LOW CYCLE FATIGUE CRACK PROPAGATION

IN STEEL UNDER VARIOUS CYCLIC LOADING

HISTORIES

By

M. Arief Rahman Panjaitan

A dissertation submitted to Nagoya University

in partial fulfillment of the requirements for

the Degree of Doctor of Engineering

Department of Civil Engineering

Graduate School of Engineering, Nagoya University

Nagoya, Japan
 ABSTRACT

Low cycle fatigue crack is a crack which propagates by the repetition of large

plastic strains. According to the researches on low cycle fatigue, the greatest of total life

is consumed in crack propagation phase and only small part of the total life is spent for

crack initiation, so that crack propagation manner in low cycle field should be carefully

determined. The ability to estimate low cycle fatigue (LCF) crack propagation in steel is

an essential aspect for assessment of the LCF lifetime or maintenance intervals. Due to

the importance of LCF crack investigation, researches in LCF propagation have been

more sophisticatedly conducted and revealed that cyclic J-integral range, ΔJ, is a

parameter that correlates to the crack growth rate and is applicable to describe its manner.

This dissertation evaluates crack propagation behavior under variable amplitude loading

through employing several LCF tests, determines parameters that correspond to the crack

growth rate and develops an estimation model to evaluate crack propagation.

In the first part of the work, LCF crack propagation under constant amplitude

loadings was experimentally investigated. Since it was already found that ΔJ is a

parameter governing the crack growth rate, the ΔJ was also resulted from FE analysis in

this study. To ensure the accuracy of estimation model, additional LCF test data from a

past study were also considered and collectively plotted with the current investigation on

the LCF. Both results indicated that the crack growth rates in the LCF correlate with the

ΔJ and were also distributed in the same region within narrow band. According to the

results, a formula for the LCF crack growth rate under constant amplitude loadings was

proposed.

As the second part of the work, evaluation was conducted under two-steps variable

amplitude conditions to simply reveal crack propagation behavior when the amplitudes

I
were changed. Accordingly, several test cases with the same value of high amplitude and

different values of low amplitude loading were employed. The results presented that crack

growth rate under high amplitude performed equivalent trend to the crack growth under

constant amplitude. Contrarily, under low amplitude loading, the majority of the crack

growth rates were noticeably to be present above the regression curve of constant

amplitude conditions indicating higher crack extension rates. To investigate accelerated

crack growth at low amplitude loading, a correction factor, which is the ratio of the crack

growth rate under variable conditions to that under constant amplitude loading was

introduced and verified to be correlated with the crack propagation manner.

At the last part, the study examines crack growth rate under random variable

amplitude loadings where all amplitudes were un-repetitive to present fully random

attributes. The difference on the COD magnitude and COD average were also designated

in these test cases. The result clarified the same crack growth manner to the two-steps

variable cases. A higher crack growth rates than those under constant amplitude loading

was identified when the amplitude was decreased from high to low levels. In contrast, the

indistinguishable manner on the crack growth rates to those under constant amplitude

were figured out when the amplitudes were escalated from low to high levels. At the final

work, the estimation model obtained from two-steps variable was then evaluated to the

random loading and resulted the good interpretation due to crack propagation manner.

Those findings confirmed that the proposed formula is applicable to the variable

amplitude loading.

II
 ACKNOWLEDGEMENT

First and foremost, praises and thanks to the God, Allah SWT, the Almighty,

for His showers of blessings throughout my research work to complete the research

successfully.

I would like to express my deep gratitude to my research supervisor, Professor

Kazuo Tateishi for providing me invaluable guidance throughout this research. He

has taught me the methodology to carry out the research and to present the research

works as clearly as possible. It was a great privilege and honor to work and study

under his guidance.

I also extend my sincere appreciation to Prof. Junji Kato and Dr. Takeshi Hanji of

Nagoya University for giving valuable advice in the seminar and dissertation. I am also

grateful to Prof. Kengo Anami of Shibaura Institute of Technology for providing

constructive comments and his valuable time in reviewing my dissertation.

I am also thankful to Dr. Masaru Shimizu for the many valuable helps, advice and

enjoyable discussions. Many thanks go to Ministry of Education, Culture, Sports, Science

and Technology (MEXT) of Japan who provided me the scholarship to support my study

in Nagoya University.

I am very much thankful to my wife, Ayu Raina Mufida and my daughters,

Qahirah and Husna, for their love, understanding, prayers and continuing support to

complete this research work. Also I express my thanks to my parents, sister, and

brother, for their support and encouragement.

III
 Many thanks also go to all my friends and colleagues, for their knowledge sharing,

helped me to deal with the administration in Japan and made my life so happy by their

warm friendship during those past years.

Finally, my thanks go to all the people who have supported me to complete the

research work and study in Nagoya University.

IV
Table of contents

ABSTRACT ...................................................................................................................... I

ACKNOWLEDGEMENT .............................................................................................. III

Table of contents ..............................................................................................................V

List of figures.................................................................................................................. IX

List of tables .................................................................................................................XIII

1. Introduction .............................................................................................................. 1

1.1. Background ........................................................................................................ 1

1.2. Previous studies ................................................................................................. 3

1.2.1. Low cycle fatigue (LCF) ............................................................................ 9

1.2.2. Crack growth rate evaluated by EPFM..................................................... 13

1.2.3. Crack closure ............................................................................................ 14

1.2.4. J-integral ................................................................................................... 17

1.2.5. Cyclic J-integral range.............................................................................. 21

1.2.6. Path independence of ΔJ .......................................................................... 22

1.3. Research objectives .......................................................................................... 23

1.4. Organization of dissertation contents............................................................... 23

2. Basic Methodology of the Study ............................................................................ 27

V
 2.1. Introduction ...................................................................................................... 27

2.2. Specimen .......................................................................................................... 27

2.3. Crack propagation measurement...................................................................... 28

2.4. FE modeling ..................................................................................................... 30

2.5. Summary .......................................................................................................... 36

3. Constant Amplitude Loading (CA) ........................................................................ 39

3.1. Introduction ...................................................................................................... 39

3.2. Previous study on the constant amplitude loading ........................................... 39

3.3. Present study on constant amplitude loading ................................................... 42

3.4. Summary .......................................................................................................... 47

4. Two-steps Variable Amplitude Loading (VA) ....................................................... 49

4.1. Introduction ...................................................................................................... 49

4.2. Load cases ........................................................................................................ 49

4.3. Load-COD relationship .................................................................................... 50

4.4. Crack propagation behavior ............................................................................. 52

4.5. Amplification factor (AF) ................................................................................ 55

4.6. Estimation model on the AF ............................................................................ 58

4.7. Validation on the estimation to the test results crack growth rate ................... 62

4.8. Crack propagation analysis .............................................................................. 64

4.9. Summary .......................................................................................................... 65

5. Random Variable Amplitude Loading (RA) .......................................................... 67

5.1. Introduction ...................................................................................................... 67

VI
 5.2. Load cases ........................................................................................................ 67

5.3. Test result arranged with ΔJ ............................................................................ 70

5.4. Summary .......................................................................................................... 77

6. Summary and Conclusions ..................................................................................... 79

6.1. Conclusion ....................................................................................................... 79

6.2. Recommendation for future study ................................................................... 81

References ...................................................................................................................... 83

APPENDIX A: Digital image of the crack ..................................................................... 89

APPENDIX B: Crack propagation analysis ................................................................... 91

VII
VIII
List of figures

Fig. 1.1 Low cycle fatigue damages in steel bridge piers in the Hyogo-ken Nanbu

earthquake ......................................................................................................................... 2

Fig. 1.2 Illustration of crack closure (Elber, 1971) .......................................................... 4

Fig. 1.3 Plasticity area around the crack tip and the plastic envelope formed in the

beginning of a crack propagation (Elber, 1971) ............................................................... 4

Fig. 1.4 Domain of the effective stress intensity which was introduced by Elber (1971) 5

Fig. 1.5 Variability of the effective stress intensity affected by the crack closure due to

the difference in applied stress range (Elber, 1971) ......................................................... 5

Fig. 1.6 Assignment of the tensile overload in steady cycles ........................................... 6

Fig. 1.7 Retardation in crack propagation due to overload employment in tensile phase 6

Fig. 1.8 Displacement normal to the crack plane (uy) at minimum load .......................... 8

Fig. 1.9 Number of cycles range for low cycle fatigue (Agrawal et al., 2014) .............. 10

Fig. 1.10 Strain-life relationship in ELCF and LCF regimes for hot-rolled carbon steel

(Gardner and Davies, 2010)............................................................................................ 10

Fig. 1.11 Idealized elastoplastic stress-strain behavior (Kim, 2015).............................. 12

Fig. 1.12 Hardening model in material behavior (Kim, 2015) ....................................... 12

Fig. 1.13 The three loading modes of a crack (Dahlin and Olsson, 2005) ..................... 13

Fig. 1.14 Fatigue crack closure mechanisms in steels (Anderson, 2011) ....................... 16

Fig. 1.15 J closed contour in a two-dimensional solid (Anderson, 2011) ...................... 18

Fig. 1.16 Two arbitrary contours Γ1 and Γ3 surrounding the crack tip. If such contours are

IX
linked by Γ2 and Γ4, a specific contour is presented (Anderson, 2011) .......................... 20

Fig. 1.17 Stress-strain curve used to calculate ΔJ (Par Ljustell, 2013) .......................... 20

Fig. 1.18 Flow chart of the research ............................................................................... 24

Fig. 2.1 Specimens (unit: mm) ....................................................................................... 28

Fig. 2.2 Experimental program and the observation ..................................................... 29

Fig. 2.3 True stress-strain relationship established with the principle of kinematic

hardening rule ................................................................................................................. 30

Fig. 2.4 Assignment of cyclic loading to the 3D-specimen model and the boundary

condition of the model .................................................................................................... 31

Fig. 2.5 Fluctuation range where ΔJ was defined ......................................................... 33

Fig. 2.6 Integration path used for independency verification (unit: mm) ...................... 33

Fig. 2.7 The ΔJ independency to integration path .......................................................... 34

Fig. 2.8 Loading cycles used for independency verification .......................................... 34

Fig. 2.9 The ΔJ independency to loading cycles ............................................................ 35

Fig. 2.10 Integration path for analysis (unit: mm) ......................................................... 35

Fig. 2.11 ΔJ-crack length relationships .......................................................................... 36

Fig. 3.1 Location of weld metal on the CT specimen..................................................... 40

Fig. 3.2 Crack growth rate and ΔJ under constant amplitude loading at the past study. 41

Fig. 3.3 Load-COD hysteresis loop ................................................................................ 42

Fig. 3.4 Load-COD verification ..................................................................................... 43

Fig. 3.5 Crack propagation behavior and ΔJ distribution under constant amplitude

loadings ........................................................................................................................... 44

Fig. 3.6 ΔJ-crack growth rate relationships under constant amplitude loadings ............ 45

Fig. 3.7 Crack length-number of cycles relationships under constant amplitude loadings

........................................................................................................................................ 46

Fig. 4.1 Loading patterns of two-steps variable amplitude loadings .............................. 50

Fig. 4.2 Load-COD hysteresis loop under VA-1 ............................................................ 51

X
Fig. 4.3 Crack growth rate behavior ............................................................................... 52

Fig. 4.4 Crack growth rate-crack length relationships.................................................... 52

Fig. 4.5 ΔJ-crack growth rate relationships under constant and two-steps variable

amplitude loadings .......................................................................................................... 54

Fig. 4.6 AF-crack length relationships............................................................................ 56

Fig. 4.7 AF-ΔJ ratio relationships under two-steps variable amplitude loadings ........... 57

Fig. 4.8 AF-ΔJ relationships under two-steps variable amplitude loadings ................... 59

Fig. 4.9 Cumulative plots of estimation and test results on AF-ΔJ ratio relationships .. 61

Fig. 4.10 Regression process to obtain estimated AF at 0.95 ≤ ΔJi/ΔJi-1 < 1 ................ 62

Fig. 4.11 Estimated AF at the ΔJi/ΔJi-1 < 1 .................................................................... 63

Fig. 4.12 Estimated AF at the 0.95 ≤ ΔJi/ΔJi-1 < 1......................................................... 63

Fig. 4.13 Verification on the estimation model to the test result crack growth rate (unit:

mm/cycle) ....................................................................................................................... 64

Fig. 4.14 Crack propagation analysis results .................................................................. 65

Fig. 5.1 Loading patterns of random variable amplitude loading .................................. 68

Fig. 5.2 Loading process used to define the ΔJ for RA-3............................................... 69

Fig. 5.3 The ΔJ for RA-3 defined based on constant amplitude principle ..................... 69

Fig. 5.4 Crack growth rate-crack length relationships under CA-1 and random amplitude

loadings ........................................................................................................................... 71

Fig. 5.5 ΔJ-crack growth rate relationships under constant and random amplitude loadings

........................................................................................................................................ 72

Fig. 5.6 AF-ΔJ ratio relationships under random amplitude loadings ............................ 73

Fig. 5.7 AF-ΔJ relationships under random amplitude loadings .................................... 74

Fig. 5.8 Verification of estimated crack growth rate to the test result (unit: mm/cycle) 75

Fig. 5.9 Crack propagation analysis results .................................................................... 76

XI
XII
List of tables

Table 2.1 Mechanical properties .................................................................................... 27

Table 3.1 Test matrix for past research work ................................................................. 40

XIII
XIV
1. Introduction

1.1. Background

Low cycle fatigue is a failure mode exhibiting a small number of large plastic

strain repetitions to failure. One type of LCF damages is fracture identified on steel bridge

components after Hyogo-ken Nanbu earthquake hit Japan in 1996 (Miki, 1997) as shown

in Fig 1.1. This figure presents LCF failures observed in welded joints in steel piers, i.e.,

beam-column connection and triangular ribs on the base joint. Related to these failures,

several researchers (Sakano et al., 1998; Sakano et al., 2001; Chen et. al, 2007; Park et

al., 2015) carried out a series of LCF test on steel-beam column joint and pier base joint

and evaluated crack propagation behavior.

Until recently, LCF crack propagation has primarily been investigated under

constant amplitude loading (Terao et al., 2015 and Hanji et al., 2017) and evaluations on

the variable amplitude loading were mainly conducted in high cycle fatigue (HCF) as

conducted by Corbly and Packman (1973), Raghuvir et al. (1996), Trebules et al. (1973),

Mc Millan and Pelloux (1967), and Matthews et al. (1971). These researches

acknowledged high amplitude stress levels in uniform low amplitude stress conditions. In

such studies, crack growth rate was noticeably retarded when amplitude decreased from

high to low levels and accelerated when the amplitude rose from low to high levels.

However, actual seismic motion are characterized by amplitude changeability from high

to low or from low to high levels, and the effect of these alterations on crack growth in

1
 (a) Beam to column connection

(b) Base joints of steel bridge pier

Fig. 1.1 Low cycle fatigue damages in steel bridge piers in the Hyogo-ken Nanbu earthquake

LCF field has not yet been clarified.

To maximize the fatigue resistance of steel member and prevent the failure in steel

caused by LCF crack propagation due to seismic motion, it is important to reveal crack

2
propagation manner under variability of cyclic strain.

1.2. Previous studies

Principally, fatigue life is introduced as summation of crack initiation and crack

propagation life. Murakami et al. (1983) discovered that crack propagation phase of LCF

holds most the total life of the specimen during large cyclic strain employment than the

phase of the crack initiation. Correspondingly, crack propagation life should be accurately

estimated so that the function of the steel member can be maximized without the failure

on the structure. At the previous research work, crack growth rate was identified to be

linearly varied with the stress intensity factor range on log–log scales in the high cycle

fatigue region known as Paris-Erdogan model (1963). However, this approach is not

applicable to the LCF field, because large scale plastic deformation occurs around the

crack tip in this region.

As far, a number studies evaluated crack growth manner due to the abrupt change

in amplitude were confirmed in HCF. At the past, Elber (1971) ascertained that in case of

the load is in tensile, a crack will not open in the following cycle until a remarkably high

tensile load is employed to the specimen as indicated in Fig. 1.2. Evaluation on the sample

of aluminum-sheet revealed the almost equivalent opening/closing stress value were

indicated to be one half than the highest employed stress in R equal to 0 (R is the ratio of

minimum to maximum stress of cyclic loading). R equal to zero means the cyclic loading

fluctuates from zero to designated tensile stress.

