of Order of Polynomial Approximation of Calibration


Strains

The calculations in this chapter, in which the order of the polynomial approximation of
the calibration strains is varied, are carried out using the results of the simulated
measurements of models fT1.5D20m, hcR3m and hchR6m. All simulation parameters
and the other influencing factors are constant (s. previous chapters). The results are
shown in the following diagrams of calculated stress distributions: Fig. 7.8 shows the
results calculated with model fT1.5D20m, Fig. 7.9 for model hcR3m and Fig. 7.10 for
model hchR6m. In addition to the diagrams, Tab. 7.3 lists the calculated stress
deviations related to the initial stress distribution σx-FEM for all three examples at
selected hole depths.
163

140
fT1.5D20m 3.th Order
σx-FEM = 100 MPa 6.th Order

120 2.th Order

σx [MPa] 5.th Order

100 4.th Order

80

60
0 0.2 0.4 0.6 0.8 1 1.2
z [mm]

Fig. 7.8: fT1.5D20m: calculated stress distribution using different orders of polynomial
approximation of calibration strains

The results of the three examples show that a stress calculation using a 2th order
polynomial approximation of the calibration strains leads to the highest stress
deviations. The stress results using a 3th order polynomial approximation also do not
reproduce sufficiently the initial stress distributions. The results using a 4th order, 5th
order and 6th order polynomial approximation, respectively, reproduces on average the
initial stress distribution with comparable results. The stress deviations using these three
polynomial orders are on the whole lower than the stress deviations calculated with the
2th and 3th order polynomial. Problems appear at the initial depth increments for the
curved models and at the final depth increments. Especially for model hchR6m, high
stress deviations are calculated at the depth increments between z = 0.8 mm and
z = 1 mm independently of the selected order of the polynomial approximation of the
calibration strain (s. Fig. 7.10, part b)). The deviations at the initial depth increments of
the curved models as well as the deviations at the final depth increments were discussed
in the previous chapter Chap. 7.3).
164

140
hcR3m
σx-FEM = 100 MPa 3.th Order
2.th Order
120 5.th Order

4.th Order
σx [MPa]

100

6.th Order
80

60
0 0.2 0.4 0.6 0.8 1 1.2
z [mm]

Fig. 7.9: hcR3m: calculated stress distribution using different orders of polynomial
approximation of calibration strains

Depth fT1.5D20m: Stress Deviation related to initial stress [%]


th
[mm] 2 Order 3th Order 4th Order 5th Order 6th Order
0.0 -52 -28 -2 1 -2
0.2 -3 2 -1 -1 -1
0.5 16 0 0 1 1
1.0 -45 -96 -2 14 33
Depth hcR3m: Stress Deviation related to initial stress [%]
[mm] 2th Order 3th Order 4th Order 5th Order 6th Order
0.0 -85 -73 -60 -60 -55
0.2 -21 -15 -17 -18 -18
0.5 18 0 0 1 1
1.0 -45 -183 0 0 -12
Depth hcR6m: Stress Deviation related to initial stress [%]
[mm] 2th Order 3th Order 4th Order 5th Order 6th Order
0.0 -88 -71 -54 -52 -50
0.2 -6 3 0 -1 -1
0.5 9 -1 5 5 5
1.0 -124 -17 -23 -17 -22
Tab. 7.3: Stress deviation using MPA Specific block and models fT1.5D20m, hcR3m and hchR6m.
Consideration of different orders of polynomial approximation of calibration strains
165

240
a)
4.th - 6.th Order
hchR6m
200 3.th Order
2.th Order
σx-FEM
160
σx [MPa]

120

80

40
0.00 0.20 0.40 0.60 0.80 1.00
z [mm]

1000
b) 4.th , 5.th Order
hchR6m
6.th Order

500
2.th Order
σx [MPa]

σx-FEM

3.th Order
-500
0.00 0.20 0.40 0.60 0.80 1.00 1.20
z [mm]

Fig. 7.10: hchR6m: calculated stress distribution using different orders of polynomial
approximation of calibration strains; a) depth z = 0-0.8 mm; b) depth z = 0-1.0 mm

As a conclusion it can be said that a 4th order polynomial approximation of the


calibration strain is sufficient in order to calculate residual stresses using the MPA
evaluation algorithm. This statement can be also found in [2]. The calibration
coefficients are calculated within the prototype program using close approximation. As
a further prospect, the effect of a possible local polynomial approximation of the
calibration strains should be verified. The way to solve this problem is the correct
assignment of the locally determined calibration functions with the measured strains as
both distributions, i.e. measured strains vs. calibration strains, may have different depth
increment distributions in the measurement practice.
166

7.6 Influence of Selected Distribution of Depth Increments

In this chapter the distribution of the depth increments of the measurement strain input
file as well as the distribution of the calibration strain file, is alternately varied, in order
to discuss its influence on the calculated stress distribution. The exemplarily used model
in this chapter is model hcR3m. All simulation parameters and influencing factors are
equal to the ones described in the previous chapter Chap. 7.4 and Chap. 7.5. The
variants of the exemplarily used depth increment distribution are listed in Tab. 7.4.
According to this, the initial variant V0 contain all strain values at the modelled depth
increments of model hcR3m. The first variant V1 omits the initial strain values at
z = 0.04 mm and z = 0.08 mm whereas the second variant V2 omits the initial strain
values as well as the final strain values at z = 0.9 mm and z = 1.0 mm. Variant no. 3
omits only the final strain values.

Variants Distribution of depth increments [mm]


Initial
0.0 0.04 0.08 0.14 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
Variant (V0)
Variant no. 1
0.0 - - 0.14 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
(V1)
Variant no. 2
0.0 - - 0.14 0.2 0.3 0.4 0.5 0.6 0.7 0.8 - -
(V2)
Variant no. 3
0.0 0.04 0.08 0.14 0.2 0.3 0.4 0.5 0.6 0.7 0.8 - -
(V3)
Tab. 7.4: Distribution of depth increments for different variants of measurement strains input files
(M) or calibration strain files (C)
167

200

180
M_V1
M_V2
160

σx [MPa] 140

120

100

80
σx-FEM = 100 MPa
60 M_V0
M_V3
40
0 0.2 0.4 0.6 0.8 1 1.2
z [mm]

Fig. 7.11: Calculated Stress Distribution using different variants of depth increment distributions
within the measurements input file; calibration file has initial variant V0

Fig. 7.11 plots the calculated stress results for model hcR3m using different variants of
measurement input files. In this calculation, the calibration file has the initial variant
V0. The distribution of variant V0 is the same result as in the previous chapter: it has
significant stress deviations between depth increments of z = 0 mm and z = 0.2 mm.
This calculated stress distribution sufficiently reproduces the initial stress of σx-
FEM = 100 MPa, between depth increments of z = 0.2 mm and z = 0.8 mm. A stress
deviation of approx. 8 % is found at z = 0.9 mm.

The calculations using the first (V1) and the second (V2) variants show also deviations
at the initial depth increments with the difference that the surface stress values are
approx. 90 % higher than σx-FEM. The reason for this is the significant change of the
strain values, which occurs from z = 0.0 mm and z = 0.14 mm, that calculates a higher
slope of the measurement strain functions. After z = 0.2 mm, both resulting stress
distributions (V1 and V2) run almost equal to the one of the initial variant V0 (variant
V2 runs, of course only, up to z = 0.8 mm). The results calculated with the variant no. 3,
which run only up to z = 0.8 mm, are almost equal to that of the initial variant V0.

