tomographic investigation of creep and creep damage under inhomogeneous loading conditions ·...
TRANSCRIPT
Tomographic Investigation of Creep and Creep Damage under Inhomogeneous Loading Conditions
Dissertation zur
Erlangung des Grades Doktor-Ingenieur
der Fakultät für Maschinenbau
der Ruhr-Universität Bochum
von
Federico Iván Sket aus Cruz del Eje, Argentinien
Bochum 2010
Dissertation eingereicht am: 13.01.2010 Tag der mündlichen Prüfung: 21.05.2010 Erster Referent: Prof. Dr.-Ing. Anke Rita Pyzalla Zweiter Referent: Prof. Dr. András Borbély
I
TOMOGRAPHIC INVESTIGATION OF CREEP AND CREEP
DAMAGE UNDER INHOMOGENOUS LOADING
CONDITIONS
Dipl.-Ing. Federico Iván Sket
ABSTRACT
Engineering components are usually subjected to inhomogeneous loading conditions,
when material behaviour cannot be easily predicted as a simple superposition of different
effects characteristic for uniform loading. To better understand complex material behaviour
under inhomogeneous loading adequate investigation methods are required. This thesis
investigates the applicability of X-ray microtomography for the study of creep and creep
damage under inhomogeneous loading. The thesis has two main parts describing effects
caused by inhomogeneous temperature and inhomogeneous stress distribution.
Exploiting the nondestructive nature of X-ray tomography the first part of the thesis
shows that in situ tomography during creep is well applicable to characterize changes in
specimen’s shape. A new method for the evaluation of the apparent activation energy of
steady-state creep is proposed, which is based on in situ monitoring the local cross-section of
a cylindrical specimen subjected to uniaxial load and linear temperature distribution. It is
shown that microtomography acting as a three dimensional extensometer enables the
evaluation of local strain-rates. Good agreement between strain distributions evaluated from
the real tomographic measurement and finite element models was obtained. Activation
energies obtained with the new method for stainless steel agree within an error of 5% with
values obtained according to the classical procedure.
The second part of the thesis describes damage distribution in a notched hollow cylinder
made of E911 steel and crept for 26,000 h under multi-axial stress. A detailed characterization
of creep cavities is presented, comprising relevant statistical information about the size,
shape, orientation, and cavity density along the notch radius. Using a proper shape descriptor,
a separation of non-coalesced and coalesced cavities could be made. The analysis of non-
coalesced cavities in terms of general power-law functions describing nucleation and growth
led to the conclusion that cavity growth in E911 steel is dominated by the constrained
II
diffusion mechanism. The cross-correlation analysis between damage and different stress
parameter distributions along the notch radius enabled emphasizing the important influence
on damage of stress triaxiality, maximum principal stress and the von Mises equivalent stress.
III
ACKNOWLEGMENTS
The present work was carried out during my three and half years activities as a scientific
co-worker in the Department of Material Diagnostics and Steel Technology of the Max-
Planck-Institut für Eisenforschung GmbH (MPIE) in Düsseldorf and at Helmholtz-Zentrum
Berlin GmbH (HZB), Berlin, Germany.
First of all I would like to express my thanks to my advisor Prof. Dr.-Ing. Anke
Kaysser-Pyzalla, now at HZB, who provided me constant guidance, fruitful suggestions and
discussions and the financial support that permitted me to carry out this work at the MPIE and
the HZB.
I am also very grateful to Prof. Dr. András Borbély, now at "Ecole des Mines de Saint
Etienne" for guiding me through the whole process of learning. As an expert in Material
Science, he gave me insightful comments and challenged me to refine thoughts. He made
possible this thesis with his guidance and the fruitful discussions.
I would also like to thank to Prof. Dr. Karl Maile, at MPA Stuttgart, for the
colaboration, discussions and creep samples provided for evaluation in this thesis.
I would also like to express my thanks to my colleagues Krzystof Dzieciol and Augusta
Isaac who have been very active discussion partners during this thesis and the synergy
generated during this time made possible the completion of this thesis. I´m very grateful to
have met you guys!
My special thank to Gerhard Bialkowski whose help in the experimental part was
essential for the development of this thesis. His good mood and efficiency was very important
to create a very nice working environment.
The friendly supportive atmosphere inherent to the whole working group contributed to
the final outcome of my studies, not only during the working hours at the institute but also
during the free time. To all of them, Dimas Souza, Rodrigo Coelho, Pedro Brito, Marcin
Moscicki, Leonardo Agudo, David Rojas, Orlando Prat, Haroldo Pinto, Carla Barbatti, Pedro
Silva, Jose García, Mauro Martin, Maitena Dumont, Aleksander Kostka, Derek Leach, Hauke
Springer, Lais Mujica, Rosario Maccio, Fernando and Andres Lasagni, Claudia Juricic and
Adelheid Adrian I am truly grateful.
I greatly acknowledge the staff of the ID15 and ID19 at the European Synchrotron
Radiation Facilities, specially Dr. T. Buslaps, Dr. di Michiel, Dr. Boller and Dr. Tafforeau,
for their helpful assistance during the tomography experiments with synchrotron radiation.
IV
My special appreciation goes to my parents, Hugo and Corina, my brothers Hugo and
Germán and my sisters in law Andrea and Luciana who were always there for me. Their love
and encouragement were the driving forces that kept me going in the difficult moments. My
girlfriend, María, who became a very important part of my life and made it easier and of
course I cannot forget my nephew and godson, Huguito, who made me part of his life even
being at 14,000 km away from me.
V
TABLE OF CONTENTS Chapter 1 – Introduction…………..………………………………...………………….
1.1 Background………………...…………………………………………………...
1.2 Preliminary works……………………………………………...……………….
1
1
3
Chapter II – Creep deformation, State-of-the-art……………………………………..
2.1 Creep of metals and alloys………………………………………………….......
2.2 Power-Law Creep…………………….…………………………………...........
2.3 Rate-controlling mechanism……………………………………………………
2.4 Methods for the determination of Activation Energy of Creep ……..…...……..
2.5 Creep damage under multiaxial stress………………………………………….
2.5.1 Experimental techniques for damage evaluation………………………….
2.5.2 Development of creep resistant steels……………………………………..
2.5.3 Diffusion models of cavity growth………………………………………...
2.5.3.1 Grain boundary diffusion controlled growth…………………………
2.5.3.2 Surface diffusion controlled growth………………………………….
2.5.3.3 Constrained diffusional cavity growth……………………………….
2.5.4 The cavity size distribution function……………………………………...
2.5.5 Prediction of creep rupture time (phenomenological approach)…………..
4
4
6
9
11
17
17
19
22
23
25
26
30
32
Chapter III – Experimental details……………………………………………………..
3.1 Uniaxial in situ experiments……………………………………………………
3.1.1 Materials and specimen geometry adapted to tomographic measurement...
3.1.2 Creep device for in situ experiments and testing conditions…………...…
3.1.3 Creep tests with nearly constant load or constant stress…………………..
3.2 Multi-axial creep experiments on E911 steel…………………………………...
3.2.1 Materials and specimen geometry adapted to tomographic measurement
conditions………………………………………………………………….
3.3 Synchrotron X-ray microtomography…………………………………………..
3.3.1 X-ray tomography: Principles…………………………………………...
3.4 Image processing………………………………………………………………..
3.4.1 Reconstruction and pre-processing………………………………………..
3.4.2 Identification of cavities…………………………………………………...
3.4.3 Image correlation for evaluation of tomographic data…………………….
37
37
37
38
40
42
42
43
44
46
46
48
50
VI
Chapter IV – Results………………………………………………………………….....
4.1 Method for evaluation of apparent activation energy of creep, aQ ……………
4.1.1 Temperature distribution calibration………………………………………
4.1.2 Image correlation for evaluation of tomographic data…………………….
4.1.3 Method for evaluation on samples with constant cross sectionaQ ......…...
4.1.3.1 In situ creep curve of brass (samples A)...............................................
4.1.3.2 aQ values according to literature and tomographic method………….
4.1.3.3 Damage development in brass (samples A)……………...…...……...
4.1.4 Method for evaluation on samples with varying cross sectionaQ ………..
4.1.4.1 aQ values according to classical and tomographic methods…………
4.1.4.2 In situ creep curves of steel (samples B and C)....…………………..
4.2 Microtomographic investigation of damage in E911 steel after long term creep
4.2.1 Conventional damage evaluation by OM………………………………….
4.2.2 Tomographic evaluation of damage……………………………………….
52
52
55
56
58
59
60
62
64
64
66
70
71
73
Chapter V – Discussion……………………………………………………...…………..
5.1 Evaluation of apparent activation energy of creep, aQ ……...................………
5.1.1 Comparison of strain distribution in the real sample and FE models…......
5.1.2 Error estimation……………………………………………………………
5.2 Damage investigation in E911 steel…………...…………………...…………...
81
81
81
84
85
Chapter VI – Conclusions and perspectives………………………............……….….. 92
Chapter VII – References……...…………..……………………………….……….….. 94
Curriculum vitae……...…………..………………………………...…….……...….….. 105
I. Introduction 1
CHAPTER I
INTRODUCTION 1.1 Background
Creep deformation of metals at temperatures higher than about 1/3rd of the melting
temperature is governed by several thermally activated processes enhancing dislocation
motion and recovery, diffusion of vacancies as well as dynamic recrystallization [
mT
1]. These
mechanisms influence mainly stage II of creep, where specimens under constant stress deform
usually at nearly constant strain-rate. The relationship between this steady-state strain-rate and
stress in pure metals and Class M alloys is often described by a power-law function [1,2]. The
key parameters of this phenomenological description are the stress exponent, , and the
apparent activation energy for steady-state creep, , which once evaluated give good hints
about the main creep mechanism. It is the aim of the first part of this work to present a new
evaluation method of based on in situ microtomography.
n
aQ
aQ
Due to the economical importance of creep at high temperatures much attention has
been given to the mechanisms of intergranular cavities growth. Creep rupture by cavity
growth and coalescence is an important failure mechanism for high temperature components
for example in power plants. Knowledge of damage development during high temperature
creep of materials is extremely important for the prediction of service lifetimes of many
engineering components. The subject was extensively studied both theoretically [3,4,5,6] and
experimentally [7,8,9]. Review articles [10,11,12] and books e.g. [13] are presenting the main
achievements in the field, but it is nevertheless quite difficult to compare existing
experimental data with model predictions. The reason is twofold:
a) There are several damage mechanisms acting simultaneously such as void nucleation,
void growth by diffusion, by plasticity or by coupling between them. Due to this complexity
global damage characterization techniques like Archimedean densitometry, optical/scanning
electron microscopy or small angle X-ray/neutron scattering [7,14,15] are usually inadequate
to conclude on the validity of a given damage model. Usually difficult assumptions have to be
made about the fraction and coupling among different mechanisms. The problem is further
complicated by the absence or inaccurate knowledge of creep material paramters, especially
for engineering materials, required for a quantitative check of the models.
I. Introduction 2
b) Due to the lack of adequate experimental techniques generally no complete three-
dimensional (3D) evaluation of microscopic damage was performed. The destructive nature of
some techniques (e.g. optical or electron microscopy) hinders a univocal characterization of
damage evolution, since data obtained from two-dimensional (2D) sections are inherently
affected by statistical fluctuations as well as by errors related to the 3D connectivity of larger
voids.
Under multiaxial and locally varying stress creep damage develops inhomogenously
and a quantitative assessment is very difficult by conventional metallographic techniques.
Creep lifetime and failure mode under multiaxial stress conditions can differ significantly
from the rules established for uniaxial loading. For example Cocks and Ashby [5] have
predicted that cavity growth is highly accelerated by high positive stress triaxiality. There are
numerous applications in which engineering components are subjected to inhomogeneous
loading conditions of temperature and stress. In a steam power plant for example the most
exposed components are usually the boiler tubes, headers and turbines, where these materials
are subjected to inhomogeneous loading. In case of a hollow cylinder for example the
multiaxiality of the stress state is imposed by internal pressure, additional longitudinal force
as well as sample geometry. The assessment of the lifetime and failure modes of such
industrial components needs the evaluation of damage on samples subjected to similar
conditions as in service. The stress parameters usually used for the assessment of rupture life
are, however, averages over the local values in the sample. Multiaxial loading of large
specimens leads, however, to inhomogenous damage indicating that local parameters should
be considered for appropriate understanding of the results. The high spatial resolution
achievable with synchrotron microtomography (of about 0.33 μm) plus the possibility to
reconstruct representative volumes makes this technique very attractive for the 3D
characterization of inhomogeneous damage distribution.
It is the aim of the second part of this thesis to assess the damage in a notched hollow
cylinder of E911 steel, which was subjected to about 26,000 h of creep of similar loads as in
service. The 3D shape, orientation and size distribution of single cavities will be described. A
comparison of local damage distributions with the distribution of selected stress parameters
(knowing to influence damage) as obtained from FE simulation will be performed. The
tomographic results will also be compared with those of conventional metallography.
I. Introduction 3
1.2 Preliminary works
This thesis was preceded by the following publications:
(1). F. Sket, A. Isaac, K. Dzieciol, A. Borbély, K. Maile, A.R. Pyzalla, Microtomographic
investigation of damage in E911 steel after long term creep, Mat. Sci. Eng. A, to be
submitted.
(2). F. Sket, K. Dzieciol, A. Isaac, A. Borbély, A.R. Pyzalla, Tomographic method for
evaluation of apparent activation energy of steady-state creep, Mat. Sci. Eng. A.
(2009) accepted for publication.
(3). F. Sket, A. Isaac, K. Dzieciol, G. Sauthoff, A. Borbély, A.R. Pyzalla, In situ
tomographic investigation of brass during high-temperature creep, Scr. Mater. 59
(2008) 558-561.
(4). K. Dzieciol, A. Isaac, F. Sket, A. Borbély, A.R. Pyzalla, Application of correlation
techniques to creep damage studies, Collected Proceedings TMS Conference, San
Francisco 15-19 February (2009), Characterization of Minerals, Metals and Materials,
pp. 15-22.
(5). A. Isaac, F. Sket, W. Reimers, B. Camin, G. Sauthoff, A.R. Pyzalla, In-situ 3D
quantification of the evolution of creep cavity size, shape, and spatial orientation using
synchrotron X-ray tomography, Mat. Sci. Eng. A 478 (1-2) (2008) 108-118.
(6). A. Isaac, F. Sket, A. Borbély, G. Sauthoff, A.R. Pyzalla, Study of cavity evolution
during creep by synchrotron microtomography using a volume correlation method,
Praktische Metallographie/Practical Metallography 45 (5) (2008) 242-245.
(7). A. Isaac, K. Dzieciol, F. Sket, M. di Michiel, T. Buslaps, A. Borbély, A.R. Pyzalla,
Investigation of creep cavity coalescence in brass by in-situ synchrotron X-ray
microtomography, In: S.R. Stock, Editor, Developments in X-Ray tomography VI,
Proc. of SPIE, 7078 (2008) pp. J1-J10.
II. State of the Art 4
CHAPTER II
CREEP DEFORMATION
STATE-OF-THE-ART
2.1 Creep of metals and alloys
The strain response of a body to the applied stress varies with the magnitude and state
of stress, temperature, and strain-rate. At homologous temperatures ( , where Tm is the
absolute melting point) above 0.3 it is a reasonable and widely used idealization to consider
the elastic-plastic behavior of metals as time-dependent. When a material undergoes
continuous deformation under constant load or stress it is said to creep and this may include
elastic, viscous and plastic deformations. The plasticity under these conditions is described in
Figure 2.1 for constant load. The instantaneous strain
m/ TT
0ε (Figure 2.1 (a), (b), and (c)) is
obtained immediately upon loading and exhibits characteristics of plastic deformation.
In Figure 2.1 (a) three regions are observed. Stage I, or primary creep, which denotes
the portion where the creep-rate (plastic strain-rate) dtd /εε =& is changing with increasing
plastic strain or time (until 1ε and time 1t Figure 2.1 (a) and (d)). The decreasing creep-rate
in the primary creep stage has been attributed to strain hardening and to the decrease of the
density of mobile dislocations.
in
Stage II, where the strain-rate has a constant value over a range of strain. This
phenomenon is also termed secondary or steady-state creep and is commonly attributed to a
state of balance between the rate of generation of dislocations contributing to hardening and
the rate of recovery contributing to softening. Beyond 2ε (Figure 2.1 (a)) or (Figure 2.1
(b)) cavitation and/or cracking increase the apparent strain-rate continuously until rupture
occurs at the strain
2t
rε and rupture time . rt
The region with increasing creep-rate is called Stage III or tertiary creep and is followed
by fracture. The increase in creep-rate with time in the tertiary creep stage can be a
consequence of increasing stress (due to necking of the specimen) or of microstructure
evolution (dynamic recovery, dynamic recrystallization, coarsening of precipitates and other
phenomena, which cause softening and result in a decrease in resistance to creep) and damage
evolution (development of creep voids and cracks, often along grain boundaries) during creep.
II. State of the Art 5
Figure 2.1. (a), (b) and (c) Creep curves of engineering materials under constant tensile load
and constant temperature and (d), (e) and (f) their creep-rate curves as a function of time [16].
The type of creep curve depicted in Figure 2.1 (a) and (d) is not always observed during
creep of metals and alloys. Whether all creep stages are observed in any test depends on
temperature, stress, and the duration of the test. Under certain conditions, the tertiary creep
stage begins immediately after the primary creep and predominates over the other stages [13],
as shown in Figure 2.1 (b) and (e). In this case, the steady-state creep-rate ssε& is defined as
minimum creep-rate minε& . The minimum creep-rate can be also explained by the process
where hardening in the primary stage is balanced by softening in the tertiary stage. In many
cases, there is substantially no steady-state stage suggesting that there is no dynamic
microstructural equilibrium during creep. At low homologous temperatures ( ), often
less than 0.3, only primary creep appears. At these temperatures diffusion is not important and
often the strain is well below 1% and it doesn’t lead to final fracture (Figure 2.1 (c) and (f)).
This deformation process is designated as logarithmic creep.
m/TT
II. State of the Art 6
2.2 Power-Law Creep
The creep-rate at constant stress usually increases exponentially with temperature
[17,18] following an Arrhenius-type law. The creep-rate also depends on the stress according
to a power law, i.e. , is the stress exponent. In pure metals and Class M alloys
there is an established, largely phenomenological relationship between the steady-state strain-
rate and stress which is often described by a power-law function [
nσε ∝ss& n
1,2]:
2.1) ⎟⎠⎞
⎜⎝⎛
⋅−⋅⎟⎟
⎠
⎞⎜⎜⎝
⎛⋅=
TRQ
TGC
n
ass exp
)(σε&
where is the apparent activation energy for creep, is the gas constant, the pre-
exponential factor C - which comprises all other structure-dependent parameters - is assumed
to be constant [
aQ R
19], )(/ TGσ is the shear modulus normalized stress (some authors are using
the Young’s modulus ), is the stress exponent, and )(TE n T the absolute temperature. The
values of are often found to be about that of lattice self-diffusion, . The stress
exponent, , is in the range of 4-7 and it is constant over a relatively large range of
temperatures and strain-rates. Eq.
aQ
n
SDQ
2.1 is often referred to as Norton’s law, power-law or five-
power-law (due to the typical value of the stress exponent). Steady-state creep is often
emphasized over primary or tertiary creep due to the relatively large fraction of creep life
within this regime. The concept of five-power-law creep, however, must be described in the
context of other (usually commercially less important) types of creep (e.g. Nabarro-Herring,
Harper-Dorn, Coble, low-temperature creep (Power Law Breakdown - PLB) as well as three-
power viscous glide creep), as illustrated in Figure 2.2.
The importance of steady-state is evidenced by the empirical relationship suggested by
Monkman and Grant [20], which shows that the overall time to fracture is controlled by the
steady-state creep-rate (
ft
ssε& )
2.2) MGssf Ct m =⋅ε&
where is a constant depending on total elongation during creep and is a constant
often nearly equal to 1. The Monkman-Grant relationship is a surprisingly simple observation
MGC m
II. State of the Art 7
which has been experimentally confirmed not only for simple metals and alloys [21,22,23]
but also for a number of engineering creep-resistant steels [16,24]. The wide range of
conditions for which the relation has been confirmed demonstrates that the Monkman-Grant
constant ( ) is independent of stress and temperature [MGC 21,22,23] and its approximate
constancy places constraints on creep cavitation theories.
Figure 2.2. Ashby deformation map of pure copper [25] with a grain size of 0.1 mm,
including power law breakdown.
The experimentally proven temperature dependence of creep-rate described by Eq. 2.1
reflects the fact that creep involves thermally activated processes and thermally activated
micromechanisms operating on the atomic scale. The first step in determining the
micromechanisms controlling the creep-rate is to obtain the activation energy. If one of these
mechanisms dominates, the activation energy of creep will be identical with the activation
energy of this mechanism. Creep, however, is a very complex phenomenon, and under certain
conditions several mechanisms may be operating simultaneously. The physical interpretation
of the activation energy determined under these conditions requires further consideration.
Here, two cases can be distinguished [26,27].
II. State of the Art 8
The first one, when the processes are independent or parallel-concurrent, they are active
simultaneously and contribute with a strain iε , the strain-rates are additive and the Arrhenius
plot is the sum of exponential functions, one for each process (given by Eq. 2.1). However,
the frequent case is that no more than two processes contribute to the creep-rate to a
comparable extent. The mechanisms usually have different activation energies and so
contribute significantly to the creep-rate only over a relatively narrow temperature interval. In
any other temperature range, the fastest process will dominates creep.
