0 Голоса «за»0 Голоса «против»

Просмотров: 3244 стр.Mar 24, 2019

© © All Rights Reserved

PDF, TXT или читайте онлайн в Scribd

© All Rights Reserved

Просмотров: 3

© All Rights Reserved

- Creep and Fatigue
- Din743 Intro English
- 2004-4052S1_12_Morris
- 1979030
- LifeFrequency Manual
- 2_1080wwwww
- Stress Strain Stuff
- Full Text
- Cycling Loading
- Lecture 8 Fatigue 2 2011
- Sample Paper Wind Eng g
- 2 Introduction RC
- A 344374
- Introduction to the Mechanical Behavior of Metals
- 12 macuta 5
- Flaws Cg 2006
- 06552600.pdf
- cyclic_loading indefinite.pdf
- Lecture Fatigue.pdf
- Efm 2011 Luke Rwa

Вы находитесь на странице: 1из 244

and

Fracture Mechanics

Edited by

Dariusz Skibicki

Fatigue Failure

and

Fracture Mechanics

Conference on XXIV Symposium on

Fatigue Failure and Fracture Mechanics,

May 22-25, 2012, Bydgoszcz-Pieczyska, Poland

Edited by

Dariusz Skibicki

Copyright 2012 Trans Tech Publications Ltd, Switzerland

All rights reserved. No part of the contents of this publication may be reproduced or

transmitted in any form or by any means without the written permission of the

publisher.

Kreuzstrasse 10

CH-8635 Durnten-Zurich

Switzerland

http://www.ttp.net

Volume 726 of

Materials Science Forum

ISSN 1662-9760

Trans Tech Publications Ltd Trans Tech Publications Inc.

Kreuzstrasse 10 PO Box 699, May Street

CH-8635 Durnten-Zurich Enfield, NH 03748

Switzerland USA

Phone: +1 (603) 632-7377

Fax: +41 (44) 922 10 33 Fax: +1 (603) 632-5611

e-mail: sales@ttp.net e-mail: sales-usa@ttp.net

Preface

The conference, one of the most important Polish conferences in the field of materials

science and engineering, is organized by the Polish Academy of Sciences and the University

of Technology and Life Sciences in Bydgoszcz. The scientific meetings have been held

every second year since 1971. The founder of the conference is the author of many

monographs, articles and manuals in the field of Fatigue, Failure and Fracture Mechanics –

Prof. Stanisław Kocańda.

The subjects covered in the conference include fatigue life of elements of structures,

research on low and high-cyclic fatigue, fatigue in conditions of complex states of stress and

strain, analysis of fatigue loads, design taking fatigue into consideration, analysis of random

fatigue loads, experimental methods in mechanics of cracking, application of fracture

mechanics in engineering issues, the impact of structural, technological and functional

factors on fatigue of elements of structures, research methods and equipment.

Table of Contents

Preface

G. Golański, S. Mroziński and K. Werner 3

Experimental Verification of the Analytical Method for Estimated S-N Curve in Limited

Fatigue Life

P. Strzelecki and J. Sempruch 11

Fatigue Life Calculation in Conditions of Wide Spectrum Random Loadings – The

Experimental Verification of a Calculation Algorithm on the Example of 41Cr4 Steel

B. Ligaj and G. Szala 17

The Fictitious Radius as a Tool for Fatigue Life Estimation of Notched Elements

G. Robak, M. Szymaniec and T. Łagoda 27

Determination of Fatigue Life on the Basis of Experimental Fatigue Diagrams under

Constant Amplitude Load with Mean Stress

A. Niesłony and M. Böhm 33

Applying a Stepwise Load for Calculation of the S-N Curve for Trabecular Bone Based on

the Linear Hypothesis for Fatigue Damage Accumulation

T. Topoliński, A. Cichański, A. Mazurkiewicz and K. Nowicki 39

Concept of Fatigue for Determining Characteristics of Materials with Strengthening

E. Marcisz, A. Niesłony and T. Łagoda 43

D. Boroński 51

Effect of Microstructure on Rolling Contact Fatigue of Bearings

T.Z. Woźniak, J. Jelenkowski, K. Rozniatowski and Z. Ranachowski 55

Determination of the Fatigue Properties of Aluminum Alloy Using Mini Specimen

T. Tomaszewski and J. Sempruch 63

Description of Cyclic Properties of Steel in Variability Conditions of Mean Values and

Amplitudes of Loading Cycles

G. Szala and B. Ligaj 69

The Comparison of Cyclic Properties of X5CrNi18-10 Steel in the Range of Low-Cycle

Fatigue in Conditions of Stress and Strain Control

B. Ligaj and G. Szala 77

Method of Determining the Initial Stiffness Modulus for Trabecular Bone under Stepwise

Load

T. Topoliński, A. Cichański, A. Mazurkiewicz and K. Nowicki 84

C. Goss and P. Marecki 93

Influence of the Notch Rounding Radius on Estimating the Elastic Notch Stress

Concentration Factor in a Laser Welded Tee Joint

K. Niklas and J. Kozak 100

Fatigue Life Tests of Explosively Cladded Steel-Titanium Bimetal

A. Kurek and A. Niesłony 106

Simulation of Tensile Test of the 1/2Y Welded Joint Made of Ultra High Strength Steel

J. Gałkiewicz 110

b Fatigue Failure and Fracture Mechanics

Ł. Blacha and A. Karolczuk 118

Residual Stresses in Steel-Titanium Composite Manufactured by Explosive Welding

A. Karolczuk, K. Kluger, M. Kowalski, F. Żok and G. Robak 125

The Impact of the Laser Welding Speed on the Mechanical Properties of Joints in

Multilayer Pipes

S. Mroziński and M. Piotrowski 133

J. Okrajni and G. Junak 143

Influence of Temperature on the Cyclic Properties of Martensitic Cast Steel

S. Mrozinski and R. Skocki 150

Use of Thermography for the Analysis of Strength Properties of Mini-Specimens

A. Lipski and D. Boroński 156

Variations of the Specimen Temperature Depending on the Pattern of the Multiaxial Load –

Preliminary Research

A. Lipski and D. Skibicki 162

Delivered and in the Annealed Condition

D. Skibicki, J. Sempruch and Ł. Pejkowski 171

Estimation of Fatigue Life of Materials with Out-of-Parallel Fatigue Characteristics under

Block Loading

M. Kurek and T. Łagoda 181

Criteria Evaluation for Fatigue Life Estimation under Proportional and Non-Proportional

Loadings

Ł. Pejkowski and D. Skibicki 189

A. Neimitz 195

Fatigue Crack Growth Rates of S235 and S355 Steels after Friction Stir Processing

D. Kocańda, V. Hutsaylyuk, T. Slezak, J. Torzewski, H. Nykyforchyn and V. Kyryliv 203

An Experimental Investigation on Crack Initiation and Growth in Aircraft Fuselage

Riveted Lap Joints

A. Skorupa, M. Skorupa, T. Machniewicz and A. Korbel 211

Application of Digital Image Correlation in Fatigue Crack Analysis

T. Marciniak, Z. Lutowski, S. Bujnowski, D. Boroński and T. Giesko 218

Dual-Band Experimental System For Subsurface Cracks Testing

T. Marciniak, Z. Lutowski, S. Bujnowski, D. Boroński and P. Czajka 222

Dual-Camera Vision System for Fatigue Monitoring

T. Giesko 226

Modeling of Crack Growth in Steels

J. Jackiewicz 233

CHAPTER 1:

Fatigue Life Prediction

© (2012) Trans Tech Publications, Switzerland

doi:10.4028/www.scientific.net/MSF.726.3

1

Czestochowa University of Technology, Armii Krajowej 19, 42 – 200 Czestochowa, Poland

2

University of Technology and Life Sciences in Bydgoszcz, Kaliskiego 7, 85-791 Bydgoszcz,

Poland

3

Czestochowa University of Technology, Akademicka 3, 42 – 200 Czestochowa, Poland

a

grisza@wip.pcz.pl, bstmpkm@utp.edu.pl, cwerner@ipp.pcz.pl

Keywords: low cycle fatigue, ageing, high – chromium cast steel, cyclic properties

Abstract. The paper presents the results of research on fatigue life of GX12CrMoVNbN9 – 1

(GP91) cast steel in delivery state and after ageing at the temperature of 600 oC. The fatigue of low

cycle scope was performed at the temperature of 600 oC at the strain amplitude level of

εac = 0.25 ÷ 0.60%. Analysis of the tests carried out has shown that the examined cast steel is

characterized by strong cyclic softening without an evident period of stabilization. Ageing of GP91

cast steel leads to the fatigue life reduction by ca. 7 to ca. 35%, depending on the level of strain

amplitude εac in comparison with the delivery state.

Introduction

The GX12CrMoVNbN9 – 1 cast steel belongs to a new group of high-chromium martensitic

materials introduced to the power industry in relation to raising the service parameters of power

units. This cast steel, whose functional properties are higher than those of previously used low alloy

cast steels, was developed on the basis of chemical composition of X10CrMoVNbN9 – 1 steel [1,

2]. During service, the power units components made of steel castings are exposed to the effect of

changeable thermal stresses due to start-ups and shut-downs of power units, as welll as mechanical

stresses which often exceed the value of yield strength. Repeated cyclic effect of temperature and

load contributes to the occurrence of deformations and fractures of fatigue character after a certain

number of cycles. Damages, and in extreme cases - complete failures of the castings developing this

way, are connected with low cycle fatigue. Damages of massive multi-ton steel castings resulting

from low cycle fatigue constitute ca. 65% of all damages in steam turbines [3].

The basic requirement put for high-temperature creep resisting steels/cast steels used in the

power industry is retaining stable microstructure for long service time, and thus maintaining certain

mechanical properties. The influence of temperature and time, and also stress in the creep

conditions, is a cause of changes in microstructure of high-chromium cast steels, which also entails

changes in their properties, including cyclic properties [4, 5].

Nowadays the tendency, both: home and abroad, is to aim for creating comprehensive

characteristics which determine the usefulness and potential possibility of using modern high-

temperature creep resisting steels and cast steels. For this purpose, proper characteristics are

necessary, such as fatigue characteristics of new grades of steels and cast steels which determine

gradual reduction of properties during the service [6 – 8]. The paper is to present and compare the

results of research on fatigue life within low cycle scope for GX12CrMoVNbN9 – 1 cast steel in

delivery state and after 8000 hours of ageing at the temperature of 600 oC.

The material under research was high-chromium martensitic GX12CrMoVNb9-1 (GP91) cast steel

of the following chemical composition (%mass): 0.12C; 0.47Mn; 0.31Si; 0.014P; 0.004S; 8.22Cr;

0.90Mo; 0.12V; 0.07Nb; 0.04N. The tests were carried out on test pieces in delivery state, i.e. after

4 Fatigue Failure and Fracture Mechanics

heat treatment (1040 oC/12h/oil + 760 oC/12h/air + 750 oC/8h/furnace) and after 8000 hours of

ageing at the temperature of 600 oC. The test pieces were exposed to low cycle fatigue at the

temperature of 600 oC at five levels of total strain amplitude εac = 0.25; 0.30; 0.35; 0.50; 0.60 %.

Description of methodology that was followed during the research on fatigue life, within low cycle

scope at elevated temperature, is presented inter alia in the works [9, 10]. The examined cast steel in

delivery state, as well as after the process of ageing, was characterized by a microstructure typical

for this grade of materials, that is the microstructure of high-temperature tempered martensite of

elongated subgrains whose shape was inherited from lath martensite with numerous precipitates. On

grain boundaries of prior austenite and on subgrain boundaries, the M23C6 carbides were

precipitated. Inside the subgrains, numerous precipitations of the MX type were observed. Examples

of microstructure of the examined cast steel in delivery state and after the ageing process are

illustrated in Fig 1.

a) b)

Fig 1. Microstructure of the cast steel: a) delivery state, b) after ageing process

Table 1 shows the characteristics of dislocation microstructure of the examined cast steel

(dislocation density, mean diameter of subgrains and shape coefficient) in delivery state and after

8000 hours of ageing at the temperature of 600 oC.

The analysis carried out by means of transmission electron microscope have revealed that as a result

of the ageing process at the temperature of 600 oC, in GP91 cast steel there is a decrease of

dislocation density and an increase in the subgrain diameters. Changes in the dislocation

microstructure are connected with the processes of recovery and polygonization of the matrix

running at elevated temperature, as a result of aiming to obtain dislocation microstructure of lower

energy.

Table 1. Quantitative parameters of dislocation microstructure of GP91 cast steel

Ageing Mean diameter Dislocation

Temperature, Shape

time, of subgrains, density,

[oC] coefficient

[h] [µm] [1014 m-2]

delivery state --- 0.708 ± 0.258 0.701 ± 0.151 2.95 ± 2.02

600 8000 0.868 ± 0.249 0.660 ± 0.173 1.59 ± 1.58

Static tests

The fatigue tests of low cycle GP91 cast steel in delivery state and after the process of ageing were

preceded by the static tests of tension performed at two temperatures. Mechanical properties of the

investigated cast steel obtained at the above-mentioned temperatures are given in Table 2.

Dariusz Skibicki 5

Table 2. Mechanical properties of GP91 cast steel at room temperature and at 600 °C, in delivery

state (sw) and after ageing (ps)

Temperature, [°C]

Parameter 20 600 600

(sw) (sw) (ps)

YS, [MPa] 503 303 297

TS, [MPa] 663 338 331

El., [%] 38.3 63.5 63.8

E, [MPa] 206870 150120 142180

Performed tests have shown that the process of ageing of the investigated cast steel during 8000

hours at the temperature of 600 °C does not have a significant influence on the basic parameters

determined by the static tensile test. Whilst a growth of the temperature of the cast steel testing

results in the occurrence of plastic strains, as the stresses decrease (there is a fall of the values of

stresses of yield strength and tensile strength). At the same time, there is an increase in elongation

from 38.3% to 63.5% (63.8% after ageing).

σ, MPa 350 1 σ, MPa 350 1

2 2

3 3

0.25 0

0

-0.3 ε, % 0.3 -0.3 ε, % 0.3

-350 -350

σ, MPa 350 1 σ, MPa 350 1

2 2

3 3

0.35 0 0

-0.5 ε, % 0.5 -0.5 ε, % 0.5

-350 -350

σ, MPa 350 1

2

σ, MPa 350 1

2

3 3

0.60 0 0

-0.8 ε, % 0.8 -0.8 ε, % 0.8

-350 -350

Fig. 2. Examples of hysteresis loops for the cast steel in delivery state (sw) and after ageing process

(ps) obtained at the temperature of 600 °C: 1 – loop at the first cycle; 2 – loop at half-life cycle; 3 –

loop for the last cycle

Fatigue tests

Analysis of the cyclic properties of GX12CrMoVNbN9 – 1 cast steel, in delivery state and after the

process of ageing in the conditions of changing load, was performed using the most important

parameters of hysteresis loop. These parameters include plastic strain amplitude εap and stress

amplitude σa. During fatigue tests, the changes in the basic parameters of hysteresis loop in the

function of the number of stress cycles were observed. These changes were very similar at all levels

of strain. In the courses of hysteresis loop parameters there were three distinguishable stages. Within

these stages, a change in the shape of hysteresis loop in the function of the number of stress cycles

was noticed. The character of these changes at all levels of strain at the temperature of testing was

comparable. Examples of the loops of hysteresis for the strain amplitude εac=0.25%, 0.35% and

0.60% at three different stages of fatigue are shown in Fig. 2.

6 Fatigue Failure and Fracture Mechanics

The process of low cycle fatigue of GP91 cast steel was characterized by strong cyclic

softening. It was recognized by the increase in the width of hysteresis loop ∆εap, with the

simultaneous intense decreasing in the values of the range of stress changes ∆σa. At the following

stages of cyclic strain, there was no period of stabilization of the hysteresis loop parameters

observed for the cast steel, neither in delivery state, nor after ageing (Fig. 3). Cyclic softening of the

examined cast steel proceeded until the moment of a crack occurring in the test pieces. The above

facts prove cyclic exhausting of fatigue life of the investigated cast steel.

a) b)

σ a, M P a σa, MPa

300 300

280 280

260 260

ε a c = 0.60%

240 240

ε a c = 0.50%

220 220

ε a c = 0.35%

200 ε a c = 0.30% 200

ε ac=0.60%

180 ε a c = 0.25% 180 ε ac=0.50%

ε ac=0.35%

160 160 ε ac=0.30%

εac=0.25%

140 140

0 1 0 00 2000 3 0 00 4000 0 500 1000 1500 2000 2500

N u m b e r o f c yc le s N Number of cycles N

Fig. 3. Changes in stress σa as the function of the number of cycles (T=600 °C): a) in delivery state;

b) after ageing process

In order to assess the influence of the strain level on the extent of changes in the loop parameters,

the coefficients of cyclic properties were introduced: δσ – for stress description, δε – for strain

description. Graphical interpretation of δσ coefficient is presented schematically in Fig. 4 [10].

σa , MPa

εac(i)= const

∆σa

σamax

σ amax − σ a min

δσ = ⋅100%

σ a min

σamin

N=1 N

Fig. 4. Graphical interpretation of δσ coefficient

The values of δε coefficient were expressed using analogous dependence. Obtained results of

calculations of coefficients δσ and δε for particular levels of strain amplitude εac and temperature of

testing are presented in Fig. 5.

On the basis of analysis of the graphs included in Fig. 5a and 5b it can be concluded that the

extent of changes in the hysteresis loop parameters in the function of the number of stress cycles for

the cast steel after ageing process is greater than in the case of material as delivered. When

comparing the two analyzed parameters of hysteresis loop (σa, εap), it can be noted that the

amplitude of plastic strain εap is characterized by smaller changes for most of the load sequences

realized in the research. This rule does not apply only for the level of strain εac=0.25% (Fig. 5b).

Smaller scope of changes in δε coefficient (the strain description) also proves that, as regards the

area of low cycle fatigue, it is justified to make calculations of fatigue life with a strain-based

approach.

Dariusz Skibicki 7

a) b)

50 δσ , % T=60 0 °C ( ps) 50 δε , %

T=60 0 °C ( sw)

40 T=600 °C (p s)

40

T=600 °C (sw)

30 30

20 20

10 10

0 0

0.25 0.30 0.35 0.50 0.60 0.25 0.30 0.35 0.50 0.60

εac, % εac, %

Fig. 5. The δσ and δε coefficients for the cast steel in delivery state (sw) and after ageing (ps):

a) δσ , b) δε

Analytical dependence between stress σa and strain εap was also described with Morrow's

equation. Fig. 6 presents the graphs obtained as a result of approximation of the loop parameters (σa

and εap). Because of the lack of a clear stabilization period of cyclic properties, the values of

hysteresis loop parameters necessary for the analytical descriptions were determined from the period

corresponding to half the fatigue life n/N=0.5.

1000

σa, MPa

lg σ a = lg K '+ n' lg ε ap

1

Nr. T, [°C] K’,[MPa] n’ R2

1 600 (sw) 496 0.1384 0.9603

2 600 (ps) 302 0.0631 0.9112

100

0.0001 0.001 0.01

Strain εap, mm·mm-1

Fig. 6. Stress – strain curves of the cast steel behaviour at the 600 °C temperature in delivery state

(sw) and after ageing process (ps)

Analysis of the tests carried out has proved that the process of ageing of GP91 cast steel causes a

significant decrease in the values of both: K’ - coefficient, and n’ - cyclic strain hardening exponent.

Moreover, graph (2), plotted for the material after ageing process, slopes more gently in comparison

with graph (1) for the material in delivery state. This proves the stress amplitude slowly decreasing

as a result of changes running in the microstructure of the aged cast steel. Cyclic softening of the

examined cast steel in delivery state, as well as after the ageing process, observed during the tests of

constant amplitude at the temperature of 600 °C, is also confirmed by the position of curves of

cyclic and static strain. Examples of the curves of static and cyclic strain obtained during the tests

are illustrated in Fig. 7.

8 Fatigue Failure and Fracture Mechanics

a) b)

350 350

σ, MPa 1 σ, MPa 1

250 2 250 2

3 3

150 150

50 50

-0.8 -0.4 -50 0 0.4 0.8 -0.8 -0.4 -50 0 0.4 0.8

ε, % ε, %

-150 -150

-250 -250

Fig. 7. Stress–strain curves of static and cyclic strain: a) test pieces in delivery state (sw), b) test

pieces after ageing (ps), 1- curve of static tension, 2- graph of cyclic strain, 3- loops of hysteresis

Regardless of the temperature of testing, the curves of cyclic strain lie below the curves of static

strain. This proves cyclic softening of the cast steel irrespective of the temperature of testing and

level of cyclic strain.

a) b)

∆ε b c

ac = ε + ε = σ f 2 N + ε ' 2 N

'

0.01 0.01

f ∆ε

ac = ε + ε = σ f

ae ap

E f f

' b c

2 2N + ε ' 2N

ae ap f

2 E f f

Strain ε

εap

Strain ε

εap

εae εae

0.001 0.001

2NT 1000 2NT 10000

1000 10000

Number of strain reversals 2Nf Number of strain reversals 2Nf

Fig. 8. Strain–life curves for GP91 cast steel: a) in delivery state (sw); b) after ageing (ps)

The results obtained in the research were used for plotting the graphs of fatigue life of GP91 cast

steel which was described with the Manson-Coffin-Basquin equation. Fatigue graphs obtained for

the examined cast steel for the temperature of 600 ºC are presented in Fig. 8, while Table 3 contains

the parameters of Manson-Coffin-Basquin equation. Analysis of the achieved characteristics (Fig. 8)

shows that the abscissa 2NT (the point of intersection of two curves εae = f(2Nf) and εap = f(2Nf)) in

both cases lies within the range of small number of cycles and amounts to about 5700 cycles for the

cast steel in delivery state, and 4558 cycles for the aged cast steel (Table 3). This proves that at the

applied levels of strain εac, the process of cyclic strain of the examined cast steel ran with the

dominant role of plastic strain component εap.

Therefore, it can be assumed that for these strain levels εac, the cyclic strain resistance of the

investigated cast steel will mostly depend on its plastic properties. The comparison between fatigue

life of GP91 cast steel in delivery state and after ageing is presented in Fig. 9.

Dariusz Skibicki 9

Table 3. Parameters of cyclic curves of GP91 cast steel at the temperature of 600°C

Temperature, σ’f ε’f b c 2NT

[°C] [MPa] - - - -

600 (sw) 248 2.21 - 0.0222 -0.8514 5700

600 (ps) 227 9.96 - 0.03014 - 1.0669 4558

1800 sw

Number of cycles to failure N

1600 sp

1400

1200

1000

800

600

400

200

0

0.25 0.30 0.35 0.50 0.60

Strain ε ac, %

Fig. 9. Fatigue life of GP91 cast steel at the temperature of 600 oC in delivery state (sw) and after

ageing (ps)

Fatigue life of GP91 cast steel after the process of ageing was lower in comparison with the life of

the cast steel in delivery state, and it was dependent on the strain amplitude εac. This decrease was

insignificant and amounted to around 7% in the area of the biggest strains realized in the research

(εac=0.60%), and it increased along with the fall of strain value (εac=0.25%) to around 35%,

compared to delivery state.

Temperature 600°C (sw) Temperature 600°C (ps)

εac, [%] σa, [MPa], σa,[MPa],

N N

(n/N=0.5) (n/N=0.5)

0.25 3548 193 2291 198

0.30 2505 206 1593 202

0.35 1945 208 1321 204

0.50 947 237 857 208

0.60 683 236 637 218

Table 4 shows the characteristics of low cycle fatigue if the examined cast steel. Due to the lack of a

clear period of stabilization of cyclic properties (Fig. 3), the value of saturation stress σa was

determined from the period corresponding to half the fatigue life (n/N=0.5).

Acknowledgements

Scientific work funded by the Ministry of Education and Science for the years 2010 - 2012 as a research

project No. N N507 510 838.

10 Fatigue Failure and Fracture Mechanics

Conclusions

1. The GX12CrMoVNbN9 – 1 cast steel is characterized by strong cyclic softening without a

clear stabilization period at both states: the delivery one and after ageing.

2. The ageing process of GX12CrMoVNbN9 – 1 cast steel contributes to a significant decrease

in the values of coefficients K’ and n’. It can be a proof of slow decreasing in the level of

stress δa of the investigated cast steel, as a result of changes running in the microstructure of

the aged cast steel through the process of the matrix softening.

3. Ageing of GP91 cast steel causes a decrease in fatigue life at the temperature of testing, and

its level is dependent on the strain amplitude εac. This decrease is slight (around 7%) in the

area of the biggest strains covered in the research (εac=0.60%) and increases as the strain

value falls (εac=0.25%) – ca. 35%, compared to the material in delivery state.

References

[1] G. Golański High chromium cast steels for power, J. Energetyka, (2010) 58 – 61 (in Polish).

[2] H. K. Mayer, H. Cerjak, P. Hofer, E. Letofsky, F. Schuster, Evolution of microstructure and

properties of 10% Cr steel castings, in Microstructural development and stability in high

chromium ferritic power plant steels, A. Strang and D. J. Gooch (Eds.) The Institute of

Materials, London, 1997, pp. 105 – 122.

[3] R. Viswanathan, Damage mechanisms and life assessment of high temperature components,

ASM International, Metals Park Ohio, USA, 1989.

[4] G. Golański, J. Kępa, The Effect of Ageing Temperatures on Microstructure and Mechanical

Properties of GX12CrMoVNbN9 -1 (GP91) Cast Steel, Archives of Metallurgy and Materials,

2012 (in print).

[5] A. Zieliński, J. Dobrzański, G. Golański, Estimation of the residual life of L17HMF cast steel

elements after long – term service, JAMME, 34 2 (2009) 137 – 144.

[6] M. Cieśla, G. Junak, Low - cycle characteristic of the latest generation of creep resistant

martensitic steels and their welded joints, in Materials and Technology for Construction of

Supercritical Boilers and Waste Plants, A. Hernas (Eds.), SITPH Publ., Katowice, 2009, pp.

378 – 399.

[7] M. Cieśla, J. Dobrzański, G. Junak, Utility characteristics of a superheater outlet chamber

material after overrunning the computational service life, J. Energetyka 1 (2012) 28 – 38 (in

Polish).

[8] J. Okrajni, M. Cieśla, K. Mutwil, Power plant component life assessment, Inżynieria

Materiałowa 1 (2005) 15 – 20.

[9] G. Golański, K. Werner, S. Mroziński, Low cycle fatigue of GX12CrMoVNbN9-1 cast steel

at 600 ºC temperature, Advanced Materials Research 396-398 (2012) 326-329.

[10] S. Mroziński, Stabilization of cyclic properties in metals and its influence on fatigue life,

Publ. House of the University of Technology and Life Sciences, Bydgoszcz, Poland 128, 2008

(in Polish).

© (2012) Trans Tech Publications, Switzerland

doi:10.4028/www.scientific.net/MSF.726.11

Curve in Limited Fatigue Life

1

University of Technology and Life Sciences in Bydgoszcz, Faculty of Mechanical Engineering,

ul. Kaliskiego 7, 85-791 Bydgoszcz

a b

p.strzelecki@utp.edu.pl, semjan@utp.edu.pl

Abstract. The paper presents the method of determining the S-N curve based on the static material

properties only. To verify the procedure algorithm, an experiment has been carried out to define the

reference fatigue characteristics of material C45+C. The research was performed with the use of the

rotary-bending test stand made according to own design. The proper operation of the materials

testing machine was verified compliant with the ISO 1143 standard. The analytical method was

verified by making statistical calculations assuming the null hypothesis of the equality of slope

coefficients of the estimated line with the analytical method and the experimental line. The statistics

values calculated have shown that there exist no grounds for rejecting the null hypothesis.

Introduction

In engineering calculations, fatigue life or strength of a design element is determined by making the

calculations based on a specific model dependent on the loading method. The calculations applied

mostly refer to the high-cycle fatigue area. To facilitate such calculations, one shall have fatigue

characteristics at their disposal. Frequently at the preliminary phase of the project, it is not possible

to perform an experiment which would aim at plotting such a curve. To do so, fatigue characteristics

are estimated based on the values of static

Stress amplitude ,σa (log)

technological process, applying analytical

1 methods. Such methods include e.g. the

ZG me FITNET method [1], the method reported in

1 the publication [2] and the algorithm

mD

proposed by the authors [3]. Verification of

these methods can be found in [4] and [5].

The aim of this paper is to verify own

proposal of the analytical method used to

106 108 determine the S-N curve. It is based on the

Life, N, cycles (log) algorithm from FITNET procedures. It has

been described in detail in the publication

Fig. 1 S-N curve according to the method proposed [3], and in a form of a diagram, see Fig 1.

The primary assumptions of the method are

based on the principle that the first step should involve determining the actual static properties of

the material (ultimate tensile strength Rm and yield stress Re). The next step is to define the equation

of the S-N curve according to equations (1), (2) and (3). Additionally an attempt has been made to

verify the method similar in nature published in [1] and [2].

12 Fatigue Failure and Fracture Mechanics

106

log

me N Re

(2)

0,9 Re

log

ZG

10

R

NRe 400 e (3)

Rm

where:

σa – stress amplitude,

N – number of cycles completed,

ZG –fatigue limit; it has been assumed according to the FITNET guidelines [after 1],

N0 – base number of cycles (it has been assumed as 106),

me – slope coefficient.

One shall mention that the line with inclination mD is plotted for the application of the

characteristics for the elements made of austenitic steel, alloys of aluminium and other metals

except for steel and cast steels. The value of slope coefficient mD is 15 for normal stresses and 25

for tangential stresses.

The FITNET method is differentiated from the method proposed by the authors by the method of

defining slope coefficient me the value of which has been determined as 5 for normal stresses and 8

for tangential stresses. The proposal given in the publication [2], on the other hand, as a different

one, defines the fatigue curve by determining the line for the range of limited fatigue life based on

two points. The points define the estimated strength for the fatigue life of 106 and 103 cycles.

Methodology

To verify the above proposals, there were applied data characteristic for the fatigue properties of

material C45 +C (Rm = 826 MPa, Re = 647 MPa). To receive that data, test has been made using the

rotary-bending fatigue test stand. The drawing of the specimen used for the tests is given in Fig. 2.

Interestingly, the materials testing

machine has been made according to own

design the diagram of which is given in

Fig. 3. The tests were made according to

norms [6] and [7].

Essentially, the test stand had been

verified earlier. The test stand verification

involved determining the maximum error

of the bending moment applied. The

calculations of that value were made

compliant with the norm [8] and it was

1.15%. The admissible value here was

1.3%. Additionally there was a

Fig. 2 Specimen. measurement of the correctness of

calculating the number of cycles. To

verify the accuracy of the cycle measurement, the Sentry ST723 tachometer was used (the

measurement accuracy of 0.01%). The maximum measurement error was 0.24%.

Dariusz Skibicki 13

l

Mg

Mo

m·g

m – mass [kg],

g – gravitational

acceleration [m/s2],

l – arm of the load [m].

Mg=m·g·l

Test results

The fatigue test results have been presented in Fig. 4. Besides, in the right lower corner (Fig. 4) you

can find the equation of the experimental line calculated according to the norm [9] and the

coefficient of determination received. The dashed line stands for the confidence interval calculated

for the confidence interval of 95%. Fig. 5 presents an estimated S-N curve according to the method

presented above (the dashed line) together with the line defined experimentally.

logN=-8.058logσ +26.33

a

2

R =0.952

14 Fatigue Failure and Fracture Mechanics

Experimental curve

Estimated curve

Fig. 5 S-N curve C45+C steel and the estimated curve according to the method proposed [3]

Verification method

To verify the method proposed, calculations were made to verify the assumption that the estimated

curve shows the slope coefficients equal with the regression curve received from the experimental

points.

The equations used for the statistical calculations are as follows [10]:

(Y Yˆ ) i i

2

Sa i 1

1 n

n 2

(n 1) X i2 X i

i 1 n i 1

(4)

a a0

ta

Sa (5)

n n

(Y Yˆ ) X

i i

2

i

2

(6)

Sb i 1 i 1

n n

2

(n 2)n X i2 X i

i 1 i 1

b b0

tb (7)

Sb

where:

X=logσ,

Y=logN,

Dariusz Skibicki 15

n – number of specimens,

=a0X+b0 – equation received with the analytical method,

Y=aX+b – equation received in the experiment.

Verification results

There exist no grounds for rejecting the null hypothesis of the equality of the slope coefficients of

the estimated equation from the experimental results and the equation received using the analytical

method when the condition below is met:

Statistics Method FITNET Method reported The results of the

reported in the method in the calculations are

publication [3] given in Table 1. To

publication [2]

Value of coefficient a0 -9.5 -5.0 -11.8 compare the method

Value of coefficient b0 29.9 18.6 36.0 proposed with other

ta -1.516 -1.953 3.585 algorithms, there

tb 1.386 1.810 -3.444 were determined

pvalue (97.5%,ta) 12.6% 6.32% 0.22% curves according to

pvalue (97.5%,tb) 15.0% 8.03% 0.31% the FITNET method

t(97.5%,19) 2.093 and the method

reported in the

publication [2]. For all those algorithms the equation of curves were determined. The coefficients of

those equations are given in Table 1.

To make a quality comparison, Fig. 6 presents the characteristics according to the experiment

(the solid line) and the analytical methods (the dashed line stands for the method reported in the

publication [3], the dotted line stands for the FITNET method, the dot-and-dashed line stands for

the method reported in the publication [2]).

Fig. 6 Diagrams of fatigue curve according to the experimental and analytical method. For the lines details,

see the text.

16 Fatigue Failure and Fracture Mechanics

Conclusions

The calculations have confirmed the assumption that the slope coefficients of the estimated curve

according to own proposal show values non-significantly different from the coefficients of the

experimental curve. A similar assumption can be assumed for the original FITNET method,

however, in that case the calculated values of statistics are close to the statistical critical interval

established at the level of confidence of α=0.05. The result is reflected in the figure presented in Fig.

6. Interestingly, the curve determined according to the method reported in the publication [3] shows

the inclination closer to the experimental curve than the curve received according to the FITNET

method.

In the statistical calculations the method reported in the publication [2] appeared to be worst,

which is connected with the fact that the fatigue limit determined according to that algorithm is

much greater than the one determined according to the FITNET method. The error in determining

that value was 14.3% and 5.3%, respectively.

Acknowledgement

The work has been co-financed by the European Union Social Fund, the state budget of Poland and

the budget of the Kujawsko-Pomorskie Province as part of the project ‘Krok w przyszłość –

stypendia dla doktorantów’, the 4th edition.

References

[1] Neimitz A., Dzioba I., Graba M., Okrajni J., Evaluation of strength, life and safety of structural

components contain defects, Politechnika Świętokrzyska, Kielce 2008, pp. 131-183.

[2] Lee Yung-Li, Pan Jwo, Hathaway R. B., Barkey M. E., Fatigue testing and analysis, University

of Alabama, Elsevier. 2005, pp. 126-140.

[3] Strzelecki P., Sempruch J., Modification of selected methods of rapid determination of fatigue

characteristics in the range of limited fatigue life, Journal of Polish Cimac, Vol. 6 No 3, Gdańsk

2011, pp. 289-296.

[4] Niesłony A., Kurek A., Chalid el Dsoki, Kaufmann H., A study of compatibility two classical

fatigue curve models based on some selected structural materials, Inernational Journal of Fatigue,

Vol. 39, June 2012, pp. 88-94.

[5] Pejkowski Ł., Skibicki D., Analysis accelerated methods for determination of fatigue curves,

Journal of Polish Cimac, Vol. 6 No 3, Gdańsk 2011, pp. 199-214.

[6] PN-H-04326:1976, Test of metal fatigue – Bending tests

[7] PN-H 04325:1976 Test of metal fatigue - Basic terms and general guidelines for preparing of

samples and carry out tests

[8] ISO 1143:2010 Metallic materials, Rotating bar bending fatigue testing.

[9] ASTM E 739-91:2004, Standard Practice for Statistical Analysis of Linear or Linearized Stress-

Life (S-N) and Strain-Life (ε-N) Fatigue Data.

[10] Krysicki W. et al, Probability theory and mathematical statistics in problem, PWN, Warszawa

2000, pp. 184-188.

© (2012) Trans Tech Publications, Switzerland

doi:10.4028/www.scientific.net/MSF.726.17

- the experimental verification of a calculation algorithm on the example

of 41Cr4 steel

1

University of Technology and Life Sciences in Bydgoszcz, Faculty of Mechanical Engineering,

Department of Machine Design, ul. Prof. S. Kaliskiego 7, 85-225 Bydgoszcz

a

email: bogdan.ligaj@utp.edu.pl, b email: grzegorz.szala@utp.edu.pl,

Keywords: random loading, fatigue life of steel, programmed fatigue testings, algorithm of fatigue

calculations.

conditions is connected with proper elaboration of loading spectrum and assumption of a proper

fatigue characteristic. On the base of literature data and own research there has been elaborated an

algorithm for fatigue life calculations in random loading conditions with wide spectrum.

Calculations were performed with the usage of chosen mathematical models of two-parametric

fatigue characteristics. Results calculated with accordance to the described procedure were

validated with experimental test results of specimens made of 41Cr4 steel with a method of

programmed fatigue life tests.

Nomenclature

C(0) – constant in the formula describing Wöhler fatigue diagram for off-zero pulsating load

(R = 0),

C(-1) – constant in the formula describing Wöhler fatigue diagram for oscillating load (R = -1),

I – coefficient characterizing width of random loading spectrum,

N – cycle number – general notation (fatigue life),

Nc exp – number of cycles determined as a result of experimental tests,

Nc cal – number of cycles determined as a result of calculations with the usage of two-parametric

fatigue life characteristic,

Nij – number of cycles to cracking determined on the base of two-parametric fatigue life

characteristic for defined Sai and Smj values,

N0 – base number of cycles corresponding to fatigue life (N0 = 106),

R – cycle asymmetry ratio (R = Smin/Smax),

S – specimen stress – general notation [MPa],

Sa – sinusoidal cycle stress amplitude [MPa],

Si – local stress values on the „i” level of loading [MPa],

Sf (-1) – fatigue limit under oscillating load (R = -1) for N0 cycle number [MPa],

Sf (0) – fatigue limit under pulsating load (R = 0) for N0 cycle number [MPa],

Sm – mean sinusoidal cycle stress [MPa],

Smax – maximum sinusoidal cycle stress [MPa],

Smin – minimum sinusoidal cycle stress [MPa],

ct – constant value in formula describing fatigue life curve,

i – general notation for the loading level (i = 1,2, ...., k),

j – index of mean value of sinusoidal cycle loading (j = 1,2, ...., p),

k – exponent in equation ψN = N-k,

nc – total number of cycles in loading spectrum,

nij – number of cycles in loading spectrum with Sa and i Sm j parameters,

m(-1) – exponent in formula describing Wöhler fatigue diagram for oscillating load (R = -1),

m(0) – exponent in formula describing Wöhler fatigue diagram for pulsating load (R = 0),

18 Fatigue Failure and Fracture Mechanics

mc cal(x) – exponent in formula describing fatigue life curve in service loading conditions,

ψ – factor of material sensitivity to cycle asymmetry, for N = N0,

ψN – factor of material sensitivity to cycle asymmetry, for N ≠ N0,

1. Introduction

There are numerous procedures of fatigue life calculations of structural elements. As an example of

such works there can be mentioned procedures published in 2006 by European Fitness-for-service

Thematic Network know as FITNET [3]. From the analysis of these procedures results that in a

simplified scope there were discussed problems of elaboration of loading spectra and programs of

loading characteristics and determination of fatigue characteristics for cases of service loadings.

Detailed analysis of the procedures indicates the need of elaboration of a fatigue life calculation

procedure of structural elements subjected to random (service) loading with wide spectrum. Set of

cycles determined on the way of schematization of such loadings is characterized by a wide range

of variability of mean values Sm and amplitude values Sa [6, 15]. That is why the undertaken task

requires the elaboration of numerous detailed methods connected among others with the elaboration

of two-parametric loading spectra (2D) [5], selection of proper fatigue characteristics (2D) [8, 13]

and selection of an appropriate cumulative fatigue damage hypothesis. [14].

The aim of this work is to present the algorithm of fatigue life calculations on the example of

41Cr4 steel in random loading conditions with wide spectrum.

The scope of the paper covers detailed review of the algorithm for the calculation on which

background there will be calculated fatigue life of 41Cr4 steel in random loading conditions with an

application of chosen two-parametric fatigue characteristics. The essential element of the work is an

experimental verification of calculated results.

Fatigue life calculations in random loading conditions require to perform numerous operations

connected with the elaboration of a model of material properties and a model of service loading.

The algorithm for fatigue life calculations in random loading conditions presented in the fig. 1

includes possibility of separation of loadings in wide and narrow ones. As a criterion for an

evaluation the coefficient I was assumed. Application of the algorithm was limited for fatigue life

calculations to macro-crack initial stage and methods of calculations in a stress approach.

The algorithm has two trails. First of them refers to service loadings while the second one to

material properties. Proceeding in the range of the first trail is connected with determination of local

extrema followed by their transformation to relative values Smax i/Smax and Smin i/Smax. Such a form

of loadings means more convenient data because of conversion of values for assumed ranges of

maximum values during calculations. The next important step for functioning of the entire

algorithm is the evaluation of width of loading spectrum. There was assumed as a criterion a value

of I coefficient on the level 0,95. Courses of loading characterized with higher value are included to

the group of loadings with narrow spectrum. In case while the value is lower the course of loading

is included to the group with wide spectrum. Classification of loadings influences the selection of a

schematization method. For loadings with narrow spectrum it is recommended to apply the peak

counting method (PCM) whereas for loadings with wide spectrum there are recommended

following methods: simple-range counting (RCM), full cycles counting (FCM), rainflow counting

(RFM) and range-pair counting (RPM) [5, 15].

Performed schamatization results as a set of sinusoidal cycles with variable Sai and Smi parameters

which is a base for determination of loading spectrum [11]. In case of loadings with narrow

spectrum calculations are based on block loading spectrum in which the mean value is constant for

individual cycles. For loadings with wide spectrum it is recommended to elaborate loading

spectrum that includes variability of loading amplitude of cycles Sai and their mean value Smi [5]. A

correlation table as Sa-Sm array [12] is the example of such a type of loading spectrum.

Dariusz Skibicki 19

START

- static (R m, Re, E, A, Z),

- cyclic (Sf (-1), Sf (0), N0, m(-1), m(0)),

Smax i /Smax i Smin i /Smax

No

I < 0,95

Determination of coefficient value

I = Ni/Ne

Yes

I < 0,95 characteristic 1D of fatigue characteristic

Yes in accordance

with one of models:

Schematization H, I, IM, II, III, IV, V

with the method: Schematization with the method:

PCM RCM, FCM, RFM, RPM

Liner cumulative fatigue Non-liner cumulative fatigue Hipothesis based of constant

damage hipothesis damage hipothesis fatigue damage lines

No Yes

I < 0,95

D = ΣDi k l

D= ∑∑ D ij

i =1 j=1

Calculation of design fatigue life Nc cal = (1/D) nc

No

Determination of design Nc cal > N proj

fatigue life Nproj

Yes

Stop

The second trail is connected with material properties. Possibility of calculations requires

knowledge on static and cyclic properties of a material. Depending on the type of service loading it

is recommended to apply suitable characteristics. For narrow spectrum loadings in calculations

there are applied one-parametric characteristics S-N (1D) while for loadings with coefficient

I < 0.95 it is recommended to apply characteristics Sa-Sm-N (2D). Such recommendations are

connected with influence of cycles with specific value of cycle asymmetry coefficient R on fatigue

life [8].

For calculations of fatigue life apart from discussed above loading spectra and two-parametric

fatigue life characteristics there should be assumed an appropriate cumulative fatigue damage

hypothesis. In the presented algorithm there was included possibility to select such a hypothesis.

20 Fatigue Failure and Fracture Mechanics

Calculating one can apply: linear cumulative fatigue damage hypothesis, non-linear cumulative

fatigue damage hypothesis or hypothesis based on idea of constant damage line [14]. In this paper

there was applied Palmgren-Miner linear cumulative fatigue damage hypothesis.

After the selection of hypothesis of accumulation of life damage calculations leading to determine

the level of fatigue damage are performed [9]. For loadings with narrow spectrum fatigue damage is

calculated from the formula

ni

D=∑ (1)

Ni

whereas for loadings with wide spectrum

k p n ij

D = ∑∑ (2)

i =1 j=1 N ij

Finally calculating fatigue life Nc cal is determined using the formula

1

N c cal = ⋅ nc (3)

D

The result of fatigue life calculations Nc cal of a structural element has to be compared with assumed

design fatigue life Nproj. In case when fatigue life Nc cal > Nproj one can state that fatigue life of the

structural element fulfills design assumptions what enables to finish calculations. In the opposite

case (Nc cal < Nproj) one has to return to the beginning of calculations assuming new properties for

structural material and/or validating service data conditions. Undertaken corrective actions enable to

recalculate and reevaluate fatigue life of the structural element.

On the base of presented algorithm for calculations fatigue life was calculated what will be

presented in the further part of the work.

3. Service loadings

Fatigue life calculations with accordance to the elaborated algorithm will be presented on the

example of service course of stress changes appearing in a dangerous section of the steering

mechanism of a passenger car. A part of the recorded loading course was presented in the fig. 2.

Following values were described in relative values Si/Smax.

0.5

Stress Si/Smax

-0.5

2s

Time, s

-1

Fig. 2 Courses of stress changes in the form of relative values for a soil roads with steering steering

maneuvers

General analysis of the loading course indicates its complex character in fields of values, time and

frequency that character should be found in factors influencing on a vehicle suspension assembly.

While steering a car it is going to be subjected to transverse swinging that occurs as loading with

the course close to sinusoidal (periodical) one on which loading caused by road surface

imperfections is overlaid [7].

Dariusz Skibicki 21

The presented course of loading changes was statistically analyzed. Values of statistical parameters

are : the root mean square ψ2 = 0.1053 (MPa/MPa)2, mean µx = 0.0138 (MPa/MPa) and variance

σx2 = 0.1051(MPa/MPa)2. There were also determined values of autocorrelation and spectral power

density that were presented in the work [6] in the form of diagrams. Their analysis enables to

classify loading service to the group of wide loading spectrum [2].

Another method to evaluate width of loading spectrum that does not require the complex statistical

analysis is determination of I coefficient value from the formula

Ni

I= (4)

Ne

In the above formula Ni jest is the number of intersections of mean values through half cycles,

increasing and decreasing ones, whereas Ne is the number of local extrema appearing in the course

(the sum of minimum and maximum values). The method was described in the work [15]. For the

assumed loading course the value of I coefficient is 0.6892 what classifies it to the group of wide

loading spectrum.

Classification of the stress course on the base of I coefficient is concurrent with classification

performed on the base of statistical functions.

Regarding to fatigue life tests there was established a service loading model in the form of the

correlation table in Sa-Sm array (2D spectrum). The set of data including following minimum values

and consequently maximum ones was the base for the undertaken actions. In tests the method of full

cycle accumulation was assumed [5] as a recommended for the wide spectrum loading analysis.