In the diagram of Fig. 7.12, similar stress calculations are shown with the difference that
in this case, the calibration file was varied according to the distribution of depth
increments listed in Tab. 7.4.
168

250

C_V2
200

C_V3
150
σx [MPa]

C_V1

C_V0
100

σx-FEM = 100 MPa


50

0
0 0.2 0.4 0.6 0.8 1 1.2
z [mm]

Fig. 7.12: Calculated Stress Distribution using different variants of depth increment distributions
within the calibration file; measurements input file has initial variant V0

The results of the stress distributions calculated using different depth distributions
within the calibration file do not differ significantly to the distribution calculated with
the initial calibration file of variant V0 at depths between z = 0.0 mm and z = 0.8 mm.
All calculations show stress deviations at the initial depth increments. At the final depth
increments only the distribution calculated with a calibration file of variant no. 2 shows
significant deviations, 127 % at z = 0.9. The reason for this is because the
approximation of the calibration functions is carried out with the variant of the depth
distribution (V2), which contains the least strain values, i.e. the polynomial
approximation does not reproduce accurately the calibration strain values at these
depths.

According to the results of this chapter, the selection of the distribution of the depth
increments within the measurement input file or within the calibration file, respectively,
does not significantly reduce the stress deviations at the initial depth increments as well
as at the final depth increments. The spatial resolution at the initial depths should be
small, i.e. the depth increments should be small and uniform, in order to avoid high
strain changes which causes higher slopes of the strain functions.
169

7.7 Interpolation between Geometry Parameters

In addition to the MPA standard block, which uses the original calibration functions
determined in [2], the prototype program provides two alternative options for the
calculation of residual stresses considering the shape of the component.

The first option is realized in the MPA expanded stress calculation block. In this block,
4th order polynomial coefficients are included as fix values. These coefficients were
determined for one strain gage rosette and considering the geometry parameters
thickness T, distance hole-edge D and hole diameter d0. If the mentioned geometry
parameters of a real component are between the geometry parameters considered in the
fixed coefficients, the program calculates the new specific coefficients using linear
interpolation.

The second optional block calculates the calibration functions “online”, i.e. a calibration
file can be imported in order to calculate the specific calibration functions. This
procedure only makes sense, if the calibration file was determined using a model or
specimen of the same shape or of a similar shape, respectively, as the shape of the
measured component. In this case, this procedure is also independent of the used strain
gage rosette.

The objective of this chapter is to compare the results calculated using the MPA
expanded block with the results using the MPA specific block. For this purpose, a
model of a flat specimen was generated. The model has a thickness of T = 1.75 mm, a
distance hole-edge of D = 3.13 mm and a hole diameter of d0 = 1.775 mm. This means
that the geometry parameters of the simulated measurement were selected to be between
the geometry parameters of the calibration coefficients of the MPA Expanded block (s.
Chap. 6.3.2). The model of the simulated measurement has elastic material constants
similar to those of aluminium (E = 70000 MPa and ν = 0.33) and was initially loaded
with a biaxial uniform stress state of σx-FEM = 80 MPa and σx-FEM = 50 MPa.

The same model geometry was used for the determination of the specific calibrations
strains, in order to import it as a calibration input file into the MPA specific stress
170

calculation block. The model of this simulated calibration (with E = 210000 MPa and
ν = 0.285) was uniaxially loaded with σc,x = 160 MPa.

130

MPA Expanded
110

MPA Specific
90
σx [MPa]

σx-FEM = 80 MPa
70

MPA Standard
50

30
0 0.2 0.4 0.6 0.8 1 1.2
z [mm]

Fig. 7.13: Calculated stress distributions in x-direction using the MPA Standard block, the MPA
Expanded block and the MPA Specific block. Geometry of model: T = 1.75 mm,
D = 3.13 mm and d0 = 1.775 mm

Fig 7.13 and Fig. 7.14 show the calculated stress results σx and σy, respectively. Both
diagrams show three different calculated stress distributions: the one calculated with the
MPA standard block (black continuous line), the one determined with the MPA
expanded block (grey dashed line) and the one using the MPA Specific block (black
dashed and dotted line). All three stress calculations were carried out using 4th order
polynomial calibration functions, the same distribution of depth increments and with no
smoothing of the measured strains. In addition, stress deviations of these calculations at
selected hole depths are listed in Tab 7.5.

For all three blocks, the calculated stress distributions in x-direction have a similar
shape than the calculated stress distributions in y-direction. In both directions the
calculated stress distributions run about the initial stress value σi-FEM. The highest
deviations were expected and calculated using the MPA Standard block. In both stress
directions, the distribution calculated with the MPA Expanded block runs almost equal
to the MPA Specific block. At a depth of z = 0.8 mm and deeper, the results calculated
with the MPA Expanded block deviate more from the initial stress values than the
results calculated with the MPA Specific block.
171

130

110

MPA Expanded

90
σy [MPa]

MPA Specific
70

σy-FEM = 50 MPa
50

MPA Standard
30
0 0.2 0.4 0.6 0.8 1 1.2
z [mm]

Fig. 7.14: Calculated stress distributions in y-direction using the MPA Standard block, the MPA
Expanded block and the MPA Specific block. Geometry of model: T = 1.75 mm,
D = 3.13 mm and d0 = 1.775 mm

Stress deviation related to initial stress σi-FEM


Depth
Deviation σx [%] Deviation σy [%]
[mm]
Standard Expanded Specific Standard Expanded Specific
0.0 -32 -7 -5 -30 -2 -1
0.2 14 -3 1 30 5 10
0.5 19 4 5 36 15 16
1.0 -14 52 16 -18 90 33
Tab. 7.5: Stress deviation in x-direction and in y-direction: comparison of stress calculation
blocks MPA Standard, MPA Expanded and MPA Specific

To summarize this chapter it can be said that both of the additionally provided stress
calculations blocks, which consider the specific geometry parameters of the component,
reproduce the initial stresses of the example model with less deviation than the
calculation procedure based on the original calibration functions. The results determined
with the MPA Expanded block show higher stress deviation at the final depth
increments than the ones calculated with the MPA Specific block.
172

7.8 Summary of the Evaluation of the Calculation Program with


Implemented Geometry Specific Calibration Functions

The stress calculations carried out in this chapter were mostly based on strain results
determined using models of specimens with non-reference shape. The sole exception
was the use of the strain distributions of the reference model at the first part of this
evaluation chapter in order to discuss general aspects of the stress calculation using the
MPA algorithm and to show the resulting calculation uncertainty. In the case of the
model of the flat reference specimen, the calculation uncertainty is on average approx.
6 %.

The parametric study has shown, firstly, the influence of the local polynomial
approximation of the measured strain compared to the close polynomial approximation.
It was shown that a stress evaluation using local polynomial approximation determines
on average a lower stress deviation as with close approximation. The possible problems
at the initial and final depth increments were discussed. The second part of the
parametric study has dealt with the effects and the importance of the initial smoothing
of the measured strains. In the third part, the influence of the order polynomial
approximation of the calibration strains was demonstrated. According to the shown
result, a 4th order polynomial approximation is sufficient for the calculation of residual
stresses using the MPA algorithm. The fourth part of the parametric study has discussed
the influence of distribution of depth increments within the measurement input file or
the calibration file.