The second case occurs when the processes are dependent and act in series (often called
series-sequential). In case of two processes acting in series-sequential, the second process
cannot operate until the first one has taken place and vice versa (mutually accommodating
processes). The time necessary for occurrence of the first and second process are additive.
Then, in any given temperature range, the slower process will control the creep.
Figure 2.3. Arrhenius plot for a) Parallel-concurrent processes and b) Series-sequential
processes [28].
A knowledge of the activation energy of creep, , and the stress exponent, n , give
indications about the main creep process and the atomic mechanisms controlling the creep-
rate under a given external condition (temperature and applied stress). In the following
sections, the rate-controlling mechanisms are briefly described and then a concisely
description of the state-of-the-art of the usual evaluation procedures for the activation energy
determination is given which will lead to the results of the first part of the thesis in which a
new method for activation energy determination based on tomographic measurements is
presented.
aQ
II. State of the Art 9
2.3 Rate-controlling mechanism
Several mechanisms can be responsible for creep; the rate-controlling mechanism
depends both on the stress and on the temperature. For the temperature range higher than
about the creep mechanisms can be divided into two major groups: m4.0 T⋅
• Grain boundary mechanisms: in which grain boundary and, therefore, grain size,
play a major role.
• Lattice mechanisms: which occur independently of grain boundaries.
Dislocation glide is the most important mechanism of plastic deformation. In the course
of plastic deformation by dislocation glide alone, the dislocation density increases, leading to
an increasing flow stress with the strain. This mechanism takes place at relatively low
temperatures and high strain-rates (in the stress range 210 ). Below the ideal shear
strength, flow by the conservative motion of dislocations (glide) is possible provided an
adequate number of independent slip systems are available. This motion is almost always
obstacle-limited, i.e. by the interaction of mobile dislocations with other dislocations, with
solute precipitates, with grain boundaries, or with the friction of the lattice itself which
determines the rate of flow and (at a given rate) the yield stress. Dislocation glide is a kinetic
process, and the strain-rate produced by the average velocity of mobile dislocations, m
/ −≥Gσ
ρ ,
moving through a field of obstacles is almost entirely determined by their waiting time at
obstacles [29]. In the most interesting range of stress, the mobility of the dislocations and thus
their average velocity is determined by the rate at which dislocation segments are thermally
activated through, or around, obstacles. Two classes of obstacles can be defined: discrete
obstacles which can be bypassed or cut by a moving dislocation, depending on the
temperature and stress; and extended obstacles, diffuse barriers to dislocation motion (e.g.
lattice friction or a concentrated solid solution).
When the homologous temperature at which plastic deformation take place is higher
than 0.4, dynamic recovery starts to play an important role. Orowan proposed that creep is a
balance between the work-hardening (due to plastic deformation) and recovery (due to
exposure at high temperatures). The recovery during creep can occur by various mechanisms.
The most important of them involves non-conservative motion of dislocations (e.g. climb) and
annihilation of dislocations, and therefore depends on diffusion which can occur either via the
lattice (higher homologous temperature) or via dislocation cores (lower homologous
temperature). The non-conservative motion of dislocations (if the dislocations are properly
arranged) can in itself represent a mechanism of plastic deformation [30,31]. In the stress
II. State of the Art 10
range , creep tends to occur by dislocation glide, aided by vacancy
diffusion (when an obstacle is to be overcome).
24 10/10 −− ≤≤ Gσ
Another deformation mechanism, which usually participates in polycrystals at
homologous temperatures higher than 0.4, is grain boundary sliding (GBS). Under the action
of applied stress, in the diffusional creep regime, GBS is accommodated by atoms transported
from boundaries subjected to compressive stress, to those of subjected to tensile stress. This
leads to changes in shape of individual grains. At higher stresses, in the dislocation creep
regime, grain boundary sliding is accommodated by dislocation glide. Dislocations move by
glide and climb, so as to remove or supply matter to the part of boundary which needs it. If
GBS is not accommodated by either diffusion of vacancies or dislocation glide, an extreme
heterogeneity of plastic deformation occurs. This leads to the reduction of polyrcystal
compatibility by formation of voids on grain boundaries.
Diffusion creep occurs for (this value to a certain extent depends on the
metal). Two mechanisms are considered important in this region. The first one takes place at
high temperatures (about 0.7·Tm) by diffusion of vacancies via lattice (Nabarro-Herring creep)
producing an increase in the length of the grain along the direction of the applied (tensile)
stress. The second one is based on diffusion along the grain boundaries (instead of bulk) and
it is called Coble creep. It occurs at lower temperatures than Nabarro-Herring creep and
results in sliding of the grain boundaries. The diffusional creep-rate is inversely proportional
to the second power of mean grain diameter when the diffusion mainly occurs via the lattice
(Nabarro-Herring creep), and to the third power of mean grain diameter when it occurs
through grain boundaries (Coble creep). Harper and Dorn observed another type of
diffusional creep in aluminium [
410/ −≤Gσ
32], which occurred at high temperatures and low stresses,
and the creep-rates were over 1000 times greater than those predicted by Nabarro-Herring
creep. They conclude that creep occurred exclusively by dislocation climb. A significant
contribution of Harper-Dorn creep occurs for large grain sizes (> 400µm).
When the integrity of the crystal is damaged, i.e. the incompatibility created by the
grain boundary sliding is not accommodated by any of the above mentioned mechanisms, the
nucleation and especially the growth of cavities or cracks appear at the grain boundaries
which lie mainly perpendicular to the tensile axis [3,33] (proceeding by grain boundary
diffusion, intragranular deformation, grain boundary sliding or by various combination of
these processes). Thus, the cavities naturally contribute to the measured strain.
In the light of the many possible mechanisms that lead to creep rupture, it follows that
the plastic strain as measured in a tensile creep test generally consists of several components:
II. State of the Art 11
2.3) vdgbndg εεεεεε ++++=
where dgε , nε , gbε , dε , vε are the strain caused by dislocation glide, non-conservative
motion of dislocations, grain boundary sliding, stress directed diffusion of vacancies and by
intercrystalline void nucleation and growth, respectively. Not all processes operating during
creep as mentioned in Eq. 2.3 are independent of each other, as frequently assumed. An idea
about the possible mechanism dominating creep under a given condition is provided by the
deformation mechanisms maps [25] on the assumption that all deformation mechanisms
considered are mutually independent, and thus operate in a parallel way. It is also based on an
application of constitutive equations describing stress, temperature, grain size and stacking
fault energy dependence of creep-rates due to individual mechanisms. Figure 2.2 shows the
deformation map for copper where several regimes are illustrated as a function of temperature
and grain size. The dislocation creep regime is indicated as power-law creep.
2.4 Methods for the determination of Activation Energy of Creep
Based on Eq. 2.1 the apparent activation energy for creep, , is defined as: aQ
2.4) ( ) sσ/G,a /1
)(ln⎥⎦
⎤⎢⎣
⎡∂∂
⋅−=T
RQ ε&
where G/σ and “s” indicate that the data should represent constant normalized stress and
constant structure conditions. In practice, the most frequently used method consists in
performing several creep tests at constant normalized-stress and different temperatures and
is obtained from the slope of the line fitting the steady-state creep-rates vs. the inverse of
temperature (1/T). Nix et al. [
aQ
34] have compiled the activation energies of creep available in
the literature for a large class of metals and compared them with the corresponding activation
energies of lattice self-diffusion, , Figure 2.4. They have shown that the two quantities
are essentially equal, which supports the idea that the mechanism of five-power-law creep
(with stress exponent ≈ 5) is strongly related to diffusion of vacancies.
SDQ
n
II. State of the Art 12
Figure 2.4. Activation energy ( ) and volume (SDQ LVΔ ) for lattice self-diffusion versus
activation energy ( ) and volume (aQ aVΔ ) for creep for various metals (from [34]).
Sherby et al. [35], using the same classical method, have obtained the activation energy
of pure aluminum over a large range of temperatures, from about m3.0 T⋅ to . Figure
2.5 a) illustrates the data obtained by Sherby et al. on a strain-rate versus modulus-
compensated stress plot allowing a direct determination of the activation energy for creep.
Their results confirm that above
m9.0 T⋅
m6.0 T⋅ , is comparable to that of lattice self-diffusion, but
below it decreases below . Luthy et al. [
aQ
m6.0 T⋅ SDQ 36] obtained similar results for aluminum
performing torsion creep tests over a range of temperatures from about to mT⋅ 029.0 m93. T⋅ .
Figure 2.5 b) shows another example for calculation of on a strain-rate versus inverse
temperature plot for silver [
aQ
37]. Above m46 T.0 ⋅ the activation energy for silver is about
195 kJ/mol. Below this temperature it undergoes a transition and it is somewhat lower. In the
temperature range from to m 63.046.0 T⋅ mT⋅ the activation energy is in close agreement with
the activation energy for lattice diffusion of silver of 190 kJ/mol. The parallel lines in this
temperature range indicate that the activation energy is not stress dependent. Below m46. T0 ⋅ ,
the lower value of activation energy is probably related with dislocation core diffusion [37]. A
fair amount of experimental evidence on different materials indicates that the effect of the
II. State of the Art 13
stress on the values of is small at high temperatures [aQ 38,39,40]. However, in the low
temperature regime the activation energy can be stress dependent as indicated for example in
[40].
An advantage of the classical method is related to the broad range of temperatures and
stresses for which the method can be applied and consecutively to the possibility to determine
an eventual stress or temperature dependence of the activation energy.
Figure 2.5. (a) Steady-state strain-rate versus modulus-compensated stress for aluminium at
various temperatures [35]. (b) Steady-state strain-rate versus inverse temperature for silver at
three different modulus-compensated stress [37].
The definition of activation energy (Eq. 2.4) indicates that only tests performed at
constant normalized stress, )(/ TGσ and characterized by identical structures can be used for
evaluation. This requirement seems to be partly satisfied already by tests performed at
constant )(/ TGσ , when the average subgrain sizes and average distance between free
dislocations were found to obey a power law relationship with )(/ TGσ [41]. Subgrain
misorientations, however, are not constant and slightly evolve with strain [42]. The
importance of subgrain boundaries in creep can be related to their role played in the recovery
of free dislocations, controlling by this the creep-rate [43]. This indicates that samples at
different temperatures and strains have slightly different structures and suggests that an
II. State of the Art 14
evaluation method based on the same specimen is more adequate [44-46]. Since temperature
appears in the exponent of Eq. 2.1 a small temperature change can have a large effect on the
strain-rate. Therefore changing the temperature by a small amount might
significantly change the creep-rate
12 TTT −=Δ
1ε& to 2ε& . If the structure does not change much over the
TΔ range the creep-rates and temperatures can be related as [44-46]:
2.5) ⎟⎟⎠
⎞⎜⎜⎝
⎛−⋅=⎟⎟
⎠
⎞⎜⎜⎝
⎛⋅
12
1 11lnTT
Rεε&
&
2aQ .
The procedure can be repeated several times for different temperature jumps and the
activation energy can be obtained from the slope of the line fitted to the logarithm of strain-
rate ratios as a function of the difference of the inverse temperatures (Eq. 2.5). Its application
is illustrated in Figure 2.6. values evaluated according to this method for Al and Ni can be
found in Refs. [
aQ
38,47].
Figure 2.6. Determination of the apparent activation energy by the “differential” method [48].
II. State of the Art 15
The activation energies obtained by Refs. [35] and [36] by the classical method for
aluminum are practically the same in the high temperature range (above ) of about
138 kJ/mol and 135 kJ/mol, respectively. Evaluations according to the “temperature jump” or
“differential” method yielded = 147 kJ/mol for aluminum in the same temperature range
[
m6.0 T⋅
aQ
47], which is higher by about 7% compared to results of the classical procedure.
The disadvantage of the differential method for the activation energy determination is
the unavoidable inertia of the furnace in which the temperature must be quickly increased or
decreased specially at high temperatures. The method is applicable at relatively high creep-
rates only, therefore, it has found relatively little use.
Another method for evaluation of relies in the experimentally proven fact that at
homologous temperatures higher than about 0.5 the creep strain is a single valued function of
the temperature-compensated time , i.e. that
aQ
Θ
2.6) constant),( =Θ= σε f .
where is defined by the equation Θ
2.7) dtTR
Qt
∫ ⎥⎦⎤
⎢⎣⎡
⋅−=Θ
0
aexp .
Differentiating Eqs. 2.6 and 2.7 with respect to time and combining them we obtain
2.8) constant,)(exp a ==Θ=⎥⎦
⎤⎢⎣
⎡⋅
⋅ σε ZFTR
Q& .
where Z is the Zener-Hollomon parameter [49] or the temperature-compensated creep-rate
[48]. In this method several creep curves at various temperatures are obtained (over not too
broad range) at a constant stress (Figure 2.7 a). A value of can be determined which
brings together the curves into a single one when plotting the creep strain,
aQ
ε , against the
temperature compensated time, , as shown in Figure 2.7 b). The value of obtained with
this method by Dorn [
Θ aQ
mT⋅48] for aluminum in the temperature range from to 45.0 m57.0 T⋅ is
142 kJ/mol. It is in agreement with values from the last two methods described. A second
variant of this method assumes a direct association between structure and creep strain in the
II. State of the Art 16
form )(exp a εε gTR
QZ =⎥⎦
⎤⎢⎣
⎡⋅
⋅= & ,[48]. Experimental data from Figure 2.7 a) plotted according
to this relation show very good correlation for the above value of Q , Figure 2.7 c). a
Figure 2.7. Determination of apparent activation energy of creep by means of temperature
compensated time Θ for aluminum. (a) Creep curves at various temperatures for a given
stress of 21 MPa. (b) Dependence on creep strain on temperature compensated time. (c)
Variation of Z parameter with creep strain [48].
Based on the concept that the effective stress (σ*) for dislocation motion is the
difference between the applied stress σ and the internal stress σi created by neighboring
dislocations and other obstacles, Ahlquist et al. [50] have introduced two other definitions of
the activation energy and at constant effective and internal stress, respectively: *Q iQ
2.9) *)/1(
ln*
σ
ε⎥⎦
⎤⎢⎣
⎡∂∂
−=T
RQ s&
.
II. State of the Art 17
2.10) i
TRQ s
iσ
ε⎥⎦
⎤⎢⎣
⎡∂∂
−=)/1(
ln &.
They suggested that the former activation energy characterizes dislocation glide,
whereas the latter the recovery process.
Later Dobeš and Milička [51] have suggested the determination of the activation energy
at the same structural state, i.e. at an identical level of the internal stress, and at the same
applied stress:
2.11) σσ
ε
,)/1(ln
iT
RQ si ⎥
⎦
⎤⎢⎣
⎡∂∂
−=∗ &.
They measure the internal stress at a temperature and applied stress 1T 1σ by the dip
test technique [52,53]. Then a stress change experiment is performed at a temperature and
applied stress
2T
2σ to which the same internal stress value corresponds. They argued (using
experimental data for the aluminum alloy Al–13.7 wt.% Zn) that the extrapolated value of
to zero applied stress becomes identical to the apparent activation energy of creep obtained
from classical tests. The constant structural stress state condition in this definition resembles
the constant structure condition of Eq.
∗iQ
2.4.
It is the aim of the first part of this thesis to demonstrate the adequacy of in situ
tomography for the evaluation of activation energy of creep based on a single specimen by
monitoring its deformation in situ. The method is exemplified on brass and stainless steel
specimens deformed under different conditions and the tomographic outcome is compared
with results from literature and obtained according to the traditional method.
2.5 Creep damage under multiaxial stress
2.5.1 Experimental techniques for damage evaluation
Several techniques have been used to study creep cavitation on grain boundaries. For
example, density measurements allow for the determination of the total cavity volume [54].
II. State of the Art 18
Small angle scattering has also been employed to measure the total cavity volume, and under
certain favourable conditions the cavity size distribution can be also obtained [7,14,15].
Optical microscopy and scanning electron microscopy (SEM) are widely used tools to study
cavity distribution and cavity shape [8,9,55-58]. Damage characterization by these techniques,
however, has the drawback that the images analyzed may contain artefacts due to sample
preparation [59]. These drawbacks were to some extent overcome by e.g. the combination of
SEM with a two stage creep device [60]. In this technique a small additional increment of
creep strain is applied to the polished specimens to open up small cavities, which may have
been closed by polishing. Needham and Gladman [61] make small cavities visible by ion
beam etching of the polished sections. For understanding the mechanism controlling damage
accumulation, creep tests have been interrupted at various stages and crept specimens were
fractured in a brittle manner at low temperatures [60,62,63,64]. These techniques have been
used to obtain information about the cavities size and growth kinetics. Since the same cavity
could not be followed due to the descructive nature of the techniques an attempt was made to
measure only the largest cavities (the ith largest or the 100 largest cavities) considering that
they have nucleated at zero time [65]. Creep damage under multiaxial stress develops
inhomogenously and a quantitative assessment is even more difficult by conventional
metallographic techniques. Therefore it has been usually done qualitatively and compared
with results from finite element models (FEM) [66-72]. Serial sectioning combined with
metallographic techniques has been used in an attempt to obtain quantitative information on
damage distribution in specimens subjected to multiaxial creep, such us number of creep void
and area fraction of voids [73-76]. These methods, however, do not allow resolving the real
three dimensional complex shape of the cavities, their connectivity and spatial distribution,
inherent three-dimensional (3D) parameters [77].
More recently, Synchrotron X-ray Microtomography (SXRM) has emerged as a new
powerful technique [78-87] that enables a 3D characterization of heterogeneous
microstructures. The high resolution (~0.33 μm) achievable with microtomography turns this
technique into an excellent tool for the characterization of creep damage at the micrometer
scale being well suited for the study of samples subjected to inhomogeneous loading
conditions. Furthermore, sufficiently large volumes can be measured, yielding relevant
quantitative information: (i) the morphology of cavities, (ii) cavity spatial orientation, (iii)
minimum distance between cavities, (iv) the connectivity between cavities can be captured,
whereas voids may appear separated on two-dimensional (2D) projections. Another advantage
of this technique is that no careful specimen preparation is needed. All the information is
II. State of the Art 19
saved in a digital file containing the data of a 3D volume. Its extraction, however, is not a
trivial task. It is the aim of the present investigation to asses the damage distribution in
notched hollow cylinder subjected to uniaxial load and inner pressure, to compare the results
obtained from tomography with those of conventional metallographic techniques and to
evaluate the role of different stress parameters on damage distribution within the notch.
2.5.2 Development of creep resistant steels
Creep cavitation concerns all industries where structural materials are subjected to
mechanical loads at high temperatures. For example, in power plants the increase in thermal
efficiency is achieved by increasing the temperature and, to a lesser extent, the pressure steam
entering in the turbine. The increase in temperature requires new concepts in alloy design. In
this section, a summary of the development of creep resistant steels and published literature
on cavitation in different steels is given.
Investigations carried out with different Mo, Cr, Ni, V, CrMo, MnSi, MoMnSi,
CrSiMo, CrNiMo, CrMnV and CrMoV contents brought forth low alloyed steels for high
temperature applications. In the 1950s, MoV steel (0.14%C-0.5%Mo-0.3%V) was developed
with higher creep strength for steam plants. Molybdenum was recognized as an important
element for improving high temperature strength if the content is about 0.5%. The strength of
these steels is due to solution hardening and Mo2C precipitation. However, a drawback of Mo
alloying over about 0.35% is a marked decrease in ductility under creep conditions as well as
graphite precipitation. This was overcome by the addition of Cr in amounts of 1% and 2.25%.
Steel with chemical composition of 0.15%C-0.3-0.5%Mo, 0.13%C-1%Cr-0.5%Mo and
0.10%C-2.25%Cr-1%Mo were developed and are still in use today. Microstructure
investigations revealed M3C, M7C3 and M23C6 precipitation in 0.13%C-1%Cr-0.5%Mo steel
and Mo2C and M23C6 in 0.10%C-2.25%Cr-1%Mo [88]. In the field of turbine manufacture,
steel with approximately 0.25%C-1.25%Cr-1%Mo-0.3%V was developed. Its resonable
strength is due to finely distributed and thermally very stable V4C3 precipitates and Mo2C.
The highest creep strength of this steel is achieved with an upper bainite structure. In the
1980s, two new low-alloyed heat-resistant steels have been developed with higher creep
strength than the aforementioned ones. The creep strength was achieved by the addition of
Nb, N and B. The names and chemical composition are HCM2S (0.06%C-2.25%Cr-2%Mo-
II. State of the Art 20
1.6%W-0.25%V-0.05%Nb-0.02%N-0.003%B) and 7CrMoVTiB (0.07%C-2.4&Cr-1.0%Mo-
0.25%V-0.07%Ti-0.01%N-0.0004%B).
Low-alloy ferritic steels are widely used in the industry at temperatures up to 550°C.
Under service conditions, these steels usually operate within a creep regime where failure
occurs at low ductilities owing to the nucleation and growth of grain boundary cavities.
Numerous investigations were performed in 1%Cr-0.5%Mo, 2.25%Cr-1%Mo and 0.5%Cr-
0.5%Mo-0.25%V steels which are still in use today [61,89,90,91].