R = -1,0

Smax = 1 Smin = -1

1 C

0.95

0.9

0.85 1

0.8 1

0.75

0.7 R R=0

Amplitude Sai / S max

0.6 1

0.55

Smax = 0 1

Smin = 0

0.5 B D

0.45 2 1 1 2 1

0.4 1 1 1 2

0.35 1 1 2

0.3 1 2 3

0.25 1 1 5 3 1 1 1

0.2 1 3 2 2

0.15 1<R<+ 1 1 2 1 4 1 0<R<1 1

0.1 1 1 5 1 2 1 1 1 1

R =1,0 0.05

0 A F E

-0.95

-0.85

-0.75

-0.65

-0.55

-0.45

-0.35

-0.25

-0.15

-0.05

0.05

0.15

0.25

0.35

0.45

0.55

0.65

0.75

0.85

0.95

-0.9

-0.8

-0.7

-0.6

-0.5

-0.4

-0.3

-0.2

-0.1

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

0

Schematization of service loading enabled to determine the set of sinusoidal cycles characterized by

variable values of maximums, minimums, amplitudes and means. In calculations and tests there was

applied 2D spectrum that is the correlation table in Sa-Sm array presented in the fig. 3. Values of

parameters of cycles were presented in relatives that refer to maximum value appearing in the

loading course. The structure of such spectra was discussed in the work [5]. Correlation tables are

applied in fatigue life tests also by other scientists what is indicated by works [4, 10]. The

22 Fatigue Failure and Fracture Mechanics

insignificantly changed coordinate system in relation to the work [5] was introduced in order to

unify the description of loading spectra 2D and two-parametric fatigue characteristics. Introduced

changes did not cause any changes in properties of tables. In such a case position o fields with

constant values of defined parameters describe suitable coordinates [6].

4. Experimental tests

Test samples were made of 41Cr4 steel classified as a steel for quenching and tempering. The

mentioned processes were performed with accordance to recommendations for the type. Surface

hardness of specimens was limited to 42÷44 HRC.

Tests in static loading conditions were conducted applying specimens made in accordance with

standard PN-EN 10002-1:2004 with diameter of a measurement part was 10 mm. Specimens for

cyclic loading conditions (for loadings with coefficient R = -1 and R = 0) made in accordance with

standard PN-84/H-04334. Diameter of a measurement part was 10 mm and its length 18 mm.

Properties of 41Cr4 steel in static loading conditions were determined in experimental tests. Mean

values for chosen parameters are following: tensile modulus (Young’s modulus) E = 203900 MPa,

plasticity limit R0,2 = 1288 MPa, tensile strength Rm = 1395 MPa, elongation A5 = 24.6 % and

contraction Z = 41.2 % .

Tests of steel in variable loading conditions with the cycle asymmetry coefficient R = -1 and R = 0

enabled to determine following parameters: m(-1) = 8.53, m(0) = 9.95, C(-1) = 4.093⋅1028,

C(0) = 3.272⋅1031, Sf (-1) = 447.6 MPa, Sf (0) = 366.7 MPa. Factors of material sensitivity to cycle

asymmetry ψ [13] were calculated from the formula

Experimental tests in service loading conditions were performer with the usage of a test stand

consisting of the Instron 8501 fatigue testing machine controlled with a dedicated software that

enables to perform programmed fatigue life tests. During tests there was applied the service loading

model in the form of 2D spectrum.

In tests of 41Cr4 steel there were applied following levels of maximum stresses in the loading

programme Smax = 800 MPa, 900 MPa, 1000 MPa and 1100 MPa. Number of specimens on

individual levels was as following: level 800 MPa – 2 pcs., level 900 MPa – 3 pcs., level 1000 MPa

– 3 pcs. and level 1100 MPa – 3 pcs.,

Fatigue test results in service loading conditions were presented as fatigue life curve (fig. 4) that

was defined with the equation

1

log Smax = − log N c exp

+ 3.6095 (7)

7.92

The square of Pearson correlation coefficient for test results is r2 = 0.9143. Moreover there was

determined a value of statistic F = 96.1. It is higher than the critical value Fkr = 5.12 read from the

Snedecor’F continuous probability distribution with a confidence interval α = 0.05 for degrees of

freedom k1 = 1 i k2 = 9. From the comparison of statistic values F with critical value Fkr it results

that equation (7) correctly describes test results what confirms its application in further analysis [1].

Dariusz Skibicki 23

3000

2000

1

1000

800

600

400

2

200

100

1.E+02 1.E+03 1.E+04 1.E+05 1.E+06 1.E+07

Fig. 4 Gassner fatigue life curve of 41Cr4 steel specimens (1) on the background of Wöhler fatigue

life curve (2) [6]

Fatigue life calculations were performed on the base of chosen mathematical models of two-

parametric fatigue life characteristics included in the work [8]. Below there are listed the final forms

of equations describing assumed models of characteristics including experimental test results of

41Cr4 steel.

(Model I (model by J. Szala). Two-parametric fatigue characteristic in accordance with the model I

has a following form:

- for the range -∞ < R ≤ 0

4,0935 ⋅10 28

N= (8)

(Sa + ψ NSm )8,53

- for the range 0 < R < 1,0

8 , 53

447,6 ⋅ (1395 + Sa − Sm )

N = 10 ⋅

6

(9)

1395 ⋅ Sa (1 + ψ N )

where

- for model I according to the formula

Model I-M (model by G. Szala). Fatigue characteristic is described by formulas (8) and (9).

Modification is on the process of calculations of the material sensitivity to cycle asymmetry factor

ψN, that was described as a following dependence [13]

24 Fatigue Failure and Fracture Mechanics

Model II (model byl J. Szala). Two-parametric fatigue characteristic in accordance with the model II

presented in the form applied in calculations for the range -∞ < R ≤ 1,0

8 , 53

447,6 S

N = 10 ⋅

6

1 − m (12)

Sa 1395

Model III (model by A. Lipski). Two-parametric fatigue characteristic form in accordance with the

model III for the range -∞ < R ≤ 1,0 is defined by the following equation

8, 53

447,6 S 2

N = 10 ⋅

6 1 − m (13)

a 1395

S

accordance with the model IV for the range of cycle asymmetry coefficient -∞ < R ≤ 1,0 is

presented by the dependence below

8, 53

447,6 Sm

2

N = 10 ⋅

6

1− (14)

Sa 1395

Fatigue life calculations were performed for loading spectrum in the form of the correlation table

(fig. 3). There was also assumed Palmgren-Miner linear cumulative fatigue damage hypothesis.

Fatigue life calculations were performer for the range of maximum stresses in loading spectrum

from 500 to 1200 MPa with 100 MPa grading. Obtained test results for assumed two-parametric

characteristics were approximated with line in logarithmic scale and described with the equation

1

log S(max

x)

=− (x)

⋅ log N c obl + c t (15)

m c cal

Exponent value mc cal(x) and ct for individual modes are following: model I – mc cal(I) = 8.86 and ct =

3.5656, model I-M – mc cal(I-M) = 9.14 and ct = 3.5250, model II – mc cal(II) = 9.63 and ct = 3.4935,

model III – mc cal(III) = 8.70 and ct = 3.5949, model IV – mc cal(IV) = 8.61 and ct = 3.6046.

Experimental verification of fatigue life test results was based on comparative analysis

of experimental and calculation fatigue life curves that led to determine the ratio Nc cal/Nc exp or

Nc exp/Nc cal in the function of maximum stress Smax (fig. 5).

Comparison of service fatigue life Nc exp with calculated fatigue life Nc cal, in accordance with

assumed models, indicates that the highest conformity of results was obtained for two-parametric

fatigue characteristic marked as a model I-M. In the range of maximum stresses from 250 MPa to

940 MPa test results are in the dangerous area while above 940 MPa in the area of secure

estimations. It results from the fig. 5 that the ratio value Nc cal/Nc exp depends on the stress level Smax.

For the level Smax = 250 MPa there was obtained the highest ratio value Nc cal/Nc exp approx. 4.2 that,

as stress Smax decreases, reaches value of 1 at the level 940 MPa. Above Smax = 940 MPa the ratio

value Nc exp/Nc cal increases reaching value approx. 1.5 for the maximum stresses 1300 MPa.

For the model II there was obtained the similar course of changes of the ratio value between

experimental and calculated values to the course characterized by I-M model. Differences among

results are basically on values of Nc cal/Nc exp and Nc exp/Nc cal on defined loading levels Sa.

Obtained test results for two-parametric fatigue life characteristics marked as model I, III and IV are

located in the dangerous calculation area for the entire range of maximum stress Smax. Similar

character of changes of values Nc cal/Nc exp characterizes models III and IV. It should be also noticed

that the smallest conformity of calculations with the experiment obtained for models IV and III.

Dariusz Skibicki 25

16

7

1 Model I 2 Model I-M 3 Model II

15

6

4 Model III 5 Model IV

Nc cal/Nc exp

5

14

1 4 5

4

13

3

12

11

2

2

1

10

Nc exp/Nc cal

29

3

38

200 400 600 800 1000 1200 1400

Stress Smax, MPa

Fig. 5 Experimental to calculated fatigue life ratio for 41Cr4 steel [6]

Numerous factors influence conformity of calculated and experimental fatigue life results, to name

the most important ones:

– conformity of mathematical models of two-parametric fatigue characteristics with characteristics

experimentally determined,

– conformity of cumulative fatigue damage hypothesis with experimental data.

Summary

a. Presented algorithm of fatigue life calculations applies to the stress approach that can be applied

in calculations for high cycle fatigue loading. Assuming I coefficient as the decision criterion

enables to chose the appropriate way of proceedings connected with elaboration of loading

spectrum and determination of necessary, because of calculations, fatigue life characteristics.

Evaluation of loading spectrum width on the base of I coefficient enables to limit the application

of complex statistical apparatus that requires detailed knowledge on interpretation of test results

as well as specialist software.

b. From the comparative analysis of fatigue life test results with calculated ones according to two-

parametric fatigue life characteristics it results that differences between calculation results

depend on the applied model and essentially on the maximum stress level in the loading

programme what has a direct connection with the evaluated fatigue life Nc cal.

c. Cumulative fatigue damage hypothesis was the constant factor in the performed fatigue life

calculations with application of assumed two-parametric fatigue life characteristics and as a

consequence it did not influence on evaluation results of individual models.

d. Regarding to the simplicity of IM and II models there are recommended especially in initial

calculations realized in design process. The above recommendation is essentials also because of

the smallest range of necessary experimental data that appear in descriptions of the mentioned

models.

e. One of the essential issues in fatigue life tests and calculations of structural materials and

elements is the appropriate elaboration of loading spectrum and on its base the proper loading

programme. For random loadings with wide spectrum the essential role is played by the

assumption of the cycle accumulation method. In the work there was assumed the method of full

26 Fatigue Failure and Fracture Mechanics

cycles. Elaborated with its application loading spectrum was presented in the form of the

correlation table including the set of sinusoidal cycles with variable parameters Sa i and Sm i.

Such a system of correlation table corresponds with descriptions applied in two-parametric

fatigue life characteristics with the form N(Sa, Sm) applied in fatigue life calculations.

Note: This work has been elaborated in the frame of the project No. 0715/B/T02/2008/35 financed

by Polish Ministry of Sciences and Higher Education.

References

[1] J.S. Bendat, A.G. Pierdol, Methods of analysis and measurement of random signals,

(in Polish), PWN, Warszawa, 1976.

[2] J. Čačko, M. Bìlý, J. Bukoveczky, Random processes: measurement, analysis and simulation,

ELSEVIER, Amsterdam – Oxford – New York – Tokyo, 1988.

[3] FITNET Fitness-for-Service Procedure – Final Draft MK7, 2006.

[4] M. Huck, W. Schultz, R. Fischer, G. Kobler, A standard random load sequence of Gaussian

type recommeded for general application in fatigue testing, LBF-Report no. 2909, IABG-

Report no. TF-570, p.21, 1976.

[5] S. Kocańda, J. Szala, Fundamentals of fatigue calculations, (in Polish), PWN, Warszawa,

1997.

[6] B. Ligaj, Experimental and calculational analysis of Steel fatigue life in random conditions of

wide range spectra, (in Polish), Monographs, 2nd part of monograph: Two-parametric fatigue

characteristics of steel and their experimental verification, Publishing House of Operation

Technology Institute - State Research Institute, Radom, 2011.

[7] B. Ligaj, G. Szala, Loading analysis in tests and calculations of fatigue life of constructional

elements – on the example of operating loadings of a car, (in Polish), Logistyka nr 6, 2009.

[8] B. Ligaj, G. Szala, Experimental Verification of two-parametic models of fatigue

characteristics by using the tests of S355J0 steel as an example, Polish Maritime Research

1/2010 (2010), 39-50.

[9] B. Ligaj, G. Szala, A fatigue life calculation method or constructional elements with a use of

two-parametric fatigue characteristics, Acta Mechanica et Automatica, vol.3 no.2 (2009), 47-

51.

[10] T. Łagoda, K. Walat, Methods of service loading performance in control systems of fatigue

machines, Acta Mechanica et Automatica, vol. 5, no. 1 (2011), 47-52,

[11] ASTM standard, Standard Practices for cykle counting in fatigue analysis, ASTM

Designation: E 1049-85 (Reapproved 1990).

[12] Description of a Fighter Aircraft Loading Standard for Fatigue, ICAF, 1976.

[13] G. Szala, Theoretical and experimental analysis of two-parametric fatigue life

characteristicsm of constr, (in Polish), Monographs, 1nd part of monograph: Two-parametric

fatigue characteristics of steel and their experimental verification, Publishing House of

Operation Technology Institute - State Research Institute, Radom, 2011.

[14] J. Szala, Hypotheses of fatigue damage accumulation, (in Polish), Monographs, University of

Technology and Agriculture, Bydgoszcz 1998.

[15] J. Szala, Loads and fatigue life of machine elements, (in Polish), University of Technology

and Agriculture, Bydgoszcz 1989.

© (2012) Trans Tech Publications, Switzerland

doi:10.4028/www.scientific.net/MSF.726.27

elements

1

Opole University of Technology, Faculty of Mechanical Engineering, ul. St. Mikołajczyka, 45-271

Opole, Poland

a b c

g.robak@po.opole.pl, m.szymaniec@wp.pl, t.lagoda@po.opole.pl

Abstract. In this paper, the fictitious radius - according to Neuber’s method for determination of

stresses at the notch root was used. Next, the fatigue lives of elements of the ring notches were

calculated, and then compared with results of experimental tests of S235JR steel samples. However,

the obtained fatigue lives did not bring satisfactory results. It has been demonstrated that the

fictitious radius strongly depends on the expected fatigue life.

Introdaction

Estimation of fatigue life of machine elements has been the subject of intensive research for

many years. Numerous failures that have been causing costly repairs of elements or even whole

structures, have forced designers to perform this type of research in search for improvement.

Moreover, in extreme cases, such failures led to life-threatening situations.

At present, the problem of lifetime prediction of machine elements and structures has been given

a proper recognition as an important and crucial issue addressed in almost every branch of a modern

industry. In particular, the problem has been occuring in industrial activities, where the life,

reliability and safety requirements are essential. One of the main reasons for conducting research on

fatigue and creating complex calculation algorithms is optimization of machine parts [1]. The

necessity of carrying out the optimization of machine elements has made the engineers to design

structures with complicated geometry while maintaining lifetime and reliability properties.

The aim of this paper is to compare the results of experimental tests performed on cylindrical

elements with ring notches to the results of lifetime prediction determined by using the fictitious

radius [2-4, 8].

Determination of the value and the fictitious radius for ring-notched elements

In order to estimate the fatigue life, a concept of the fictional radius by Neuber was used [2]. The

fictitious radius at the notch root allowed to calculate the maximum stresses. Determination of this

radius results from the mean stresses at the notch root. Fictitious radius at the notch root was

described in the following relation:

ρ f = ρ + sρ * , (1)

where:

s – multiaxial coefficient

ρ* - equivalent microstructural length,

ρ - real radius at the notch root.

The multiaxial coefficient s for round elements under tension was calculated by referring to the

Huber-Mises-Hencky’s criterion, according to the equation:

28 Fatigue Failure and Fracture Mechanics

5 − 2ν + 2ν 2

s= , (2)

2 − 2ν + 2ν 2

The equivalent microstructural length as proposed by Neuber ρ * for the steel is ca. 0.1 mm.

Verification of the fictitious radius assumptions was conducted by comparison of the calculated

fatigue life with test results for the notched specimens (Fig. 1) made of stainless S235JR steel [5]. In

Table 1, the mechanical properties of S235JR steel are presented. The theoretical stress

concentration factor was calculated, according to Nody [6] K t = 3.288, and with the calculations

obtained from FE analysis K t = 3.330. However, it must be remembered that the notch factor is

determined for the fatigue limit. In other cases, its value strongly depends on fatigue life [3, 4].

Re, Rm,

ν E, GPa

MPa MPa

299 423 0.29 200

Experimental tests were performed on the fatigue stand type UFP ± 400 – specimens were

subjected to cyclic loading under tension-compression [7]. In view of the fact that the real radius ρ =

0.8 mm, and using equations (1) and (2), the value of fictitious radius ρf = 1.088 mm. Based on the

obtained radius, a fictitious model of a specimen was created, then used for determination of

stresses at the notch root by a finite element method. The specimen model was limited to only a

quarter of the entire specimen, which enabled a greater grid density within the notch root. Software

“COMSOL” was applied for calculations. The calculations were performed for five levels of

nominal stresses in the notch cross-sections, at which the proper experimental tests were carried out:

125, 160, 200, 230, 300 and 330 MPa. The calculations were made for an elastic range. Figure 2

shows distribution of the applied finite elements at the notch root.

Dariusz Skibicki 29

Fig. 2. Distribution of finite elements at the notch root for the geometry of a specimen with the

fictitious radius.

Based on Neuber’s assumptions, stress values obtained for a modified geometry of the specimen

should be compared with fatigue characteristics of smooth specimens. The obtained values of the

fatigue life should correspond to the results of experimental tests. Figure 3 shows two fatigue

characteristics for notched specimens and smooth specimens made of S235JR steel. Relevant

fatigue characteristic for smooth (3) and notched (4) specimens can be written in S-N as

Fig. 3. Fatigue characteristics for smooth and notched specimens made of S235JR steel.

30 Fatigue Failure and Fracture Mechanics

The obtained fatigue characteristics were then compared with the experimental ones - presented

in figure 4. As can be seen, the values of calculated fatigue lives are much lower than the

experimental ones. Observing the distribution of characteristics in Fig. 3, note that the higher the

amplitude of the load, the difference between the calculated and experimental lives increases. In

case of fatigue at about 106, the difference between the obtained results decreases. It should be seen,

however, that in this case the model proposed by Nueber for the material analysed is not

appropriate. This suggests that Neuber’s assumptions are not correct for the steel for stress values at

the fatigue limit, and even more to stresses above this level. In figure 3 comparison of experimental

lives for nominal stresses equal to 300 and 330 MPa was not included, because the obtained values

of calculated fatigue lives took values below one cycle. From figure 3, it is confirmed that

distribution of fatigue characteristics for smooth and notched specimens, which coincide to the point

below 3000 cycles, where S235JR material ceases to be sensitive to the notch effect. In such a case,

the fictitious radius should aim at infinity, and the fatigue coefficient of the notch effect K f = 1.

Fig. 4. Comparison of the experimental fatigue lives with the calculated ones for specimens made

of S235JR steel.

According to the remarks discussed in the previous chapter, using fatigue characteristics for

smooth and notched specimens, the geometry of the specimen was changed by modeling different

fictitious radiuses, which resulted in obtaining various equivalent microstructural length, according

to the equation (1) we obtain new formula for equivalent microstructural length :

ρf −ρ

ρ* = . (5)

s

Dariusz Skibicki 31

The geometry of a specimen was modified in such a way so that the stresses for notched

specimens corresponded to stresses that occured for the same values as for smooth obtained from

the fatigue characteristics. In the result, an ideal compliance of the estimated fatigue characteristics

with experimental ones were therefore obtained. Based on the fatigue characteristics for

calculations, two additional levels of nominal stresses of 80 and 100 MPa were determined. In

figure 5 were estimated the obtained microstructures radiuses, according to the equation (5) as

loading cycles function. From the analysis of this figure, it can be seen that the obtained values of

the radius of microstructures are arranged according. to a linear function. This relation was

described in the following equation:

It should be noted that values of mircrostructure radius take quantities of over 1 millimeter. Only

in the case of a nominal stress equal to 80 MPa, the obtained value was 0.85 mm. This result means

that a ring notch causes much greater changes in microstructure of a material as compared to

Neuber’s assumptions.

Summary

From the obtained calculations, it can be observed that:

1. Application of the fictitious radius, according to Neuber’s solution, for steel in elastic range

under loadings above the fatigue limit does not allow to obtain results comparable with the

test results.

2. The equivalent radius of microstructure strongly depends on the fatigue lifetime.

3. The microstructure radius are arranged in a linear function depending on the number of

cycles.

32 Fatigue Failure and Fracture Mechanics

The project has been finance from the National Center for Science No 2011/01/B/ST8/06850

REFERENCES

[1] Gasiak G., Robak G.: Simulation of fatigue life of constructional steels within the mixed

modes I and III loading, Vol. 34, 2011, pp. 389-402

[2] Neuber H.: Über die Berücksichtigung der Spannungskonzentration bei

Festigkeitberechnungen. Konstruktion 20, 1968, SS. 245-251

[3] Łagoda T.: Lifetime estimation of welded joint, Springer 2008.

[4] Biłołus P., Lagoda T.: Structural notch effect in steel welded joints, Materials and Design,

Vol. 30, 2009, pp. 4562-4564

[5] Słowik J., Łagoda T.: The fatigue life estimation of elements with circumferential notch under

uniaxial sate of loading, Int. J Fatigue, Vol.33, 2011, pp.1304-1312

[6] Noda N. A., Takasa Y.: Stress concentration formula useful any shape of notch in a round test

specimen under tension and under bending, Fatigue, Fract. Engng Mater. Structur., Vol. 22,

1999, pp.1071-1082

[7] Blacha Ł., Karolczuk A., Łagoda T.: Assessment of multiaxial fatigue behaviour of welded

joint under consideration of plastic strains in fatigue life calculations, Materials Testing, Vol.

53, 2011, No 6, pp. 339-343

[8] Sonsino C. M., Łagoda T., Demofonti G.: Damage accumulation under variable amplitude

loading of welded medium and high-strength steels, Int. J. Fatigue, Vol. 26, No 5, 2004, pp.

487-495

© (2012) Trans Tech Publications, Switzerland

doi:10.4028/www.scientific.net/MSF.726.33

FATIGUE DIAGRAMS UNDER CONSTANT AMPLITUDE LOAD WITH

MEAN STRESS

Adam Niesłony1,a and Michał Böhm1,b

1

Department of Mechanics and Machine Design, Opole University of Technology, Mikołajczyka 5,

45-271 Opole, Poland

a

a.nieslony@po.opole.pl, bm.bohm@doktorant.po.opole.pl

Abstract. The paper deals with a comparison of fatigue life calculations, obtained on the basis of

classical Basquin diagrams and approximated diagrams acquired with the use of the section method,

which allows us to fit the diagrams shape to real material properties. While comparing the

calculation results, literature data concerning fatigue tests of welded cruciform specimens from high

performance low-alloy steel HSLA-80 presented by Kihl and Sarkani as well as of smooth

specimens out of the aluminum alloy 75S-T6 by Grover et al. has been used. It has been noticed that

the calculations performed with the use of fatigue diagrams approximated using the section method

reflect the true behavior of the material. The models by Niesłony-Böhm and Smith-Watson-Topper

compensating the influence of the mean stress gave similar results.

Introduction

Mechanical fatigue of the material is a process resulting from a time varying load and

manifested by the loss of cohesion of the material and its ability to carry loads. The mechanism of

this phenomenon is complex and depends mainly on the type of material, the nature of changes in

load, the surface condition and shape of the element. In engineering practice, this phenomenon is

described by the fatigue diagrams, which are the result of statistical processing of test results of

samples of material at strictly defined variable loading conditions. We obtain in this way

information sufficient to properly describe the phenomenon without going into the macroscopic

fatigue mechanisms that are difficult to describe in a deterministic manner. For this purpose most

often are used stress fatigue diagrams, which relate the number of cycles till destruction with the

stress amplitude. These diagrams are modeled with exponential functions obtaining the

corresponding sections of simple graphs with logarithmic axes. One of the most popular types of

these diagrams is the Wöhler and Basquin diagram [1,2]. Unfortunately, they describe the real

diagram only for a narrow interval of cycle number, in which we observe approximately a straight

line. Therefore the focus of this paper is set on the correct description of the experimental results

obtained from literature [3,4] with classical Basquin diagrams for the stress ratio R = −1 and R = 0

as well as with the section method, and later on these diagrams have been used to calculate the

fatigue life taking into account the mean stress value of the load . This is the simplest practical case

in which the particular role is played by the quality of the fit to the actual characteristics of the

fatigue properties of the material. The main objective of this work is to investigate whether and to

what extent the stress diagram modeling of fatigue by the classical method and sectional influences

the fatigue life calculation.

Stress fatigue diagrams represent material resistance to variable loads and are used in the

calculation process of fatigue life. Due to the exponential nature of the changes of this type the

fatigue diagrams axes are scaled logarithmically. According to the literature review, the true fatigue

curves form the shape of a descending wave, which is schematically presented in Fig.1.

34 Fatigue Failure and Fracture Mechanics

Figure 1. True shape of a fatigue curve and the approximation with the Wöhler and Basquin

diagram [5].

Due to the fact that most fatigue calculations are made for a range of limited or unlimited

fatigue life, a number of proposals for the approximation of these ranges with corresponding

functions have been presented [6], the most popular is the Wöhler diagram

σ a = CN b . (2)

Due to the high ease of use of numerical methods, it is becoming increasingly popular to

represent a function with straight sections. This method, called section method is very common in

FEM calculation programs, and allows the approximation of experimental data showing no fatigue

limit or non-linear nature of the stress fatigue diagram [7]. In this case, to define the further sections

is enough to provide pairs of stress amplitude σa and the corresponding number of cycles to

destruction N, which define a series of points connected by straight lines on the graph. The number

of sections is chosen to suit your needs. The obtained diagrams approximated using this method is

presented in Fig. 2 and Fig.3 respectively for the fatigue tests of welded cruciform specimens made

out of high-strength low alloy steel HSLA-80 performed by Kihl and Sarkani and for smooth

specimens out of the aluminum alloy 75S-T6 performed by Grover et al..

3

10

R = -1; experiment

R = -1; Basquin graph

R = -1; section graph

R = 0; experiment

R = 0; Basquin graph

R = 0; section graph

σmax, MPa

2

10

3 4 5 6 7 8

10 10 10 10 10 10

N, cycle

Figure 2. Stress fatigue diagram on the basis of research by Kihl and Sarkani [3] approximated with

the Basquin and section method diagrams.

Dariusz Skibicki 35

3

10

R= -1; experiment

R= -1; Basquin graph

R= -1; section graph

R= 0; experiment

R= 0; Basquin graph

R= 0; section graph

σ max, MPa

2

10

3 4 5 6 7 8

10 10 10 10 10 10

N, cycle

Figure 3. Stress fatigue diagram on the basis of research by Grover et al. [4] approximated with the

Basquin and section method diagrams

Table 1. Points defining the stress fatigue diagrams in the section method.

Steel HSLA-80, Kihl and Sarkani [3] Aluminum alloy 75S-T6, Grover et al. [4]

R = −1 R=0 R = −1 R=0

i σai N i σai Ni σai N i σai Ni

3

1 380 5⋅10 275 5⋅103 385 5⋅10 3

295 5⋅103

4

2 304 1,8⋅10 205 3⋅104 325 1,8⋅10 4

235 2⋅104

4

3 200 8⋅10 100 3⋅105 245 8⋅10 4

165 6⋅104

6

4 92 1⋅10 62 3⋅106 212 3⋅10 5

140 2⋅105

6

5 77 3⋅10 50 1⋅108 180 3⋅10 6

122 3⋅106

6 70 1⋅107 150 1⋅108 108 1⋅108

7 60 1⋅108

Determination of fatigue life on the basis of σa-N diagrams taking into account the influence of

mean stress

Reliable prediction of fatigue life is inherently bond with taking into account the mean stress

value of the load on the endurance. A multitude of combinations of amplitude loads and mean value

found in the actual runs is the main reason for the development of models describing these effects in

an analytical way. Various equations created over the years present fatigue life as a function of

amplitude, mean value and the selected material constants. As an example we can mention the

models proposed by Gerber [8], Goodman [9] or Soderberg [10]. In the context of this work a

special attention deserves models that also allow the calculation according to the characteristics

described by the section method. Two models have been chosen for this reason. The first proposed

by Niesłony and Böhm (NB) [11]:

σm

σ aT = σ a + [σ afR =−1 ( N ) − σ afR =0 ( N )] , (3)

σ af , R=0 ( N )

where: σaf,R=-1(N) and σaf,R=0(N) are values of the amplitudes read from the stress fatigue diagrams

for a certain number of cycles N for the stress ratio R = −1 and R = 0. The second one is the well

known and popular Smith-Watson-Topper model (SWT) [12] in the form

36 Fatigue Failure and Fracture Mechanics

σ aT = (σ a + σ m )σ a . (4)

Transformed stress amplitudes obtained on the basis of the equation (3) and (4) have been used

to calculate the number of cycles Ncal on the basis of the Basquin stress fatigue characteristics and

characteristics described by the section method, which are presented in Fig.2 and Fig.3 and

described earlier. Of course for reading the number of cycles, the characteristic with the stress ratio

R = −1 has been used. In the case of the NB model, while performing calculations the number of

cycles N had to be chosen for the proper values of the σaf,R=-1(N) and σaf,R=0(N), so that it would be

equal to the calculated number of cycles Ncal. Since before computing the transformed amplitude

according to equation (3) the number of cycles was not known, N was determined by solving the

optimization problem in which the objective function was defined as follows

N = N cal . (5)

The fzero function of the MATLAB environment has been used for that reason. The calculated

numbers of cycles Ncal were compared with the ones gained from the experimental research Nexp and

presented in Fig.4 for the research conducted by Kihl and Sarkani [3] and for Grover et al. [4] in

Fig.5.

a) b)

8 8

10 10

R=0 R=0

R=1/3 R=1/3

7 7

10 R=2/3 10 R=2/3

R=-1/3 R=-1/3

R=-2 R=-2

6 6

10 R=-3

Ncal , cycle

10 R=-3

Ncal , cycle

5 5

10 10

4 4

10 10

3 3 3 3

Basquin graph Basquin graph

3 NB model 3 SWT model

10 10

4 6 8 4 6 8

10 10 10 10 10 10

Nexp , cycle Nexp , cycle

c) d)

8 8

10 10

R=0 R=0

R=1/3 R=1/3

7 7

10 R=2/3 10 R=2/3

R=-1/3 R=-1/3

R=-2 R=-2

6 6

10 R=-3 R=-3

Ncal , cycle

10

Ncal , cycle

5

10 5

10

4

10 4

10

3 3

Section graph 3 3

NB model Section graph

3

10 3 SWT model

4 6 8 10

10 10 10 4 6 8

10 10 10

Nexp , cycle Nexp , cycle

Figure 4. Comparison of the calculated number of cycles Ncal on the basis of the NB model (a) and

(c) and SWT model (b) and (d) with the number of cycles obtained from the experimental

research Nexp for different values of the stress ratio R. Results of the cyclic research by Kihl

and Sarkani R = 0 and R = −1 approximated with the Basquin graph (a) and (b) as well as

with the section method (c) and (d).

Dariusz Skibicki 37

a) b)

8 8

10 10

R=0.7 R=0.7

R=0.6 R=0.6

7 7

10 R=0.5 10 R=0.5

R=0.4 R=0.4

6

R=0.25 6

R=0.25

10 R=0.1 10 R=0.1

R=0.02 R=0.02

Ncal , cycle

Ncal , cycle

5 R=-0.6 5 R=-0.6

10 10

R=-0.8 R=-0.8

R=-1.0 R=-1.0

4 4

10 10

3 3 3 10

3 3 3

10

2 NB model 2 SWT model

10 10

2 4 6 8 2 4 6 8

10 10 10 10 10 10 10 10

Nexp , cycle Nexp , cycle

c) d)

8

10 10

8

R=0.7 R=0.7

R=0.6 R=0.6

7

10 R=0.5 10

7

R=0.5

R=0.4 R=0.4

6 R=0.25 R=0.25

10 6

R=0.1 10 R=0.1

R=0.02 R=0.02

Ncal , cycle

Ncal , cycle

5 R=-0.6 5 R=-0.6

10 10

R=-0.8 R=-0.8

R=-1.0 R=-1.0

4 4

10 10

3 3 3 3 3 3

10 10

2 NB model 2 SWT model

10 10

2 4 6 8 2 4 6 8

10 10 10 10 10 10 10 10

Nexp , cycle N

exp

, cycle

Figure 5. Comparison of the calculated number of cycles Ncal on the basis of the NB model (a)

and (c) and SWT model (b) and (d) with the number of cycles obtained from the

experimental research Nexp for different values of the stress ratio R. Results of the cyclic

research by Grover et al. R = 0 and R = −1 approximated with the Basquin graph (a) and

(b) as well as with the section method (c) and (d).

The application of experimental fatigue diagrams approximated with the section method in the

process of fatigue determination affected the accuracy of the calculated results. This is particularly

visible for the extreme values of the range of the number of cycles, where the classical Basquin

diagram doesn’t describe sufficiently accurately the behavior of the material The section method

allows with a sufficiently large number of experimental studies, a better fit to the actual fatigue

characteristics. NB and SWT models used to determine the transformed amplitudes due to mean

stress allow to compensate the effect of the mean stress value on fatigue life in a satisfying way. We

can notice that while using the NB model we obtain better results comparing to SWT taking into

account the range of research results in the number of cycles. This can be explained by the fact that

there is a dependence of the sensitivity of the material on the mean value due to the number of

cycles. The NB model accounts for this change by determining the transformed amplitudes on the

basis of amplitudes read from the fatigue diagrams for a fixed number of cycles. There is a need to

conduct further review of selected models on the basis of fatigue diagrams predicted approximated

by sections to demonstrate their usefulness in predicting the fatigue life in a wide range of number

of cycles.

38 Fatigue Failure and Fracture Mechanics

References

[1] Kocańda S. Fatigue Failure of Metals. 1st ed. Springer, 1978.

[2] Kocańda S, Szala J. Basics of fatigue calculations. Warsaw: PWN (in Polish), 1997.

[3] Kihl DP, Sarkani S. Mean stress effects in fatigue of welded steel joints. Probabilistic

Engineering Mechanics (1999), 14:97–104.

[4] Grover HJ, Bishop SM, Jackson LR. Fatigue strengths of aircraft materials: axial load fatigue

tests on unnotched sheet specimens of 24S-T3 and 75S-T6 aluminum alloys and of SAE 4130 steel.

(1951).

[5] Basquin OH. The Exponential Law of Endurance Tests. Am. Soc. Test. Mater. Proc. (1910),

10:625–30.

[6] Schütz W. A history of fatigue. Engineering Fracture Mechanics (1996), 54:263–300.

[7] Berger C, Pyttel B, Schwerdt D. Beyond HCF – Is there a fatigue limit? Materialwissenschaft

Und Werkstofftechnik (2008), 39:769–76.

[8] Gerber WZ. Bestimmung der zulässigen spannungen in eisen-constructionen (Calculation of

the allowable stresses in iron structures). Z Bayer Archit. Ing-Ver (1874), 6:101–10.

[9] Goodman J. Mechanics applied to engineering. Longmans, Green & Co.; 1899.

[10] Soderberg CR, Sweden V. Factor of safety and working stress. ASME Trans, AER-IS

1930;52:13–28.

[11] Niesłony A, Böhm M. Fatigue life of cast iron ggg40 under variable amplitude tension with

torsion with mean stress, Engineering Modeling (in Polish) (2011), 41:299–306.

[12] Smith KN, Watson P, Topper TH. A stress-strain function for the fatigue of metals. Journal

Materials (1970), 5:767–76.

© (2012) Trans Tech Publications, Switzerland

doi:10.4028/www.scientific.net/MSF.726.39

Applying a stepwise load for calculation of the S-N curve for trabecular

bone based on the linear hypothesis for fatigue damage accumulation

Adam Mazurkiewicz1,c, Krzysztof Nowicki1,d

1

Faculty of Mechanical Engineering, University of Technology and Life Sciences,

Kaliskiego 7 Street, 85-789 Bydgoszcz, Poland

a

tomasz.topolinski@utp.edu.pl, bartur.cichanski@utp.edu.pl,

c d

adam.mazurkiewicz@utp.edu.pl, krzysztof.nowicki@utp.edu.pl

Keywords: trabecular bone, stepwise loading, linear hypothesis of fatigue damage accumulation.

Abstract: In this work were presented calculated fatigue curves based on fatigue tests of trabecular

bone under stepwise load with the application of a linear hypothesis accumulation of fatigue

damage. The investigation was performed on 61 cylindrical bone samples obtained from the neck of

different femur heads. The bone sample fatigue tests were carried out under compression with

stepwise increases of the applied load. The fatigue calculation assumed the Palmgren-Miner (P-M)

linear hypothesis accumulation of fatigue damage and the associated modified formulae. The

obtained mean fatigue curves were based on the modified stress σ/E0 (E0 – initial stiffness) for the

assumed rule-determined slope or y-intercept. The highest agreement with the literature was

obtained for Σn/N=10.

Introduction

Due to the complexity of the fatigue processes, physical descriptions are often impossible to

define, and thus, in most methods that allow for a calculated evaluation of fatigue life, the fatigue

damage accumulation hypothesis is used. Publications [1,2] refer to several hypotheses of fatigue

damage accumulation for different materials. The simplest and the oldest hypothesis is the Palmgren

hypothesis supported by the results reported by Miner and referred to as the P-M hypothesis. The

modifications to this hypothesis involve introducing more complex forms of the fatigue life

relationships, including many of the load parameters and environmental factors. There are also

nonlinear hypotheses based on the assumption of equi-damage fatigue lines [3].

The aim of the authors is to present a method for the estimation of the S-N curve for the

description of fatigue properties of bone using as the base Palmgren–Miner hypothesis (and its

modifications), literature data and a single sample. The authors focused on the presentation of this

method and calibration its parameters.

Methods

The paper uses the research results of 61 cylindrical bone samples that were 10mm in diameter

and 8.5mm in length and were obtained from the neck of femur heads. The samples were obtained

from 21 men and 40 women undergoing hip joint alloplasty. The samples were stored in a 10%

formalin solution at room temperature.

All of the samples were scanned with a desktop microCT system (µCT-80, SCANCO Medical

AG, Bruettiselllen, Switzerland) with a distance of 36µm between. When scanning the values of

many bone structure indices were obtained: trabecular number Tb.N, trabecular thickness Tb.Th,

trabecular separation Tb.Sp, bone volume fraction BV/TV, surface fraction BS/BV and the number

of joints between individual trabeculae per unit volume of specimen Conn.D.

The bone sample fatigue tests were carried out under compression with stepwise increases in the

load using the testing machine, INSTRON 8874 (Instron, High Wycombe, England). The minimum

loading for all of the loading levels was 5N. The maximum loading started at 10N with a gain every

40 Fatigue Failure and Fracture Mechanics

10N at each successive step. At each loading level, 500 cycles were completed under constant-

amplitude loadings at a frequency of 1Hz. During the fatigue test were recorded values of

displacements and forces with a frequency of 100Hz. Fatigue test termination criterion was the step

increase of the recorded displacement.

The calculations assumed a linear hypothesis of fatigue damage accumulation.

k

n

∑ N

i =1

=D (Eq. 1)

In the first case, it was assumed that the sum of the quotient of the executed cycles, n, at a

specific load level to the fatigue life, N, at the same load level at the moment of damage occurrence

equals D=1. In the second case, the same linear hypothesis was considered, however parameter D

can take values different than 1.

In both cases, the fatigue life, N, was derived from the S-N curve of the form shown below:

σ E0 = aN b (Eq. 2)

where: σ is the maximum stress, a is the y-intercept of the obtained straight line, and b is the slope

after linearization in bilogarythmic coordinates.

Founded in the work method of calculating the S–N curve based on fatigue test results with

stepwise load requires the determination of N in equation 1, based on the equation 2. Then:

k

n

∑ 1

=D (Eq. 3)

i =1 (σ / aE0 )b

Based on the equation 3, for the assumed values of D and experimental values of σ were

calculated coefficients a or b. The calculations were based on constant values of the equations

coefficients for two cases. In first case, all of the fatigue curves assumed the same value of the y-

intercept, a. In second case, assumed that the value of the slope b from did not change in all

calculations.

The following values from literature [4,5,6,7,8] were assumed for the main calculations in the

case of invariability of the y-intercept: 0.0098 - the mean value a for human bones, in case of

invariability of the slope: -0.1094 – the mean value b for human bones. Calculations were also made

for the following values of the y-intercept 0.0121 – the maximum value for human bone (same as

mean value a for bovine bone) and 0.0241 – the maximum value a for bovine bone.

Results

As an example, the results of calculations, made for D=1, and the changes in the values of the

coefficients for the fatigue equation were a=0.00980, 0.0121 and 0.0241 and b=-0.1094 (mean) are

shown on the Fig. 1.

Fig. 1 presents a set of fatigue curves in a different form; as an example, for the sum of damage

D, which equals 1, the curves for all of the samples show the range of variation in the location; the

mean curve (for the mean value of a, Fig. 1a-c, or the mean value of the slope b, Fig. 1d); and the

area defined by the intervals of confidence for the mean value. In all cases, α=0.05 was assumed.

This distribution demonstrates how the locations of the curves change with increasing values of the

y-intercept (Fig. 1a-c) when compared with the curves calculated for the mean value of the slope b.

Dariusz Skibicki 41

a) b) c) d)

0,1 0,1 0,1 0,1

σ /Ε 0

σ/Ε 0

σ/Ε 0

σ/Ε 0

0,01 0,01 0,01 0,01

10 100000 10 100000 10 100000 10 100000

Fig. 1. The results of the calculations of the S-N curve: extreme (represented by thin lines), mean

(thick line) and their intervals of confidence for human trabecular bone samples (dashed line),

assuming a-c) the constant value of the coefficient of the y-intercept and d) the constant value of the

slope b

The fatigue curve calculation results for all the conditions assumed are given in Table 1. An

increase in the absolute value of the slope bm with an increase in the value of the y-intercept and an

increase in the value of the damage parameter D was clearly visible. The damage parameter D is

well described by the logarithmic function of the mean value of the slope bm and changing in the

value of a. The minimal value of the coefficient of determination R2 compounds to 0.97, whereas

the relationship between the value of the mean y-intercept and the value of the sum of the damage

was a power function (R2=0.99).

Table 1. The mean values of the coefficients, bm and am, for the fatigue curve equation determined

for three different values of the y-intercept a and the slope b for four values of D, 0.5, 1, 2 and 10.

D=0.5 D=1 D=2 D=10

bm

0.0098 -0.0509 -0.0558 -0.0617 -0.0810

a 0.0121 -0.0644 -0.0704 -0.0775 -0.1007

0.0241 -0.1002 -0.1088 -0.1189 -0.1511

am

b -0.1094 0.0420 0.0389 0.0361 0.03026

The results of the experiments for fatigue life, Ns, when exposed to a stepwise load in reference

to their structure index BV/TV for all of the samples are given in Fig. 2. This demonstrates the

plotted linear approximating function together with the value of the coefficient of determination.

The results constitute a basis for the calculations of the fatigue curves.

Discussion

This paper is based on the results of research conducted on a different samples of the bone,

which were clearly confirmed by the results of the investigations made for the structure indices. In

paper [5], the scatter BV/TV for 35 samples from 9 donors (relative standard deviation -

RSD=38.5%) was comparable with the scatter obtained in the current experiments (61 donors and

RSD=37.1%). The coefficient of determination for the fatigue curve in the same paper [5] was not

particularly high R2=0.54). In paper [4], the scatter for 29 samples from 4 donors was higher than in

our experiment (as much as RSD=42.5%); however, introducing the stress modification to the

fatigue equation with the use of the volume fraction and the fabric eigenvalue resulted in a high

correlation between the stress and the fatigue life at a constant load amplitude (R2=0.95).

42 Fatigue Failure and Fracture Mechanics

0,50

0,40

BV/TV

0,30

0,20

y = 6E-06x + 0,0895

0,10 R2 = 0,689

0,00

0 10000 20000 30000 40000 50000 60000

Ns cycle

Fig. 2. The relationship between the fatigue life Ns and the volume fraction BV/TV for the samples

of bone that were loaded with a stepwise increasing amplitude: the determination coefficient was

0.689 and the p-value was 1.32×10-14

The relationship between the fatigue life results and the volume fraction, and thus one of the

structure indices, does exist. It was not a strong correlation (R2=0.689); however, it should be noted

that this relationship covers not only the bone properties (including the structure with its individual

characteristics, damage and the effects of remodeling) but also the specificity of the fatigue damage

process for stepwise loading with the dynamics also associated with the bone properties.

Reference

[1] Manson S.S., Halford G.R.: Practical Implementation of the Double Linear Damage Rule and

Damage Curve Approach for Treating Cumulative Fatigue Damage. International Journal of

Fatigue 1981, 17: 169-192

[2] Hashin Z., and Rotem A.: A Cumulative Damage Theory of. Fatigue Failure, Mats. Sci and

Eng. 1978, 34: 147-160

[3] Subramanyan S.: A cumulative damage rule based on the knee point of the S-N curve. J. Engng

Mater. Technol. 1976, 98: 316-21

[4] Rapillard L., Charlebois M., Zysset P.K.: Compressive fatigue behavior of human vertebral

trabecular bone, Journal of Biomechanics 2006, 39(11): 2133-2139

[5] Haddock S.M., Yeh O.C., Mummaneni P.V., Rosenberg W.S., Keaveny T.M.: Similarity in the

fatigue behavior of trabecular bone across site and species. Journal of Biomechanics 2004,

37(2): 181-187

[6] Kosmopoulos V., Schizas C., Keller T.S.: Modeling the onset and propagation of trabecular

bone microdamage during low-cycle fatigue. Journal of Biomechanics 2008, 41: 515-522

[7] Bowman S.M., Guo X.E., Cheng D.W., Keaveny T.M., Gibson L.J., Hayes W.C., McMahon

T.A.: Creep contributes to the fatigue behavior of bovine trabecular bone. Journal of

Biomechanical Engineering 1998, 120(5): 647-654

[8] Winwood K., Zioupos P., Currey J.D., Cotton J.R., Taylor M.: Strain patterns during tensile,

compressive, and shear fatigue of human cortical bone and implications for bone biomechanics.