The last part compared the two provided calculation procedures which consider
geometry specific calibration functions. Both calculation procedures (MPA Expanded
and MPA Specific) lead to similar results: they reproduce in average the initial stress
distributions with less stress deviation than the MPA Standard block, which uses the
original reference calibration functions.
173

8 Conclusion
This work investigated the geometry influence on the stress calculation according to the
differential MPA evaluation algorithm for the hole drilling method, which is an
accepted and widely used technique for the residual stress measurement. The available
evaluation methods of the hole drilling method are based on a previous calibration of
the method on a known stress state under reference geometrical conditions, which are
realized usually with the FEM model of a thick and wide flat plate. Because of this, the
violation of recommended geometrical restrictions, e.g. a measurement on a thin tube,
influences the calculated residual stresses. In this work, exemplarily calculations with
current evaluation software using geometrical non-reference specimens were carried
out. The resulting stress distributions of the selected non-reference specimens show a
significant overestimation of the initially presented stress distribution at the
measurement point. In order to compensate these geometry effects some authors
recommend the determination of geometry specific calibration functions. Following this
recommendation, this work dealt with two main topics.

The first topic was the systematic investigation of the individual contribution of the
violation of single geometry parameters. For this purpose, calibration experiments as
well as calibration FEM calculations were carried out, in which the following geometry
parameters were systematically varied: the specimen’s thickness T, the distance
between the measurement point and a free specimen’s edge D and the cylinder
curvature radius R. The calibrations were compared by means of relieved strain
deviations and calculated stress deviations, which were both related to the results
obtained with the reference calibration. The influence of the single violation of
geometry parameters can be summarized with the following statements:

The boundary condition “thickness” is proposed with T = 3 x d0, according to the


authors of the MPA algorithm. The results considering the geometry parameter
thickness (and also the distance hole-edge) were determined using uniformly loaded flat
specimens or flat finite element models, respectively. They showed that this geometry
limit may be minimized up to a thickness of T = 1.66 x d0. Stress deviations of 18 % for
T = 0.83 x d0 = 1.5 mm and about 53 % for T = 0.55 x d0 = 1 mm were determined on
average over the hole depth.
174

The distance from the center of the measurement point (hole center) to the specimen’s
free edge did not influence the hole drilling results as significantly as the specimen’s
thickness. A maximum stress difference of approx. 5 % on average related to the
reference was found for the results of the model with D = 1.11 x d0 = 2 mm. The
boundary condition for this special geometry parameter, which is set with min.
D = 5 x d0 in literature, may be minimized up to D = 1.94 x d0.

The boundary condition radius of surface curvature is originally set with T = 3 x d0. The
results considering this geometry parameter were determined using uniformly loaded
cylindrical specimens or cylindrical finite elements models, respectively. It was shown
that the boundary condition could not be minimized and should be even displaced to a
higher value. The maximal stress deviations in x-direction were 12 % for the solid
cylinder and 39 % for the hollow cylinder. These results were already calculated for the
cylinders with R = 3.33 x d0, which already satisfies the set boundary condition. In y-
direction, which was transverse to the load direction, the maximum stress differences
were significantly high for the results with R = 1.67 x d0.

An uncertainty analysis was carried out for the experimental calibration as well as for
the numerical calibration. According to this uncertainty analysis, the experimentally
determined calibration strains showed a tolerance range, as many influencing factors
have an effect on the calibration results. The distributions of the numerical calibration
strain ran within the determined tolerance range. In the case of the differential
evaluation method, the proper determination of calibration functions should be carried
out using the results of the finite element calibration, because there is almost no
scattering. The scattering of the experimental calibration strain distribution could lead to
high or low values of the calculated strain derivations, which are part of the equation of
the calibration function. Thus, the stress calculation could lead to erroneous values.

The second topic was the implementation of geometry specific calibration functions into
an evaluation program using the differential MPA method. For this purpose, a prototype
program was developed. This program allows the calculation of residual stresses
according to the MPA method using three options. The first option uses the original
calibration functions of the authors of the MPA method. The second option considers
175

geometry specific calibration functions, which are included as fixed polynomial


coefficients. In this block, the fixed coefficients consider the geometry parameters
thickness, distance hole-edge and hole diameter. The third option calculates the
calibration functions “online”, i.e. the calibration strain distributions are imported and
the calibrations functions are then calculated for these specific calibration strains. This
procedure can be beneficial only if the geometry of the calibration model is equal or
similar to the geometry of the measured component.

The prototype program was evaluated by means of a parametric study. It was shown
that the calculation procedure itself leads to small stress deviations even in the case of
the evaluation of the reference specimen. Moreover, the importance of the strain
smoothing, the importance of the used differentiation procedure and the influence of
different selected depth distributions within the strain files were shown and discussed.
The results obtained with the prototype program confirm previous investigations
concerning the use of a 4th order polynomial approximation of the calibration strain
distributions.

To summarize the evaluation of the prototype program, it can be said that the
calculation using geometry specific calibration functions minimizes on average the
stress deviations compared to a calculation using the original reference calibration
functions. Stress deviations still appear at the first depth increments of curved surfaces
and at the final depth increments. The deviations at the near surface depth increments
were caused due to the fact that the introduced hole did not have an ideal blind-hole
(with d0) within the first increments whereas the deviations at the last increments were
related to the general specification of the differential method, which does not consider
stresses above the actual hole depth.

The two presented calculation procedures determine similar results for the selected plate
model with minor stress deviation using the MPA Specific block. The MPA Expanded
calculation block with fixed calibration coefficients has the advantage that the user does
not require a previous calibration. The disadvantage of this procedure is that it is always
limited by the considered range of the geometry parameters implemented within the
calibration coefficients. In this calculation block, three geometry parameters for one
specific strain gage rosette are implemented within the calibration functions but it is
176

possible, in principle, to consider more geometry variables. This could be carried out at
the cost of calculation accuracy as the evaluation becomes more complex. The MPA
Specific block has the advantage that the exact geometry of the component, the used
strain gage rosette and the stress conditions at the measurement point are considered
during the determination of the calibration file, i.e. there is no need for the
determination of the specific calibration coefficients by means of an interpolation
between calibration coefficients, which consider single geometry parameters.
177

9 References
[1] T. Schwarz; H. Kockelmann (1993). Die Bohrlochmethode – ein für viele
Anwendungsbereiche optimales Verfahren zur experimentellen Ermittlung
von Eigenspannungen
Messtechnische Briefe 29, Heft 2, pp. 33-38
Darmstadt (D): HBM GmbH

[2] T. Schwarz (1996). Beitrag zur Eigenspannungsermittlung an isotropen,


anisotropen sowie inhomogen, schichtweise aufgebauten Werkstoffen
mittels Bohrlochmethode und Ringkernverfahren
Dissertation, Fakultät Energietechnik der Universität Stuttgart

[3] J. Lu (1996). Handbook of measurement of residual stresses / Society for


experimental mechanics, Inc.
Lilburn (USA): The Fairmont Press, Inc.

[4] E. Macherauch; K.H. Kloos (1987). Origin, Measurement and Evaluation of


Residual Stresses
Residual Stresses in Science and Technology, pp. 3-26
DGM-Informationsgesellschaft

[5] E. Macherauch; H. Wohlfart; U. Wolfstieg (1973). Zur zweckmäßigen


Definition von Eigenspannungen
Härterei-Technische Mitteilungen HTM, Vol. 28-3 (1973), pp. 201-211

[6] P.J. Withers; H.K.D.H. Bhadeshia (2001). Residual Stress Part 1 -


Measurement Techniques
Material Science and Technology Vol. 17 (April); pp. 355-365
IoM Communications Ltd.