With further increase of service temperature, ferritic-martensitic steels with higher Cr
content (typically 9-12%) were developed operating up to 600-650°C. The X22CrMoV-12-1
steel was developed in the 1950s for thin- and thick-walled components up to 566°C. Its
strength is based on solution hardening and on the precipitation of M23C6 carbides. In the
1970s, a newer steel generation the 9CrMo or P91 (designation of the ASTM specifications)
has been developed at the Oak Ridge National Laboratories (USA) [92] for the manufacture
of pipes and vessels for a fast breeder and found broad application in power stations with
steam temperature up to 600°C. The increase of the creep strength in comparison with the
12%CrMoV steel was obtained by secondary MX precipitation of the type VN and Nb(C,N)
due to the addition of 0.05% Nb, 0.2%V and 0.05% N. More recent developments in Japan,
Nippon Steel Corporation, led to the steel NF616 (9%Cr, 0.5%Mo, 2%W) – also designated
as P92 in the ASTM specifications. A further increase in the rupture strength was obtained by
the addition of 1.8% W, 0.003% B and the reduction of the Mo from 1% to 0.5%. The
addition of boron gives thermally stabile M23(C,B)6 precipitates whereas the higher W content
leads to a higher amount of the Laves phase (Fe,Cr)2(Mo,W). In Europe, similar research
activities of the Cooperation in Science and Technology (COST) Action 501 [93] led to the
development of steel E911, which with a similar amount of Cr (9%), less W (1%) and more
Mo (1%) offers similar rupture strength as P92. The Ni content is also low (0.07%), in the
steels P92 and E911, since due to the low Cr content there is no risk of the occurrence of delta
ferrite. Further ferritic 9-10%Cr steels are under development for steam temperature up to
650°C [16]. Figure 2.8 shows the maximum operating temperature of different ferritic and
austenitic steels for 100,000 h creep rupture strength of 100 MPa. The rupture strengths of
high Cr ferritic-martensitic steels are comparable to those of austenitic stainless steels.
Creep failure of high Cr steels is generally caused by loss of strength from accumulated
microstructural damage such as coarsening of precipitates and critical decrease in dislocation
density due to thermally-activated recovery processes accompanied by growth and
II. State of the Art 21
coalescence of creep cavities. Publications on cavitation in 9-12% Cr steels are, however, less
frequent [66,74,75,76,94,95].
Austenitic steels were developed for chemical plant equipments and used in various
corrosion and oxidation environments. Austenitic steels exhibit not only good corrosion
resistance but also very high creep rupture strength and therefore they have found applications
for pressure vessels of fast breeder reactors. They can be used at higher temperatures than
ferritic steels, up to approximately 750°C, since the diffusion coefficient in the face-centered-
cubic (austenitic) lattice is about two orders lower than in the body centered (ferritic). Many
but not all austenitic steels develop intergranular cavities depending on the heat treatment and
the impurity content, see for example ref. [60] in 304 austenitic steels at 50% of the melting
temperature. It was also reported that grain boundary cavitation at carbides is one of the most
important factors influencing the degradation of austenitic stainless steels at high temperature
[96,97].
Figure 2.8. Maximum operating temperature of the currently used and the newly developed
power station steels for 100,000 h average rupture strength of 100 MPa [98].
II. State of the Art 22
2.5.3 Diffusion models of cavity growth
Due to the economical importance of creep at high temperatures much attention has
been given to the mechanisms of intergranular cavities growth. However, as several
mechanisms can be responsible for creep, as discussed in section 2.3, a practically large
number of cases results which cannot be analyzed deterministically. Therefore, certain
idealizations have to be made giving rise to a certain number of models of nucleation and
growth of intergranular cavities under creep conditions. Cavity growth mechanisms are
usually grouped into four categories: plasticity-controlled cavity growth [5,99], diffusion-
controlled cavity growth [3,4,10,33], constrained cavity growth [11,100,101,102] and
coupling of diffusion and plasticity (power-law creep) controlled cavity growth [6,11,99,103].
A survey of these models can be found in [11,12,13,104,105]. At high stresses and relatively
low temperatures, cavities grow by power-law creep, whereas at low stresses and relatively
high temperatures, growth is determined by diffusional flow.
The following sections focus on the diffusional cavity growth models during creep in
polycrystalline metals and alloys. Since, theoretically, the cavity size at nucleation range from
2-5 nm to about 100 nm [13,106] and plasticity-controlled growth models are applied at high
stresses, they will be not discussed here (this subject is described in, e.g. [12,13]). Cavity
growth by diffusion is discussed starting from the diffusion-controlled cavity growth model
proposed by Hull and Rimmer [3], followed by refinements of this model. This will be
followed by the constrained cavity growth model initially proposed by Dyson [100].
Cavity growth at grain boundaries at elevated temperatures has been suggested to
involve vacancy diffusion. According to this mechanism, cavity growth rate is determined by
the gradient of chemical potential of vacancies, f∇ , in the plane of the grain boundary.
Cavities grow due to migration of vacancies under the influence of this gradient. Cavity
growth rate is also influenced by the shape of voids and the diffusion process. During the
growth of a cavity, atoms are transported from the surface of the cavity to the adjoining grain
boundary where they are deposited. The growth rate is thus expected to be controlled by the
slower of the two mechanisms, namely surface diffusion or grain boundary diffusion.
Therefore, it is necessary to distinguish between diffusive growth mechanisms controlled by
grain boundary diffusion or controlled by surface diffusion.
II. State of the Art 23
Grain boundary diffusion controlled growth 2.5.3.1
In the case when grain boundary diffusion is slower than surface diffusion, the growth
process is governed by the first mechanism. The voids retain their spherical cap shape since
surface diffusion rapidly redistributes the matter within it (equilibrium growth). Hull and
Rimmer [3] first proposed a model based on diffusion-controlled cavity growth. The
equations that Hull and Rimmer and subsequently others [4,33] derived for diffusion-
controlled cavity growth are similar. The basic form of their result can be expressed as:
2.12) aTk
aD
dtda
⋅⋅⋅⋅
⎟⎠⎞
⎜⎝⎛ ⋅
−⋅Ω⋅⋅≅
s
mgb
2
2
λ
γσδ.
where δ is the grain boundary width, Ω is the atomic volume, a is the cavity radius, σ is
the remotely applied stress normal to the grain boundary, is the Boltzmann constant, k T is
the absolute temperature, is the grain boundary diffusivity, gbD sλ is the mean separation
between cavities, and mγ is the surface tension of the metal. These parameters are described
in Figure 2.9 which shows the basic geometry for diffusive cavity growth.
Figure 2.9. Cavity growth by diffusion along cavity surface and through the grain boundaries
due to stress gradient.
II. State of the Art 24
By integrating Eq. 2.12 between the critical radius (below which sintering occurs) and
the coalescence condition ( 2/sλ=a ), a linearly inverse stress dependence for the rupture
time σ/1r ∝t is obtained.
Later, improvements were made by [4,13,107,108,109] including the influence of
diffusion lengths (the entire grain boundary is a source of vacancies), stress redistribution (the
integration of the stress over the entire boundary should equal the applied stress), cavity
geometry (cavities are not perfectly spherical) and the “jacking” effect (atoms deposited on
the boundaries cause displacement of the grains). Riedel [13] proposed an equation for the
growth rate of unconstrained widely-spaced cavities as:
2.13) ( )
2s
0gb
)24.4/ln(22.1 aaTkD
dtda
⋅⋅⋅⋅
−⋅Ω⋅⋅=
λσσδ
.
where 0σ is the sintering stress. The same inverse stress dependence of the rupture time is
obtained by integrating Eq. 2.13. Despite these improvements, the basic description suggested
by Hull and Rimmer is largely representative of unconstrained cavity growth. An important
point is the predicted inverse stress and activation energy (of grain boundary diffusion)
dependence of cavity growth. When cavities are subjected to a triaxial stress field with
3,21 , σσσ being the principal stresses, the stress used in Eq. 2.13 should be 1σ (maximum
principal stress), the other components 32 ,σσ having less influence on void growth by this
mechanism [11].
The prediction of the stress dependence based on the proposed equations has been
frequently tested. In polycrystals, it has been found that the stress dependency is often
stronger than the predicted one, and is compatible with the Monkman-Grant relationship, i.e.
. Some examples are the studies of Pavinich and Raj [nt σε /1/1r ∝∝ & 110] on copper
polycrystals and of Raj [111] on copper bicrstals. In polycrystals, stronger stress dependence
and ruptures times much longer than the predicted ones were observed. In bicrstals, however,
the rupture time was inversely proportional to the stress, consistent with the diffusion
controlled cavity growth model. In α-brass, Svensson and Dunlop [112] found a linear
dependence of cavity growth with stress, however, the fracture time was consistent with the
Monkamn-Grant and continuous nucleation was observed. Cho et al. [113] and Needham and
Gladman [90] have measured the rupture times and cavity growth rates and found consistency
with the cavity diffusion growth model of Hull and Rimmer if allowance for continuous
II. State of the Art 25
nucleation was made. These ambiguous results have led to continuing investigations
regarding the possible reasons of such discrepancies.
Surface diffusion controlled growth 2.5.3.2
When surface diffusion is slow compared with diffusion in the grain boundary, the
cavity ceases to grow as a spherical cap. Matter at the tip of the cavity is removed at a faster
rate and it becomes flatter and more crack-like until the curvature difference between the
poles and the equator (tip of the cavity) is sufficient to drive a surface flux which matches the
boundary diffusion in the grain boundary. The growth of the cavity is then controlled by
surface diffusion. This is also known as non-equilibrium cavity growth. This problem was
first studied by Chuang and Rice [114] and Chuang et al. [10]. They determine the three-
power stress-relationship for the cavity growth at low stresses as:
2.14) 32
m
s
2σ
γδ⋅⋅⋅Ω⋅⋅
≅Tk
Ddtda
where is the surface diffusion coefficient, and the other terms have their usual meaning.
The result is an increasingly crack-like cavity whose growth rate depends on the third power
of the stress. At higher stresses, the growth rate varies as [
sD
2/3σ 13].
Nieh and Nix [115] measured the activation energy for cavity growth in copper
(assuming unconstrained growth) in which vapour bubbles were implanted on grain
boundaries before creep tests in order to bypass the nucleation stage. They also assumed that
continuous nucleation was not important during creep. Their measurements are inconclusive
as to whether it better matches versus . However, the stress and temperature
dependences of the measured lifetimes are in good agreement with an inverse cubic law
( ). Later, Stanzl et al. [
sD gbD
3r /1 σ∝t 116] confirmed the observations of Nieh and Nix on the
same material by performing tension and torsion creep tests. They also found that rupture
lifetime was determined by the maximum principal stress only, but not by the von Mises
equivalent stress.
A problem associated with the surface-diffusion controlled growth model (even when
the inverse cubic law fits the data for copper) is that it is not clear in the experiments why
II. State of the Art 26
surface diffusion should be much slower than grain boundary diffusion. For most of the
materials, is larger than [sD gbD 13].
Constrained diffusional cavity growth 2.5.3.3
For the case of grain boundary diffusion or surface diffusion controlled growth it was
assumed that the cavities are uniformly distributed on all boundary facets normal to the
external stress direction. However, over a substantial fraction of the creep lifetime, cavitation
is confined to relatively isolated grain boundary facets which are surrounded by undamaged
material, as illustrated in Figure 2.10. According to experiments the tendency of a grain
boundary to cavitation depends on its orientation to the stress axis and on the crystallographic
orientation of the adjacent grains (i.e. high or low angle grain boundary). If a cavity grows by
diffusion, the material surrounding the cavitating facet should deform also in order to
accommodate the excess of volume of the cavity. Thus, the cavity growth rate may be
controlled by the deformation rate of the surrounding matrix. This was called by Dyson
[100,117] “constrained cavity growth”. If the surrounding material were rigid, the cavity
growth would come to a standstill. On the other hand, if the surrounding material is relatively
soft, the accommodation process occurs readily and cavity growth rate approximate to that of
diffusion-controlled (unconstrained limit). The constraint can be interpreted as the
surrounding material exerts a back stress on the cavitating grain boundary facet. The resulting
stress on the grain boundary, bσ , adjusts itself such that the rate of volume increase by
diffusive cavitation is compatible with the deformation rate of the surrounding material. Here
two extreme cases can be distinguished: (i) the unconstrained limit, in which the cavitated
facet is subjected to the whole applied stress, i.e. , and (ii) the constrained limit, in
which the stress on the boundary is reduced to the sintering stress.
∞=σσ b
Figure 2.10. Uniformly distributed cavitation (a) and heterogeneously distributed cavitation
(b) at transverse grain boundary facets.
II. State of the Art 27
Rice [101] developed a quantitative model based on the idea of constrained cavity
growth, which was further improved by other authors (e.g. [13]). The model was idealized by
cavitating boundary facet described as a penny-shaped crack embedded in a creeping matrix
described by Norton’s power law (Figure 2.11). An axisymmetric loading was considered
with being the normal applied stress, and the transverse stress. ∞1σ
∞Tσ
Under the conditions specified above, the cavity growth rate can be described by the
relationship derived by Riedel [13]:
2.15)
⎟⎟⎠
⎞⎜⎜⎝
⎛
⋅⋅⋅
+⋅⋅Ω⋅⋅⋅
⋅
−−=
∞
∞
dq
DTkqah
dtda
2se
e
gb
2
0
'2
)()(
)1(
λεσ
δωψ
σωσ
&
where d is the facet size, is the cavity area fraction on grain boundary facets,
is an abbreviation,
2s )/2( λω a=
2/12 )/31(' nq +=π ∞∞∞ −= T1e σσσ is the von Mises equivalent stress,
is the equivalent strain-rate, and the functions ∞eε& )(ψh and )(ωq account for the cavity
shape and the cavity area fraction, respectively. The other terms have their usual meaning.
Figure 2.11. Constrained cavity growth model on a grain boundary facet.
Needham and co-workers [73,118] have carried out an extensive investigation on low
alloyed ferritic steels (1Cr-½Mo and 2¼Cr-1Mo). In their study, the growth rate of the ith
largest cavity (with i≅1000) has been quantified for both steels. The experimental cavity
II. State of the Art 28
growth rates were then compared with the theoretical predictions. The comparison is shown in
Figure 2.12. For the predictions, the sintered stress, 0σ , was neglected and the expression
2/)()( ωψ qh ⋅ was approximated by 1 as a representative average value. The measured
strain-rates and the values for grain size (d=18µm) and for the cavity spacing ( m2.3s μλ = for
the 1Cr-½Mo and 4.5µm for the 2¼Cr-1Mo) were also introduced in Eq. 2.15. The cavity
diameter chosen was m12 μ=a and the material parameters were taken from tabulations
[119]. As it is shown in Figure 2.12, the unconstrained diffusion model overestimates the
growth rate considerably and leads to incorrect stress dependence. However, agreement of the
constrained cavity growth model with the data is very good for the 2¼Cr-1Mo and fair for the
1Cr-½Mo. It should be pointed out, however, that a and sλ vary during the test substantially
whereas the calculations were done using average values. The agreement of the constrained
growth model with Needham’s data suggests that diffusive cavity growth is indeed
constrained under the condition of these experiments.
Cane [62] arrived to a similar conclusion by analysing cavitation in 2¼Cr-1Mo steel at
565°C. The material was heat treated to produce a coarse-grained bainitic structure. The
thermal treatment produced a prior austenite grain size of d=150 µm and fine sulphides on
grain boundaries which act as nucleation sites for cavities. Cavity spacing and cavity radii
were m5.4s μλ = and m8.1 μ=a , respectively. With these parameters, the constrained
growth model agrees with the measurements within a factor of 2.3, while the unconstrained
model strongly over predicts the cavity growth rate at low stresses.
In a later work, Cane [120] investigated notched specimens of the same steel. For the
notch geometry considered , and according with Eq. 2/ e1 =∞∞ σσ 2.15 the growth rate is
expected to be twice larger than in the uniaxial tension case for the same equivalent von
Mises stress. Although few measurements were presented, a reasonable close agreement was
obtained, being the growth rate under multiaxial stress state a factor of 1.5 larger than the data
for uniaxial tension.
II. State of the Art 29
Figure 2.12. Cavity growth rate for two ferritic steels (1Cr-½Mo and 2¼Cr-1Mo). Predictions
from unconstrained and constrained models are depicted (adapted from [13]).
The time to cavity coalescence on the cavitating boundary facet can be obtained by
integrating Eq. 2.15 between (cavity radius at time t=0, considering nucleation to occur at
the beginning of the test) and
0a
2/sλ=a (cavity coalescence) as:
2.16) d
nhD
Tkht⋅⋅⋅+⋅
+⋅⋅⋅Ω⋅⋅⋅
= ∞∞∞∞ )/()/31()(4.0)(006.0e1e
s2/1
1gb
3s
c σσελψ
σδλψ
&
where n is the steady state stress exponent.
The time to coalescence, , is composed of two terms, the first representing the time to
coalescence in the absence of constraint (i.e. only diffusion controls growth) while the second
term represents the constraint. At small strain-rates, the second term leads to a very long time
to coalescence. In the uniaxial case, the constrained term predicts a much stronger stress
dependence, . It should be emphasised that failure is not expected by coalescence of
ct
nt σ/1c ∝
II. State of the Art 30
cavities on isolated facets, and some additional time may be required to join facet-size
microcracks. Further nucleation may be necessary to produce the rupture (e.g. local nucleation
ahead the crack or nucleation on non-cavitated facets). Therefore, continuous nucleation
should be considered in addition to the constrained case. If this is done, the stress dependence
of the rupture time takes the form [5/)23(r /1 +∝ nt σ 13]. The additional effect of continuous
nucleation is illustrated in Figure 2.13 for the rupture life of 2¼Cr-1Mo steel measured by
Cane [120]. For unconstrained growth the difference between predicted values and measured
values of times to rupture is significant, while it is nearly negligible for constrained growth.
No adjustable parameters were used in this study. Figure 2.13 shows an excellent agreement
between and for constrained cavity growth indicating that the time to coalescence, , is
most of the specimen lifetime, in this case.
ct rt ct
rt
Figure 2.13. Time to rupture as a function of stress for constrained and unconstrained cavity
growth with instantaneous and continuous nucleation for 2¼Cr-1Mo steel measured by Cane
(adapted from [13]).
2.5.4 The cavity size distribution function
As it was previously mentioned, cavities nucleate continuously during creep of metals
and engineering alloys and this fact needs to be considered in the lifetime calculation.
II. State of the Art 31
Experimental results on cavity nucleation are scarce due to difficulties related to the small
cavity size. The cavity size distribution function provides, however, a link between
experimental data and theories on cavity nucleation and growth. The cavity distribution
function is denoted by , where ),( taN daN ⋅ is the number of cavities per unit grain
boundary area having a radii between a da+ . In this formulation, only those boundary facets
which effectively contribute to the rupture process are considered in the cavity distribution
function. In the case of failure by diffusive cavity growth, the boundary facets that should be
considered are those which have orientations between 60° and 90° with respect to the
principal stress axis, since these boundaries cavitate preferentially. Nevertheless, this freedom
in definition does not affect the results.
The cavities are considered to fit in a certain size class which is determined by its
radius, , but they can pass from one size class to the next one,a daa + , by growing. The
coalescence event during the late stages of the ruptured process is ignored. Riedel [13]
proposed a cavity distribution function by imposing a continuity condition in the size space,
and assuming a power-law functions for the cavity growth rate, , and the
nucleation rate, , as:
dtda /=a&
*J
2.17) . αβ −− ⋅⋅= taAa 1&
2.18) γtAJ ⋅= 2* .
where , , 1A 2A α , β and γ might depend on stress, on temperature, on material parameters
as well as on the actual microstructure, but not on time nor on the cavity size. The final result
for the distribution function proposed by Riedel [13] is:
2.19) )1/()(
11
1
1
2
111),(
αγα
α
βγαβ
βα
−+
−
++
⎟⎟⎠
⎞⎜⎜⎝
⎛+−
−=tA
ataAAtaN
The solution for different values of α , β and γ is shown in Figure 2.14 for
. If 121 == AA 1<α , the distribution function must be cut off at a maximum . If a 1>α , the
distribution extends to infinite , because the first cavities nucleated grow infinitely fast for a
value of
a
1>α . For the case of 1=α , Eq. 2.19 can be rewritten as:
II. State of the Art 32
2.20) ⎟⎟⎠
⎞⎜⎜⎝
⎛++−
=+
+
1
11
1
2
)1()1(exp),(
Aata
AAtaN
βγ β
γβ
It is important to note in this case, that the maximum of the distribution function, as
well as the average cavity size, remain at a fixed value of . a
Figure 2.14. Evolution of the cavity size distribution function (adapted from [13]).
2.5.5 Prediction of creep rupture time (phenomenological approach)
Creep rupture by cavity growth and coalescence is an important failure mechanism for
high temperature components. Cavity growth is a very complex phenomenon involving
kinetic and mechanical processes which are manifold (e.g. lattice, grain boundary and surface
diffusion, dislocation creep, grain boundary sliding). Furthermore, the cavity size distribution
is a function of the time and varies with the local microstructure and with the heterogeneous
and continuous nucleation of new cavities during creep. Also, the models which describe the
intergranular cavity growth under creep conditions are based on idealization of the real
structure. The cavity growth, the assessment of the life time and the failure modes under creep
condition have been extensively studied both theoretically and experimentally, e.g.
[5,10,11,100,109,117,121-124] mostly under uniaxial stress conditions. However, in most of
the cases the engineering components are subjected to inhomogeneous loading conditions of
temperature and stress. To be able to predict the failure mode and life time of such industrial
components the assessment of damage on specimens deformed under conditions similar to
those in service is necessary [73]. This is done for example by performing creep tests on
hollow cylinder samples, where the stress multiaxiality is imposed by an internal pressure, a
II. State of the Art 33
longitudinal force and the notched sample geometry. Creep lifetime and failure mode under
multiaxial stress conditions can differ significantly from the rules established for uniaxial
loading. For example Cocks and Ashby [5] have predicted that cavity growth is highly
accelerated by high positive stress triaxiality.