Explanation of the mechanism on crack-closure is described. As depicted in Fig.

1.3, at the area around the crack tip, the existence of the plastic zone is indicated. The

figure described that when the crack propagation passes this plasticity area (the size is

developed to be wider together with crack length), such wake of growing crack will form

the deformed material envelope as illustrated in the left side of the figure

3
 Fig. 1.2 Illustration of crack closure (Elber, 1971)

Fig. 1.3 Plasticity area around the crack tip and the plastic envelope formed in the

beginning of a crack propagation (Elber, 1971)

In this envelope, the existence of residual tensile deformations are representation

of residual compressive stresses. In this process, residual tensile deformation triggers the

crack to be close until a remarkably high tensile loading was applied to the specimen,

eventhough in tensile portion. Study conducted by Elber (1971) by X-ray-diffraction

measurements also ascertained the presence of residual compressive stresses in the

beginning of a crack.

Elber also discovered that the crack propagation rate are more satisfyingly to be

related to an effective stress-intensity range, ΔKeff = Kmax - Kop, as shown in Fig. 1.4, since

a crack is open in the phase of the tensile load application, rather than the total ΔK which

has the range from minimum to the maximum applied stress. The significance in

4
 Fig. 1.4 Domain of the effective stress intensity which was introduced by Elber (1971)

Fig. 1.5 Variability of the effective stress intensity affected by the crack closure due to the

difference in applied stress range (Elber, 1971)

explaining evaluated load sequence effects in crack progress was then figured out from

the effective stress intensity principle.

Both crack retardation and acceleration effects may be accounted on the crack

closure as presented in Fig. 1.5. In figure, If A is defined as balanced closure stress

intensity associated to the cyclic load at the lower level. The stabilized stress intensity

addressed to the cyclic load at the higher level is indicated with B. In the process of

transition from low to high, at the high level, the ΔKeff is un-permanently to be higher

5
than B. Hence, at the high level, acceleration of propagation is obtained. Contrarily, much

less value in ΔKeff is indicated for the conversion from high to low level than the A-

stabilized value and contribute the retardation in crack. Such behaviors have also been

observed experimentally by Elber (1971).

Ling and Schijve (1992) and Shercliff and Fleck, (1990) undertook the tests to

evaluate crack propagation behavior when one or more overloads were put in steady

cycles as expressed in Fig. 1.6. These studies confirmed that crack growth rates of steady

cycles behind overload were less than constant cycles without overload as shown in Fig

1.7. However, the behavior of the crack extension observed from those study are still

Fig. 1.6 Assignment of the tensile overload in steady cycles

Fig. 1.7 Retardation in crack propagation due to overload employment in tensile phase

6
needed to be clarified in LCF domain since it is acknowledged that there is the difference

on the plasticity area surrounding the crack tip under LCF and HCF in which this

distinction may also lead to give the different crack growth manner.

Shahani et al. (2009) evaluated fatigue crack growth rate behavior under variable

stress ratio and proposed four prediction models that contained some parameters, i.e.,

cyclic J-integral, crack tip opening displacement, crack mouth opening displacement and

stress intensity factor, to estimate crack growth rate.

Gonzales-Herrera and Zapatero (2008) evaluated CT aluminum specimen model tri-

dimensionally with considering different specimen thickness and stress range. This study

discovered that crack closure and crack opening mechanism caused abrupt transition contour of

deformation at the external side of the crack tip on the specimen slice model as expressed in Fig 1.8.

This contour area was identified has similar size with other specimen regardless the thickness.

Relied on this study, in 2D model, this significant abruption cannot be identified.

Tateishi and Hanji (2004) introduced LCF strength curves for plain steel material,

deposit metal and heat affected zone (HAZ), which were based on local strain amplitude

at a cracking point. Accordingly, relationships between local strain amplitude and LCF

life were presented in this study.

At the past study, LCF crack behavior in steel under multiaxial loading was also

investigated (Itoh and Miyazaki, 2003). Examinations were done with developing

correlation between two LCF life values from experiment and analytical study.

Equivalent strain parameter resulted from analytical study was confirmed to give good

estimation indicated by fatigue life distributed within narrow band in relationship of

estimated-experimental fatigue life. Past studies (Tanaka et. al, 2006 and Tanaka et. al,

2003) also revealed crack propagation behavior from hole in tubular specimens subjected

to cyclic torsional loading and identified that cyclic J-integral was an appropriate

parameter to estimate crack growth. Dong et al. (2016) tested CT specimen made of Q345

steel under constant amplitude conditions and discovered linear relationship between

7
 Fig. 1.8 Displacement normal to the crack plane (uy) at minimum load

crack growth rate and the cyclic crack tip opening displacement (CTOD).

Former research works in low cycle fatigue also found that J-integral calculated

from the area under a load–displacement hysteresis loop, called the cyclic J-integral range,

8
correlates with the crack growth rate (Usami and Kumar, 1996 and Masatoshi, 2002).

Jono et al. (1993) carried out the load-controlled elasto-plastic fatigue crack growth tests

under variable amplitude loading condition. It was found that crack growth rate can be

estimated by cyclic J-integral range and also by fracture mechanics parameter Jmax which

is the maximum J-integral. Terao et al. (2015) tested CT specimen made of steel and

deposit metal under constant amplitude conditions and examined crack propagation rate.

Besides, the difference in average of diplacement was also considered. This study

discovered that crack growth rate correlates with cyclic J-integral range regardless the

specimen steel grades and the average of cyclic displacement. Hanji et al. (2017)

examined LCF crack propagation rate tendency under constant amplitude loading and

established estimation model by acknowledging cyclic J-integral range (ΔJ) parameter.

This study also considered ΔJ to reveal crack growth behavior in a corner welded joints.

At the past research work, Tanuma and Kobayashi (2002) also applied the cyclic J-

integral concept to crack growth estimation under extremely high-strain.

In this present research work, examination on crack growth under constant

amplitude conditions was firstly conducted to obtain rigorous formula to estimate crack

propagation. Next, crack growth rates under variable amplitude loading particularly when

the amplitude loadings were changed, were evaluated to provide prediction model and

then its applicability was verified.

1.2.1. Low cycle fatigue (LCF)

LCF crack is a typical crack which propagates at the high stress levels and a low

number of cycles to failure. In all fatigue cracks, the weakening material caused by cyclic

loading is started from the crack initiation and continued with the crack propagation. LCF

is addressed with the fatigue life consumed in the crack propagation phase. As expressed

in Fig. 1.9, at the stress range from zero to endurance strength, the occuring stress will

9
 Fig. 1.9 Number of cycles range for low cycle fatigue (Agrawal et al., 2014)

Fig. 1.10 Strain-life relationship in ELCF and LCF regimes for hot-rolled carbon steel

(Gardner and Davies, 2010)

not cause the fatigue failure on the steel component, meanwhile when the stress range

10
over than endurance strength, it is possible to give fatigue failure on steel components.

LCF is attributed with the number of cycles less than 105 cycles for its fatigue life,

meanwhile high cycle fatigue (HCF) addressed its fatigue life over than 105 cycles. LCF

could also be sub divided into:

 LCF for less than 105 cycles to failure.

 LCF for less than 100 cycles to failure.

The second part of the LCF at above classification is extremely low cycle fatigue

(ELCF). The discussions on the (ELCF) have been made at the several past studies

(Coffin, 1971; Shimada et al., 1987; Komotori and Shimizu; 1998; Chung and Abel,

1988). Fig. 1.10 presented the difference on the strain-fatigue life behavior between LCF

and ELCF. From the figure, the plastic strain-life data from LCF tests scatters

approximately on a straight line when plotted on a log-log scale. This observation

established the basis of the Coffin-Manson relationship (Coffin, 1954; Manson, 1965).

The figure also indicates that ELCF fatigue data do not obey the Manson-Coffin Curve,

where such curve was too over to estimate the fatigue life in ELCF. The failure

mechanisms are also different in LCF and ELCF. For example, the center of the

cylindrical bars in push-pull fatigue test is the common location of the fracture due to

ELCF test. Meanwhile, at the same type of the test, the crack often starts from the surface

under LCF test (Shimada et al., 1987; Kanvinde and Deierlein, 2005). Related to Manson-

Coffin’s law (strain-life relationships) application in ELCF, several previous studies

(Tateishi et al., 2007, Hatanaka and Fujimitsu, 1984, Kuroda 2002) were carried out to

control the disadvantage of formula.

1.2.2. Idealized stress-strain relationship

Generally, analytical work assumed material in idealized elastoplastic stress strain

behavior due to employment of the cyclic loading as shown in Fig. 1.11. At the

application of the tensile load, the stress increases with slope of E (elastic modulus) until

11
 Fig. 1.11 Idealized elastoplastic stress-strain behavior (Kim, 2015)

(a) Kinematic hardening model (b) Isotropic hardening model

Fig. 1.12 Hardening model in material behavior (Kim, 2015)

yield stress, σy (line o-a). When the load is continued, after yielding, the stress experiences

the plastic phase with slope Et (tangent modulus) (line a-b). Upon removing the load, the

stress perform elastic unloading with slope E (line b-c). Due to loading in the opposite

12
direction, the material will eventually yield in that direction (point c). At the strain

hardening phase, line a-b or c-d, the specimen requires more force to continuously deform

in plastic region.

The material behavior in hardening model is outlined in Fig. 1.12. Fig. 1.12(a)

performs kinematic hardening model. In such model, the elastic range remain constant,

where line b-c is equal to line d-e. The model also adheres the principle where the center

of the elastic region, symbolized with dash-dot line moves parallelly to the work

hardening line. The center of elastic domain is also utilized as an evolution variable of

the model.

Fig. 1.12(b) exhibits isotropic hardening model. The model considers that elastic

range (yield stress) proportionally increases to the plastic strain. Besides, it also

contemplates the yield stress for the reversed loading is equal to the previous yield stress.

Due to sustainability of cyclic loading, the plastic strain is used as an evolution variable

of the model.

1.2.2. Crack growth rate evaluated by EPFM

Crack growth rate can be simply defined as the difference of the crack length at

Fig. 1.13 The three loading modes of a crack (Dahlin and Olsson, 2005)

13
the peak amplitude of the current cycle to the crack length at the one cycle previously.

Under large cyclic strains, Elastic Plastic Fracture Mechanics (EPFM) has proven to be

an appropriate domain for the characterization of the fatigue process at the area

surrounding the crack tip. In EPFM, specimen responses such as stress, strain, and

deformation are assigned to evaluate crack growth rate. Such investigation is needed

since the details of the fatigue mechanism under large cyclic strains are largely unknown.

At the very low loading, EPFM may not be appropriate for fatigue crack growth analyses.

Such low loads are not considered in this present study.

Fatigue crack growth behavior was usually evaluated through three types of

mode, as expressed in Fig. 1.13. In most cases, mode 1 was selected to evaluate the crack

growth rate. Mode 2 and 3 were usually employed with the other modes to obtain crack

growth under combined loadings.

1.2.3. Crack closure

Crack closure is the specimen behavior where the crack surfaces experiences

contact due to cyclic loading application. The closure is identified when the external load

acting on the material. At the large cyclic loading, due to surfaces contact, the inflection

points where the sustainability of the hysteresis loop is changed, will be indicated in the

unloading process.

The crack closure can be simulated by the finite element method. At the 1970s,

the sophisticated development of the computer allows the possibilities of using finite

element models to evaluate the crack closure caused by plastic deformation at the crack

tip. A very simple and small models were applied at the earlier studies but recently more

complicated model can be examined due to rapid development of the computer capability.

At the past, a related paper on plasticity caused crack closure investigation by the finite

element analysis was carried out by Solanki et al. (2004). In such study, the crack is

modelled and cyclic load was employed to the component as the basic principle. By

14
freeing the crack tip node at the several stages in the load cycle, crack growth is

established. Consideration of a constitutive model of elastic plastic on the plastically

deformed material is firstly applied in tensile load and followed by the cyclic loading

employment. The crack closure is able to be evaluated when the first node behind the

crack tip was released where the crack face leaves contact (crack opening) and followed

by contact at the opposing crack surfaces. Several simulation were provided by several

researches (McClung and Sehitoglu, 1989; Solanki et al., 2004; and Gonzales and

Zapatero, 2005) and obtained the dependence of the crack closure influence relied on the

finite element size. Additionally, at the separate studies, the results were also influenced

by the choice of constitutive model (Jiang et al., 2005; Pommier and Bompard, 2000; and

Pommier, 2001).

Fig. 1.14(a) expresses compelling evidence for plasticity-induced closure from

the experiment was conducted by James and Knott (1985). In that study, they presented

crack from the test in a standard specimen and to eliminate a little part of material from

the surfaces of the crack approaching 0.5 mm from the tip, they utilized electron

machining. They located the specimen back to the machine for test, and identified that

accelerated crack propagation is indicated than the prior measurement. They asserted that

the plastic wake triggered the closure and decreasing the crack propagation. Contrarily,

acceleration in crack propagation is indicated since the residual stress in plastic area had

been eliminated by machining.

Fig. 1.14(b) outlines roughness-induced closure which is affected by the

microstructure. Relied on the observation, combined condition on the microscopic view

is possibly caused by the deflection form of crack in case of un-homogeneity of the

material, even though on a global scale, fatigue cracks propagate in pure Mode 1

conditions. Illustration about the typical modes are previously given in Fig. 1.13. When

the crack is subjected to Mode 2, the crack leads to give different pattern from the Mode

1. The closure form can also be shown in Fig. 1.14(b). The displacement-controlled

15
 (a) Closure caused by plasticity (b) Closure caused by roughness

(c) Closure caused by oxide (d) Viscous fluid caused closure

(e) Closure caused by transformed zone

Fig. 1.14 Fatigue crack closure mechanisms in steels (Anderson, 2011)

applied to the specimen causes un-proper condition in both crack surfaces, which yields

16
crack faces contact when the load in tensile phase. In case of coarse-grained materials,

the higher closure load in fatigue can be produced due to higher degree of surface

roughness (Gray et al., 1983).

An aggressive environment is usually corresponded with the oxide-induced

closure as shown in Fig. 1.14(c). In this closure, between crack faces, oxide debris or

other corrosion products become wedged. In many cases, the crack closure may also be

present by a viscous fluid, as presented in Fig. 1.14(d). As presented by the oxide

mechanism, between the crack surfaces, such fluid behaves as a block. Fig. 1.14(e)

outlines the closure caused by transformed zone that wakes up of process zone and causes

the closure. From the described typical closures, the microstructure, yield strength, and

environment are the determining factors that lead to cause the closure.

The crack wedging is noticeably present on the several of the closure process

outlined in Fig. 1.14 (closure caused by viscous fluid, roughness and product of

corrosion). The obstruction of some types is figured out as the reason due to the

prevention of the crack closure. In this mechanism, the term residual crack opening is

considered to be an appropriate approach to give the description. The closure caused by

transformed zone can be classified to be a real mechanism in crack closure since its crack

faces received the closure force from the residual stress.

1.2.4. J-integral

The J-integral is the parameter used to evaluate crack growth in the EPFM field.

By such definition, the parameter is capable to specify the crack tip level, although the

plasticity area is not limited to a wide zone if it is measured to the dimension of in-plane

specimen. In another condition, the plastic zone may not interact with any in-plane

dimension for the theory to be applicable.

17
 Fig. 1.15 J closed contour in a two-dimensional solid (Anderson, 2011)

Rice (1968) presented the J contour integral. Idealization to the non-linear elastic

behavior from the elastic plastic nonlinear condition was conducted in this study. Rice

figures out that J, is a path independent line integral and introduced as follows:

𝛿𝑢𝑖
𝐽 = ∫ (𝑊 𝛿𝑦 − 𝑇𝑖 𝑑𝑠) (1.1)
𝛤 𝛿𝑥
𝜀𝑖𝑗
(2.1)
𝑊 = ∫ 𝜎𝑖𝑗 . 𝛿𝜀𝑖𝑗 (1.2)
0

Where W is the strain energy density, Ti are the traction vector components, ui(2.2)
are the

displacement vector components in xi-axis, ds is the differential length along the contour

Γ surrounding the tip, δx and δy are differential length along the contour in x and y-axis,

σij and εij are stress and strain in the loading process, respectively.