[7] V. Hauk (1986). Residual Stresses, Their Importance in Science and


Technology
Proceedings of the 1th European Conference on Residual Stresses,
Karlsruhe, Germany (4, 1983); pp. 9-45
DGM-Informationsgesellschaft

[8] G. Maeder (1992). How to account residual stresses: examples in the


automotive industry
Proceedings of the 3th European Conference on Residual Stresses,
Frankfurt, Germany (11, 1992); pp. 3-21
DGM-Informationsgesellschaft

[9] E. Garcia-Sobolevski; J. Gibmeier; B. Scholtes; A. Gebauer-Teichmann; R.


Herzog; A. Egner-Walter; E. Hepp; E. Stark (2005). Production Optimisation
of Thin Walled Light Metal Die Casting Components
Proceedings of the 1. International Conference on Distortion Engineering, DIE
2005, Bremen, Germany, September 14.-16. 2005, pp. 159-166
178

[10] F.A. Kandil, J.D. Lord, A.T. Fry, P.V. Grant (2001). A Review of Residual
Stress Measurements Methods – A Guide to Technique Selection
NPL Report MATC(A)04, Feb. 2001
Teddington (GB): NPL

[11] N.N. (2001). Standard Test Method for Determining Stresses by the Hole-
Drilling Strain-Gage Method
ASTM Designation E837 – 01

[12] N.N. (2005). Measurement of Residual Stresses by the Hole-Drilling Strain


Gage Method
Technical Note: TN-503-6
(USA): Vishay – Measurement Group, Inc.

[13] P. V. Grant, J. D. Lord, P. S. Whitehead (2002). The Measurement of Residual


Stresses by the Incremental Hole Drilling Technique
Measurement Good Practice Guide No. 53
Teddington (GB): NPL

[14] J. Mathar (1933). Ermittlung von Eigenspannungen durch Messung von


Bohrloch-Verformungen
Archiv für das Eisenhüttenwesen, Vol. 6, pp. 277-281

[15] J. Mathar (1934). Determination of Initial stresses by Measuring the


Deformation Around Drilled Holes
Trans., ASME 56, No. 4, pp. 249-254

[16] J. H. Kirsch (1898). Die Theorie der Elastizität und die Bedürfnisse der
Festigkeitslehre
VDI-Z 42, pp. 797-809

[17] W. Soete; R. Vancrombrugge (1950). An Industrial Method for the


Determination of Residual Stresses
Proceedings SESA 8 (1); pp. 17-28

[18] R. A. Kelsey (1965). Measuring Non-Uniform Residual Stresses by the Hole


Drilling Method
Proceedings of the Society for Experimental Stress Analysis (SESA) 14,
No. 1, pp. 181-1194

[19] J. Rendler; I. Vigness (1966). Hole-drilling Strain-gage Method of Measuring


Residual Stresses
Experimental Mechanics, December 1966, pp. 577-586

[20] M. Flaman (1982). Brief investigation of induced drilling stresses in the


center-hole mthod of residual-stress measurement
Experimental Mechanics, January 1982, pp. 26-30
179

[21] Bijak-Zochowski, M (1978) A Semi-Destructive Method of Measuring


Residual Stresses.
International Conference on Experimental Stress Analysis
VDI-Verlag Nr. 313, pp. 469-476

[22] G. S. Schajer (1981). Application of Finite Element Calculations to Residual


Stress Measurements
Journal of Engineering Materials and Technology Vol. 103, No. 2, pp. 157-163
(USA): ASME International

[23] G. Schajer (1988). Measurement of Non-Uniform Residual Stresses Using


the Hole-Drilling Method. Part I – Stress Calculation Procedures
Journal of Engineering Materials and Technology, Vol. 110-10, pp. 338-343
Transactions of the ASME

[24] G. Schajer (1988). Measurement of Non-Uniform Residual Stresses Using


the Hole-Drilling Method. Part II – Practical Application of the Integral
Method
Journal of Engineering Materials and Technology, Vol. 110-10, pp. 344-349
Transactions of the ASME

[25] H. Kockelmann; T. Schwarz (1993). Vergleich praktizierter


Auswerteverfahren der Bohrlochmethode
Experimentelle Spannungsanalyse, GMA-Bericht 22
VDI/VDE-Gesellschaft, Düsseldorf

[26] P.S. Prevey (1988). Residual-Stress Distributions Produced by Strain-Gage


Surface Preparation
Experimental Mechanics, Vol. 28-1, pp. 92-97
Sage Science Press

[27] S. Rendner; C. Perry (1982). Factors Affecting the Accuracy of Residual


Stress Measurements Using the Blind-Hole Drilling Method
7th International Conf. on Experimental Stress Analysis
Haifa, Israel, August 1982, pp. 604-613

[28] O. Sicot; X. Gong; A. Cherouat; J. Lu (2003). Influence of experimental


parameters on determination of residual stress using the incremental hole-
drilling method
Composites Science and Technology 64, pp. 171-180
(GB): Elsevier Science

[29] H. Häusler, G. König, H. Kockelmann (1987). On the accuracy of determining


the variation with depth of residual stresses by means of the hole drilling
method
Residual Stresses in Science and Technology
Ed. E. Macherauch, V. Hauk DGM Informationsgesellschaft, pp. 257-264
180

[30] M. Kaibiri (1986). Toward More Accurate Residual-Stress Measurement by


the Hole-Drilling Method: Analysis of Relieved-Strain Coefficients
Experimental Mechanics, March 1986, pp. 14-21

[31] G. Schajer (1993). Use of displacement data to calculate strain gauge


response in non-uniform strain fields
Strain 02/1993; pp. 9-13
Newcastle (GB): British Society for Strain Measurement

[32] Nickola, W. E. (1986) Practical Subsurface Residual Stress Evaluation by


the Hole-Drilling Method
Joint Conference on Experimental Mechanics
Society for Experimental Mechanics New Orleans LA, pp. 47-58

[33] H. Kockelmann; G. König (1987). Eignung der Bohrlochmethode zur


Ermittlung der Tiefenverläufe von Eigenspannungen
Final report: research project of the German Research Foundation (DFG) Ko
896/2-1
Stuttgart (D): MPA

[34] G. Schajer (2003). H-DRILL 2.30 – User Guide

[35] S. Altrichter (1960). Die mathematische Grundlagen der


Eigenspannungsmessung mit dem Bohrlochverfahren
ZIS-Mitteilungen 12, pp. 843-854

[36] H. Motzfeld (1961). Fehlerquellen beim Messen von


Schweißeigenspannungen nach dem Trepanier- und Bohrlochverfahren
Schweissen und Schneiden, Bd. 13, H. 10, pp. 464-470

[37] A. Nawwar; K. McLachlan; J. Shewchuk (1976). A Modified Hole-drilling


Technique for Determining Residual Stresses in Thin Plates
Experimental Mechanics, June 1976, pp. 226-232

[38] J. Münker (1995). Untersuchung und Weiterentwicklung der Auswertungs-


methoden für teilzerstörende Eigenspannungsmessverfahren
Dissertation, Universität – Gesamthochschule Siegen

[39] J.-N. Aoh; C.-S. Wei (2002). On the Improvement of Calibration Coefficients
for Hole-Drilling Integral Method: Part I – Analysis of Calibration
Coefficients Obtained by a 3-D FEM Model
Journal of Engineering Materials and Technologies, April 2002, Vol. 124, pp.
250-258

[40] J.-N. Aoh; C.-S. Wei (2003). On the Improvement of Calibration Coefficients
for Hole-Drilling Integral Method: Part II – Experimental Validation of
Calibration Coefficients
Journal of Engineering Materials and Technologies, April 2003, Vol. 125, pp.
107-115
181