Data obtained from uniaxial stress experiments have led to a good understanding of the
physical processes involved, although they do not always provide sufficient information to
predict failure under multiaxial stress conditions. Based on principles of damage mechanics
[125,126] Hayhurst et al. [127-130] have studied the effects of stress multiaxiality on creep
rupture. For a smooth round bar under uniaxial loading, the rupture lifetime, , at a given
temperature is usually expressed as
ft
2.21) . χσ −⋅= Mt f
where σ is the stress an M and χ are temperature dependent parameters. Hayhurst [130]
has shown that Eq. 2.21 does not correctly predict the creep rupture properties of notched
rounded bars and generalized it by including the stress invariants, and [1J 2J 130]:
2.22) . χγβσα −⋅+⋅+⋅⋅= )( 211f JJMt
where 1σ is the maximum principal stress with 321 σσσ >> , H1 3 σ⋅=J ( Hσ is the
hydrostatic stress) and is equal to 2J eσ , the equivalent von Mises stress. The coefficients α ,
β and γ are weighing factors describing the relative contribution of the different stress
parameters to rupture life ( 1=++ γβα ). The physical basis for including various multiaxial
stress terms in Eq. 2.22 is based on the following facts. First, intergranular fracture usually
occurs by diffusive growth of intergranular cavities (driven by the tensile stresses) at
boundaries perpendicular to the maximum principal stress, 1σ . Second, cavity growth at high
temperature can be also driven by the tensile hydrostatic stress component, thus H1 3 σ⋅=J .
Third, high stress concentration produced by inhomogeneous plastic deformation such us
grain boundary sliding or slip band formation are required for the nucleation of cavities.
These deformation processes are driven by shear stresses, therefore, some form of the shear
stress should enter in the equation. Therefore, the effective von Mises stress was also included
in Eq. 2.22. Reviews of multiaxial creep rupture data for several metals suggest that the
II. State of the Art 34
maximum principal stress 1σ , and the von Mises equivalent stress eσ are more important than
the hydrostatic stress Hσ in determining the creep lifetime [130]. Then, Eq. 2.22 can be
rewritten as
2.23) . χσασα −⋅−+⋅⋅ ))1(( e1=ft M
where α is a single parameter that describes relative importance of 1σ in the creep rupture
time. The value of α varies for different materials and fracture modes [131]. Assuming that
cavitation is concentrated on grain boundaries (GB) perpendicular to the maximum principal
stress [60,62,116,131-134], Nix et al. [135] have developed a model based on the “principal
facet stress”, which predicts good results in cases when the grain boundary sliding (GBS)
mechanism is active. They argue that GBS relieves the shear stress along the grain boundaries
and the stress causing cavitation is not the nominal applied stress but the concentrated normal
stress acting on the transverse or nearly transverse grain boundary facets. Based on three-
dimensional numerical analysis performed by Anderson and Rice [136], who have used the
Wigner-Seitz cell of an fcc lattice as the grain for modeling, Nix et al. define the average
principal facet stress, , as Fσ
2. 24) )(62.024 321.2fσ = σσσ +−⋅ .
with 321 σσσ >> being the principal stresses that exist in macroscopic sense. It should be
emphasized that in this approach the principal stress is expected to be valid after grain
boundary sliding has caused redistribution of the stresses but before cavitation started.
The hydrostatic stress (or mean normal stress) is known to influence cavity growth [5]
under power law creep. However, Sakane et al. [137] have confirmed its contribution to void
nucleation, too, during experiments with equi-triaxial stress state, i.e. zero von Mises
component. Watanabe et al. [66] and Li et al. [75] also reported that in 9Cr-1Mo-V-Nb steel
welded joints the distribution of the stress triaxiality factor coincided better with the creep
damage distribution rather than the equivalent creep strain. They argued that the local
concentration of both triaxiality factor and equivalent creep strain accelerate void formation
and growth.
More recently, various experiments have been done to test materials under multiaxial
loading conditions. Niu et al. [69,70] analysed creep rupture using tubular and notched
II. State of the Art 35
specimens of austenitic steel SUS310S with high ductility, in tension, torsion and combined
tension–torsion stress states at 700°C. They have found that the maximum principal stress
determines the multiaxial creep rupture life of the steel. They further suggest that the fracture
mode is related to the magnitude of Hσ with respect to 1σ and eσ , because is supposed that
Hσ promotes the growth of voids. Sakane et al. [137,138] analysed two- and three-cruciform
specimens subjected to axial forces. They conclude that the von Mises equivalent stress and
the equivalent stress based on crack opening displacement were a suitable parameter to assess
the biaxial creep damage in 304 type stainless steel in which the fracture mode was mostly
transgranular. Hsiao et al. [67] have compared three different stress parameters, the von Mises
effective stress, the maximum principal stress and the principal facet stress to correlate the
local multaxial stress with the local creep damage distribution and failure lifetime on
weldments of 316 steel. They showed that the principal facet stress parameter gave the best
prediction of the creep damage distribution in the weldments.
Based on these experiments, it is suggested that the von Mises stress, eσ , predicts better
the rupture time and creep damage when the fracture mode is transgranular, which is usually
found in short time creep. It has been also reported that the von Mises equivalent stress will
determine the rupture life for metals which show virtually no cracking [139]. The maximum
principal stress, 1σ , on the other hand, will determine the rupture life time of metals pre-
cavitated or continuously cracking, in which many voids have nucleated in the early stages of
creep [70,116]. Furthermore, for metals and alloys which exhibit grain boundary sliding and
fail by cavitation at grain boundaries perpendicular to the maximum principal stress, the
principal facet stress, Fσ , will correctly predict creep rupture life under various stress states.
However, analysis of creep damage under a multiaxial stress state has led to ambiguous
results of the parameters defining creep lifetime and damage distribution. The choice of the
best damage indicator tended to be material, stress and temperature dependent.
The stress parameters usually used for the assessment of rupture life are, however,
averages of the local values in the sample. Under multiaxial loading the damage develops
inhomogeneously indicating that local parameters should be considered for evaluation of the
damage distribution. As mentioned in section 2.5.1, tomography has evolved as a powerful
tool for such evaluations. It is the aim of the second part of this thesis to assess the damage in
a notched hollow cylinder of E911 steel, which was subjected to about 26,000 h of creep of
similar loads as in service. The 3D shape and size distribution of single cavities will be
II. State of the Art 36
described. A comparison of local damage distributions with the distribution of selected stress
parameters, knowing to influence damage, will be performed. The tomographic results will
also be compared with those of conventional metallography.
III. Experimental Details 37
CHAPTER III
EXPERIMENTAL DETAILS
3.1 Uniaxial in situ experiments
3.1.1 Materials and specimen geometry adapted to tomographic measurement
For in situ tomographic investigations a common easily machinable brass alloy and a
commercial stainless steel (AISI 440B) were chosen as model materials. Their chemical
compositions are given in Table 3.1.
The brass alloy was fabricated by indirect extrusion (1023 K -750°C- billet temperature,
19:1 extrusion ratio, 285 mm/s extrusion rate) at the Extrusion Research and Development
Center of Technische Universität Berlin (TU Berlin) [140]. In order to promote the damage
during creep the sample was pre-strained in tension (4.5% in total) with the specimen axis
parallel to the extrusion direction [141].
Brass (wt.%)
Cu Zn Pb
58 40 2
Steel AISI 440B (wt.%)
C Si Mn P S Cr Mo V
0.85 - 0.95 ≤ 1.00 ≤ 1.00 ≤ 0.04 ≤ 0.015 17.0 - 19.0 0.90 - 1.30 0.07 - 0.12
Steel grade E911 (wt.%)
C Si Mn Cr Mo Ni W V Nb N B
0.11 0.2 1.00 9.1 0.96 0.25 0.98 0.21 0.082 0.091 0.0071
Table 3.1. Chemical composition of the investigated brass alloy, stainless steel AISI 440B
used for activation energy determination and that of E911 steel used for damage evaluation
after 26,000 h of creep.
Specimens for in situ tests were specially developed to meet the requirements of the
tomographic set-up at ID15A of the European Synchrotron Radiation Facility (ESRF). The
III. Experimental Details 38
geometry of the investigated specimens is presented in Figure 3.1. The sample has a diameter
and gauge length (L) of about 1 mm. Figure 3.1 also shows the position of the illuminated
volume in the steel sample B (see Table 3.2) and its reconstruction by tomography. The
specimens are designed with threads at both ends which are screwed at one end into the load
cell and at the other end into the reaction yoke of the creep machine. The specimens have two
small holes placed 4 mm apart from the center of the sample where thermocouples (T1 and
T2) for temperature measurement are placed.
Figure 3.1. Schematic drawing of the specimen used for in situ investigations. The
illuminated volume in the sample and its reconstruction for the steel sample B by tomography
are also shown.
3.1.2 Creep device for in situ experiments and testing conditions
A small creep machine for in situ experiments was specially developed to fit on the
goniometer table. The machine had a reduced weight to allow for the proper rotation during
tomographic scans. The design of the machine allows a free path for the X-ray beam to reach
the sample during 360° turn around a vertical axis. The loading system is supported by a
ceramic tube made of Macor (Figure 3.2) [142], which has good creep resistance at high
III. Experimental Details 39
temperature and transparent to high energy X-rays. The lower end of the specimen is screwed
in the reaction yoke of the machine and the upper part is connected to the loading system. The
load is applied to the specimen via two springs (material spring steel 1.1200 Class C) with
proper constant to achieve the required stress condition. When the springs are compressed the
load is transferred to the sample through the load cell, which measures the actual load applied
to the specimen. The high temperature necessary for creep is obtained by resistive heating
coils placed around the ends of the sample. At the position of each heating coil thermocouples
measure the temperature and a proportional-integral-derivative (PID) controller is used to
keep the temperature constant within ±1 K. Asymmetric heating of the sample becomes
possible by setting different temperatures for the two coils, enabling tests with a controlled
temperature gradient along the gauge length. Asymmetric heating produces a temperature
distribution along the stress axis, which is linear in the region of the sample with constant
cross-sectional area. The elongation of the sample is measured with an inductive displacement
transducer attached to the head of the load cell. Recording of data as well as the control of the
machine are made through computer.
Figure 3.2. Creep machine used for in situ measurements.
III. Experimental Details 40
Experiments with asymmetric heating were performed with brass and AISI 440B
stainless steel. For the brass sample (named as sample A in Table 3.2) the load was supplied
by a spring with small spring constant [87], the stress at the beginning of the experiment being
about 25 MPa and the temperature at the bottom of the sample 673 K. Two experiments were
performed with the AISI 440B stainless steel: the first test at 180 MPa and 985 K at the
bottom of the sample, and a second test at 210 MPa at 1013 K also at the bottom of the
sample. Sample designations (B and C) and corresponding deformation conditions are given
in Table 3.2. In all cases the upper coils were switched off. The asymmetric heating caused a
temperature difference between the two thermocouples (separated by a distance of 8 mm
length) of about 390±2 K (sample A), 624±2 K (sample B), and 645±2 K (sample C).
Sample σ
[MPa]
bT
[K]
TΔ
[K]
T∇
[K/mm]
ssε&
[1/s]
rt
[h]
A (brass) 25 673 ± 1 117 ± 2 37 2.1·10-6 7.3
B (steel) 180 985 ± 1 351 ± 2 99 3.1·10-6 24.7
C (steel) 210 1013 ± 1 372 ± 2 105 1.2·10-5 7.9
Table 3.2. Measurement conditions for in situ experiments. Applied stress σ , temperature at
the bottom of the sample bT , measured temperature difference between calibration points
TΔ , temperature gradient in region with constant cross section T∇ , minimum steady-state
strain-rate ssε& and rupture tim e rt .
3.1.3 Creep tests with nearly constant load or constant stress
A proper selection of the spring constant enables creep tests at nearly constant load or
constant stress. Evidently the elongation of the sample produces a load decrease by the
expansion of the springs, however, this change can be kept below a certain tolerance by
selecting springs with proper constants. The admissible relative load change or tolerance is
defined as follows:
3.1) tolerancex
xxF
FFF ≤−
=−
=Δ0
r0
0
r0 ,
III. Experimental Details 41
where F0 and Fr are the loads at the beginning and at the end of the creep test and and
the corresponding compression lengths of the springs. The maximum spring constant
fulfilling this criterion is obtained as:
0x rx
3.2) )1( r
0s
00max −⋅⋅
⋅⋅= εelN
toleranceAk σ ,
where is the initial length of the sample, the initial cross section, σ0 the initial stress, Ns
the number of springs and εr the strain to rupture. In our experiments usually a load tolerance
of 5% was selected.
0l 0A
A nearly-constant stress test becomes also possible if the relaxation of the load due to
the elongation of the springs equals the decrease in load due to contraction of the cross
sectional area (assuming the stress constant) of the specimen. The equal stress condition will
be valid at two strains, which can be adequately selected for example at the beginning and at
the end of secondary creep regime. The following condition can be written:
3.3) )()( 21 εσεσ = ,
where 1ε and 2ε are the selected strains. Assuming that the volume of the sample remains
constant during deformation, the initial compression of the spring can be worked out: 0x
3.4) 12
1122 )1()1(00 εε
εεεε
eeeeeelx
−⋅−−⋅−
= ,
with being the initial gauge length. 0l
The spring constant is determined by the selected stress 0σ and is given by:
3.5) ))1(( ii
00s
00
−⋅−⋅⋅⋅
= εεσ
elxeNAk ,
where iε denotes either 1ε or 2ε . Selecting for example the strains 08.01 =ε and 3.02 =ε
where the equal stress condition is imposed it can be shown that the relative stress variation in
III. Experimental Details 42
the [0, 0.35] strain interval is less than 2%. Figure 3.3 shows the relative stress deviation from
the imposed value of 25 MPa vs. true strain. The stress changes less than 2% during the creep
test if the values of ε1=0.08 and ε2=0.3 are chosen to fulfill the Eq. 3.3.
Figure 3.3. Relative stress deviation during creep test for selected parameters ε1=0.08 and
ε2=0.3, , mm10 =l MPa250 =σ and . 20 mm78.0=A
3.2 Multi-axial creep experiments on E911 steel
3.2.1 Materials and specimen geometry adapted to tomographic measurement
conditions
Creep experiments under multi-axial loading conditions were performed at the MPA
Stuttgart using the sample geometry shown in Figure 3.4. Notched hollow cylinder samples
made of E911 steel (chemical composition is given in Table 3.1) were subjected to a uniaxial
load of 8543 N applied along the cylinder axis and an internal pressure of 17.5 MPa. The
temperature was set to 848 K (575°C), which is the usually operating temperature of this
material in a power plant. After a creep time of 26,000 h the test was stopped and a small
cylindrical sample of about 0.6 mm in diameter (area of about 0.28 mm²) and approximately
2 mm in length (this being the minimum width at the notch) was extracted for tomographic
measurements from the middle plane of the notch. Further metallographic investigations were
III. Experimental Details 43
performed on the same sample at three cross-sections: near the outer, the inner surface and in
the middle of the small cylindrical specimen. A careful preparation with grinding and
polishing was performed. Polishing was done in different steps with graded grain sizes up to 1
µm and a regular change of polishing direction. In addition no etching was applied in order to
prevent the extraction of particles by the reagent. The identification of the cavities was made
by an experienced metallographer using a light-optical microscopy with up to 1000 times
magnification. A representative area of approximately 0.14 mm² was evaluated.
Figure 3.4. Schematic drawing of the notched hollow cylinder specimen creep deformed
under multiaxial loading conditions. For tomographic investigations a small sample from the
middle part of the notch was extracted.
3.3 Synchrotron X-ray microtomography
The experiments in the first part of the present thesis are focusing on in situ monitoring
by microtomography the geometry of creeping specimens subjected to uniaxial load. The
basis of this technique is briefly explained in the next subsection. In situ fast tomography
measurements during creep were performed at beamline ID15A of the ESRF, which was
equipped with a CCD camera having a restricted field of view (FOV) of about 1.2 mm x
1.1 mm. The effective pixel size of the detector was 1.6 µm and had a resolution of about
2.1 µm (the full width at half maximum of the point spread function). The tomograms were
recorded using a high-energy beam (~80 keV) with a large bandwidth (~50 %). The
reconstructed volume is constituted of voxels, each voxel representing a volume of 1.63 µm3.
The acquisition time of a complete tomographic scan was about 3 min. Due to the highly
III. Experimental Details 44
brilliance beam available at the ID15A, the recorded radiographs had good statistics, which
allowed for good quality reconstructions.
The second part of this thesis focuses on the characterization of damage developed
during multiaxial creep test using X-ray microtomography. A sample extracted from a crept
hollow cylinder was investigated by microtomography at the ID19 beamline of the ESRF
using a monochromatic beam of 51 keV. The beamline was equipped with a CCD camera
having a restricted FOV of about 0.7 mm x 0.7 mm, the effective pixel size of the detector
being 0.33 µm. In this case each voxel represents a volume of 0.33³ µm³. The acquisition time
of a complete tomographic scan was about 2 h. Due to the restricted FOV of the CCD camera
three volumes were measured in order to cover the whole length of the sample of about 2 mm.
3.3.1 X-ray tomography: Principles
X-ray tomography is a non-destructive imagining method in which a cross-sectional
view of an object can be obtained from transmission data collected at many different angles.
The basis of X-ray tomography is X-ray radiography, which represents a “projection” of the
absorption coefficient of the investigated material. The mathematical formulation for
reconstructing an object from multiple projections is based in the work of Radon [143], who
demonstrated the possibility of replicate an object from a set of its projections. X-ray
radiography physics is based on the Lambert-Beer law which relates the ratio of transmitted
(I) to incident (I0) intensity (number of photons) to the integral of the linear absorption
coefficient of the material μ along the path L :
3.6) dxEyxII
L),,(ln
0∫=⎟⎟
⎠
⎞⎜⎜⎝
⎛− μ
μ depends on the material and the X-ray energy, therefore for a polychromatic beam, Eq. 3.6
has to be integrated over the whole energy spectrum.
Because the projections contain superimposed information of a volume in a 2D plane
the estimation of the absorption coefficient distribution of the scanned object needs many
projections to be obtained. For every angular position, θ , a 2D projection image is recorded.
The cross-section of the object (slice) to be reconstructed is called and a parallel ),( yxf
III. Experimental Details 45
projection of this object at the angle θ is denoted by ),( θtp , where t is the distance from the
projection ray to the center of rotation (Figure 3.5).
Once the projections are recorded, the next step is to obtain the tomographic
reconstruction itself. This is the inverse of the Radon transform. In the case of parallel beam
geometry, a condition that can be met at synchrotron sources, the slices of the sample
corresponding to different heights in the sample can be treated independently and it is also
sufficient to record the projections for half turn due to mirror symmetry, i.e.
),(),( θπθ tptp −=+ .
The theory governing the tomographic reconstruction is generally known as the Fourier
slice theorem [144,145]. It states that the Fourier transform of a parallel projection ),( θtp of
an object obtained at an angle ),( yxf θ is identical to a section in the two-dimensional
Fourier transform of taken at the same angle. ),( yxf
3.7) θθω )(),( FP = ,vu
where ),( θωP denotes the Fourier transform of ( ),θtp . This is illustrated in Figure 3.5.
Figure 3.5. Principle of tomography and illustration of the Fourier slice theorem. The object
is represented in the rotated coordinate system by . ),( yxf ),( ts'f
III. Experimental Details 46
Although the Fourier slice theorem provides a straightforward solution to the
reconstruction problem, it has some inconvenient in actual implementation such us
interpolation in the frequency domain or implementation of the targeted reconstruction [146].
Therefore, alternative implementations of the Fourier slice theorem were explored being the
most popular one the so-called Filtered Back Projection (FBP) algorithm [144,145]. The idea
of the backprojection relies in assigning to each point of the object the average value of all the
projections that pass through that point. The backprojected image is, however, a blurred
version of the original object. An exact mathematical correction of the backprojection
smoothing effect can be performed by an appropriate pre-filtering of the projections leading to
the Filtered Back Projection algorithm. Based on the Fourier slice theorem and symmetry
property given by the parallel sampling geometry it can be written:
3.8) θωωθωπ θθπω ddePyxf yxj∫ ∫
∞
∞−
+=0
)sincos(2),(),(
The inside integral is the inverse Fourier transform of the quantity ωθω ),(P which
represents a Fourier transform of a projection filtered by a function whose frequency domain
is ω (ramp filter). The ramp filter emphasizes high-frequency contents and consequently the
high-frequency noise. Different types of windows can be applied to shape the filter’s
frequency response and hence modify the noise characteristic to the reconstructed images.
Some often used window functions are Hanning or sinc functions [146]. All tomographic
volumes evaluated in this thesis were reconstructed using the FBP algorithm with a
Butterworth filter [144].
3.4 Image processing
3.4.1 Reconstruction and pre-processing
The reconstructions of tomographic measurements were performed with in-house
developed software and their processing was performed using algorithms written in
Interactive Data Language (IDL) and Matlab. For damage evaluation purposes histogram
equalization was applied to each slice in the volume. When necessary, the remaining ring
artifacts in the reconstruction (visible at the center of Figure 3.6a) were removed by an
III. Experimental Details 47
algorithm based on local threshold and finally a band pass filter was applied to enhance the
contrast of the cavities and make them easier to segment. The band pass filter procedure was
applied, however, only for counting the cavities. A mask situated 50 pixels inside the border
of the sample was then used to avoid picking up artifacts coming from the band pass filtering
(light border at sample surface visible in Figure 3.6c). The sequence of the procedure applied
for one slice is exemplified for the E911 steel sample in Figure 3.6.