Rice (1968) then employed the divergence theorem to convert Eq. (1.1) into area

integral as illustrated in Fig. 1.15, the equation is constructed as follows:

18
 𝛿𝑊 𝛿 𝛿𝑢𝑖
𝐽=∫ [ − (𝜎𝑖𝑗 )] 𝛿𝑥𝛿𝑦 (1.3)
𝐴 𝛿𝑥 𝛿𝑥𝑗 𝛿𝑥
(2.3)
in which A is the specific area for computing J, δxj is the differential coordinate scalar

component in xj-axis.

Employing the strain energy density characteristics, presented by Eq. (1.2), we

can change the first expression in brackets of Eq. (1.3) to be Eq. (1.4).

𝛿𝑊 𝛿𝑊 𝛿𝜀𝑖𝑗 𝛿𝜀𝑖𝑗
= . = 𝜎𝑖𝑗 ( ) (1.4)
𝛿𝑥 𝛿𝜀𝑖𝑗 𝛿𝑥 𝛿𝑥

Relationship between strain-displacement (for small strains) is applied and(2.4)

gives,
𝛿𝜀𝑖𝑗 1 𝛿 𝛿𝑢𝑖 𝛿 𝛿𝑢𝑗
= [ ( ) + ( )] (1.5)
𝛿𝑥 2 𝛿𝑥 𝛿𝑥𝑗 𝛿𝑥 𝛿𝑥𝑖

in which δxi is the differential coordinate scalar component in xi-axis and δx(2.4)
j is the

differential coordinate scalar component in xj-axis. The xi and xj-axis are provided in Fig.

1.15 to contruct definition of strain in respect of the traction.

Eq. (1.4) is applicable when elastic potential refers to W. Substituting correlation

of strain-displacement (Eq. (1.5) to Eq. (1.4)) gives,

𝛿𝑊 1 𝛿 𝛿𝑢𝑖 𝛿 𝛿𝑢𝑗
= 𝜎𝑖𝑗 [ ( ) + ( )] (1.6)
𝛿𝑥 2 𝛿𝑥 𝛿𝑥𝑗 𝛿𝑥 𝛿𝑥𝑖

Since σij is equal to σji, Eq. (1.6) is derived as follows: (2.6)

𝛿𝑊 𝛿 𝛿𝑢𝑖
= 𝜎𝑖𝑗 ( ) (1.7)
𝛿𝑥 𝛿𝑥𝑗 𝛿𝑥
Invoking the equilibrium condition: (2.7)
𝛿𝜎𝑖𝑗
=0 (1.8)
𝛿𝑥𝑗
Eq. (1.8) can be fulfilled, if σij is constant, so that:
(2.8)
𝛿 𝛿𝑢𝑖 𝛿 𝛿𝑢𝑖
𝜎𝑖𝑗 ( )= (𝜎𝑖𝑗 ) (1.9)
(2.9)
𝛿𝑥𝑗 𝛿𝑥 𝛿𝑥𝑗 𝛿𝑥
In Eq. (1.9), the term is equivalent to square brackets in second expression(2.6)
in Eq.
(2.9)
(1.3). Thus, Eq. (1.3) integrand vanishes and J = 0 for every closed contour.

19
Fig. 1.16 Two arbitrary contours Γ1 and Γ3 surrounding the crack tip. If such contours are

linked by Γ2 and Γ4, a specific contour is presented (Anderson, 2011)

Fig. 1.17 Stress-strain curve used to calculate ΔJ (Par Ljustell, 2013)

Around the tip, suppose two evaluated contours Γ1 and Γ3, as outlined in Fig. 1.16.

When Γ1 and Γ3 are associated by path along the surface of crack (Γ2 and Γ4), a closed

contour is presented. Accumulation from each path in closed contour is equal to total J

and written as follows:

20
 𝐽 = 𝐽1 + 𝐽2 + 𝐽3 + 𝐽4 = 0 (1.10)

On the crack face, the traction vector (Ti) and the differential length in vertical(2.10)
axis (δy)

are equal to zero, Ti = δy = 0. From these results, J-integrals at the particular path, J2 = J4

= 0 and J1 = −J3. Here, the equivalent J will be yielded at the any counterclockwise path

surrounding the tip. From this result, that J is path-independent can be proven from this

derivation.

1.2.5. Cyclic J-integral range

Lamba (1975) firstly introduced the definition of in connection to stress

concentration factor (SCF), where SCF is determined as the ratio of maximum stress to

the nominal stress (Bhandari, 2008). The definition is established since the stress

concentration is indicated at each employment of tensile loading and cyclic stress-strain

ranges are also characterized on the material in front of the crack tip. Accordingly, Δσij

and Δεij is used to result the ΔJ.

A curve of cyclic stress strain in which the initial value is presented with 1 in

numbering and final value is symbolized with 2 in numbering is expressed in Fig 1.17.

The range of the J-integral is introduced by Lamba (1975), Dowling and Begley (1976)

and Lambert et al. (1988) as follows:

𝛿∆𝑢𝑖
∆𝐽 = ∫ (𝑊 ′ 𝛿𝑦 − ∆𝑇𝑖 𝑑𝑠) (1.11)
𝛤 𝛿𝑥
2 2 (2.11)
𝜀𝑖𝑗 𝜀𝑖𝑗
𝑊′ = ∫ Δ𝜎𝑖𝑗 . 𝛿(∆𝜀𝑖𝑗 ) = ∫ (𝜎𝑖𝑗2 − 𝜎𝑖𝑗1 ). 𝛿(∆𝜀𝑖𝑗 ) (1.12)
(1.12)
1
𝜀𝑖𝑗 1
𝜀𝑖𝑗

(2.11)
In which, Γ is the integration path surrounding the tip, ΔTi and Δui are traction
(2.13)
and

displacement vector change from the initial point 1 to the final point 2. W’ is addressed

to the strain energy density.

21
 Generally, in computing process of ΔJ, the loading phase of the cyclic curve from

initial point 1 to point 2 is merely involved in the integration of the strain energy density

rather than the overall loop. In case of initial point 1 is positioned at zero stress and strain,

the ΔJ is equivalent to J. Eq. (1.3) is a result of Eq. (1.1) to involve the condition in which

the stress and strain at point 1 is not zero.

1.2.6. Path independence of ΔJ

In case of W’ exhibits the strain energy density, the stresses can be stated by

evaluating W’ with considering the strains:

𝑊′
Δ𝜎𝑖𝑗 = (1.13)
𝛿(∆𝜀𝑖𝑗 )

(2.14) ΔJ
Eq. (1.13) is required for evaluating the path independence of ΔJ. Evaluating

along a closed contour, Γ, as formerly shown in Fig. 1.15 and employing divergence

theorem to convert Eq. (1.13) into an area integral gives:

𝑊′ 𝛿 𝛿∆𝑢𝑗
∆𝐽 = ∫ [ − (𝛥𝜎𝑖𝑗 )] 𝛿𝑥𝛿𝑦 (1.14)
𝐴 𝛿𝑥 𝛿𝑥𝑗 𝛿𝑥
(2.14)Eq.
where A is the area defined by Γ. By assuming W’ displays the strain energy density,

(1.13), the first term of integrand can be written as:

𝑊′ 𝑊′ 𝛿(∆𝜀𝑖𝑗 ) 𝛿(∆𝜀𝑖𝑗 )
= = Δ𝜎𝑖𝑗 (1.15)
𝛿𝑥 𝛿(∆𝜀𝑖𝑗 ) 𝛿𝑥 𝛿𝑥

(2.15)
It can be shown that Eq. (1.15) is similiar to the second expression in the integrand

in Eq. (1.14) by employing the strain-displacement relationships for small strains. Hence,

for any closed contour, the ΔJ is equal to zero. By considering the contour previously

illustrated in Fig. 1.16, the path independence of ΔJ was evaluated surrounding the crack

tip and yielded ΔJ1 = −ΔJ3. .

In case of proportional loading, an individual variable of crack tip is expressed.

22
Hence, due to the existence of the proportional loading in the area around the tip, the ΔJ

specifically introduces the changes in stress and strain around this zone. Under the

monotonic loading, due to large plastic distribution or the large crack propagation, the J

integral discontinue to contribute a single-parameter of crack tip condition. Similarly,

there is only limited studies expecting ΔJ to specify crack propagation beyond an

undoubtable level of deformation. Accordingly, evaluations of ΔJ have to be conducted.

1.3. Research objectives

The purpose of this research is to investigate crack propagation manner in steel

due to the employment of plastic strains variability.

To achieve it, the basic characteristic of crack propagation under constant

amplitude was firstly revealed and then to obtain crack propagation behavior when the

amplitudes were reduced or increased, a simple variable amplitude loading, two-steps

variable condition was employed. Applicability of estimation model established from

two-steps variable was then evaluated to the random variable amplitude loading.

1.4. Organization of dissertation contents

The flowchart of this study is presented in Fig. 1.18. The content of this paper are

concisely described as follows.

Chapter 1 describes introduction of this study, including research background,

previous studies, the basic literature elaborating LCF characteristics, specimen behavior

due to applied cyclic loading, i.e., crack propagation and crack closure, and basic theory

of EPFM parameters used to evaluate crack propagation rate, i.e., cyclic J-integral and

cyclic J-integral range, research objectives and organization of the dissertation.

23
 Fig. 1.18 Flow chart of the research

Chapter 2 elaborates the way to carry out the research which consists of

explanation about the specimen characteristics, test set-up, the way to apply cyclic

loading, and to measure crack length. Besides, this chapter also discusses the method to

24
define the ΔJ, i.e., loading process and the path consideration to compute ΔJ. In this

chapter, the ΔJ independency due to integration path and loading cycle was also verified

before calculating ΔJ at the designated crack length.

Chapter 3 elaborates crack propagation behavior under LCF constant amplitude

loadings. In this chapter, firstly crack growth rates information from the past research

work (Hanji et al. 2017), the study which also examined crack growth rate under constant

amplitude loading was briefly explained. Prior to the discussion for variable amplitude

loadings, verification on the analysis to test was also done in this typical loading to

confirm whether the analysis can satisfyingly figure out the test. Next, the load-COD

behavior in entire hysteretic loop and crack growth rate-ΔJ relationship were evaluated.

To ensure the accuracy, crack growth rates from Hanji et al. (2017) were collectively

plotted with the crack growth rates of present LCF tests to yield estimation formula.

Chapter 4 performs LCF two-steps variable amplitude loading tests in steel. To

compare the behavior, two-steps variable’s crack growth rates were cumulatively plotted

with those under constant amplitude loading. In this chapter, a ratio of crack growth rate

of variable amplitude to that under constant amplitude loading, named as ‘amplification

factor’, was presented and identified to be correlated with crack growth rate. Due to the

uncomplicated attributes of two-steps cases, estimation model was constructed from these

typical loadings. At the end of the work, verifications were made with outlining the value

of crack propagation rate and the crack length originated from the analysis and the test.

Chapter 5 discusses crack growth rate behavior under LCF random variable

amplitude loadings. In this chapter, random loading’s crack growth rate manners were

also characterized to that under constant amplitude loading. In such cases, observations

were emphasized on the crack growth at the domain of amplitude change, i.e., from high

to low levels and from low to high levels. Since the estimation model had been acquired

from two-steps variable cases, such formula was also employed to the random variable

amplitude loading and then validated.

25
26
2. Basic Methodology of the Study

2.1. Introduction

Study methodology which involves the way to do experiment and analysis is

elaborated in this chapter. Experimental study describes specimen characteristics, test set

up configuration, the loading mode, typical of loading histories, technical measurement

of crack opening displacement (COD) and crack length. This chapter also discusses about

the way to perform analyses to compute the ΔJ since the crack growth rate was evaluated

through the ΔJ. On the analysis, explanation about model characteristics, crack growth

simulation on the model and the ΔJ independency due to loading cycle and path was

evaluated prior to the ΔJ computation which relied on the applied COD.

2.2. Specimen

The specimen tested in this experiment was a compact tension specimen shown

in Fig. 2.1. The configurations were arranged based on ASTM E1820-031 (2008). The

mechanical properties of tested steel obtained from mill sheet are shown in the Table 2.1.

Table 2.1 Mechanical properties

Steel Yield Tensile Elongation
type strength strength
(N/mm2) (N/mm2) (%)
SS400 296 431 32

27
 (a) Specimen (b) Detail of side groove

Fig. 2.1 Specimens (unit: mm)

The specimen was cut from a steel plate of SS400 and thickness of all specimens was

arranged in 12.5 mm. Side grooves were introduced at both side surfaces of the specimen

in order to guide a crack path as straight as possible.

2.3. Crack propagation measurement

A test machine, which is type of Shimadzu EHF-UB300KN-40N, was employed

to evaluate specimen behavior under LCF tests. The maximum loading capacity is 300

kN. Experimental set up is presented in Fig. 2.2(a). The specimen was fixed to a testing

machine and the cyclic displacement was applied. The cyclic loading was carried out with

controlling the crack opening displacement (COD). Three loadings of the COD were

applied to the specimens, i.e., constant amplitude, two-steps variable amplitude and

random variable amplitude loadings. Detail information on the loading sequence will be

given in each chapter later. Loading speed was approximately 0.001mm/sec to

0.05mm/sec.

The applied cyclic loading refers to the mode 1 as explained at the Section 1.2.2.

Due to the given load, at the employment of tensile loading, a crack occurs from the notch

28
point and propagates along the grooves.

The propagation was observed with a high-resolution microscope at each peak

tensile load. The microscope has the magnification of 50 to 100. The crack tip was

rigorously identified relied on digital image of crack as shown in Fig. 2.2(b). To measure

the crack opening displacement (COD), a clip gauge was mounted at the sharp projection

in the specimen notch as illustrated in Fig. 2.2(c). This clip gauge can exactly measure

maximum opening of 5.0 mm. The cyclic loadings were continued until the crack growth

exhibited insignificant rate. From the observation of the crack extension, crack growth

(a) Test set up (b) Digital image of crack

(c) Mounted position of the clip gauge

Fig. 2.2 Experimental program and the observation

29
rate is introduced as:
𝑑𝑎 𝑎𝑖 −𝑎𝑖−1
= (2.1)
𝑑𝑁 1𝑐𝑦𝑐𝑙𝑒
where da/dN is the crack growth rate (mm/cycle), ai is the crack length at an observed

cycle (mm), ai-1 is that at one cycle prior to the observed cycle (mm).

2.4. FE modeling

In this study, the specimen at the ligament was modeled by fine solid element

mesh with 1.0 mm x 0.5 mm x 0.625 mm. Solid element type is eight-node, isoparametric

and arbitrary hexahedral elements. Analysis was performed using Abaqus FEA (Ver.

6.13) to obtain ΔJ. A quarter model of the CT specimen was utilized to conduct the

analysis. This model was constructed with symmetric supports assigned at the thickness

center (XY-plane). The material properties of the model were assigned using the same

characteristics listed in Table 2.1 where kinematic hardening rule based on true stress-

strain was applied. Fig. 2.3 expressed the difference slope between E and Et (E and Et are

Fig. 2.3 True stress-strain relationship established with the principle of kinematic

hardening rule

30
Fig. 2.4 Assignment of cyclic loading to the 3D-specimen model and the boundary

condition of the model

the first slope initial elastic and the second slope of plastic phase, respectively). E and Et

was obtained from the mill sheet and can be calculated as follows:

31
 𝜎𝑦
𝐸= (2.2)
𝜀𝑦
𝜎𝑢 − 𝜎𝑦
𝐸𝑡 = (2.3)
𝜀𝑢 − 𝜀𝑦
where σy is the yield strength (MPa), εy is the yield strain (mm/mm), σu is the ultimate

strength (MPa), εu is the ultimate strain (mm/mm).

According to the true stress-strain relationship, E is 200,000 MPa and Et is 670.23

MPa. Those two moduli were assigned as the mechanical properties of the material.

Fig. 2.4 figures out cyclic loading application to the pin center and to ensure the

contact, rigid connection was established between pin and CT specimen. Abaqus standard

module was used to conduct the analysis until 1.5 cycles. At the 0-cycle, a ligament was

initially supported in the crack opening direction. Then, at the first 0.5 cycle-tensile

loading, the crack was modelled by releasing supports from the notch point to crack tip

position. Accordingly, several models with different crack length, 8 to 56 mm, were

provided. The crack length is fixed until 1.5 cycles in the analysis although the crack

propagates in every cycle during the test. The compressive load applied after tension

caused crack surface to be pressed and experiences contact. To ensure the closure, a rigid

contact element was established along the crack surface-ligament and placed right below

the model as shown in Fig. 2.4.