[41] F. Haase (1998). Eigenspannungsermittlung an dünnwandigen Bauteilen


und Schichtverbunden
Dr. Ing. Dissertation, Universität Dortmund

[42] U. Preckel (1986). Erweiterung der Verfahrensgrenzen beim


Bohrlochverfahren zur Ermittlung von Eigenspannungen 1. Art
Industrie Anzeiger, Kurzberichte der Hochschulgruppe Fertigungstechnik 108,
Nr. 8, pp.18/26

[43] Wu-Xue Zhu; D.J. Smith (1994). Residual Stresses by hole-drilling in curved
componements
Recent Advances in Experimental Mechanics, Silva Gomes et al. (eds) pp. 777-
782

[44] G. Montay; A. Cherouat; C. Garnier; J. Lu (2004). Determining Residual


Stress in Spherical Componements: A New Application of the Hole-Drilling
Method
Journal of Testing and Evaluation Vol. 32-1, pp. 73-79
ASTM International

[45] J. Luo; G. Montay; J. Lu (2005). An Advanced Residual Stress Determination


for 3D Cylinder Structure
Proceedings of the 7th International Conference on Residual Stresses, Xi’an-
China (6, 2004); Materials Science Forum 490/491, pp. 62-66

[46] G. Montay; A. Cherouat; J. Lu (2006). Residual Stress Analysis in a


Crankshaft Using the Hole Drilling Method
Proceedings of the 7th European Conference on Residual Stresses, Berlin-
Germany (9, 2006); Materials Science Forum 524/525, pp. 537-542

[47] Werkstoffblatt 210 (1981). Hochfeste vergütete Feinkornbaustähle


(D) Duisburg: Thyssen AG

[48] Measurements Group (1976). Surface Preparation for Strain Gage Bonding.
Instruction Bulletin B-129-7
Raleigh, USA, Measurements Group, Inc.

[49] Measurements Group(1979). Strain Gage Installations with M-Bond 200


Adhesive.
Instruction Bulletin B-129-13
Raleigh, USA, Measurements Group, Inc

[50] H. B. Hibbit; B. J. Karlsson; E. P. Sorensen (2000). ABAQUS Standard


Manual, Version 6.1
(USA): HKS Inc.

[51] R. E. Peterson (1974). Stress Concentration Factors


(USA): John Wiley & Sons, Inc.
182

[52] K. Hoffmann (1989). An Introduction to Measurements using Strain Gages


Darmstadt (D): Hottinger Baldwin Messtechnik GmbH

[53] R. Oettel (2000). The Determination of Uncertainties in Residual Stress


Measurement (using the hole drilling technique)
Manual of Codes of Practice for the Determination of Uncertainties in
Mechanical Tests on Metallic Materials
Standards Measurement & Testing Projekt No. SMT4-CT97-2165
NPL

[54] N.N. (2005). Strain Gage Selection - Criteria, Procedures,


Recommendations
Technical Note: TN-505-4
(USA): Vishay – Measurement Group, Inc.

[55] N.N. (1992). Model RS-200 - Milling Guide


Instruction Manual
(USA): Vishay – Measurement Group, Inc.

[56] J. Gibmeier; M. Kornmeier; B. Scholtes (2000). Plastic Deformation during


Application of the Hole-Drilling Method
Proceedings of the 5th European Conference on Residual Stresses,
Delft-Netherlands (6, 2004); Materials Science Forum 347/349, pp. 131-136

[57] D. Vangi; M. Ermini (2000). Plasticity Effects in Residual Stress


Measurements by the Hole Drilling Method
Strain, Vol. 36-2, pp. 55-59
Newcastle (GB): British Society for Strain Measurement

[58] G. Schajer; E. Altus (1996). Stress Calculation Error Analysis for


Incremental Hole-Drilling Residual Stress Measurements
Journal of Engineering Materials and Technology; Vol. 118; pp. 120-126
Transactions of the ASME

[59] E. Garcia Sobolevski (2005). Eigenspannungsanalysen mit dem


Bohrlochverfahren an Bauteilen mit komplexen Geometrien
Speech at the Association for Materials and Heat Treatment, Section of Residual
Stress: AWT-FA 13 (10.05.2005), Karlsruhe

[60] H. Späth (1978). Spline-Algorithmen zur Konstruktion glatter Kurven und


Flächen
München (D): R. Oldenburg Verlag

[61] B. Scholtes (1990). Eigenspannungen in mechanisch randschicht-verformten


Werkstoffzuständen. Ursachen, Ermittlung und Bewertung.
Oberursel (D): DGM Informationsgesellschaft mbH
183

10 Zusammenfassung in deutscher Sprache

Titel:

Eigenspannungsanalyse an Bauteilen mit realen Geometrien


unter Verwendung der inkrementellen Bohrlochmethode und
eines differentiellen Auswerteverfahren

Eigenspannungen entstehen als Folge von Fertigungsprozessen in nahezu allen


Bauteilen. In manchen Fällen kann dadurch das Bauteilverhalten je nach Höhe und
Wirkrichtung der vorhandenen Eigenspannung günstig oder gar nachteilig beeinflusst
werden. Aus diesem Grund ist die Nachfrage nach zuverlässigen und kostengünstigen
Eigenspannungs-analysemethoden von besonderem Interesse in der Industrie. Die
Bohrlochmethode ist in diesem Zusammenhang ein übliches und weit verbreitetes
Messverfahren, das auf dem folgendem Prinzip beruht: Beim Einbringen eines
Sackloches in die Bauteiloberfläche wird durch das Entfernen spannungsbehafteter
Werkstoffbereiche das innere Gleichgewicht lokal gestört und es stellt sich in der
Bohrlochumgebung ein neuer Gleichgewichtszustand ein. Diese
Gleichgewichtsänderungen führen zu entsprechenden Verformungen, die mittels
speziell für die Bohrlochmethode entwickelten Dehnungsmessstreifen (DMS) an der
Oberfläche ermittelt werden können. Aus den gemessen Dehnungsänderungen lassen
sich dann die im Bauteil vorhandenen Spannungen auswerten. Auswertemethoden, die
u.a. die Bohrungstiefe explizit berücksichtigen, um z.B. einen
Eigenspannungstiefenverlauf berechnen zu können, basieren auf vorher ermittelten
Kalibrierfunktionen, die bei einem bekannten Spannungszustand aufgenommen worden
sind. Die Ermittlung dieser Kalibrierfunktionen erfolgt dabei unter Einhaltung
geometrischer Randbedingungen, wie z.B. der Bauteildicke, dem Abstand der Bohrung
zur Bauteilberandung, dem Radius der Oberflächenkrümmung. Zahlreiche Bauteile aus
der Praxis, an denen eine Eigenspannungsmessung mit dem Bohrlochverfahren sinnvoll
erscheint, weisen oft Abweichungen von den in der Literatur angegebenen
erforderlichen geometrischen Randbedingungen auf. Eine Eigenspannungsberechnung
mit den kommerziell erhältlichen Auswertemethoden könnte z.B. zu einer
184

Überschätzung der Eigenspannungswerte führen. In solchen Fällen wird die Ermittlung


von geometriespezifischen Kalibrierfunktionen empfohlen, um die
Spannungsabweichung korrigieren zu können.