Figure 3.6. Sequence of image analysis applied for pore number extraction. a) Original slice
as obtained from reconstruction, b) slice after histogram shift and ring artifacts removal, c)
band pass filtering, d) mask and threshold applied for pore extraction.
III. Experimental Details 48
3.4.2 Identification of cavities
The extraction of the cavities was performed by identifying voxels as belonging either
to a cavity or to the matrix material based on their gray level. For this, an algorithm based on
local grey level threshold was written in IDL. A cavity is considered as a configuration of
voxels, which share a common face, edge, or corner. Cavities with a size of 2 x 2 x 2 or more
connected voxels were considered. Smaller cavities can result from noise and were neglected.
In order to characterize their shape and orientation, creep cavities were approximated by
ellipsoids with the same volume and moment of inertia [147,148]. A diagonalization of the
momentum of inertia tensor, I, can be performed in order to obtain the eigenvalues, which
allow calculation of the length of the semi-axes, , and the eigenvectors, representing the
spatial orientation of the semi-axes. The semi-axes length can be calculated as [
ja
149]:
3.9) 1,2,3j,2
)2)((5 jj =
⋅
⋅−⋅=
VIITr
a
where V is the cavity volume and is the trace of the momentum of inertia tensor. It is
also assumed that the mass density is equal to unity. A description of the cavities fitted by
their equivalent ellipsoid can be given by the definition of 3D shape descriptors namely
elongation, , and flatness, [
)(ITr
e f 150]. The elongation parameter used in this work, e , has been
slightly modified in order to obtain the ratio between the length of the maximum semi-axis,
, and the average of the two minor semi-axis, and , respectively. Elongation and
flatness are then defined as:
1a 2a 3a
3.10) 32
12aa
ae+⋅
=
3.11) 3
2
aaf =
With these two parameters, the shape of nearly regular cavities can be described and
quantitatively separated in different cavity shapes such us spheres, ellipsoids and rods [83].
However, these shape descriptors cannot provide information about the complexity of the
III. Experimental Details 49
cavity shape, therefore and additional parameter (complexity factor, CF ) was proposed [83].
The parameter CF indicates the deviation of the cavity shape from its equivalent ellipsoid. It
is related to the exclusion volumes resulting from the intersection of the real cavity with its
equivalent ellipsoid. Figure 3.7 illustrates this definition in a 2D case. The complexity factor
is defined as:
3.12) 21TT
31 , VVVV
VVCF +=
+=
A large CF indicates that the void’s shape is different from that of an ellipsoid, while a
small value means a good similarity. CF varies between zero and about two and offers a
good possibility to distinguish between regular cavities and those affected by coalescence.
Figure 3.7. Complexity factor, CF , definition based on a 2D projection, a) 2D projection of a
real cavity, b) equivalent ellipsoid, d) both.
The orientation of the cavities was characterized by the angle between the major axis of
the equivalent ellipsoid and an arbitrary axis which can be defined conveniently in each
experiment. Usually for analysis of cavities under creep condition the loading axis is chosen
as the reference axis [83]. However for the E911 sample, the reference axis was considered
along the notch radius.
III. Experimental Details 50
3.4.3 Image correlation for evaluation of tomographic data
To monitor changes in the sample’s shape, material slices lying perpendicular to the
external stress axis were chosen, which due to the axial symmetry of the experiment were
supposed to move during creep along this axis. Slice displacement was evaluated by cross
correlating a selected slice from the initial volume with slices from consecutive
reconstructions.
The classical bi-dimensional cross-correlation method quantifies the degree of fit of two
matrices, and , to a linear model using the correlation coefficient. For image-
processing applications, variation in the brightness of the image can lead to erroneous values
of the correlation coefficient. In order to decrease this effect, the images are normalized by
subtracting the mean and dividing by the standard deviation. The cross-correlation coefficient
is defined as:
nm × mnX mnY
3. 13) [ ][ ]
∑∑∑∑
∑∑−
=
−
=
−
=
−
=
−
=
−
=
−−
−−
−−−=
1
0
1
0
21
0
1
0
2
1
0
1
0
)(1
1)(1
11
1
N
m
N
nmn
N
m
N
nmn
N
mmn
N
nmn
YYN
XXN
YYXXNr
where is the number of pixels in one image, and are the pixel intensities of the
images and
N mnX mnY
X and Y their average intensities, respectively.
The correlation coefficient, r , is a scalar quantity which varies in the interval [-1,1]. A
value of 1+=r indicates a perfect fit to a positive linear model whereas 1−=r is a perfect fit
to a negative linear model. Although conventional correlation can provide information about
similarities between two images, it is inefficient in the presence of horizontal shifts (image not
in phase with each other), which is the common situation in deformed samples due to material
flux towards the center of the sample. Moreover, the standard cross-correlation technique is
sensitive for luminance variations, which requires pre-processing of the images.
An advantage for in plane shift detection with respect to the classical cross-correlation
is provided by the Phase Correlation method, based on the translation property of the Fourier
Transform and expressed by the Fourier-shift theorem [151]. The theorem states that a
translation of a given function in the space domain corresponds to a phase shift in the
frequency domain, which is represented by Eq. 3.14 and 3.15.
III. Experimental Details 51
3.14) ),(),( 0012 yyxxfyxf −−=
3.15) ),(),( 1)(2
200 vuFevuF yvxuj ⋅= ⋅+⋅⋅⋅⋅− π
where and represent two grey level images, respectively, is the
shift of the function with respect to
( yxf ,1 ) )
)
( yxf ,2
( yxf ,2
),( 00 yx
( )yxf ,1 in the space domain and
denotes the Fourier transform of , = 1, 2.
),( vuFi
),( yxfi i
Then, the cross-power spectrum is defined as:
3.16) )(2*
12
*12 00
),(),(),(),( yvxuje
vuFvuFvuFvuFG ⋅+⋅⋅⋅⋅−=
××
= π
where * indicates the complex conjugate. The inverse Fourier transform (IFT) applied to
Eq. 3.16 gives the Dirac delta function ),( 00 yyxx −−δ centered at expressed in
Eq.
),( 00 yx
3.17.
3.17) ( ) ),(),(),(),(),(
00)(21
*12
*121 00 yyxxeF
vuFvuFvuFvuFF yvxuj −−==⎟⎟
⎠
⎞
⎜⎜
⎝
⎛
×× ⋅+⋅⋅⋅⋅−−− δπ
Then, from the inverse Fourier transform ( 1−F ) of the cross-power spectrum a peak
corresponding to the Dirac delta function, is obtained at the position of the shift between the
images. Other advantages of phase correlation are the possibility to detect “sub-pixel”
changes [152], and the small sensitivity to luminance variations. The most sensitive
correlation indicator was found to be, however, the product between the amplitude of the
inverse cross-power spectrum and the classical cross-correlation coefficient [153]. The
presented algorithm was applied to monitor sample shape changes during creep.
IV. Results 52
CHAPTER IV
RESULTS
4.1 Method for evaluation of apparent activation energy of creep, a Q
The asymmetric heating applied to the samples in the in situ experiments produces a
temperature gradient (Table 3.2) which can be used to determine the temperature dependence
of strain and strain-rate along the stress axis. This dependence can be exploited to evaluate
as it will be explained next. However, the evaluation of the local strain and strain-rate is
only possible if we can correlate or follow the displacement of slices perpendicular to the
external stress during the creep process. This was done using the cross correlation method
explained in section
aQ
3.4.3. The temperature distribution in the specimen can be determined by
calibration measurements combined with finite element (FE) calculations. A description of
the calibration procedure will be given in section 4.1.1. Figure 4.1 shows the reconstruction of
a brass sample after 307 min of creep. The slab, composed of 10 slices, indicates that the local
strain is a function of temperature coupled with the position along the Z axis.
Figure 4.1. Tomographic reconstruction of brass sample (sample A) after 307 min of creep
showing a selected slab. The temperature increases from the top to the bottom.
IV. Results 53
Applying Eq. 2.1 to two slabs one located at a reference temperature (usually the
smallest temperature in the analyzed region) and the second at temperature T, the ratio of the
corresponding strain-rates can be written as follows:
0T
4.1) ⎟⎟⎠
⎞⎜⎜⎝
⎛⋅⋅
Δ⋅=
⎥⎥⎦
⎤
⎢⎢⎣
⎡⎟⎟⎠
⎞⎜⎜⎝
⎛⋅⎟⎟
⎠
⎞⎜⎜⎝
⎛⋅
TTRTQ
TGTG
tTtT
tTtT
nn
0a
0
0
0 )()(
),(),(
),(),(ln
σσ
εε&
&,
where . The local strains (and strain-rates) are obtained from tomographic
reconstructions performed at the reference time and current time t. Assuming that the
volume of the creeping material is constant the local strain in the slab can be related to the
local change in sample’s cross section as:
0TTT −=Δ
0t
4.2) ⎟⎟⎠
⎞⎜⎜⎝
⎛=
),(),(
ln),( 0
tTAtTA
tTε ,
while the local strain-rate in the time interval 12 ttt −=Δ (and associated to time ) is given
by:
2t
4.3) ⎟⎟⎠
⎞⎜⎜⎝
⎛Δ
=⎥⎦
⎤⎢⎣
⎡⎟⎟⎠
⎞⎜⎜⎝
⎛−⎟⎟
⎠
⎞⎜⎜⎝
⎛Δ
=Δ−
=),(),(
ln1),(),(
ln),(),(
ln1),(),(),(
2
1
1
0
2
0122 tTA
tTAttTA
tTAtTAtTA
tttTtT
tTεε
ε& ,
where is the average area of the undamaged material (without voids) in the slab at
temperature T and time t. It is important to note that the evaluation of the strain-rate at time
does not require the measurement of the initial state (at time ) but only the state at time .
However, if the local creep curves are of interest then the reference state is also required (this
will be shown later).
),( tTA
2t
1t0t
Figure 4.2 exemplifies the position change of three slabs during creep
and the methodology used for strain evaluation.
Considering a sample with constant cross-section in the reference state it becomes
possible within certain conditions to evaluate without knowing the stress exponent n. The
method applied to a constant cross sectional area will be exemplified for the evaluation of
in brass (sample A) in section
aQ
aQ
4.1.3. Usually in practice not only volumes with constant cross-
IV. Results 54
section are reconstructed, but also regions with varying section (see Figure 4.2) where the
stress is slightly changing. The stress ratio in Eq. 4.1 can be expanded in a series of the strain
difference between slabs at temperatures T and from which keeping only the first terms we
get:
0T
4.4) )),(),((1n
≅),(),(
),(),(
00
0 tTtTntTAtTA
tTtT
n
εεσσ
−+⎥⎦
⎤⎢⎣
⎡=⎥
⎦
⎤⎢⎣
⎡ ,
This equation suggests that the stress ratio approaches 1 if the strain difference between
slabs is small. This condition is usually fulfilled in case of small temperature gradients, when
the ratio of the shear moduli approaches unity, too. Considering the stress exponent n to be
known, Eq. 4.1 suggests that can be determined even in such cases, however, only with
the restriction that the cross-sectional area of the slices does not vary strongly within short
time intervals and the assumption of a uniaxial stress state is approximately fulfilled. In case
of the present experiments the stress triaxiality as determined from FE calculations in the
investigated region with varying cross section was below 0.38.
aQ
Figure 4.2. Examples of slab displacement along the tensile axis and the methodology of
strain and strain-rate calculation for the slab. N represents the number of slices in the slab.
IV. Results 55
4.1.1 Temperature distribution calibration
The temperature distribution in the specimen was determined by finite element (FE)
calculations taking into account the accurate shape of the sample and the temperatures
measured by the thermocouples. The calculations were performed for thermal steady-state
using the following material constants for brass (sample A): heat conductivity λ = 115
W/m K, specific heat c = 380 J/kg K and mass density ρ = 8530 kg/m3. For steel (sample B
and C) the material constants used were: heat conductivity λ = 24.2 W/m K, specific heat c =
460 J/kg K and mass density ρ = 7650 kg/m3. The FE results were confirmed by temperature
measurments performed by thermocouples welded at different hights for all samples. Figure
4.3 shows the temperature distribution in samples A (brass), B (steel), and C (steel) as a
function of the distance from the center of the specimens. There is a non-linear temperature
variation in regions with varying cross section, while in the middle of the samples, where the
cross-section is practically constant the temperature changes linearly with distance. Based on
temperature profiles obtained from FE simulations a certain average temperature and a
corresponding tolerance can be now ascribed to each slab. It will be shown later that the
temperature change of one slab during the short strain interval considered for the evaluation
of is small and the error in temperature is mainly determined by its variation over the
slab’s height.
aQ
IV. Results 56
Figure 4.3. Temperature profiles (a) and contour plots (b) along the stress axis as obtained
from FE simulations for samples A (brass), B (steel) and C (steel). Lines and number in (a)
indicate the linear temperature gradient at the center of the samples.
4.1.2 Image correlation for evaluation of tomographic data
Considering the tomographic reconstruction similar to data delivered by a 3D
extensometer it becomes possible in case of in situ tests to trace the movement of each
material slice or volume. The cross-correlation method explained in section 3.4.3 was used for
that purpose. Once the displacement of a slice was found, the corresponding local true strain
could be calculated as shown in Eq. 4.2 and depicted in Figure 4.2. For higher accuracy an
average strain over 10 slices making up a slab was calculated. Selecting a slab as structural
IV. Results 57
unit for evaluations was imposed also by the nearly constant temperature condition required
in the slab in the time interval selected for evaluation. Usually the difference in temperature
between the lower and upper slice in the slab is larger than the average temperature change of
the slab over the selected time interval. Figure 4.4 shows for example the displacement of
three selected slices (with regard to their original position) during the creep test. The material
slices were selected at the top, in the middle and at the bottom of the volume measured and
correlated for the whole creep test (steel, sample C). Since the bottom of the sample was kept
fixed all material slices move upwards. A slice at a higher position (lower temperature)
experiences therefore a larger displacement since the displacements of all slices situated
bellow are added together. Since the cross section of the material slightly varies along the
vertical axis ( ) the displacements of the slices are different and characteristic
for the local inhomogeneous deformation. For example, in the interval from 234 min to 300
min the slice 70 (
19.1/ minmax =dd
Figure 4.4) experience a temperature change of ~0.68 K which is small
compared to the total temperature difference in the analyzed volume of about 50 K.
Figure 4.4. Displacement of three slices along the loading direction as obtained from cross-
correlation of tomographic slices. Sample C (steel).
IV. Results 58
4.1.3 Method for aQ evaluation on samples with constant cross section
The structure of the sample A (brass) before fracture is shown in Figure 4.5a), where the
gradual increase of damage with increasing temperature is evident. Most quantitative
evaluations were performed on the central part of the sample, which fulfilled the constant area
criterion required by the constancy of the temperature gradient and stress. The selected ROI
indicated in dark grey (Figure 4.5a) was a cylindrical volume of 480 µm (300 voxels) in
height and of almost constant cross-sectional area (relative deviation less than 0.5%). Figure
4.5b) shows the spatial distribution and evolution of cavities within a subvolume of the ROI
after creep times of 52 and 110 min. Compared to samples tested in the laboratory, the
diameter of the in situ investigated specimen was small (~1 mm); however, the resolution of
the tomographic reconstructions (~0.002 mm) yields reliable local geometric information for
calculation of local strains [154] and consequently of local creep curves for different
subregions of the sample.
Figure 4.5. (a) Tomographic reconstruction of the sample A (brass) after 440 min creep. The
darker region indicates the selected ROI. (b) Spatial distribution and evolution of pores within
a subvolume of the ROI after creep times of 52 and 110 min.
IV. Results 59
4.1.3.1 In situ creep curve of brass (samples A)
The overall creep curves for sample A (brass, see Table 3.2) is plotted in Figure 4.6a),
together with the local creep curves of the material slabs. The black solid line is the curve
calculated from the total displacement of the sample (during the in situ measurement)
considering an effective length of 1.8 mm as defined in section 5.1.1. Local creep curves for
nine slabs (one slab was obtained as an average over 10 slices) positioned 96 µm from each
other along the tensile axis are shown in Figure 4.6a). The curves for the slabs begin at
52 min, when the first deformed state was reconstructed, and show a quite different behavior
as a function of time. The curve associated with the slab at the lowest temperature of 611 K
shows a nearly constant slope, characteristic of steady-state creep, while the slab with the
highest temperature of 641 K experiences the transition from steady-state creep to accelerated
creep; their average creep-rates at 52 min differ within a factor of 2, they are ~1.5·10-6 and
~3.0·10-6 s-1, respectively. It is important to mention that the local steady-state creep-rates in
the ROI of the pre-deformed sample are about one order of magnitude higher than the steady-
state creep-rate of large samples without pre-deformation and tested under similar conditions
in the laboratory [58].
Figure 4.6b) shows the global strain-rates as a function of time for sample A, with the
typical behavior of a dual phase α/β brass. After the “pseudo-minimum” Iε& the strain-rate
oscillates around a more or less constant value (as obtained by averaging the data in the
corresponding interval and shown in red) until a time of about 200 min, followed by tertiary
creep. Due to oscillations it is difficult to see that creep of the measured tomographic sample
has reached the steady state or not. However, from the point of view of the results presented
here, which correspond for the selected slabs this has no significance. The creep curves of the
slabs restricted to the ROI (Figure 4.6a) clearly show (lower curves, maximum T=628K) that
the local strain is linear with time.
Similar creep curves for a brass alloy were shown by Willis and Jones [141]. They
showed creep curves, which were generally tertiary-stage dominated, and are very similar to
the creep curve presented here.
IV. Results 60
Figure 4.6. (a) Local creep curves of slabs at different temperatures compared to the creep
curve of the tomographic sample A recorded during in situ measurement. (b) Variation of
strain-rate with time during creep of Cu-40Zn-2Pb at 25 MPa and 673 K at the bottom of the
sample.
4.1.3.2 aQ values according to literature and tomographic method
According to the data in Figure 4.6, steady-state creep prevails in the slabs of the ROI in
the time interval 52–137 min. In the particular case in which a volume with constant cross
section is analyzed the ratio of stresses and shear moduli in Eq. 4.1 approaches unity if the
temperature gradient between slabs is small compared to the applied temperature. For these
IV. Results 61
cases, the logarithm of the normalized local strain-rate referring to the lowest temperature, 0T ,
becomes a linear function of 0TTT −=Δ . Eq. 4.1 can be rewritten as:
TRTQ
TT a Δ≈ 2
00 )()(ln
εε&
& 4.5)
or the states investigated in sample A the stress ratio (given by Eq. 4.4 has only a
secon
F
d-order contribution and can be considered equal to 1. Figure 4.7 shows the normalized
strain-rates as a function of TΔ . Both data sets for time intervals 52–110 min (at 52 min in
Figure 4.7) and 110–137 m (at 110 min in in Figure 4.7) fit Eq. 4.5 with regression
coefficients of 0.89 and 0.93, respectively. The normalized strain-rates at 52 and 110 min
yield apparent activation energies for steady-state creep of 101 ± 12 and 117 ± 12 kJ/mol,
respectively, which can be considered equal within the limits of their standard deviations
considered as errors. Their average value of 109 kJ/mol differs by about 20% from the value
of 133 ± 9 kJ/mol obtained from laboratory tests on large samples with somewhat different
chemical composition (CuZn36Pb2.5) [58].
igure 4.7. Logarithm of average strain-rate ratios at temperatures T and T0 = 611 K at two F
creep times (52 and 110 min) as a function of temperature difference TΔ (the error of the
strain-rate ratios was estimated from the standard deviation of the area ch ge between the 10 an
cross-correlated slices making up the slabs; their relative error is about 15–20%).
IV. Results 62
4.1.3.3 Damage development in brass (samples A)
The temperature dependence of damage can simply be characterized by calculating the
area fraction of voids on slices perpendicular to the temperature gradient. The results
show
AA
n in Figure 4.8 indicate an almost constant AA in the initial stage (and also at 52 min),
which is followed by a progressive increase as a function of temperature and time, A
be
A
becoming as large as 7% before final rupture. According to Figure 4.8b (corresponding to the
ROI) the average AA
ch
stant along the sample length with an average area fraction of 0.019%. There exist,
howe
– disregarding its fluctuations – increases with temperature. It should
noted that in the present experiment voids are observed only after nucleation and growth,
once they have rea ed the resolution limit. Thus there are two possible contributions to the
observed inhomogeneous void-area increase: (i) voids at higher temperatures grow at higher
rates and (ii) in material regions with higher temperatures, voids have nucleated at higher
rates.
Despite of the resolution limit mentioned above the damage in the initial state seems to
be con
ver, regions in which damage is more localized, Figure 4.8b. After 52 min creep, the
average area fraction is 0.024% and it shows a faster development towards higher
temperatures. It should be also noticed that the initial distribution of the damage plays an
important role in the final fracture [155], since practically all the regions with a higher area
fraction of cavities developed faster than the less cavitated ones at all temperatures. It is
observed that even small variation in AA in the initial state can lead to large differences in
damage accumulation at the end of the creep lifetime. Even in the case when a temperature
gradient is applied, the influence of the itial damage on its further development is observed
over the whole temperature range. In the case of brass the regions where the A
in
A has
developed more rapidly can be already recognized after 137 min of creep.