Journal of Biomedical Materials Research 2006, 79A(2): 289-297

© (2012) Trans Tech Publications, Switzerland

doi:10.4028/www.scientific.net/MSF.726.43

MATERIALS WITH STRENGTHENING

Ewa MARCISZ 1,a, Adam NIESŁONY 2,b, Tadeusz ŁAGODA 3,c

1,2,3

Technical University of Opole, Faculty of Mechanical Engineering, ul. Mikolajczyk 5,

45-271 Opole

a

marciszewa@wp.pl, b a.nieslony@po.opole.pl, c t.lagoda@po.opole.pl

Abstract. The paper presents the concept of division of the total strain amplitudes. Simulations

were performed for high-alloy steel X6NiCr3220 for proposing a new curve of cyclic strain based

on the best fit to the experimental points and plotted the hysteresis loop. Proposed division on the

total strain amplitude of three parts: plastic strain amplitude, amplitude of the linear elastic strain

and strain amplitude of the coupled. In order to preserve the forms of popular formula Ramberg-

Osgood and Manson-Coffin-Basquin modified them in such a way that added to their member

responsible for the description of the coupled strain. Inclusion of additional term leading to closer

representation of the actual material properties.

Introduction

In most engineering constructions, machines and technical equipment we find the problem

of fatigue materials [3]. Fatigue in a small number of cycles in uniaxial loading condition can be

presented with the two basic characteristics: the curve Manson-Coffin-Basquin strain amplitude

describing the dependence of the number of cycles and cyclic strain curve, the relationship between

stress amplitude and the amplitude of the strain during the load-amplitude was. Ramberg-Osgood

equation describing the curve of cyclic strain, this equation can be divided into two parts of the

elastic and plastic. There are different ways to obtain experimental data to determine the cyclic

strain curve: you can use the tests performed with the use of multiple samples or one sample at

many levels of stress or strain or fatigue on the basis of the uniaxial loading of a sample of a

random process [4]. The most popular method, however, is the first of these, which also allows you

to get the results needed to make a chart of fatigue plot strain.

The paper presents results of high-alloy steel X6NiCr3220, which is taken from the

literature [1]. On their basis, was designated strain amplitude dependence of the amplitude of elastic

and plastic, and then plotted the hysteresis loop. Through simulations conducted seen some

belonging to the plastic strain amplitude, the amplitude of linear-elastic strain and the third part,

which is defined in this paper as a coupled and results from the curvature of the branches of the

hysteresis loop during the unloading of the material before reaching a point σ= 0.

Division the total strain amplitude of the plastic, linear-elastic and coupled

The results of fatigue tests for high-alloy steel subjected to uniaxial tensile X6NiCr3220-

compression at elevated temperatures are presented in paper [1]. In the literature, we find the

description of cyclic strain curve Ramberg-Osgood equation in the form

1

σa σ n'

ε ac = + a

E K'

, (1)

where:

εac - total strain amplitude,

σa - the amplitude of stress,

E - Young's modulus,

K’ – cyclic strength coefficient,

44 Fatigue Failure and Fracture Mechanics

Ramberga-Osgood equation can be separated into two parts, the first of which defines the amplitude

of elastic

σa

ε as =

E (2)

and the second part defines the plastic strain amplitude

1

σ n'

ε ap = a

K' (3)

We can also be written as

σa

ε ap = ε ac −

E . (4)

In order to model a hysteresis loop we use the formula (1), in which we move from the amplitudes

of stress and strain ranges. Equation takes the form

1

∆σ ∆σ n '

∆ε = + 2

E 2K ' . (5)

Strain amplitude dependence of the number of cycles to damage is given by the Manson-Coffin-

Basquin as

σ 'f

( 2N f )

b

ε ac = + ε ' f (2 N f )c , (6)

E

where:

σ’f – coefficient of fatigue strength under tension-compression,

b – fatigue strength exponent,

ε’f – coefficient of the fatigue plastic strain,

c – exponent of fatigue plastic strain,

Nf – number of cycles to failure.

Equation (6) can also be separated into two parts, the first of which shows the amplitude the elastic

strain

σ 'f

( 2Nf )

b

εas =

E , (7)

and the other for the plastic strain amplitude

ε ap = ε ' f ( 2N f )

c

. (8)

In order to illustrate the amplitudes of elastic and plastic strain of cyclic strain curve was plotted

and modeled on the basis of the hysteresis loop. Figure 1 illustrats by an example of hysteresis loop

together with the designated parts of the elastic and plastic. Plastic deformation, as defined, are

those strain that are in the material after removing the load, ie when the stress again reaches the zero

value [2].

Dariusz Skibicki 45

Fig. 1.The hysteresis loop and the cyclic strain curve with the selected deformation plastic, linear-

elastic and coupled.

The figure shows that if the dual cyclic yield stress (Re’) is less than the amplitude of stresses, i.e.

2 R e' < σ a (9)

then we can identify some plastic ɛ ap, linear-elastic ɛas part conjugate and a new ɛsp. It follows that

the newly defined amplitude of strain is described in the coupled model

ε sp = ε ac − ε as − ε ap

(10)

In order to illustrate the idea of sharing the total strain amplitude into three parts modeled

hysteresis loop [1]. The analysis was performed for steel X6NiCr3220, for which the

calculated: K '= 1455 MPa, n' = 0.242 according to (1) and σ 'f = 1295MPa, b = -0,1460,

ε'f = 0.4768, c = -0.57 according to the formula (6). Literature results and their analysis allowed us

to determination the cyclic strain curve

Cyclic strain curve was determined based on the best fit to the experimental points. Obtained

in this way, the parameter values K*=1720 MPa and n* = 0.2728 different from those designated by

the classical method. Following the analysis suggested different light patterns describing the cyclic

strain curve for the material shown in Figure 2 against plastic strain amplitude, amplitude of the

linear elastic strain and strain amplitude of the coupled. The curve is plotted according to the

formula Ramberg-Osgood (1) calculating the new material constants, Young's modulus was

obtained statically on 193GPa

From the figure 1 it can be observed non-linear elastic character. Proposed division of the

total strain amplitude of the ɛac for some ɛas linearly elastic, plastic and part of the ɛap coupled ɛsp.

In order to preserve the forms hitherto used Ramberg-Osgood equation describing the cyclic

deformation of the full use of them by adding the member responsible for the description of the

coupled strain. The plot takes the form of cyclic strain

1 1

σ a σ a n ' σ a n ''

ε ac = ε as + ε ap + ε sp = + +

E K' K '' , (11)

where the parameter for the studied steels are as follows: K * '= 2846MPa, n *' = 0.2728, K *'' =

1525MPa, and n *'' = 0.2442.

46 Fatigue Failure and Fracture Mechanics

Fig. 2 The curves of cyclic deformation on the background of elastic, plastic and coupled.

Fig. 3 Cyclic strain of the total, linear-elastic, plastic and incorporated in the system (ɛa - σa)

Cyclic strain the total, linear-elastic, plastic and coupled are shown in Figure 3 in the system (ɛa -

σa). In order to preserve the forms hitherto used equation Manson-Coffin-Basquin strain amplitude

describing the total used by placing states responsible for the description of the strain coupled

σ 'f

(2N f ) ( ) ( )

b c d

ε ac = + ε ' f 2N f + ε '' f 2 N f (12)

E

where the parameter for the tested steels are: σ 'f = 1191MPa, b = -0.1359, ɛ' f = 0.0411, c = -

0.4984, ɛ'' f = 0.3369, d = -0.5486 .

gdzie parametr dla badanej stali wynoszą: σ’f = 1191MPa, b = -0,1359, ε’f = 0,0411, c = -0,4984, ε’’f

= 0,3369, d = -0,5486.

Dariusz Skibicki 47

Fig.4 The amplitude of strain of the total, elastic, plastic and coupled depending on the number of

cycles for the damaged steel X6NiCr3220.

In Tables 1 and 2 are summarized the calculated coefficients and exponents of the study of cyclic

deformation strengthening of steel, used in the tests carried out.

Strengthening of the cyclic

deformation coefficient, MPa

K’ K* K*’ K*’’

1455 1720 2846 1525

The exponent of cyclic

deformation strengthening

n’ n* n*’ n*’’

0.242 0.2728 0.2728 0.2442

Conclusinon

In this paper we analyzed the plot of cyclic strain on the example of steel X6NiCr3220.

Modeled curve of cyclic strain and stable course of the hysteresis loop by dividing the total strain

amplitude in the linear elastic, the coupled and the plastic strain. Following conclusions are

detailed:

1. Observed for the analyzed material nonlinear elastic character, which led to propose the

division the amplitude of elastic strain on the amplitude of linear-elastic and coupled.

2. For materials modeled by equation (11), for which the coupled strain takes negligibly small

values, the equation reduced to the well-known Ramberg-Osgood proposal.

3. For further analysis is necessary to perform experiments for different materials and to

analyze the results on the basis of experimental and modeled hysteresis loop.

4. The proposed model is valid for materials which have a double cyclic yield strength is less

than the applied stress amplitude.

48 Fatigue Failure and Fracture Mechanics

Acknowledgements

This work was supported by the National Centre for Science, contract No. 2011/01/B/ST8/06850.

References

[1] BÄUMEL A., SEEGER T.: Material Data for Cyclic Loading, Materials Science Monographs,

Vol. 42A – E, 1987, Supplement 1, 1990, Elsevier Science Publishers.

[2] NIESŁONY A., EL DSOKI Ch., KAUFMANN H., KRUG P.: New mathod for ewaluation of

the Manson-Coffin-Basquin and Ramberg Osgood equations with respect to compatibility,

International Journal of Fatigue, Vol. 30, 2008, pp. 1967-1977.

[3] ROZUMEK D., MARCINIAK Z., LACHOWICZ C. T.: The energy approach in the calculation

of fatigue lives under non-proportional bending with torsion. Int. J. of Fatigue, Vol. 32, No. 8, 2010,

pp. 1343-1350.

[4] WALAT K., ŁAGODA T., KAROLCZUK A.: Fatigue life according to cyclic strain

characteristics determined from variable amplitude loading, Materials Testing, (Materialprufung),

Vol. 51, No. 5, 2009, pp. 286-290.

CHAPTER 2:

Fatigue Properties of Materials

© (2012) Trans Tech Publications, Switzerland

doi:10.4028/www.scientific.net/MSF.726.51

BOROŃSKI Dariusz1, a

1

University of Technology and Life Sciences, Faculty of Mechanical Engineering, Department of

Machine Design, al. prof. S. Kaliskiego 7, Bydgoszcz, Poland

a

dariusz.boronski@utp.edu.pl

Abstract. In the paper there are discussed possibilities of performing static and cyclic investigations

of a material properties with the use of small size specimens. An original research system

developed at the University of Technology and Life Sciences in Bydgoszcz has been used for static

and fatigue testing. Thanks to tests performed on microspecimens it is possible, among others,

determining the local properties of material in objects with the material discontinuities. This

research provides exemplary results of fatigue tests carried out on microspecimens taken from a

laser welded joint.

Introduction

Development of fatigue design methods, based on the local approach, involves an increasing

demand for information on local, cyclic properties of structural materials. It refers both to non-

homogeneous and homogeneous materials which can change their basic properties in result of

technology used for production of structural parts or under the influence of loads. Joints made using

welding techniques constitute a special group of such elements [1]. However, it was the

development of nuclear energy that initiated tests on microspecimens, including materials exposed

to radiation [2]. There have been developed different methods for testing [3,4] and preparation of

specimens. Also the conference entitled: ‘Small Specimen Test Techniques’ was dedicated to tests

on ‘small specimen’. Testing of mechanical properties on a small specimen is also necessary in case

of materials used in MEMS electromechanical microstructures [5,6,7].

The worked out technical solutions in the field of test equipment as well as specimen preparation

methods are discussed more thoroughly in monograph [8].

Thorough knowledge of material properties is especially useful for a fatigue analysis of structural

parts involving the local approach which provides the possibility to study the fatigue process

directly in the zone of fatigue crack initiation. In such a case, the quality of a structure fatigue

analysis largely depends on the familiarity with the material local properties in these areas. It refers

mainly to two types of fatigue characteristics: fatigue life curves describing the dependence between

fatigue life and the level of load (stress or strain) and cyclic stress-strain curves characterizing the

relation between strains and stresses for cyclically variable loading. Both characteristics are most

often determined on the basis of the analysis of the stress-strain hysteresis loop parameters recorded

during cyclically variable loading applied to the material specimen.

However, the analysis of local fatigue properties often requires application of very small size

specimens which causes that the so far existing test methods and research instruments are

insufficient to perform efficient tests in this field.

Apart from performing tests of the material local properties, conducting tests on small size

objects, is of great significance for microelements and microstructures which in many cases are

produced with the use of materials whose characteristic dimension is considerably decreased [9].

52 Fatigue Failure and Fracture Mechanics

In order to perform fatigue tests with the use of microelements there has been developed an

original research system MFS. The system includes: primary loading unit ‘nanodrive’, secondary

loading unit ‘microdrive’, unit for strain measurement by digital image correlation method (CID),

unit for measurement of strain by laser grating interferometry technique (moiré interferometry)

(LFI), units for precise alignment and fastening of objects, computerized control an supply system.

A scheme of the system is shown in Figure 1a. An overall view of the system is shown in Figure 1b.

a) b)

11

1

7

5 9 2

4 10 10 4 3

8 6

7

12

Fig.1. Scheme of MFS system (a): carrying base (1), primary loading unit „nanodrive” (2), (3), force

measurement unit (4), nanoscale displacement measurement unit (5), microscale displacement

measurement unit (6), strain measurement unit by the digital image correlation technique (7), strain

measurement unit by the laser grating interferometry technique (8), alignment unit (9), fastening

unit (10), thermographic analysis (11), control and supply unit (12). The overall view of MFS

system (b)

Thanks to the worked out solutions, including a doubled loading system basing on piezo actuator

and microstepping motor with a precise ball screw, MFS system enables determination of many

static and fatigue properties of the material, including static tension curves, cyclic stress-strain

curves and different types of fatigue life curves: stress-life, strain-life. Optoelectronic measurement

units used in this system enable measurement of strains on measurement bases with micrometric

length.

Microspecimen preparation

On the basis of carried out research, including an assessment of the influence of the specimens

preparation technology on their microstructure, the method of laser cutting was chosen to prepare

them. First, plates with thickness 0,2 ± 0,02 mm were cut from a 4 mm thick joint by WEDM

technique (wire thickness 0,2mm.), according to the scheme presented in Figure 2. From so

prepared plates, micro-specimens were cut from particular zones of the joint which were localized

by micro-etching. A laser with power range from 1 to 20W and frequency 20-80 kHz, equipped with

heads moving with velocity 20 000 mm/s was used to cut out microspecimens.

base

0.2

metal

4

weld microspecimen

heat

affected

zone (HAZ)

Fig.2. Scheme of specimen preparation

Dariusz Skibicki 53

Two types of specimens were prepared for the tests: dumbbell shape specimen (Fig.4) and

specimens with triangular griping part. Specimens were cut with the use of a small power laser in

order to avoid the impact of heat on the specimen structure, which was verified by their

metallographic analysis (Fig.5).

15.5 mm

50 µm

50 µm

Fig.5. Microspecimen microstructure: specimen taken from base metal zone

Exemplary results of fatigue tests performed on microspecimens taken from a laser welded joint,

made of the S355J2G3 steel (low-alloy steel with improved mechanical properties primarily

designed for welded structures), have been shown in Figure 6. In Figure 6a there are exemplary

hysteresis loops recorded for different phases of constant amplitude loading course, and in Figure

6b, selected courses of hysteresis loops recorded for two zones of the welded joint: base metal and

the weld zone.

The presented courses show both diversification of the material properties and their variability

under the influence of cyclically variable loading. Further analysis of hysteresis loops recorded for

different loading levels make possible determination of local cyclic material properties for particular

zones of laser welded joints.

Summary

The presented system MFS makes it possible to perform fatigue tests with the use of specimens with

dimensions accounting for local differences in material properties.

The possibility to determine the material local properties with the use of microspecimens allows to

significantly extend the range of fatigue analysis methods based on the local approach.

54 Fatigue Failure and Fracture Mechanics

Application of small specimens and test methods enabling performance of tests in micro-scale,

opens new grounds for the development of design methods for prevention against fatigue cracks in

objects with different material properties.

a) 600 1900 3200 cycle b)

20

force, N weld

stress, MPa

15

400

10

200 base

5 metal

0 0

-0.005 -0.003 -0.001 0.001 0.003 0.005 -25 -15 -5 5 15 25

strain -5 displacement, µm

-200

-10

-400

-15

-600 -20

Fig.6. Exemplary results of fatigue test of microspecimens: a) hysteresis loops recorded for different

phases of constant amplitude loading, b) hysteresis loops recorded for two zones of the welded

joint: base metal and the weld zone

This research work is financially supported by the Polish state budget for science as a research

project

References

[1] D. Boroński, Cyclic material properties distribution in laser-welded joints, International Journal

of Fatigue, Vol 28/4 (2006) 346-35.

[2] G.H. Lucas, Review of small specimen test techniques for irradiation testing, Journal Me-

tallurgical and Materials Transactions A, 21, 5 (1990).

[3] S. Saito, K. Kikuchi, Y. Onishi, T. Nishino, Development of piezoelectric ceramics driven

fatigue. Journal of Nuclear Materials 307-311 (2002) 1609-1612.

[4] W.N. Sharpe, Tensile Testing at the Micrometer Scale: Opportunities in Experimental

Mechanics, Experimental Mechanics, 43, 3 (2003) 228-237.

[5] S.M. Allameh, An introduction to mechanical-properties-related issues in MEMS structures.

Journal of Materials Science, 38 (2003) 4115-4123.

[6] Ando T., Shikida M., Sato K. Tensile-mode fatigue testing of silikon films as structural

materials for MEMS. Sensors and Actuators, A 93 (2001) 70-75.

[7] Cho H.S., Hemker K.J., Lian K., Goettert J., Dirras G. Measured mechanical properties of

LIGA Ni structures. Sensors and Actuators, A 103 (2003) 59-63.

[8] D. Boroński, Local material properties in fatigue analysis, Publishing House of ITeE-PIB,

Bydgoszcz-Radom (2009) (in polish).

[9] S.M. Allameh, J. Lou, F. Kavishe, T. Buchheit, W.O. Soboyejo, An investigation of fatigue in

LIGA Ni MEMS thin films, Materials Science and Engineering, A 371 (2004) 256-266.

© (2012) Trans Tech Publications, Switzerland

doi:10.4028/www.scientific.net/MSF.726.55

FATIGUE OF BEARINGS

Tadeusz Z. Wozniak1, a, Jerzy Jelenkowski2, b, Krzysztof Rozniatowski3, c,

Zbigniew Ranachowski4, d

1

Kazimierz Wielki University, Chodkiewicza 30, 85-064 Bydgoszcz, Poland

2

Institute of Precision Mechanics, Duchnicka 3, 01-796 Warsaw, Poland

3

Faculty of Materials Science and Engineering, Warsaw University of Technology,

Woloska 141, 02-507 Warsaw, Poland

4

Institute of Fundamental Technological Research, Polish Academy of Sciences,

Pawinskiego 5B, 02-103 Warsaw, Poland

a

wozniak@ukw.edu.pl, b jjele@imp.edu.pl,

c

krozn@inmat.pw.edu.pl, d zranach@ippt.gov.pl

Keywords: Rolling contact fatigue; Bearing steels; Pitting; Isothermal heat treatment; Bainite;

Midrib; Acoustic methods

Abstract. There has been proposed an innovative thermal treatment of bearing steel 100CrMnSi6-

4, where the existing standard heat treatment has been replaced by austempering. The structure of

low-temperature tempered martensite has been replaced by a microstructure composed of

martensite and lower bainite with midrib. The kinetics of bainitic transformation and isothermal

martensitic transition at selected austempering temperatures was controlled by acoustic emission.

The research on contact strength was made under the conditions of rolling-sliding friction. The

microstructure was revealed with the use of a light microscope and the forms of pitting wear were

displayed by a scanning electron microscope. It was found that the optimum microstructure

providing the best used contact strength of the tested steel is conditioned by the formation of a

lower bainite with midrib at the temperatures near MS. A plausible cause of the increased resistance

to pitting is bifurcation of fatigue cracks on dispersion bainitic carbides in combination with

primary carbides, in bainitic-martensitic matrix.

Introduction

The main causes of destruction, friction nodes in machinery parts in working conditions of

variable contact loads are seizing and pitting [1-3]. That failure takes place mostly when

cooperating surfaces touch each other in rotary motion (bearings, gears, etc.). Pitting is a

catastrophic destruction and appears suddenly, after many millions cycles of loading. Starting from

first spalling, a further process of the destruction of top layers of material is very fast. Pitting

depends on the top layer of material structure, endurance material properties, surface roughness and

the values of the friction coefficient. Hardening parameters determine micro-structural habitus, a

dispersion of carbides, carbon diffusion and internal stresses. In recent years, the issues on

improving the mechanical properties of materials by changing their microstructure have been

developed. On the one hand, it is related to an increase in knowledge of the relationship: structure-

properties and to the development of new technologies that provide a better control of the material

microstructure. There exists both an optimal volume fraction and their optimal dispersity for a

definite type of matrix and a definite type of carbides [4, 5]. A material with a uniform carbides

distribution or in the form of clusters and thin layers shows higher crack resistance than the

microstructures with thick bands or random distribution [5]. This research aimed to verify the

possibilities of replacing the standard heat treatment of steel 100CrMnSi6-4 at the isothermal heat

treatment at the elevated temperature of austenitising, controlled by the method of sound emission

[6, 7, 8]. Both the Hertz stress values and the number of cycles up to the first registered damage of

sample were adopted as a criterion for the correctness of the modified heat treatment.

56 Fatigue Failure and Fracture Mechanics

Steel 100CrMnSi6-4 (ISO EN 683-17:1999) was used in the research. The content of sulphur

and phosphorus was accordingly: 0.014% P and max. 0,007% S. For steel used for the research the

degree of carbide banding was 7.3 acc. to Stahl-Eisen-Prüfblatt SEP 1520. The tested steel was

delivered as rolled products of 46 mm in diameter in a softened state. The analysis of chemical

composition was carried out using an instrument made by Spectrolab. The results of the analysis for

chemical composition are shown in Table 1.

C Cr Mn Si Cu Ni Al Mo P S

0.95 1.47 1.10 0.57 0.21 0.07 0.02 0.01 0.014 0.007

In order to determine the temperature MS in the tested steel, dilatometric tests were made for six

specimens, 12mm long and 2 mm in diameter. Austenitising of specimens was carried out in an

oven of dilatometer at temperature 950°C in a time period of 30 minutes, with registering accuracy

of ± 5K. The tests of fatigue contact strength were made with the use of series of specimens

austempered and then grinded. Specimens used in the tests were austenitised at 950°C during 0.5 h

and quenched in hot oil, maintaining constant temperature (30 - 180°C). The tests were carried out

using appliances type ULP-2. There were used specimens of size Φ 8 x 40mm cooperating with two

clamping rollers: driven and driving of 150mm in diameter, made of steel ŁH15, of hardness

60HRC. The measure of pitting resistance was the number of cycles until the moment when first

spallings occur. The construction of the appliance used and the specimen size enable us to carry out

5 measurements. The tests were made with Hertzian contact pressure 3440 MPa, frequency f = 300

Hz to max. testing limit NG = 2 x 107 cycles, acc. to PN-83/H-04324.

Specimens for microscopic research were prepared in a standard way: they were etched with

Nital and with Vilell's reagent [9]. Microscopic examination was carried out using a metallographic

inverted light microscope type EPIPHOT 200 made by NIKON equipped with a digital colour CCD

camera, resolution 2 Mp. Surface geometry of rollers after the tests was imaged thanks to laser

interference profilometry, by using a laser appliance WYKO 9300 made by Veeco, a scanning

microscope TM-1000 Hitachi in an observation mode BSE [10].

The test method EA [10, 11] made it possible to determine precisely the optimum temperature

for austempering (Fig. 1). Testing EA effects were carried out by using modern instrumentation that

was able to make austempering in the range of temperature of martensite and bainite formation 30-190ºC.

For experimental tests, specimens of 45 mm in diameter and 2 mm thick were used [7, 8, 9].

AE signals sent while austempering were recorded using a special instrumentation to which an

ultrasonic transducer WD (20 kHz-900 kHz) was connected. The signals were registered by AE

Signal Analyser 10/20 kHz-800 kHz and recorded in PC computer memory by a card ADLINK

9812 with the frequency of 1200 kHz. In order to make spectrogram graph, the Short Time Fourier

Transform (STFT) algorithm with the Hamming window was used.

Dariusz Skibicki 57

A) B)

Figure 1. A) The stand for testing austempering using the method of acoustic emission (AE),

B) Exemplary spectrograms of AE signals for temperatures a) 30 ºC, 160 ºC, 190 ºC.

Research Results

Examples of spectrogram graphs for AE signals received during the process of austempering are

given in Figure 1 B) in the system coordinates: time - amplitude. In the initial stage of

transformation, about 4 minutes, the austempering temperature increases and the individual AE

events are gradually expanded. Midribs, formed at the start of the process, do not show any large

dilatometric effects, but they are very significant as the structure precursors of bainitic

transformation [12, 13]. Thin platelets martensite nucleation in the form of midribs during the

quench time makes the first stage of the process. The second stage is related to the enrichment of

adjacent austenite in carbon, which leads to its further transition into lower bainite.

a) b)

Figure 2. The microstructure of steel after austempering at different temperature: a) bainite plate

of lower bainite with midrib at 160°C for 3 min, white areas of residual austenite with athermal

martensite revealed with Vilell's reagent, b) martensite at 100°C during 1450 s, etching with Nital.

At temperature 30°C, an intensive range of signal emissions lasts up to approx. 60s. The lower

bainite with midrib is formed the fastest at temperature 130°C, which is due to the formation of

large quantities of midribs in a shorter time. The longest process of the midribs formation occurs at

58 Fatigue Failure and Fracture Mechanics

temperature 160°C (Figure 2a), which is connected with the diffusion of bainite carbon and the

enrichment of austenite in carbon. When the temperature is raised to 190°C, i.e. above MS, the rates

of midribs formation and bainitic transition are significantly reduced. Midribs are not observed at

temperature below MS of a massive character and above MS (Fig. 2b). Temperature MS determined

in dilatometric tests as the deviation from linearity is MS =157°C, while the substantial martensitic

transformation of massive character starts at 110°C.

a) b)

Figure 3. The impact of the various options of austempering on contact strength: a) the number of

load cycles, 1) quenching below MS, 2) 160°C during 3 min., 3) 160°C during 60 min,, (4) 180°C

during 3 min,, (5) 180°C during 10 min., (b) the share in % of population in the number of cycles

above the experiment limit: 20x 106 cycles, according to the variant of heat treatment.

The research results of fatigue contact strength as a function of temperature and austempering

time is shown in Figure 3 a). The research results of austempered specimens at 160ºC and 180ºC

significantly exceed the results for steels ŁH15 obtained after customary quenching with tempering

at 200ºC [10, 14]. This negative impact of customary quenching is probably caused by blocking of

displacement movement in the structure with the martensitic phase prevalence, which delays the

relaxation of stresses and thus facilitates the nucleation of micro-cracks. If no pitting was observed

after completing the tests, i.e. after 20 million cycles, a limit value was taken to calculate the

average value of the test results. It made a certain flattening of average results for different variants

of the treatment, which can be seen in Figure 3 a). For this reason, Figure 3 b) additionally shows a

share of population in % concerning the number of cycles above the experiment limit 20x 106

cycles for each variant of heat treatment. The largest population reaching as much as 83% was

obtained at austempering temperature 160°C during 3 minutes. (variant no. 2 of heat treatment)

i.e. on the limit near temperature MS.

The fatigue contact strength is conditioned by the presence of elastic and plastic strains as well as

by the factors accompanying rolling-sliding friction in the system of cooperating elements. Intense

lubrication causes spalling (delamination) of about 0.1µm thick shells and approx. 3µm in diameter.

Larger damages arise in heavier loads and faulty lubrication, when thin metallic layer flakes and

hard carbides, nitrites, etc. are revealed (Fig. 4 a). An intensive and variable load may generate

pitting, (Fig. 4 b)), with characteristic elongated strips of larger width and depth. The stria layouts

suggest the depth of craters, traces of carbides presence and the speeds with which the generated

cracks spread towards the surface. Pitting is connected with the appearance of cracks on or under

the surface (Fig. 5).

Dariusz Skibicki 59

a) b)

Figure 4. Examples of rolling wear owing to joining fatigue cracks near the surface: a) spalling

from the strengthened top layer with participation of shear stresses, carbides contact directly with

the damaged layer, b) the area formed around the carbides induces pitting of particles of a few

micrometers in diameter [10].

First cracks of material appear mostly on a certain depth below the surface near the so-called

Bielajew point in the largest material effort area [1, 14-16]. The increase in shear stresses causes

displacement of the Bielajew point towards the surface. The highest durability occurs for pitting

initiating on the surface (Fig. 4 a, Fig. 5). For pitting initiation on a certain depth below the surface,

these areas are ovally shaped and the service life of bearing elements is lower (Fig. 4 b). The

differentiation in pitting morphology was also found in the research on other bearing steel 100Cr6

[17], similarly like in the tested here steel 100CrMnSi6-4 after austempering.

Figure 5. The surface geometry of the roller after the tests, treatment variant no. 2, Fig. 3. The

image achieved by laser interference profilometry. The results of tests of roller surface geometry for

five measurements of wear, marking: Ra - the average arithmetic deviation from the roughness

profile, Rq - the average square deviation from the roughness profile, Rz - maximum height of the

roughness profile elevation, Rt - maximum depth recesses of the roughness profile [10].

While moving the rolling body, the strains migrate wavely and micro-glides occur as a result of

elastic and often plastic strains [14, 18]. Additional strains during the rolling body movement also

result from transiting austenite into martensite under the load. The development of pitting cracks is

also dependent on the material strengthening. The plastic strain and the growth in dislocation

density make carbon diffusion in these areas easier, which in turn affects the formation of areas that

are not susceptible to etching, the so-called white etching areas (WEA). In the maximum shear

strain area, the layers with a modified structure are formed, the thickness of which increases from

about 0.1 mm to 0.7mm with the increase in the value of contact stress and load cycles number.

60 Fatigue Failure and Fracture Mechanics

Apart from temperature and time, a form of cementite in annealed steel is an important

factor, affecting the kinetics of austenitising. The steel structure fulfilling the optimum of

various requirements in respect of technological and functional properties of bearing steel is an

equally dispersed spheroidal cementite (Fig. 6 a). In carbides (Fe, Cr, Mn) 3 C, carbon and a

great proportion of chromium and manganese are bounded in almost all their amounts [1].

The shape and the dispersion degree of cementite activate carbides dissolution processes, thus

affecting the related changes in carbon contents and in alloy elements in matrix. After

hardening from 850°C (after austenitising time of 30 min.), approx. 7% of carbides remained

and the martensitic matrix contained approx. 1% Cr and 0.6% C [14]. In the areas with

segregation and carbide banding, chromium or carbon concentrations may differ in certain areas

from their average values. The presence of chromium results in reducing carbon solubility in

austenite, which consequently affects the increase in the amount of carbides in steel.

a) b)

annealing, the structure of spheroidal cementite dispersed in a ferrite matrix, SEM image. On the

etched surface of scarp, the carbides of different dispersion are revealed. b) Microstructure of

steel 100CrMnSi6-4 austempered at 130°C, the inhomogeneity caused by carbide banding along

rolling direction, banding degree 7.3 acc. to Stahl-Eisen-Prüfblatt SEP 1520. Etching with Nital

reagent [10].

In the tested steel 100CrMnSi6-4 after hardening, inhomogeneous dispersion of primary carbides

was observed. An additional factor favourable to inhomogeneous distribution of carbides appearing

as clusters is structural heterogeneity caused by diverse fraction of bainite and martensite in micro-

areas. Metallographic analyses of steel 100CrMnSi6-4 revealed the banding of microstructure (Fig.

6b). Banding is defined as light bands of almost pure carbide-poor martensitic structures. Light

martensitic bands occur next to dark bands of martensitic-bainitic matrix (lower bainite with a

midrib) with the increased concentration of carbides. In bands of bearing steel with the elevated

content of carbides, the solid solution after austenitising is richer in carbon, chromium and

manganese than in the adjacent bands. This is caused by considerable differentiation in matrix

hardenability and microhardness of bands, which affects the resistance to fretting, fatigue and

contact wear. The bonded structure also shows a favourable influence on a use life of tools for

plastic hot treatment [19]. Complex microstructures containing clusters and laminated arrangement

of carbides may improve crack resistance [20, 21]. A positive effect of the impact of complex

microstructure is a crack direction path deflection from the initial direction, which increases both

the length of the crack path and the total area of the crack. This zone develops further through the

coalescence of microcracks in carbides.

Dariusz Skibicki 61

Summary

1. It was manifested that one of the factors to reduce the adverse impact of primary carbide

banding on contact strength of bearing steel 100CrMnSi6-4 may become properly selected heat

treatment. Not eliminating the beneficial effect of martensitic transformation, we suggest the heat

treatment with isothermal hardening between MS and Mf with the use of bainitic transformation,

which resulted in eliminating the necessity of tempering.

2. A thermal treatment controlled by the method of acoustic emission, which made it possible to

control the kinetics of forming midribs, was suggested. This, in turn, allowed us to achieve the

microstructure with a complex dispersion of hard particles in the form of clusters in bainitic-

martensitic matrix. In many cases, the level of fatigue contact strength was achieved for above 20x

106 cycles with no macroscopic traces of surface damage in the form of cracks or spallings.

3. Local damages in the tested paths can be divided into two main groups: surface destructions of

fraction micrometer size, and pittings i.e. failures in the form of spallings of the mm size. With the

initiation of pitting at a certain depth below the surface, the life of bearing elements is lower. The

highest life occurs with the initiation of damages on the surface. For austempered specimens at 160-

180°C, cracks did not propagate deep into material.

References

[1] H. K. D. H. Bhadeshia, Steels for Barings, Progress in Materials Science. 57 (2012) 268-435.

[2] S. Pytko, M. Szczerek, Pitting – the form of damaging rolling elements, Tribology. 4/5 (1993)

317-334 (in Polish).

[3] Z. Gawroński, Technological toplayer in gears and cam mechanisms, Monographies. Lodz

University of Technology, 2005 (in Polish).

[4] J. Pacyna, Metallurgy of tool steel cracking, Scientific Bulletins, Metallurgy and Foundry

Engineering. No. 120, AGH Krakow, 1988 (in Polish).

[5] H. Berns, A. Melander, D. Weichert, N. Asnafi, C. Broeckmann, A. GroB-Weege. A new

material for cold forging tools, Computational Materials Science. 11 (1998) 166-180.

[6] T.Z. Wozniak, K. Rozniatowski, J. Jelenkowski, New Technologies in Heat Treatment of

Steel Rolling Bearings 100CrMnSi6-4, Toolmaker (in Polish). 3 (2009) 17-20.

[7] T.Z. Wozniak, K. Rozniatowski, Z. Ranachowski, Acoustic Emission in Bearing Steel during

Isothermal Formation of Midrib, Metals and Materials International. 17(3) (2011) 365-373.

[8] T.Z. Wozniak, K. Rozniatowski, Z. Ranachowski, Application of Acoustic Emission to

Monitor Bainitic and Martensitic Transformation, Metallic Materials -Kovove Materialy. 49(5)

(2011) 319-331.

[9] C.K. Shui, W.T. Reynolds Jr., G.J. Shiﬂet, H.I. Aaronson, Etchants for Quantitative

Metallography of Bainite and Martensite Microstructures in Fe–C–Mo Alloys, Metallography. 21

(1) (1988) 91-102.

[10] Report on implementing the Research and Development Project No. R15 010 02 financed by

Polish Ministry of Science and Higher Education, System for determinating an optimal

austempering temperature by applying the method of acoustic emission. Head of the Project: prof.

dr hab. inż. Rożniatowski K. Faculty of Materials Science and Engineering, Warsaw University of

Technology, made by: J. Jelenkowski, T. Z. Wozniak (2007-2010), in Polish.

[11] T.Z. Wozniak, J. Jelenkowski, K. Rozniatowski, Z. Ranachowski, Station for austempering of

steel with the use of acoustic emission, Engineering Quaterly - Metal Treatment, in Polish. 1 (2010)

2-8.

62 Fatigue Failure and Fracture Mechanics

[12] T.Z. Wozniak, Z. Ranachowski, Acoustic Emission During Austenite Decomposition into

Lower Bainite with Midrib, Archives of Acoustics. 31(3) (2006) 1-15.

[13] T.Z. Wozniak, Acoustic Phenomena Near MS in Hypereutectoid Steels, Materials

Characterization. 59/6 (2008) 708-716.

[14] W. Luty, Bearing Steels, WNT, Warszawa, 1969 (in Polish).

[15] S. Pytko, Bases of Tribology and Lubricating Engineering, AGH University Course Books (in

Polish), No. 1164, 1989.

[16] T. Smolnicki, Physical aspects of the coherence of large-size rolling bearings and deformable

supporting structures, Monographies 28. Publishing House of the Wroclaw University of

Technology, Wroclaw, 2002 (in Polish).

[17] F. Akbasoglu, D. Edmonds, Rolling contact fatigue and fatigue crack propagation in 1C-1.5Cr

bearing steel in the bainitic condition. Metall. Mater. Trans. A, 21A (1990) 889-893.

[18] S. Pytko, Structures in machine construction and their effect on the devalopment of

machinery construction in Poland, Polish and Worldwide Achievements of Science, Gliwice, 2010,

pp.123–191 (in Polish).

[19] M. Hawryluk, M. Zwierzchowski, Structural analysis of dies used for hot forging in the aspect

of their reliability, Maintenance and Reliability. 2B (2009) 31-41 (in Polish).

[20] L. Mishnaevsky Jr, U. Weber, S. Schmauder, Numerical analysis of the effect of

microstructures of particle-reinforced metallic materials on the crack growth and fracture resistance,

International Journal of Fracture. 125 (2004) 33-50.

[21] K. Rozniatowski, Methods of the arrangement inhomogenity characterization of the structural

elements in multiphase materials, Scientific Papers of the Warsaw University of Technology.

Materials Engineering. 22, 2008 (in Polish).

© (2012) Trans Tech Publications, Switzerland

doi:10.4028/www.scientific.net/MSF.726.63

using mini specimen

Tomasz Tomaszewski1, a, Janusz Sempruch1, b

1

Uniwersytet Technologiczno-Przyrodniczy, al. Prof. S. Kaliskiego 7, 85-791 Bydgoszcz, Poland

a

tomaszewski@utp.edu.pl, bsemjan@utp.edu.pl

Abstract. There are situations where taking normative specimens is impossible due to the

dimensions of the objects investigated (e.g. extruded sections) and one of the solutions is to use

mini specimens. As for non-standard specimen testing, it is essential to define the effect of size on

fatigue strength. The research methodology facilitates the determination of fatigue characteristics

(S-N) for EN AW-6063 aluminum alloy. The material is used to manufacture the extruded section

in the process of extrusion of the material through the extruding die. The methodology assumes the

geometry of the mini specimen and the normative specimen. As for the material strength

identification, a static tensile test for the specimens made directly from finished elements and

preliminarily strained in cycles was carried out. As a result of the cyclic material reinforcement, an

increase in yield strength Re was observed, which, in turn, rejects Re as the upper criterion of the

high-cycle fatigue range. The essential fatigue tests were performed based on unilateral cyclic

tension (R = 0.1). The effect of size on fatigue strength was defined. Theoretically aluminum alloy

non-sensitive to changes in the size of the cross-section showed a different strength in mini and

normative specimens.

Introduction

The size effect is known in the science of fatigue material failure and design elements. It is

usually perceived in the way that we determine the testing area in which that effect is unobserved.

The results are objective.

In the present paper, the starting point is the statement that there exist situations where the

normative specimen (located by its dimensions in the area of a lack of effect/ a limited size effect) is

uncomfortable or unfeasible. This is true for the case in which the dimensions of the objects

investigated make it possible to manufacture mini specimens only rather than specimens with the

geometry recommended by norms [1].

mini specimen

64 Fatigue Failure and Fracture Mechanics

The tests reported were performed for the aluminum alloy from which the extruded sections are

made (Fig. 1). The dimensions of the objects facilitate making the specimens smaller (mini

specimens) than normative ones. They are produced in the process of extrusion of the material

through the opening of the extruding die, defining the final shape of the cross-section of the

product. Due to the high plastic strain accompanying the process, the initial material shows other

strength and cyclic properties than the initial material. It is justifiable to perform tests which involve

the specimens taken from finished elements.

As was demonstrated in an earlier publication [2], tests with mini specimens are a solution

reported in a number of papers which are experimental in nature. These have been applied for

example in the nuclear industry. The effect of irradiation on fatigue strength of the steel from

which the shields for nuclear reactors are made was investigated [3]. The experiments with the use

of mini specimens were also performed in the range of gigacycle strength for high-strength steel;

such experiments were performed with the use of non-standard test stands with ultrasonic

application where the frequencies of the change load were at the level of 20 kHz [4]. A more

detailed description of the research methodology and result is covered in a number of other papers

[5, 6]. The review presented in [2] points to a lack of firm research methodology concerning mini

specimens.

As for the performance of tests with mini specimens, it is essential to define the effect of the size

of the cross-section on fatigue strength. That effect is accounted for by the probability of failure of

the weakest link of the material structure in the cross-section analyzed. The greater the material

volume, the greater the probability of material defects triggered by focal points for fatigue cracking.

It is assumed that the size effect is characterized by coefficient [7]:

Ζd

Κd = (1)

Ζ

where:

- Zd – fatigue strength of specimen of any cross-section,

- Z – fatigue strength of specimen of the same material, cross-section area 20 ÷ 80 mm2.

a) it is possible to develop a methodology for a specific group of structural materials to address

fatigue tests performed with the use of a mini specimen. The methodology refers to:

- defining the specimen development method (the geometry),

- defining the range of feasible loads,

- identifying groups of materials showing a varied sensitivity to the size effect;

b) performing comparative tests (between the normative specimen and the mini specimen) will

make it possible to determine the relationships, defined for a given group of materials, between the

values characteristic for fatigue properties for both specimen development methods compared;

c) there exists a group of structural materials for which the above relationships have a simplified

form (a low level of sensitivity to the size effect) and the performance of tests with mini specimens

made from those materials is especially justifiable and cost-effective.

This paper focuses on the verification of the research methodology by determining the fatigue

characteristics in the high-cycle range. Mini specimens made from EN AW-6063 aluminum alloy

were used. Additionally, the effect of cross-section size on fatigue strength was defined (between

the mini specimen and the normative specimen).

Specimens

The mini specimen geometry assumed as part of the methodology showed the dimensions

defined below as recommended to be applied by norms [1]. The mini specimens were machined in

packages in the process of milling, and the post-treatment surface was neither ground nor polished.

Dariusz Skibicki 65

The specimens were made from EN AW-6063 aluminum alloy. The material to be tested was

provided in the form of flat bars of the specimen thickness. The mini specimen geometry assumed

for testing was compared with the specimen for static tests and the normative specimen for fatigue

test in Fig. 2. Both specimens for fatigue testing show the same value of theoretical stress

concentration factor αk.

a)

130

R25

R25

14

7 ± 0,05

4

38

b) c)

100 50

R50

R25

14

7

7 ± 0,05

3,5 ± 0,05

4 1

Fig. 2. Geometry of: a) monotonic test specimen [8]; b) fatigue test specimen [1]; c) fatigue test

mini specimen

To identify the mechanical properties of material, a static tensile test compliant with the PN-EN

ISO 6892-1:2010 norm was performed. The test was made for 5 specimens of the geometry

compliant with Fig. 2a. The tension diagram is provided in Fig. 3 (the solid line), while the

mechanical properties determined with them are given in Table 1.

As for the metals showing a lack of stability of the cyclic properties, it is assumed that the yield

strength (Re) is not a credible criterion of the upper range of high-cycle fatigue strength. As for the

aluminum alloys, changes in the mechanical properties are connected with the cyclic hardening of

the material [9].

Table 1. Mechanical properties of Table 2. Cyclic load levels of specimens

EN AW-6063 aluminum alloy

σa, MPa σm, MPa σmax, MPa σmin, MPa

Rm, MPa Re, MPa Ru, MPa E, MPa A, % Z, %

P1 75,5 94,5 170 18,9

200 167 350 61458 16,7 61,8

P2 84,5 105,5 190 20,9

determined. Specimens preliminarily loaded (the geometry compliant with Fig. 2a) with fatigue

cycle tested for static tension were investigated. The experiments were made for two load levels

(P1, P2 according to Table 2). A sinusoidal cycle, unilateral cyclic tension with cycle asymmetry

coefficient R = 0.1 was applied. Stress constituted the control parameter. The specimens were

66 Fatigue Failure and Fracture Mechanics

number of cycles Nf = 1000 at 250

the frequency of f = 5 Hz. Then a

static tensile test was made. Both 200

the monotonic and the fatigue

Stress, σ [MPa]

tests were performed at room 150

temperature.

An effect of cyclic strain on

the mechanical properties of 100

material was observed. Fig. 3 for σmax 190 MPa

presents stress–strain curves for 50 for σmax 170 MPa

the specimen without the cyclic As-receive

prestrain and for the specimen 0

cyclic prestrain at the maximum 0 2 4 6 8 10 12 14

stress levels of 170 MPa and 190 Strain, ɛ [%]

MPa of fatigue cycle.

Fig. 3. Stress–strain curves for as-received and

prestrained aluminum alloy at two levels

0,5 200

σmax 190 MPa

Strain increment, ∆ɛ [%]

0,4

σmax 170 MPa 160

0,3 3 cycle 10 cycle 30 cycle

Stress, σ [MPa]

120

0,2

80

0,1

0 40

-0,1 0

1.E+00 1.E+01 1.E+02 1.E+03 1.E+04 0 1 2 3

Number of cycles, Nf Strain, ɛ [%]

Fig. 4. Strain increment with the number of Fig. 5. Fragments of the hysteresis loop for

cycle at different stress amplitudes maximum stress of 190 MPa

170 197 174 342 63022 16,5 62,1 0,96

190 198 192 352 57294 17,2 64 4,1

The material exhibits a change in mechanical properties as a result of cyclic loads. At the initial

state, it showed the yield strength at the level of 167 MPa (plastic strain of 0.2 %). As a result of the

initial cyclic strain, an increase in the yield strength by about 15 % was observed (Table 3). There

occurred a short period of stability (30 – 40 cycles) of cyclic properties. At the first stage of cyclic

load, the greatest strain increment occurred which decreased until saturation was achieved (Fig. 4).

Fig. 5 presents changes in the shape of the hysteresis loop for single cycles recorded in various

fatigue life periods. The loops were plotted at a single level of the maximum stress (190 MPa).

Dariusz Skibicki 67

The studies of cyclic prestrain specimens showed a possibility of rejecting the ‘static’ yield

strength as the upper one admissible for fatigue test (σa, σm), which widened the scope of high-cycle

studies above the yield strength (167 MPa) up to the value of 190 MPa of maximum stresses.