In der vorliegenden Arbeit wird der Einfluss der Bauteilgeometrie auf die
Spannungsberechnung nach dem differentiellen MPA Verfahren [2.44] für die
Bohrlochmethode untersucht. Zuerst soll dabei der Einfluss der einzelnen
geometrischen Randbedingungen anhand von Dehnungs- und Spannungsabweichungen
bezogen auf Ergebnisse, die mit der Referenzgeometrie erzielt worden sind, ermittelt
werden. Die drei untersuchten geometrischen Parameter sind die Bauteildicke (T für
engl. „Thickness“), der Abstand der Bohrung zum Bauteilrand (D für engl. „Distance“)
und der Radius der Oberflächenkrümmung (R für engl. „Radius“). Als zweites Ziel
sollen geometriespezifische Kalibrierfunktionen in dem differentiellen MPA-
Algorithmus implementiert werden, um eine mögliche Spannungskorrektur zu zeigen.
Für beide Zielsetzungen werden sowohl experimentelle Kalibrierungen als auch
Kalibrierungen mittels FEM an Proben bzw. Modellen, welche die empfohlenen
Randbedingungen systematisch verletzen, durchgeführt. Die Ergebnisse der
durchgeführten Kalibrierungen werden anhand einer Unsicherheitsanalyse diskutiert. Im
letzen Teil der Arbeit wird die erste Version eines Auswerteprogramms nach der
differentiellen MPA-Methode beschrieben und evaluiert. Im Gegensatz zu
kommerziellen Bohrloch-Auswerteprogrammen ermöglicht das vorgestellte Programm
die Berechnung von spezifischen Kalibrierfunktionen, welche die tatsächliche
Bauteilgeometrie implizit berücksichtigen.

Die erzielten Ergebnisse können wie folgt zusammengefasst werden:

Die Randbedingung „Bauteildicke“ wird in [2.44] T = 3 x d0 empfohlen (d0 ist dabei der
Bohrungsdurchmesser). Die Ergebnisse, die den Geometrieparameter Bauteildicke
berücksichtigen (und auch den Parameter Abstand Bohrung-Kante) wurden mit
einachsig belasteteten Flachzugproben bzw. Modellen ermittelt. Für diesen Fall zeigte
sich, dass diese Randbedingung auf eine Dicke von T = 1.66 x d0 verkleinert werden
kann. Für Proben mit T = 0.83 x d0 = 1.5 mm und T = 0.55 x d0 = 1 mm wurden im
Durchschnitt Spannungsabweichungen bezogen auf die Referenzergebnisse von 18 %
bzw. 53 % ermittelt.
185

Die Ergebnisse bezüglich des Parameters „Abstand Bohrung-Kante“ zeigen keinen


signifikanten Einfluss auf die Kalibrierung im Vergleich mit dem der Kalibrierung unter
Berücksichtigung der Bauteildicke. Die maximale Spannungsabweichung bezogen auf
die Referenz beträgt durchschnittlich 5 % für das Modell der Probe mit
D = 1.11 x d0 = 2 mm. Die Randbedingung „Abstand Bohrung-Kante“ kann daher von
empfohlenen D = 5 x d0 auf D = 1.94 x d0 minimiert werden.

Die Randbedingung „Radius der Oberflächenkrümmung“ ist ursprünglich mit


D = 3 x d0 angesetzt. Die Ergebnisse unter Berücksichtigung dieses Parameters wurden
unter Verwendung von einachsig belasteten zylindrischen Proben oder dessen FEM
Modelle ermittelt. Es zeigt sich, dass die empfohlene Randbedingung nicht minimiert
werden kann und sogar erweitert werden sollte. Schon bei einem Zylinderradius von
R = 3.33 x d0 = 6 mm (die Randbedingung ist dabei erfüllt) betragen die maximale
Spannungsabweichungen in Belastungsrichtung 12 % für Vollzylinderproben und 39 %
für Hohlzylinderproben.

Die Unsicherheitsanalyse wurde durchgeführt um die Güte der FEM Ergebnisse,


bezogen auf die experimentellen Kalibrierergebnisse, zu vergleichen. Demzufolge
zeigen die experimentell ermittelten Verläufe einen Toleranzband, welches aus den
zahlreichen versuchstechnischen Einflussfaktoren verursacht wird. Die simulierten
Ergebnisse verlaufen innerhalb des errechneten Toleranzbandes. Aus der Gewonnen
Erfahrung in dieser Arbeit wird für die differentielle MPA-Auswertemethode eine
Kalibrierung mittels FEM empfohlen, um die Streuung der Messdehnung zu vermeiden.
Diese Streuung der gemessenen Kalibrierdehnung kann dazu führen, dass die
berechnete Dehnungsableitungen der Kalibrierung, auf die das MPA-Verfahren basiert,
zu hoch oder zu niedrig ausfallen und dadurch fehlerbehaftete Eigenspannungswerte
berechnet werden.

Das Berechnungsprogramm, das im zweiten Teil der Arbeit entwickelt worden ist
basiert auf dem MPA-Verfahren und hat drei Optionen für die Spannungsberechnung.
Die erste Option verwendet die originalen Kalibrierfunktionen des MPA-Verfahrens aus
[2.44] („MPA Standard“), die unter Einhaltung der geometrischen Randbedingung an
einer flachen und ausreichend dicken und breiten Platte ermittelt worden sind. Die
186

zweite Berechnungsoption („MPA Expanded“) berücksichtigt drei geometrische


Parameter (die Bauteildicke, den Abstand Bohrung-Kante und den
Bohrungsdurchmesser) in Form von festgelegten geometriespezifischen
Kalibrierfunktionen. Die dritte Berechnungsoption ermittelt die Kalibrierfunktionen
während der eigentlichen Spannungsberechnung durch Import der spezifischen
Kalibrierdehnungen und anschliessender Ermittlung der Kalibrierfunktionen („MPA-
Specific“). Diese Option ist nur dann sinnvoll, wenn die Geometrie der Kalibrierprobe
bzw. des Modells ähnlich oder gleich der Geometrie des gemessenen Bauteiles ist. Das
Programm wurde anhand einer Parameterstudie evaluiert. Es hat sich dabei gezeigt, dass
eine geometriespezifische Berechnung im Durchschnitt die Spannungsabweichung im
Vergleich mit der Berechnung unter Verwendung der Referenzkalibrierfunktionen
minimiert.
187

11 Appendix
11.1 FEM Post Processor Plots

Fig. 11.1: fhT6D2m: stress distribution σc,x before drilling

Fig. 11.2: fhT6D2m: stress distribution σc,y before drilling


188

Fig. 11.3: fhT1.5D2m: stress distribution σc,x before drilling

Fig. 11.4: fhT1.5D2m: stress distribution σc,y before drilling


189

Fig. 11.5: hchR6m: stress distribution σc,x before drilling

Fig. 11.6: hchR6m: stress distribution σc,y (tangential to cylinder surface) before drilling
190

Fig. 11.7: Inhomogeneity in hcR6m model: stress distribution in x-direction before drilling. Initial
stress σc,x = 160 MPa

Fig. 11.8: Inhomogeneity in hcR4m model: stress distribution in x-direction before drilling. Initial
stress σc,x = 160 MPa
191

Fig. 11.9: Inhomogeneity in hcR3m model: stress distribution in x-direction before drilling. Initial
stress σc,x = 160 MPa

Fig. 11.10: Plastification in hchR6m model: stress distribution in x-direction before drilling. Initial
stress σc,x = 210 MPa
192

Fig. 11.11: Location of strain gage no. 1 in hcR6m model: strain distribution tangential to cylinder
surface. Hole depth z = 1 mm, Initial stress σc,x = 160 MPa

Fig. 11.12: Location of strain gage no. 1 in hcR4m model: strain distribution tangential to cylinder
surface. Hole depth z = 1 mm, Initial stress σc,x = 160 MPa
193