The value of ΔJ was calculated based on specimen responses at the final process

of cyclic loading as shown in Fig. 2.5. The ranges of each parameter shown in Eq. (1.11)

and (1.12) were taken as the difference from the minimum point to the maximum point,

during the loading process from 1.0 to 1.5 cycles as shown in figure. The figure also

describes the different loading portions to compute ΔJ. Fig. 2.5(a) expresses the loading

process from 0-COD to tensile-COD and Fig. 2.5(b) outlines the process from

compressive-COD to tensile-COD. The loading processes presented in such figures were

utilized to calculate ΔJ at the different averages of COD.

32
 a) 0-COD to tensile-COD b) Compressive-COD to tensile-COD

Fig. 2.5 Fluctuation range where ΔJ was defined

Fig. 2.6 Integration path used for independency verification (unit: mm)

To investigate the independency of ΔJ, several integration paths surrounding the

crack tip as shown in Fig. 2.6 were considered, prior to the ΔJ calculation. The

33
 Fig. 2.7 The ΔJ independency to integration path

Fig. 2.8 Loading cycles used for independency verification

verification was done for 0.5 mm-COD at 16 mm-long crack. For overall analyses,

loading portion of 1.0 to 1.5 cycles was chosen. Fig. 2.7 provides the ΔJ independency in

the integration paths where negligible difference of ΔJ value is obtained.

To figure out independency of ΔJ to loading cycles, at at 6th-path number, several

patterns of the loading were considered, i.e., 1.0 to 1.5-loading cycle, 2.0 to 2.5-loading

cycle, etc. The detail information about considered loading cycles is provided in Fig. 2.8.

34
 Fig. 2.9 The ΔJ independency to loading cycles

Fig. 2.10 Integration path for analysis (unit: mm)

Similar to the path independency clarification, the analysis was also conducted for 0.5

mm-COD at 16 mm-long crack. As shown in Fig. 2.9, un-obvious differences of ΔJ value

are confirmed in loading cycles.

An integration path for calculating ΔJ is exampled in Fig. 2.10 for the 8 mm-long

crack. The same paths were taken for other size cracks. For the efficiency, the portion of

1.0 to 1.5 cycle is selected since another cyclic loading, i.e., response of 2.0 to 2.5 cycle

results similar ΔJ value. It has been previously explained at the independency of ΔJ to

35
 2400
: −1.5 to 2.5 mm−COD
: 0 to 2.0 mm−COD
1800 : 0 to1.5 mm−COD
: 0 to 1.0 mm−COD
J (N/mm)

: 0 to 0.5 mm−COD
1200

600

0
16 32 48 64
Crack length (mm)

Fig. 2.11 ΔJ-crack length relationships

loading cycles. The analyses were carried out with assigning fixed crack length (ranging

from 8 to 56 mm at 8 mm intervals) and fixed constant displacement amplitude (ranging

from -1.5 to 2.5 mm at 0.25 mm intervals). In figure, un-symmetrical range of -1.5 to 2.5

mm-COD was considered since the minimum COD more than -1.5 mm is difficult to be

reached due to the high compressive load. The obtained ΔJ values are summarized in Fig.

2.11. The value of ΔJ for arbitrary crack length and amplitude was linearly interpolated

from two nearest points shown in the figure.

2.5. Summary

This chapter comprises the study methodology elaborating the way to carry out

experiment and analysis. In the experiment description, explanations about the way to

arrange experimental setting up, to employ the load and to measure crack propagation

36
have been provided. As for the following step, on the analysis section, the independency

of ΔJ due to integration path and loading cycle was verified. It can be showed by

negligible differences on the obtained ΔJ values. Moreover, the ΔJ computation results

related to the crack length are constructed and perform declining manner along with the

crack extension due to displacement-controlled employment. The ΔJ obtained from this

chapter is utilized as the parameter to examine crack growth rate which will be elaborated

later at the following contents in this study.

37
38
3. Constant Amplitude Loading (CA)

3.1. Introduction

This chapter evaluates crack growth rate behavior under constant amplitude

loading with using the ΔJ. In the previous chapter, the methodology to acquire the crack

propagation rate due to applied cyclic loading has been explained. To obtain the accurate

description of the behavior, the results under constant amplitude loadings obtained at the

previous study were also contemplated in this present study. To describe the specimen

responses, load-COD relationships, verification on the analysis to the test result, and crack

growth rate-ΔJ under constant amplitude loading (CA) are in orderly performed. At the

final part, relying on the cumulative crack growth rate plots from past and current work,

estimation model was provided and validation was conducted with showing crack

propagation analysis results.

3.2. Previous study on the constant amplitude loading

At the previous research work, Hanji et al. (2017) evaluated LCF crack growth

rate under constant amplitude loading using compact tension (CT) specimens. The

dimensions of the specimen were established referring to the ASTM standard E1820-08.

To assign the crack origin in specimen, an artificial notch was built. To obtain crack

propagation behavior, three specimen types were provided to identify crack propagation

39
 Fig. 3.1 Location of weld metal on the CT specimen

Table 3.1 Test matrix for past research work

Case 1 2 3 4 5 6 7 8 9 10 11
Material JIS-SM400A JIS-SM400A Weld metal
Displ. range 0.8 1.5 3.0 0.8 1.5 3.6 1.5 0.4 0.8 1.5 3.0
(mm)
Min. displ. 0 0 0 0 0 0 1.5 0 0 0 0
(mm)

rate in different material, plain steel and weld metal. To fabricate specimen, steel plates

of JIS-SM490A and JIS-SM400A (steel grade of 490 and 400 N/mm2) and a weld metal

of JIS Z 3183 S501-H were used. As for the welding part, its position can be shown on

the Fig. 3.1. The materials were originated from the different grades and all specimens

had the uniform dimensions of the 60 mm-height, 62.5 mm-width, 12.5 mm-thickness.

Several constant amplitude loadings applied to the CT specimen in different

material characteristics are outlined on the Table 3.1 above. As shown on table, the load

40
 Fig. 3.2 Crack growth rate and ΔJ under constant amplitude loading at the past study

cases also contemplate different average of cyclic displacement. This past study presented

cyclic J-integral range, ΔJ, a LCF parameter obtained through 2D FEA. In definition,

these ΔJ values were calculated based on plane strain analysis on the considered path

where the 2D plane was assumed as middle plane in CT specimen thickness of 3D-model.

The test results were arranged with establishing the relationships between crack

growth rate, da/dN and ΔJ. As expressed in Fig 3.2, although there were the differences

in displacement range and displacement mean, the crack growth rates are dispersed within

a relatively narrow area. Additionally, all plots were distributed in the same area, as shown

in figure, meaning that the cyclic J-integral range correlates with the crack growth rate

regardless of the material. Based on the results, the following formula can be derived

assuming that the relationship yields a straight line on log–log scales that can be

approached by Eq. (3.1) as follows:

41
 𝑑𝑎
= 9.6 × 10−6 . ∆𝐽1.67 (3.1)
𝑑𝑁
where da/dN is the crack growth rate (mm/cycle), and ΔJ is cyclic J-integral range
(4.1)
(N/mm).

Evaluation of the crack growth performed by Hanji et al. (2017) interpreted that

Eq. (3.1) is applicable to examine crack extension on the different grade of the steel

material.

3.3. Present study on constant amplitude loading

For constant amplitude loadings, two COD amplitudes were taken, CA-1 and CA-

2. The maximum CODs for CA-1 and CA-2 were 1.0 mm and 0.5 mm respectively, while

Fig. 3.3 Load-COD hysteresis loop

42
the minimum COD was set to be 0 mm for both cases.

Fig. 3.3 shows the measured load-COD relationships for CA-1 and CA-2 during

entire loading histories. In the graph, the maximum load of each cycle continually

decreased because of the crack extension. This behavior is more noticeable as the COD

(a) 0 mm-long crack (b) 10 mm-long crack

(c) 20 mm-long crack

Fig. 3.4 Load-COD verification

43
Fig. 3.5 Crack propagation behavior and ΔJ distribution under constant amplitude

loadings

range increased, where CA-1 presents more clear strength degradation pattern than that

under CA-2. Crack closure was also noticeably present, indicated by inflection points in

unloading process where the crack surfaces have already experienced contact. This

inflection point is identified where the hysteresis loop change its continuity on the

unloading process.

Several crack lengths, i.e., 0, 10, 20 mm, applied under CA-1, for 1.5 cycles were

selected for analysis-test verification since such crack lengths were also presented from

experiment. The validation was done at the same crack lengths. As presented in Fig. 3.4,

analysis agreed well with the experiment where the maximum and minimum loading

point are almost same regardless the unidentified closure in analysis. Since for considered

loading cycle, the ΔJ was principally computed from minimum to the maximum loading

point, the closure effect at the unloading can be considered to be insignificant.

Fig. 3.5 exhibits the crack growth curves, the relationship between the number of

44
cycles and crack length. The crack length is defined by accumulating crack propagation

rates from each cycle. In the figure, the fluctuation of ΔJ is also shown. Because of the

displacement-controlled condition, the inclination of the crack growth curves gradually

decreases. The change of curve inclination indicates that the crack propagation rates at

the earlier cycles gradually decreased along with the crack extension. The same trend is

also identified in ΔJ, where the ΔJ inclination curve is reduced with the crack length

Fig. 3.6 expresses he crack growth rates from both studies, this present and past

study (Hanji et al., 2017), which are distributed in the same region within a narrow band.

This tendency indicates that ΔJ correlates with crack growth rates regardless of the

material and specimen size scale. Based on crack growth rate-ΔJ pattern, the regression

curve for constant amplitude loading is reconstructed as follows:

Fig. 3.6 ΔJ-crack growth rate relationships under constant amplitude loadings

45
 𝑑𝑎
= 1.74 × 10−6 . ∆𝐽2.00 (3.2)
𝑑𝑁
The regression lines by Eq. (3.1) and (3.2) are also shown in Fig. 3.6. A(4.1)
small

difference is observed in their slopes, but it is not critical. Since Eq. (3.2) is established

(a) CA-1

(b) CA-2

Fig. 3.7 Crack length-number of cycles relationships under constant amplitude loadings

46
on a larger number of test data than Eq. (3.1) and considered to be reliable, Eq. (3.2) is

used to provide the crack growth rate under constant amplitude loading in this study.

Crack propagation analysis and validation were conducted for constant amplitude

loadings, and the crack growth rates were defined by Eq. (3.2). Total crack length are

obtained by accumulating crack growth rates during each cycle. The crack propagation

analysis results are expressed in Fig. 3.7. From the figure, it can be shown that estimation

model by Eq. (3.2) perform similar estimated crack lengths to those observed from the

experiment and confirm the applicability of Eq. (3.2).

3.4. Summary

This chapter discloses crack growth rate behavior under constant amplitude

loadings. Even though at the past research work (Hanji et al. 2017), the crack propagation

was observed under the same typical of loading, this present research attempted to employ

constant amplitude loading to test the specimen with different material and specimen size

from the past study to result rigorous estimation model for the crack propagation. The

results from both studies, past and this current study, figured out that crack growth rates

were distributes on the same region within narrow band and yield the estimation model.

Crack propagation analysis presented acceptable results shown by identical manner of the

crack length resulted from the analysis and experiment.

47
48
4. Two-steps Variable Amplitude Loading (VA)

4.1. Introduction

This chapter aims to evaluate crack propagation rate under variable amplitude

loadings, particularly when the amplitudes was changed. Due to its simple loading

attributes, two-steps variable was selected to examine amplitude alteration influence to

the crack extension. The study was firstly done with evaluating crack growth behavior

among test cases and investigating the crack growth patterns when the loading amplitudes

were altered. Accordingly, collective plots of the crack growth under constant and

variable amplitude loading at any crack length were outlined. After the crack growth

patterns of two-steps variable cases were characterized, EPFM parameter computed from

FE analysis, i.e., the ΔJ, was employed to specify crack growth manners. At the following

step, estimation model was established and validated to the test result.

4.2. Load cases

In this load cases, two different constant COD amplitudes were mixed for each

specimen. Fig. 4.1 shows the loading patterns where each loading history consists of high

amplitude and low amplitude loading cycles. The magnitude of the high amplitude was

1.0 mm through VA-1 to VA-4. The magnitude of the low amplitude was variably

employed in order to examine the effect of different amplitude loading. Besides the steps

number at high and low amplitude loading was also considered to identify its influence

49
 (a) VA-1 (b) VA-2

(c) VA-3 (d) VA-4

Fig. 4.1 Loading patterns of two-steps variable amplitude loadings

to the crack propagation rate. This arrangement is exampled at the VA-1 and VA-2 where

VA-1 and VA-2 have the same loading amplitude but have the difference in the number

of steps at high and low amplitude loading.

4.3. Load-COD relationship

The behavior of the specimen under two-steps variable loadings was exampled by

50
evaluating VA-1 test case until 6 cycles. The distinction was made for high amplitude in

red line and low amplitude in blue line. As expressed on the Fig. 4.2, maximum loads of

each cycle experience exhibited declining manner because of the crack growth. Due to

the higher crack propagation rate, such tendency is more clearly shown at the higher

amplitude steps than that under low amplitude loading employment.

In high and low amplitude loading, the crack closure was also figured out. As

shown in figure, at the unloading process, the inflection points were indicated due to crack

surfaces contact. In such process, when the compressive load was acted to the specimen,

the crack faces are contacted together and cause un-continuity of curve where the

compressive load significantly increases before it reaches 0-COD.

Fig. 4.2 Load-COD hysteresis loop under VA-1

51
4.4. Crack propagation behavior

Crack propagation behavior is presented with providing crack length and number

of cycles as presented in Fig 4.3. As expressed in constant amplitude loading, the curve

Fig. 4.3 Crack growth rate behavior

Fig. 4.4 Crack growth rate-crack length relationships

52
inclination gradually decreased in tandem with the crack extension due to the

displacement-controlled conditions. Due to the lowest low amplitude among other two-

steps variable cases, VA-4 experienced the most significant reduction in crack growth

when the amplitude was dropped. VA-1 and VA-2 performed almost similar crack growth

behavior even though their cumulative cycle number in each step are different. VA-3

performed the most identical crack growth behavior to the constant amplitude due to the

close magnitude of high and low amplitude steps.

An example of the crack growth rate at any crack length is shown in Fig. 4.4. For

the comparison, the crack growth rates under constant amplitude loading are also plotted.

Comparing high amplitude loading in VA-1, VA-2 with CA-1, the difference identified

in the crack growth rate is negligible, meaning the effect of variable amplitude loading

can be neglected in high amplitude loading. Meanwhile, low amplitude loading in VA-1,

VA-2 represents higher crack propagation rates from those under CA-2 even though COD

amplitude is equivalent. This behavior implies that crack propagation rates are accelerated

than those under constant amplitude loading when the amplitudes are decreased.

The accelerated crack propagation rates are highly expected by the remained

plasticity area at the end of high amplitude loading employment. In such process, a drop

of the amplitude is supposed to leave wider plastic zone than that under low amplitude.

Accordingly, that wider plasticity area is possible to accelerate the crack growth. Those

mechanism are hypothetically provided and need to be clarified for the future research

work.

Fig. 4.5 shows the relationships between the crack growth rate and ΔJ under two-

steps variable loadings. For high amplitude loading cycles, such relationships are shown

in Fig. 4.5(a). The equivalent trend of crack growths to those under constant amplitude

loading is exhibited in this figure. Thus, such crack propagation can be predicted from

the regression curve for constant amplitude loading. In contrast, the crack growth rates

for low amplitude loading cycles, shown in Fig. 4.5(b), are plotted above the regression

53
 (a) High amplitude

(b) Low amplitude

Fig. 4.5 ΔJ-crack growth rate relationships under constant and two-steps variable

amplitude loadings

curve especially in relatively high ΔJ region. Hence, the constant amplitude regression

curve provides the crack growth rate in non-conservative side. According to the findings,

54
the crack growth rate in low amplitude loading should be carefully investigated.