IV. Results 63
Figure 4.8. (a) Variation of the area fraction of voids in slices perpendicular to the
temperature gradient as a function of temperature at various creep times (b) Enlarged view of
the marked region in (a).
AA
IV. Results 64
4.1.4 Method for aQ evaluation on samples with varying cross section
The structure of sample B (steel, K985 MPa,180 b == Tσ ) and C (steel,
K1013 MPa,210 b == Tσ ) after 951 and 194 min of creep is shown in Figure 4.9a) and b),
respectively. The correlated volume at these two states (region of interest -ROI-) is depicted
by the dark grey region in the middle of the volumes. The quantitative evaluations were
performed on the ROI. It can be observed in Figure 4.9a) and b) that the cross section slightly
varies from top (lower temperature) to bottom (higher temperature) for both samples,
however, for sample B the variation is less significant. The selected ROI for samples B and C
has a height of 699 µm (437 voxels) and 672 µm (421 voxels), respectively.
Figure 4.9. Tomographic reconstruction of the sample B (steel, K985 MPa,180 b == Tσ )
after 951 min creep (a) and of the sample C (steel, K1013b MPa,210 == Tσ ) after 194 min
creep (b). The darker regions indicate the selected and cross correlated regions of interest.
4.1.4.1 aQ values according to classical and tomographic methods
The creep behavior of the AISI 440B steel was first investigated by the conventional
method on large specimens. The stress exponent was evaluated from the slope of n )ln( minε&
vs. )ln(σ obtained from tests at constant temperature 873 K (see Figure 4.10a). Additionally,
IV. Results 65
the activation energy corresponding to the minimum creep-rate was obtained from tests
performed at different temperatures for two different normalized stress values of 1.4·10-3 and
3.1·10-3 (σ of about 77 MPa and 180 MPa, respectively). The corresponding linear fits
performed on the plot of )ln( minε&⋅− R
aQ
as function of yielded the values of
kJ/mol and 389±17 kJ/mol for the small and large normalized stress,
respectively. This means that does not depend on stress in the studied interval, which is in
good agreement with literature data for stainless steel [
T/1
15±388a =Q
45].
Figure 4.10. Evaluation of the stress exponent (a) and activation energy of steady-state
creep (b) from laboratory tests according to the conventional method.
n
IV. Results 66
4.1.4.2 In situ creep curves of steel (samples B and C)
The overall creep curves for samples B (steel, K985 MPa,180 b == Tσ ) and C (steel,
K1013 MPa,210 b == Tσ ) are plotted in Figure 4.11a) and c), respectively, together with
the local creep curves of the material slabs. The thick solid lines are the curves calculated
from the total displacement of the samples (during the in situ measurement) considering an
effective length of 1.8 mm as defined in section 5.1.1. The local curves (gray lines with
symbols) were obtained from tomographic evaluations by averaging the local strain over the
slices making up the slab. The reference state for calculation of the local creep curves was
considered at time 852 min and 156 min for sample B and C, respectively. These points on
the creep curve coincided with the beginning of the secondary state. For easier comparison
the local creep curves where shifted vertically to bring them together with the overall curves.
The insert in Figure 4.11a) serves this purpose, too. 44 and 42 slabs (each slab obtained as an
average of 10 slices) were followed from the beginning of the secondary state until the
beginning of the tertiary state for sample B and C, respectively. For a better representation,
only local creep curves for fifteen and sixteen slabs separated 48 µm from each other are
depicted in Figure 4.11a) and c) for sample B and C, respectively. For sample B, the average
creep-rates of the local creep curves differ by a factor of about 8 varying between
and for the coldest slab and hottest slab, respectively. The global
strain-rate was determined by hotter regions not captured by tomography. For
sample C, the average creep-rate for the local creep curves differ within a factor of about 11
and they lie between and for the coldest and hottest slabs,
respectively. The global strain-rate was .
-18 s100.8~ −⋅ 105.6~ ⋅
6 s102.1~ −⋅
.1~
-17 s−
-1
101⋅ -16 s− -15 s−
-1
102.1~ ⋅
6 s103. −⋅5~
The different creep behavior of the slabs is primarily due to the temperature gradient
leading to inhomogeneous strain distribution in the sample, i.e. the local creep behavior of
material slabs will differ from the overall behavior of the sample. Figure 4.11b) and d) show
the global strain-rates as a function of time for samples B and C, respectively. Sample B
experiences a long primary creep regime in which the strain-rate decreases continuously
reaching a steady-state after about 15 hours. For sample C, deformed at higher stress and
higher temperatures, a short transient regime with increasing and decreasing strain-rate is
observed between 60 and 120 min of the creep test. Such transient was also observed during
creep of a small brass sample [156]. All our evaluations were focused on the secondary
IV. Results 67
regime, when a nearly steady-state in the whole sample and in local material slabs was
reached.
It continues in the next page
IV. Results 68
Figure 4.11. Overall creep curves for samples B and C (thick solid lines) together with creep
curves of local material slabs (gray lines with symbols) obtained from tomographic
evaluations (a) and (c). Overall creep-rate vs. time for sample B (b) and C (d).
IV. Results 69
For the analysis of tomographic data it is preferred to plot the left hand side of Eq. 4.1
against )( 0 TTRT ⋅⋅Δ . When TTT ≅<<Δ 0 then the left hand side of Eq. 4.1 becomes a
linear function of TΔ and the fitted line should intersect the origin, too. Figure 4.12a) and b)
show the variation with temperature of the parameter-ratios entering Eq. 4.1 namely, that of
the strain-rates, stresses and shear moduli. It is evident that the main contribution to comes
from the ratio of the strain-rates and the other two should be considered as correction factors
with decreasing importance. Neglecting for example the shear modulus correction, has an
effect of less than 5% on . The slopes of the fitted lines for sample B and C (
aQ
aQ Figure 4.12a
and b) fit Eq. 4.1 with regression coefficients of 0.988 and 0.995, respectively, yielding
apparent activation energies equal to 377 ± 30 kJ/mol and 402 ± 12 kJ/mol for samples B and
C, respectively. These values are equal within experimental error with each other and with the
result obtained according to the conventional method. It should be noted that the error of
can be further reduced by increasing the resolution of the tomographic reconstruction. For
high energy beams, however, detectors with better efficiency are needed to pass the actual
resolution of about 2 µm.
aQ
IV. Results 70
Figure 4.12. Linear regressions performed on tomographic data to obtain ; (a) sample B
and (b) sample C. The diagrams also show the contributions of different ratios entering
Eq.
aQ
4.1. The sum of the three functions gives the left hand side of Eq. 4.1.
4.2 Microtomographic investigation of damage in E911 steel after long term creep
This part of the thesis is focused on the microtomographic evaluation of damage
distribution along the notch radius as well as cavities shape. The 3D shape, orientation and
size distribution will be described. Results from conventional metallography and tomography
IV. Results 71
will be compared. Also, a comparison of local damage distribution with the distribution of
selected stress parameters as obtained from FE simulation will be performed.
4.2.1 Conventional damage evaluation by OM
The cylinder extracted from the notch region of the hollow cylinder sample (E911 steel)
shown in Figure 3.4 was investigated by conventional optical microscopy (OM). Cavities
were analyzed on 2D sections perpendicular to the notch radius at three different distances
from the inner surface. Figure 4.13 shows examples of findings by optical microscopy. At the
magnification of the pictures, the visible cavities show an irregular shape, typical for large
cavities. From the 2D sections, however, it cannot be stated if the irregular cavity shape is due
to coalescence or to the growing mechanism. Also the coalescence of cavities of different
sizes could be observed (middle surface).
Figure 4.13. Micrographs of cavities at the different cross sections of the specimen (polished
surface).
IV. Results 72
The microstructure of the E911 steel is shown in Figure 4.14. It consists of tempered
martensite, the primary austenite grains are still visible. The cavities however are linked with
martensite laths. It is obvious that the shape of the cavities is influenced by the random
orientation of the single martensite laths in their direct vicinity.
Figure 4.14. Optical image of the E911 steel after 26,000 h of creep (uniaxial load of 8543 N
and internal pressure of 17.5 MPa) showing a martensitic microstructure. The voids are
mainly aligned with the martensite laths.
The evaluation of cavity density, area fraction and size was done by means of Digital
Image Processing (DIP). The results are given in Table 4.1.
Distance from
inner surface
Size (μm) Number
(1/mm2)
Area fraction
(%) Minimum Maximum Equivalent
diameter
400 μm 0.30 6.65 1.48 861 0.22
1000 μm 0.27 8.52 1.50 1848 0.53
1600 μm 0.21 8.36 1.77 2101 0.73
Table 4.1. Evaluation of cavity size, equivalent diameter, cavity density and area fraction at
three different distances from the inner surface.
IV. Results 73
4.2.2 Tomographic evaluation of damage
The microtomographic evaluation was focused on damage distribution along the notch
radius as well as on cavity shape. In the investigated state the voids appear to be
homogenously distributed on slices perpendicular to the notch radius (Figure 3.6), but their
number and average size varies along the radius (Z direction). Figure 4.15 shows qualitatively
the increase of pore density from the inner side towards the outer surface of the notch.
Figure 4.15. Tomographic reconstruction of the E911 steel specimen after 26,000 h of creep.
The section reveals inhomogeneous distribution of cavities along the notch radius (Z
direction).
Figure 4.16 shows two sub-volumes containing some local accumulation of voids of
various sizes probably situated at grain boundaries, information impossible to obtain from
tomography. They are located in the highly damaged region of the sample, where also the
largest pores were found.
IV. Results 74
Figure 4.16. Two sub-volumes taken from the region of the sample with maximum damage
(at 1600 µm from the inner surface of the notch) showing a region with high void density.
To compare the tomographic and metallographic data the cavity density on 2D slices
was first evaluated. Figure 4.17a) shows the variation of this density along the notch radius.
Each point corresponds to one slice with thickness equal to the voxel size of the
reconstruction (0.33 µm). An abrupt increase1 in cavity density is observed at about 200 µm
from the inner surface of the notch (where the pressure was applied). The cavity density
increases monotonically and reaches a maximum at about 1600 µm after which a slight
decrease towards the outer notch surface is observed. The maximum density is about
1700 pores/mm². The area fraction of voids (area of pores/total area investigated) shows
similar distribution as the cavity density and has its maximum value of 1.5% also around
1600 µm (Figure 4.17b). A fair agreement between tomographic and metallographic results
was obtained, the metallographic pore densities being somewhat larger.
1 An influence of the pressurizing medium (air) could be assumed: at the inner surface higher
contents of C and N have been found.
IV. Results 75
Figure 4.17. Cavity density (a) and area fraction (b) of cavities along the notch radius. The
results shown in three different gray levels were obtained from three tomographic
reconstructions coupled into one large volume. The circles show the cavity density obtained
from metallography.
The 3D tomographic reconstructions allowed the analysis of cavity shape, their
orientation and spatial distribution. Cavity shape was characterized by a complexity factor
( ) related to the exclusion volumes resulting from the intersection of the real cavity with
its equivalent ellipsoid as defined in section
CF
3.4.2. The varies between zero and two and
offers a good possibility to distinguish between regular cavities and those affected by
coalescence.
CF
The distribution of the complexity factor (CF ) and elongation ( e ) as a function of
volume is presented in Figure 4.18. Both the CF and e decrease at small cavity volumes
IV. Results 76
reaching a minimum value, followed by a further increase. This effect is related to the cavity
size and the error introduced when the equivalent ellipsoid is fitted to the cavity. This error
decreases with the increase of cavity size. However, as it is improbable that cavities with a
small size are affected by coalescence, these cavities will be considered as non-coalesced ones
for the calculation of the cavity size distribution function. For the calculation of cavity
orientation, a minimum cavity of 125 voxel and a minimum elongation of 1.4 will be
considered. With further increase of cavity size (from 125 to about 3000 voxel) both CF and
increase. For cavity sizes larger than 3000 voxel a large scatter is observed for both
parameters (CF and ) which is due to the smaller number of cavities sizes in that range.
e
e
Figure 4.18. Complexity factor CF , and elongation , as a function of the cavity volume.
The bin size for the volume was 20 voxels.
e
Based on a visual analysis and on the distribution functions of and , cavities with
have been considered regular and those with as complex.
CF
CF
e
5
5.0≤CF 5.0>CF Figure 4.19a) to
d) show selected cavities and their equivalent ellipsoids. Figure 4.19a) shows a very large
pore of 13,679 voxels which has experienced coalescence with several smaller pores located
along a given direction. Figure 4.19b) illustrates a case when the , however, it is
visible that the cavity is affected by coalescence indicating that the CF criteria alone may fail
in cases of coalesced cavities with broad interconnecting bridges (indicating that the
coalescence process is not recent).
.0≤
Figure 4.19c) shows the case of a recent coalescence where
IV. Results 77
the interconnecting bridges are narrow, hence the CF is larger than 0.5. Figure 4.19d)
represents the most typical case, a pore with a 5.0≤CF and 9.1≤e .
Figure 4.19. a) to c) Cavities with volume larger than 3000 voxels and their equivalent
ellipsoids. Volume, complexity factor (CF ) and elongation ( e ) are indicated in the picture. d)
Spheriod cavity shape typical for non-coalesced cavities.
Since coalescence generally leads to long cavities a new criteria based on the aspect
ratio of the equivalent ellipsoid was introduced. From the definition of elongation of the
cavity presented in section 3.4.2 the value of 9.1max =e was selected on a visual basis to
discriminate between coalesced and non coalesced cavities. Finally, a population of about
69,100 regular cavities could be separated, which fulfilled the condition to be affected only by
nucleation and growth. The histograms of voids major axis and equivalent radius are
shown in
1a ea
Figure 4.20 together with a fitted distribution function proposed by Riedel [13]
(presented in section 2.5.4).
IV. Results 78
The parameters βα ,,, 21 AA and γ of the size distribution function (Eq. 2.19) can be
related to general power-law functions describing non-stationary laws for the growth rate of
cavity radius a& nd the nucleation rate *J as expressed by Eq. a
o
2.17 and 2.18 [13]. According
to Figure 4.20 the function defined by Eq. 2.19 describes well the distribution of cavity sizes,
however, it approximates better the distribution of the equivalent radius ea than that f the
major axis 1a (only f results for ea will be discussed in the following). Since only one
deformation state was investigated the time t in Eq.
it
2.19 was considered constant, which
permitted the determination of exponents α and β characterizing cavity growth. The value
of 05.95.1= 0±β is close to 2 as predicted by the constrained diffusion mechanism [12,13].
The fitted exponent 200)( 1/() −+ ≅αγα is a very large number, which in case of realistic α
and γ of the order of unity, suggests 1≅α .
Figure 4.20. Probability density functions of cavity’s equivalent radius ea and major
ellipsoid axis 1a . The continuous lines represent the fit of Eq. 2.19. Only c with
5. and 9.1≤e were considered (cavities before coalescenc
avities
e happened). 0≤CF
The orientation of cavities with 5.0≤CF has been characterized based on the direction
of the eigenvector pointing along the major axis of the equivalent ellipsoid. For this purpose
the usual spherical coordinates, the polar angle θ and the azimuthal angle ϕ were used.
Figure 4.21 shows a system of curvilinear coordinates that are natural for describing positions
IV. Results 79
on a sphere. θ was defined with respect to the Z axis of a Cartesian system pointing along
the radius of the notch. The polar angle θ varies between πθ ≤≤0 . The azimuthal angle ϕ
varies between πϕ ≤≤0 in the xy-plane from the x-axis.
Figure 4.21. Definition of spherical coordinates used for describing the spatial orientation of
creep cavities.
The azimuthal angle ϕ , did not show any preferred orientation as a function of the
distance from the inner surface ( Z ) or cavity volume. Selecting a bin size of °=Δ 5ϕ the
corresponding number histogram for ϕ showed a nearly constant behavior. This means that
the voids are randomly oriented on planes lying perpendicular to the notch radius. These
planes (X-Y) contain, however, the direction of the applied axial load. On the other hand the
distribution of the polar angle θ , shows a peak at 90° with respect to the Z axis (Figure
4.22). This maximum is characteristic for almost all θ distributions calculated for
subvolumes situated at various distances from the inner surface. The maximum of the
distributions becomes sharper with increasing Z .
IV. Results 80
Figure 4.22. Distribution of the polar angle θ at different Z locations along the notch radius.
Z bin size = 33 µm (100 slices), not all the curves are plotted. θ bin size = 5°. Selection
cavity condition: 5.0,9.11 vox,125umecavity vol min4. ≤≤≤≥ e CF .
V. Discussion 81
CHAPTER V
DISCUSSION
5.1 Evaluation of apparent activation energy of creep, aQ
5.1.1 Comparison of strain distribution in the real sample and FE models
The overall strain is a macroscopic quantity characteristic for the whole sample. Due to
the restricted field of view of the CCD camera, tomography gives information only from a
portion of the gauge length and the average “tomographic” strain might not coincide with the
strain calculated from the measured displacement of the sample’s top (Figure 4.11a, c and
Figure 4.6a). In order to check if the local strain distribution delivered by tomography
matches the strain distribution in homogeneous material, FE simulations have been
performed. The comparison is shown in Figure 5.1a) and b) for the sample B and C,
respectively. Similar strain profiles were obtained for sample A with the maximum of the
strain distribution function closer to the center of the sample due to the smaller temperature
gradient applied. Therefore, the discussion will be focussed on sample B and C. The
conclusions are, however, general and applicable also to sample A. The contour plots and the
above located diagrams in Figure 5.1 clearly show that the maximum strain regions are shifted
towards higher temperatures (negative position values in Figure 5.1a) and b) which were not
captured by tomography. The regions evaluated are indicated by vertical lines and are
confined to the interval [-0.67, -0.05] mm for sample B and [-0.90, -0.41] mm for sample C.
The inserts in the upper right corner of the figures show the strain distribution obtained from
tomography at different creep times. The continuous black line with full squares depicts the
distribution at the end of steady-state. The behavior of material slabs at different positions
with respect to the maximum of the strain distribution is reflected in the local creep curves
shown in Figure 4.11a) and c). Hence a slab situated closer to the maximum of the strain
distribution has not only a larger strain, but also elongates at a larger strain-rate. In FE
simulations a Norton law describing steady-state creep with appropriate constants for the
investigated steel was used and deformation was simulated until the elongation of the sample
reached the value describing the end of secondary stage. The strain distribution obtained from
FE simulations facilitates a better understanding and design of the in situ experiment. It
becomes evident that the field of view of the CCD should be positioned in the sample region
V. Discussion 82
with the highest strain gradient, where the ratio of strain-rates )),(/),(( 0 tTtT εε && is the largest
and less affected by experimental errors.
The strain distribution obtained from FE calculations allows introducing an “effective
length”, a quantity that facilitates the calculation of the average strain and the overall creep
curve. In case of tests performed at constant temperature the strain distribution along the
sample axis can be well described by a Gaussian function. For asymmetric heating the
distribution becomes asymmetric, which suggests selecting a criterion based on the relative
amount of strain present in a given volume around the center of the strain profile. The average
strain in the sample of length is obtained as follows: L
5.1) L
Udzz
Lexp)(1
== ∫εε . (14)
where the integration is performed over L and is the measured displacement. Since the
strain is localized in a small region, using L as the effective length is not adequate. Therefore
we define an effective length equal with the interval around the center of the strain
profile, which contributes by 95% to the total elongation of the sample. The was
evaluated from strain profiles obtained from FE simulations and its value is about 1.8 mm,
almost twice the gauge length of the sample with constant cross section. The average strain
(and strain-rate) shown in
expU
effL
effL
Figure 4.11 (thick solid lines) were calculated according to Eq. 5.1
considering . It should be noted, however, that the value of is not affected by
the selected , which affects only the overall curves.
mm8.1=L
effL
aQ
V. Discussion 83
Figure 5.1. Strain distribution (obtained from FE simulations) along the load axis of the specimens characterizing the end of steady-state (a) sample
B and (b) sample C. The upper part in each figure shows the asymmetric strain distribution. The volume measured by tomography is marked by
vertical lines. The inserts in the upper right corners show the strain distribution in the sample at stages where the tomographic reconstructions were
performed.
VI. Discussion 84
5.1.2 Error estimation
For every new evaluation method it is important to characterize the sources of error
influencing the final results. The resolution of the tomographic reconstruction is of about
2 µm, which in case of a sample with diameter equal to 1 mm, gives a relative error of the
local strain and strain-rate below 1% (the uncertainty in determining the diameter is
considered equal with the tomographic resolution). This is much smaller than the standard
deviation of the average strain (and strain-rate) of one slab, of about 12%, determined by the
different creep behavior of the slices making up the slab. The stress ratio in Eq. 4.1 reduces
also to the ratio of two areas and consequently it has the same relative error.