Fatigue test

Complete fatigue characteristics (S-N) were provided with the use of the mini specimen

(Fig. 2c). To define the size effect (at the level of maximum stress of 180 MPa), additionally

3 normative specimens were investigated (Fig. 2b). The experiments were made for the high-cycle

fatigue range (104 ÷ 2x106). Drawing on the cyclic prestrain specimen test results, the range of the

stress levels applied was defined. Due to the specimens buckling, loads with the tensile mean

component of cycle asymmetry coefficient R = 0.1 (sinusoidal cycle) were used. The frequency of

the specimen load change was 5 Hz.

Both the monotonic and the fatigue tests were made using the Instron 8874 servo-hydraulic

material testing machine, and applying the extensometer of the measurement of 25 mm (monotonic

tests).

Based on the fatigue tests with a controlled load, it was possible to plot a fatigue curve (S-N) in

the high-cycle fatigue range. Fig. 6 breaks down the results recorded for the mini specimen

(complete characteristics) and for normative specimens (3 specimens). Only those results where the

specimen was destroyed in the smallest cross-section of the test section have been considered. In yet

another case, a crack was developing in the area of the material defect decreasing the actual

material fatigue life by making the result non-objective.

97,9 220

96,7

93,5 R2 = 0,93 210

Stress amplitude, σa [MPa]

91,7

89 200

86,7

84,5 190

81,7

80 180

76,7

75,5 170

Mini specimen

71,7

66,7 66,7

150

1.E+04 1.E+05 1.E+06 1.E+07

Number of cycles, Nf

For the results a regression line with slope coefficient m = 12.8 and the value of the coefficient

of determination of 0.93 was plotted. As for the specimens with a smaller cross-section, higher

values of fatigue life were reported than the normative specimens. The relationship can be defined

with the coefficient of the cross-section size which, for the results reported, was Kd = 1.17.

68 Fatigue Failure and Fracture Mechanics

Summary

The research suggests the following conclusions:

- it was possible to develop fatigue testing methodology for the high-cycle range for the

aluminum alloy applied; that development concerned the specimen geometry, the method and the

range of the specimen loads;

- the results in the form of an S-N curve for the mini specimen, as compared with the literature

data [10], shows a satisfactory compliance of the value of coefficient m of the regression line; for

the range of the verification, all that points to the credibility of the result received for the evaluation

of the EN AW-6063 aluminum alloy;

- the preliminary studies of the size effect showed the values of coefficient Kd = 1.17 the value of

which, contrary to the literature data, e.g. [11], must be considered essential for strength analysis.

References

[1] PN-74/H-04327. Badanie metali na zmęczenie. Próba osiowego rozciągania – ściskania przy

stałym cyklu obciążeń zewnętrznych.

[2] J. Sempruch, T. Tomaszewski, Application of mini specimens to high-cycle fatigue tests,

Journal of Polish Cimac (2011) 279-287.

[3] T. Hirose, H. Sakasegawa, A. Kohyama, Y. Katoh, H. Tanigawa, Effect of specimen size on

fatigue properties of reduced activation ferritic/martensitic steels, Journal of Nuclear Materials

(2000) 283-287.

[4] Y. Furuya, Notable size effects on very high cycle fatigue properties of high-strength steel,

Materials Science and Engineering A (2011) 5234-5240.

[5] M.D. Callaghan, S.R. Humphries, M. Law, M. Ho, K. Yan, W.Y. Yeung, Specimen-size

dependency and modelling of energy evolution during high-temperature low-cycle fatigue of

pressure vessel steel, Scripta Materialia (2011) 308-311.

[6] D. Boroński, Lokalne własności materiałowe w analizie zmęczeniowej, 2009.

[7] S. Kocańda, J. Szala, Podstawy obliczeń zmęczeniowych, 1997.

[8] PN-EN ISO 6892-1:2010. Metale - Próba rozciągania - Część 1: Metoda badania w

temperaturze pokojowej.

[9] J. Szala, Hipotezy sumowania uszkodzeń zmęczeniowych, 1998.

[10] A.A. Luo, R.C. Kubic, J.M. Tartaglia, Microstructure and fatigue properties of hydroformed

aluminum alloys 6063 and 5754, Metallurgical and Materials Transactions A (2003) 2549-2557.

[11] A. Neimitz, I. Dzioba, M. Graba, J. Okrajni, Ocena wytrzymałości, trwałości i bezpieczeństwa

pracy elementów konstrukcyjnych zawierających defekty, 2008.

© (2012) Trans Tech Publications, Switzerland

doi:10.4028/www.scientific.net/MSF.726.69

values and amplitudes of loading cycles

1

University of Technology and Life Sciences in Bydgoszcz, Faculty of Mechanical Engineering,

Department of Machine Design, ul. Prof. S. Kaliskiego 7, 85-225 Bydgoszcz

a

email: grzegorz.szala@utp.edu.pl, b email: bogdan.ligaj@utp.edu.pl,

(TFC) on the base of experimental test results and description of mathematical models of these

characteristics. Mathematical models were verified in C45 and 41Cr4 tests that are essentially

different from the point of view of mechanical properties. On the base of analysis of verification

test results there were determined ranges of applications of TFC models described in the work.

Nomenclature

C(0) – constant in the formula describing Wöhler fatigue diagram for off-zero pulsating load

(R = 0),

C(-1) – constant in the formula describing Wöhler fatigue diagram for oscillating load (R = -1),

N – cycle number – general notation (fatigue life),

N0 – base number of cycles corresponding to fatigue life (N0 = 106),

R – cycle asymmetry ratio (R = Smin/Smax),

Re – material yield point [MPa],

Rm – material tensile strength [MPa],

Sf – fatigue limit - general notation [MPa],

Sf (0) – fatigue limit under pulsating load (R = 0) for N0 cycle number [MPa],

Sf (-1) – fatigue limit under oscillating load (R = -1) for N0 cycle number [MPa],

Sf( T( 0) ) – fatigue life for sinusoidal off-zero pulsating loading (R=0) for the number of cycles N

(T = logN) [MPa],

Sf( T( −) 1) – fatigue life for sinusoidal oscillating loading (R=-1) for the number of cycles N (T = logN)

[MPa],

m(0) – exponent in formula describing Wöhler fatigue diagram for pulsating load (R = 0),

m(-1) – exponent in formula describing Wöhler fatigue diagram for oscillating load (R = -1),

ψ – factor of material sensitivity to cycle asymmetry, for N = N0,

ψN – factor of material sensitivity to cycle asymmetry, for N ≠ N0,

Abbreviations:

TFC – Two-parametric fatigue characteristics,

LCF – Low-cycle fatigue,

HCF – Hight-cycle fatigue.

1. Introduction

In general service loadings of structural elements are random loadings with wide spectrum [1, 2, 3].

In fatigue life calculation methods and in programmed fatigue life tests random loadings are

replaced as a result of application of accumulation methods of cycles [4, 5] with a set of sinusoidal

cycles that is called loading spectrum [6, 7]. In case of random loadings with wide spectrum it is the

set of cycles with parameters changeable in wide limits: amplitude and mean value of loading

cycles.

70 Fatigue Failure and Fracture Mechanics

Variability of amplitudes and mean values in loading requires to apply, in fatigue life calculations,

an appropriate fatigue characteristic for description of material cyclic properties. Two-parametric

fatigue characteristics (TFC) N(Sa, Sm) [8] or N(Smin, Smax) [9] correspond to the mentioned

conditions. Small set of experimental two-parametric fatigue life characteristic N(Smin, Smax) of

steels, aluminum alloys and titanium can be found in the work [9]. Mathematical model of

characteristics N(Smin, Smax) was published by Bolotin [10] whereas models in the approach N(Sa,

Sm) were given by Heywood in his book [11].

New mathematical concepts of TFC models were published in works [12, 13, 14]. The lack of wide

experimental verification of these concepts causes that, for special cases connected with the type of

materials and loading conditions, there were obtained diversified conformity of test results with

calculated ones.

In the presented work there was attempted to describe mathematically TFC that fulfilling the

condition of description universality and acceptable conformity of test results with calculated ones.

From the analysis of data included in reference materials among others in works [8] and [9] and

publications of own tests [13], [14] and [15] it results that depending on the type of material,

especially its cyclic properties, curves of experimental characteristics TFC have different forms.

Schematic illustration of typical forms of TFC was shown in Fig. 1.

a) b)

c)

fatigue lives (A type), b – reentrant lines of constant fatigue lives (B type), c – mixed case of

constant fatigue lives ((C type) reentrant for high values N = const., salient fot small values

N = const.).

Lines of constant values of fatigue life N = const. are set in a coordinate system Sa, Sm in the area

limited by lines: Sm = Rm, Sa = Rm i -Sm = -Rm, fulfilling the condition Sa + Sm ≤ Rm. The mentioned

area was divided into 4 fields (numbers in brackets) in dependence on the range of the cycle

asymmetry coefficient R = Smin/Smax. The first field (no.1) covers cycles with the coefficient

variability 0 ≤ R < 1, second field (no.2) – -1 ≤ R < 0, third field (3) – -∞ < R < -1, while the fourth

field (no.4) covers cycles 1 < R < +∞. Mentioned ranges of the R coefficient variability close all

Dariusz Skibicki 71

possible cases of sinusoidal cycle parameters Sm and Sa separated from random loading. From the

analysis of reference data [9] it results that most of experimental TFC is determined in fields (no.1)

and (no.2).

In the monograph [19] there was presented description of TFC determination, TFC mathematical

models and experimental verification of the analyzed models. Structural steels C45, S355J0 and

41Cr4 that are different from the point of view of static and cyclic mechanical properties were

analyzed as well as models formulated by Heywood (H) [11] and models marked with Roman

numerals I – simplification of Haigh curve, II – generalized on the range of limited fatigue life

(Goodman formula), III – model based on the parabola equation, IV – model described with the

equation of an eclipse and V – model described with the extended parabolic equation [14], [15] and

[17].

From the analysis of data included in the paper [17] it results that results of calculations in

accordance with the model I the most correspond to the experimental ones. In the model good

conformity of test results was on the range of variability of the cycle asymmetry coefficient R from

the fields (1) and (2), whereas in fields (3) and (4) of R coefficient variability the conformity, for

increased mechanical properties (C45) and low-alloy (S355J0) steels, was significantly lower. The

mentioned disadvantage was not observed in case of 41Cr4 alloy steel.

Analyzing reference data on steel there can be observed similar phenomenon in a slightly smaller

scale because classification of TCF type for alloy steels covers variability of the R coefficient only

in fields (1) and (2) excluding fields (3) and (4) where one can observe the clearly diversified form

of TFC diagrams.

Therefore a conception of TFC modification in accordance with the model I was created towards its

universality. The conception was illustrated in Fig. 2.

Fig. 2 Diagram of the model I (line ABCD) and model IM (line ABCEF) of TFC

Line of constant value (N = const.) in accordance with the model I with reference to description

included in works [15] and [17] is extended between points A B C D (Fig. 2) and described with

formulas:

m

N 0 ⋅ Sf ((−−11))

N= for -∞ < R ≤ 0 (1)

(S (T)

fa ( R ) + ψ NSfm

(T)

(R ) )

m( −1)

72 Fatigue Failure and Fracture Mechanics

and

m( −1)

Sf ( −1) (R m + Sfa( T()R ) − Sfm

(T)

(R ) )

N = N0 for 0 < R ≤ 1,0 (2)

Sfa ( R ) R m (1 + ψ N )

(T )

Mathematical model I described with formulas (1) and (2) refers to TFC of the C type for resilient

materials. Model IM is more suitable for plastic-resilient materials. For the model lines of constant

value are extended between points A B C E F. These lines are described with formulas: (1) and (2)

for the range -1,0 < R ≤ 0 and 0 < R < 1,0 (fields (1) and (2) in the fig. 2). Whereas for the range

-∞ < R < -1,0 (segment C E – field (3)) and 1,0 < R < +∞ (segment E F – field (4)) formulas are as

following:

m

N 0 ⋅ Sf ((−−11))

N= for -∞ < R < -1 (3)

(S (T)

fa ( R ) − ψ NSfm

(T)

(R ) )

m ( −1 )

and

m ( −1)

Sf ( −1) (R m + Sfa( T()R ) + Sfm

(T )

(R ) )

N = N0 for 1 < R < +∞ (4)

Sfa( T()R ) R m (1 + ψ N )

From the above results that for fields (1) and (2) mathematical models I and IM have the same form

essentially differing in fields (3) and (4).

Parameters appearing in formulas from (1) to (4) were explained in the diagram (Fig. 3) whereas the

cycle asymmetry coefficient for metals that appear in formulas from (1) to (4) can be determined in

accordance with the method described in works [17] and [19]. Line A C E is a curve of limit

stresses whereas the line B D E is an exemplary line of constant fatigue lives (N < N0).

models

ψ N = tg Θ N = (5)

Sfa( T()0)

Dariusz Skibicki 73

Experimental verification of analyzed models I and IM described with formulas in the section 2 will

be based on results of C45 and 41Cr4 steel described in details in the work [17].

Wöhler fatigue life curves determined for different values of the cycle asymmetry coefficient R are

the base of experimental TFC. These curves are described with following formulas:

1

log Saf ( R ) = − log N + b (6)

m(R )

List of parameters from Wöhler fatigue life curves of tested steels for variable values of the cycle

asymmetry coefficient R were inserted in Table 1.

Table 1. Wöhler fatigue life curves and the ones experimentally determined fatigue for variable

values of the cycle asymmetry coefficient R

Equation describing Wöhler fatigue life curves

Cycle 1 Fatigue limit Sfa(R)

logSaf ((RT )) = − log N + b ( R )

asymmetry m(R ) for N0 = 106

coefficient R m(R) b(R)

C45 41Cr4 C45 41Cr4 C45 41Cr4

0 17,64 9,95 2,6517 3,1671 204,7 366,7

-0,5 12,02 8,97 2,8188 3,3145 208,7 442,4

-1,0 9,80 8,53 2,9611 3,3546 223,5 447,6

-2,0 12,38 9,17 2,8614 3,3612 238,0 509,8

-3,0 19,80 15,87 2,7440 3,1900 276,0 648,5

Conformity of test and calculation results was evaluated by comparison of fatigue life amplitude

value Sfa( T()R ) for individual lines of constant fatigue lives N and variable values of the cycle

asymmetry coefficient R. Experimental data and test results in accordance with modified

mathematical models were inserted in Table 2 and 3.

As a measure of conformity there was obtained the relative value of test and calculation results

calculated from the formula:

Sfa( T()R ) ex − S(faT()R ) obl

δ= (7)

Sfa( T()R ) ex

Calculation results of relative values were presented in the form of diagrams in individual figures:

for C45 steel in Fig. 4a and for 41Cr4 steel in Fig. 4b. In these figures individual diagram lines

stand for dependence of relative differences δ from N cycle number that characterizes lines of

constant fatigue lives TFC. Lines 1, 2, 3 and 4 correspond to the mathematical model I before

modification, in sequence for following values of the cycle asymmetry coefficient R = 0; -0,5; -2

and -3. Lines 5, 6, 7 and 8 correspond to the mathematical model IM (modified) for values of the

cycle asymmetry coefficient as above. From assumptions of the analyzed model described in the

section 2 and illustrated in the fig. 2 results that diagram lines 1 line up with diagram lines 5 and

diagram lines 2 line up with diagram lines 6.

In the conformity analysis of test and calculation results for both discussed model diagrams for

R = -1,0 were excluded because from assumptions of models results that conformity is total because

formulas (3) and (4) are based on Wöhler curve parameters for R = -1,0.

74 Fatigue Failure and Fracture Mechanics

Table 2. List of data on fatigue life Saf ((RT )) experimentally determined (Ex) and calculated in

accordance with the mathematical models I and IM for C45steel

R=0 R = -0,5 R = -2,0 R = -3,0

N

Ex I IM Ex I IM Ex I IM Ex I IM

2

10 345 345 374 449 469 486 501 732 483 440 852 449

103 303 303 326 371 388 400 416 541 399 391 600 452

4

10 266 266 279 306 321 326 345 404 325 348 432 311

105 233 233 235 253 264 264 287 304 262 310 316 251

106 205 205 195 209 217 213 238 230 211 276 234 206

7

10 180 180 159 172 178 171 198 175 168 246 175 164

Table 3. List of data on fatigue life Saf ((RT )) experimentally determined (Ex) and calculated in

accordance with the mathematical models I and IM for 41Cr4steel

R=0 R = -0,5 R = -2,0 R = -3,0

N

Ex I IM Ex I IM Ex I IM Ex I IM

102 925 925 760 1235 1155 1059 1390 1535 1052 1159 1672 956

103 734 734 618 955 895 832 1082 1148 823 1002 1234 755

104 582 582 499 739 694 651 842 859 767 867 913 602

5

10 426 426 402 571 538 508 645 644 506 750 677 474

6

10 367 367 321 442 417 396 510 483 395 649 503 373

107 291 291 256 342 323 307 396 363 305 561 374 342

a) b)

Fig. 4. Diagrams of relative differences between experimental and test results in accordance with

models I and IM for: a – C45 steel, b – 41Cr4 steel

Test and calculation results described in the section 3 enable the qualitative and quantitative

analysis of mathematical models based on the conception of Haigh stress limit curves (model I and

IM).

Dariusz Skibicki 75

Analysis of the basic model I presented in the work [17] indicated that the form of the line of

constant fatigue lives N of TFC essentially differs from that the form of the line of constant fatigue

lives N experimentally determined. Mentioned differences were observed for the coefficient R for

the range (-1,0; -∞) especially for plastic-resilient C45 steel, less in 41Cr4 alloy steel. The above

statement finds its confirmation on the level of relative differences between calculation and test

results illustrated in Fig. 4. These curves indicate that range of differences for C45 steel and the

model I is from -0,75 to -0,3 for R = -3,0 and from -0,35 to 0 for R = -2,0 (diagram lines 3 and 4 in

Fig. 4a). It should be emphasized that higher of presented values concern smaller number of cycles

N (high stress values).

Implementation of modification of the model described in the section 2, changing the course of

lines of constant fatigue lives N, as in Fig. 2, enabled to obtain much more higher conformities what

is reported by the course of lines 7 and 8 in Fig. 4a. Relative values of δ for these cases

respectively are from 0,06 to 0,12 for R = -2,0 and from 0,08 to 0,3 for R = -3,0 for this model (IM).

It should be emphasized that smaller of presented values correspond to small number of cycles N

(high stress values) what, from the point of view of fatigue life calculations of structural elements,

is a beneficial case.

From the analysis of lines of constant experimental fatigue lives TFC for analyzed steels presented

in the work [17] and the analysis of criteria for classification of lines to ranges of LCF and HCF

results that the type A dominates in the LCF range while the type B of lines of constant fatigue lives

N dominates in the HCF range.

The above statement indicates the general TFC form of the C type. Practically in the algorithm of

fatigue life calculations of structural elements there should be introduced LCF and HCF criterion

and in the first range (LCF) assumed following formulas for calculations (1), (2), (3) and (4) (model

IM) whereas for the second range (HCF) there should be assumed formulas (1) and (2) (model I).

From HCF and LCF criteria analyzed in works [20] and [21] the criterion of number of cycles Nt

from Manson-Coffin curve it is recommended as the most efficient one.

From above considerations results the following general conclusion on the application of discussed

models in fatigue life calculations of steel elements. In case of calculations in the range of HCF it is

recommended to apply the basic model I in accordance with the scheme in the fig. 2 that is

described with formulas (1) and (2).

In case of loadings in the LCF range it is recommended to apply the modified model marked as IM

(Fig. 2) and formulas: form the model IM – (1), (2), (3) and (4).

In the analysis of service loadings of structural elements there are cases of loadings with both HCF

and LCF values. In these cases in the calculation algorithm there has to be implemented the

criterion discussed above and for calculations for cycles from the HCF range the model I has to be

applied while for cycle from the LCF range the modified model IM should be applied.

Note: This work has been elaborated in the frame of the project No. 0715/B/T02/2008/35 financed

by Polish Ministry of Sciences and Higher Education.

References

[1] O. Andersen, W. Popp, M. Schaffrabek, D. Stenmetz, H. Stenger, Schaetzen und Testen,

Springer Verlag, Berlin, 1976.

[2] J.S. Bendat, A.G. Pierdol, Methods of analysis and measurement of random signals,

(in Polish), PWN, Warszawa, 1976.

[3] L. Gajek, M. Kałuszka, Statistical conclusion validity, (in Polish), WNT, Warszawa, 2000.

[4] ASTM standard, Standard Practices for cykle counting in fatigue analysis, ASTM

Designation: E 1049-85 (Reapproved 1990).

[5] S. Kocańda, J. Szala, Fundamentals of fatigue calculations, (in Polish), PWN, Warszawa,

1997.

76 Fatigue Failure and Fracture Mechanics

[6] J. Szala, Programmed fatigue tests of materials and reveted structural components in aviation

industry – selected problems, Monographs, 4nd part of monograph: Improvement of fatigue

life of rivet connections applied in aviation structures – selected problems, Publishing House

of Operation Technology Institute - State Research Institute, Radom, 2010.

[7] P. Heuler, H. Klätschke, Generation and use of standardized load spectra and load – time

historie, International Journal of Fatigue, 27 (2005).

[8] E. Haibach, Betriebsfestigkeit – verfahren und daten zur bauteilberechnung, VDI Verlag,

1989.

[9] Atlas of Fatigue Curves, ASM International, The Materials Information Society, Publishing

House of Howard E. Boyer, 2003.

[10] W.W. Bołotin, Applied statistical methods in mechanics of buildings, Publishing House of

Arkady, 1968.

[11] R.B. Heywood, Designing Against Fatigue, Chapman Hall, Londyn, 1962.

[12] J. Szala, G. Szala, Two-parametric fatigue characteristics formulating problem, Maintenance

problems, no. 3 (2001), 287-296.

[13] J. Szala, G. Szala, Comparative analysis of two-parametric fatigue characteristics and their

experimental verification, Maintenance problems, no. 3 (2001), 297-304.

[14] J. Szala, A. Lipski, Conception of description of material fatigue properties in calculations of

structural elements (in Polish), Problems of Machines Operation and Maintenance, no. 2

(2005).

[15] B. Ligaj, G. Szala, Experimental verification of two-parametric models of fatigue

characteristics by using the tests of S355J0 steel as an example, Polish Meritime Research,

no.1 (2010), 39-50.

[16] B. Ligaj, Experimental and calculational analysis of Steel fatigue life in random conditions of

wide range spectra, (in Polish), Monographs, 2nd part of monograph: Two-parametric fatigue

characteristics of steel and their experimental verification, Publishing House of Operation

Technology Institute - State Research Institute, Radom, 2011.

[17] G. Szala, Theoretical and experimental analysis of two-parametric fatigue life

characteristicsm of constr, (in Polish), Monographs, 1nd part of monograph: Two-parametric

fatigue characteristics of steel and their experimental verification, Publishing House of

Operation Technology Institute - State Research Institute, Radom, 2011.

[18] I.E. Figle, An empirical wquation relating fatigue limit and mean stress, NASA Technical

note TND-3883, Washington, 1967.

[19] G. Szala, Stress sensivity coefficient of a material in range of high – cycle fatigue,

Maintenance problems, no. 4 (2010), 7-16.

[20] G. Szala, B. Ligaj, Evaluation criteria of low- and high-cycle fatigue in fatigue life

calculations of constructional elements, Logistyka, no.6 (2009).

[21] S. Mroziński, J. Szala, Problem of cyclic hardening or softening in metals under programmed

loading, Scientific Problems of Machines Operation and Maintenance, no. 4 (164), vol. 45

(2010), 83-96.

© (2012) Trans Tech Publications, Switzerland

doi:10.4028/www.scientific.net/MSF.726.77

of low-cycle fatigue in conditions of stress and strain control

Bogdan Ligaj1, a, Grzegorz Szala1, b

1

University of Technology and Life Sciences in Bydgoszcz, Faculty of Mechanical Engineering,

Department of Machine Design, ul. Prof. S. Kaliskiego 7, 85-225 Bydgoszcz

a

email: bogdan.ligaj@utp.edu.pl, b email: grzegorz.szala@utp.edu.pl,

Abstract. Service loading of constructional elements are caused by mass forces (dynamic loading)

or displacement (kinematic loading). There are methods of estimation of cyclic properties of

material based on stress or strain range control corresponding to the mentioned loading types. The

paper includes an analysis of differences between diagrams determined in controlled stress and

strain conditions as well as relations among: Wöhler curves (W), Manson-Coffin curves (M-C) and

Ramberg-Osgood curves (R-O). Basing on the performed analysis there have been formulated

conclusions that generally are on connecting loading character with conditions of determination of

atigue curves.

Nomenclature

K’ – strength coefficient of cyclic strain curve [MPa],

2Nf – the number of reversals to failure,

N(s) – fatigue life read from Wöhler curve determined with stress controlled conditions for defined

value of stress Sa,

N(ε) – fatigue life read from Wöhler curve determined with strain controlled conditions for defined

value of stress Sa,

S – specimen stress – general notation [MPa],

Sa – sinusoidal cycle stress amplitude [MPa],

Sa(ε) – stress aplitude read from Ramberg-Osgood cyclic strain curve determined with strain

controlled conditions for defined εac value [MPa],

Sa(σ) – stress aplitude read from Ramberg-Osgood cyclic strain curve determined with stress

controlled conditions for defined εac value [MPa],

b – fatigue strength exponent in Manson-Coffin equation,

c – fatigue plastic deformation exponent in Manson-Coffin equation,

m(-1) – exponent in formula describing Wöhler fatigue diagram for oscillating load (R = -1),

n' – cyclic hardening exponent in Ramberg-Osgood equation,

ε – strain – general notation,

εac – total strain amplitude,

εac(ε) – total strain amplitude amplituda read from Manson-Coffin curve with strain controlled

conditions for the defined number of reversals of loading 2Nf,

εac(σ) – total strain amplitude amplituda read from Manson-Coffin curve with stress controlled

conditions for the defined number of reversals of loading 2Nf,

εf’ – coefficient of plastic fatigue deformation,

δ(Sa) – relative difference of stress amplitude value for Ramberg-Osgood curves determined with

stress and strain controlled conditions,

δ(ε) – relative difference of total strain amplitude value for Manson-Coffin curves determined with

stress and strain controlled conditions,

δ(N) – relative difference of fatigue life for Wöhler curves determined with stress and strain

controlled conditions,

σf’ – fatigue strength coefficient [MPa].

78 Fatigue Failure and Fracture Mechanics

1. Introduction

Design analysis of different types of machines enables to state that constructional elements are

subjected to loads appearing due to control: dynamic (stress) or kinematic (strain) ones. The

dynamic one appears when a force (moment) acting on a defined section area leads to changes of

strain values while stress values remain constant. As examples of design elements subjected to

stress control one can define subassemblies working in engineering designs i.e. aircrafts, seagoing

vessels, working stands. The kinematic control appears when value of stresses in the defined cross-

sectional area is connected with a constant value of strain. As an example of such a case one can

define off-centre machine shafts connected with a rigid coupling. A lack of alignment causes the

constant value of deflection curve to with the corresponding value of stress. While a shaft turns a

rotational bending appears characterized by a constant value of strain during the entire service life

while stress changes [2].

The presented method of control of constructional elements finds its refection in experimental

research procedures. It is assumed that results of fatigue life tests in stress-controlled conditions are

presented as Wöhler curves (W) whereas results determined in strain-controlled conditions are

presented as following diagrams: Manson-Coffin fatigue life curve (M-C) and Ramberg-Osgood

cyclic strain curve (R-O). Detailed description of the mentioned diagrams was included in papers

[1, 6].

Fatigue life diagrams W and M-C and R-O cyclic deformation curve are determined on the base of

research conducted in constant-amplitude loading characterized by asymmetry coefficient R = -1

[5]. Multiple repetition of sinusoidal loading cycle enables to observe changes of cyclic properties

of a material in specified periods of life. In the fig. 1 there are presented exemplary hysteresis loops

recorded during tests of X5CrNi18-10 steel. Superficial analysis of shape and parameters of the

loops for different periods of life indicates differences resulting from control methods and cyclic

change of material properties. The change of the loop parameters influences stabilization curves of

strain amplitude Sa (for the strain-controlled tests) and total strain amplitude εac (for the stress-

controlled tests) pinpointing lack of stabilization of the mentioned parameters in the entire life.

Discussed issues were presented in details in works [3,7].

a) b)

600 Stress S, MPa 400 Stress S, MPa

0,5N 0,9N

0,5N

300 200

0,1N 0,1N

0 0

-1 -0.5 0 0.5 1 -0.6 -0.3 0 0.3 0.6

-300 -200

-600 -400

Fig. 1 Exemplary hysteresis loops for X5CrNi18-10 steel in control conditions: a – strain, b – stress

Lack of stablilization of loop parameters in the entire life indicates the problem of choosing the life

period to determine curves: M-C fatigue and R-O cyclic deformation. It is essential because of

application of the above mentioned curves in fatigue life calculations.

Dariusz Skibicki 79

With reference to presented problems connected with a method of control and change of cyclic

properties the following hypothesis can be formulated: determination of cyclic properties of metals

should be applied in dependence on the character of variable loading (for dynamic loading in a

stress picture, and for kinematic one in a strain picture).

The evaluation of kinematic and dynamic control for test results was performed by comparison of

W, M-C and R-O curves. The mentioned curves were determined on the base of data obtained from

tests realized in constant amplitude sinusoidal loading conditions (R=1) with stess and strain

control. Hysteresis loop parameters recorded for half-life were the base to determine M-C and R-O

curves

Tests in static loading conditions were performed on round cross-section specimens in accordance

with EN ISO 6892-1:2010 standard with a diameter of a measurement base 10 mm. Geometrical

features of specimens that are applied in fatigue tests were assumed on the base of PN-84/H-04334

standard. Measurement base of the specimen was 18 mm while its diameter was 10 mm.

Tests were conducted with an application of austenitic chrome-nickel steel classified as stainless

1.4301 (designation of X5CrNi18-10) with accordance to PN-EN 10088-1:2007 standard. Material

for specimens was bought as a round drawn bar with 24 mm diameter. Delivered bars were in the

annealed state.

X5CrNi18-10 steel was tested ub static and variable constant amplitude loading conditions. Tests in

variable conditions were conducted in strain and stress controlled conditions.

Tests in static loading conditions was conducted in accordance with PN-EN ISO 6892-1:2010

standard. Average values of chosen parameters of a tensile test for X5CrNi18-10 steel are as

following: yield strength R0,2 = 322 MPa, tensile strength Rm = 652 MPa, tensile modulus (Young's

static modulus) E = 166632 MPa, elongation A5 = 60.6 %, contraction Z = 59.8 %, ratio

Rm/R0.2 ≈ 2.0.

Tests of cyclic properties, strain and stress-controlled, were conducted in sinusoidal conditions

characterized by asymmetry coefficient R=-1.

Obtained fatigue test results enabled to determine Ramberg-Osgood (Sa-εac) cyclic deformation

curve described with the formula:

1

S S n'

ε ac = a + a (1)

E K'

Manson-Coffin fatigue life curve (εac-2Nf)

∆ε ac σ'

= ε 'f (2 N f ) + f (2 N f )

c b

(2)

2 E

and Wöhler fatigue life curve (Sa-N)

1

log Sa = − log N + b (3)

m ( −1)

Values of parameters appearing in equations (1), (2) and (3) were presented in the table 1.

80 Fatigue Failure and Fracture Mechanics

for X5CrNi18-10 steel

Test conditions

Stress-controlled (Sa = const.) Strain-controlled (εac = const.)

E MPa 166632 166632

Equation parameters

c -0.3106 -0.3597

b -0.1560 -0.0987

M-C

of curves

σf' MPa 1540 929

n' 0.3165 0.2066

R-O

K’ MPa 2188 1257

m(-1) 9.05 11.56

W

b 2.9606 2.9051

Evaluation of obtained test results strain and stress controlled conditions was based on the

comparative analysis of Ramberg-Osgood, Manson-Coffin and Wöhler curves that led to determine

differences of relative values.

Analysis of Ramberg-Osgood cyclic deformation was conducted on the base of relative difference

of amplitudes of nominal stress δ(Sa) determined from the equation:

Sa ( σ ) − S a ( ε )

δ (Sa ) = (4)

Sa ( ε )

Results of amplitudes Sa(σ) and Sa(ε) were read from the Ramberg-Osgood curve for εac specific

values. Strain-controlled test results were a reference point in the conducted analysis.

Comparison of Mason-Coffin fatigue life curves was conducted on the base of the analysis

of difference value of relative total strain amplitude δ(ε) determined from the equation

ε ac ( σ ) − ε ac ( ε )

δ(ε) = (5)

ε ac ( ε )

Values of total strain amplitudes εac(σ) and εac(ε) were read from the Mason-Coffin curve for specific

values of the number of reversals of loading 2Nf. Strain-controlled test results were a reference

point in the conducted analysis.

Differences among test results, shown in a form of Wöhler curves for stress and strain-controlled

conditions, were analyzed on the base of a value of relative difference of fatigue lives N calculated

from the equation

N ( σ) − N (ε )

δ( N) = (6)

N (ε)

Values of fatigue life N(s) and N(ε) were read from Wöhler curves for specific value of stress

amplitude Sa. Strain-controlled test results were a reference point in the conducted analysis.

In the fig. 2 there are presented test results in the form of R-O, M-C and W diagrams and results of

their comparison. Comparison of cyclic strain curves (fig. 2a) determined in stress and strain-

controlled conditions indicates essential differences in the range of diagram courses and value of

stress amplitude Sa for specific value of total strain amplitude εac. Analysis of changes of Sa values

for the range of εac from 0.001 to 0.03 was performed on the base of relative amplitude difference

diagram (fig. 2b). The range of changes of δ(Sa) values is from -0.15 to 0.16 that indicates an

Dariusz Skibicki 81

intersection of the analyzed diagrams R-O. The intersection point refers to the value εac = 0.009. For

values εac < 0.009 there are obtained lower values of Sa, whereas for εac > 0.009 higher values of

Sa with dynamic control.

a) b)

720 Stress Sa, MPa 0.2 Relative difference δ(Sa)

0.15

540

0.1

0.05

εac = const.

360

0

Sa = const. 0 0.01 0.02 0.03

-0.1

0

-0.15

0 0.01 0.02 0.03

Fig. 2 Comparison of Ramber-Osgood cyclic hardening curves in stress and strain controlled

conditions (a) leading to determination of relative difference of nominal stress amplitude Sa (b)

a) b)

0.1 Strain εac 0.06 Relative difference δ(ε)

εac = const. 0

1.E+02 1.E+03 1.E+04 1.E+05 1.E+06 1.E+07

-0.06

0.01

-0.12

Sa = const.

-0.18

0.001

1.E+02 1.E+03 1.E+04 1.E+05 1.E+06 1.E+07

Number of reversals of loading 2Nf

Number of reversals of loading 2Nf -0.24

Fig. 3 Comparison of Mason-Coffin fatigue life curves in stress and strain controlled conditions (a)

leading to determination of relative difference of total strain amplitude εac (b)

Mutual position of Mason-Coffin fatigue life curves for dynamic and static control was presented in

the fig. 3a whereas values of relative difference calculated from the formula (6) in the fig. 3b. For

the entire range of the number of reversals 2Nf higher values of total strain amplitude εac are

82 Fatigue Failure and Fracture Mechanics

connected with the period of fatigue life determined in strain-controlled conditions. The highest

values of relative difference were obtained for 2Nf = 107 that is δ(ε) = -0.2. On the other hand the

lowest value of δ(ε) that is 0.02 applies to the number of reversals from the range 2·104. The

diagram of changes of relative difference value δ(ε) resembles a parabola in its form.

a) b)

1000 Stress Sa, MPa 0.6 Relative difference δ(N)

800

0.4

600 εa = const.

0.2

500

400

0

-0.2

200

-0.4

Sa = const. Stress Sa, MPa

-0.6

100

-0.8

1.E+02 1.E+03 1.E+04 1.E+05 1.E+06 1.E+07

Number of cycles N -1

Fig. 4 Comparison of Wöhler fatigue life curves in stress and strain controlled conditions (a)

leading to determination of relative difference of number of cycles N (b)

Analysis of mutual position of Wöhler curves (fig. 4a) determined in dynamic and kinematic

control indicates essential differences in their courses. Determined values of relative difference δ(N)

(fig. 4b) for the analyzed range of stress amplitude changes Sa are limited from δ(N) = -0.9 to δ(N)

= 0.55. Fatigue life curves intersect in a point for which Sa = 510 MPa. For values Sa < 510 MPa the

diagram determined in strain-controlled conditions is characterized by higher values whereas for

Sa > 510 MPa higher values of N were obtained for the diagram determined with stress-controlled

conditions.

5. Summary

Presented test results in the form of curves: Ramberg-Osgood cyclic strain curve (R-O), Manson-

Coffin fatigue life curve and Wöhler fatigue life curve determined in stress and strain controlled

conditions indicates differences. Value of differences is connected with the type of material [8] and

value of stress amplitudes Sa or total strain amplitude εac. Performed analysis emphasizes the need

of sensible choice of a testing method suitable for stress or strain control conditions.

The comparative analysis of curves: R-O, M-C and W that were determined in stress and strain

controlled conditions indicates that differences are connected with properties of the tested material.

The range of changes of relative difference are as following:

- cyclic strain curve: δ(Sa) = -0.15 ÷ 0.16,

- Manson-Coffin fatigue life curve: δ(ε) = -0.20 ÷ -0.02,

- Wöhler fatigue life curve : δ(N) = -0.90 ÷ 0.55.

Application of the above mentioned characteristics in fatigue life calculations has to be connected

with the analysis of loading control conditions. Appearing differences between Manson-Coffin and

Wöhler curves can lead to essential differences in fatigue life calculations resulting from multiple

addition of errors in the range of fatigue estimations from 106 to 108 cycles.

Dariusz Skibicki 83

Note: This work has been elaborated in the frame of the project No. 2221/B/T02/2010/39 financed

by Polish Ministry of Sciences and Higher Education.

References

[1] S. Kocańda, J. Szala, Fundamentals of fatigue calculations, (in Polish), PWN, Warszawa,

1997.

[2] B. Ligaj: Selected problems of service load analysis of machine components, Journal of

Polish Cimac, vol.6, no.1 (2011), 125-131.

[3] S. Mroziński, Stabilization of cyclic properties in metals and its influence od fatigue life, (in

Polish), Monographs no. 128, Publishing House University of Technology and Life Sciences,

Bydgoszcz, 2008.

[4] J. Nemec, Strength and stifness of steel elements, (in Polish), WNT, Warszawa, 1968.

[5] G. Szala, Theoretical and experimental analysis of two-parametric fatigue life

characteristicsm of constr, (in Polish), Monographs, 1nd part of monograph: Two-parametric

fatigue characteristics of steel and their experimental verification, Publishing House of

Operation Technology Institute - State Research Institute, Radom, 2011.

[6] J. Szala, Hypotheses of fatigue damage accumulation, (in Polish), Monographs, University of

Technology and Agriculture, Bydgoszcz, 1998.

[7] J. Szala, S. Mroziński, The problem of cyclic hardening or softening of metals in programmed

loading conditions, (in Polish), Acta Mechanica et Automatica, vol.5, no.3 (2011), 99-107.

[8] B. Ligaj, G. Szala, The comparison of cyclic properties of metals in the range of low-cycle

fatigue in conditions of stress and strain control, (in Polish), Conference proceedings XXIV

Sympozjum Zmęczenie i Mechanika Pękania, Bydgoszcz (2012), 89-90.

© (2012) Trans Tech Publications, Switzerland

doi:10.4028/www.scientific.net/MSF.726.84

for trabecular bone under stepwise load

Tomasz Topoliński1, a, Artur Cichański1,b,

Adam Mazurkiewicz1,c, Krzysztof Nowicki1,d

1

Faculty of Mechanical Engineering, University of Technology and Life Sciences,

Kaliskiego 7 Street, 85-789 Bydgoszcz, Poland

a

tomasz.topolinski@utp.edu.pl, bartur.cichanski@utp.edu.pl,

c

adam.mazurkiewicz@utp.edu.pl, dkrzysztof.nowicki@utp.edu.pl

Abstract. In this work was presented method of initial stiffness modulus E0 calculation based on

fatigue tests of trabecular bone under stepwise load. The investigation was performed on 61

cylindrical bone samples obtained from the neck of different femur heads. The bone sample fatigue

tests were carried out under compression with stepwise increases of the applied load. The obtained

values of the initial stiffness modulus E0 were consistent with literature data and can be used to

determine the S-N curve for trabecular bone using the hypotheses of fatigue damage accumulation.

It was also an unsuccessful attempt to find a statistical relationship between the values of the initial

stiffness modulus E0 and indices of bone structure.

Introduction

Damage can be defined in various ways. According to the study [1], there are four typical

definitions for the characterization of damage:

̶ defects at the microscale described by the number/density of the cracks;

̶ changes in the physical properties, such as material density, acoustic emission recordings,

electrical resistivity, ultrasonic waves, and micro-hardness measurements;

̶ the remaining lifetime of the material;

̶ variations in the macromechanical behavior, such as changes in the elastic, plastic or

viscoplastic properties.

When investigating bones, damage is most often defined as D = E/E0, where E is the current

value of the stiffness modulus, E0 is the initial value [2, 3], and by convention, D = 1 – E/E0 [4].

The description D = AD/A is also used, where AD is the bone damage area, A is the initial area [2],

and D = DV/BV ,where DV is the damaged volume, BV is the undamaged bone volume [4].

The values of the initial stiffness modulus were determined using different methods, as described

below. In paper [4], the tangent modulus (initial tangent stiffness) was the maximum slope of an 11-

piece linear regression of the stress–strain curve between the applied apparent strains of 0.1% and

0.8%. In paper [6], to determine E0 the slope of the stress–strain curve of the tenth cycle for

compression at 1500 microstrain was used. As reported in [7], E0 is defined in a manner analogous

to paper [6]; however, the strain ranges from 0.6% to 2.1%. The tangent modulus in [8] was

determined prior to fatigue testing using a 100 N (10 MPa) load applied at 2 Hz tensile and

compressive ramps. According to work [9], the initial modulus was measured by taking the slope of

the best linear fit for the final preconditioning loading cycle (ε=0.001 to 0.003). The initial modulus

reported in [10] was measured by taking the slope of the best linear fit of the tenth loading cycle

from 0.1% to 0.3% strain. In the work [5] tested each sample of the stress–strain curve at the origin.

Samples were loaded cyclically at 2 Hz to 100 N for approximately 20 cycles, which was equivalent

to a stress of approximately 12 MPa for the samples. For this test, the slope of the stress-strain curve

was linear; therefore, the tangent modulus was calculated by dividing the pretest stress by the

strains. This strategy was associated with calculating the standardized stress with the quantity E0

used in the creep and fatigue studies of minimized bone scatter. It appeared that most of the effect

on the structure and complexity of the process fell within the quantity E0.

Dariusz Skibicki 85

Evaluating the risk of bone cracking for real-life applications is a very complex procedure. The

evaluation critically depends on ability to define the fatigue life of bone tissue; thus, there is a

significant part of published work in that field [3, 6, 7, 9, 10] The majority of the research was

performed in vitro and using only sinusoidally-variable constant-amplitude loads, most often with

compression loads [6, 7, 9, 10], particularly for trabecular bone, whereas tension [5] was the more

rarely used type of load [1]. The quantity used to describe the results was generally the input stress

associated with the initial stiffness, E0. Similarly, our experiment was developed for cylindrical

samples of trabecular bone exposed to sinusoidally-variable loads; however a stepwise loads were

applied, which has not been applied in bone experiments, but is commonly used when investigating

other materials and structures [11, 12, 13, 14, 15]. This kind of loading makes it possible for the

fatigue damage accumulation to involve loads that start from the lowest value to those for which the

damage will be visible in a few dozen to a few hundred cycles. Such a load model should allow

more information to be obtained about the fatigue process than in the one-step load experimental

procedure.

Methods

The paper uses the research results of 61 cylindrical bone samples that were 10mm in diameter

and 8.5mm in length and were obtained from the neck of femur heads. The samples were obtained

from 21 men and 40 women undergoing hip joint alloplasty. The samples were stored in a 10%

formalin solution at room temperature.

All of the samples were scanned with a desktop microCT system (µCT-80, SCANCO Medical

AG, Bruettiselllen, Switzerland) with a distance of 36µm between. When scanning the values of

many bone structure indices were obtained: trabecular number Tb.N, trabecular thickness Tb.Th,

trabecular separation Tb.Sp, bone volume fraction BV/TV, surface fraction BS/BV and the number

of joints between individual trabeculae per unit volume of specimen Conn.D.

The bone sample fatigue tests were carried out under compression with stepwise increases in the

load using the testing machine, INSTRON 8874 (Instron, High Wycombe, England). The minimum

loading for all of the loading levels was 5N. The maximum loading started at 10N with a gain every

10N at each successive step. At each loading level, 500 cycles were completed under constant-

amplitude loadings at a frequency of 1Hz.

As stated in the literature, we have used the description of the results for the standardized load,

where the standardizing quantity is the quantity, E0. Depending on the requirements and the

assumptions made in literature, the standardizing quantity value is either the initial tangent stiffness

or the initial secant stiffness. In our work, the calculations of the tangent modulus, E0, involved

defining the angle of inclination of the line best adjusted to the pattern of the upper branch of the

loop in the range from 85% to 100% of the maximum stress (referenced to the absolute value). The

tangent modulus was determined for all of the recorded fatigue loops. This paper makes use of the

mean value of the tangent modulus defined by the tangent modulus of the hysteresis loop recorded

at the first load step. Full information about sample preparation and recorded during test values

contains paper [17].

86 Fatigue Failure and Fracture Mechanics

stress range

tg(α) - stiffness

modulus E

15% of stress

range

Fig. 1. The hysteresis loop with marked: stress range, 15% of the stress, line of best fit to the upper

branch of the hysteresis loop, stiffness modulus E

Results

Fig. 2a shows the stiffnes modulus values for the first block of loading. The continuous lines

mark the initial modulus. Fig. 2b shows the stiffnes modulus values for all recorded hysteresis loop.

a) b)

Fig. 2. Stiffness modulus calculated for a) the hysteresis loop recorded in the first block of the load,

with a line marked the initial modulus E0 b) all recorded hysteresis loop

The results of investigating the structure indices are given in Table 1, which presents the mean

values of the selected indices, the standard deviation SD and the relative standard deviation

(RSD=SD/mean) for all samples. The considerable variation in the structure of the bone samples

investigated should be noted. The relative standard deviation ranges from 19.6% to as much as

48.9%, which must affect the scatter of the parameters recorded. The table was supplemented by the

results of calculations of E0, according to the method proposed in our work and layout as for

indicators of the structure.

Dariusz Skibicki 87

Table 1. The values of the mean, standard deviation (SD) and relative standard deviation (RSD) for

selected indices of the structure and E0 of the bone samples investigated.