4.5. Amplification factor (AF)

To simulate crack growth under variable amplitude conditions, accelerated crack

growth is expressed as the amplification factor (AF) that denotes the ratio of crack growth

rate under variable amplitude loading to constant amplitude loading. Here, AF is

expressed as:

(𝑑𝑎⁄𝑑𝑁)
𝑣𝑎𝑟
𝐴𝐹 = (4.1)
(𝑑𝑎⁄𝑑𝑁)
𝑐𝑜𝑛𝑠𝑡
(5.2)
where (da/dN)var is the crack growth rate under variable amplitude loading, (da/dN)const is

the crack growth rate obtained from the regression curve presented in Eq. (3.2).(5.2)

the ΔJ is
To obtain crack propagation rate under constant amplitude in Eq. (4.1), (5.2)

substituted to the Eq. (3.2) in which, this ΔJ is defined from the equivalent COD and

equivalent crack length to those under variable amplitude.

Fig. 4.6 shows a scatter of AF during each loading history. The test data for high

and low amplitude loading are individually presented. In both figures, between the results

for VA-1 and VA-2, no obvious difference can be observed, meaning that the number of

constant loading cycles has little effect on the AF. As shown in Fig. 4.6(a), during high

amplitude loading, the AF is scattered below three. In contrast, a high AF can be seen

under low amplitude loading conditions, as shown in Fig. 4.6(b). Generally, these high

AFs are noticeably decreased with crack extension, and become 1.0 or less when the crack

length reaches approximately 25 mm in each loading case, although AF values strongly

depend on the loading case, i.e., the magnitude of low amplitude loading.

The AF is supposed to be influenced by a combination of the previous and the

current loading amplitudes. Therefore, a new parameter of ΔJ ratio, ΔJi/ΔJi-1, which is

55
 (a) High amplitude

(b) Low amplitude

Fig. 4.6 AF-crack length relationships

the ratio of ΔJ at an observed cycle to ΔJ at one cycle before the observed cycle, is

examined. Fig. 4.7 shows the relationships between the AF and the ΔJ ratio. The figures

are separately indicated for the first loading cycle after amplitude alteration and others.

56
 (a) First cycles, ΔJi/ΔJi-1 ≥ 1

(b) First cycles, ΔJi/ΔJi-1 < 0.95

(c) Second and steady following cycles, 0.95 ≤ ΔJi/ΔJi-1 < 1

Fig. 4.7 AF-ΔJ ratio relationships under two-steps variable amplitude loadings

As shown in Fig. 4.7(a), at the first cycle just after the amplitude increases from low to

57
high level, the AF is almost 1.0 constantly, which means that it can be disregarded when

the ΔJ ratio is larger than 1.0. Meanwhile, as shown in Fig. 4.7(b), at the first cycle when

the amplitude is reduced from high to low level, the AF tends to increase with the decrease

of the ΔJ ratio. At the second and steady following cycles, the ΔJ ratio was gradually

decreased even though the COD amplitude was kept constant because of the displacement

controlled condition, where the ΔJ ratios were distributed in the range from 0.95 to 1.0,

and they have an approximately linear relation with the AF as shown in Fig. 4.7(c). Relied

on this figure, the inclination of each curve tends to be changed depending on the

magnitude of the COD amplitude or ΔJ.

Fig. 4.8 presents the relationships between AF and ΔJ. Fig. 4.8(a) exhibits AF

plots related to the ΔJ at the ΔJ ratio higher than 1.0, where the AFs lead to give the value

of 1.0 constantly at the entire ΔJ values. Fig. 4.8(b) demonstrates the AF values at the ΔJ

ratio less than 0.95, where the AFs perform elevating trend during reducing ΔJ values.

Fig. 4.8 (c) shows the AF patterns of each test cases on the range within 0.95 to 1.0. The

figure describes that AFs of high amplitude loadings are distributed within 300 to 800

N/mm-ΔJ and AFs of low amplitude loadings are scattered within, 0 to 300 N/mm-ΔJ.

The figure exhibits that the low amplitude loading experience more narrow range of AFs

scatter than those under high amplitude loading. This tendency is obtained since the crack

rapidly propagates at the high amplitude and then dropped when the COD amplitude were

decreased.

4.6. Estimation model on the AF

AF estimation formula was firstly established with considering a parameter related

to the AF, ΔJ ratio. This parameter was selected since it is dimensionless and applicable

to be used in any circumstances of ΔJ value. Accordingly, the estimation was attempted

58
 (a) First cycles, ΔJi/ΔJi-1 ≥ 1

(b) First cycles, ΔJi/ΔJi-1 < 0.95

(c) Second and steady following cycles, 0.95 ≤ ΔJi/ΔJi-1 < 1

Fig. 4.8 AF-ΔJ relationships under two-steps variable amplitude loadings

to be constructed relied on Fig 4.7. Estimated AF for ΔJ ratio higher than 1.0 was

59
obtained from Fig. 4.7(a) where the AFs are clearly dispersed at the value equals to 1.0.

Fig. 4.7(b) and Fig. 4.7(c) illustrate that even though the ΔJ ratios are nearly equal, AFs

are widely scattered. Due to such sprinkled plots, additional descriptions comparing the

test and estimation are required to clarify the accuracy of the estimation formula.

A linear regression is chosen to reflect the AFs plots since such regression line has

the best coefficient determination (R squared) than other regression trend lines. However,

as expressed on the Fig 4.9(a) and Fig. 4.9(b), the linear regression is not able to

satisfyingly represent overall dispersed AFs.

According to those findings, both parameters, ΔJ ratio and an absolute ΔJ are

required to estimate the AF. Based on the scatter patterns of ΔJ and ΔJ ratio, a formula to

estimate the AF is established and proposed as follows:

787 0.4 ∆𝐽𝑖 2.5 Δ𝐽𝑖

( ) + 1.8 (1 − ) + 0.5 ( < 0.95)
𝛥𝐽𝑖 ∆𝐽𝑖−1 Δ𝐽𝑖−1
5670 1.77 ∆𝐽𝑖 Δ𝐽𝑖
𝐴𝐹𝑒𝑠𝑡 = ( ) (1 − ) ( 0.95 ≤ < 1) (4.2)
𝛥𝐽𝑖 ∆𝐽𝑖−1 Δ𝐽𝑖−1
Δ𝐽𝑖
1.0 ( ≥ 1) (5.2)
{ Δ𝐽𝑖−1

Fig. 4.7(b) and Fig 4.8(b) are utilized to obtain estimated AF on the ΔJ ratio range
(5.2)
less than 0.95. For the accuracy, the coefficients for ΔJ ratio and ΔJ in Eq. (4.2) are
(5.2)
obtained by Microsoft Excel’s regression analysis relying on AFs scatter. Accordingly,

the power for each parameter is firstly determined and by that feature the coefficients then

are solved.

Estimated AF on the ΔJ ratio range within 0.95 to 1.0 is derived from the scatter

plots of AFs presented in Fig. 4.7(c) and Fig. 4.8(c). The equation also considers ΔJ ratio

and ΔJ to determine estimated AF where it is resulted with assigning relationship between

AF/(1-ΔJ ratio) as the vertical axis and ΔJ as the horizontal axis. Thus, relying on the

60
 (a) First cycles, ΔJi/ΔJi-1 ≥ 1

(b) Second and steady following cycles, 0.95 ≤ ΔJi/ΔJi-1 < 1

Fig. 4.9 Cumulative plots of estimation and test results on AF-ΔJ ratio relationships

plots of AF/(1-ΔJ ratio), the regression equation to obtain AF/(1-ΔJ ratio) is stated as the

function of ΔJ as shown in Fig. 4.10. For the validation of Eq. (4.2), estimated AF by the

above equation was utilized to yield estimated crack growth rate which is acceptable for

fatigue life estimation as presented in later section.

61
 Fig. 4.10 Regression process to obtain estimated AF at 0.95 ≤ ΔJi/ΔJi-1 < 1

To illustrate the pattern of estimated AF at the ΔJ ratio less than 1.0, the plots

then was established within range 0 to 1.0 relied on two-steps variable amplitude loading

as shown in Fig. 4.11. In general, this figure indicates that the estimated AFs perform the

discontinuity in the plots and specific patterns depending on the ΔJ ratio range. At the

range within 0.95 to 1.0, the detail pattern of AFs is given in Fig. 4.12. The Eq. (4.2) for

the range within 0.95 to 1.0 is derived from two-steps cases, so that this formula is only

valid for variable amplitude loading at the low cycle fatigue condition.

4.7. Validation on the estimation to the test results crack growth rate

In this research, the estimated crack growth rate (da/dN)est is expressed as follows:
𝑑𝑎 𝑑𝑎
( ) = 𝐴𝐹𝑒𝑠𝑡 . ( ) (4.3)
𝑑𝑁 𝑒𝑠𝑡 𝑑𝑁 𝑐𝑜𝑛𝑠𝑡
The estimated crack growth rate for constant amplitude loading presented in Eq.
(5.3)

(5.3)

62
(5.3)
 Fig. 4.11 Estimated AF at the ΔJi/ΔJi-1 < 1

Fig. 4.12 Estimated AF at the 0.95 ≤ ΔJi/ΔJi-1 < 1

(4.3), (da/dN)const, was calculated from Eq. (3.2) by substituting the ΔJ based on constant

amplitude, as explained in Section 3. As shown in Fig. 4.13, estimated and test result

crack growth rate were distributed in the same region within narrow band and confirmed

applicability of Eq. (3.2) corrected by Eq. (4.2) to estimate crack propagation.

63
 Fig. 4.13 Verification on the estimation model to the test result crack growth rate (unit:

mm/cycle)

4.8. Crack propagation analysis

Crack propagation analysis and validation were conducted for all test cases, and

the crack growth rates were defined under the two-steps variable amplitude loadings in

Eq. (4.3). In this equation, the crack growth rate under constant amplitude loading is

calculated from Eq. (3.2), while the estimated AF is calculated from Eq. (4.2). Total crack

lengths are obtained by accumulating crack growth rates during each cycle. The crack

growth curves, the relationships between the crack length and the number of cycles, are

shown in Fig. 4.14. For comparison, the prediction curve neglecting the AF is also

presented. The figures indicate that disregarding the AF estimates the crack length shorter

than the actual one obtained from the experiment, and their gaps are significant. Upon

consideration of the AF, the estimated crack length presents close pattern to the test results.

64
 (a) VA-1 (b) VA-2

(c) VA-3 (d) VA-4

Fig. 4.14 Crack propagation analysis results

Therefore, the formula presented in Eq. (3.2) corrected by Eq. (4.2) is applicable to

estimate the crack growth rate under two-steps variable amplitude loading.

4.9. Summary

This chapter elaborates crack propagation behavior under two-steps variable

amplitude loading. The observation was emphasized on the loading amplitude alteration

65
from high to low levels and from low to high levels. When the amplitudes were dropped

from high to low levels, it was found that the crack growth rates increase than those

observed under constant amplitude loadings. Meanwhile, when amplitude loading

increased from low to high levels, the crack growth rates were similar to those observed

under constant amplitude conditions. Futhermore, during the second and steady

subsequent cycles, the crack growth rate decreased continuously in the same manner as

those under constant amplitude loading.

The crack growth rates under two-steps variable amplitude loading conditions can

also be simulated by AF, which is the ratio of the crack growth rate under two-steps

variable amplitude loading conditions to that under constant amplitude loading conditions.

Estimation model are also presented from this study and yields the formula which is

considered acceptable, as shown by the close pattern between the estimation and

experimental results.

66
5. Random Variable Amplitude Loading (RA)

5.1. Introduction

Crack propagation behavior under random amplitude loading was discussed in

this chapter. This random amplitude loading considered different amplitude loading

ranges and the average of COD. Firstly, the amplitudes of random loading was determined

where they were un-repetitive, to obtain value of fully random. Next, to identify the

behavior of random and constant amplitude conditions, crack growth rates from both

typical loadings were togetherly plotted. Since at two-steps variable amplitude loading,

crack growth rates were governed by ΔJ ratio, the growth rate under random loading was

also distinguished based on such parameter. At the final work, estimation model from

two-steps variable was employed to the random amplitude loading and then validated.

5.2. Load cases

The COD amplitudes were obtained by arranging each peak COD in range within

0 to 300-level, in which 0-levels refers to minimum COD and 300-level is addressed to

maximum COD.

Three different ranges of COD were employed: RA-1, RA-2, and RA-3, with

range values of 0.5 to 1.5 mm, 0.5 to 2.0 mm, and -1.5 to 2.5 mm, respectively for random

amplitude loadings, as shown in Fig. 5.1. All CODs are un-repetitive and attributed to be

67
 (a) RA-1

(b) RA-2

(c) RA-3

Fig. 5.1 Loading patterns of random variable amplitude loading

random entirely. The RA-1 and RA-2 are employed to examine the effect of amplitude

68
range differences to the crack extension. Both have minimum COD set to be 0 mm. In the

RA-3, the minimum COD are set to be -1.5 mm to evaluate effect of the average COD on

the crack propagation. As for the RA-3, the minimum COD value more than -1.5 mm is

not recommended since it is required very high compressive load which is difficult to be

reached on the test.

Fig. 5.2 Loading process used to define the ΔJ for RA-3

Fig. 5.3 The ΔJ for RA-3 defined based on constant amplitude principle

69
5.3. Test result arranged with ΔJ

In random loading, the ΔJ is calculated in each cycle in the loading process as

previously illustrated in Fig. 2.5. Loading simulation on the model is applied until 1.5

cycles. For RA-1 and RA-2, the ΔJ is defined with the same principle as constant

amplitude with minimum COD equals to 0. Meanwhile, for RA-3, the ΔJ is computed

and refers to an illustration in Fig. 5.2. The figure outlines several cycles to define the ΔJ

in a loading history. Line a-b, c-d, e-f are the loading processes that need to be defined

for the ΔJ. Fig. 5.3 outlines each loading process in which ΔJ is computed from the

minimum COD to maximum COD.

It is obtained from two-steps variable amplitude cases, that the crack growth rate

is governed by a combination between the previous and the current loading amplitudes.

Therefore, the ΔJ ratio, ΔJi/ΔJi-1, which is the ratio of ΔJ for an observed cycle to ΔJ at

one cycle before the observed cycle, was also employed in these typical loadings.

Fig. 5.4 shows an example of the crack growth rate at any crack length which

provides comparison between random amplitude loadings and CA-1. The figures are

individually presented for the ΔJ ratio equal to or larger than 1.0 and for the ΔJ ratio less

than 1.0. To obtain proportional comparison on the crack growth rates, 0.9 to 1.1 mm-

COD range is selected for random loading and then compared to the CA-1. Since the

COD of both types of loadings are almost equivalent, the crack growth comparison can

be considered reliable. In figure, the differences in crack growth rates between CA-1 and

random amplitude conditions at ΔJ ratio higher than 1.0 are indistinguishable as shown

in Fig. 5.4(a). Hence, the influence of random amplitude loadings can be neglected when

the amplitudes increase from low to high levels. In contrast, as expressed in Fig. 5.4(b),

the crack growth rate observed during ΔJ ratio less than 1.0 shows considerable

distinctions from those observed during CA-1, even though the COD amplitude is

equivalent. This implies that accelerated crack growth should be considered when the

70
 (a) ΔJ ratio ≥ 1

(b) ΔJ ratio < 1

Fig. 5.4 Crack growth rate-crack length relationships under CA-1 and random amplitude

loadings

amplitude was dropped. Additionally, the graph also figured out that the random

amplitude range affects the crack growth rates, i.e., the RA-3 provides higher crack

growth rate than those observed in the RA-1 and RA-2.

71
 (a) ΔJ ratio ≥ 1

(b) ΔJ ratio < 1

Fig. 5.5 ΔJ-crack growth rate relationships under constant and random amplitude loadings

Fig. 5.5 shows relationships between the crack growth rate and the ΔJ under

random variable amplitude loading where the relationships are distinguished based on ΔJ

72
 (a) ΔJ ratio ≥ 1

(b) ΔJ ratio < 1

Fig. 5.6 AF-ΔJ ratio relationships under random amplitude loadings

ratio range, as previously explained. Similar pattern for all cases of random amplitudes

loading to the constant amplitude conditions was indicated at the ΔJ ratio higher than 1.0

as shown in Fig. 5.5(a). This behavior indicated that crack growth rate for ΔJ ratio higher

than 1.0 can be interpreted by utilizing regression curve originated from constant

amplitude loadings. On the contrary, as shown in Fig. 5.5 (b), at the ΔJ ratio less than 1.0,

a trend of the crack growth rate which is obviously different than that in constant

amplitude loadings was interpreted where most of the plots are scattered above the

regression curve. Therefore, the crack growth rates have to be corrected by the AF for the

case of the ΔJ ratios less than 1.0.