Another source of uncertainty is related to the local temperature calibration obtained
from FE simulations. Selecting a slab (instead of a slice) as structural unit the error of the
temperature associated to the slab depends on the location of the slab (through the local
temperature gradient and local displacement) and the number of slices included. In case of
samples B and C the slabs were considered to be made up of 10 slices, which gave maximum
temperature errors of ±0.8 K and ±0.9 K, respectively. The temperature error due to sample
elongation during steady-state creep was estimated based on the experimentally detected
displacement for the uppermost slab in the reconstruction and the temperature gradient in the
specimen at the beginning of steady-state. The total analyzed creep strain in sample B was
2.3% ( =310 min) and 5.5% ( = 174 min) in sample C. These led to a temperature change
of the uppermost slab of about 0.3 K and 1.5 K for specimens B and C, respectively. Since the
slabs are moving upwards the temperature error becomes asymmetric. The total temperature
errors are equal to +0.8 K and -1.1 K for sample B and +0.9 K and -2.4 K for sample C. These
errors correspond to the slab with the highest location and compared to the total temperature
difference in the analyzed volume of about 50 K are small. The acceptable accuracy of the
local temperature and strain-rate ratios, as well as the large number of slabs analyzed, makes
an accurate evaluation of the activation energy possible. Both values of obtained from in
situ tests as well as the values obtained from conventional laboratory tests are equal within an
error of 5%.
tΔ tΔ
aQ
VI. Discussion 85
5.2 Damage investigation in E911 steel
Damage dependence on different stress parameters under multiaxial creep was usually
studied on notched specimens [73] and the stress parameters (maximum principal stress,
equivalent von Mises stress) used for interpretation of damage were characteristic for the
entire sample. Nix et al. [135] have also considered the principal facet stress prevailing at the
center of the notch, their argument being that the stress state around the notch center remains
constant in a reasonably large volume. As shown in Figure 4.17 microtomography is able to
characterize damage at the mircometer scale giving not only an average value, but also the
distribution of it along the notch radius. Since the stress state varies along the radius the
availability of local damage distribution offers a more accurate possibility to check the
influence of different stress parameters on creep damage. The local stress and stress state were
obtained from finite element (FE) simulations considering the viscoplastic Norton-law to
describe creep of the notched hollow cylinder. However, considering only steady-state creep
means that strain hardening effects are neglected. The experimentally determined parameters
used in the creep law were the stress exponent n = 7 and the apparent activation energy
. Elastic properties of the E911 steel were taken into account by considering a
Young’s modulus of 177 GPa and a Poisson’s ratio of 0.31. A constant load of 8543 N and a
pressure of 17.5 MPa were applied at one extreme of the sample (along the Y axis) and at the
inner surface of the hollow cylinder, respectively. After loading four stress parameters were
calculated at each integration point: the maximum principal stress (MPS), the equivalent von
Mises stress (VMS), the principal facet stress (PFS) and the stress triaxiality. The distribution
of the MPS, VMS and PFS in the notch after steady-state creep deformation of 0.30%
corresponding to a creep time of 20,000 h are shown in
kJ/mol458a =Q
Figure 5.2 (0.32% in 6,000 h
correspond to primary creep and were not considered for the calculation of the stress
parameters). Although the distributions of the MPS and VMS are different, their values are
nearly of the same order of magnitude, all increasing towards the outer surface of the notch.
The PFS, however, has an average value higher than the VMS and MPS by a factor of about
1.9 and it doesn’t show a monotonic increase towards the outer surface of the sample.
VI. Discussion 86
Figure 5.2. Stress distribution inside the notch after 20,000 h of creep ( %30.0=ε ). The
sample region measured by tomography is marked by the dotted line.
According to the FE results neglecting transient creep should have only a minor effect
of the final stress distribution in the notch. The contour-plots of the stress parameters evolve
differently during the first 8,000h. For example, the position of the kink of the MPS and PFS
(see Figure 5.4b and c) varies during this time from Z = 1900 µm until 1350 µm, as shown in
Figure 5.3. However, after a creep time of 8,000 h the contour-plots of the stress parameters
evolve in a nearly self-similar manner, i.e. the highest and lowest values remain practically
unchanged and in the case of MPS and PFS the location of kink position remains between
Z = 1350 µm and 1250 µm for over 12,000 h.
VI. Discussion 87
Figure 5.3. Variation with time of kink location of the MPS and PFS parameters (Figure 5.4b
and c).
The distributions of the stress parameters along the notch radius after 20,000 h of creep
are plotted in Figure 5.4a) to d) and compared with the cavity density distribution. The VMS
and MPS (Figure 5.4a and b, respectively) have similar values, however, their distributions
are quite different. The VMS increases continuously along the radius, while the MPS shows a
kink at Z ≈ 1255 µm. The PFS shows also a kink at the same position which is, however, the
minimum of the distribution.
To obtain a more accurate damage parameter the volumetric cavity density
corresponding to local volumes made up of 30 slices was evaluated. The result shown in
Figure 5.4a) to d) is, however, similar to the distribution obtained from 2D sections. None of
the parameters selected is 100% similar to the cavity density, which can have two reasons: a)
Due to its limited resolution, of about 0.5 μm, microtomography does not detect all cavities.
This hypothesis can, however, be dropped since the modus of cavities size distribution was
captured by the technique (Figure 4.20), suggesting that the proportion of non-detected
cavities (assuming unimodal distribution) is small, which will not change significantly the
found distribution. b) The stress parameters selected does not influence equally the
mechanisms of nucleation and growth, which both contribute to the observed distribution.
An eventual correlation between stress parameters and creep damage can be obtained
based on the comparison between the corresponding distributions (Figure 5.4a to d). The
cross-correlation coefficients of the selected stress parameters and the volumetric cavity
density are given in Table 5.1.
VI. Discussion 88
Cross-correlation Coefficient
Stress triaxiality
factor, ft
Maximum Principal
Stress, MPS
Von Mises
Stress, VMS
Principal Facet
Stress, PFS
Cavity density
function 0.978 0.917 0.872 -0.594
Table 5.1. Cross-correlation coefficients obtained from comparison of the cavity density
function and the value of stress parameters.
It continues in the next page
VI. Discussion 89
Figure 5.4. a) Variation of the von Mises stress (VMS), b) maximum principal stress (MPS),
c) principal stress facet (PFS), and d) stress triaxiality factor ( ) compared with volumetric
cavity density along the notch radius (The stress parameters were obtained from FE
simulation).
tf
The stress parameter showing the best similarity with cavitiy density is the triaxiality
factor, . tf Figure 5.4d) shows the comparison of both distributions. Even though the
distributions do not begin in a similar manner, a similar slope and maximum can be observed
VI. Discussion 90
on the distribution of the stress triaxiality factor. The correlation coefficient of these two
functions is the highest of 0.978. The VMS and the MPS have similar correlation coefficients,
however they are significantly smaller than that of . Both, the VMS and the cavity density
distribution begin with a plateau, which changes into a linear increase at about Z ≈ 200 μm.
The slope at this point, however, is higher for the distribution of real damage. At higher Z the
VMS increases monotonously, while the experimental cavity density reaches a maximum
plateau at Z ≈ 1600 μm, followed by a slight decrease towards the external surface. Assuming
that cavity nucleation is primarily affected by the VMS and cavity growth by the stress
triaxiality the whole damage distribution curve could be explained as a superposition of the
two mechanisms. Most of damage found at the maximum of stress triaxiality is in good
qualitative agreement with the theory describing vacancy diffusion under the action of
hydrostatic stress gradient [
tf
157]. Watanabe [66] concluded also that the crack initiation site
and crack growth path in 9Cr–1Mo–V–Nb (P91) steel welded joints coincided well with the
distribution of the stress triaxiality factor. Comparing experimental creep damage
distributions with computed FE results Li et al. [75] have found that both creep strain
concentration and high stress triaxiality have to be considered to explain creep void formation
and growth; a result similar to that obtained in this work.
In case of intergranular fracture literature data suggest that the cavity density and the
growth rate of grain boundary voids is described either by the maximum principal stress
[70,73,116,158] or by the principal facet stress [135]. Both parameters show a kink at about
1255 μm, which is difficult to recognize on the measured distribution due to experimental
scatter. We note, however, that the PFS model is expected to be valid for uniform grain
structures showing grain boundary sliding [135]. The E911 martensitic steel reveals in
contrast a band like granular structure, when GBS is more difficult. The fairly high correlation
coefficient obtained for the MPS indicates that it has an important effect on the growth of
cavities. Thus, the maximum of the cavity density is well described by (tf Figure 5.4d), the
possible kink position of the cavity density function is observed at the same position of the
kink of the MPS (Figure 5.4b) which according to the FE simulation it remains fairly stable
for over 60% of the creep time. These two parameters also account fairly well for the slope of
the distribution under the multiaxial stress state applied. However, the beginning of the
distribution where the density of cavities is small the VMS, considered to affect the
nucleation, correlates better with the cavity density distribution. Furthermore, the direction of
the MPS changes over 90° along the sample length, however, no preferential orientation of
the voids was found for the azimuthal angle as a function of Z.
VI. Discussion 91
An indirect hint about the cavity growth mechanism is obtained by comparing the
experimental size distribution with the function proposed by Riedel [13], which takes into
account both continuous cavity nucleation and growth. The agreement is remarkable and the
exponents α and β determined from the fit have a high degree of confidence. 2≅β is in
good agreement with the constrained diffusional mechanism [12,13], which predicts a growth
rate for cavity radius proportional to . The second exponent -2ea 1≅α indicates that the size
distribution function depends on time. This non-stationary feature is related to continuous
nucleation, which decreases the average distance between cavities.
VI. Conclusions and perspectives 92
CHAPTER VI
CONCLUSIONS AND PERSPECTIVES
It was shown that in situ microtomography is an excellent method for the investigation
of specimens undergoing inhomogeneous deformation allowing a good characterization of the
kinematics of steady-state creep. The good accuracy of the method is related to the use of
powerful image correlation algorithms, which applied to 3D tomographic images precisely
capture changes in specimen’s shape. The possibility to apply different loads on the sample
like temperature, electric or magnetic fields, shows the flexibility of the experiment, which in
present case was performed in a 5 dimensional parameter space (three spatial dimensions, the
temporal dimension and a temperature gradient).
Activation energies of steady-state creep obtained with the new tomographic method on
small samples are in good agreement with values from literature and those obtained by
conventional methods on large specimens. The good agreement has two reasons: i) the
tomographic method uses the true values of strain-rate and stress and ii) the slabs chosen as
structural unit to characterize the inhomogeneous deformation still contain a large number of
dislocations to fulfill the conditions necessary to treat them on the basis of continuum
mechanics. The newly developed tomographic method should be attractive for the evaluation
of activation energy in expensive materials or in cases when only a limited amount of material
is available.
The method allows also evaluating the apparent activation energy in steady-state creep
as a function of time or strain. Then, any change in the activation energy can be related with
microstructural changes such us precipitation of new phases, coarsening of the already
existing ones or changes in the dislocation density. In this case, however, additional
investigation of the microstructure evolution is needed. With the increase of the tomographic
resolution, the error of can be further reduced, which is mainly determined by the
resolution of the tomographic technique. However, to increase the resolution detectors with
better efficiency at energies above 50 keV are needed.
aQ
From a single in situ experiments lot of information is available, from the growth
evolution of single cavities presented in [159] to damage development as function of
temperature and time or the extraction of material parameters, such as apparent activation
energies of steady-state creep. Therefore, it would be important to plan long-term creep
VI. Conclusions and perspectives 93
experiments in cooperation with synchrotron radiation laboratories in order to study the creep
damage of engineering materials, under conditions characteristic of engineering applications.
In situ experiments with varying stress and constant temperature could be also studied
to evaluate the stress exponent, , in case of a homogeneous material undergoing creep.
Microtomography could be also applied to the study of creep behavior of inhomogenous
materials (i.e. composition or grain size gradients). The possibility to apply electric or
magnetic fields to specimens studied in situ opens up a broad range of experiments delivering
new information about the behavior of engineering materials under complex inhomogeneous
loading conditions.
n
It has been shown that cavities density evaluated from microtomographic
reconstructions is in fair agreement with metallographic results obtained from 2D sections.
Compared to the 2D technique, however, microtomography allows a more detailed
characterization of creep damage. The cavities shape can be well estimated based on the
ellipsoid fit and the shape descriptors defined from it, namely complexity factor (CF ),
elongation ( ) and flatness ( ). From this approximation, statistically relevant information
on the size, shape, orientation and number of cavities in bulk sample could be obtained. Based
on the selection of proper CF and values, the non-coalesced cavity population could be
identified and used for a more straightforward analysis. The study of pore size distributions
suggest that growth of cavities in the E911 steel is governed by the constrained diffusional
mechanism. The study of different deformation states will allow obtaining reliable
information about cavity nucleation, which was very difficult until now.
e f
e
Tomography delivered additionally the continuous distribution of damage along the
notch radius, which could be directly compared to the distribution of stress parameters
obtained from FE modeling. This comparison indicates that cavities density found by
microtomography is most probably affected by the combined action of the equivalent von
Mises stress and stress triaxiality. However, due to its high cross-correlation coefficient the
maximum principal stress can also influence damage development. Although its effect might
be recognized from the location of the kink present in both the MPS and cavity density
distribution, a random spatial orientation of cavities was found in the XY plane.
VII. References 94
7. References [1] M.E. Kassner, M.T. Pérez-Prado, Fundamentals of Creep and Alloys, Elsevier, San
Diego (2004).
[2] M.E. Kassner, M.T. Pérez-Prado, Five-power-law creep in single phase metals and
alloys, Prog. Mater. Sci. 45 (2000) 1-102.
[3] D. Hull, D.E. Rimmer, The growth of grain-boundary voids under stress, Philos. Mag.,
4 (1959) 673.
[4] R. Raj, M.F. Ashby, Intergranular fracture at elevated temperature, Acta Metall. 23 (6)
(1975) 653–666.
[5] A.C.F. Cocks and M.F. Ashby, Intergranular fracture during power-law creep under
multiaxial stresses, Met. Sci., 14 (1980) 395-402.
[6] A. Needleman, J.R. Rice, Plastic creep flow effects in the diffusive cavitation of grain
boundaries, Acta Metall. 28 (10) (1980) 1315–1332.
[7] M.S. Yang, J.R. Weertman, M. Roth, A test of grain boundary void growth theories by
small angle neutron scattering, Scr. Metall. 18 (1984) 543-548.
[8] X.-W. Du, J. Zhu, Y.-W. Kim, Microstructural characterization of creep cavitation in a
fully-lamellar TiAl alloy, Intermet. 9 (2) (2001) 137-146.
[9] G. Dimmler, W. Schalk, P. Weinert, H. Gruber, S. Wieser, H. Cerjak, Sample
preparation for high-resolution damage examinations in the SEM, Praktische
Metallographie/Pratical Metallography 39 (12) (2002) 619-633.
[10] T-J. Chuang, K.I. Kagawa, J.R. Rice, L.B. Sills, Overview no. 2: Non-equilibrium
models for diffusive cavitation of grain interfaces, Acta Metall. 27 (1979) 265-284.
[11] A.C.F. Cocks, M.F. Ashby, On creep fracture by void growth, Prog. Mater Sci., 27
(1982) 189-244.
[12] M.E. Kassner, T.A. Hayes, Creep cavitation in metals, Int. J. Plast., 19 (10) (2003)
1715-1748.
[13] H. Riedel, Fracture at High Temperatures, Springer Verlag, Berlin (1987).
[14] R. Page, Small-angle neutron scattering investigation of creep damage in type 304
stainless steel and alloy 800, Metall Trans. A 16 (12) (1985) 2283-2289.
[15] P. J. Bouchard, P. J. Withers, S. A. McDonald, R. K. Heenan, Quantification of creep
cavitation damage around a crack in a stainless steel pressure vessel, Acta Mater. 52
(1) (2004) 23-34.
[16] F. Abe, T.-U. Kern, R. Viswanathan, Creep-resistant steels, Woodhead publishing
limited and CRC Press LLC, England (2008).
VII. References 95
[17] E.N. da C. Andrade, On the viscous flow in metals and allied phenomena, Proc. Roy.
Soc. London, A.84 (1910) 1-12.
[18] E.N. da C. Andrade, The flow of metals under large constant stresses, Proc. Roy. Soc.
London, A.90 (1914) 329-342.
[19] V. Levitin, High temperature strain in metals and Alloys, Wiley-VCH Verlag GmbH
& Co. KGaA, Weinheim (2006).
[20] F.C. Monkman, N.J. Grant, An empirical relationship between rupture life and
minimum creep rate in creep-rupture tests, Proc ASTM 56 (1956) 593-620.
[21] P. Feltham, J.D. Meakin, Creep in face-centered-cubic metals with special reference to
copper, Acta Metall 7 (1959) 614-627.
[22] H.E. Evans, Mechanisms of Creep Rupture, Elsevier Applied Science Publishers,
London and New York (1984).
[23] D.C. Dunand, B.Q. Han, A.M. Jansen, Monkman-Grant analysis of creep fracture in
dispersionstrengthened and particulate-reinforced aluminium, Metall. Trans. 30A
(1999) 829-838.
[24] F. Doběs, K. Milička, The relation between minimum creep rate and time to fracture,
Met. Sci. 10 (1976) 382-384.
[25] H.J. Frost, M.F. Ashby, Deformation-Mechanism Maps, The Plasticity and Creep of
Metals and Ceramics, Pergamon Press, Oxford, New York (1982).
[26] R.C. Gifkins, Transitions in creep behaviour, J. Mater. Sci., 5 (1970) 156-165.
[27] T.G. Langdon, F.A. Mohamed, The characterization of independent and sequential
creep processes, J. Austr. Inst. Met., 22 (1977) 189-199.
[28] J.P. Poirier, Creep of crystals. High-temperature deformation processes in metals,
ceramics and minerals, Cambridge University Press, Great Britain (1990).
[29] E. Orowan, Problems of plastic gliding, Proc. Phys. Soc., 52 (1940) 8.
[30] J. Weertman, Theory of steady-state creep based on dislocation climb, J. Appl. Phys.,
26 (1955) 1213.
[31] J. Weertman, Steady-state creep through dislocation climb, J. Appl. Phys., 28 (1957)
362.
[32] J. Harper, J.E. Dorn, Viscous creep of aluminum near its melting temperature, Acta
Met., 5 (1957) 654.
[33] M. V. Speight, J. E. Harris, Kinetics of stress-induced growth of grain-boundary voids,
Met. Sci. J., 1 (1967) 83.
VII. References 96
[34] W.D. Nix, B. Ilschner, In: P. Haasen, V. Gerold, G. Kostorz, editors. Strength of
metals and alloys, Oxford: Pergamon Press (1980).
[35] O.D. Sherby, R.J. Klundt, A.K. Miller, Flow stress, subgrain size, and subgrain
stability at elevated temperature, Metall. Trans. A 8A (1977) 843.
[36] H. Luthy, A.K. Miller, O.D. Sherby, The stress and temperature dependence of steady-
state flow at intermediate temperatures for pure aluminum, Acta Metall., 28 (1980)
169-178.
[37] S.H. Goods, W.D. Nix, The coalescence of large grain boundary cavities in silver
during tension creep, Acta Metall., 26 (1978) 753-758.
[38] P.R. Landon, J.L. Lytton, L.A. Shepard, J.E. Dorn, The activation energies for creep
of polycrystalline copper and nickel, Trans. ASM, 51 (1959) 900.
[39] C.M. Sellars, A.G. Quarrell, The high temperature creep of gold-nickel alloys, J. Inst.
Met., 90, (1961-62), 329-336.
[40] M. Pahutová, J. Čadek, Interpretation of high temperature creep in alpha-zirconium in
terms of effective stress and dislocation dynamics, Mater. Sci. Eng. 11 (1973) 151-
162.
[41] W. Blum, Creep of crystalline materials: experimental basis, mechanisms and models,
Mater. Sci. Eng., A 319–321 (2001) 8–15.
[42] K.-G. Tak, U. Schulz, G. Eggeler, On the effect of micrograin crystallography on
creep of FeCr alloys, Mater. Sci. Eng., A 510–511 (2009) 121–129.
[43] W. Blum, P. Eisenlohr, Dislocation mechanics of creep, Mater. Sci. Eng., A 510–511
(2009) 7–13.
[44] R.E. Reed-Hill, R. Abbaschian, Physical Metallurgy, third ed., PWS-Kent, Boston
(1992).
[45] F. Garofalo, Fundamentals of creep and creep-rupture in metals, Macmillan, New
York (1965).
[46] O.D. Sherby, T.A. Trozera, J.E. Dorn, Effects of creep-stress history at high
temperatures on creep of aluminum alloy, Proc. Am. Soc. Test. Mat., 56 (1956) 789-
804.
[47] O.D. Sherby, J.L. Lytton, J.E. Dorn, Activation energies for creep of high-purity
aluminum, Acta. Metall., 5 (1957) 219.
[48] J.E. Dorn, in Creep and Fracture of metals at High Temperatures, Proc. NPL
Symposium, HMSO, London (1956), p.89.
VII. References 97
[49] C. Zener, J.H. Hollomon, Effect of strain rate upon plastic flow of steel, J. Appl.
Phys., 15 (1944) 22-32.
[50] C.N. Ahlquist, R. Gasca-Neri, W.D. Nix, A phenomenological theory of steady-state
creep based on average internal and effective stresses, Acta Metall. 18 (1970) 663–
671.
[51] F. Doběs, K. Milička, Internal stress and activation energy of creep, Mater. Sci. Eng.,
A 462 (2007) 380–383.
[52] C.N. Ahlquist, W.D. Nix, A technique for measuring mean internal stress during high
temperature creep, Scr. Metall. 3 (1969) 679-682.
[53] W. Blum, A. Finkel, New technique for evaluating long range internal back stresses,
Acta Metall. 30 (1982) 1705-1715.
[54] M.D. Hanna, G.W. Greenwood, Cavity growth and creep in copper at low stresses,
Acta Metall. 30 (1982) 719-724.
[55] J. Lapin, M. Nazmy, Microstructure and creep properties of a cast intermetallic Ti–
46Al–2W–0.5Si alloy for gas turbine applications, Mater. Sci. Eng. A 380 (1) (2004)
298-307.