BV/TV Conn.D. N

indices [1/mm] [1/mm] [mm] [mm] [MPa]

Mean 0.204 11.998 1.133 0.176 0.786 2.764 388,95 20630

SD 0.076 2.747 0.222 0.045 0.385 0.921 211,05 12078

RSD [%] 37.1 22.9 19.6 25.6 48.9 33.3 54,3 59

Discussion

Our results are consistent with the calculation of the initial stiffness modulus E0 for human

vertebral trabecular bone that has been shown in [3.16] for samples cut at an angle of 0°. Table 2

contains the summary of our results and literature data. A higher value for the standard deviation of

our results is the effect of much larger number of samples from a larger number of donors. This

indicates the correctness of the proposed method for determining the value of E0. The initial

stiffness of the cancellous bovine bone samples is much higher, regardless of the direction of the cut

samples [3, 10, 16]. Stiffness of cortical bone is also much higher [1].

The obtained values of the initial stiffness modulus E0 were consistent with literature data and

can be used to determine the S-N curve for trabecular bone using the hypotheses of fatigue damage

accumulation [18].

Table 2 Values of E0 values obtained by method presented in the work and data from the literature

E0 [MPa]

No. References Bone

mean±SD min÷max

1 [1] human cortical - 12370÷15470

2 [6] bovine trabecular 2690±900 1190÷4150

3 [3] human vertebral trabecular 251±137 44÷497

4 [10] bovine trabecular - 1466÷2732

5 0° 2709±548 -

bovine trabecular

6 90° 1227±725 -

7 0° 447±117 -

8 [16] 22° 159±11 -

human vertebral trabecular

9 45° 111±68 -

10 90° 98±78 -

11 human femoral trabecular 0° 1031±461 -

12 our resultes human head femoral trabecular 389±211 69÷867

In addition, an attempt was made to determine the relationship between E0 and calculated indices

of the bone structure. Fig. 3 shows scatterplots of the relationship between the values of the sample

structure indicators (BV/TV, BMD) and the initial values of stiffnes modulus E0. Regression lines

in Fig. 3 are omitted because the coefficient of determination values R2 for all structure indicators in

Table 1 according to the E0 did not exceed 0.14. Therefore, no relationships were found between the

E0 and bone structure. In our opinion, this relationship should take into account characteristics of

strength of individual trabeculae.

88 Fatigue Failure and Fracture Mechanics

a) b)

Fig. 3. Scaterrplots of the relationship between the values of the sample structure indicators BV/TV

a), BMD b) and the initial values of stiffness modulus E0

References

[1] Zioupos P., Casinos A.: Cumulative damage and the response of human bone in two-step

loading fatigue, Journal of Biomechanics 1998, 31(9): 825-833

[2] Kosmopoulos V., Keller T.S: Predicting trabecular bone microdamage initiation and

accumulation using a non-linear perfect damage model. Medical Engineering and Physics

2008, 30: 725-732

[3] Rapillard L., Charlebois M., Zysset P.K.: Compressive fatigue behavior of human vertebral

trabecular bone, Journal of Biomechanics 2006, 39(11): 2133-2139

[4] Tang S.Y., Vashishth D.: A non-invasive in vitro technique for the three-dimensional

quantification of microdamage in trabecular bone. Bone 2007, 40: 1259-1264

[5] Cotton J.R., Winwood K., Zioupos P., Taylor M.: Damage rate is a predictor of fatigue life and

creep strain rate in tensile fatigue of human cortical bone samples. Journal of Biomechanical

Engineering 2005, 127: 213-219

[6] Haddock S.M., Yeh O.C., Mummaneni P.V., Rosenberg W.S., Keaveny T.M.: Similarity in the

fatigue behavior of trabecular bone across site and species. Journal of Biomechanics 2004,

37(2): 181-187

[7] Michel M.C., Guo X.E., Gibson L.J., McMahon T.A., Hayes W.C.: Compressive fatigue

behavior of bovine trabecular bone. J Biomechanics 1993, 26(4/5): 453-463

[8] Pattin C. A., Caler W. E., Carter D. R.: Cyclic mechanical property degradation during fatigue

loading of cortical bone, Journal of Biomechanics 1996, 29(1): 69-79

[9] Ganguly P., Moore T.L. A., Gibson L.J.: Analysis of fatigue damage in bovine trabecular bone.

www.asbweb.org

[10] Moore T.L., Gibson L.J.: Fatigue of Bovine Trabecular Bone. Journal of Biomechanical

Engineering 2003, 125: 761-768

[11] Landgraf R.W., Morrow J., Endo T.: Determination of the cyclic stress-strain curve. J of

Materials 1969, 4: 1621-1653

[12] Janzen W., Ehrenstein G.W.: Bemessungsgrenzen von glasfaserverstärktem PBT bei

schwingender Beanspruchung. Kunststoffe 1991, 81(3): 231-236

[13] Orth F., Hoffmann L., Zilch-Bremer H., Ehrenstein G.W.: Evaluation of composites under

dynamic load. Composite Structure 1993, 24(3): 265-272

Dariusz Skibicki 89

Elastomery 2004, 8(5): 15-19

[15] Casado J.A., Carrascal I., Polanco J.A., Gutiérrez-Solana F.: Fatigue failure of short glass fibre

reinforced PA 6.6 structural pieces for railway track fasteners. Engineering Failure Analysis

2006, 13(2): 182-197

[16] Dendorfer S., Maier H.J., Taylor D., Hammer J.: Anisotropy of the fatigue behaviour of

cancellous bone. J Biomech. 2008; 41(3): 636-41

[17] Topoliński T., Cichański A., Mazurkiewicz A., Nowicki K., Study of the behavior of the

trabecular bone under cyclic compression with stepwise increasing amplitude, Journal of the

Mechanical Behavior of Biomedical Materials 2011, Vol. 4, No. 8: 1755-1763

[18] Topoliński T., Cichański A., Mazurkiewicz A., Nowicki K., Applying a stepwise load for

calculation of the S-N curve for trabecular bone based on the linear hypothesis for fatigue

damage accumulation, Fatigue Failure and Fracture Mechanics 2012, In Press

CHAPTER 3:

Fatigue of Welded Structures

© (2012) Trans Tech Publications, Switzerland

doi:10.4028/www.scientific.net/MSF.726.93

GOSS Czesław1a, MARECKI Paweł1b

1

Military University of Technology, Faculty of Mechanical Engineering, Kaliskiego Street

2, 00-908 Warsaw

a b

cgoss@wat.edu.pl, pmarecki@wat.edu.pl

Abstract. Test results of residual stresses in welded butt joints were also presented. Finite Elments

Methods and Jewdokimow and Lawrance’a methods of stress concentration factor αk calculation

were comparative sumarized. Presents results is a part more considerable research attendant bridge

MS-20.

Introduction

In last year’s, a investigated of the attendant bridge MS-20, constructed in the Department of

Mechanical Engineering Military University of Technology. Load tests were carried out

successfully. The research confirmed a sufficient fatigue life of the bridge (5000 cycles). Loads with

a larger number of cycles showed cracks occurring in welded joints. Therefore, it was further tests

of welded joints of steel from which the bridge was produced.

Test results

In the first place in order to determine the basic material properties the flat specimen static

tensile tests were carried out. The specimens were made in accordance with appropriate [1]. Table 1

presents the chemical composition and Figure 1 shows an example chart of steel tensile test. Table 2

shows the typical mechanical properties obtained during testing.

Table 1. Steel S960QL chemical composition (melting analysis)

C Si Mn P S Cr Mo Ni V

0,18 0,50 1,60 0,02 0,01 0,80 0,60 2,00 0,10

Table 2. Mechanical properties Steel S960QL

R0,2 [MPa] Rm [MPa] Ru [MPa] A [%] Z [%] E [GPa]

1000 1090 2209 14 38 208

1100

σ [MPa]

1000

900

800

700

600

500

400

300

200

100

0

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15

ε [%]

94 Fatigue Failure and Fracture Mechanics

The base material flat specimens and steel S960QL specimens with welded V butt joints made

using TIG method, coalescence X96-IG, in accordance with appropriate [1] were research subject.

The tests were carried out on a machine Instron 8802 with maksimum load 250kN. The specimens

were testing using cycling loads with a constant maximum stress value. A fatigue tests were carried

out in full cycles in order to obtain a complete steel characteristics. This allowed to obtain Wöhler

curve including high-cycle range.

1100

σ a [MPa]

1000

95%

900

95%

800

POŁĄCZENIESPAWA

WELDED JOINT NE

log Nˆ = 8, 5943 − 0,0045σ

BASE MATERIAL

MATERIA ŁRODZIMY

600

500

400

1,00E+02 1,00E+03 1,00E+04 1,00E+05 1,00E+06 N f 1,00E+07

Figure 2. Wöhler curve for base material and welded joint specimens

The test results presents a significant decrease in fatigue life of welded specimens compared to

base material specimens. Fatigue life in low-cycle range decreased up to 90% (figure 2). It follows

that the steel S960QL should not be used for fatigue loaded welded structures. Furthermore a

tendency to cyclical strengthening of steel S960QL are presented in following figures (figure 3 and

figure 4).

1100

σ[MPa]

1000

900

800

700

600

500

400

300

200

100

1 2 3 4 ε[%]

Figure 3. Chart of stresses and strains changes as cycle number function at a constant stress

σmax = 1080 MPa for a base material specimen (Nf = 5770)

Dariusz Skibicki 95

1100

σ[MPa]

1000

900

800

700

600

500

400

300

200

100

1 2 3 4 5 6 7 8 ε[%]

Figure 4. Chart of stresses and strains changes as cycle number function at a constant stress

σmax = 1080 MPa for a welded butt joint specimen (Nf = 5770)

Residual stresses measurements in the welded joints were also performed [3]. X-ray diffraction

measurements were performed using the Rigaku diffractometer STRAINFLEX PSF-2M.

The chart illustrated the following types of stresses: x - along the specimen longitudinal axis,

y – perpendicular to the specimen longitudinal axis and the average stresses of these directions

(figure 5).

The measuring points were placed in the center of the welded joint (point 1), on the melting line

(point 2) and on the border of heat-affected zone (point 3) (Figure 6).

200

100

0

1

-100

Stresses [MPa]

direction x

-200 direction y

average value

-300

-400

-500

-600

Tests points 1,2,3

S960QL)

3 2 1

96 Fatigue Failure and Fracture Mechanics

The Lawrance and Jewdokimow methods were used to a shape factor determination [2].

The Lawrence method

The measurements of a testing welded joint geometry were performed in order to determining of

a shape factor αk. The angle weld slope Θ and transition welding radius ρ in base material were

measured (figure 7). Measurements were made from the weld cap and from the root of weld. The

obtained values are presents in table 3 [4].

Table 3. Angle weld slope Θ and transition welding radius ρ in base material

Parameter Value

Θ 26° - 27°

Weld cap

ρ 4,8° - 7,5°

Θ 32° - 35°

Root of weld

ρ 2,3° - 3°

t

C

α k = B ⋅ 1 + A ⋅ (1)

ρ

where: A, B, C - constants depends on the type of welded joint, the way it loads and the local

geometric shape,

t - jointed elements thickness (t = 6 mm),

ρ - transition welding radius,

Θ - angle weld slope.

The values of a coefficients A, B, C are shown in following table (Table 4).

Coefficient Value

A (weld cap) A = 0,27(tg Θ)0,25 = 0,226 – 0,228

A (root of weld) A = 0,27(tg Θ)0,25 = 0,240 – 0,247

B 1

C 0,5

The shape factor values were obtained using equation 1 with appropriate values of coefficients:

weld cap - αk = 1,202 – 1,255,

root of weld - αk = 1,339 – 1,399.

The maximum value was taken for further calculations.

The Jewdokimow Method

The shape factor value αk in Jewdokimow method was obtained using a following equation [2]:

α k = α kc ⋅ α p (2)

Dariusz Skibicki 97

α kc –tensile and bending welded butt joint shape factor of circle cap weld with a width

equal to the thickness of the joint elements

α p – correction factor taking into account the actual weld width [2].

αk = 1,35 · 1,03 = 1,39 - weld cap,

αk = 1,46 · 0,98 = 1,53 - root of weld.

The following value was taken for further calculations αk = 1,53.

Stress distributions were obtained using by finite element modelling. There was obtained a value

of a stress concentration factor:

αk = 1,08 - weld cap,

αk = 1,20 - root of weld.

The shape factor values obtained during numerical analysis were lower than shape factor values

obtained during calculations. The stresses and strains distributions in steel S890QL welded

specimens during tensile were calculated.

Analysis Object

The analysis object was flat specimen with welded butt joint in the middle (Figure 8a).

a) b)

Computational Model

Due to longitudinal and transverse plane of symmetry the computational model was ¼ section of

the specimen (only work part) (Figure 8b).

C

Figure 9. FEM model with an indication of the elements, boundary conditions and loads

Model consisted of 10 371 spatial elements (3D), including the 3462 tetrahedral 10-nodes

elements and 6909 hexahedral 20-nodes elements (Figure 9). Model mesh consisted of 33232

nodes. The boundary conditions indications in Figure 10 are: „C” – no displacement in y direction

„D” – no displacement in x direction. The elastic material model with nonlinear strengthening

developed on the tensile test results was assumed.

The Calculation Results

The results are shown in colored contour lines represented von Mises stresses, displacements and

charts along selected lines (figure 10, 11, 12).

98 Fatigue Failure and Fracture Mechanics

Figure 13. Reduced stress along the bottom edge of the model (middle of the specimen) for

three load levels: 850, 600 and 500 MPa

Dariusz Skibicki 99

The stress distributions showed in Figures 10-13 were obtained during numerical analysis. The

highest stresses occurred on the melting line. Numerical analysis shows that these stresses approach

the yield strength of this steel, although the nominal stresses are much lower.

The stress concentration factor was determined on that basis:

- root of weld - αk = 1,20,

- cap weld - αk = 1,08.

Therefore, these values are lower than those obtained using computational methods. All

specified values of αk will be used in further work to analyze fatigue life of welded joints.

Summary

Steel S960QL has cyclic strengthening with the tendency for fast stabilization. This applies both

to base material specimens and welded butt joint specimens. During testing using cycling loads with

a constant maximum stress value σmax a decrease of hysteresis loop width and amplitude of plastic

deformation were showed. The test results showed a significant decrease up to 90 % in fatigue life

of welded specimens compared to base material specimens. This tendency is maintained also in

high-cycle range.

The measurements of residual stresses in welded butt joints showed their considerable value.

This fact results in summing of residual stresses and stresses from external forces during tension.

This has a substantial effect on the durability of tested elements.

The results confirmed a great influence of welded joints on mechanical properties of structures

made using this steel. In the case of a bridge we are mainly dealing with similar cyclic loads.

Therefore a authors attention will be focus on a such kind of loads.

This research are continued mainly for the modification of welded bridge nodes to increase the

structure durability.

References

[1] Polish Standart PN-88/M-69710. (in polish)

[2] Goss Cz., Kłysz S., Wojnowski W.; The low cycle fatigue life selected steel AT the welded

joint; ITWL, Warszawa 2004 (in polish).

[3] Lech-Grega M., Raport Nr 3129/N/1/4 – The residual stress measured, Skawina 2011 (in

polish).

[4] Goss Cz., Marecki P., The durability fatigue problems welded joint with steel S890QL,

Biuletyn WAT, Warszawa 2012 (in polish).

© (2012) Trans Tech Publications, Switzerland

doi:10.4028/www.scientific.net/MSF.726.100

stress concentration factor in a laser welded tee joint

1, 2

Gdansk University of Technology, The Faculty of Ocean Engineering and Ship Technology,

Narutowicza 11/12 St., Gdansk, Poland

a

karnikla@pg.gda.pl, bkozak@pg.gda.pl

Keywords: laser welds, fatigue, local stress approach, fem, steel sandwich panel

Abstract. In recent years an increased interest of industry in sandwich-type metal structures can be

observed. These structures consist of thin plates of 2.5 mm in thickness separated by stiffeners of

different shapes and forms. Welds joining the plates and stiffeners are made on the outer side of the

plates using laser welding technique. A locally focused source of heat causes the plate to melt

creating a very narrow and elongated joint. As a result, sharp geometric notches are formed on the

side of the root of a weld – a place which is inaccessible and cannot be checked. Geometries of

individual welded joints vary, sometimes considerably, and this makes their analysis even more

complicated. Additionally, the use of laser welding technique influences the formation of untypical

distribution of changes in material properties in weld zones. The effect is a joint whose behaviour

under load is significantly different from the behaviour of a welded tee joint made with the use of

classical methods. Fatigue strength calculations for this type of joints can be conducted based on

local stress values in notches, which can be determined with the use of Finite Element Method

(FEM). This article analyses the influence of the notch rounding radius on the elastic notch stress

concentration factor Kt The aim of the analysis is to evaluate the notch stress concentration

according to local notch stress approach.

Introduction

The growth of transport operating costs causes an increase of interest in new materials and

structural solutions. There is an on-going search for innovative structural solutions characterised by,

among others, a better strength to mass coefficient and a longer service life. One of the more

interesting from the many possible new solutions are steel sandwich panels. These are multi-layered

structures consisting of cover plates, stiffeners and a core. The panels are prefabricated as large-size

structures made of plates with different thickness, stiffeners of various height and form located

between the panels, and different forms of filler material. The stiffeners are flat bars perpendicular

to plates, most often in a shape of V, C, O Z or X. Standard plate thicknesses are 1–3 mm, stiffener

thicknesses equal 4 mm and the distance between plates is from 20 mm to 250 mm. Overall

dimensions of a panel reach from 1.5 m to 10 m. An example of a panel with flat bar stiffeners and

other most popular forms of stiffeners is shown in Fig. 1.

Dariusz Skibicki 101

Fig. 1 An example of a steel sandwich panel with flat bar stiffeners (I-core) [1]

In all practical applications panel structures must comply with a number of rules (regulations of

classification societies, standards, etc.). One of the requirements that welded structures must fulfil is

an adequate fatigue strength. The analysed elements of a steel sandwich panel structure that are

prone to fatigue failure are welds connecting plates with stiffeners. These joints are laser welded in

a very untypical way – by melting a plate from the outer side and forming a joint with a stiffener.

The joint created this way has a unique geometry and its fatigue strength is hard to estimate. It

should be noted that these joints play a key role in terms of the ability of a multi-layered structure to

transfer loads. A theoretical analysis of fatigue strength for this type of joints may be carried out

with the use of different criteria:

- structural stresses and strains criterion

- local criteria (local stresses and local strains)

- criteria based on fracture mechanics theory

- mixed criteria.

Due to the specifics of the analysed laser welds, and especially their geometry, the most advisable

fatigue approach seems to be the local stresses criterion. The sections below will discuss the

geometries of the analysed laser welds, and than numerical modelling of notch stress concentrations.

Numerical modelling of laser welds is performed in order to use local notch stresses to evaluate

fatigue strength according to local stress criterion.

Welds of steel sandwich panels are prefabricated with a rather untypical technique of laser welding.

Plates are joined with stiffeners from their outer side creating joints with a very unique shape. The

heat source is a 14 kW CO2 laser whose beam is focalized by parabolic mirrors with a spot size of

0.5 mm and directed onto the outer surface of the panel's plate. The focused laser beam of a very

high power density (10e9 W/mm2) makes the plate and stiffener material fuse. A very narrow and

elongated joint is created, but without a full penetration on the stiffener edges. On both sides of the

weld, between the plate and the stiffener there are technological gaps left. It should be noted that the

unpenetrated region of a joint often covers more than a half of the stiffener's width. An example of a

laser joint created with the use of this technology and its geometry is shown in Fig. 2. This figure

includes mean values of microscopic measurements performed on samples collected from the

examined panels. Detailed measurements of SAW and laser welded butt joints on 12 mm thick

sheets are published in [3, 4, 5]. However, it must be noted that in case of the analysed joints

between plates and stiffeners of steel sandwich panels the dispersion of parameter values describing

face and root notches is much greater, in proportion to the thickness of the analysed sheets, than in

case of laser joints of about 10 mm thick sheets.

102 Fatigue Failure and Fracture Mechanics

Influence of the notch rounding radius on the elastic notch stress concentration factor

The aim of numerical modelling of laser welds is to determine the distribution and, as a

consequence, to calculate stress concentrations present in geometrical notches. Determination of

stress concentrations is crucial for evaluating the fatigue strength of welds. Following the

methodology of numerical calculations according to the local stress criterion, notches should be

modelled by fictitious roundings or alternative geometries [6, 7]. However, application of a standard

rounding radius ρf = 1 mm, recommended in a number of publications for joints of elements that are

over 5 mm thick, is not possible in this case since the elements being joined are much thinner and

dimensions of the notches are smaller. The few publications describing the way of calculating

elastic stress concentration factors for welds joining very thin sheets (of about 1.5 mm) concern spot

welds of car bodies [8, 9]. For this type of intermittent welds the proposed rounding radius is

0.05 mm. Considering the technology of spot welding, different geometry and properties of such

welds, generalization of the calculation method with a 0.05 mm radius for other welding

technologies is very risky. Other authors suggest the radius value of 0.1 mm for lap joints of thin

laser welded sheets [10]. As can be seen from a limited number of publications on the subject, small

amount of empirical data and a considerable spread of the proposed values, methods for calculating

stress and strain concentrations of welds joining thin elements are a poorly explored research area.

Therefore, in the conducted calculations values of rounding radii of face and root notches were

adopted as a parameter that will be analysed. The assumptions for the FEM model were: plane stress

state, loads in the plane of panel plates, weld symmetry axis. The model geometry with loads and

boundary conditions is shown in Fig. 3. Linear elastic material model with Young's modulus

E = 2e5 MPa and Poisson ratio ν = 0.3 were assumed. 8-node quadrilateral elements with quadratic

shape function were used.

Dariusz Skibicki 103

Fig. 3 Geometry of laser weld model with loads and boundary conditions

The examination of the influence of the notch alternative geometry radius ρf on the elastic stress

concentration factor Kt was conducted for 6 different values of alternative radii (called ‘fictitious’ in

the literature and denoted with ρf ): 0.025 mm, 0.05 mm, 0.1 mm, 0.15 mm, 0.2 mm and 0.25 mm. It

was assumed that the values of face and root notch rounding are equal in all analysed cases. Model

geometries with face notch rounding and root notch alternative circular geometry assumed for

calculations are shown in Fig. 4.

Fig. 4. Model geometries with different values of radius ρf for alternative geometry:

a) ρf = 0.025 mm, b) ρf = 0.05 mm, c) ρf = 0.1 mm, d) ρf = 0.15 mm, e) ρf = 0.2 mm,

f) ρf = 0.25 mm

104 Fatigue Failure and Fracture Mechanics

The influence of the notch rounding value ρf on the elastic notch stress concentration factor Kt is

shown in Fig. 5. Stress concentration values, calculated as the ratio of maximum notch stress to the

nominal stress, observed in the root notch are significantly greater than in the face notch. For the

value of rounding radius above ρf = 0.15 changes of Kt factor are insignificant. For the face notch

stress concentration values decrease in the whole range of the notch rounding radius ρf and for the

higher values of this radius the changes are more and more gentle. The differences between

concentration curves for face and root notches result from their different alternative geometries. For

the face notch the radius ρf is a rounding, and for the root notch it is the radius of an alternative

circular geometry. Looking at the greater values of stress concentrations in the root of a weld and

the course changes in the value of Kt factor presented in Fig. 5 it can be said that for numerical

modelling of stress and strain concentrations with the use of local approaches a value of alternative

geometry radius equal to ρf = 0.15 mm can be adopted. This is the smallest radius value for which

the concentration level can be uniquely determined. Application of a greater radius value does not

influence the result of concentration, but only causes artificial decrease of cross-section due to

removal of weld material.

Fig. 5. Influence of the notch rounding radius ρf on the elastic stress concentration factor Kt

Summary

One of the key issues connected with the ability of sandwich-type structure to transfer loads is

the fatigue strength of laser welds joining plates and stiffeners. Due to the difference of laser welds

present in the panel construction it is very hard to reliably estimate their fatigue strength. One of the

possible methods for evaluating the fatigue strength for this type of joints is the local stress

criterion, for which the elastic notch stress concentration factor can be determined with the use of

numerical Finite Element Method. However, the methodology for notch stress concentration

calculations assumes the use of a notch rounding or an alternative circular geometry with a specified

radius. For most joints of over 5 mm thick sheets a fictitious radius of 1 mm is used. However, for

Dariusz Skibicki 105

the analysed laser welds rounding the notch with a 1 mm radius is not possible as dimensions of the

joined elements, welds and notches are substantially smaller. Therefore, an attempt has been made

to determine the value of fictitious notch rounding radius. The influence of radius value on stresses

was analysed. As a result, curves illustrating the influence of radius value on elastic stress

concentration factor were determined. Based on the analysis of the received results a rounding

radius of 0.15 mm was proposed as recommended for modelling weld notches for steel structures

with 2.5 mm thick cover plates.

References

[1] Kozak J., Fatigue durability estimation problems of all steel sandwich panels, Gdansk

University of Technology, ISBN 83-7348-136-2, Gdansk, 2005

[2] Reinert, T., Mühlenplatzabdeckung mit I-core Sandwich Paneelen. Papenburg: Öffentliche

Dokumentation, I-core Panels Sales & Production, 2006

[3] Remes, H., Strain-based approach to fatigue strength assessment of laser-welded joints,

Doctoral Dissertation. Espoo: Helsinki University of Technology, ISBN 978-951-22-9189-2, 2008

[4] Laitinen, R., Results of the hardness tests no. 580/01, 622/01, 468/02, 583/02. Ruukki: Raahe,

2003

[5] Laitinen, R., Kujala, P., Remes, H. i Nielsen, S., CO2-laser MAG Weldability of Laser Cutting

LASER RAEX Steels, Hull Structural Steel Grade A and High Strength Formable Steel OPTIM

RAEX 700 MC. In: Halmøy. Trondheim: E. Proceedings 9th Conference on Laser Materials

Processing in the Nordic Countries, Norwegian University of Technology, 2003

[6] Radaj, D. i Helmers, K., Bewertung von Schweisverbindungen hinsichtlich Schwingfestigkeit

nach dem kerbspannungskonzept. Konstruktion 49, p. 41–27, 1997

[7] Radaj, D., Sonsino, C. M. i Fricke, W., Fatigue assessment of welded joints by local

approaches, 2nd Ed. Cambridge, England: Woodhead Publishing Limited, ISBN-13:978-1-85573-

948-2, 2006

[8] Eibl, M., Sonsino, C., Kaufmann, H. i Zhang, G., Fatigue assessment of laser welded thin sheet

aluminium. International Journal of Fatigue 25, p. 719–731, 2003

[9] Sonsino, C., Kueppers, M., Eibl, M. i Zhang, G., Fatigue strength of laser beam welded thin

steel structures under multiaxial loading. Elsevier, International Journal of Fatigue 28, 657–662,

2006

[10] Pinho da Cruz, J., Costa, J., Borrego, L. i Ferreira, J., Fatigue life prediction in AlMgSi1 lap

joint weldments. International Journal of Fatigue 22, p. 601–610, 2000

© (2012) Trans Tech Publications, Switzerland

doi:10.4028/www.scientific.net/MSF.726.106

Andrzej Kurek1,a, Adam Niesłony1,b

1

Opole University of Technology, Faculty of Mechanical Engineering

Department of Mechanics and Machine Design

ul. Mikołajczyka 5, 45-271 Opole, POLAND

a b

a.kurek@po.opole.pl, a.nieslony@po.opole.pl

Abstract. The paper contains a description of fatigue life tests of titan-steel bimetal. The study

involved specimens made of bimetal which was a combination of S355J2 steel and SB G1 265

titanium, which was imposed in the material by explosive cladding method. The research shows that

the fatigue life of specimens made of native material, derived from cladded plate is less than the life

of specimens of titanium-steel bimetal.

Introduction

Energy of explosion has been applied for peaceful aims for a long time. Explosive materials can

be applied for mineral crushing, demolition of buildings, fire suppression (blowing the flame away),

stretching the belts and filling the airbags, shooting rackets or rescue cartridges, treatment of

metallic and non-metallic materials. Intense energy released in a short time while explosion gives a

possibility of realization of technological processes which could not be realized in typical

conditions. The mentioned technologies using the explosion force have one common feature –

suitable application of extreme values of velocity and pressure accompanying the explosion [1].

Explosive cladding allows joining materials with completely different properties which are difficult

to obtain by means of other methods of joining [2]. Clad obtained with this method are materials of

strongly gradient properties and they have complex joining zones. The clad materials are often

applied in processing apparatus (chemical and power industries). Wide application of titanium and

its alloys in power engineering (condensers, steam condensers, heat exchangers and steam turbines

in power plants and thermal-electric power stations) causes that the problem of fatigue life of

bimetallic clads, for example, those of steel-titanium type, becomes more and more important. As

for many applications, fatigue life of clads is the most important parameter. Taking the specify of

the problem into account, the tests are performed according to suitable standards, but the tests not

included into the standards are also realized. The strength tests presented in this paper concern the

metallic composite, so-called clad, obtained during so-called explosive cladding [3]. This paper is a

continuation of the previous papers concerning strength of clad materials [4], and special attention

is paid to fatigue life of such materials.

This paper presents results of the fatigue tests of the clad material being a joint of steel S355J2

with titanium SB265 G1, being the overlaid material, and the steel being a native material.

Thickness of the overlaid material (titanium) was 6 mm, so the specimens of untypical, determined

with FEM, shape and dimensions were used for fatigue tests.

Next, the specimens were prepared. The specimens 100 x 9 x 9 mm were tested at the fatigue test

stand existing at Department of Mechanics and Machine Design, Opole University of Technology.

In the tested specimens, the bimetal joint was located exactly in a half of their section. Thus, each

specimen was divided into two equal parts – steel and titanium (Fig.1a). Three series of tests were

performed. The specimens were subjected to a traditional fatigue test, i.e. alternating bending under

sinusoidally changing loading with constant amplitude of bending moment.

Dariusz Skibicki 107

TITANIUIM

Fig. 1 Scheme of attachment and loading of specimens (a) and a scheme of the loading method

depending on the specimen attachment: perpendicular (b) or in parallel (c) [4]

In the first series of tests, the specimen was located at the test stand in a way allowing to realize

bending according to the scheme (Fig. 1b) on the plane perpendicular to the bimetal joint. In such a

case, three fatigue tests were performed. They were presented also in the paper [4], and next seven

more tests were realized. The results of those fatigue tests were presented in Fig. 3. The specimens

are shown as loaded perpendicular to the bimetal joint plane.

The next series of tests was performed under different configuration of the clade loading. The

specimen was mounted at the stand in the way allowing the vector of the bending force to act on the

plane of bimetal joint (Fig. 1c). The test results are presented in Fig. 4 where the specimens are

loaded in parallel to the bimetal joint plane. In the third series of tests, the specimens were made of

the native material (steel S355J2), and they were obtained from the sheets after the explosive

welding. The used specimens had the same shapes and dimensions as the specimens tested in the

previous series, so it was possible to relate the obtained results to the results of previous tests. These

results are presented in Fig. 5.

The fatigue test stand MZGS-100 allows to realize tests under a constant value of the bending

moment, so the stress amplitudes used for plotting the presented fatigue characteristics have been

determined by means of calculations with FEM, and the linearly-elastic model of the material was

assumed.

Fig. 2 presents the fatigue test results obtained for the steel-titanium bimetal, where σa is the

maximum value of the stress amplitude, and Nf, is a number of cycles to the fatigue crack initiation

under loading on the plane perpendicular to the bimetal joint (Fig. 1a – the first series of tests)

including the stress amplitude used during the tests, occurring in both steel S355J2, and in titanium

SB265 G1. Drop of the specimen stiffness by 30% was assumed as the main criterion of the fatigue

crack initiation. Fig. 3 shows the fatigue test results for the steel-titanium bimetal for the stress

amplitude σa depending on a number of cycles Nf, under loading on the plane parallel to the bimetal

joint (Fig. 1c – the second series of tests) including the stress amplitude applied in tests, occurring

in both steel S355J2 and titanium SB265 G1. The double-logarithmic curve shown on Fig. 4

presents the fatigue test results showing dependence between the stress amplitude σa (in steel) and

108 Fatigue Failure and Fracture Mechanics

the fatigue life Nf for two kinds of the clad material loading and the native material together with

characteristics. Scatters of results are very important for fatigue tests, they also occur for

homogeneous materials. In the case of bimetallic materials clad with the explosive method we have

heterogeneous structure of the material, and differences in the test results can be much

different even for similar loadings.

Fig. 2 Graph of dependence of the stress amplitude σa = f(Nf) occurring in steel and titanium

Fig. 3 Graph of dependence of the stress amplitude σa=f(Nf) occurring in steel and titanium

From the tests it appears that titanium SB265 G1 was very often the material in which the crack

initiation occurs. It concerns the bimetallic specimens loaded on the plane perpendicular to the joint.

In spite of great difference in Young moduli of the joined materials, determining a lower stress

value, titanium has a lower fatigue life. On the other hand, in the case of specimens subjected to

bending on the plane perpendicular to the joint plane, the crack initiation occurred in the steel layer

of the bimetal. Thus, we can state that for a given specimen shape, loading by alternating bending

and the resulting stress distribution, the fatigue life of both materials is similar, and the final life is

influenced by different random factors caused by technology of materials joining (inclusions,

Dariusz Skibicki 109

heterogeneity of the joining cone etc.). Thus, while designing of machine elements made of bimetals

a designer should use fatigue characteristics including fatigue properties of the basic material and

the overlaid material [5]. The Wöhler’s curve for the considered material (see Fig.4) is an example

of such characteristic. The expected fatigue life of both steel and titanium should be checked, and

the lower fatigue life should be assumed as the life of the element.

If the tested specimens are loaded in parallel to the joint plane, the difference of stress amplitude

occurring in steel and titanium is greater, so the crack initiation process will proceed in the steel

layer at first. From the tests it also appears that the fatigue life of the specimens made of the native

material obtained from the clad is lower than the fatigue life of steel-titanium specimens. Moreover,

it has been proved that a change of a way of the bimetal loading influences its fatigue life. It can be

also seen that the specimen loaded in parallel to the joint plane has a little higher fatigue life.

References

[1] Niesłony A., Kurek A., Bański R., Čižek L. 2010. Static and fatigue tests of explosively

cladded materials – titanium-steel, Scientific Papers Opole University of Technology, Series

Mechanics z. 97, nr 337/2010 (in polish)

[2] Mckenney C.R., Banker J.G. 1971. Explosion-Bonded Metals for Marine Structural

Applications, Marine Technology, pp. 285-292

[3] Szulc Z., Gałka A., Bański R., Pocica A. 2007. Explosive cladding with titanium - the

development of technologies and areas of industrial applications, XII Scientific and Technical

Welding Conference "Progress, innovation and quality requirements of welding processes,",

Międzyzdroje 29-31.05.2007, s. 13-14 (in polish)

[4] Kurek A., Żok F., Niesłony A., Bański R. 2011. Strength tests of steel-titanium explosively

cladded bimetal, XXIV Conference "Problems of Development of Machines", s. 47-49 (in

polish)

[5] A. Niesłony and A. Kurek, “Influence of the Selected Fatigue Characteristics of the Material

on Calculated Fatigue Life under Variable Amplitude Loading,” Applied Mechanics and

Materials, vol. 104, pp. 197–205, Sep. 2011.

© (2012) Trans Tech Publications, Switzerland

doi:10.4028/www.scientific.net/MSF.726.110

Strength Steel

GALKIEWICZ Jaroslaw1, a

1

Politechnika Świętokrzyska w Kielcach, Wydział Mechatroniki i Budowy Maszyn, A. 1000-lecia PP

7, 25-314 Kielce, Poland

a

jgalka@tu.kielce.pl

Abstract. The detailed analysis of the tensile properties of the 1/2Y welded joint made of ultra-high

strength steel S 960 QC was carried out. The analysis concerned various parts of welded joint and

has been carried out with the help of both experiment and numerical simulation. Results were

compared with the data measured using the ARAMIS system. The purpose of the analysis was to

provide the constitutive relations for detailed analysis of the welded joint by finite element method.

Introduction

The analysis of the welded joint strength is a seemingly simple task. It is probably so when the

conventional structural steels are welded. It is certainly not so when the ultra-high strength steel are

welded. It is not easy to find a proper a high-strength binder to make the joint as strong, or stronger

than the welded material (BM). One may observe a complex distribution of the mechanical

properties as well as the secondary stresses within the weld material (WM) and heat affected zone

(HAZ). Classical structural integrity assessment proposed in the procedures such as FITNET [1, 2]

or API [3] may not be applicable. Often the question arises which materials should be taken into

account to decide if the problem of interest concerns undermaching (M<1), overmatching (M>1) or

evenmaching (M=1)situation. M=(σeWM)⁄(σeBM) where σe is a yield strength. Namely, the FITNET

does not take into account the HAZ in the analysis. It not necessarily is correct in the case of the

ultra-high strength steels.

It may happen in the case of ultra high strength steel that the WM is the weakest one. In such a

case we face the undermatching situation. This problem will be discussed in the paper. In such a

case it turns out that the HAZ influences strongly the behavior of the whole welded joint [4],[5].

The micro-structure and the properties of the HAZ change in transverse direction and through the

thickness of the joint. Another problem in the strength analysis is to decide on the HAZ size [6].

In the paper the finite element analysis of the welded joint is presented. The undermatching

welded joint is considered. Computations were carried out using the ADINA code. Various

configurations (models) of the joint were analyzed and compared from the point of view of the

benefits reached. Both two-dimensional and tridimensional models were used.

First the surface of the welded joint was polished and observed under the microscope in order to

define roughly different zones in the welded joint. In order to define the HAZ more precisely the

hardness measurements were made (Fig.1). Then, the mini-specimens (rectangular section 2x4mm

and initial length of extensometer 25mm) for tensile tests were machined from different zones of the

welded joint. The results of tensile tests are shown in Fig.2 [7].

Dariusz Skibicki 111

Fig.1. Results of the hardness measurements Fig.2. Results of tensile tests. Different curves

concern the specimens depicted in Fig.1

In the next step the specimens of the cross-section 6x3mm, and the axis perpendicular to the

weld were cut out, loaded and the strains were measured using ARAMIS video-extensometer [8].

Then results recorded by ARAMIS were used to produce the constitutive relations in different zones

of the welded joint.

Numerical simulation of the welded joint based on the data recorded in the tensile test

The purpose of this step of analysis was to simulate numerically the behavior of different zones

of the stretched welded joint. Computations were made using 3D models field with the four-nodes

elements of the tetra-type. To reduce the number of nodes only a half of the specimen was modeled,

due to the assumed symmetry. Geometry of the joint is shown in Fig.3.

Fig.3a. Specimen with the boundary conditions Fig,3b. Three different zones of the welded

and load joint

Boundary conditions (Fig. 3a) are introduced as a displacement applied to the selected nodes at

the selected cross-section. Displacements were measured by ARAMIS. The nodes at the opposite

side of the specimen are fixed in all directions. First, the simulation was carried out before the

necking of the weakest element. Until this moment one can assume that the stresses are relatively

uniform in the specimen. The strains measured at different points were later used to estimate the

constitutive relationships, locally in different zones.

112 Fatigue Failure and Fracture Mechanics

However, in the first model numerically tested, the constitutive relationships obtained using the

micro-specimens (Fig.2) were used. The BM is characterized by curve A4, the WM is characterize

by curve A1 and the HAZ by curve A2. The width of the HAZ was assumed 1.8mm. The tensile

properties are summarized in Table 1.

Material ε0 σ0 [MPa] E [GPa] ν

BM 5.62E-03 1090 194 0.3

HAZ 3.13E-03 607 194 0.3

WM 2.60E-03 505 194 0.3

First, the global responses of the specimen to the external loading measured experimentally and

computed numerically are compared. The result is shown in Fig. 4a.

Fig.4a. Comparison of the response of the Fig. 4b. Comparison of the strains measured and

specimen to the external loading recorded computed along the selected line within the

experimentally and computed numerically; welded joint (horizontal line in Fig. 5a)

Model 1

The responses measured and computed are close each other. However, the trend of changes

measured by the finite element method is not satisfactory. The curves do not converge. One may

expect that when the external loading increases the difference between computed and measured

values will also increase.

The strain distributions were also compared. Qualitatively the strain maps are shown in Fig.5.

They look similar to each other. However, when the strain values are computed and measured along

the same line (horizontal line in Fig. 5a) the differences are observed and at certain domains are too

large.

Next, the stress-strain curves recorded in three different points, shown in Fig.5, were compared.

One curve was computed numerically, using constitutive relationships received by testing the

micro-specimens, the second curve was recorded experimentally (strains were measured directly by

ARAMIS, stresses were computed dividing the external force be the actual cross-section). The

results are shown in Fig.6 and they are not satisfactory.

Dariusz Skibicki 113

a. b.

Fig. 5a. Strain map computed by FEM Fig. 5b. Strain map measured by ARAMIS

a. b. c.

Fig. 6. Comparison of the stress-strain curves: a) point close to the weld root, b) point close to the

specimen axis, c) point at the weld face.

In this Section more precise simulation of the welded joint is presented. The volume of the weld

was divided into ten strips of different tensile properties (Fig.7). The tensile properties of the

materials within each strip was determined using ARAMIS.

Fig. 7. Different zones within the weld material Fig.8. Tensile curves of different zones in WM

and experimentally measured tensile curves

within BM and HAZ

114 Fatigue Failure and Fracture Mechanics

Numerical procedure remains the same as in Model 1, except constitutive equations within the

WM. Tensile curves are shown in Fig.8. The global responses of the material measured and

computed are compared in Fig.9a. Strains along the same line as in the Model 1 are compared in

Fig, 9b. Force-elongation curves are not as close as in Fig.4a. However, now, they are almost

parallel. Computed strains are closer to the measured one than in the Model 1.

Fig. 9a. Comparison of the response of the Fig. 9b. Comparison of the strains measured and

specimen to the external loading recorded computed along the selected line within the

experimentally and computed numerically; welded joint; Model 2

Model 2

In the Model 2 the local stress-strain curves measured in computed in the selected points

converge very well. It can be observed in Fig. 10.

a. b. c.

Fig. 10. Comparison of the local stress-strain curves: a) point close to the weld root, b) point

close to the specimen axis, c) point at the weld face

Numerical simulation of the welded joint - tensile properties averaged within the HAZ

Model 2 provided better results of simulation than Model 1. Thus, in the next step the HAZ was

differently modeled. Instead of tensile characteristics of the HAZ, recorded experimentally, using

micro-specimens, the constitutive relationship recorded by ARAMIS was used. Quite arbitrarily the

representative point within the HAZ was selected. It was assumed that within this point the

constitutive relation is representative (average) for the whole HAZ. Such a curve is shown in Fig.

11.

Dariusz Skibicki 115

Fig. 11. Stress-strain curves used in Simulation according to Model 3. The new curve representing

the HAZ is the red curve. The previous curve representing the HAZ is denoted by rhombuses.

Numerical simulation according to the Model 3 provided a good agreement between the global

responses of the welded joint recorded experimentally and computed by FEM (Fig.12a). Strains are

distributed along almost parallel curves. However, these curves do not converge (Fig.12b). The

local stress-strain curves recorded and computed in three selected points converge perfectly (stress-

strain curves are similar to Figs. 10a-c).

Fig. 12a. Comparison of the response of the Fig. 12b. Comparison of the strains measured

specimen to the external loading recorded and computed along the selected line within the

experimentally and computed numerically; welded joint; Model 3

Model 3

116 Fatigue Failure and Fracture Mechanics

Fig. 13a. Comparison of the response of the Fig. 13b. Comparison of the strains measured

specimen to the external loading recorded and computed along the selected line within the

experimentally and computed numerically using welded joint. Computations were made using

plane strain and plane stress models plane strain and plane stress models of the

welded joint

Numerical simulations using plane strain and plane stress models of the welded joint

In the next step, the numerical simulation was made using simple plane stress and plane strain

models of the analyzed welded joint. The material data tested in Model 3 were used. Results of the

analysis along with the experimental results are shown in Figs 13a and 13b. One may notice a very

good agreement between experimental and numerical results for the plane stress model of the

specimen. Certainly, this observation can be used in the simulation of other welded joints of more

complex shape.

Conclusions

In the paper numerical simulation of the welded joint is presented. The undermatching case was

analyzed. It may happen when the ultra high strength steel is welded. The structure of the welded

joint is very complex. Modeling of the WM and HAZ is very difficult because of varying micro-

structure and tensile properties. It was assumed that more detailed map of the tensile properties

should be limited to the weakest zone. In the analyzed case it was the WM. In other zones the

"average" properties were adopted. They were either measured using the micro-tensile-specimens or

estimated using video-extensometer ARAMIS, which was useful in the analysis. The global and

local responses of the welded joint received by the numerical simulations were satisfactory if the

HAZ tensile properties were estimated by ARAMIS technique at the selected "representative

average" point. In this case, the computed numerically strain distribution along selected lines is

similar to the experimentally measured strains. However, the computed strains for final stage of

loading are always several per cent lower than the strains measured by ARAMIS.

The numerical results received by using the plane stress model of the welded joint provide good

results, which are close to the measured experimentally.

Acknowledgements: The financial support from Polish Ministry of Science and Higher

Education under contract N N501 199640 is gratefully acknowledged. The ARAMIS testing system

was purchased within the European Union Grants to Kielce University of Technology: Modin II and

Molab.

Dariusz Skibicki 117

References

[1] FITNET Report, (European Fitness-for-service Network). Edited by M.Kocak, S.Webster,

J.J.Janosch, R.A.Ainsworth, R.Koers, Contract No. G1RT-CT-2001-05071, (2006)

[2] Neimitz A., Dzioba I., Graba M., Okrajni J.: Ocena wytrzymałości, trwałości i bezpieczeństwa

pracy elementów konstrukcyjnych zawierających defekty, Wydawnictwo Politechniki

Świętokrzyskiej, 2008

[3] American Petroleum Institute, API 579: Recommended practice for fitness-for-service.

Washington DC, 2000.

[4] Rodgigues D.M., Menezes L.F., Loureiro A., The influence of the HAZ softening on the

mechanical behavior of welded joints containing cracks in the weld metal, Engineering Fracture

Mechanics, 71, 2053-2064, 2004

[5] Rodgigues D.M., Menezes L.F., Loureiro A., Fernandes J.V., Numerical study of the plastic

bahaviour in tension of welds in high strength steels, International Journal of Plasticity, 20, 1-18,

2004

[6] Zhang Z.L., Hauge M., Thaulow C., Odegard J., A notched cross welds tensile testing method

for detrmining thrue stress-strain curves for weldments, Engineering Fracture Mechanics, 69, 353-

366, 2002

[7] Gałkiewicz J., Dzioba I., Pała R., Właściwości mechaniczne złączy spawanych z ultra

wysokowytrzymałych stali ferrytycznych, XXIV Sympozjum Zmęczenie i Mechanika Pękania,

Bydgoszcz - Pieczyska, maj 2012

[8] ARAMIS User Manual, GOM mbH

© (2012) Trans Tech Publications, Switzerland

doi:10.4028/www.scientific.net/MSF.726.118

Opole University of Technology, Faculty of Mechanical Engineering, Department of Mechanics

and Machine Design, ul. Mikołajczyka 5, 45-271 Opole, Poland

a b

l.blacha@doktorant.po.opole.pl, a.karolczuk@po.opole.pl

Keywords: welded joints, fatigue failure, weakest link concept, failure probability analysis efficient

material.