73
 In random amplitude cases, the AF was also considered since such parameter was

noticeably to be influenced by the ΔJ ratio. Hence, relationship between AF and ΔJ ratio

are presented in Fig. 5.6. When the amplitude increases from low to high level, the

constant value of 1.0 in AF is exhibited as shown in Fig. 5.6(a). This means that the

contribution of the AF is meaningless when the ΔJ ratio is larger than 1.0. On the contrary,

as shown in Fig. 5.6(b), when the amplitude is reduced from high to low level, AF tends

to increase with the decrease of the ΔJ ratio.

Fig. 5.7 expressed relationships between AF and ΔJ. The AFs lead to give the

value of 1.0 constantly during entire ΔJ values at the ΔJ ratio higher than 1.0, as presented

(a) ΔJ ratio ≥ 1

(b) ΔJ ratio < 1

Fig. 5.7 AF-ΔJ relationships under random amplitude loadings

74
 Fig. 5.8 Verification of estimated crack growth rate to the test result (unit: mm/cycle)

in Fig 5.7(a). On the contrary, as exhibited in Fig 5.7(b), the AFs performs elevating

manner together with decreasing value of ΔJ at the ΔJ ratio less than 1.0.

As an estimation of the AF, the same equation to Eq. (4.2), derived for two-

steps variable amplitude loading, is adopted even for random variable amplitude loading.

Since the random loading has un-repetitive COD, estimated AF only adopts two formulas

from two-steps cases. The estimation formula for ΔJ ratio less than 0.95 and the

estimation formula for ΔJ ratio higher than 1.0. Here, the estimation formula for ΔJ ratio

less than 0.95 for two-steps is subjected to estimation formula for ΔJ ratio less than 1.0

in random loading. The comparison of the crack growth rate under random amplitude

loading, shown in Fig. 5.8, elucidates the crack growth rate obtained by Eq. (3.2) and

corrected by the AF in Eq. (4.2) is close to the test result, even though there is the scatter

on the plots. This condition is identified since the regression line is established mostly

from the crack propagation rate less than 1.0 mm/cycle. Hence, it may result a more wide

range plots for crack propagation rate higher than 1.0 mm/cycle. The relationships

75
 (a) RA-1 (b) RA-2

(c) RA-3

Fig. 5.9 Crack propagation analysis results

between the crack length and the number of cycles for random variable amplitude loading

are shown in Fig. 5.9. In the figure, a prediction curve neglecting the AF is drawn together.

It is confirmed that the neglection presents shorter crack length than the test results. The

consideration of the AF can improve the accuracy so that the crack growth curve expresses

the pattern similar to the test results. Consequently, it is confirmed that the AF by Eq.

76
(4.2), derived from two-steps variable amplitude loading, is also applicable for random

variable amplitude loading regardless the magnitude of COD and average of COD.

5.4. Summary

On the purpose to investigate crack propagation behavior under random amplitude

loading, an experimental work and analytical study were conducted. As identified from

two-steps variable loadings, crack growth rates were governed by ΔJ ratio. Accordingly,

ΔJ ratio which is less than 1.0 and higher than 1.0 were considered and such distinction

manners were employed to the random variable loadings. Relied on the experiment, crack

growth rates increased when the amplitudes were reduced from high to low levels or at

ΔJ ratio less than 1.0, than those observed under constant amplitude loading. Contrarily,

the equivalent tendency of the crack growth rates to that under constant amplitude

conditions were identified when the amplitudes were elevated from low to high levels or

at the ΔJ ratio higher than 1.0. Estimation formula yielded from two-steps variable cases

was applied to the random variable loading expressed that the estimation model can be

accepted in domain of random loading regardless the COD magnitude and COD mean.

77
78
6. Summary and Conclusions

6.1. Conclusion

In this study, the author employed ΔJ as a parameter to evaluate crack growth rate.

Under variable amplitude loading, LCF crack propagation was investigated through

experiments and analysis. By those ways, the observations were emphasized on the crack

growth behavior when the loading amplitudes were altered. The conclusions derived from

this study are included as follows.

1. As a general summary, as discovered from the variable amplitude loading, there

is a different tendency between the crack growth rates at the decreasing and at the

increasing phases of amplitude when those two crack growth manners were

evaluated to those under constant amplitude conditions.

2. Chapter 2 contemplates the different integration paths and loading cycles to

examine the independency of ΔJ and relying on the analysis, the negligible

difference on the ΔJ was verified. ΔJ which correlates to the crack length were

also outlined in this chapter where the ΔJs performed declining manner along with

the extension of the crack due to displacement-controlled condition.

3. Chapter 3 reveals crack propagation pattern under constant amplitude loading. To

ensure the accuracy of the estimation, study results from the past study (Hanji et

al. 2017), also tested crack growth under constant amplitude conditions, were

involved and collectively plotted with the present study. As for the validation,

cumulative plots of load-COD response at the same crack lengths, resulted from

79
 experiment and analysis, figure out that maximum and minimum loading points

are identical which confirm that the analysis can interpret the experimental results.

The present study employed constant amplitude with different material and

specimen size from the previous research (Hanji et al., 2017). However, from both

studies, the crack growth rates were distributed within narrow band regardless

material grades, specimen dimensions and COD averages.

4. In chapter 4, two-steps variable amplitude loadings were employed to the

specimen and crack growth rates were evaluated. At the amplitude change from

high to low levels, escalation rates of the crack growth than those observed under

constant amplitude condition were confirmed. In contrast, the close pattern of

crack growth rates to those under constant amplitude loadings was indicated when

the amplitudes were elevated from low to high levels. The second and following

cycles on the two-steps variable loading performed crack growth rates

sustainability decreased in the same manner as constant amplitude. In this chapter,

an amplification factor (AF) which is the ratio of crack growth rate under variable

amplitude to that under constant amplitude loading was introduced, and confirmed

to be useful parameter to describe the crack growth rate manners. At the final work,

estimation model was resulted from two-steps variable amplitude and then

validated to be acceptable.

5. In chapter 5, random variable loading was employed to investigate crack

propagation rate. Since the crack growth rates were noticed to be correlated to the

ΔJ ratio from two-steps variable study, the plots were also distinguished based on

ΔJ ratio in these random variable cases. In such cases, at the ΔJ ratio less than 1.0,

the higher rates of crack growth than those observed under constant amplitude

loadings is clarified. Meanwhile, at the ΔJ ratio higher than 1.0, an equivalent

manner of crack growth rates as shown by constant amplitude loading was

exhibited. At the final discussion, the acceptable results due to crack propagation

80
 estimation was yielded when the formula acquired from two-steps variable study

was also employed to the random variable cases.

6.2. Recommendation for future study

In this study, at the discussion of two-steps variable and random amplitude, a

higher crack growth rate than those under constant amplitude was obviously clarified

when the amplitudes were reduced from high to low levels. This behavior is

hypothetically expected caused by wide plasticity area formed at the end of the higher

amplitude loading employment. However, the reason of those accelerated crack growth

rates has not been discussed in this dissertation, yet. Correspondingly, at the future study,

it is suggested to establish a modelling that can completely simulate the crack progress to

reveal the contribution of the plasticity to the crack growth rate manners.

As expressed in Chapter 1, LCF failures were identified on the bridge pier

particularly at the beam column connection and base joint of steel bridge piers. However,

the number of LCF crack propagation tests on this bridge pier is very limited and as far

the tests have mainly been conducted under constant amplitude loading. At the real

condition, the bridge pier interacts with seismic loading which is attributed to the

variability of the cyclic strains. Therefore, to discover crack propagation tendency under

the real loading, it is suggested to test the steel bridge pier under variable amplitude

conditions to characterize crack growth manner due to seismic loading employment.

81
82
References

Agrawal, R., Uddanwadiker, R., Padole, P. (2014): Low Cycle Fatigue Life Prediction,

International Journal of Emerging Engineering Research and Technology, Vol. 2(4),

pp. 5-15.

Anderson, T. L. (2011): Fracture Mechanics: Fundamentals and Applications (3rd ed.),

Taylor and Francis Group.

Bhandari V. B. (2008): Design of Machine Elements (2nd ed.), 3rd Reprint, Mc Graw Hill

Education.

ASTM E1820-031. (2008): Standard Test Method for Measurement of Fracture

Toughness, pp. 1-48.

Chen, T. and Tateishi, K. (2007): Extremely Low Cycle Fatigue Assessment of Thick

Walled Steel Pier Using Local Strain Approach, Journal of Structural Engineering,

Vol. 53A, pp. 485-492.

Chung, Y. S and Abel, A. (1988): Low Cycle Fatigue of Some Aluminium Alloys,

American Society for Testing and Materials STP 942, pp. 94–106.

Coffin, L. F. (1954): A Study of the Effect of Cyclic Thermal Stresses on a Ductile Metal,

Transactions of the American Society of Mechanical Engineers, Vol. 76, pp. 931–950.
Coffin, L. F. (1971): A Note on Low Cycle Fatigue Laws, Journal of Material Science,
Vol. 6(2), pp. 388–402.
Corbly, D. M., Packman, P. F. (1973): On the influence of single and multiple peak
overloads on fatigue crack propagation in 7075-T6511 aluminum, Engineering
Fracture Mechanics, Vol. 5(2), pp. 479-497, 1973.

83
Dahlin, P., and Olsson, M., (2007): Fatigue crack growth – Mode I cycles with periodic

Mode II loading, International Journal of Fatigue, Vol. 30(5), pp. 931-941.

Dong, Q., Yang, P., Xu, G. and Deng, J. (2016): Mechanisms and Modeling of Low Cycle

Fatigue Crack Propagation in A Pressure Vessel Steel Q345, International Journal of

Fatigue, Vol. 89, pp. 2-10.

Dowling, N. E., Begley, J. A. (1976): Fatigue crack growth during gross plasticity and

the J-integral, ASTM STP 590, pp. 82-103.

Gardner, L. and Davies, C.M., (2010): Extremely Low Cycle Fatigue Tests on Structural

Carbon Steel and Stainless Steel, Journal of Constructional Steel Research, Vol. 66(1),

pp. 96-110.

Gonzalez-Herrera, A., and Zapatero, J. (2008): Tri-dimensional numerical modelling of

plasticity induced fatigue crack closure, Engineering Fracture Mechanics, Vol.

75(15), pp.4513–4528, 2008.

Hanji, T., Tateishi, K., Terao, N. and Shimizu, M. (2017): Fatigue Crack Growth

Prediction of Welded Joints in Low-Cycle Fatigue Region. International Journal of

Material Joining, Vol. 61(6R), pp. 1189-1197.

Hatanaka, K., and Fujimitsu, T. (1984): Some considerations on cyclic stress–strain

relation and low cycle fatigue life, Transactions of the Japan Society of Mechanical

Engineers A., Vol. 50(451), pp. 291–300.

Gray G. T., Williams J. C., and Thompson A. W. (1983): Roughness-Induced Crack

Closure: An Explanation for Microstructurally Sensitive Fatigue Crack Growth

Metallurgical Transactions A, Vol. 14, pp. 421-433.

Gonzalez-Herrera, A. and Zapatero, J. (2008): Tri-dimensional numerical modelling of

plasticity induced fatigue crack closure, Engineering Fracture Mechanics, Vol. 75(15),

pp.4513–4528.

Itoh, T. and Miyazaki, T. (2003): A Damage Model for Estimating Low Cycle Fatigue

Lives under Non proportional Multiaxial Loading, European Structural Integrity

84
 Society Journal, Vol. 31, pp. 423-439.

James, M. N. and Knott, J. F. (1985): Critical aspects of the characterization of crack tip

closure by compliance techniques, Materials Science and Engineering, Volume 72(1),

1985, pp. L1-L4.

Jiang, Y., Feng, M., and Ding, F. (2005): A reexamination of plasticity-induced crack

closure in fatigue crack propagation. International Journal of Plasticity, Vol. 21(9),

1720–1740.

Jono, M., Sugeta, A. and Ohta, S. (1993): Prediction Method of Elastic-Plastic Fatigue

Crack Growth Rate under Variable Amplitude Loadings, The Japan Society of

Mechanical Engineers, Vol. 59, No. 560, pp. 939-945 (in Japanese).

Kanvinde, A. M. and Deierlein, G. G. (2005): Continuum based micro-models for ultra

low cycle fatigue crack initiation in steel structures, Forensic Engineering Symposium,

Vol. 19(171), pp. 20–24.

Kim, N., (2015): Introduction to Nonliear Finite Element Analyis, Springer Book.

Komotori, J. and Shimizu, M., (1998): Fracture mechanism of ferritic ductile cast iron in

extremely low cycle fatigue, Elsevier Science Ltd, Vol. 39, pp. 39–44.

Kuroda M. (2002): Extremely low cycle fatigue life prediction based on a new cumulative

fatigue damage model, International Journal of Fatigue, Vol. 24(6),

pp.699–703.

Lamba, H. S. (1975): The J-integral applied to cyclic loading, Engineering Fracture

Mechanics, Vol. 7(4), pp. 693-703.

Lambert, Y., Sailiard, P., Bathias, C. (1988): Application of the J concept to fatigue crack

growth rate in large-scale yielding, ASTM STP 969, pp. 318-329.

Ling, M. R. and Schijve, J. (1992). The Effect of Intermediate Heat Treatments on

Overload Induced Retardation during Fatigue Crack Growth in An Al-Alloy, Fatigue

and Fracture Engineering Materials and Structures, Vol. 15, pp. 421–430.

Manson, S. S. (1965): Fatigue: A complex Subject – Some Simple Approximation,

85
 Experiment Mechanical, Vol. 5, pp.193–226.

Masatoshi, K. (2002): Extremely Low Cycle Fatigue Life Prediction based on a new

cumulative fatigue damage model. International Journal of Fatigue, 24(6), 699-703.

Matthews, W. T., Barratta, F. I., Driscoll, G. W. (1971). Experimental observations of a

stress intensity history effect upon fatigue crack growth rate, International Journal of

Fracture Mechanics, Vol. 7 (2), pp. 224-228.

McClung, R. C. and Sehitoglu. (1989): On the finite element analysis of fatigue crack

closure-2. Numerical results, Engineering Fracture Mechanics, Vol. 33, pp. 253–272.

McMillan, J. C. and Pelloux, R, M. N. (1967): Fatigue Crack propagation under program

and random loads, Fatigue Crack Propagation, ASTM STP 415, Vol. 505.

Miki, C. (1997): Fractures in Seismically Loaded Bridges. Progress in Structural

Engineering and Materials, Vol. 1(1), pp. 115-121.

Murakami, Y., Harada, S., Tani-ishi, H., Fukushima, Y., and Endo, T. (1983):

Correlations among Propagation Law of Small Cracks, Manson-Coffin Law and

Miner Rule under Low-Cycle Fatigue, Japan Society of Mechanical Engineers, Vol.

A49, pp. 1411-1419. (in Japanese).

Ljustel, P. (2013): Fatigue crack growth experiments on specimens subjected to

monotonic large scale yielding, Engineering Fracture Mechanics, Vol. 110, pp. 138-

165.

Paris, P. and Erdogan, F. (1963): A Critical Analysis of Crack Propagation Laws,

Transactions of the American Society of Mechanical Engineers Journal of Basic

Engineering, Vol. 85(4), pp.528–533.

Park, J. E., Hanji, T., and Tateishi, K. (2015): Extremely Low Cycle Fatigue Assessment

of Corner Crack in Concrete-Filled Steel Piers based on Effective Notch Strain

Approach, Journal of Structural Engineering, Vol. 61A, pp. 662-671.

Pommier, S. N. (2001): A study of the relationship between variable level fatigue crack

growth and the cyclic constitutive behaviour of steel, International Journal of Fatigue,

86
 Vol. 23, pp.111-118

Pommier, S. N., Bompard, P. (2000): Bauschinger effect of alloys and plasticity-induced

crack closure: A finite element analysis, Fatigue and Fracture of Engineering

Materials and Structures, Vol. 23(2), pp.129-139.

Raghuvir K., Arbind K., Sunil, K. (1996): Delay effects in fatigue crack propagation,

International Journal of Pressure Vessels and Piping, Vol. 67(1), pp. 1-5.

Rice, P. (1968): A Path Integral and the Approximate Analysis of Strain Concentration

by Notches and Cracks, Journal of Applied Mechanics, Vol. 35(2), pp. 379-386.