Fig. 11.13: Location of strain gage no. 1 in hcR3m model: strain distribution tangential to cylinder
surface. Hole depth z = 1 mm, Initial stress σc,x = 160 MPa
194

11.2 Program Listings


11.2.1 Smoothing using Weighted Average Algorithm

% Smoothing of measured strains according to weighted means algorithm


% B- strain data matrix, c-column
function gw1 = gewichtfactor(B,c)

[r1,c1]=size(B);
index1=r1;
for i = 1:index1
Tiefe1(i)=B(i,1);
Eps901(i)=B(i,c);
Y(i)=B(i,c);
end;
%Weight factors
cim1=0.1;
ci12=0.2;
ci=0.4;
ci23=0.2;
cip1=0.1;
for i = 2:index1-1
Yi12(i)=0.5*(Y(i-1)+Y(i));
Yi23(i)=0.5*(Y(i+1)+Y(i));
gw1(i)=(cim1*Y(i-
1)+ci12*Yi12(i)+ci*Y(i)+ci23*Yi23(i)+cip1*Y(i+1));
end;
gw1(1)=(0.95*Y(1)+0.05*Y(2));
gw1(index1)=(0.9*Y(index1)+0.1*Y(index1-1));

11.2.2 Smoothing using Compensative Cubic Splines

% Smoothing of measured strains according to compensative cubic


splines
% B- strain data matrix, c-column, fil - smoothing factor

function yw = CubicSpline(B,c,fil)

[r1,c1]=size(B);
n=r1;
for i = 1:n
x(i)=B(i,1);
y(i)=B(i,c);

end;

p=1000/fil^2;
dx=x(3)-x(2);
dxp=x(4)-x(3);
dxm=x(2)-x(1);
dy=y(3)-y(2);
dym=y(2)-y(1);
dxmm=.00001;

if (dx == 0)
dx=.00001;
195

end
if(dxm == 0)
dxm=.00001;
end

for i=2:(n-1)
if (dxp == 0)
dxp=.00001;
end
ak(i)=dxm -(6/ (p*dxm))*( 1/dxmm+1/dxm)-(6/(p*dxm))*( 1/dx+1/dxm);
bk(i)=2*(dxm+dx)+ 6/(p*dxm^2) + 6/(p*dx^2) + (6/p)*(1/dx+1/dxm)^2;
ck(i)=dx-(6/(p*dx))*(1/dx+1/dxm)-(6/(p*dx))*(1/dx+1/dxp);
dk(i)=6/(p*dxm*dxmm);
ek(i)=6/(p*dx*dxp);
yk(i)=6*(dy/dx - dym/dxm);

dxmm=dxm;
dxm=dx;
dx=dxp;
dym=dy;
if (i == (n-1))
continue
else
dy= (y(i+2)-y(i+1));
end

if(i==(n-2))
continue
else
dxp=(x(i+3)-x(i+2));
end
end

ak(2)=0;
dk(2)=0;
dk(3)=0;
ek(n-2)=0;
ek(n-1)=0;
ck(n-1)=0;
zn(3)=ck(2)/bk(2);
za(3)=ak(3);
bk(3)=bk(3)-za(3)*zn(3);

if (bk(3) == 0)
bk(3)=.00001;
end

for i=4:n-1
zo(i)=ek(i-2)/bk(i-2);
zn(i)=(ck(i-1)-za(i-1)*zo(i))/ bk(i-1);
za(i)=ak(i)-dk(i)*zn(i-1);
bk(i)=bk(i)-dk(i)*zo(i)-za(i)*zn(i);
if (bk(i) == 0);
bk(i) = .00001;
end
end

ak(2)=yk(2)/bk(2);
ak(3)=(yk(3)-za(3)*ak(2))/bk(3);
196

for i=4:n-1
ak(i)=(yk(i)-za(i)*ak(i-1)-dk(i)*ak(i-2))/bk(i);
end

yz(1)=0;
yz(n)=0;
yz(n-1)=ak(n-1);
yz(n-2)=ak(n-2)-zn(n-1)*yz(n-1);

for i=3:n-2
in=n-i;
yz(in)=ak(in)-zn(in+1)*yz(in+1)-zo(in+2)*yz(in+2);
end

dxi=(x(2)-x(1));
if(dxi ==0)
dxi=.00001;
end

yw(1)=y(1)-yz(2)/(p*dxi);

for i=2:n-1
dxm=dxi;
dxi=(x(i+1)-x(i));
if (dxi == 0)
dxi=.00001;
end

yw(i)=y(i)-((yz(i-1)/dxm -(1/dxm+1/dxi)*yz(i) + yz(i+1)/dxi)/p);

end

yw(n)=y(n)-(yz(n-1)/(p*dxi));

11.2.3 Local Polynomial Approximation of Measured Strains

% Local Polynomial Approximation of Measured Strains (2th order


% polynomials) and Differentiation

function [De11, De21, De31] = polynomialapproximation(ksi, e1, e2, e3,


index);

for i = 1:index

% Surface Value of Differentiated Strain (z=0)


if(i == 1)

ek1 = polyfit(ksi(i:i+2), e1(i:i+2), 2);


ek2 = polyfit(ksi(i:i+2), e2(i:i+2), 2);
ek3 = polyfit(ksi(i:i+2), e3(i:i+2), 2);

De1 = polyder(ek1);
De2 = polyder(ek2);
De3 = polyder(ek3);

De11(i)=polyvalm(De1, ksi(i));
De21(i)=polyvalm(De2, ksi(i));
197

De31(i)=polyvalm(De3, ksi(i));

end

% Final Value of Differentiated Strain (z=n)


if( i== index)

ek1 = polyfit(ksi(i-2:i), e1(i-2:i), 2);


ek3 = polyfit(ksi(i-2:i), e3(i-2:i), 2);
ek2 = polyfit(ksi(i-2:i), e2(i-2:i), 2);

De1 = polyder(ek1);
De2 = polyder(ek2);
De3 = polyder(ek3);

De11(i) = polyvalm(De1, ksi(i));


De21(i) = polyvalm(De2, ksi(i));
De31(i) = polyvalm(De3, ksi(i));

end
% Values of Differentiated Strain (z>=0 & z<=n)
if(i ~= 1 & i ~= index)

ek1 = polyfit(ksi(i-1:i+1), e1(i-1:i+1), 2);


ek3 = polyfit(ksi(i-1:i+1), e3(i-1:i+1), 2);
ek2 = polyfit(ksi(i-1:i+1), e2(i-1:i+1), 2);

De1 = polyder(ek1);
De3 = polyder(ek3);
De2 = polyder(ek2);

De11(i) = polyvalm(De1, ksi(i));


De31(i) = polyvalm(De3, ksi(i));
De21(i) = polyvalm(De2, ksi(i));

end
end

11.2.4 Calculation of Geometry Specific Calibration Coefficients via Linear


Interpolation between Single Geometry Specific Coefficient

%MPA Expanded: Calculation of Geometry Specific Calibration


Coefficients via linear Interpolation between Single Geometry Specific
Coefficient

%Example for one coefficient


function [AFinal ,BFinal] = InterpThreeFinal()

% A1's
DataBaseA1=zeros(5,5,5)
%DataBaseA1(T, D ,d)
DataBaseA1(1,1:5,1) = [-3700.2705 -1912.7868 -1639.2446 -1531.3581 -1549.9594]
DataBaseA1(2,1:5,1) = [-3698.3435 -1907.2014 -1634.4170 -1527.7767 -1548.6013]
DataBaseA1(3,1:5,1) = [-3647.7449 -1874.9862 -1608.8592 -1510.5137 -1541.4919]
DataBaseA1(4,1:5,1) = [-3579.3598 -1846.8651 -1586.6564 -1495.3431 -1535.1242]
DataBaseA1(5,1:5,1) = [-3482.2677 -1838.8630 -1567.9683 -1479.0370 -1528.3653]