Abstract. The paper presents results of simulation and experimental tests carried out in order to

identify the S-N curve introduced for the evaluation of fatigue life for various types of steel welded

joints. Such curve is explicitly defined through the corresponding values of Cf and mf parameters.

The identification process involved finite element analysis, undertaken for different discrete models

varying in notch radius. The obtained results served as an input into iterative calculations carried for

different values of Cf and mf parameters. Such calculations were aimed at minimization of the

estimator in the form proposed for the process of identification.

Introduction

Fatigue of steel welded joints in the as-welded condition is an issue of a complex matter which is

being attributed to the presence of geometric and material inhomogeneity. Variability in shape of the

weld face, especially the notch radius, implies difficulties during the assessment of stress fields.

Changes in material microstructure are due to localised melting of the material, corresponding to the

presence of heat affected zone (HAZ), transient between the microstructures of parent material and

weld. Within the HAZ the following microstructural areas tend to develop: recrystallization area,

normalization area, overheated area and the area of fusion. Widmannstätten microstructural pattern

observed in the overheated zone indicate HAZ as the area where the damage process of the entire

structure is initiated.

Scatter of weld profile parameters and inhomogeneous stress fields resulting from varying

material structure substantiate the probabilistic approach to fatigue assessment of steel welded

joints, e.g. the weakest link concept.

The aims of this paper are to: (i) define the efficient material S-N curve, formulated as a part of

probabilistic computational model for wide variety of steel welded joints; (ii) introduce the

algorithm for identification of such curve; (iii) present results derived from identification process.

A computational model based on extension of the weakest link concept into area of steel welded

joints was presented in the literature [1]. The weakest link concept is in the group of non-local stress

approaches to fatigue assessment, in the sense that it utilizes the complete stress fields instead of the

local values. Computational model defines the structural element as a serial system of smaller,

independent elements, having its origin in the reliability theory.

In the proposed method division of structural element into smaller elements is performed within

the frames of meshing operations during the finite element (FE) model generation. Each element is

described by failure probability Pf defined by the Weibull distribution. Once the Pf probability for

each element is known, application of the weakest link concept in the volume of the material

defines the cumulative failure distribution function Pf on a certain loading level as a function of

number of cycles to failure N:

Dariusz Skibicki 119

p

1 log N

− ∫ dV

V0 H

Pf ( N ) = 1 − e V , (1)

where:

V – material volume,

V0 - reference volume,

N - number of cycles to failure,

H – scale parameter,

p – shape parametr, p = 66.82.

This form of distribution (1) is described by two parameters: shape parametr p and scale

parameter H. The p parameter is interpreted as an exponent corresponding to the volume effect,

defined as decrease in fatigue life with increasing volume of the joint. The shape parametr p was

determined on the basis of two series of fatigue tests on welded elements submitted to fully reversed

loading of a tension-compression type [1]. Volume of the joint in second serie of tests was set at the

level of eight times smaller than in the first serie. Identification of the shape parameter p was based

on comparison between S-N curves derived from each serie of tests. Obtained results allowed to

estimate the value of p = 66.82. The shape parameter is considered as a constant value within the

certain range of number of cycles to failure and under the assumption of being independent from

stress range ∆σ. Such assumption appears to be valid for cycles to failure that are not exceeding

values close to the fatigue strength.

The H parameter normalizes logN variable for a given loading level. For a certain range of cycles

the scale parameter can be described by the form of H = logNf , where Nf is the number of cycles to

failure of an efficient material, for a certain value of Ps:

where:

Cf, mf - fatigue parameters of efficient material,

∆σ - stress range according to the chosen multiaxial failure criterion.

In case of joints in as-welded condition grade of the steel (parent material) is of secondary

importance for fatigue strength [2]. This feature arises from the joining process: consumable

electrode arc welding in shielding gas atmosphere. Cyclic properties of the overheated material

sections dominate over the properties of parent material. Cyclic behaviour of a material with this

microstructure is described by the scale parameter: S-N (∆σ - Nf ) curve for efficient material.

The well established approaches to fatigue design of welded structures presented in numerous

standards and recommendations [3, 4] are based on a series of classified S-N curves corresponding

to each class of joint geometry. S-N curves in the form as in recommendations from International

Institute of Welding (IIW) [3] are identified by the stress range ∆σ at 2.106 cycles, known as the

FAT class. Each curve corresponds to survival probability Ps equal of 95%.

The process of identification of the Cf and mf parameters proceeds through minimization of the

E(Cf,mf) estimator for efficient material parameters. It leads through iterative selection of the pair of

values that determine Eq. 2 and utilizes FAT curves and the nominal stress approach [3]. The idea

of identification requires knowledge on number of cycles to failure of a welded joint, obtained for

specific value of survival probability; this assumption can be ideally met by application of FAT

curves classified by IIW. The performed process was based on calculations that were carried out

according to the algorithm shown in Fig. 1. Identification estimator for the efficient material

parameters is described by Eq. 3:

120 Fatigue Failure and Fracture Mechanics

p

1

− (log N FAT ) p ∫ dV

log C − m f log ∆σ

E (C f , m f ) = 0,95 − e V f

, (3)

where:

NFAT – number of cycles to failure according to the nominal stress approach; referential volume

V0 (Eq. 1) for efficient material is assumed to be equal of 1 mm3.

nominal stress approach

corresponding to the chosen nominal stress ranges

of E(Cf,mf) identification estimator

Fig. 1. Algorithm for identification of the efficient material S-N curve parameters

In the assumptions underlying to the model, grade of the steel elements being joined, as well as

their geometry, are not affecting the efficient material characteristic. From the standpoint of

validation, the identification process was carried for two welded elements of a different geometry:

transverse butt weld (type a joint) and a transverse stiffener (type b joint). Geometry of the analysed

elements is shown in Fig. 2. In the nominal stress approach [3] these elements are classified as

structural details of no. 213 and 511, respectively.

(a) (b)

Fig. 2. Geometry of the investigated elements: a) transverse butt weld [3], b) transverse stiffener [3]

Finite element analysis was carried out for FE models of a geometry being representative for both

welded elements shown in Fig. 2. Models for the analyzed elements are shown in Fig. 3.The

elements were subjected to axial loading. Analysis was undertaken for the loadings corresponding to

the stress ranges at the level of ∆σ = 250 [MPa] and ∆σ = 300 [MPa], imposed for the model of 10

variations in notch radius ρ, located along the weld line (ρ = {0.1-1.0} [mm]). Models were meshed

by Hex8 [5] elements. Global edge length of the elements in the notch section (section I, Fig. 3) was

set to be equal of 0.1ρ. The linear elastic material properties for steel with Young’s modulus equal

of 201 GPa and Poisson’s ratio equal of 0.3 were used throughout the calculation process.

Dariusz Skibicki 121

Fig. 3. Partial FE models, representative for the following welded elements: a) transverse butt weld,

b) transverse stiffener

Calculated components of ∆σij(x,y,z) stress fields acting in vicinity of the welded joint were

transformed according to the maximum principal stress criterion. In order to compare the value of

the estimators derived for two loading levels, the following transformation formula was used:

where:

E(1), E(2) – estimator values derived for 1st and 2nd loading level, respectively.

Results obtained from the performed calculations were further used to determine the map of

values of identification estimator E for the efficient material Cf, mf parameters. Example map is

shown in Fig. 4. The derived maps allowed to initially assess the location of minimun of the

function (4).

Retrieved data served as a starting point for the process of minimization through the downhill

simplex optimization technique proposed by Nelder and Mead and implemented in MATLAB

environment (R2011b). Identification process involved finite elements within the distance x = t/3

from the root of the notch, where t is the thickness of welded plates (in both cases t = 10 mm). The

log(Cf), mf parameters extracted for each variation of FE model and joint type are presented in

Table 1.

122 Fatigue Failure and Fracture Mechanics

Fig. 4. An example map of values of identification estimator E for efficient material Cf, mf

parameters

Table 1. Efficient material log(Cf), mf parameters for different values of notch radius ρ and joint

geometry

Type a joint (butt weld, Type b joint (transverse stiffener,

ρ, [mm] IIW no. 213) IIW no. 511)

E, ·10-7 log(Cf) mf E, ·10-7 log(Cf) mf

1,0 6 13.77 2.98 7 13.68 3.11

0,9 6 13.70 3.04 6 13.71 3.11

0,8 7 13.68 3.07 1 13.69 3.12

0,7 2 13.66 3.10 5 13.69 3.13

0,6 5 13.66 3.12 2 13.68 3.13

0,5 6 13.65 3.14 6 13.67 3.14

0,4 7 13.65 3.16 7 13.66 3.14

0,3 4 13.65 3.17 4 13.66 3.15

0,2 9 13.65 3.19 6 13.65 3.15

0,1 3 13.65 3.20 7 13.65 3.16

Influence of proper selection of Cf and mf parameters on the estimated fatigue life was tested

through sensitivity analysis. Sensitivity was estimated through total differential of Eq. (2):

1 C f ⋅ ln (∆σ )

∆N f = ∆C f + ∆m f , (5)

m m

(∆σ ) f (∆σ ) f

where:

∆Cf – overestimation of Cf parameter,

∆mf – overestimation of mf parameter.

Calculations were carried out for stress range ∆σ = 100 MPa and Cf, mf values as referred to the

FAT curve for structural details no. 213 and 511 (transverse butt weld and transverse stiffener,

respectively). Results were used to generate the map of sensitivity assessment in the form shown in

Fig. 5. The mf parameter has proved to have greater influence on the resulting number of cycles to

failure; 10% overestimation in the value of mf parameter demonstrated 50% overestimation in

fatigue life N.

Dariusz Skibicki 123

Fig. 5. Map of sensitivity assessment, where: ∆mf – relative overestimation of mf parameter; ∆Cf –

relative overestimation of Cf parameter; ∆N – relative overestimation of fatigue life

Summary

Based on the results of fatigue tests and iterative calculations, fatigue parameters were extracted

that determined the efficient material ∆σ - N curve, introduced as a representation of the

microstructure in overheated area of a welded joint. Calculations were carried for two FE models, in

10 variants each, representative for two axially loaded steel welded joints. From the results, the

following conclusions were drawn:

- the mf and logCf parameters are changing with increasing value of notch radius ρ; this effect is

evident since when the notch rounding becomes smoother stresses in the root of the notch are

approaching nominal values when ρ →∝ and mf and logCf parameters will gain their values

according to the nominal ∆σ -N curve (in the analysed cases impossible due to the modeled shape

of joint);

- for the notch radius ρ = {0.1-0.2}[mm] logCf value is identical for both joint types while mf

values differ by 1.25% which leads to less than 10% change in fatigue life N;

- the undertaken analysis over the values of efficient material parameters yielded the following

results for the variant of notch radius being modeled as ρ = 0.1 [mm]: logCf = 13.65 and mf = 3.18

(average value derived from two variants, leading to several percent deviation in fatigue life in the

analysed cases).

Acknowledgements

This paper is realized within the framework of research project No. DEC-2011/01/N/ST8/02566

funded by the National Science Centre in Poland.

124 Fatigue Failure and Fracture Mechanics

References

[1] Ł. Blacha, A. Karolczuk, Application of the weakes link analysis to the area of fatigue design of

steel welded joints, Fifth International Conference on Engineering Failure Analysis, Elsevier

Publishing, The Hague, The Netherlands, 1-4 July 2012.

[2] C. M. Sonsino, H. Kaufmann, G. Demofonti, S. Rifisculi, G. Sedlacek, C. Müller, F. Hanus, H.

G. Wegmann, High-Strength steels in welded state for light-weight constructions under high and

variable stress peaks, ESCC Steel Research Programme, CSM – Roma, LBF – Darmstadt,

Published by the European Commission, Brussels 1999.

[3] A. Hobbacher, Recomendations for fatigue design of welded joint and components, IIW

document XIII-2151-07/XV-1254-07, Paris, 2007.

[4] American Bureau of Shipping, Guide for the fatigue assessment of offshore structures, ABS

document, Houston, 2003 - 2010.

[5] MSC/PATRAN, MSC. The MacNeal-Schwendler Corporation, ver. 2005.

© (2012) Trans Tech Publications, Switzerland

doi:10.4028/www.scientific.net/MSF.726.125

explosive welding

Aleksander Karolczuk1,a, Krzysztof Kluger1,b, Mateusz Kowalski1,c,

Fabian Żok2,d and Grzegorz Robak Author1,e

1

Opole University of Technology, ul. Mikołajczyka 5, 45-271 Opole, Poland

2

Z.T.W Explomet s.c., ul. Oświęcimska 100H, 45-641 Opole, Poland

a b c d

a.karolczuk@po.opole.pl, k.kluger@po.opole.pl, mat7711@gmail.com, fabianzok@interia.pl,

e

g.robak@po.opole.pl

Abstract. The main aim of the paper is determination of residual stresses in explosively welded

steel-titanium bimetal. The analysis considers two bimetallic specimens: before and after the heat

treatment. In residual stress determination the hole drilling method along with finite element

analysis were applied. The results show different residual stress states depending on the heat

treatment. The obtained results are confirmed by thermal stress calculation.

Introduction

Increasing demands in area of efficiency, strength, reliability and fatigue life of modern

engineering structures requires application of non-typical materials. Those non-typical materials

must satisfy different mechanical properties such as, e.g. high corrosive and temperature resistance

along with high strength. In order to meet all the requirements the bimetallic composite are often

taken into account. Explosive welding is one of manufacturing technology that allows joining

structurally dissimilar metals such as steel and titanium [1]. The joint of materials is obtained during

detonation in which one metal (the flyer) undergoes a large deformation. Typical interface in case of

explosive welding has wavy shape although flat shape is also possible. The quality of obtained joint

depends on many factors [2]. The most important are as follows: detonation velocity, standoff

distance between metals, structural and mechanical properties of bonded materials. The joint is

obtained only under proper range of welding parameters (welding window) [3]. Detonation imposes

a high velocity collision of metals in which a high velocity jet is formed. This air jet swept away the

oxide films and other impurities making the bond possible. The research of joint area shows that

explosive welding is a solid state metal joining process [4]. Because of unique advantages the

bimetals are used in chemical and energy process apparatus, e.g. for heat exchangers, reactors, pipes

and others devices that require high resistance against applied fluids [1, 4, 5]. Sensitivity to

inappropriate selection of welding parameters is a disadvantage of the process. The dynamic

character of explosive welding still prevents full understanding of its mechanisms. The research of

bimetallic composites is focused on inhomogeneous structure and optimization of welding process.

The problem of monotonic mechanical properties is also raised [6, 7]. Unfortunately, there is not

enough information concerning cyclic properties of bimetallic composites. Some limited

information could be found in [8, 9]. However, the tests presented in [8] concern only cyclic three

point bending and influence of heat treatment on fatigue life of titanium-steel composite. The cyclic

stress-strain relation was not investigated. The specific character of explosive welding process is

reflected in micro-structural properties [10]. The metallographic research exposes changes in

microstructure of subsequent layers of bimetal [10]. Large deformation that undergoes the cladding

metal during explosive could result in high residual stresses [5, 11]. The residual stresses could have

detrimental effect on fatigue life. Preliminary fatigue tests [12] performed under tension-

compression loading of explosively welded steel-titanium components demonstrate two unexpected

fatigue phenomena: (i) ratcheting (accumulation of strain) and (ii) cyclic instability (fig. 1).

126 Fatigue Failure and Fracture Mechanics

Fig. 1. Preliminary fatigue test results of steel-titanium bimetal (S355J2+N and Titanium ASTM

Grade 1) under push-pull loading with zero mean value of force. Hysteresis loops registered under

different ratio of current and total number of cycles, n=N/Nf

In or order to explain these effects the elastic-plastic simulations are planed but firstly the initial

stress conditions must be known. Thus, the residual stresses must be determined. In the present

paper the procedure of measurement of residual stresses is presented.

Research setup

In the residual stress measurement two specimens were used: S1 and S2 (210x180x46 [mm])

made of bimetal obtained by explosive welding process of steel (S355J2+N) and titanium (ASTM

Grade 1) plates. Titanium plate with thickness of 6 [mm] was welded to the base plate made of steel

with thickness of 40 [mm]. The specimen S1 and S2 were cut out from one welded plate

(4330x3150x46 [mm]) designated for a heat exchanger construction. The place of cutting is

presented in figure 2.

Fig. 2. Photograph of explosively welded plate with visible place where the specimens were cut out

The specimen marked by S1 was cut directly after the welding process and specimen marked by S2

was cut after the heat treatment and flattening process of the original plate. The heat treatment

consisted in heating for 90 minutes at 600[oC] and then cooling down to 300[oC] at cooling velocity

100[oC/hour]. The final cooling was performed in the air outside the furnace. The chemical

composition and basic mechanical properties of bonded metals are presented in tables 1 and 2,

respectively.

Table 1. Chemical composition of S355J2 steel (EN 10025-2:2004) and titanium ASTM Grade 1

Steel S355J2 (EN 10025-2: 2004)

Chemical element: C Si Mn P S Cu

In weight, %: 0.23 0.60 1.70 0.035 0.035 0.45

Titanium Grade 1 (ASTM Grade 1)

Chemical element: C Fe H N O Ti

In weight, %: 0.10 0.20 0.015 0.03 0.18 99.5

Dariusz Skibicki 127

Table 2. Mechanical properties of S355J2 steel (EN 10025-2:2004) and titanium ASTM Grade 1

Material Re, MPa Rm, MPa E, MPa G, MPa ν, - A5, %

* *

S355J2 382-395 598-605 220000 84000 0.3 24-34*

Grade 1 189-215

308-324* 100000 38000 0.37** 43-56*

(R02)*

*

- according to manufacturer certificates, **- in titanium after welding

Where: E, G – Young and Kirchhoff modules , Re (R02) – yield stress,

Rm- ultimate strength, ν - Poisson ratio, A5 - elongation

Measurement of residual stresses. The residual stresses were measured by incremental hole-

drilling method. The method consists in strain relaxation measurements on the specimen surface

around progressive drilled hole. The strain relaxations were measured by a special type strain gauge

rosette [13]. In figure 3a a scheme of applied type A strain gauge rosette is presented. The

experimental strain results were obtained by using TFrw-1.5/120 gauge [14] with a gauge circle

diameter of D = 5.6 [mm] (fig.3a). The TFrw-1.5/120 gauge is characterized by resistance equal to

120 [Ω] and strain sensitive coefficient k = 2.15. Strain values were registered with SCXI-1600

strain gauge (National Instruments manufactured) and LabView Signal Express application. In a

hole drilling process a helix drill BOSH HSS-R with diameter 1,5 [mm] (diameter tolerance h8) was

used. In figure 3b an example of specimen during drilling operation is presented.

(a) (b)

Fig.3. (a) Scheme of strain gauges type A, (b) Photo of strain gauge under drilling

Constant drilling speed equal to 2850 rpm was used. Drilling depths were measured with the help of

dial gauge (±0,01 [mm]) which was coupled with drill chuck. Since the value of calibration

constants used in standard procedure described in [13] are unknown the finite element method

(FEM) was applied to calculate residual stresses (Comsol 3.5). Values of residual stresses in

material are determined based on changes in elastic part of strains around a drilling hole. Because

the strain gauge location is shifted from the drilled hole the strain relaxation does not correspond

directly (through Hook's law) to residual stresses. This discrepancy depends on size and location of

strain gauge, and also on the drilling hole depth. These factors could be taken into account using

FEM simulation (correction parameter p). The applied algorithm for residual stresses determination

is as follows:

- strain measurements in three directions (A, B, C) for drilling hole depth h,

- change of strain signs,

- calculation of the principal strains ε1,2,

- calculation of principal stresses σ1, σ2 from the Hook’s law for a plane stress state,

- multiplication of the principal stress values by correction parameter p adequate to the hole depth

h.

Values of correction parameter p were determined with FEM simulation for different values of the

hole depth h. The correction parameter value p is simply calculated by division of simulated value

of residual stress σr and calculated value of stress σcal, where σcal is computed using the Hook's law

128 Fatigue Failure and Fracture Mechanics

for a plane stress state from determined by FEM strain relaxation. The strain relaxation determined

by FEM is an averaged value of strain relaxation over the area of virtual strain gage (fig. 4). In order

to increase the computations the real shape of drill is modeled (fig. 4). Figure 4 presents an

exemplary strain distribution εx in cross section of modeled specimen (solid 3D) for the hole depth

of 1.5 [mm] (where x – direction of simulated residual stresses, σx = 1 [MPa]). The position and

length of the virtual strain gauge along with computed stress field σx is presented in figure 5a.

Figure 5b presents exemplary strain distribution εx over the area of the virtual strain gauge. It is

evident that the strain field over the area of the virtual strain gauge must be averaged for practical

usage since the strain gradient is meaningful.

along with virtual strain gauge location (h=1.5 [mm])

(a) (b)

Fig. 5. (a) Exemplary stress field σx along with virtual strain gauge location; (b) Strain distribution

εx on specimen surface along axis symmetry of virtual strain gauge

For a given (simulated) value of residual stress (σr = 1 [MPa]) several FEM models were created

with different hole depth h. For each h value the correction parameter p was computed (fig.6).

Furthermore, the distribution of correction parameter p over h was investigated separately for steel

and titanium using appropriate linear elasticity material constants which results with the same

relation p(h).

Dariusz Skibicki 129

Measured and computed results. Figure 7 presents results of strain measurements in three

directions (A,B,C) and residual principal stress (σ1, σ2) computations for specimen S1 (before the

heat treatment) and for specimen S2 (after the heat treatment). The strain measurements were

performed in two places (point I and point II) for each specimen. The distance between two points

where strain gage rosette were attached is around 55 [mm] (fig. 8).

(a) (b)

(c) (d)

Fig. 7. Measured strains in three directions (A,B,C) in function of depth of drilled hole and

calculated principal stresses: (a) without the heat treatment – point I, (b) without the heat treatment

– point II; (c) after the heat treatment – point I, (d) after the heat treatment – point II.

130 Fatigue Failure and Fracture Mechanics

(a) (b)

Fig. 8. (a) Photo of the specimen S1 with two strain gauge rosettes; (b) Photo of the specimen S2

with two strain gauge rosettes

For the specimen S1 without the heat treatment measurements and calculations show tensile stresses

in titanium. Whereas, in the specimen S2 that is after the heat treatment the residual stresses are

negative (compressive). The applied algorithm of residual stress calculations assumes homogeneous

field of stresses. The evaluated residual stresses depend on the depth h of drilled hole (fig.7) that

needs a proper interpretation. The value of stresses σ(h) for given depth h is the mean value of

residual stresses in material layer of thickness h. The values of residual stresses stabilizes for all

measured points for drilling depth of h = 3 mm. The stabilized values are the most representative

regarding to fact of large sensibility of p parameter to the hole depth h up to 2 mm (fig. 6). The

applied residual stress calculations do not consider the influence drilling operation on stress state

which can be significant because of the low yield stress of titanium (tab. 2). However, the

introduction of persistent deformations as a result of drilling process has the same effect for all

measured points. The qualitative estimation of the residual stresses is correct, i.e.: (i) tensile stresses

for titanium without the heat treatment, (ii) compressive stresses for titanium after the heat

treatment. The tensile stresses appearance in titanium without the heat treatment can be explained

by detailed analysis of explosive welding (fig. 9). The titanium layer undergoes tension in x axis

direction as results of detonation pressure because of standoff distance existence between bonded

layers (fig. 9). Whereas, appearance of the compressive stresses after the heat treatment is result of

different thermal expansion coefficient for bonded metals: for titanium αTi = 8,6·10-6 [1/K] and for

steel αSt = 13,0·10-6 [1/K]. The heat treatment causes titanium recrystallization and stress relaxation

but only at temperature 600 [oC]. During cooling process down to temperature 20oC titanium and

steel reduces theirs volumes. Because the steel has larger thermal expansion coefficient its change

of dimensions is larger that initiates the compressive stress for the titanium and tensile stress for the

steel. Assuming: (1) the linear dimensions change range (∆L = L·∆T·α); (2) the elastic strain state;

(3) deformation compatibility; (4) a plane stress state; it is calculated that as a result of temperature

decrease about 580 [oC] stresses for the titanium are σx=σy= -377 [MPa] and for the steel

σx=σy=56 [MPa]. The stresses in the titanium exceed the yield stress which means that the above

mentioned assumption of the elastic strain state is not satisfied. The assumption of constant thermal

expansion coefficients α (assumption 1) is also only an approximation. However, the calculated

signs of thermal stresses correspond to measured-computed signs of residual stresses.

Dariusz Skibicki 131

Summary

Basing on the obtained experimental results and performed theoretical analyses the following

conclusions are drawn:

(i) The performed heat treatment of bimetal steel - titanium does not remove residual stresses but

only changes their values and directions.

(ii) After the heat treatment the residual stresses in the titanium are compressive that could be

beneficial for fatigue loading but introduces tensile stresses in the steel. Initial tensile stresses

are unbeneficial for fatigue loading and could be the reason of ratcheting effect observed in

preliminary fatigue tests [12].

(iii) Future tests should take into account appearance of persistent strain change caused by drilling

process.

References

[1] J. Banker E., Reineke, Explosion Welding, ASM Handbook, Vol 6 Welding Brazing and

Soldering ASM International, 1993, pp. 303-305.

[2] S.A.A. Akbari Mousavi, S.T.S. Al-Hassani, A.G. Atkins, Bond strength of explosively welded

specimens, Materials and Design 29 (2008) 1334–1352.

[3] S.A.A. Akbari Mousavi, P. Farhadi Sartangi, Experimental investigation of explosive welding

of cp-titanium/AISI 304 stainless steel, Materials and Design 30 (2009) 459–468.

[4] S.A.A. Akbari Mousavi, P. Farhadi Sartangi, Effect of post-weld heat treatment on the interface

microstructure of explosively welded titanium–stainless steel composite, Materials Science and

Engineering A 494 (2008) 329-336.

[5] L.B. Pervukhin, Y.U. Mal’tsev, A. Konon, B.D. Tsemakhovich, A.D. Chydnovskii,

Distribution of internal stresses in bimetal steel 22K+steel Kh18N10T obtained by explosive

welding, Metal Science and Heat Treatment 17/11 (1976) 934-937.

[6] R. Kacar, M. Acarer, An investigation on the explosive cladding of 316L stainless steel-din-

P355GH steel, Journal of Materials Processing Technology 152 (2004) 91–96.

[7] Mustafa Acarer, Bilge Demir, An investigation of mechanical and metallurgical properties of

explosive welded aluminum–dual phase steel, Materials Letters 62 (2008) 4158–4160.

[8] L. Čižek, D. Ostroushko, Z. Szulc, R. Molak, M. Prażmowski, Properties of sandwich metals

joined by explosive cladding method, Archives of Materials Science and Engineering, 43(1)

(2010) 21-29.

[9] D. Ostroushko, E. Mazancová, Chosen properties of sandwich CrNi steel-Ti material after

explosive cladding, Conference proceedings of 19th International Conference on Metallurgy

and Metals, Metal 2010, Czech Republic.

132 Fatigue Failure and Fracture Mechanics

[10] S.Król, R. Bański, Z. Szulc, A. Gałka, Practical aspects of structural tests of titanium-steel

bonds made by explosive cladding and exposed to thermal process loads, Advances in Material

Science, 7 (2007) 50-56.

[11] M. Sedighi, M. Honarpisheh, Experimental study of through-depth residual stress in explosive

welded Al–Cu–Al multilayer, Materials and Design 37 (2012) 577–581.

[12] A. Karolczuk, M. Kowalski, R. Bański, Ż. Żok, Fatigue tension-compression testing of steel-

titanium bimetal produced by explosive welding, XI Scientific National Polish Conference

"Titanium and its alloys - 2011", Rzeszów University of Technology, p. 4 (in Polish).

[13] ASTM E837-08. Standard test method for determining residual stresses by the hole drilling

strain-gauge method. West Conshohocken: American Society for Testing and Materials; 2008.

[14] Information on http://www.tenmex.pl

© (2012) Trans Tech Publications, Switzerland

doi:10.4028/www.scientific.net/MSF.726.133

Joints in Multilayer Pipes

Stanisław Mroziński 1,a, Michał Piotrowski 2,b

1

University of Technology and Life Sciences In Bydgoszcz, Prof. Kaliskiego 7 85-796 Bydgoszcz,

Poland

2

University of Technology and Life Sciences In Bydgoszcz, Prof. Kaliskiego 7 85-796 Bydgoszcz,

Poland

a

stmpkm@utp.edu.pl, bm.piotrowski@utp.edu.pl

Abstract. The paper assessed the impact of the laser welding speed on the strength and fatigue

properties of the aluminum layer found in multilayer pipes. The conducted experiment has shown

that during the adjustment of the welding speed one has to take into account not only the results of

static tests, but also the results of fatigue tests. The impact of the welding speed on fatigue life

depends on the level of stress σmax. This level is slight in the area of the biggest stresses and

increases along with the decrease in stresses.

1. Introduction

A large number of advantages of composite materials causes that their application in many

industries rises very quickly. One example of such an application are multilayer pipes used both for

the distribution of drinking water, usable water and as well as in the installations of central and

underfloor heating systems. The multilayered structure of the pipe consists of a few layers: high-

density, high-temperature resistant polyethylene, aluminum and glue (Fig. 1). Because of the many

advantages, including: resistance to low and high temperatures, stone overgrowth resistance and the

ability to self-compensate thermal elongations multilayer pipes find wide applications.

Laser beam

PE Glue Al Glue PE

Aluminum plate

a) b)

polypropylene and polybutene in temperatures exceeding 60°C is oxygen permeability [1]. The

presence of oxygen in installation water is the main cause of corrosion of metal elements in central

heating systems. For the production of plastic pipes the application of non-metal layers is made on

the pipes limiting the stream size of diffusing oxygen. The more effective way of eliminating this

dangerous occurrence is the application of an aluminum layer in pipes 100% reducing the inflow of

oxygen to the installations. During the design process of multilayer pipes it is assumed that the

pipes during use are exposed to variable loads and creep deformation. Considering the low

thickness of the aluminum layer and the necessary tightness, the aluminum layer on a polyethylene

pipe is laser-welded. The welding optimization process requires to appropriately adjust welding

parameters, which include welding speed and laser beam power. Subjecting the pipe during a

technological process to the influence of strong heat stream causes the occurrences of material

134 Fatigue Failure and Fracture Mechanics

heterogeneity (structural notches). These apply to both the polyethylene layer and the aluminum

layer. During operation there can appear in these areas the beginnings stages of cracks, which

further lead to the damage of the pipes.

The researched problem is the impact that laser welding parameters have on the fatigue life of

multilayer pipes. The presented paper deals with the assessment of the impact of the laser welding

speed on the mechanical properties of the welded joints used in multilayer pipes.

2. Experiment methodology

Test samples were taken out of ⌀50mm multilayer pipes, where during the process of joining the

0,6mm thick aluminum layer seven speeds of laser welding were applied ranging from 3,6m/min to

6m/min (3,6; 4,0; 4,4; 4,8; 5,2; 5,6; 6,0m/min). Constant value of the laser beam power equaled

4kW was applied for these welding speeds. The way of extracting the test samples from the pipe is

shown below in Fig. 2.

Weld joint

Weld joint

Test sample

Test sample

a) b)

After cutting off 10mm wide rings from the pipe a cut was made alongside the weld (Fig. 2a) and

on the opposite side from the weld (Fig. 2b). As a result, test samples with and without a weld were

obtained. For the purpose of comparison, other samples from a plate used for the pipe braid were

also prepared for the experiment. The dimensions of the plate samples matched those samples

shown in Fig. 2. Static and fatigue tests were preceded by a metallographic analysis of the laser

welds. In Fig. 3 there are shown three selected microsections of laser joints obtained from three

different welding speeds.

a) b) c)

Fig. 3. Welding speed impact on the dimensions of the laser weld: a) v=3,5; b) v=4,4;

c) v=6m/min

On the basis of the conducted analysis of the microsections it can be stated that the welding speed

influences the basic weld dimensions. In Fig. 4 there are diagrammatically shown the characteristic

dimensions which were analyzed in detail. Values of the characteristic dimensions of the laser

welds in relation to the welding speed were compiled in the form of graphs in Fig. 4.

Dariusz Skibicki 135

7

6

Dimenison, mm 5

C D B

4

3

A

2 B

C

1

D

0

3,6 4,0 4,4 4,8 5,2 5,6 6,0

Welding speed, m/min

Analysis of the given microsections of the laser welds makes it valid to state that the welding speed

has its impact mainly on the weld width. As expected, along with the increase of the speed the

width of the weld decreases. At the speed of 6m/min the weld width is approx. 30% smaller

compared to the weld width obtained with the speed of 3,6m/min, which simultaneously causes the

heat affected zone to shrink, alongside with the possibility of crack initiations characterized to that

area.

Measurements of microhardness were taken on the microsections (both on the weld itself and in the

heat affected zone) using the Hanneman method with the help of a Neophot 2 microscope. The

spots which were measured for microhardness are diagrammatically shown on the section below in

Fig. 5.

A BC D

67

Microhardness, uHV0,1

62

57

52

47

3,6 4,0 4,4

42

4,8 5,2 5,6

37

6,0

32 1 2 3 4

A B C D

Analyzing the results of the obtained measurements makes it possible to state that hardness

increases along with the approaching the center of the weld. It is hard to assess the effect of the

welding speed on the weld hardness.

During the static tests the samples were being overloaded with the speed of the piston feed of the

tester machine, which equaled 0,1mm/s. The elongations of the samples were measured using an

extensometer with the base of 12,5mm and the measuring range of ±40%. The measurements were

taken at the temperature of 21oC. The tests were being conducted up to the point of the sample split

within the measuring range of the extensometer. During the static elongation test the temporary

values of load force affecting the sample were registered and its elongation.

136 Fatigue Failure and Fracture Mechanics

For the purpose of fatigue tests only two test samples were used which were taken out of the

pipes that constitute the two ends of the welding speed range (v=3,6m/min and v=6m/min). The test

samples were subjected to variable tensile stress. The parameters of the tensile stress were compiled

in Table 1. and in Fig. 6. The fatigue tests were conducted using the Instron 8501 tester machine

equipped with a force gauge head of the ±10 kN measuring range. The experiment was conducted at

the temperature of 21oC. Stress frequency f equaled 15Hz at the time.

Level Stress

σa

No. σmax σmin σa ∆σa

σmax

MPa MPa MPa MPa

1 70 0 35 7

σm

2 80 0 40 80 σmin=0 t

3 90 0 45 90

a) b)

Fig. 6. Stress program: a) stress parameters, b) stress graph

The criterion assumed for ending the fatigue tests was a crack of the test sample. The strains of the

samples were measured using a dynamic extensometer with the range of ±1mm and the base of

10mm. During the experiment the tests samples were secured in hydraulic clamps with a regulated

force of the grip.

3.1. The impact of the plate strengthening on the static properties

All the test samples (with and without a weld) fractured in the area of the extensometer’s

measuring base. The results of the static tensile stress test is shown in Fig. 7 in the form of graphs

on the coordinate system: test sample elongation ε − stress σ. The stresses in the test sample

subjected to tensile load were calculated dividing the temporary values of load force by the initial

cross-sectional area of the sample. The results are grouped in Table 2.

Test sample no.3.0

120 test of the sample from the

Original material

100

aluminum plate

Stress, MPa

60 A12,5mm Rm RP0,2

40 [%] [MPa] [MPa]

20

Plate sample 29 112 70

0 Pipe sample 38 117 85

0 5 10 15 20 25 30 35 40 45

Strain, %

on the strength parameters

After examining the graphs in Fig. 7 it can be stated that, as expected, the process of wrapping the

polyethylene pipe with a layer of an aluminum plate increases the pipe’s strength parameters. The

phenomenon of the strength parameters’ increase proceeding the earlier plastic deformations is

Dariusz Skibicki 137

related to the so-called material strengthening phenomenon. The graphs in Fig. 7 show that

performing the plastic strain process of the aluminum plate during the wrapping of the polyethylene

pipe causes an increase in tensile strength Rm and yield strength Rp02. The quantitive changes of the

listed parameters proceeding the wrapping process can be found in Table 2.

3.2. The impact of the welding speed on the static properties of the weld joints

Regardless of the welding speed the fracture of all the test samples occurred always in the laser

weld area. In Fig. 8 there are illustrated example graphs of tensile tests of samples with the welds

obtained at various welding speeds. For the purpose of comparison, there is also another graph of a

tensile test of the sample without the weld. The graphs depicted in Fig. 8 were made using the basic

strength parameters of the studied material. The results of the calculations assembled in Table 3.

120

Without the weld

3.6

4.0

100

4.4

4.8

80

5.2

5.6

Stress, MPa

60

6.0

40

20

0

0 10 20 30 40

Strain, %

Fig. 8. Static tensile tests’ graphs of the welded samples at various speeds

Item v Defined parameters

no. m/min A12,5 [%] Rm [MPa] RP0,2 [MPa]

1 Without the

weld 38,1 117 85

2 3,6 25,4 122 92

3 4,0 23,5 122 95

4 4,4 23,1 124 97

5 4,8 21,3 124 96

6 5,2 20,8 123 97

7 5,6 19,3 122 95

8 6,0 18,7 121 94

On the basis of the obtained results it can be stated that the weld presence, regardless of the welding

speed, always caused a slight increase in strength parameters (Rm and Rp02) compared to the test

samples without the weld.

138 Fatigue Failure and Fracture Mechanics

125 without

the weld

3,6 4 4,4 4,8 5,2 5,6

115 6

original

105 material

Stress, MPa

95 without

the weld 4 4,4 4,8 5,2 5,6 6

85 3,6

Rp0.2

75 original Rm

material

65 2,8

Comparative analysis of tensile strength Rm and yield strength Rp02 illustrated in Fig.9 allows it to

state that the welding speed has little effect on the mentioned parameters. As expected, the presence

of the weld has its impact on the total elongation of the test samples up to the point of fracture.

Regardless of the pipe’s state, elongation A12,5 of the test samples containing the welds is

considerably lower than those samples without the welds or made out of the original material. In

Fig. 10 there is shown the impact of the welding speed on elongation A12,5 up to the moment of

fracture.

without the

40 weld

35

30

Strain, %

25 original

material 3,6

20 4 4,4

4,8 5,2

15 5,6 6

10 2,8

It can be concluded from the graph that the welding speed has little effect on the total elongation.

The obtained results of elongation A12,5 are very similar for all the given speeds.

During the fatigue experiment the test samples always fractured outside of the weld.

No. Number of cycles until fracture N

Stress σmax Speed v=3,6m/min Speed 6 m/min

1 70 450000 593253

2 80 301127 481281

3 90 133217 138756

Dariusz Skibicki 139

After an analysis of the obtained results in Table 4 it can be stated that the laser welding speed

has an impact on the fatigue life. In order to make the analysis easier the results from Table 4. are

also depicted in Fig. 11 in the form of fatigue graphs on the semi-logarithmic coordinate system:

stress σmax – number of cycles N. The fatigue results were approximated using the following

equation:

b- y-intercept of simple regression for the fatigue graph.

95 v, m/min a b

3,6 -0,1973 934,1

90 6 -0,1053 314,5

Stress σmax MPa

85

80 v=6 m/min

V=3,6 m/min

75

70

65

60

100000 1000000 10000000

Number of cycles

Fig. 11. The impact of the welding speed on the fatigue life of the weld joints

After an analysis of the results it can be stated that the weld joint obtained at the speed of v=6m/min

has higher fatigue life compared to that joint performed with the speed of v=3,6 m/min. The degree

of the impact of the speed is dependent on the level of stress σmax. It is slight in the areas of the

biggest stresses and increases along with the decrease of stresses. On the level of stresses

σmax=70MPa the fatigue life of the weld joints obtained at the speed of 6m/min is approx. 100%

greater compared to those joints at the speed of 3,6m/min.

4. Conclusions

The results obtained during the laboratory experiment of the multilayer pipes, which consist

mainly of cross-linked polyethylene and aluminum plate 8006 make it possible to formulate the

following general conclusions:

- an increase in the welding speed while retaining constant laser power causes a considerable

decrease in the crosswise dimensions of the weld, the almost double increase in the speed

causes an approx. 30% decrease of the dimensions, which has its effect on the size of the heat

affected zone in the welded object,

- a decrease in microhardness of approx. 8% can be observed in the weld obtained at the speed of

3,6m/min compared to the weld at 6m/min, which is a result of a smaller amount of energy that

is distributed to the weld during the welding and it ultimately lowers the temperature of the

welding process itself,

- the aluminum plate during the process of its wrapping and pressing to the polyethylene pipe

strengthens itself which causes an increase in strength parameters,

- the presence of the weld, regardless of the welding speed, always caused an increase in

strength parameters (Rm and Rp02) compared to the test samples without the weld,

140 Fatigue Failure and Fracture Mechanics

- regardless of pipe’s state, elongation A12,5 of the test samples containing the weld was

significantly lesser than the elongation of the samples without the weld and those from the

original material

- the welding speed has little impact on tensile strength Rm, yield strength Rp02 and elongation

A12,5,,

- the impact of the speed on the fatigue life depends on the level of stress σmax, it is slight in the

area of the biggest stresses and increases along with a decrease in stresses.

The obtained results of the conducted tests of the polyethylene–aluminum multilayer pipes

illustrated that the welding speed of the aluminum layer of the polyethylene pipe has little impact on

the tensile strength of the joints. The welding speed considerably increases the fatigue strength on

the low levels of stresses σmax.

References:

[1] Rutkiewicz. A, Pomiar szybkości przenikania tlenu przez ścianki rur z tworzyw sztucznych,

Prace Instytutu Techniki Budowlanej, R38, (2009).

[2] PN-74/H04327- Badanie metali na zmęczenie. Próba osiowego rozciągania ściskania przy

stałym cyklu obciążeń zewnętrznych

CHAPTER 4:

Temperature and Thermo-Mechanical Fatigue

© (2012) Trans Tech Publications, Switzerland

doi:10.4028/www.scientific.net/MSF.726.143

under thermo-mechanical fatigue conditions

Jerzy Okrajni 1, a, Grzegorz Junak 1, b

1

Department of Materials Technology, Silesian University of Technology,

Krasinskiego 8, 40-019 Katowice, Poland

a

jerzy.okrajni@polsl.pl, b grzegorz.junak@polsl.pl

Abstract.

The paper focuses on the development of a mathematical representation of deformation

characteristics under the conditions of an elevated changeable temperature and mechanical loads.

The method proposed in the paper is based on the use of characteristics determined in low-cycle

fatigue tests at constant temperatures. Hysteresis loops reflecting the behaviour of a material under

the conditions of low-cycle loads at an elevated temperature were primarily used. The effect of

relaxation on the course of the hysteresis loop was taken into account. The steady state of the

material is referred to in the proposed representation.

A calculation algorithm was developed to enable the determination of the stress value for

subsequent increases of strain over time. The results obtained were compared with experimentally

determined characteristics.

Introduction

The course of the deformation process in the conditions of simultaneous action of a variable

temperature and of mechanical and thermal loads has been described in many papers concerning

solid body mechanics. Some papers attempted to generalise the description of the behaviour of

materials through constitutive equations for complex stress states that take into account rheological

phenomena and dependencies of material properties on the temperature, including Leimatre and

Chaboche models [1-3] which are commonly used to describe the properties of materials applied for

operation at an elevated temperature or, for example the model of Kichenin and co-authors [3] or

the one developed by Figiel and Günter [4]. Therefore, this paper naturally is not an attempt at such

generalisation. At the current stage, it only attempts to describe the behaviour of a material

subjected to thermo-mechanical fatigue, while taking into account a number of constraints. This is

because it assumes that the deformation process occurs in a steady state and that the reaching of the

steady state in low-cycle fatigue conditions is tantamount to reaching such a state in the case of

thermo-mechanical fatigue. The discussion is referred to the uniaxial state of stress.

The proposed method of describing dependences between stress and strain under thermo-

mechanical fatigue conditions is based on the use of characteristics determined during tests of low-

cycle fatigue at constant temperatures [5,6]. Basic dependences used to describe the mechanical

behaviour of a material in such tests are characteristics in the form of hysteresis loops whose

courses depend on the value of temperature. The shape of the loops and their parameters may

change in successive load cycles as a result of cyclic strengthening or weakening of the material,

usually reaching a steady sate that covers a majority of load cycles until failure of the tested

specimen. At the current stage, the discussion refers to a steady state. In the developed model, the

course of the deformation process is described based on the stabilised hysteresis loops which had

been experimentally determined. Function σ`=f(ε`, T), ε`∈(0, ∆ε) describing the course of a

hysteresis loop in the state of saturation (Fig. 1a) was introduced.

144 Fatigue Failure and Fracture Mechanics

a) b)

Fig. 1. The courses of steady hysteresis loops, depending on the temperature. schematic

representation –a) and a diagram of a hysteresis loop branch for the method of approximating its

course used in the paper –b)

Function f(ε`, T) can be adopted in various forms. In papers [5,6], the dependence between stress

σ`, deformation ε` and temperature T was adopted in the form of the following equation:

A number of other methods can be proposed to represent the f`(ε’, T) function. The isolating

of elastic and elastic–plastic deformation ranges on the characteristics under consideration seems to

be an appropriate method. In this case, the curve describing the dependence between stress, strain

and temperature will be comprised of two sections. A linear dependence between strain and stress is

assumed in the range of values below the strain corresponding to the transition between the elastic

R (T )

and elastic-plastic range (Fig. 1b). Above the value of ε ′ = e , strengthening of the material

E (T )

occurs, which can be described for example with an exponential dependence. Consequently, we

receive the following function:

R (T )

E( T )ε ' for ε ' < e

E( T )

f ( ε ' ,T ) = n( T ) (2)

R ( T ) + C( T ) ε ' − Re ( T ) , for ε '

≥

Re ( T )

e E( T )

E( T )

Re is in this case a conventionally treated “yield point” separating the elastic from the elastic-

plastic range (Fig. 1b).

Dariusz Skibicki 145

Other dependences describing the course of the f`(ε`, T) curve above the yield point Re(T) can

also be adopted in the range of elastic-plastic deformations. One can, for example, apply the

previously used arctan(x) function. In this way, the following equation can be obtained:

Re ( T )

E ( T )ε '

for ε '

<

E( T )

f ( ε ,T ) =

'

(3)

Re ( T ) + B( T ) arctan ε ' − Re ( T ) m( T ), for ε ' ≥ Re ( T )

E( T ) E( T )

The values B(T), C(T), n(T), m(T), E(T) and Re(T), which are dependent on temperature,

are determined based on low-cycle fatigue tests. Figure 2 presents examples of σ`(ε`)

characteristics determined experimentally and via approximation with the use of dependences (3).