Sakano, M., Kishigami, N., Ono, T., and Mikami, I. (1998): Super-Low-Cycle Fatigue

Behavior of Steel Pier Base Joint with Triangular Ribs, Journal of Structural

Engineering, Vol. 44A, pp. 1281-1288 (in Japanese).

Sakano, M. and Wahab, M. A. (2001): Extremely Low Cycle (ELC) Fatigue Cracking

Behavior in Steel Bridge Rigid Frame Piers, Journal of Materials Processing

Technology, Vol. 118, pp. 36-39.

Shahani, A. R., Kashani, H. M., Rastegar, M., and Dehkordi, M. B. (2009): A unified

model for the fatigue crack growth rate in variable stress ratio, Fatigue and Fracture

Engineering Materials and Structures, Vol.32(2), pp.105–118.

Shercliff, H. R. and Fleck N. A. (1990): Effect of Specimen Geometry on Fatigue Crack

Growth in Plane Strain-II. Overload Response, Fatigue and Fracture Engineering

Materials and Structures, Vol. 13, No. 3, pp. 297–310.

Shimada, K., Komotori, J., and Shimizu, M. (1987): The applicability of the Manson–

Coffin law and Miner’s law to extremely low cycle fatigue, Transactions of the Japan

Society of Mechanical Engineers A., Vol. 53(491), pp. 1178–85.

Solanki, K., Daniewicz, S. R., and Newman, J. C. (2004): Finite Element Analysis of

Plasticity-Induced Fatigue Crack Closure: An Overview, Engineering Fracture

Mechanics, Vol. 71, pp. 149–171.

Tanaka, K., Takahashi, H. and Akiniwa, Y. (2006): Fatigue Crack Propagation from A

87
 Hole in Tubular Specimens under Axial and Torsional Loading, International Journal

of Fatigue, Vol. 28, No. 4, pp. 324-334.

Tanaka, K., Akiniwa, Y., Takahashi, A., Mikuriya, T. (2003): Fatigue Crack Propagation

from A Hole in Thin-Walled Tubular Steel Specimens under Torsional-Axial Loading,

Transactions of the Japan Society of Mechanical Engineers, Vol. 69, pp.1001–1008.

Tanuma, Y. and Kobayashi, H. (2002): Study on Ultra Low Cycle Fatigue Crack Growth

Characteristics of High Strength Steel, Journal of Structural and Construction

Engineering, Vol. 67, No. 553, pp. 105–112. (in Japanese).

Tateishi, K., Hanji, T., Minami, K. (2007): A prediction model for extremely low cycle

fatigue strength of structural steel, International Journal of Fatigue, Vol. 29(5), pp.

887–896.

Terao, N., Hanji, T. and Tateishi, K. (2015): Crack Propagation Behavior of Structural

Steel in Extremely Low Cycle Fatigue Region, IABSE Conference NARA.

Terao, N., Hanji, T., Tateishi, K. and Shimizu, M. (2015): A Prediction Method for

Extremely Low Cycle Fatigue Crack Propagation of Structural Steel, Proceedings of

the 8th International Symposium on Steel Structures, pp. 387-388.

Trebules, V. W., Roberts, R., Hertzberg, R. W. (1973): Effects of multiple overloads on

fatigue crack propagation in 2024-T3 aluminum alloy, Progress in Flaw Growth and

Fracture Toughness Testing, ASTM STP 536, Vol. 113.

Usami, T., and Kumar, S. (1996): Damage Evaluation in Steel Box Columns by

Pseudodynamic Tests. Journal of Structural Engineering, Vol. 122, No. 6, pp. 635-

642.

88
APPENDIX A: Digital image of the crack

(a) 0-cycle number (b) 0.5-cycle number

(c) 1.5-cycle number (d) 2.5-cycle number

(e) 3.5-cycle number (f) 4.5-cycle number

Fig. A1 Digital image of CA-1 crack

89
90
APPENDIX B: Crack propagation analysis

Table A1 Crack propagation analysis of CA-1

Test result Estimation
Cycle Crack Crack Crack Crack
ΔJ
Number growth rate length growth rate length
(mm/cycle) (mm) (N/mm) (mm/cycle) (mm)
0 0 0 0 0 0
0.5 0 0 0 0 0
1.5 0.68 0.68 805.11 1.13 1.13
2.5 0.69 1.37 791.20 1.09 2.22
3.5 0.64 2.01 778.27 1.05 3.27
4.5 0.79 2.80 762.40 1.01 4.28
5.5 0.95 3.75 743.29 0.96 5.24
6.5 1.17 4.92 719.80 0.90 6.15
7.5 1.05 5.97 698.62 0.85 6.99
8.5 0.61 6.58 686.43 0.82 7.81
9.5 0.80 7.37 670.44 0.78 8.60
10.5 0.93 8.30 651.80 0.74 9.34
11.5 0.67 8.97 638.25 0.71 10.04
12.5 0.47 9.45 628.75 0.69 10.73
13.5 0.47 9.91 619.34 0.67 11.40
14.5 0.89 10.81 601.42 0.63 12.03
15.5 0.65 11.46 588.33 0.60 12.63
16.5 0.48 11.94 578.60 0.58 13.21
17.5 0.55 12.49 567.49 0.56 13.77
18.5 0.37 12.86 560.07 0.55 14.32
19.5 0.45 13.32 550.93 0.53 14.85
20.5 0.36 13.68 543.61 0.51 15.36
21.5 0.38 14.06 535.96 0.50 15.86
22.5 0.37 14.44 528.45 0.49 16.35
23.5 0.35 14.79 521.39 0.47 16.82
24.5 0.48 15.27 511.66 0.46 17.28
25.5 0.42 15.69 503.19 0.44 17.72
26.5 0.58 16.28 492.85 0.42 18.14
27.5 0.70 16.97 482.40 0.40 18.54
28.5 0.76 17.74 470.90 0.39 18.93
… … … … … …
104.5 0.10 33.40 274.34 0.13 34.09

91
 Table A2 Crack propagation analysis of CA-2
Test result Estimation
Cycle Crack Crack Crack Crack
ΔJ
Number growth rate length growth rate length
(mm/cycle) (mm) (N/mm) (mm/cycle) (mm)
0 0 0 0 0 0
0.5 0 0 0 0 0
1.5 0.68 0.68 805.11 1.13 1.13
2.5 0.69 1.37 791.20 1.09 2.22
3.5 0.64 2.01 778.27 1.05 3.27
4.5 0.79 2.80 762.40 1.01 4.28
5.5 0.95 3.75 743.29 0.96 5.24
6.5 1.17 4.92 719.80 0.90 6.15
7.5 1.05 5.97 698.62 0.85 6.99
8.5 0.61 6.58 686.43 0.82 7.81
9.5 0.80 7.37 670.44 0.78 8.60
10.5 0.93 8.30 651.80 0.74 9.34
11.5 0.67 8.97 638.25 0.71 10.04
12.5 0.47 9.45 628.75 0.69 10.73
13.5 0.47 9.91 619.34 0.67 11.40
14.5 0.89 10.81 601.42 0.63 12.03
15.5 0.65 11.46 588.33 0.60 12.63
16.5 0.48 11.94 578.60 0.58 13.21
17.5 0.55 12.49 567.49 0.56 13.77
18.5 0.37 12.86 560.07 0.55 14.32
19.5 0.45 13.32 550.93 0.53 14.85
20.5 0.36 13.68 543.61 0.51 15.36
21.5 0.38 14.06 535.96 0.50 15.86
22.5 0.37 14.44 528.45 0.49 16.35
23.5 0.35 14.79 521.39 0.47 16.82
24.5 0.48 15.27 511.66 0.46 17.28
25.5 0.42 15.69 503.19 0.44 17.72
26.5 0.58 16.28 492.85 0.42 18.14
27.5 0.70 16.97 482.40 0.40 18.54
28.5 0.76 17.74 470.90 0.39 18.93
29.5 0.44 18.18 464.26 0.38 19.31
30.5 0.64 18.82 454.61 0.36 19.67
… … … … … …
104.5 0.10 33.40 274.34 0.13 34.09

92
 Table A3 Crack propagation analysis of VA-1
Test result Estimation
Crack
Crack Crack Crack
growth rate
Cycle Crack Crack ΔJ growth rate AF Eq. length length
ΔJ (var.
Numb. growth rate length ratio (const.ampl) (4.2) (Without (With
ampl.) Eq.
Eq. (3.2) AF) AF)
(4.3)
(mm/cycle) (mm) (N/mm) - (mm/cycle) - (mm/cycle) (mm) (mm)
0 0 0 0 0 0 0 0 0 0
0.5 0 0 818.69 0 0 0 0 0 0
1.5 1.04 1.04 797.77 0.97 1.11 0.82 0.91 1.11 0.91
2.5 1.26 2.30 772.41 0.97 1.04 1.08 1.12 2.15 2.03
3.5 0.44 2.74 276.88 0.36 0.13 2.61 0.35 2.28 2.38
4.5 0.40 3.14 273.64 0.99 0.13 2.50 0.33 2.41 2.71
5.5 0.35 3.49 270.80 0.99 0.13 2.26 0.29 2.54 3.00
6.5 0.95 4.44 729.41 2.69 0.93 1 0.93 3.46 3.92
7.5 0.86 5.30 712.09 0.98 0.88 0.93 0.82 4.34 4.74
8.5 0.80 6.11 695.91 0.98 0.84 0.93 0.78 5.19 5.53
9.5 0.47 6.58 245.53 0.35 0.10 2.70 0.28 5.29 5.81
10.5 0.25 6.82 243.50 0.99 0.10 2.17 0.22 5.40 6.04
11.5 0.36 7.19 240.53 0.99 0.10 3.27 0.33 5.50 6.36
12.5 1.21 8.40 649.79 2.70 0.73 1 0.73 6.23 7.10
13.5 1.13 9.53 627.17 0.97 0.68 1.71 1.17 6.92 8.27
14.5 1.06 10.59 605.77 0.97 0.64 1.79 1.14 7.55 9.41
15.5 0.47 11.06 208.89 0.34 0.08 2.83 0.21 7.63 9.63
16.5 0.46 11.51 205.14 0.98 0.07 6.38 0.47 7.70 10.09
17.5 0.40 11.92 201.86 0.98 0.07 5.85 0.42 7.77 10.51
18.5 0.87 12.79 561.55 2.78 0.55 1 0.55 8.32 11.06
19.5 1.08 13.86 539.93 0.96 0.51 2.47 1.25 8.83 12.31
20.5 0.98 14.85 520.16 0.96 0.47 2.51 1.18 9.30 13.49
21.5 0.34 15.19 175.10 0.34 0.05 2.97 0.16 9.35 13.65
22.5 0.28 15.47 172.83 0.99 0.05 6.24 0.32 9.41 13.97
23.5 0.18 15.64 171.39 0.99 0.05 4.08 0.21 9.46 14.18
24.5 0.67 16.31 492.33 2.87 0.42 1 0.42 9.88 14.60
25.5 0.43 16.74 485.88 0.99 0.41 1.01 0.42 10.29 15.02
26.5 0.23 16.97 482.46 0.99 0.41 0.55 0.22 10.69 15.24
27.5 0.15 17.12 161.36 0.33 0.05 3.04 0.14 10.74 15.38
28.5 0.18 17.30 160.23 0.99 0.04 3.88 0.17 10.78 15.56
29.5 0.11 17.41 159.52 1.00 0.04 2.44 0.11 10.83 15.66
… … … … … … … … … …
101.5 0.01 27.24 104.61 1.00 0.02 0.58 0.01 21.35 26.06

93
 Table A4 Crack propagation analysis of VA-2
Test result Estimation
Crack
Crack Crack Crack
growth rate
Cycle Crack Crack ΔJ growth rate AF Eq. length length
ΔJ (var.
Numb. growth rate length ratio (const.ampl) (4.2) (Without (With
ampl.) Eq.
Eq. (3.2) AF) AF)
(4.3)
(mm/cycle) (mm) (N/mm) - (mm/cycle) - (mm/cycle) (mm) (mm)
0 0 0 0 0 0 0 0 0 0
0.5 0 0 818.69 0 0 0 0 0 0
1.5 1.03 1.03 798.07 0.97 1.11 0.81 0.90 1.11 0.90
2.5 1.25 2.27 773.00 0.97 1.04 1.07 1.11 2.15 2.01
3.5 1.06 3.34 751.61 0.97 0.98 0.99 0.97 3.13 2.98
4.5 1.06 4.39 730.38 0.97 0.93 1.06 0.99 4.06 3.96
5.5 0.88 5.27 712.74 0.98 0.88 0.95 0.84 4.94 4.80
6.5 0.29 5.56 253.84 0.36 0.11 2.67 0.30 5.06 5.10
7.5 0.29 5.85 251.48 0.99 0.11 2.31 0.25 5.17 5.36
8.5 0.29 6.13 249.14 0.99 0.11 2.34 0.25 5.27 5.61
9.5 0.26 6.40 246.99 0.99 0.11 2.21 0.23 5.38 5.84
10.5 0.25 6.64 244.97 0.99 0.10 2.12 0.22 5.48 6.07
11.5 0.22 6.87 243.13 0.99 0.10 1.97 0.20 5.59 6.27
12.5 0.75 7.62 665.49 2.74 0.77 1 0.77 6.36 7.04
13.5 0.70 8.32 651.41 0.98 0.74 0.97 0.72 7.10 7.76
14.5 0.70 9.02 637.27 0.98 0.71 1.04 0.73 7.80 8.49
15.5 0.67 9.69 623.82 0.98 0.68 1.05 0.71 8.48 9.20
16.5 0.65 10.34 610.84 0.98 0.65 1.07 0.70 9.13 9.90
17.5 0.55 10.89 599.80 0.98 0.63 0.96 0.60 9.75 10.50
18.5 0.27 11.16 208.07 0.35 0.08 2.82 0.21 9.83 10.71
19.5 0.18 11.34 206.56 0.99 0.07 2.55 0.19 9.90 10.90
20.5 0.29 11.63 204.21 0.99 0.07 4.09 0.30 9.98 11.20
21.5 0.32 11.95 201.56 0.99 0.07 4.75 0.34 10.05 11.53
22.5 0.30 12.25 199.11 0.99 0.07 4.56 0.31 10.12 11.85
23.5 0.21 12.47 197.37 0.99 0.07 3.34 0.23 10.18 12.08
24.5 0.77 13.24 552.51 2.80 0.53 1 0.53 10.72 12.61
25.5 0.67 13.91 538.96 0.98 0.51 1.58 0.80 11.22 13.40
26.5 0.69 14.61 525.00 0.97 0.48 1.75 0.84 11.70 14.24
27.5 0.77 15.37 509.60 0.97 0.45 2.08 0.94 12.15 15.18
28.5 0.68 16.05 496.17 0.97 0.43 1.96 0.84 12.58 16.02
29.5 0.69 16.74 485.86 0.98 0.41 1.61 0.66 12.99 16.68
… … … … … … … … … …
101.5 0.07 27.52 337.11 1.00 0.20 0.36 0.07 23.26 27.73