[56] D.W. Livesey, N. Ridley, Cavitation and cavity growth during superplastic flow in
microduplex Cu-Zn-Ni alloys, Metall. Trans., A 13 (1982) 1619–1626.
[57] G. Eggeler, Microstructural parameters for creep damage quantification, Acta Metall.
Mater., 39 (1991) 221–231.
[58] V. Buršíková, J. Buršík, V. Navrátil, K. Milička, Creep behaviour of leaded brass,
Mater. Sci. Eng., A 324 (2002) 235–238.
[59] L. Salvo, P. Cloetens, E. Maire, S. Zabler, J.J. Blandin, J.Y. Buffière, W. Ludwig, E.
Boller, D. Bellet, C. Jossenrond, X-ray micro-tomography an attractive
characterisation technique in material science, Nucl. Instrum. and Methods Phys. Res.,
B 200 (2003) 273–286.
[60] I.-W. Chen, A.S. Argon, Creep cavitation in 304 stainless steel, Acta metall. 29 (1981)
1321–1333.
[61] N.G. Needham, T. Gladman, Nucleation and growth of creep cavities in a Type 347
steel, Metal Sci. 14 (1980) 64-73.
[62] B.J. Cane, Interrelationship between creep deformation and creep rupture in 2¼ Cr-
1Mo steel, Met. Sci., 13 (1979) 287–294.
[63] B.J. Cane, G.W. Greenwood, The nucleation and growth of cavities in iron during
deformation at elevated temperatures, Metal Sci. 9 (1975) 55-60.
VII. References 98
[64] R. Pilkington, D.A. Miller, D. Worswick, Cavitation damage and creep crack growth,
Metall. Trans. A 12 (1981) 173-181.
[65] D.W. Livesey, N. Ridley, Superplastic deformation, cavitation, and fracture of
microduplex copper-nickel-zinc alloys, Metall. Trans. A 9A (4) (1978) 519-26.
[66] T. Watanabe, M. Tabuchi, M. Yamazaki, H. Hongo, T. Tanabe, Creep damage
evaluation of 9Cr–1Mo–V–Nb steel welded joints showing Type IV fracture, Int. J.
Press. Vessels Pip., 83 (2006) 63-71.
[67] Y.H. Hsiao, H. Zhang, G.S. Daehn, Effect of multiaxial stresses on creep damage of
316 stainless steel weldments, Metall. Mater. Trans., A 27 (4) (1996) 891–900.
[68] V. Gaffard, J. Besson, A.F. Gourgues-Lorenzon, Creep failure model of a tempered
martensitic stainless steel integrating multiple deformation and damage mechanisms,
Int. J. Fract. 133 (2005) 139-166.
[69] L.-B. Niu, M. Nakamura, A. Futamura, M. Kobayashi, Effect of stress states on creep
fracture modes of an austenitic steel with high ductility, ISIJ Int. 40 (5) (2000) 511-
518.
[70] L.-B. Niu, M. Kobayashi, H. Takaku, Creep rupture properties of an austenitic steel
with high ductility under multi-axial stresses, ISIJ Int., 42 (10) (2002) 1156–1161.
[71] L. Falat, A. Výrostková, V. Homolová, M. Svoboda, Creep deformation and failure of
E911/E911 and P92/P92 similar weld-joints, Eng. Fail. Anal. (2009), article in press.
[72] S. Zahng, T. Wakai, M. Sakane, Creep rupture life and damage evaluation under
multiaxial stress state for type 304 stainless steel, Mat. Sci. Eng. A 510-511 (2009)
110-114.
[73] N.G. Needham, T. Gladman, The effect of stress-state on the processes controlling
creep fracture in low alloy ferritic steels, In: D. Wilshire, D.R.J. Owen (Eds), Creep
and Fracture of Engineering Materials and Structures. Part II, Pineridge Press,
Swansea (1984), pp. 1263–1276.
[74] T. Ogata, T. Sakai, M. Yaguchi, Damage characterization of a P91 steel weldment
under uniaxial and multiaxial creep, Mat. Sci. Eng. A 510-511 (2009) 238-243.
75] Y. Li, H. Hongo, M. Tabuchi, Y. Takahashi, Y. Monma, Evaluation of creep damage
in heat affected zone of thick welded joint for Mod. 9Cr–1Mo steel, Int. J. Press.
Vessels Pip., (2009) 1-8, article in press.
[76] M. Tabuchi, H. Hong, Creep damage behavior of welded joints for high Cr steels,
Proc. of 3rd Symposium on heat resistant steels and alloys for high efficiency USC
power plants, Japan (2009).
VII. References 99
[77] A. Isaac, K. Dzieciol, F. Sket, M. di Michiel, T. Buslaps, A. Borbély, A. Pyzalla,
Investigation of creep cavity coalescence in brass by in-situ synchrotron X-Ray
microtomography, In: S.R. Stock, Editor, Developments in X-Ray tomography VI,
Proc. of SPIE, 7078 (2008) pp. J1-J10.
[78] L. Babout, E. Maire, R. Fougères, Damage initiation in model metallic materials: X-
ray tomography and modeling, Acta Mater., 52 (2004) 2475-2487.
[79] C.F. Martin, C. Josserond, L. Salvo, J.J. Blandin, P. Cloetens, E. Boller,
Characterisation by X-ray micro-tomography of cavity coalescence during superplastic
deformation, Scripta Mater., 42 (2000) 375-381.
[80] A. Borbély, F.F. Csikor, S. Zabler, P. Cloetens, H. Biermann, Three-dimensional
characterization of the microstructure of a metal–matrix composite by
holotomography, Mat. Sci. Eng., A 367 (2004) 40–50.
[81] G.C. Requena, P. Degischer, E.D. Marks, E. Boller, Microtomographic study of the
evolution of microstructure during creep of an AlSi12CuMgNi alloy reinforced with
Al2O3 short fibres, Mat. Sci. Eng., A 487 (2008) 99-107.
[82] T. Link, S. Zabler, A. Epishin, A. Haibel, M. Bansal, X. Thibault, Synchrotron
tomography of porosity in single-crystal nickel-base superalloys, Mat. Sci. Eng., A
425 (2006) 47-54.
[83] A. Isaac, F. Sket, W. Reimers, B. Camin, G. Sauthoff, A.R. Pyzalla, In situ 3D
quantification of the evolution of creep cavity size, shape, and spatial orientation using
synchrotron X-ray tomography, Mat. Sci. Eng. A 478 (1-2) (2008) 108-118.
[84] J. Hsieh, Computed Tomography: Principles, Design, Artifacts and Recent Advances,
SPIE Press, Bellingham (2003).
[85] D. Bernard, D. Gendron, J.-M. Heintz, S. Bordère, J. Etourneau, First direct 3D
visualisation of microstructural evolutions during sintering through X-ray computed
microtomography, Acta Mater. 53 (1) (2005) 121–128.
[86] K.E. Thompson, C.S. Wilson, W. Zhang, Quantitative computer reconstruction of
particulate materials from microtomography images, Powder Technol. 163 (3) (2006)
169–182.
[87] A. Pyzalla, B. Camin, T. Buslaps, M. Di Michiel, H. Kaminski, A. Kottar, A. Pernack,
W. Reimers, Simultaneous Tomography and Diffraction Analysis of Creep Damage,
Science 308 (2005) 92-95.
[88] J. Orr, F.R. Beckitt, A. Met, G.D. Fawkes, The physical metallurgy of chromium–
molybdenum steels for fast reactor boilers, Proc. of BNES (British Nuclear Energy
VII. References 100
Society) Conference on Ferritic Steels for Fast Reactor Steam Generator, London
(1977).
[89] H.R. Tipler, B.E. Hopkins, The creep cavitation of commercial and high purity Cr-
Mo-V steels, Met. Sci. 10 (2) (1976) 47-56.
[90] N.G. Needham, T. Gladman, Intergranular cavity damage and creep fracture of 1Cr-
0.5Mo steels, Mater. Sci. Tech. 2 (4) (1986) 368-373.
[91] D. Lonsdale, P.E.J. Flewitt, Damage accumulation and microstructural changes
occurring during creep of a 2¼ % Cr-1% Mo steel, Mater. Sci. Eng. 39 (1979) 217-
229.
[92] V.K. Sikka, C.T. Ward, K.C. Thomas, Modified 9Cr−1Mo steel — an improved alloy
for steam generator application. In: A.K. Khare, Editor, Ferritic steels for high
temperature applications, Proc. of ASM International Conference on Production,
Fabrication, Properties and Application of Ferritic Steels for High Temperature
Applications, Warren, PA, 6–8 October 1981, ASM, Metals Park (OH) (1983) pp. 65–
84.
[93] M. Staubli, W. Bendick, J. Orr, F. Deshayes, Ch. Henry. In: J. Lecomte-Beckers, F.
Schubert, P.J. Ennis, Editors, Materials for advanced power engineering. FZ Jülich
(1998).
[94] G. Eggeler, J.C. Earthman, N. Nilsvang, B. Ilschner, A Microstructural Study of Creep
Rupture in a 12% Cr Steel, Acta Metall 37 (1) (1989) 49-60.
[95] J.S. Lee, H.G. Armaki, K. Maruyama, T. Muraki, H. Asahi, Causes of breakdown of
creep strength in 9Cr–1.8W–0.5Mo–VNb steel, Mater. Sci. Eng. A 428 (2006) 270-
275.
[96] K.S. Min, J.K. Kim, N. Nam, Investigation of the effect of the types and densities of
grain boundary carbides on grain boundary cavitation resistance of AISI 321 stainless
steel under creep–fatigue interaction, J. Alloys Compd. 370 (2004) 223-229.
[97] K.J. Kim, H.U. Hong, K.S. Min, S.W. Nam, Correlation between the carbide
morphology and cavity nucleation in an austenitic stainless steels under creep-fatigue,
Mater. Sci. Eng. A 387-389 (2004) 531-535.
[98] P.J. Ennis, A. Czyrska-Filemonowicz, Recent advances in creep resistant steels for
power plant applications, Operation Maintenance and Material Issues (OMMI) 1
(2002) 1-28.
[99] G.H. Edward, M.F. Ashby, Intergranular fracture during power-law creep, Acta
Metall. 27 (1979) 1505-1518.
VII. References 101
[100] B.F. Dyson, Constraints on diffusional cavity growth rates, Met. Sci., 10 (1976) 349-
353.
[101] J.R. Rice, Constraints on the diffusive cavitation of isolated grain boundary facets in
creeping polycrystals, Acta Metall. 29 (4) (1981) 675–681.
[102] V. Tvergaard, On the creep constrained diffusive cavitation of grain boundary facets,
J. Mech. Phys. Solids 32 (5) (1984) 373–393.
[103] W. Beere, M.V. Speight, Creep cavitation by vacancy diffusion in plastically
deforming solid, Metal. Sci. 12 (4) (1978) 172–176.
[104] T.J. Delph, Some selected topics in creep cavitation, Metall. Mater. Trans. A 33A
(2002) 383-390.
[105] H.-T. Yao, F.-Z. Xuan, Z. Wang, S.-T. Tu, A review of creep analysis and design
under multi-axial stress states, Nucl. Eng. Des. 273 (2007) 1969-1986.
[106] J. Cadek, Creep in Metallic Materials, Elsevier, Amsterdam (1988).
[107] J. Weertman, Hull-Rimmer grain boundary void growth theory—a correction, Scr.
Metall. 7 (10) (1973) 1129–1130.
[108] R. Raj, H.-M. Shih, H.H. Johnson, Correction to: ‘Intergranular fracture at elevated
temperature’, Scr. Metall. 11 (10) (1977) 839–842.
[109] M.V. Speight, W. Beere, Vacancy potential and void growth on grain boundaries, Met.
Sci., 9 (1975) 190-191.
[110] W. Pavinich, R. Raj, Fracture at elevated temperature, Metall. Trans. 8A (12) (1977),
1917–1933.
[111] R. Raj, Intergranular fracture in bicrystals, Acta Metall. 26 (2) (1978) 341–349.
[112] L.-E. Svensson, G.L. Dunlop, Growth mechanism of intergranular creep cavities in
alpha-brass, Met. Sci. 16 (1) (1982) 57–64.
[113] H.C. Cho, J. Yu, I.S. Park, Creep cavitation in a NiCr steel, Metall. Trans. 23A (1)
(1992) 201–210.
[114] T.-J. Chuang, J.R. Rice, The shape of intergranular creep cracks growing by surface
diffusion, Acta Metall. 21 (12) (1973) 1625–1628.
[115] T.G. Nieh, W.D. Nix, The formation of water vapor bubbles in copper and their effect
on intergranular creep fracture, Acta Metall. 28 (5) (1980) 557–566.
[116] S.E. Stanzl, A. S. Argon, E.K. Tschegg, Diffusive intergranular cavity growth in creep
in tension and torsion, Acta Metall., 31 (1983) 833-843.
[117] B.F. Dyson, Constrained cavity growth, its use in quantifying recent creep fracture
results, Can. Metall. Q., 18 (1979) 31-38.
VII. References 102
[118] N.G. Needham, Cavitation and Fracture in Creep Resisting Steels, In: Report EUR
8121 EN Commission of the European Communities, Brussels, Contract No.
7210MA/802, (1983).
[119] H.J. Frost, M.F. Ashby, In: Fundamental Aspects of Structural Alloy Design, R.I.
Jaffee and B.A. Wilcox, Eds., Plenum Press, New York and London, (1977) pp.27-58.
[120] B.J. Cane, Mechanistic control regimes for intergranular cavity growth in 2.25 Cr-1
Mo steel under various stresses and stress states, Metal. Sci. 15 (7) (1981) 302–310.
[121] J.W. Hancock, Creep cavitation without a vacancy flux, Met. Sci., 10 (1976) 319-325.
[122] X-W. Du, J. Zhu, Y-W. Kim, Microstructural characterization of creep cavitation in a
fully-lamellar TiAl alloy, Intermet., 9 (2) (2001) 137-146.
[123] G. Dimmler, W. Schalk, P. Weinert, H. Gruber, S. Wieser, H. Cerjak, Sample
preparation for high-resolution damage examinations in the SEM, Prakt. Metallogr.,
39 (12) (2002) 619-633.
[124] J. Lapin, M. Nazmy, Microstructure and creep properties of a cast intermetallic Ti–
46Al–2W–0.5Si alloy for gas turbine applications, Mater. Sci. Eng. A, 380 (1) (2004)
298-307.
[125] L.M. Kachanov, Time of the rupture process under creep conditions. Izv. Akad. Nauk.
SSR Otd., Teckh. Nauk , 8 (1958) pp. 26–31.
[126] Y.N. Robotnov, Creep Problems in Structural Members (English translation edited by
F.A. Leckie). North Holland, Amsterdam (1969).
[127] D.R. Hayhurst, Creep Rupture under Multi-Axial Stress, J. Mech. Phys. Solids, 20
(1972) 381–390.
[128] D.R. Hayhurst, F.A. Leckie, The effect of creep constitutive and damage relationships
upon rupture time of solid circular torsion bar, J. Mech. Phys. Solids, 21 (6) (1973)
431–446.
[129] D.R. Hayhurst, F.A. Leckie, C.J. Morrison, Creep rupture of notched bars, Proc. R.
Soc. London, A 360 (1978) 243–264.
[130] D.R. Hayhurst, F.A. Leckie, Behaviour of materials at high temperatures. In: J.
Carlsson and N.G. Ohlson, Editors, Mechanical Behaviour of Materials. Proc. of
ICM4 vol. 2, Pergamon Press, Oxford (1984) pp. 1195–1212.
[131] D.R. Hayhurst, P.R. Dimmer, C.J. Morrison, Development of Continuum Damage in
Creep Rupture of Notched Bars, Phil. Trans. R. Soc. London, A 311 (1984) 103–129.
[132] P.W. Davies, K.R. Williams, B. Wilshire, On the distribution of cavities during creep,
Phil. Mag., 18 (1968) 197–200.
VII. References 103
[133] R.T. Chen, J.R. Weertman, Grain boundary cavitation in internally oxidized copper,
Mater. Sci. Eng., 64 (1984) 15–25.
[134] K.S. Yu, W.D. Nix, The effects of pre-existing grain boundary cavities on the creep
rupture life of a copper-antimony alloy, Scripta Metall., 18 (1984) 173–178.
[135] W.D. Nix, J.C. Earthman, G. Eggeler, B. Ilschner, The principal facet stress as a
parameter for predicting creep rupture under multiaxial stresses, Acta Metall., 34 (4)
(1989), 1067–1077.
[136] P.M. Anderson, J.R. Rice, Constrained creep cavitation of grain boundary facets, Acta
Metall., 33 (3) (1985) 409–422.
[137] M. Sakane, T. Hosokawa, Biaxial and triaxial creep testing of type 304 stainless steel
at 923 K. In: S. Murakami, N. Ohno, Editors, IUTAM Symposium on creep in
structures, Kluwer, Dordrecht, (2001) pp. 411–418.
[138] M. Sakane, H. Tokura, Experimental study of biaxial creep damage for type 304
stainless steel, Int. J. of Damage Mech., 11 (2002) 247–262.
[139] J. Henderson, An investigation of multi-axial creep characteristics of metals, ASME
Trans. 101 (1979) 356-364.
[140] K. Holler, B. Reetz, K.B. Müller, A. Pyzalla, W. Reimers, Microstructure and
properties of hot extruded brass CuZn40Pb2, Mat. Sci. Forum 426-432 (2003) 3667-
3672.
[141] M.R. Willis, J.P. Jones, Creep mechanisms in a dual-phase brass, Scripta Mater. 44
(2001) 31–36.
[142] www.schroederglas.com
[143] J. Radon, Über die Bestimmung von Funktionen durch ihre Integralwerte längs
gewisser Mannigfaltigkeiten, Ber. Verh. Sächs. Akad. Wiss. Leipzig, Math. Nat. kl., 69
(1917) 262-277.
[144] G.T. Herman, Image Reconstruction from Projections: The Fundamentals of
Computerized Tomography, Academic, New York (1980).
[145] A.C. Kak, M. Slaney, Principles of computerized tomographic imaging, IEEE Press,
(1987).
[146] J. Hsieh, Computed tomography, Principles, design, artifacts, and recent advances,
SPIE Press, (2003).
[147] A. Bedford, W. Fowler, Engineering Mecanics Static, second ed., Addison-Wesley
(1995).
VII. References 104
[148] R. Karnesky, C. Subrack, D. Seidman, Best-fit ellipsoids of atom-probe tomographic
data to study coalescence of γ’ (L12) precipitates in Ni-Al-Cr, Scripta Materialia 57
(2007) 353-356.
[149] A. Borbély, F.F. Csikor, S. Zabler, P. Cloetens, H. Biermann, Three-dimensional
characterization of the microstructure of a metal–matrix composite by
holotomography, Mater. Sci. Eng. A 367 (2004) 40-50.
[150] S. Endhoh, In: H. Masuda, K. Higashitani, H. Yoshida (Eds.), Powder Technology
Handbook, Florida, CRC Press; 2006.
[151] L.G. Brown, A survey of image registration techniques, ACM Comput. Surv. 24
(1992) 325–376.
[152] H. Foroosh, J. Zerubia, M. Berthod, Extension of phase correlation to subpixel
registration, IEEE Trans. Image Process. 11 (3) (2002) 188-200.
[153] K. Dzieciol, A. Isaac, F. Sket, A. Borbély, A.R. Pyzalla, Application of correlation
techniques to creep damage studies, Collected Proceedings TMS Conference, San
Francisco 15-19 February (2009), Characterization of Minerals, Metals and Materials,
pp. 15-22.
[154] L. Babout, E. Maire, J.Y. Buffière, R. Fougères, Characterization by x-ray computed
tomography of decohesion, porosity growth and coalescence in model metal matrix
composite, Acta Mater. 49 (2001) 2055-2063.
[155] A. Bodnar, M. Chrzanowski, K. Nowak, P. Latus, Influence of small variations of
initial defects upon crack paths in creeping plates – Continuum Damage Mechanics
description, Eng. Fract. Mech. 75 (2008) 526–533.
[156] A. Isaac, F. Sket, A. Borbély, G. Sauthoff, A.R. Pyzalla, Study of cavity evolution
during creep by synchrotron microtomography using a volume correlation method,
Practical Metallography 45 (5) (2008) 242-245.
[157] H.P. Leeuwen, The application of fracture mechanics to creep crack growth, Eng.
Fract. Mech. 9 (1977) 951–974.
[158] G.W. Greenwood, Creep and cavity growth under multiaxial stresses, Phil. Mag. A 43
(2) (1981) 281-290.
[159] A. Isaac, In-situ tomographic investigation of creep cavity evolution in brass, PhD
Thesis, Ruhr-Universität Bochum (RUB), Fakultät für Machinenbau (2009).
VIII. CV 105
Curriculum Vitae
PERSONAL INFORMATION
Last Name SKET
First Name FEDERICO
Nationality Austrian
Date and place of birth 11th of June 1979, Cruz del Eje, Argentina
EDUCATION AND TRAINING
Since March 2006 Max-Planck Institute for Iron research, Düsseldorf
(Germany) Department of Material Diagnostics and Steel Technology, PhD studies on Tomographic characterization of creep damage.
December 2004 to November 2005
Vienna University of Technology, Vienna (Austria), Institute of Materials Science and Technology, Diploma Thesis on Characterization of corrosion resistance of functionally graded hardmetals.
March 1998 to February 2006 National University of Comahue, Neuquén (Argentina), Faculty of Engineering, Electrical Engineering Degree.