800

700

600

500

STRESS, MPa

400

300

200

100

0

0 0.002 0.004 0.006 0.008

STRAIN

Fig. 2. Experimentally determined (E) dependences between strain ε’ and stress σ’ for P91 steel

and courses corresponding to them, determined based on a model presentation (M)

After obtaining functions (2) or (3), the method described in papers [5,6] can be used to characterise

the course of dependences between stresses and strains in the process of thermo-mechanical fatigue.

Functional dependences describing the course of hysteresis loop branches with increasing (4) and

decreasing (5) strain were substantiated in these papers:

1 1

σ= f [(ε − ε R ),T ]− f [(ε − ε R ),TR ]+ σ R 0 (4)

2 2

f [ε − ε C ,T ]+ f [ε − ε C ,TC ]+ σ C 0

1 1

σ =− (5)

2 2

146 Fatigue Failure and Fracture Mechanics

TR is the temperature at the beginning moment of the part of the cycle with increasing strain. TC is

the temperature at the beginning moment of the part of the cycle with decreasing strain.

Constants ε R ,ε C ,σ RO ,σ CO are coordinates of the points of hysteresis loop (Fig. 3a). Figure 3

presents a diagram of a loop which can be obtained in thermo-mechanical fatigue tests (Fig. 3a) and

the characteristics of this type of fatigue process which was experimentally determined for the P91

steel (Fig. 3b). This characteristics was compared to a model presentation which uses equations (3)-

(5). Compliance of the course of both characteristics was achieved, which seems sufficient for

technical applications.

MODEL EXPERIMENT

500

400

300

200

STRESS, MPa

100

-100

-200

-300

-0.006 -0.004 -0.002 0 0.002 0.004 0.006

STRAIN

a) b)

– a) and a comparison of an experimentally determined hysteresis loop with a hysteresis loop

determined based on a model representation –b); steel P91 (Tmax=650oC, Tmin=200oC)

The issue becomes much more complex when periods of a steady value of total mechanical

strain occur in the deformation process. Under fatigue conditions at elevated temperatures,

relaxation of stresses occurs during deformation induced by changes of external factors, especially

in periods with a fixed strain value. After a period of relaxation at a fixed strain value, another

period of its increase or decrease occurs, which compensates for the effects of the relaxation

process. An effect of the return of the material to the state from before the rheological processes can

be observed in metal alloys. After a short return period, the material is deformed once again in

accordance with the course of the hysteresis loop characteristic for the steady state (Fig. 4). When

describing the course of the hysteresis loop under such conditions, a mathematical representation of

the deformation process, both during relaxation and under the conditions of “return” to the state

preceding this process, is necessary.

Dariusz Skibicki 147

400

0.006

300

0.004

200

0.002

100

STRESS, MPa

STRAIN

0 0

-100

-0.002

-200

-0.004

-300

-0.006

-400 2500 2700 2900 3100 3300 3500

-0.006 -0.004 -0.002 0 0.002 0.004 0.006

TIME, s

STRAIN

a) b)

Fig. 4. A hysteresis loop determined in an isothermal low-cycle fatigue test (T=550oC) with the

total strain under control –a ) whose course over time is shown in the figure –b); steel P91

Having adopted a power law creep, the creep strain is expressed [7] with the following dependence:

ε c = Kσ n tm (6)

By applying the ageing theory and assuming a constant creep rate for the relaxation process, the

following equation is obtained:

1 1 1

t= −

KE (n − 1) σ n −1 σ on −1 (7)

Where E is the elasticity modulus and σo is the initial stress value. Using equation (7), the increase

of stress caused by relaxation may be written down as follows:

dσ r = − KEσ n dt (8)

In this case, constant K and elasticity modulus E should be treated as constants depending on

temperature. Due to the sign of stress, equation (8) is valid for odd integer values of exponent n. In

or sgn(σ )σ .

n−1 n

remaining cases, σ should be replaced with expression σ σ

n

The paper was an attempt at expanding the scope of the application of dependence (8) to the

range of elastic-plastic deformations. Product KE was treated as a constant depending on

temperature S (T ) = K (T )E (T ) . The overall effect of relaxation in time t can be calculated by

adding increases of stresses in individual moments of time. Consequently, we receive:

σ r = −∫t S (T )σ n dt (9)

148 Fatigue Failure and Fracture Mechanics

While treating the relaxation process as “reversible”, a function expressing the effects of the

“return" in the form of changes of the value of the total effect of relaxation σr caused by the strain

was introduced in the paper. It was assumed that the intensity of the “return” measured using the

increase of stress to the increase of strain ratio depends on stress value σr and temperature.

The increase of stress caused by the effects of the “return” was expressed in the first approximation

with the following dependence:

dε

dσ r = M (T )σ r dε = M [T (t )]σ r (t ) dt (10)

dt

Consequently, we receive the following equation:

dε

σ r = − ∫t S [T (t )]σ n − M [T (t )]σ r (t ) dt (11)

dt

Taking into consideration dependences (4) and (5) and equation (11), a recurrent computational

algorithm which allows calculating stress values for next strain increases over time was developed.

In this algorithm, stresses in next calculation stages are determined as a sum of the results of strain

increase, determined from (4) and (5) dependences, and stress changes caused by relaxation - σr

(11), while taking into account the effects of the “relaxation return”. An example of thus determined

hysteresis loop is presented in Figure 5a which shows an experimentally determined loop and a loop

determined based on the developed model representation. Constants S and M were assumed to be

dependent on temperature and were determined by trial and elimination based on conducted

experiments. Diagrams shown in Figure 5a correspond to the characteristics of changes in stresses

as a function of time, shown in Figure 5b. Hysteresis loops determined experimentally and those

determined based on a model representation of the characteristics of changes in stresses as a

function of time show the conformity of the course, both in the periods in which no discernible

influence of rheological processes was observed and during relaxation.

MODE L E X P E R IME NT

400

300

300

200

200

S T R E S S , MP a

100 100

S T R E S S , MP a

0 0

-100 -100

-200

-200

-300

-300

-400

-400 0 100 200 300 400

-0.006 -0.004 -0.002 0 0.002 0.004 0.006 TIME , s

S TR AIN

a) b)

Fig. 5. A hysteresis loop determined in an isothermal low-cycle fatigue test (T=500oC) with total

strain control and a hysteresis loop determined based on a corresponding model representation –a )

for those whose courses of stress over time are shown in the figure –b); steel P91

Dariusz Skibicki 149

Conclusion

The aim of the tests was to develop a method of the mathematical representation of deformation

characteristics under the conditions of an elevated temperature and mechanical loads. At the current

stage, calculation methodology was developed and verified for selected cases. The main focus was

on representing the relaxation phenomenon. Characteristics shown in Figure 5 refer to fatigue under

the conditions of action of constant temperature and variable in time strain whose cycle contained a

time interval during which strain remained constant. Dependences assumed in the paper require

further verification, to check in particular whether constants determined for particular cases of

relaxation itself and isothermal low-cycle fatigue without the participation of rheological processes

may serve as a basis for the generalisation of the description of material behaviour under the

conditions of cyclic deformation at variable temperature and mechanical strain, while taking into

consideration time intervals of relaxation and creep. Currently this type of tests is performed while

taking into account various materials and cycles of thermo-mechanical loads and taking into

consideration the specificity of operation of devices used under the conditions of mechanical and

thermal loadings [8-10], in particular currently developed power engineering devices with enhances

operating parameters in the case of which phenomena of fatigue induced by cyclical changes in

temperature become more and more important.

The results presented in this paper were obtained from research work co-financed by the National

Centre of Research and Development in the framework of Contract SP/E/1/67484/10 - "Strategic

Research Programme - Advanced Technologies for obtaining energy: Development of a technology

for highly efficient zero-emission coal-fired Power units integrated with CO2 cap”

References

[1] J. Lemaitre, J.-L. Chaboche, Mechanics of solid materials, Cambridge University Press,

Cambridge 1990.

[2] J.-L. Chaboche, Cyclic viscoplastic equations part I: e thermodynamically consistent

formulation, Journal of Applied. Mech., 1993, vol. 60, pp. 813-821.

[3] J. Kichenin, K. Dang Van, K. Boytard, Finite element simulation of a new two-dissipative

mechanisms model for bulk medium-density polyethylene, J. Mater. Sc 31(1996), pp.1653-

1661.

[4] Ł. Figiel, B. Günter, Modelling the high-temperature longitudinal fatigue behaviour of metal

matrix composites (SiC/Ti-6242): Nonlinear time-dependent matrix behaviour, International

Journal of Fatigue, v. 30, issue 2, January 2008, pp. 268-276.

[5] J. Okrajni, G. Junak, A. Marek, Modelling of the deformation process under thermo-

mechanical fatigue conditions, International Journal of Fatigue, v. 30, issue 2, January 2008,

pp. 324-329.

[6] J. Okrajni, G. Junak, A. Marek, Description of the deformation process under thermo-

mechanical fatigue, Journal of Achievements in Materials and Manufacturing Engineering,

vol. 21 issue 2, 2007, 15-24.

[7] G.A. Webster, R.A. Ainsworth, High Temperature Component Life Assessment, Chapman &

Hall, London 1994.

[8] J. Okrajni, W. Essler, Computer models of steam pipeline components in the evaluation of

their local strength, Journal of Achievements in Materials and Manufacturing Engineering,

vol. 39 issue 1, 2010, 71-78.

[9] J. Okrajni, Thermo-mechanical conditions of power plant components, Journal of

Achievements in Materials and Manufacturing Engineering, vol. 33 issue 1, 2009, 53-61.

[10] J. Okrajni, K. Mutwil, M. Cieśla, Steam pipelines’ effort and durability. Journal of

Achievements in Materials and Manufacturing Engineering, vol. 22, issues 2, June 2007, pp.

63 – 66.

© (2012) Trans Tech Publications, Switzerland

doi:10.4028/www.scientific.net/MSF.726.150

of martensitic cast steel

1,2

University of Technology and Life Sciences in Bydgoszcz,

Faculty of Mechanical Engineering,

A. Prof. S. Kaliskiego 7, 85-789 Bydgoszcz

tel.: 48 52 340-82-64, fax: 48 52 340-82-71

a b

mrozinski.stanislaw@utp.edu.pl, rskocki@gmail.com

Abstract. This paper attempts to describe changes of the cyclic properties of martensitic cast steel

in the function of the number of loading cycles under temperature of 25 and 600oC. The cyclic

properties were described by means of three hysteresis loop parameters: stress amplitude σa, plastic

strain amplitude εap, plastic strain energy ∆Wpl. It was stated that martensitic cast steel always

undergoes clear softening which is independent of the temperature and level of total strain.

Introduction

The issue of this paper concerns the analytical description of changes of the martensitic cast steel

cyclic properties namely the changes of the hysteresis loop parameters in the function of the number

of loading cycles. The problem complicates significantly when the course of the changes aside from

the mechanical loading is influenced by the changeable temperature [1,2]. At elevated temperature

the range of changes of the hysteresis loop parameters is wider than at room temperature [3,4,5].

Due to this reason more propositions are made considering the changes of cyclic properties

influenced by the temperature in calculations of fatigue life [6,7].

Basic aim of this work is to determine the influence of the temperature on changes of hysteresis

loop parameters in the function of the number of loading cycles.

Descriptions of tests

Specimens for the tests were made of GX12CrMoVNbN9-1 martensitic cast steel. The shape and

dimensions of the specimens were in agreement with the standard [8] (Fig. 1).

φ 8 h6

0.32

φ 16.5 h6

M 20

φ 16

R10 R10

2

15 7

2 27

105

The fatigue tests were preceded by carrying out the static tensile tests. The specimens used in the

tests are shown in Fig. 1. The specimens underwent increasing loading with the rate of machine

displacement speed of 0.05 mm/s. Specimen’s elongation was measured by a 12.5 mm gauge length

Dariusz Skibicki 151

axial extensometer with measuring range of 3.75 mm. The static tensile tests were carried out under

the temperature of 25oC and 600 oC. During these tests momentary loading forces and elongation of

the specimen were recorded. After analysing the static tensile tests five levels of total strain εac were

accepted in low cycle tests mainly: εac=0,25; 0,30; 0,35; 0,50; 0,60.

LCF tests were conducted under controlled total strain εac =const. The same procedure of measuring

strain was employed in static tensile test. Test temperature of 25oC and 600 oC and frequency of 0.2

Hz were employed. Accepted sampling frequency of force signal and strain signal allowed to

describe loading cycles with set of 200 points. As the end criterion of the fatigue test, the

deformation of hysteresis loop (during semi cycle of compression) is accepted. During the tests

momentary values of loading force and strain for selected loading cycles were recorded.

The examined material cyclically softened independently of the temperature. To analyse this

process three hysteresis loop parameters were assumed namely: stress amplitude σa, plastic strain

amplitude εap, plastic strain energy ∆Wpl (Fig. 2).

σ

+σ

∆Wpl

2σa=∆σ

−εa + εa ε

-σ

∆εap εae

2εac= ∆εac

Fig. 2. Hysteresis loop with its basic parameters

Fig. 3 presents three graphs of the analysed parameters in the function of the number of loading

cycles for the two levels of total strain εac=0,25 and εac=0,60%. According to the results it was

stated that the temperature clearly influences fatigue life and the course of the analysed parameters.

152 Fatigue Failure and Fracture Mechanics

εac=0,25%

500 500

400 400

300 300

250 250

T=600 ºC

200 200 T=600 ºC

150

150

0 200 400 600 800 1000

0 10000 20000 30000 40000

n

n

a) b)

0,0014 εap, mm/mm εac=0,25% 0,0050 εap , mm/mm εac=0,6%

0,0012 0,0045

0,0040 T=600 ºC

T=600 ºC

0,001

0,0035

0,0008 0,0030 T=25 ºC

0,0006 0,0025

0,0020

0,0004 T=25 ºC

0,0015

0,0002 0,0010

0 0,0005

0 10000 20000 30000 40000 0 200 400 600 800 1000

n n

c) d)

3

∆ Wpl, MJ/m3 ε ac =0,25 %

6 ∆ Wpl, MJ/m εac=0,6 %

0, 8

5

T=600 ºC

T=25 ºC

0, 7

4

0, 6

T=25 ºC 3 T=600 ºC

0, 5

2

0 10000 20000 30000 40000

0 200 400 600 800 1000

n

n

e) f)

Fig. 3. Changes of the hysteresis loop parameters (σa, εap, ∆Wpl) depending on the total strain

level and the temperature: a, c, e) εac=0,25%; b, d, f) εac =0,6%

On the basis of the presented graphs it can be stated that for each level of total strain there can be

distinguished stages with a different softening speed. Fig. 4 presents the course of changes of stress

amplitude σa in the function of loading cycles for the level of total strain of εac =0,6% and the

temperature of T=20 oC. Three distinctive stages can be observed (A, B, C).

Dariusz Skibicki 153

530 σa , MPa

510

490 σa =aσ⋅ n +b

470

450

430

A B C

410

390 A – changeable softening speed

B – constant softening speed

370 C – changeable softening speed

350

0 500 1000

n

Fig. 4. Consecutive stages martensitic cast steel softening

In stage “A” cast steel clearly softens. The speed of softening decreases with the increase of the

number of loading cycles. In stage “B” the speed of softening is constant. In case of tested cast steel

it was the longest stage, independently on the level of total strain and the temperature of the test.

In the last stage “C” the speed of softening increases. Crack initiation takes place here and

eventually leads to the fatigue failure. It should be noted that the stages mentioned above can be

clearly observed for the stress amplitude σa curve. It is more difficult to distinguish these stages for

the plastic strain amplitude εap and plastic strain energy ∆Wpl curve, particularly in the temperature

of 600oC. This paper attempts to analyse the influence of the temperature on the speed of softening

only for the one stage (the longest stage “B”). The hysteresis loop parameters were approximated by

the linear regression line with regression equations presented in Fig. 4. To illustrate this

approximation in Fig. 5 there were shown the courses of changes of the stress amplitude σa in the

temperature of 600oC with all stages and also the courses of this parameter constrained only to the

second stage “B”.

300 σa, MPa 300 σa, MPa

280 280 σa=aσ⋅n +b

260 εac=0,6% 260

240 εac=0,5% 240 εac=0,6%

εac=0,35% εac=0,5%

220 εac=0,3% 220 εac=0,35%

200 200 εac=0,3%

εac=0,25%

εac=0,25%

180 180

160 160

0 2000 4000 10 100 1000 10000

n log n

a) b)

Fig. 5. Changes of stress amplitude σa in the function of the number of loading cycles in

the temperature of 600o: a) all stages, b) stage “B” – constant softening speed

154 Fatigue Failure and Fracture Mechanics

In Fig. 6 there were collected slopes of the regression lines aσ , aε , a∆W according to the level of total

strain εac in the temperature of a) T=25 ºC b) T=600 ºC

0,1 - aε 0,1

- aε

- a∆W - a∆W

- aσ - aσ

0,05

aσ, aε, a∆W,

0,05

0 0

-0,05

-0,05 0,25 0,3 0,35 0,5 0,6

0,25 0,3 0,35 0,5 0,6

εac, % -0,1

-0,1 εac, %

a) b)

Fig. 6. Slopes of the regression line describing hysteresis loop parameters according to the

level of total strain a) T=25 ºC b) T=600 ºC

The analysis of the slopes indicates that the course of tested cyclic properties (stress amplitude σa,

plastic strain amplitude εap, plastic strain energy ∆Wpl ) is significantly influenced by the level of

total strain εac and the temperature. Increasing the speed of softening depending on the temperature

can be particularly noticed in the stress amplitude σa curve. Obtained tests results concerning the

larger changes in cyclic properties at elevated temperatures confirm reports shown in literature

[9,10,11].

Conclusions

On the basis of carried out LCF tests it was stated that martensitic cast steel is a material which

significantly softens independently of the temperature. The softening process occurs in all realized

levels of total strain. Three characteristic stages with different speed of softening can be determined.

Independently of the total strain level there is always a stage with a constant softening speed (Fig. 4

stage “B”). In this stage changes of the cyclic properties of the martensitic cast steel (σa, εap, ∆Wpl)

in the function of loading cycles can be approximated by the linear regression lines. The analysis of

the linear regression line slopes aσ, aε, aW leads to the conclusion that the speed of softening for all

stages of softening process (A, B, C) is significantly influenced by the temperature and level o total

strain. It can be stated that plastic strain energy ∆Wpl is the least sensitive hysteresis loop parameter

to changes of the cyclic properties independently of the temperature and level of total strain.

References

[1] Collins J.A., Failure of Materials in Mechanical Design, Analysis, Prediction, Prevention.

John Wiley & Sons, New York 1993.

[2] Kocańda S., Kocańda A., Niskocyklowa wytrzymałość zmęczeniowa metali. PWN Warszawa

1989.

[3] Vani Shankar, Valerij Bauer, R. Sandya, M.D. Mathew, H.-J. Christ.: Low Cycle fatigue and

thermo-mechanical fatigue behavior of modified 9Cr-1Mo ferritic steel at elevated

temperatures, Journal of Nuclear Materials vol. 420 (2012), pp. 23-30.

Dariusz Skibicki 155

[4] Mroziński S., Skocki R., Softening of Martensitic Cast Steel, Journal of Polish CIMAC.

Volume 5 No 3, 2011, pp. 173-180.

[5] Mroziński S., Golański G.: Low cycle fatigue of GX12CrMoVNbN9-1 cast steel at elevated

temperature. Journal of Achievements in Materials and Manufacturing Engineering. Vol 49

ISSUE 1, November 2011, pp. 7-16.

[6] Nagode M., Hack M.: An online algorithm for temperature influenced fatigue life estimation:

stress-life approach, International Journal of Fatigue 26 (2004), pp.163-171.

[7] Nagode M., Zingsheim M.: An online algorithm for temperature influenced fatigue life

estimation: strain-life approach, International Journal of Fatigue 26 (2004), pp. 155-161.

[8] ASTM E606-92: Standard Practice for Strain -Controlled Fatigue Testing.

[9] Mroziński S., Golański G.: Fatigue life of GX12CrMoVNbN9-1 cast steel in the energy-based

approach. Advanced Materials Research. Vol 396-398 (2012), pp. 446-449.

[10] Kaae J.L.: High-temperature low-cycle fatigue of Alloy 800H. International Journal of Fatigue

31 (2009), pp. 332–340.

[11] Okrajni J., M. Cieśla M., Mutwil K.: Power plant component life assessment, Inżynieria

Materiałowa 1 (2005), 15 – 20.

© (2012) Trans Tech Publications, Switzerland

doi:10.4028/www.scientific.net/MSF.726.156

of Mini-Specimens

Adam Lipski1,a, Dariusz Boroński1,b

1

The University of Technology and Life Sciences in Bydgoszcz,

Faculty of Mechanical Engineering, Al. Prof. S. Kaliskiego 7, 85-789 Bydgoszcz, Poland

a

Adam.Lipski@utp.edu.pl (corresponding author), bDariusz.Boronski@utp.edu.pl

Abstract. This paper presents sample applications of passive infrared thermography for research

on temperature changes of mini-specimens resulting from monotonously increasing or cyclically

variable mechanical load. The MFS system developed in the Department of Machine Design

at the University of Technology and Life Sciences in Bydgoszcz (Poland) and designed for testing

mechanical properties of microelements were used for tests. The MFS system ensures nanometric

measurement accuracy of many static and fatigue-related material properties, including, i.a., static

tension curves, cyclic strain curves, fatigue life curves as a function of force, stress and strain.

Measurements of the mini-specimens temperature were performed using thermographic camera

equipped with microscope lens. The tests have shown that research on the passive infrared

thermography may be successfully applied for determining strength properties of materials in micro

scale. The used research instrumentation is characterized by sufficient sensitivity and resolution

(camera with the microscope lens), while the MFS system ensures accurate load and position control.

Introduction

The fast development of micro- and nanotechnology as well as material engineering and welding

engineering that has recently been observed allows to use completely new design solutions based on

advanced structural materials and new methods of their joining. However, those possibilities are

often limited by lack of knowledge on specific properties of those materials such as mechanical

properties in their broad meaning.

One of the main properties of technical objects that may have significant effect on the reliability

and safety of the object operation is its ability to operate under variable load, including, in

particular, cyclically variable load [1, 2, 3, 4].

Thanks to fast development of material engineering and new manufacturing methods, technical

objects are more and more often developed using solutions requiring use of new methods of fatigue

analysis, including methods based on local approach. A significant limitation here is often

insufficient knowledge on local material properties, including, in particular, fatigue properties. One

of the reasons for such situation is high variability resulting mainly from applied material

technologies or the object manufacturing processes, generating areas of different mechanical

properties located in the close vicinity. Good examples of such structures are e.g. laser-welded

joints or friction stir welded (FSW) joints, for which size of zones with different properties often

does not exceed several decimal fractions of millimetre. Thus it is necessary to perform fatigue tests

using very small specimens, i.e. so called mini-specimens.

Strength properties tests (both monotonic and cyclic) can be successfully performed using

infrared thermography. Examples of its application are provided e.g. in papers [5, 6, 7]. However,

they concern macro scale tests in the broad sense. The professional literature seems not to include

any articles concerning that problem as regards micro-scale specimens or structural components.

This paper presents sample applications of passive infrared thermography for research on

temperature changes of mini-specimens resulting from monotonously increasing or cyclically

variable mechanical load. The assessment of the possibility to perform research in that scope is the

precondition for further activities concerning passive thermography applied mainly for quick

determining of strength properties of materials under cyclic load in micro scale.

Dariusz Skibicki 157

Test description

Test station. For those tests, the researchers used the MFS system designed for testing mechanical

properties of microelements. The system was developed in the Department of Machine Design at

the University of Technology and Life Sciences in Bydgoszcz as part of the research and

development project funded by The Polish National Centre for Research and Development. The

MFS system consists, among others, of such components like: nanodrive load system, microdrive

load system, strain measurement system using digital image correlation method (DIC), strain

measurement system using laser grating interferometry (LFI) method, precise object alignment and

fixing systems, computerized power supply and control system.

Overview of the system is shown in Figure 1 next to the drawing of the fatigue test specimen.

Thanks to the applied solutions including double load system based on piezoelectric and micro-

stepper actuator equipped with the precise ball screw, the MFS system ensures nanometric

measurement accuracy of many static and fatigue-related material properties, including, i.a., static

tension curves, cyclic strain curves, fatigue life curves as a function of force, stress and strain. As

the system is equipped with optical-electronic measuring instrumentation, it provides micrometric

accuracy of strain measurement at measuring points.

a) b)

Fig. 1. Overview of MFS system (a) and a mini-specimen compared with the smallest Polish coin (b)

Continuous measurement of the mini-specimen temperature (Fig. 2) was performed during the

research, using CEDIP Silver 420M thermographic camera equipped with G3 microscope lens

(viewing area of about 3.2 mm x 2.6 mm – image pixel size of about 0.01 mm x 0.01 mm) with the

frequency of 100 Hz.

Camera images were transmitted via USB 2.0 interface to PC with appropriate software, where

they were digitally recorded directly on HDD in form of PTW format files.

The following software was installed in the PC computer:

– VirtualCAM allowing two-way communication between PC and the thermographic camera via

USB 2.0 interface,

– CIRRUS Front End which was the user interface of the thermographic camera,

– ALTAIR allowing downloading, storage and advanced processing of thermographic images.

Main parameters of CEDIP Silver 420M thermographic camera include:

– resolution: 320×256 pixels,

– pixel size: 25 µm,

– pixel matrix pitch: 30 µm,

– spectral range: 3.6÷5.0 µm,

– sensitivity: below 20 mK (available: 8 mK),

– recording frequency at maximum resolution and digital image transfer via USB 2.0 interface:

up to 140 Hz (up to 25 kHz at the resolution of 64x8 pixels),

– programmable integration time in the range from 10 µs to 10 ms.

158 Fatigue Failure and Fracture Mechanics

camera

microscope lens

specimen

with the microscope lens

Specimens. Samples for strength properties tests at monotonically and cyclically variable load

(Fig. 3) were cut with laser from 0.2 mm thick sheet plate made of S355J2G3 (18G2A) steel.

a) b)

,5

R0

60°

2,35

0,2

0,3

2

4,3

Fig. 3. Dimensions of mini-specimens used in the research (thickness 0.2 mm) (a) and a sample

thermogram of a mini-specimen tested under monotonically variable load (b)

Monotonic tensile test. Figure 4 shows sample tensile force curve and average temperature curve

of the mini-specimen at the measuring part as the function of time, recorded during the monotonic

tensile test.

The analysis of the obtained average temperature curve of the specimen shows gradual

temperature decrease at the first phase of loading (by about 0,5°C), due to the thermal-elastic effect.

Minimum temperature defines so called thermal-elastic-and-plastic strength σθ. Thermal-elastic

cooling effect prevails up to the value of the yield strength, while once the yield strength

is exceeded, thermal-plastic heating of the material becomes more and more significant. Those two

effects are balanced at the point corresponding to the thermal-elastic-and-plastic strength, i.e. the

heating balances the cooling resulted from the thermal-elastic effect. The shift from temperature

decrease to increase during the test is smooth and corresponds to properties of metals characterized

by absence of explicit yield strength, which is confirmed by the shape of the tensile force curve.

The average specimen temperature change recorded during the test amounted to about 2,81°C.

Dariusz Skibicki 159

Fig. 4. Sample curve of the force and temperature change in the mini-specimen as the function

of time, recorded during the monotonic tensile test

No significant temperature change was observed at the moment of the specimen failure,

as compared to curves recorded for specimens of standard size. This may result from the small size

of the mini-specimen and, as a consequence, low energy released at the moment of sample failure.

Cyclic test. Figure 5 shows fragment (seconds 345 to 355 of the test) of the load force curve

and the average temperature curve of the mini-specimen measuring part as the function of time,

recorded at variable load.

Fig. 5. Sample fragment of the load force curve and temperature change curve of the mini-specimen

as the function of time, recorded during the variable load test

160 Fatigue Failure and Fracture Mechanics

Cyclic variation of the mini-specimen load force with average value of about 6.5 N

and amplitude value of about 10.8 N was assisted by cyclic temperature variation

with the amplitude of about 0.39°C and gradually increasing average temperature value which had

not stabilized until the sample failure moment, when the variation achieved its maximum value

of about 0.86°C. Gradual increase of the average temperature value during the test resulted from

plastic strain causing energy dissipation and heat generation in the specimen, which could be

noticed on the basis of the hysteresis loop (Figure 6).

The frequency of the recorded sample temperature curve corresponds to the frequency

of constant-amplitude variable load of f = 0.5 Hz.

The specimen temperature variation curve is phase shifted with regard to the load curve,

i.e. maximum force value in the load cycle corresponds to the minimum temperature value, while

the minimum force value in the load cycle corresponds to the maximum temperature value.

331 load cycles were performed during the fatigue test until the specimen failure moment.

Summary

The tests discussed above have shown that research on the passive infrared thermography may be

successfully applied for determining strength properties of materials in micro scale. The described

research instrumentation is characterized by sufficient sensitivity and resolution (the thermographic

camera equipped with the microscope lens), while the MFS system ensures accurate load control

and readout of the selected parameters.

The load frequency typical for low-cycle fatigue tests was used in this work. The low load

frequency (f = 0.5 Hz) made possible plastic strain development. The presented method allows

realising load with nanometric measurement accuracy in range of displacement (minimal extension

about 1.7 nm). The measured temperature change for this extension was ca. 0.4°C. This means that

it is possible to determining strength properties of objects in micro scale with measure base even

less than 200 µm by using coupled mechanical and temperature fields. For example the strain

resolution is 0.001 % for 0.17 mm measure base. Thus it is possible to determine the local material

characteristics, e.g. for different zones of welded joints.

Dariusz Skibicki 161

Results of further research on that area will be presented in future conferences dedicated

to fatigue and experimental mechanics.

This research work is financially supported by the Polish state budget for science in 2010-2013

as a research project.

References

[1] S. Kocańda, J. Szala, Basics of the fatigue calculations, PWN, Warsaw, 1997 (in Polish).

[2] J. Szala, Fatigue damage accumulation hypothesis, University of Technology and Agriculture

in Bydgoszcz, 1998 (in Polish).

[3] S. Kocanda, Fatigue Failure of metals (Fatigue and Fracture), Springer, 1978.

[4] J. Schijve, Fatigue of Structures and Materials, Kluwer Academic Publishers, Dordrecht-

Boston-London, 2001.

[5] N. Harwood, W.M. Cummings (eds), Thermoelastic Stress Analysis, IOP Publishing Ltd.,

Bristol, 1991.

[6] G. La Rosa, A. Risitano, Thermographic methodology for rapid determination of the fatigue

limit of materials and mechanical components, Int. J. Fatigue, 22 (2000), pp. 65–73.

[7] A. Lipski, The use of passive infrared thermography for tests of materials and riveted joints

used in aviation industry- selected problems. Part II of the collection of monographs

(edited by J. Szala): Experimental methods in studies of materials and riveted joints used

in aviation industry - selected problems. Institute For Sustainable Technologies –National

Research Institute in Radom, Bydgoszcz-Radom, 2010 (in Polish).

© (2012) Trans Tech Publications, Switzerland

doi:10.4028/www.scientific.net/MSF.726.162

Of The Multiaxial Load - Preliminary Research

Adam Lipski1,a, Dariusz Skibicki1,b

1

University of Technology and Life Sciences in Bydgoszcz,

Faculty of Mechanical Engineering, Al. Prof. S. Kaliskiego 7, 85-789 Bydgoszcz, Poland

a

Adam.Lipski@utp.edu.pl (corresponding author), bDariusz.Skibicki@utp.edu.pl

Abstract. This paper provides the results of research on temperature changes of cylindrical

specimens depending on the pattern of the multiaxial load. The research were made by using

passive infrared thermography. It was found out that the average temperature value is significantly

dependent on the plastic strain energy and that the temperature change amplitude depends

on the nominal normal stress (except for torsion).

Introduction

Many physical phenomena related to mechanical load cause temperature changes. Those changes

can be observed using passive infrared thermography. Many of those phenomena have already been

subject to research (e.g. temperature changes caused by monotonically increasing mechanical load,

discussed in the papers [1] and [2]. However, temperature measuring instrumentation that is usually

used is designed for spot temperature measurement and, moreover, it requires direct contact with

the surface of the tested object, which can influence the temperature of the object and the test result.

Many of those phenomena are also related to surface, hence determining temperature at selected points

of a specimen or a structure is not sufficient. It is necessary to determine distribution of temperature

over the entire surface instead. Temperature distribution recording rate together with the accuracy

of the temperature measuring system is of key importance, particularly for variable load tests.

The above specified conditions for surface temperature measurement using contactless method

and the temperature distribution recording rate and accuracy are met by modern scientific matrix

thermographic cameras. Their application in passive infrared thermography research on general

strength of materials and structures was presented in the papers [3] and [4].

Thanks to modern research instrumentation such as thermographic cameras, new information can

be obtained on processes that have already been thoroughly analysed and described from another

point of view. Such instrumentation also often allows to reveal completely new phenomena. Such

processes include multiaxial fatigue, for which it is hard to find any papers in the available

professional literature, providing examples of thermography use in the analysis of, for example,

how the extent of load nonproportionality influences dissipation processes associated with the

degree of the specimen or the structural component effort, manifesting in the temperature changes.

This paper provides the results of research on temperature changes of cylindrical specimens

depending on the pattern of the multiaxial load. That research should provide grounds for broader

analysis of the effect of nonproportional load on fatigue properties, which manifests

in the temperature changes.

Test description

Test station. The research was carried out using biaxial testing machine Instron 8874 with the load

range ±25 kN for tension-and-compression and ±100 Nm for torsion (fig. 1).

The extensometric research was performed using 8-channel, versatile extensometric bridge

NI SCXI-1520 manufactured by National Instruments – and NI SCXI-1600 USB module for data

acquisition (fig. 2). 1-RY85-0.6/120#-3-3m strain rosettes manufactured by HBM (fig. 3),

with measuring base of 0.6 mm and quick-drying adhesive 1-Z70 were also used.

Dariusz Skibicki 163

Fig. 1. CEDIP Silver 420M thermographic camera aiming at the specimen installed in the handle

of the testing machine Instron 8874

Fig. 2. The extensometric bridge and the computer controlling Instron 8874 machine as well

as computers retrieving data from the extensometric bridge and the thermographic camera.

a) b) c)

y

Fig. 3. The specimen installed in the handle of the testing machine Instron 8874 (a), arrangement

of extensometers over the rosette (b) and the assumed reference coordinate system (c)

164 Fatigue Failure and Fracture Mechanics

Signals from three extensometric channels of the rosette were recorded during tests together

with force and torque signals from Instron machine. Data was recorded using NI LabVIEW Signal

Express software.

The surface temperature distribution of the specimens was additionally recorded

with the frequency of 50 Hz using thermographic camera CEDIP Silver 420M (fig. 1). Camera

images were transmitted via USB 2.0 interface to PC with appropriate software, where they were

digitally recorded directly on HDD in form of PTW format files.

The following software was installed in the PC computer:

– VirtualCAM allowing two-way communication between PC and the thermographic camera

via USB 2.0 interface,

– CIRRUS Front End which was the user interface used to control the thermographic camera,

– ALTAIR allowing downloading, storage and advanced processing of thermographic images.

Main parameters of CEDIP Silver 420M thermographic camera include:

– resolution: 320×256 pixels,

– pixel size: 25 µm,

– pixel matrix pitch: 30 µm,

– spectral range: 3.6÷5.0 µm,

– sensitivity: below 20 mK (available: 8 mK),

– recording frequency at maximum resolution and digital image transfer via USB 2.0 interface:

up to 140 Hz (up to 25 kHz at the resolution of 64x8 pixels),

– programmable integration time in the range from 10 µs to 10 ms.

Specimen. As nonproportional load was planned to be applied in the research, specimens made

of austenitic steel X2CrNiMo17-12-2 characterized by high sensitivity to nonproportional fatigue

load were used.

The results of the chemical composition test were presented in table 1. Basic mechanical

properties of the tested steel were also determined with the following results: Ultimate Tensile

Strength UTS = 416 MPa, Tensile Yield Strength TYS = 210 MPa and Vickers Hardness

HV10 = 234.

with the requirements of the standard PN-EN 10088-1

Fe C Si Mn P S Cr Mo Ni

[%]

specimen 63.9 0.028 0.25 1.48 0.033 <0.005 18.03 2.22 12.3

min - 0 0 0 0 0 16.5 2.0 10

max - 0.03 1 2 0.045 0.015 18.5 2.5 13

Test specimens were prepared in accordance with the standard ASTM E2207-02 (fig. 4).

Ra

Dariusz Skibicki 165

Loading conditions. The following types of oscillatory, sinusoidal load (see tab. 2 and fig. 5) were

used for the discussed tests: tension-and-compression (1), torsion (2), combined proportional load

including tension-and-compression and tension with the amplitude ratio of τa/σa = 0.5(3) as well

as combined nonproportional load including tension-and-compression and torsion

with the amplitude ratio of τa/σa = 0.5(4) or 0.826 (5) and the phase angle of ϕ = 90°.

The amplitude ratio values were assumed based on desired features of the nonproportional load.

For τa/σa = 0.5 and ϕ = 90° the value of the maximum tangent stress vector does not change over

the entire fatigue cycle. This pattern of load is very popular in many studies available in the

professional literature thus results of individual studies can be easily compared. Whereas due to the

amplitude ratio τa/σa = 0.826, the value of the equivalent stress according to Zenner formula applied

in this research, is constant over the entire fatigue cycle. Thus, this is the most nonproportional load

and one should expect that the effect of the nonproportionality shall be most evident for that load.

Amplitudes of normal and tangent stress were selected so as to ensure that all load patterns

are characterized by equal equivalent stress value amounting to σeq = 330 MPa. In case of that

equivalent stress value, average lives for uniaxial and proportional loads amount to about

150 thousand cycles, while for the nonproportional load (4) about 55 thousand cycles, whereas

for the most destructive load (5) – about 25 thousand cycles.

The tests were performed for four levels of the equivalent load σeq = 300, 310, 320 and 330 MPa

at the load frequency of 1 Hz.

Table 2. Main load parameters assumed in the research for the equivalent stress level of σeq = 330 MPa

σeq σa τa ϕ N

load type

[MPa] [°] [cycles]

1 tension-and-compression 330.0 0.0 - 150 000

2 torsion 0.0 273.1 - 150 000

3 proportional 0.5 330 282.5 141.2 0 150 000

4 nonproportional 0.5 330.0 165.0 90 55 000

5 nonproportional 0.8 330.0 272.6 90 25 000

1) 2) 3) 4) 5)

MY MY MY MY MY

PY PY PY PY PY

Fig. 5. Load patterns used in the research: tension-and-compression (1), torsion (2), proportional

0.5 (3), nonproportional 0.5 (4) and nonproportional 0.8 (5)

Test results. Figure 6 shows the curve of the average specimen temperature T in the measuring part

for following load patterns: nonproportional 0.8 (fig. 6a), proportional (fig. 6b), tension-and-

compression (fig. 6c) and torsion (fig. 6d). Performed analysis of the amplitude and average

temperature recorded during tests suggests that the temperature significantly depends on the pattern

of the load applied to the specimen. The highest change of the average temperature over time was

observed for nonproportional load, where the temperature change stabilized at the level of about

1.15°C after 60 seconds of load application. While in case of proportional load, the average

166 Fatigue Failure and Fracture Mechanics

temperature change value stabilized at the level of about 0.48℃, for tension-and-compression

– at the level of about 0.55°C and for torsion - at the level of about 0.40°C. Moreover, it was found

out that the nature of the average temperature value change corresponds to the nature of the change

of the highest principal strain. The highest value of the temperature amplitude was recorded

for the nonproportional load. While the value of the recorded temperature amplitude for torsion

was negligible.

a)

0,25 ε1 ε2

ε% t, s

0

60 63

-0,25

b)

0,2 ε1

ε% ε2

0 t, s

60 63

-0,2

c)

0,1 ε2

ε%

0 t, s

60 63

-0,1 ε1

d)

ε1 ε2

0,2

ε%

0 t, s

60 63

-0,2

Fig. 6. Sample curves of the specimen temperature change during first 70 seconds of variable load

for the equivalent stress level of σeq = 330 MPa for nonproportional load 0.8 (a),

for proportional load (b), for tension-and-compression (c) and for torsion (d)

Dariusz Skibicki 167

The following parameters were calculated for the values constituting curves presented in test results

paragraph: amplitudes and average temperature values, principal strain vectors εI and corresponding

normal stress vectors σε, plastic strain values σPL and the plastic strain energy EPL. Authors analysed

recorded hysteresis loops, to determine plastic strain value and plastic strain energy. Sample loops

are presented in figure 7. It is worth to analyse the way of the hysteresis loop change with the

increase of load nonproportionality (1) – zero value of nonproportionality, (4) – low

nonproportional load and (5) – the load of the maximum nonproportionality.

400

σε, MPa

300

200

(1)

100

εI, %

0

-0.2 -0.15 -0.1 -0.05 0 0.05 0.1 0.15 0.2

-100

(4)

-200

(5)

-300

-400

Fig. 7. Sample hysteresis loops for loads (1), (4) and (5)

The authors analysed the possibility of correlation between the mean value of the average

temperature curve of the measured specimen part surface TM and the maximum average principle

strain εI (fig. 8a), plastic strain εPL (fig. 8b), plastic strain energy EPL (fig. 8c), as well as between

the amplitude TA of the average temperature curve of the measured specimen part surface

and the amplitude of the minimum normal stress σY (fig. 8d).

It was found out that the average temperature value TM is significantly dependent on the plastic

strain energy EPL and that the temperature change amplitude TA depends on the nominal normal

stress σY (except for torsion). No TA change for torsion is the subject of further work.

Summary

The preliminary research results provided in the paper herein showed that it is possible to perform

research on temperature changes of variably loaded specimens, depending on the nature

of the multiaxial load. Cyclic temperature changes corresponding to the load were observed.

That phenomenon currently is subject to further research, the results of which shall be presented

in future conferences dedicated to fatigue.

168 Fatigue Failure and Fracture Mechanics

a) b)

c) d)

Fig. 8. Graphical presentation of the correlation between parameters determined for the following

individual multiaxial load patters (1-5): average temperature TM and maximum average

principle strain εI (a), average temperature TM and plastic strain εPL (b), average temperature

TM and plastic strain energy EPL (c) as well as amplitude of temperature TA and amplitude

of nominal normal stress σY (d)

The scientific research financed by funds of the Polish National Centre for Science.

References

[1] N. Harwood, W.M. Cummings (eds), Thermoelastic Stress Analysis, IOP Publishing Ltd.,

Bristol 1991.

[2] G. Rudowski, Infrared and its application, Wydawnictwo Komunikacji i Łączności, Warszawa

1978 (in Polish).

[3] R. Litwinko, W. Oliferuk, Yield Point Determination Based On Thermomechanical Behaviour

Of Polycrystalline Material Under Uniaxial Loading, Acta Mechanica et Automatica, Vol. 3,

No.4 (2009), pp. 49-51.

[4] A. Lipski, The use of passive infrared thermography for tests of materials and riveted joints

used in aviation industry- selected problems. Part II of the collection of monographs

(edited by J. Szala): Experimental methods in studies of materials and riveted joints used

in aviation industry - selected problems, Institute For Sustainable Technologies –National

Research Institute in Radom, Bydgoszcz-Radom, 2010 (in Polish).

CHAPTER 5:

Multiaxial Fatigue

© (2012) Trans Tech Publications, Switzerland

doi:10.4028/www.scientific.net/MSF.726.171

and Non-Proportional Loads as delivered and in the Annealed Condition

Dariusz Skibicki1, a, Janusz Sempruch1,b and Łukasz Pejkowski1,c

1

University of Technology and Life Science, Faculty of Mechanical Engineering, Kaliskiego 7,

85-796, Bydgoszcz, Poland

a

dariusz.skibicki@utp.edu.pl, b semjan@utp.edu.pl, c lukasz.pejkowski@utp.edu.pl

Abstract. The article presents the results of fatigue life and fractographic testing of steel

X2CrNiMo17-12-2 exposed to proportional and non-proportional fatigue loads. The following load

types were applied: tension-compressive strength, torsion, proportional combined/complex loads

produced by tension-compressive strength and torsion as well as non-proportional combined load –

by tension-compressive strength and torsion by the phase shift angle φ=90°. The paper analyses the

effect of the load method on the fatigue life and fractography of fatigue fractures recorded, and

especially the effect of non-proportional load.

Introduction

The results presented in the present article refer to the first stage of the research project which aims

at developing the methodology of research programmed for multiaxial load with non-proportionally

changing components by the adaptation of uniaxial ideas of programmed fatigue research. The first

stage of the project programme was planned to include the plotting of reference fatigue life plots for

the uniaxial, proportional and non-proportional loading. The article also presents the fractographic

analysis of the fatigue specimen fractures produced in the above-mentioned fatigue specimen types.

Testing conditions

Loads. The research involved the application of the following types of fully-revered, sinusoidally-

variable loads: tension-compressive strength, torsion, proportional combined loads produced by

tension-compressive strength and torsion with the amplitude ratios of τa ⁄σa =0.5 and 0.826 as well

as non-proportional combined loads produced by tension-compressive strength and torsion with the

amplitude ratios of τa ⁄σa =0.5 and 0.826 and the phase shift angle of φ=90°. The values of the

quotient of amplitudes were assumed by determining the desired non-proportional load features. For

τa ⁄σa =0.5 and φ=90° the rotating vector of maximum tangent stress has the same value throughout

the fatigue cycle. It is a very frequently applied type of load in many literature reports. With that in

mind, it offers a possibility of a comparison of the research results. On the other hand, the ratio of

amplitudes τa ⁄σa =0.826 results in the value of equivalent stress according to the Zenner formula

applied in this research is constant throughout the fatigue cycle. It is therefore the load of the

highest degree of non-proportionality and one must expect that for that type of load the effect of

non-proportionality on the fatigue life will be more visible.

Material. The research material was selected according to its sensitivity to non-proportionality.

Out of 6 preliminarily selected materials (Table 1.), there was chosen the steel of the highest

sensitivity to non-proportionality; X2CrNiMo17-12-2 as delivered and complete annealing.

The material selection was made based on the results of specimens of monotonic tension using

the relationships proposed by Borodii and Shukaeva [1]. In this paper the authors report in the

relationship between the quotient of ultimate tensile strength and the yield strength σu ⁄σy , and

coefficient DN expressing the change in the fatigue life under non-proportional load (the higher the

value of the fraction, the lower the fatigue life under non-proportional load in relation to fatigue life

under proportional load).

172 Fatigue Failure and Fracture Mechanics

Complete annealing was performed in electric vacuum oven with argon protection atmosphere.

The samples were heated up to the temperature of 1050°C, annealed in this temperature for 2 hours,

after which it was cooled to the temperature of 620°C. Then, having been taken out from the

oven/furnace, the specimen were cooled in the open air.

Table 1. Calculation results for the evaluation of the effect of non-proportionality on fatigue life

σu Nn

Material β = -1 ln|α| = 0.705β-1.22 DN = p -1

σy N

C45 0.3302 0.1030 0.5

Cu-ETP 0.0958 0.0704 0.5

MO58 0.7800 0.2137 0.55

S355J2G3 0.1679 0.0791 0.5

X2CrNiMo17-12-2 0.5924 0.1576 0.58

X2CrNiMo17-12-2

1.4570 0.6414 0.75

(annealed)

The selected steel was identified in terms of its monotonic properties, chemical properties and

microstructure.