DataBaseA1(1,1:5,2) = [-3906.7247 -2051.2754 -1764.5050 -1648.9371 -1667.9301]


198

DataBaseA1(2,1:5,2) = [-3904.9939 -2045.5976 -1758.9237 -1645.2296 -1666.6122]


DataBaseA1(3,1:5,2) = [-3853.2428 -2011.7834 -1732.3980 -1627.9806 -1658.9324]
DataBaseA1(4,1:5,2) = [-3782.5792 -1982.6426 -1709.3333 -1612.2845 -1652.7124]
DataBaseA1(5,1:5,2) = [-3686.4660 -1975.2597 -1689.7591 -1594.7679 -1645.4352]

DataBaseA1(1,1:5,3) = [-4052.2732 -2173.3406 -1871.8268 -1750.4798 -1770.3993]


DataBaseA1(2,1:5,3) = [-4050.0980 -2166.4257 -1866.3050 -1746.4591 -1768.6442]
DataBaseA1(3,1:5,3) = [-4000.3242 -2131.7552 -1838.9090 -1728.2361 -1760.2929]
DataBaseA1(4,1:5,3) = [-3928.7754 -2101.8966 -1814.8983 -1711.5301 -1753.4989]
DataBaseA1(5,1:5,3) = [-3832.4550 -2093.5125 -1793.4843 -1692.6975 -1746.1197]

DataBaseA1(1,1:5,4) = [-4347.4414 -2357.3631 -2042.2573 -1912.4525 -1930.1884]


DataBaseA1(2,1:5,4) = [-4345.5617 -2350.8198 -2036.4484 -1908.3985 -1928.7200]
DataBaseA1(3,1:5,4) = [-4294.1687 -2315.5693 -2008.7075 -1889.2160 -1920.2157]
DataBaseA1(4,1:5,4) = [-4219.5929 -2285.3094 -1984.2833 -1872.6718 -1913.3006]
DataBaseA1(5,1:5,4) = [-4125.7105 -2278.5124 -1963.0949 -1854.1975 -1906.6113]

DataBaseA1(1,1:5,5) = [-4583.5832 -2526.0317 -2196.3158 -2058.9393 -2076.2847]


DataBaseA1(2,1:5,5) = [-4581.8166 -2520.0205 -2190.8460 -2054.6214 -2074.0630]
DataBaseA1(3,1:5,5) = [-4529.3496 -2483.9705 -2162.4532 -2035.0887 -2065.5472]
DataBaseA1(4,1:5,5) = [-4454.2168 -2452.7721 -2137.1478 -2018.1857 -2058.4440]
DataBaseA1(5,1:5,5) = [-4363.1939 -2446.7651 -2115.3559 -1998.7067 -2051.0379]

% Generating A,D and d Arrays


A=zeros(5,5,5)
D=zeros(5,5,5)
d=zeros(5,5,5)

temp=[20 10 5 3.5 2] %Range of distance parameters


for k=1:5
for j=1:5
for i=1:5
A(j,i,k)=temp(j)
end
end
end

temp1=[1 1.5 2 3 6]%Range of thickness parameters

for k=1:5
for j=1:5
for i=1:5
D(j,i,k)=temp1(i)
end
end
end

temp2=[1.7 1.75 1.8 1.85 1.9]%Range of hole diameter parameters

for k=1:5
for j=1:5
for i=1:5
d(j,i,k)=temp2(k)
end
end
end

global BHD %Measurement Hole Diameter


global THICKNESS %Specimens Thickness of Measurement
global EDGEDISTANCE % Distance Hole-Edge of Measurement

gd=BHD
gD=THICKNESS
gA=EDGEDISTANCE
199

A1=interp3(D,A,d,DataBaseA1,gD,gA,gd,'linear')
AFinal=[A1]

%Stress Calculation function testdifferential_std.m

testdifferential_std(AFinal,BFinal)
200
201

Acknowledgements

This thesis represents my learning and work at the Institute of Materials Engineering –
University of Kassel.

Thanks begin with Prof. Dr.-Ing. Habil B. Scholtes, my doctoral thesis supervisor and
first reviewer. He gave me the opportunity to work at the institute and motivate me to
begin, to proceed and to finish this doctoral thesis. I will always remember his
invaluable scientific and professional advices as well as the kind personal contact with
him. I hold him with high esteem.

Next, I would like to thank Prof. Dr. rer. nat. Dipl.-Ing. A. Wanner from the Institute of
Materials Science and Engineering I at the University of Karlsruhe for his interest for
this work and his acceptance to be the second reviewer.

My gratitude is also to Prof. Dr.-Ing. A. K. Błedzki and Prof. Dr.-Ing. J. Hesselbach,


both from the University of Kassel, for being members of the committee of my
disputation.

I deeply appreciate the significant input in the form of useful discussion and help
concerning practical and theoretical problems of my work provided by Dr.-Ing. J.
Gibmeier.

Thanks are also due to the undergraduate students, who contribute with their work and
effort to the success of this thesis. Davide Iacovelli, Imad Alfituri, Pedro Pimentel,
Matthias Ohlwein and Nils Heinemann worked on substantial topics of the thesis. The
evaluation program, which was developed during this project, is a co production with
the assistant programmers Rajarajan Balraj, Nikunjkumar Parmar and Eduard Jäckel.

I would like to express sincere thanks to all my colleagues and all my friends for their
support, courtesy and the pleasant working atmosphere. Special thanks to
Christian Franz and Günther Fehrl for their valuable technical support. My friend and
“compadre” Hagen Peukert accompanied me with understanding and involved me in
202

interesting scientific and non-scientific discussions. Thanks to my friend Olga Drotlew


for her indescribable support and encouragement during the last year.

In closing, my parents deserve special acknowledgement for their love, support and
guidance during my life as well as during my whole educational procedure.
In der Reihe Forschungberichte
aus dem

der

ist bisher erschienen:

Stefan Jägg Rissspitzennahe Eigenspannungen und Ermüdungsrissausbreitung


des Stahls S690QL1 bei unterschiedlichen Beanspruchungsmoden

Martin Kornmeier Analyse von Abschreck- und Verformungseigenspannungen


mittels Bohrloch- und Röntgenverfahren

Igor Altenberger Mikrostrukturelle Untersuchung mechanisch


randschichtverfestigter Bereiche schwingend beanspruchter
metallischer Werkstoffe

Gerd Zöltzer Einfluß von Mikro- und Makroeigenspannungen auf das


Deformationsverhalten bauteilähnlicher Proben

Ulf Noster Zum Verformungsverhalten der Magnesiumbasislegierungen


AZ31 und AZ91 bei zyklischen und quasi-statischen
Beanspruchungen im Temperaturbereich von 20°C – 300°C

Jens Gibmeier Zum Einfluss von Last- und Eigenspannungen auf die Ergebnisse
instrumentierter Eindringhärteprüfungen

Juijerm Pathiphan Fatigue behaivor and residual stress stability of deep-rolled


aluminium alloys AA5083 and AA6110 at elevated temperature

Martin Krauß Zur thermischen Ermüdung der Magnesiumbasislegierung AZ31


und AZ91

Das könnte Ihnen auch gefallen