94
 Table A5 Crack propagation analysis of VA-3
Test result Estimation
Crack
Crack Crack Crack
growth rate
Cycle Crack Crack ΔJ growth rate AF Eq. length length
ΔJ (var.
Numb. growth rate length ratio (const.ampl) (4.2) (Without (With
ampl.) Eq.
Eq. (3.2) AF) AF)
(4.3)
(mm/cycle) (mm) (N/mm) - (mm/cycle) - (mm/cycle) (mm) (mm)
0 0 0 0 0 0 0 0 0 0
0.5 0 0 818.69 0 0 0 0 0 0
1.5 0.18 0.18 815.10 1.00 1.16 0.14 0.16 1.16 0.16
2.5 1.02 1.20 794.66 0.97 1.10 0.81 0.89 2.25 1.05
3.5 0.82 2.02 778.17 0.98 1.05 0.70 0.73 3.31 1.78
4.5 0.93 2.95 759.40 0.98 1.00 0.85 0.85 4.31 2.63
5.5 0.60 3.55 747.32 0.98 0.97 0.57 0.56 5.28 3.19
6.5 1.04 4.59 726.41 0.97 0.92 1.06 0.97 6.20 4.16
7.5 0.97 5.56 706.93 0.97 0.87 1.07 0.93 7.07 5.09
8.5 1.10 6.66 684.81 0.97 0.82 1.32 1.08 7.89 6.17
9.5 0.64 7.30 671.90 0.98 0.79 0.82 0.65 8.67 6.81
10.5 0.53 7.83 435.21 0.65 0.33 1.90 0.63 9.00 7.44
11.5 0.72 8.55 425.08 0.98 0.31 2.28 0.72 9.32 8.16
12.5 0.51 9.06 417.87 0.98 0.30 1.71 0.52 9.62 8.68
13.5 0.27 9.33 414.10 0.99 0.30 0.93 0.28 9.92 8.95
14.5 0.41 9.74 408.34 0.99 0.29 1.46 0.42 10.21 9.38
15.5 0.32 10.06 403.80 0.99 0.28 1.19 0.34 10.49 9.72
16.5 0.74 10.80 393.41 0.97 0.27 2.89 0.78 10.76 10.49
17.5 0.49 11.29 386.46 0.98 0.26 2.05 0.53 11.02 11.03
18.5 0.21 11.50 383.53 0.99 0.26 0.89 0.23 11.28 11.26
19.5 0.18 11.68 380.93 0.99 0.25 0.80 0.20 11.53 11.46
20.5 0.73 12.41 569.22 1.49 0.56 1 0.56 12.09 12.02
21.5 0.72 13.13 554.75 0.97 0.54 1.55 0.83 12.63 12.85
22.5 0.51 13.64 544.47 0.98 0.52 1.17 0.60 13.15 13.46
23.5 0.30 13.94 538.38 0.99 0.50 0.72 0.36 13.65 13.82
24.5 0.35 14.29 531.44 0.99 0.49 0.85 0.42 14.14 14.24
25.5 0.23 14.52 526.78 0.99 0.48 0.59 0.28 14.62 14.52
26.5 0.32 14.84 520.32 0.99 0.47 0.84 0.40 15.10 14.92
27.5 0.33 15.17 513.64 0.99 0.46 0.90 0.41 15.55 15.33
28.5 0.39 15.56 505.87 0.98 0.45 1.09 0.49 16.00 15.82
29.5 0.45 16.01 496.83 0.98 0.43 1.33 0.57 16.43 16.39
… … … … … … … … … …
109.5 0.09 29.70 312.60 1.00 0.17 0.55 0.09 30.18 30.35

95
 Table A6 Crack propagation analysis of VA-4
Test result Estimation
Crack
Crack Crack Crack
growth rate
Cycle Crack Crack ΔJ growth rate AF Eq. length length
ΔJ (var.
Numb. growth rate length ratio (const.ampl) (4.2) (Without (With
ampl.) Eq.
Eq. (3.2) AF) AF)
(4.3)
(mm/cycle) (mm) (N/mm) - (mm/cycle) - (mm/cycle) (mm) (mm)
0 0 0 0 0 0 0 0 0 0
0.5 0 0 818.69 0 0 0 0 0 0
1.5 0.26 0.26 813.47 0.99 1.15 0.20 0.23 1.15 0.23
2.5 0.29 0.55 807.69 0.99 1.14 0.22 0.25 2.29 0.48
3.5 1.43 1.98 778.84 0.96 1.06 1.20 1.26 3.34 1.75
4.5 1.50 3.48 748.62 0.96 0.98 1.40 1.36 4.32 3.11
5.5 0.51 3.99 738.43 0.99 0.95 0.50 0.48 5.27 3.58
6.5 0.77 4.76 723.03 0.98 0.91 0.80 0.73 6.18 4.31
7.5 1.96 6.72 683.65 0.95 0.81 2.30 1.87 6.99 6.18
8.5 0.35 7.06 676.70 0.99 0.80 0.44 0.35 7.79 6.53
9.5 0.61 7.67 664.42 0.98 0.77 0.81 0.62 8.55 7.15
10.5 0.06 7.74 70.60 0.11 0.01 4.48 0.04 8.56 7.19
11.5 0.05 7.79 70.43 1.00 0.01 5.60 0.05 8.57 7.24
12.5 0.04 7.83 70.31 1.00 0.01 4.16 0.04 8.58 7.27
13.5 0.04 7.87 70.19 1.00 0.01 3.91 0.03 8.59 7.30
14.5 0.03 7.90 70.09 1.00 0.01 3.41 0.03 8.60 7.33
15.5 0.03 7.93 70.00 1.00 0.01 3.02 0.03 8.61 7.36
16.5 0.03 7.95 69.92 1.00 0.01 2.77 0.02 8.61 7.38
17.5 0.02 7.98 69.85 1.00 0.01 2.56 0.02 8.62 7.40
18.5 0.02 8.00 69.78 1.00 0.01 2.39 0.02 8.63 7.43
19.5 0.02 8.02 69.72 1.00 0.01 2.22 0.02 8.64 7.44
20.5 0.13 8.15 654.81 9.39 0.75 1 0.75 9.39 8.19
21.5 0.60 8.75 642.77 0.98 0.72 0.87 0.62 10.10 8.81
22.5 0.69 9.44 628.87 0.98 0.69 1.06 0.73 10.79 9.54
23.5 0.81 10.25 612.53 0.97 0.65 1.33 0.87 11.45 10.41
24.5 0.96 11.22 593.19 0.97 0.61 1.71 1.05 12.06 11.46
25.5 0.92 12.13 574.78 0.97 0.57 1.78 1.02 12.63 12.49
26.5 0.27 12.40 569.34 0.99 0.56 0.55 0.31 13.20 12.80
27.5 0.22 12.63 564.82 0.99 0.56 0.47 0.26 13.75 13.06
28.5 0.34 12.97 557.99 0.99 0.54 0.73 0.40 14.29 13.46
29.5 0.43 13.40 549.30 0.98 0.53 0.97 0.51 14.82 13.96
… … … … … … … … … …
119.5 0.00 29.40 23.71 1.00 0.00 0.38 0.00 26.01 30.34

96
 Table A7 Crack propagation analysis of RA-1
Test result Estimation
Crack
Crack Crack Crack
growth rate
Cycle Crack Crack ΔJ growth rate AF Eq. length length
ΔJ (var.
Numb. growth rate length ratio (const.ampl) (4.2) (Without (With
ampl.) Eq.
Eq. (3.2) AF) AF)
(4.3)
(mm/cycle) (mm) (N/mm) - (mm/cycle) - (mm/cycle) (mm) (mm)
0 0 0 0 0 0 0 0 0 0
0.5 0 0 754.74 0 0 0 0 0 0
1.5 1.05 1.05 917.95 1.22 1.47 1 1.47 1.47 1.47
2.5 1.69 2.74 1130.35 1.23 2.22 1 2.22 3.69 3.69
3.5 0.64 3.38 596.81 0.53 0.62 1.89 1.17 4.31 4.86
4.5 0.34 3.72 418.89 0.70 0.31 1.87 0.57 4.61 5.43
5.5 1.59 5.31 947.28 2.26 1.56 1 1.56 6.18 7.00
6.5 0.21 5.52 262.77 0.28 0.12 2.85 0.34 6.30 7.34
7.5 1.83 7.34 1019.76 3.88 1.81 1 1.81 8.11 9.15
8.5 1.15 8.49 687.34 0.67 0.82 1.66 1.37 8.93 10.52
9.5 0.26 8.75 313.60 0.46 0.17 2.34 0.40 9.10 10.92
10.5 0.25 9.00 292.86 0.93 0.15 1.99 0.30 9.25 11.21
11.5 0.27 9.27 320.87 1.10 0.18 1 0.18 9.43 11.39
12.5 0.17 9.44 280.70 0.87 0.14 2.02 0.28 9.56 11.67
13.5 0.22 9.66 291.15 1.04 0.15 1 0.15 9.71 11.82
14.5 0.17 9.83 256.68 0.88 0.11 2.07 0.24 9.83 12.05
15.5 1.18 11.01 604.73 2.36 0.64 1 0.64 10.46 12.69
16.5 1.18 12.19 532.22 0.88 0.49 1.68 0.83 10.96 13.52
17.5 1.09 13.27 645.41 1.21 0.72 1 0.72 11.68 14.24
18.5 0.27 13.54 275.27 0.43 0.13 2.47 0.33 11.81 14.57
19.5 0.21 13.76 197.84 0.72 0.07 2.31 0.16 11.88 14.73
20.5 0.94 14.69 575.66 2.91 0.58 1 0.58 12.46 15.30
21.5 0.40 15.10 355.41 0.62 0.22 2.04 0.45 12.68 15.75
22.5 0.39 15.48 313.42 0.88 0.17 1.95 0.33 12.85 16.08
23.5 0.24 15.72 203.55 0.65 0.07 2.35 0.17 12.92 16.25
24.5 0.23 15.95 181.08 0.89 0.06 2.31 0.13 12.98 16.38
25.5 0.25 16.20 247.91 1.37 0.11 1 0.11 13.08 16.49
26.5 0.57 16.78 470.27 1.90 0.38 1 0.38 13.47 16.88
27.5 0.25 17.03 248.88 0.53 0.11 2.36 0.25 13.58 17.13
28.5 0.22 17.25 195.93 0.79 0.07 2.28 0.15 13.64 17.28
29.5 0.19 17.43 227.83 1.16 0.09 1 0.09 13.73 17.37
… … … … … … … … … …
119.5 0.05 32.73 91.57 0.65 0.01 2.99 0.04 25.51 33.36

97
 Table A8 Crack propagation analysis of RA-2
Test result Estimation
Crack
Crack Crack Crack
growth rate
Cycle Crack Crack ΔJ growth rate AF Eq. length length
ΔJ (var.
Numb. growth rate length ratio (const.ampl) (4.2) (Without (With
ampl.) Eq.
Eq. (3.2) AF) AF)
(4.3)
(mm/cycle) (mm) (N/mm) - (mm/cycle) - (mm/cycle) (mm) (mm)
0 0 0 0 0 0 0 0 0 0
0.5 3.44 3.44 1254.65 0 2.74 1 2.74 2.74 2.74
1.5 2.28 5.72 1068.64 0.85 1.99 1.40 2.78 4.73 5.52
2.5 0.23 5.95 407.35 0.38 0.29 2.34 0.68 5.01 6.20
3.5 2.02 7.97 1310.48 3.22 2.99 1 2.99 8.00 9.19
4.5 0.81 8.78 475.59 0.36 0.39 2.31 0.91 8.40 10.09
5.5 2.60 11.38 1162.13 2.44 2.35 1 2.35 10.75 12.44
6.5 1.97 13.35 995.44 0.86 1.72 1.42 2.46 12.47 14.90
7.5 0.74 14.10 476.06 0.48 0.39 2.08 0.82 12.87 15.72
8.5 2.74 16.84 1254.42 2.63 2.74 1 2.74 15.60 18.46
9.5 1.18 18.02 475.21 0.38 0.39 2.27 0.89 16.00 19.35
10.5 1.16 19.18 726.08 1.53 0.92 1 0.92 16.91 20.27
11.5 0.97 20.15 605.07 0.83 0.64 1.63 1.04 17.55 21.30
12.5 0.49 20.65 225.04 0.37 0.09 2.71 0.24 17.64 21.54
13.5 0.61 21.26 419.06 1.86 0.31 1 0.31 17.94 21.85
14.5 0.68 21.94 812.02 1.94 1.15 1 1.15 19.09 23.00
15.5 0.20 22.14 132.14 0.16 0.03 3.70 0.11 19.12 23.11
16.5 0.70 22.84 762.94 5.77 1.01 1 1.01 20.13 24.12
17.5 0.35 23.19 162.84 0.21 0.05 3.37 0.16 20.18 24.28
18.5 0.46 23.65 340.91 2.09 0.20 1 0.20 20.38 24.48
19.5 0.48 24.14 257.67 0.76 0.12 2.12 0.24 20.50 24.72
20.5 0.40 24.53 209.04 0.81 0.08 2.23 0.17 20.57 24.89
21.5 0.36 24.89 182.92 0.88 0.06 2.30 0.13 20.63 25.03
22.5 0.16 25.05 118.25 0.65 0.02 2.77 0.07 20.66 25.09
23.5 0.44 25.49 358.16 3.03 0.22 1 0.22 20.88 25.32
24.5 0.46 25.95 381.77 1.07 0.25 1 0.25 21.13 25.57
25.5 0.47 26.42 612.91 1.61 0.65 1 0.65 21.79 26.22
26.5 0.61 27.03 936.69 1.53 1.53 1 1.53 23.31 27.75
27.5 0.31 27.34 161.54 0.17 0.05 3.51 0.16 23.36 27.91
28.5 0.34 27.68 301.35 1.87 0.16 1 0.16 23.52 28.07
29.5 0.34 28.02 253.64 0.84 0.11 2.09 0.23 23.63 28.30
… … … … … … … … … …
119.5 0.06 43.81 155.77 0.81 0.04 2.44 0.10 36.64 44.56

98
 Table A7 Crack propagation analysis of RA-3
Test result Estimation
Crack
Crack Crack Crack
growth rate
Cycle Crack Crack ΔJ growth rate AF Eq. length length
ΔJ (var.
Numb. growth rate length ratio (const.ampl) (4.2) (Without (With
ampl.) Eq.
Eq. (3.2) AF) AF)
(4.3)
(mm/cycle) (mm) (N/mm) - (mm/cycle) - (mm/cycle) (mm) (mm)
0 0 0 0 0 0 0 0 0 0
0.5 0.00 0.00 1128.47 0 0.00 0 0.00 0.00 0.00
1.5 2.32 2.32 1445.70 1.28 3.64 1 3.64 3.64 3.64
2.5 1.15 3.46 730.49 0.51 0.93 1.84 1.71 4.57 5.35
3.5 3.30 6.77 1841.15 2.52 5.90 1 5.90 10.46 11.24
4.5 0.81 7.58 339.73 0.18 0.20 2.98 0.60 10.66 11.84
5.5 1.10 8.68 486.12 1.43 0.41 1 0.41 11.08 12.25
6.5 1.08 9.75 697.51 1.43 0.85 1 0.85 11.92 13.10
7.5 0.94 10.69 521.34 0.75 0.47 1.74 0.82 12.39 13.92
8.5 1.27 11.96 620.08 1.19 0.67 1 0.67 13.06 14.59
9.5 1.21 13.18 583.16 0.94 0.59 1.63 0.96 13.66 15.55
10.5 1.56 14.74 966.24 1.66 1.62 1 1.62 15.28 17.18
11.5 1.25 15.99 519.25 0.54 0.47 1.94 0.91 15.75 18.09
12.5 1.74 17.73 1093.19 2.11 2.08 1 2.08 17.83 20.17
13.5 1.68 19.41 1337.02 1.22 3.11 1 3.11 20.94 23.28
14.5 0.70 20.10 369.22 0.28 0.24 2.66 0.63 21.18 23.91
15.5 0.49 20.59 200.41 0.54 0.07 2.48 0.17 21.25 24.08
16.5 0.99 21.58 489.13 2.44 0.42 1 0.42 21.66 24.50
17.5 0.69 22.28 384.06 0.79 0.26 1.87 0.48 21.92 24.98
18.5 1.11 23.39 603.15 1.57 0.63 1 0.63 22.55 25.61
19.5 0.59 23.98 224.99 0.37 0.09 2.71 0.24 22.64 25.85
20.5 0.57 24.55 224.08 1.00 0.09 2.15 0.19 22.73 26.04
21.5 1.54 26.09 1033.76 4.61 1.86 1 1.86 24.59 27.90
22.5 0.76 26.84 314.99 0.30 0.17 2.67 0.46 24.76 28.36
23.5 1.23 28.08 586.44 1.86 0.60 1 0.60 25.36 28.96
24.5 1.48 29.56 882.97 1.51 1.36 1 1.36 26.71 30.31
25.5 1.30 30.87 707.93 0.80 0.87 1.57 1.37 27.59 31.69
26.5 1.27 32.13 580.25 0.82 0.59 1.65 0.97 28.17 32.66
27.5 1.60 33.73 976.60 1.68 1.66 1 1.66 29.83 34.32
28.5 1.47 35.20 780.96 0.80 1.06 1.54 1.63 30.89 35.95
29.5 1.11 36.31 404.12 0.52 0.28 2.10 0.60 31.18 36.54
… … … … … … … … … …
119.5 0.21 55.43 443.22 1.22 0.34 1 0.34 47.24 56.03

99
100