For the material non-annealed and annealed, the following were reported:

σu = 416 MPa, σy = 210 MPa, HV10 = 234 as well as σu = 589 MPa, y = 240 MPa, HV10= 153. The

results of research of the chemical composition are given in Table 2.

Table 2. Chemical composition recorded for steel X2CrNiMo17-12-2 and the comparison with

norm PN-EN 10088-1

Fe C Si Mn P S Cr Mo Ni

63.9 0.028 0.25 1.48 0.033 <0.005 18.03 2.22 12.3

Min 0 0 0 0 0 16.5 2.0 10

Max 0.03 1 2 0.045 0.015 18.5 2.5 13

The steel microstructure is given in Fig. 1. The metallographic specimens were digested with a

water solution of the mixture of hydrofluoric and nitric acids (Mi16Fe) - 1 volumetric part of nitric

acid. 2 volumetric parts of hydrofluoric acid. 3 volumetric parts of glycerine. The enlargement

applied during the registration of microstructure was identical and it was 250x. The composition of

a visible austenitic structure includes austenite grains with visible twins and grain borders. The

austenite grain size differs considerably. According to norm PN-84/H04507-01, the grain size was

determined at the enlargement of 100x. As for the specimen as delivered the grain size is g=8, while

for the specimen after complete annealing g=5.

a) b)

Fig. 1. Microstructure of steel X2CrNiMo17-12-2 as delivered (a) and after complete annealing (b)

Specimen. The research was made using biaxial strength machine Instron 8874. Performing

tension-compressive strength in the range of +/-25 kN and torsion – 100 N·m. The specimens for

research were designed according to norm ASTM E2207-02 (Fig. 2).

Dariusz Skibicki 173

Uniaxial Loads. For the steel selected there were made plots of fatigue life for tension and torsion

(Figs 3 and 4).

The results were approximated in a double-logarithm design using linear functions. The forms of

the functions are given in Table 3.

Table 3. Approximating functions of fatigue life under tension-compressive strength and torsion for

steel X2CrNiMo17-12-2 as delivered and in the annealed condition.

σ=S·Nb Tension-Compressive

Torsion

log(σ)=b·log N +log (S) strength

S=561, b=-0.0477, S=478, b=-0.0504,

Steel as delivered

R2=0.88 R2=0.91

S=833, b=-0.1119, S=3518, b=-0.2757,

Annealed steel

R2=0.91 R2=0.71

The fatigue plots for torsion in both cases are located under fatigue plots of tension-compressive

strength. As for the material as delivered both plots are approximately parallel (close to the value of

coefficient b). As for annealed material, the torsion plot is more inclined than the tension plot. The

problem of out-of-parallelism of the fatigue characteristics is widely discussed by Kurek and

Łagoda [2].

A comparison of the plots for the specimens from the samples of the same load, however, for the

material as delivered and in the annealed condition, one can observe that in both cases of load of the

plots for annealed steel are found below the plots for steel as delivered Figs 5 and 6). The plots also

show a greater inclination.

174 Fatigue Failure and Fracture Mechanics

Fig. 3. Fatigue life for tension and torsion specimens for nominal stresses as delivered. Symbol ‘x’

stands for non-cracked specimens.

Fig. 4. Fatigue life for tension and torsion specimens for nominal stresses for annealed material.

Symbol ‘x’ stands for non-cracked specimens. ‘A’ stands for material annealed.

Dariusz Skibicki 175

Fig. 5. Fatigue life for specimens exposed to tension for nominal stresses as delivered and in the

annealed condition

Fig. 6. Fatigue life for specimens torsion for nominal stresses as delivered and in the annealed

condition

Proportional Load. Based on the results of tests of tension-compressive strength and torsion, the

analysis of multiaxial fatigue criteria was made in terms of the accuracy of fatigue life evaluation

[3]. The best criterion was the Zenner criterion [4]. Applying that criterion, there have been

proposed the values of amplitudes for proportional combined loads – tension-compressive strength

with torsion with two different ratios of amplitudes τa ⁄σa =0.5 and 0.826. The results of fatigue life

for uniaxial and proportional loads have been compiled in Figs 7 and 8 for steel as delivered and in

Fig 9 for material after annealing. In the first case almost all the results fall within the scatter bound

of factor 2, while in the other one – in the scatter bound of factor 3.

176 Fatigue Failure and Fracture Mechanics

Fig. 7. Fatigue life for material as delivered for the specimen exposed to tension, torsion and

proportional specimens τ ⁄σ = 0.5 and 0.826, for equivalent stresses according to Zenner.

Symbol ‘x’ stands for non-cracked specimens.

Fig. 8. Comparison of calculated fatigue life with the experimental ones for the specimen of tension,

torsion and proportional specimens τ ⁄σ = 0.5 and 0.826, for equivalent stresses according to

Zenner, for the material as delivered.

Dariusz Skibicki 177

Fig. 9. Comparison of calculated fatigue life with the experimental ones for the specimen of tension,

torsion and proportional load τa ⁄σa =0.5 and 0.826, for equivalent stresses according to Zenner, for

the material after annealing.

Non-proportional load. Then the fatigue life was tested under non-proportional load. The research

was performed for tension with torsion with the nominal amplitude ratios of τa ⁄σa =0.5 and 0.826

and for the phase shift angle φ=90°.

The results are presented in Fig. 10. The fatigue life reported under non-proportional load are

much lower than the uniaxial and complex proportional loads and are decreasing with an increase

in the degree of non-proportionality – for loads ⁄ = 0.826 life times are lower than for the

ratio of amplitudes 0.5. Fig. 11 demonstrates fatigue life for τa ⁄σa =0.5 are located in the scatter

bound of factor 3, and for τa ⁄σa =0.5 they are found beyond the bound.

Fig. 10. Fatigue life for non-proportional loads τa ⁄σa =0.5 and 0.826 and the phase angle shift

between components φ=90°.

178 Fatigue Failure and Fracture Mechanics

Fig. 11. Comparison of calculated fatigue life with the experimental ones for non-proportional load

τa ⁄σa =0.5 and 0.826 and phase angle shift between components φ=90°.

Fractographic analysis

As for tension-compressive strength, the fatigue fracture is perpendicular to the direction of the

effect of load (Fig. 12). Cracks were initiated from a single focus found on the specimen surface.

As for torsion, for high values of stresses the direction of macro-cracking coincided with the

direction of the effect of maximal tangent stresses (Fig. 13). A decrease in the amplitude of stress

resulted in a change in the direction of macro-cracking by 45°, which means that they were

perpendicular to the directions of principal stresses. The occurrence of cracks in both planes was

characteristic. For an even lower amplitude of stress, the crack maintained the direction

perpendicular to the direction of the effect of principal stresses, however, it occurred only in one of

the planes of the effect of that stress.

As for the proportional load with the amplitude ratios of τa ⁄σa =0.5 the cracks were perpendicular

to the specimen axis (Fig. 14), while in case of load of a greater share of tangent stress, namely for

the ratio of amplitudes of τa ⁄σa =0.826, the cracks were turned by about 20° as compared with the

specimen axis. The direction is perpendicular to the direction of the effect of principal stresses. In

both cases fatigue fractures were similar to those reported for tension-compressive strength.

Non-proportional loads resulted in the pattern of the trace of the crack on the specimen surface

being very irregular. As for the load with the amplitude ratios of τa ⁄σa =0.5, the directions of cracks

were perpendicular to the direction of the effect of principal stresses, whereas for the load with the

amplitude ratios of τa ⁄σa =0.826 the crack directions were perpendicular to the specimen axis.

Conclusions

The Zenner criterion allows for a satisfactory forecast of fatigue life exposed to proportional

multiaxial load. This criterion, however, cannot be applied under non-proportional load.

According to the forecasts, steel X2CrNiMo17-12-2 is a material which is sensitive to non-

proportionality. The material sensitivity to non-proportionality is high enough to allow for revealing

the effect of non-proportional load of a varied degree of non-proportionality.

The fractography of fatigue fractures for uniaxial and proportional loads complies with the

forecasts. The directions of cracks, on the other hand, for non-proportional load require further

analyses.

Dariusz Skibicki 179

180 Fatigue Failure and Fracture Mechanics

Acknowledgements

The project has been financed from the National Centre for Science

References

[1] M.V. Borodii, S.M. Shukaev. Additional cyclic strain hardening and its relation to material

structure, mechanical characteristics and lifetime. International Journal of Fatigue. 29 (2007)

1184–1191.

[2] M. Kurek, T. Łagoda. Fatigue life estimation under cyclic loading including out-of-parallelism

of the characteristics. Applied Mechanics and Materials. 104 (2012) 125-132.

[3] D. Skibicki, Ł. Pejkowski. Integrals fatigue criteria evaluation for life estimation under

uniaxial, combined proportional and non-proportional loadings, Journal of Theoretical and

Applied Mechanics. 4 (2012)

[4] H. Zenner, A. Simbürger, J. Liu. On the fatigue limit of ductile metals under complex

multiaxial loading, International Journal of Fatigue. 2 (2000) 137-145.

© (2012) Trans Tech Publications, Switzerland

doi:10.4028/www.scientific.net/MSF.726.181

characteristics under block loading

Marta Kurek1, a, Tadeusz Łagoda1,b

1

Department of Mechanics and Machine Design, Faculty of Mechanical Engineering , Opole

University of Technology,

ul. Mikołajczyka 5, 45-271 Opole, POLAND

a

m.kurek@doktorant.po.opole.pl, b t.lagoda@po.opole.pl

Abstract. The paper presents the algorithm of fatigue life estimation for materials with out-of-

parallel fatigue characteristics under block loading. Brass CuZn40Pb2, medium-alloy steel

30CrNiMo8 and high-alloy steel 35NCD16 belong to such materials. Brass CuZn40Pb2 was used

for analysis. The experimental results were compared with those calculated according to the

assumed model, and satisfactory results were obtained.

Introduction

All machinery, equipment and structures must be characterized by the greatest durability and

reliability while maintaining safety. However, there are difficulties in determining the fatigue life

resulting from the inability to determine the stability using a single equation. In [1] presents an

analysis of the relationship between fatigue strength of pure bending and pure torsion of selected

construction materials, which amounts to determining the relative value of normal stress to shear

stress

σ

af

B = , (1)

τ

af

where: σaf – fatigue limit for bending, τaf – fatigue limit for torsion.

Result of work [1] to review and distribution of construction materials due to the variability of the

parameter (1) depending on the number of cycles to destruction. For most materials, this

relationship shows constancy. However, there is a group of materials, for which relation (1) is

variable and dependent on the number of cycles. These are materials which are characterized by a

lack of parallelism of the characteristics of fatigue in pure bending and pure torsion such as brass

CuZn40Pb2, medium-alloy steel 30CrNiMo8 and high-alloy steel 35NCD16. For such materials,

estimation of fatigue life is complicated because it must take into account the variability of the

parameter B in the multiaxial fatigue criteria which take into account this parameter. They include:

the criteria of Gough and Pollard [2], Nisihary and Kawamoto [3], Lee [4], Findley [5], Carpinteria,

and Spagnoli [6], Ogonowski [7,8] and Walat [9]. Aim of this work is the presentation an algorithm

for estimating the fatigue life under loads block for materials characterized by a lack of parallelism

of the mutual characteristics of fatigue. All calculations were done for brass.

Experimental research

CuZn40Pb2 brass was used for the analysis, the results of fatigue presented in [10]. Block loads are

increasingly used in fatigue tests as a kind of random loads [11]. Figure 1 shows the fatigue graph

for pure bending and pure torsion material analyzed.

182 Fatigue Failure and Fracture Mechanics

500

400

logNf=19.9-5.86logσa

300

200

logNf=45.31-17.18logτa

100 4 5 6 7

10 10 10 10

N , cycles

f

Fig. 1. Fatigue curves for pure bending and pure torsion for CuZn40Pb2

The authors of [10] tested 36 samples at various load levels and different order of these levels. Fig.

2 shows the geometry of the specimens used for testing.

Figure 3 shows an example fragment of the course stresses.

Four specimens were used in the experiment for each combination of alternating blocks of torsion

and bending. The block was made up of nτ cycles of torsion and nσ cycles of bending, each

sinusoidal. Periods that have been adopted to study accounted for 10% durability at this level charge

for NI, NII and NIII (Fig. 4). Therefore, the moments causing bending and torsion were alternately

and lasted respectively, nI = 104, nII = 3 ·104 or nIII = 105 cycles, until the destruction of the sample,

where:

nI=0.1·NI . (2)

Dariusz Skibicki 183

Table 4 contains detailed characteristics of levels and amplitudes, which were calculated by using

the characteristics of fatigue.

Tab. 1. Levels of research alternating bending and torsion and the corresponding number of cycles

in the blocks [10]

type of load load level

nσ nτ σ a , MPa τ a , MPa

Ig 10 4 - 364 -

bending II g 3·10 4 - 302 -

5

III g 10 - 245 -

4

Is - 10 - 222

4

torsion II s - 3·10 - 209

5

III s - 10 - 194

Stages of the algorithm of fatigue life determination under block loading are shown in Fig.5. The

initial number of cycles taken for the calculation of the first cycle algorithm is Ni = 106. The first

stage of proceedings is to record the input data, which are the stress courses coming from bending

and torsion in their blocks:

σ xx (t ) = σ a sin(ωt ) , (3)

τ (t ) = τ sin(ωt )

xy a , (4)

where: σa – amplitude of normal stress coming from bending, τa – amplitude of shear stress coming

from torsion, ω – angular frequency, t – time.

184 Fatigue Failure and Fracture Mechanics

The next important step in fatigue life determination is determination of the critical plane

orientation angle corresponding to the maximum effort of the material. In the paper, the critical

plane position was determined with the method of damage accumulation. For loads of block

accumulation was calculated respectively for bending:

j 1

Sσ (α ) = ∑ n(σ ) (5)

i =1 N f

j 1

Sτ ( α ) = ∑ n( τ ),

(6)

i =1 N f

where Nf is the number of cycles corresponding to the shear stress from the bending, respectively:

1

τ ηs (t , α ) = − σ xx (t ) sin 2α (7)

2

τ ηs (t , α ) = τ xy (t ) cos 2α , (8)

according to the standard of ASTM [12]. The course of stress (7) and (8) set the amplitude. Then, in

the case of bending fatigue life are determined from the equation (7)

, (9)

Dariusz Skibicki 185

In the case of torsion τηsa substitute for the value of the formula (8) and fatigue life are determined

from the equation:

After summing up the cumulative (5) and (6) search angle, for which the cumulative value was the

maximum:

S (α ) = Sσ (α ) + Sτ (α ) , (11)

α = max(S (α )) . (12)

The criterion on the plane of maximum shear stresses [7] was applied in analysis. According to this

criterion, the history of equivalent stress can be written as:

σ eqa = (2 − B) ⋅ σ ηa + B ⋅τ ηsa ,

(13)

where:

1

τ ηsa = − σ xxa sin 2α + τ xya cos 2α . (15)

2

Variability parameter B (Nf) is included in the multiaxial fatigue criterion (13). To calculate the

fatigue life (Ncal) the method of the average amplitude of the damage [13]. Modified amplitude was

calculated according to the expression:

σ md = . (20)

n(σ ) ⋅ σ σeqa (σ ) mσ + n(τ ) ⋅ σ τeqa (τ ) mσ

186 Fatigue Failure and Fracture Mechanics

Then, looking for fatigue life are determined from the formula [13]:

mσ mσ

σ σeqa σ τeqa

N f = n(σ ) ⋅

+ n(τ ) ⋅

. (21)

σ md σ md

The presented algorithm is based on the method of iterations, so the rest of the work to calculate the

ratio between constant obtained by initial:

N i +1

∆= . (22)

Ni

It should be noted that Ncal calculated from equation (21) is the sum of two blocks of fatigue life

originating from the bending and torsion which represents 20% of the total according to equation

(2). This procedure is repeated for successive calculated fatigue lives to the moment when the

following condition is satisfied:

0.99<∆<1.01. (23)

Thus, the error at the level 1% was assumed, what is satisfactory in the case of fatigue calculations

of machine element and structures. If the condition (23) is fulfilled, the obtained fatigue life is the

searched quantity.

Fig. 6. presents comparison of the calculated and experimental fatigue lives if the calculated fatigue

life was calculated according to the expression:

7

10

IIt-Ib

IIt-IIb

IIt-IIIb

6

It-IIb

10 IIIt-Ib

IIIt-IIb

It-IIIb

IIIt-IIIb

5 It-Ib

10

2.3

4

10 4 5 6 7

10 10 10 10

Fig.6. Comparison of the calculated fatigue life Ncal according to the criterion in the plane of

maximum shear stresses and the experimental fatigue life Nexp for CuZn40Pb2 when B=const

Dariusz Skibicki 187

Fig.7 shows comparison of calculated and experimental fatigue lives for brass CuZn40Pb2 when B

depends on (1).

7

10

IIt-Ib

IIt-IIb

IIt-IIIb

6 It-IIb

10 IIIt-Ib

Ncal, cycles

IIIt-IIb

It-IIIb

IIIt-IIIb

5 It-Ib

10

2.3

4

10 4 5 6 7

10 10 10 10

Nexp, cycles

Fig.7. Comparison of the calculated fatigue life Ncal according to the criterion on the plane of

maximum shear stresses with the experimental fatigue life Nexp for CuZn40Pb2 when B(Nf)

From the figures it appears that in the case of including variability of the coefficient B(Nf) all the

results are included into the scatter band of the coefficient 2.3. If B is constant, the results are worse.

Conclusion

1. From analysis of the obtained results it appears that the proposed algorithm can be applied

for calculations of the fatigue life under block loadings for the materials with out-of-parallel

characteristics for pure bending and pure torsion. The performed verification gave

satisfactory results.

2. From analysis of simulation tests it appears that the best results are obtained when B is a

function of a number of cycles.

3. The proposed model should be a subject of further analyses for other materials and loadings.

"Project financed from the funds of the National Center of Science, awarded according to the

decision No. DEC-2011/01/N/ST8/06900"

References

[1] Kurek M., Łagoda T., Comparison of fatigue characteristics for some selected structural

materials under bending and torsion, Materials Science, Vol. 47, No. 3, November, 2011, pp.

334-344.

[2] Gough H.J., Some Experiments on the Resistance of Metals to Fatique under Combined

Stresses, London: His Majesty’s Stationery Office, 1951.

[3] Nishihara T, Kawamoto M., The strength of Metals under Combined Alternating Bending and

Torsion with Phase Difference. Memoirs of the College of Engineering, Kyoto Imperial

University, Vol.X, No. 6, 1941.

[4] Lee S.B., A criterion for fully reversed out–of–phase torsion and bending, Multiaxial fatigue

ASTM STP 853, Philadelphia1985, pp.553–568.

[5] Findley W.N., A theory for the effect of mean stress on fatigue of metals under combined

torsion and axial load or bending, Journal of Engineering for Industry, 1959pp.301–306.

188 Fatigue Failure and Fracture Mechanics

[6] Carpinteri A., Spagnoli A., Multiaxial high–cycle fatigue criterion for hard metals, Int J Fatigue

23, 2001, pp.135–145.

[7] Łagoda T., Ogonowski P., Criteria of multiaxial random fatigue based on stress, strain and

energy parameters of damage in the critical plane, Mat.-wiss. u. Werkstofftech, Vol.36, No 9,

2005 pp. 429-437.

[8] Walat K., Kurek M., Ogonowski P., Łagoda T.: The multiaxial random fatigue criteria based on

strain and energy damage parameters on the critical plane for the low-cycle range, International

Journal of Fatigue, 37, 2012 ss. 100-111

[9] Walat K., Łagoda T., Application of the covariance on the critical plane for determination of

fatigue life under cyclic loading, Procedia Engineering, Vol. 2, 2010, pp. 1211–1218

[10] Kohut M., Łagoda T., Trwałość zmęczeniowa elementów wykonywanych z mosiądzu MO58 w

warunkach blokowych obciążeń skręcających i zginających, Problemy rozwoju maszyn

roboczych Konferencja Naukowa Zakopane, 2007, pp.165-167 (in Polish).

[11]Skibicki D. Experimental verification of fatigue loading nonproportionality model, Journal of

theoretical and applied mechanics, 2007, 45, 2, 337-348.

[12] ASTM E 739-91, Standard practice for statistical analysis of linearized stress–life (S–N) and

strain life (ε–N) fatigue data, in: Annual Book of ASTM Standards, Vol. 03.01, Philadelphia,

1998, pp.614–620.

[13] Łagoda T., Sonsino C.M., Comparison of different methods for presenting variable amplitude

loading fatigue results, Matt. – wiss. u. Werkstofftech, 2004, 35, No.1 pp. 13-19.

© (2012) Trans Tech Publications, Switzerland

doi:10.4028/www.scientific.net/MSF.726.189

non-proportional loadings

Łukasz Pejkowski1,a, Dariusz Skibicki1,b

1

Faculty of Mechanical Engineering, University of Technology and Life Sciences in Bydgoszcz,

Poland

a

lukasz.pejkowski@utp.edu.pl (corresponding author), bdariusz.skibicki@utp.edu.pl

Keywords: multiaxial fatigue, non-proportional loading, fatigue criteria, fatigue life estimation

Abstract. The paper presents the fatigue criteria, representing the integral criteria, most frequently

reported in literature. They were verified for uniaxial loadings and for combined: tension-

compression with torsion both proportional and non-proportional. The verification involved a

comparison of the fatigue life reported based on the criterion with the experimental fatigue life.

Introduction

A special case of a varied state of stress is the state causing a change in the directions of the

principal axes. The loading triggering such a state is referred to as non-proportional. The

experiments show that this type of loading results in a considerable decrease in fatigue strength and

fatigue life. The paper analyses the criteria representing the group of integral criteria in terms of the

applicability for the fatigue calculations in the cases of proportional and non-proportional loadings.

Integral approach to the Huber-Mises-Hencky hypothesis. The idea of integral criteria was

initiated by Novoshilov [1] who proved that both the second invariant of the deviator of the stress

state, [2] as well as the equivalent stress according to the HMH hypothesis [3] can be interpreted

as the mean squared value from stresses tangential for all the physical planes passing through the

material point considered:

15

= sin . (1)

8

Symbols and in the above notations denote the angles of the polar system in which the location

of the physical plane is described [1, 2].

The Zenner Criterion. The integral approach was developed by Zenner. In his papers [2,3], he

. According to HMH, it is 1⁄√3, and in real life it falls within the range 0.5 <

claims that the HMH hypothesis assumes a constant ratio of the fatigue limits for tension and

( ⁄ ) < 0.8 for ductile materials. The HMH hypothesis provided as in the formula (2) suggests

torsion

that the value of tangential stress acting on the physical plane is the only one which affects the

material fatigue behaviour. In the real life also the changing value of normal stress affects fatigue.

The author of the criterion claims that those facts should be reflected in the calculations. The effect

of those considerations is the criterion which is expressed in the following formula:

15

= ($ '1 + ) + +, ,& '1 −. ,& + ) sin ≤ . (2)

#

8 ,& ,*

190 Fatigue Failure and Fracture Mechanics

The first Papadopoulos criterion was formulated based on the considerations on the crystalline

microstructure of metals. He claims that the main reason for the initiation of the fatigue crack are

plastic slips in metal grains [4,5]. They can be triggered even in the case of slight external loadings,

e.g. during high-cycle fatigue. It happens so since the grains are provided with the systems of easy

slip. According to the author of the criterion, the tangential stresses acting in the directions of slips

are the main quantities with the effect on the initiation of the crack. The criterion proposed is

expressed in the relationship [4,5,6]:

In the above formula 〈3& 〉 stands for the mean squared value of the generalised amplitude of

tangential stress 3& , )$56 〈7〉 stands for the maximum value the mean normal stress reaches during

the loading cycle and coefficient $ is calculated based on the material constants.

Papadopoulos Criterion 2 (2001). The second of the Papadopoulos criteria assumes widening of

the term of uniaxial fatigue limit by the cases of mulitaxial loading [7]. At the same time the author

claims that the criterion of functionality in the engineering term should divide the stress levels into

safe and unsafe and to estimate the levels based on simple material data without going into

microstructural aspects. The criterion has been written below:

where 3& stands for the generalised amplitude of tangential stress, , *&= is as maximum

hydrostatic stress and ;< - the coefficient determined based on material constants.

Criteria analysis method. The paper compares the experimental fatigue life values with the

calculated ones. The equivalent stress values, calculated based on the criteria, have been related

with fatigue life values based on the Basquin equation [8]. Coefficients of the equation were

parameters in a form of mean dispersion of fatigue life 3? and mean squared error fatigue life

determined drawing on the adequate uniaxial tests [9]. Then there were defined the statistical

7

∑QM R WXY Z K=L,M [

1 7K=L,M 7N&O,M

Q

. 7N&O,M A A

.

M R

a comparison of the values of mean squared error 3@ A . If, for a given type of loading, the mean

Figures providing a comparison of the values of mean squared error. The figures below present

dispersion of fatigue life assumed the negative value, which points to the overestimation of fatigue

life, indicator showing the value of mean squared error was directed towards the negative side,

Dariusz Skibicki 191

6

4

2 T-C

0 T

-2

P

-4

HMH INT Zenner Papadopoulos 1 Papadopoulos 2 N

-6

-8

-10

Fig. 1. Comparison of the mean squared error of fatigue life estimation for aluminium alloy 7075-

T651 [11].

10

7,5 > 6000

5 T-C

2,5 T

0 P2

-2,5 N2

HMH INT Zenner Papadopoulos 1 Papadopoulos 2

-5 N0,5

-7,5

< -130

-10

Fig. 2. Comparison of the mean squared error of fatigue life estimation for 1045 steel [12].

50

T-C

40 >100 T

30 P0,5

20 P1

10 P10

P2

0

HMH INT < -200 Zenner Papadopoulos 1 Papadopoulos 2 P4

-10

Fig. 3. Comparison of the mean squared error of fatigue life estimation for steel 1045 [13].

100

90

80 > 1900 T-C

70

60 T

50

40 P 0,5

30

20 P 0,8

10

0 N 0,5

-10

-20 < -1000 N 0,8

-30

HMH INT Zenner Papadopoulos 1 Papadopoulos 2

Fig. 4. Comparison of mean squared error of fatigue life estimation for X2CrNiMo17-12-2 steel

[14].

192 Fatigue Failure and Fracture Mechanics

Conclusions

To use the experimental fatigue life data, one can state that:

1. Relatively the best results were reported by applying the second Papadopoulos criterion and

the Zenner criterion, and the worst – by applying the integral approach according to the

Huber-Mises-Hencky hypothesis.

2. None of the criteria analysed can be applied to estimate the fatigue life under non-

proportional loadings for materials that are sensitive to non-proportional loads.

3. The integral approach can be effective under non-proportional loadings, however the range

of use should be proven by comparision to wide number of experimental results..

References

[1] H. Zenner, I. Richter, Eine Festigkeitshypothese für die Dauerfestigkeit bei beliebigen

Beanspruchungskombinationen, Konstruktion. 29 (1977) 11-18.

[2] H. Zenner, R. Heidenreich, I. Richter, Schubspannunsintensitätshypothese – Erweiterung und

experimentelle Abstützung einer neuen Festigkeitshypothese für schwingende

Beanspruchung, Konstruktion. 32 (1980) 143-152.

[3] H. Zenner, A. Simbürger, J. Liu, On the fatigue limit of ductile metals under complex

multiaxial loading, International Journal of Fatigue. 22 (2000) 137-145.

[4] I. V. Papadopoulos, A high-cycle fatigue criterion applied in biaxial and triaxial out-of-phase

stress conditions, Fatigue & Fracture of Engineering Materials & structures. 18 (1995) 79-91.

[5] I. V. Papadopoulos, A new criterion of fatigue strength for out-of-phase bending and torsion

of hard metals, Fatigue. 16 (1994) 377-384.

[6] I. V. Papadopoulos et al., A comparative study of multiaxial high cycle fatigue criteria for

metals, International Journal of Fatigue. 19 (1997) 219-235.

[7] I. V. Papadopoulos, Long life fatigue under multiaxial loading, International Journal of

Fatigue. 23 (2001) 839-849.

[8] R. I. Stephens, A. Fatemi, R. R. Stephens, H. O. Fuchs, Metal fatigue in engineering, Willey

Interscience, New York, 2001.

[9] D. Skibicki, Ł. Pejkowski, Integrals fatigue criteria evaluation for life estimation under

uniaxial, combined proportional and non-proportional loadings, Journal of Theoretical and

Applied Mechanics. 50 no 4 (2012).

[10] K. Walat, T. Łagoda, Fatigue life of machine elements on the critical plane determined by the

stress covariance extremum (in Polish), Oficyna wydawnicza Politechniki Opolskiej, Opole,

2011.

[11] E. N. Mamiya, F. C. Castro, R. D. Algarte, J. A. Araujo, Multiaxial fatigue life estimation

based on a piecewise ruled S - N surface, International Journal of Fatigue. 33 (2011) 529-540.

[12] Y. Verreman, H. Guo, High-cycle fatigue mechanisms in 1045 steel under non-proportional

axial-torsional loading, Fatigue & Fracture of Engineering Materials & Structures. 30 (2007)

932-946.

[13] D. L. McDiarmid, Multiaxial fatigue life prediction using a shear stress based critical plane

failure criterion, Fatigue design Vol. 1, Technical Research Center of Finland. (1992) 21-33.

[14] D. Skibicki, J. Sempruch, Ł. Pejkowski, Steel X2CrNiMo17-12-2 Testing for Uniaxial,

Proportional and Non-Proportional Loads as delivered and in the Annealed Condition,

Materials Science Forum. (2012).

CHAPTER 6:

Fatigue Crack Growth

© (2012) Trans Tech Publications, Switzerland

doi:10.4028/www.scientific.net/MSF.726.195

Neimitz Andrzej

Kielce University of Technology, Al. 1000 lecia P.P.,25-314 Kielce

neimitz@tu.kielce.pl

Abstract. In the paper several formulae to compute the fracture toughness are presented. The

formulae include either parameter characterizing the in-plane constraint or out-of-plane constraint

or both. The formulae are based on different assumptions and approaches to fracture mechanics.

Namely, small or finite strains were assumed, global or local approach was adopted. In all cases the

standard, plain strain fracture toughness was used as a reference state.

Introduction

It is well known that fracture toughness is not a material property. It depends on the shape and size

of the structural element. Fracture toughness measured according to the proper standards [1],[2],[3]

(KIC, JIC and δTC) is most often used in fracture criteria. It is so, because such measured a critical

value belongs to the smallest from the all other fracture toughness’s measured using specimens with

a shorter crack and a smaller thickness than defined in the standards. It was shown in Fig.1, where,

the fracture toughness’s measured according to the standards requirements are those values for the

relative crack length, a/W, from the range (0.45 – 0.65). However, using KIC,

40H SEN(B) T TEMP=680 0C

B=16mm, W=25mm

2000 potential drop

compliance change

1600

JQ [kN/m]

1200

800

400

0

0 0.2 0.4 0.6 0.8

a0/W

Figure. 1. The influence of the relative crack length on fracture toughness. Triangles denote results

obtained using potential drop method to measure the crack extension. Rhombuses denote results

received using the compliance change technique.

JIC or δTC one receives conservative; thus, save results; they are not always economically

acceptable.

In this paper several models are presented to compute the “real” fracture toughness of the

structural elements. They will be based on the parameters which characterize the stress field in front

of the crack, namely Q – introduced by O’Dowd and Shih [4], and = introduced by Guo

[5]. The former one represents the influence of relative crack length; the later one represents

element’s thickness.

Models concern ductile materials, which are the most often used for structural elements. Thus,

not only results concerning the stress fields received using the assumption of the small strains were

utilized. Also, the stress analysis using finite strains, more realistic one, was used.

196 Fatigue Failure and Fracture Mechanics

Recent engineering procedures, SINTAP [6], and later FITNET [7] adopted the formula, received

by approximation of experimental results [8]

c

K mat [

= K mat 1 + α (−Q) k ] (1)

where Kmat is the critical stress intensity factor (SIF) computed from the formula: = ,

( )

and coefficients α, k were given for selected steels in SINTAP or they can be found in the look-up

tables after the time consuming calibration procedure which must be performed to estimate the

parameters of the Weibull distribution [9],[10], J is J-integral, E is Young modulus, ν is Poisson

ratio.

Classical Hutchinson [11], Rice and Rosengren [12] solution (HRR) for stress distribution in

front of the crack was corrected by O’Dowd and Shih [4] to take into account the in-plane

constrains:

1

EJ 1+n

σ ij = σ 0 2 σ~ij (θ , n ) + Qσ 0σˆ ij (θ ) (2)

ασ 0 I n r

where σ0 is yield stress, α and n are Ramberg-Osgood coefficients, r and θ are polar coordinates,

and are functions which can be found using computer program [13].

Guo [5] derived approximate formula for the stress and strain fields in front of the crack in

tridimensional geometry. Guo’s formula can be extended to the form similar to Eq. (2) [14].

Functions and depend now on Tz parameter and they can be also found using program [13].

Function Q should be replaced by Q* and the later one can be computed similarly to Q [4] but the

reference HRR state must be replaced by the Guo state. Tz parameter may characterize the out-of-

plane constraint.

O’Dowd [15] assumed that the critical state in front of the crack is met when the opening stress

σ22, riches the critical value, σc, which was assumed to be a material constant, at certain distance rc,

which also was assumed to be a material constant. Selecting the specimen geometry, thickness and

crack length, according to standards [2] and assuming that for this reference state the Q parameter is

equal to zero, O’Dowd received a very simple formula:

n +1

Q

J C = J IC 1 −

(3)

σC /σ0

When in the reference state Q≠0, Eq.(3) can be written in the form:

n +1

σ − Qσ 0

J C = J IC C (4)

σ C − Qref σ 0

Using Guo’s approach without the second term correction, assuming that the Q*=0, the appropriate

formula is as follows:

(1+ n )

I n (Tm , n) σ~22 (Tm = 0.5, n)

J C = J IC (5)

I n (Tm = 0.5, n) σ~22 (Tm , n)

In Eq.(5) it was assumed that at the reference state the average through the thickness value of Tz

function, Tm is equal to 0.5. When one wishes to take into account both the in-plane and out-of-

plane constraints may use another form

Dariusz Skibicki 197

(1+ n )

I n (Tm , n) σ~22 (Tm = 0.5, n) Qm* (n, Tm ,σ 0 , a W )

J C = J IC 1−

I n (Tm = 0.5, n) σ~22 (Tm , n) σC σ0 (6)

In Eq.(6) the reference state (standard specimen) assumes that Q*=0 and Tm=0.5, If in the standard

specimen ∗ ( ) ≠ 0 the appropriate formula is as follows:

(1+ n )

σ~ (T = 0.5, n) Q* (n, T , σ , a W )

22~ m 1 − m m 0 ⋅

σ σ σ

I n (Tm , n) (T

22 m , n ) C 0 (7)

J C = J IC

I n (Tm = 0.5, n) σ − Q* σ

C m 0

σ

C − Q *

σ

m ( ref ) 0

Usually, structural steels are ductile at the service temperature. In such a case the strains within

plastic domain in front of the crack are large, on the order of several or more than 20 per cent. Thus,

the assumption, that the strains are finite seems to be quite obvious. However, there are no closed

form analytical solutions for the stress-strains fields in front of the crack in the case of finite strains.

Nevertheless, the finite element solutions provide an interesting picture concerning the stress fields.

Stresses are finite and the maxima of all the normal components are observed at certain small

distance from the crack tip. When the external loading increases, the opening stress maximum

moves out from the crack tip. If plasticity is constrained, it means that the Q-parameter does not

drop much below zero and the average value of Tz parameter at the maximum location does not

drop much below 0.4, the stress maximum remains constant. In such a case the following formula

can be derived [16]

n +1

Q

J C = J IC 1 − max (8)

σ

22 / σ 0

where "" # is the maximum opening stress. It should be computed numerically or approximated

using formula [16]:

""

#

' $ -('⁄(, $ ⁄* )

= %& , ++

(9)

$ ( * /"

where coefficients c, and d can be found in [16]. If plasticity is not constrained, the plastic zones are

large – even the total yielding is observed, the stress maximum drops slightly (4%-5%) when

external loading increases. In such a case one can use the following formula.

#) ( )

78$ − ( ""

#)

$ − 6( "" 79$ :

01 = 021 31 − ;

( "" )79$

# (10)

Numerical and experimental investigations allowed to formulate the fracture criterion [16]

similar but not identical to the famous Ritchie, Knot and Rice [17] postulate. It was assumed that

cleavage fracture may take place if the opening stress in front of the crack assumes higher value

than the critical stress σc, over the critical length lc. The critical length was assumed to be a material

constant, independent of the temperature, also the critical stress σc, was assumed to be a material

constant; however, this quantity depends on the temperature. Example values of the critical stress

and the critical distance are listed in Table 1.

198 Fatigue Failure and Fracture Mechanics

Table 1. An example values of the critical stress and critical distance [18]

Temperature critical stress critical distance critical stress critical distance

(MPa) (µm) (MPa) (µm)

-200C 1300-1350 ~250 970 ~270

In Fig.2 the stress field is shown for two different specimens made of the same steel. Specimens

contain cracks of different relative length a/W=0.5 and a/W=0.2. Rhombs denote the state right

before stable crack growth, triangles show the state right before the unstable crack growth for

a/W=0.5 and the state during stable, ductile crack propagation for a/W=0.2. For the long crack the

stress maximum moves from the crack tip until stresses are higher than the critical value over the

critical length lc. If it happens, the unstable crack jump is observed. For the short crack situation

changes because the plasticity is free to expand and the Q-parameter drops down. As a result the

opening stresses decrease below the critical value and cleavage is not observed. However, if

plasticity evolution is limited, e.g. by microstructure and/or temperature the stress maximum

remains constant when external loading increases (Fig.3).

Fig.2. Material: ferritic bainitic, Temp. -20oC, a/W=0.5, solid symbols, a/W=0.2 empty symbols,

rhombuses – crack before the stable growth, triangles – crack before the cleavage jump. Critical

stress, σcrit=1300 MPa, lcrit=200 µm.

Dariusz Skibicki 199

Fig.3 Material: martensitic, Temp. -80oC, a/W=0.2, rhombuses: opening stress, circles: effective

stress, triangles: accumulated effective plastic strains.

The characteristic features of the stress field in front of the crack, measured numerically using the

assumption of finite strains, allow for a simple geometric model, which in turn allows formulating

the fracture criterion (Fig.4).

Fig.4. Material: ferritic, Temp. -20oC, a/W=0.5. Broken line: beginning of the stable crack growth.

Solid line: before the cleavage crack

200 Fatigue Failure and Fracture Mechanics

Symbols in Figs 4 and 5 denote: lcrit - critical length (material constant), lact - actual length where

opening stresses are higher than the critical value, ψ - coefficient, computed numerically which is

used to define the distance of the stress maximum location from the crack tip. If extension of the

plastic zone is large this coefficient is usually constant and it is close to 1. For low in-plane

constraint it can slightly decrease when the external loading increases.

Simple analysis of the similar triangles in Fig.5 leads to the following formulae:

lcσ 0 (σ max − σ 0 )

J crit =

2ψ (σ max − σ c ) (11)

or using the relationship between the external loading and the J-integral [19]

1

l n +1 (12)

Pcrit = Pact crit

lact

where Pcrit, and Pact are critical and actual external loadings respectively.

Next we compare two states; the critical one reached for standard specimen and the critical one

for the specimen containing shorter crack than a/W=0.45. To simplify computations we replace the

σmax value with the opening stress value computed using HRR formula at the location of the stress

maximum, (σ 22 )HRR(ψ ) .

Qσ 0

1 +

ψ (σ 22 )HRR(ψ ) − σ 0

J c = J IC (13)

ψ* Qσ 0

1 +

(σ 22 )HRR(ψ ) − σ c

where ψ* is the value computed for shorter crack. Usually both ψ* and ψ are close each other and

the ratio ψ/ ψ* can be neglected. In deriving Eq.13 the self-similarity of the HRR fields was

utilized, ( "" )<==(>) = ( "" )<==(>∗ .

Dariusz Skibicki 201

Using the same arguments as above the formula can be derived for the situation in which non-

standard specimen is not dominated by plane strain state.

J c = J IC

[ ][

ψ (σ 22 )ψpstrain − σ c Qσ 0 + {(σ 22 )m }GUOψ * − σ 0 ] (14)

[ ][

ψ * (σ 22 )ψpstrain − σ 0 Qσ 0 + {(σ 22 )m }GUOψ * − σ C ]

In Eq.14, ( "" >?@ is computed from the HRR formula at the stress maximum, A( "" BCDE>∗

is computed from the Guo formula at the stress maximum, and average value of this maximum is

computed.

Summary

In the paper several different approaches to determine the real fracture toughness of the structural

element were presented. They take into account both the in- and out-of plane constraint measures.

All approaches require numerical computation to determine both the Q parameter and Tz (Tm)

parameters. Finite strains analysis requires, in addition, computation of the maximum opening stress

in front of the crack and location of this maximum.

Acknowledgements: The financial support from Polish Ministry of Science and Higher Education

under contract N N501 199640 is gratefully acknowledged.

Literature

[1] E 1820-05 Standard Test Method for Measurement of Fracture Toughness. ASTM, 2005.

[2] ASTM (2002), ASTM E 399: Standard Test Method for Plain-Strain Fracture Toughness of

Metallic Materials, ASTM International.

[3] BS 5762:1979, Method of crack opening displacement (COD) Testing, (1979)

[4] O’Dowd, N.P., Shih, C.F., „Family of crack-tip fields characterised by a triaxiality parameter-I.

Structure of fields”, Journal of the Mechanics and Physics of Solids, Tom 39, str. 898-1015 (1991

[5] Guo W. Elastoplastic three dimensional crack border field - I. Singular structure of the field.

Engineering Fracture Mechanics 1993; 46(1):93-104.

[6] FITNET, Report (European Fitness-for-service Network). Edited by Kocak M., Webster S.,

Janosch J.J., Ainsworth R.A., Koers R., Contract No. G1RT-CT-2001-05071, 2006

[7] SINTAP: Structural Assessment Procedure for European Industry, Final Procedure, 1999, Brite-

Euram Project No. BE95-1426, British Steel.

[8] Ainsworth R.A., "A constraint based failure assessment diagram for fracture assessment".

International Journal of Pressure Vessels and Piping, 64, 277-285,(1995)

[9] Sherry, A.H., France C.C., Goldthorpe M.R., „Compendium of T-stress solutions for two and

three dimensional cracked geometries”, Fatigue and Fracture of Materials and Structures, 1995,

Vol. 18, No. 1, str. 141-155.

[10] Sherry, A.H., Wilkes M.A., Beardsmore D.W., Lidbury D.P.G. (2005), “Material constraint

parameters for the assessment of shallow defects in structural componenets – Part I: Parameter

solutions”, Engineering Fracture Mechanics, &@, str. 2373-2395

[11] Hutchinson, J.W., „Singular behaviour at the end of a tensile crack tip in a hardening

material”, Journal of the Mechanics and Physics of Solids, Tom 16, str. 13-31 (1968).

[12] Rice, J.R., Rosengren, (1968), G.F., Plane Strain Deformation Near a Crack Tip in a Power-

law Hardening Material, Journal of the Mechanics and Physics of Solids, 16, pp.1-12,

202 Fatigue Failure and Fracture Mechanics

M., Algorithm for Determination of σ~ij (n,θ ) , ~εij (n,θ ) , u~i (n,θ ) , d n (n) , I n (n) Functions in Hutchinson-

Rice-Rosengren Solution and its 3d Generalization, Journal of Theoretical and Applied Mechanics,

44, 1:19-30, (2006).

[14] Neimitz A., Graba W., Analytical-Numerical Hybrid Method to Determine the Stress Field

in Front of the Crack in 3D Elastic-Plastic Structural Elements, Proceedings of XVII European

Fracture Conference. Brno, Electronic version, Book of Abstracts p.85 (2008)

[15] O’Dowd N.P. (1995), “Applications of two parameter approaches in elastic-plastic fracture

mechanics”, Engineering Fracture Mechanics, Vol. 52, No. 3, 445-465.

[16] Neimitz A., Graba, M., Gałkiewicz J. „An alternative formulation of the Ritchie-Knott-Rice

local fracture criterion”, Engineering Fracture Mechanics, 74, 8, str. 1308-1322, (2007)

[17] Ritchie, R.O., Knott, J.F., Rice, J.R., (1973), “On the Relationship Between Tensile Stress

and Fracture Toughness in Mild Steels”, Journal of the Mechanics and Physics of Solids, 21, pp.

395-410.

[18] Neimitz A., Gałkiewicz J., Dzioba I., “The ductile to cleavage transition in ferritic Cr-Mo-V

steel: A detailed microscopic and numerical analysis”, Engineering Fracture Mechanics, vol.77,

pp.2504-2526, 2010

[19] Kumar V., German M.D., Shih C.F., “An Engineering Approach for Elastic-Plastic Fracture

Mechanics”, Electric Power Research Institute. Palo Alto, Ca (1981) EPRI Report NP-[67]1931

© (2012) Trans Tech Publications, Switzerland

doi:10.4028/www.scientific.net/MSF.726.203

after friction stir processing

KOCANDA Dorota1,a, HUTSAYLYUK Volodymyr1, SLEZAK Tomasz1,

TORZEWSKI Janusz1, NYKYFORCHYN Hryhorij2,b , KYRYLIV Volodymyr2

1

Military University of Technology, 2 gen. S. Kaliski Str., 00-908 Warsaw, Poland

2

G.V.Karpenko Physico-Mechanical Institute of NASU, 5 Naukova Str., 79060 Lviv, Ukraine

a b

dkocanda@wat.edu.pl, nykyfor@ipm.lviv.ua

Keywords: carbon steel, fatigue crack growth rate, friction stir processing

Abstract. In the study, there were investigated the effects of friction stir processing (FSP) which

was applied in order to improve the surfaces of notched specimens made of S235JR and S355J2

carbon steels, on their fatigue crack growth rates in the air. There were presented the results of com-

parative fatigue tests conducted at asymmetric tension (R= -0.2) for these steels treated by means of

FSP and for the ones in the delivery state. The method of successive etched material layers used re-

vealed the presence of internal tensile stresses in the surface layers of treated specimens. Crack

growth rates were described on the basis of non-linear fracture mechanics, taking the effects of in-

ternal stresses into account.

1. Introduction

Over the last decade, there has been observed strong interest in friction treatment as a new tech-

nology used mainly to bond metallic materials (Friction Stir Welding, abbr. FSW) that are difficult

to join by means of traditional methods without reducing their fatigue properties. Another prospec-

tive application of friction treatment is the possibility of modifying the surfaces of components and

forming multi-layer structures of various fatigue strength and operating properties by means of

combining the friction method with mechanical stirring of substrates. This modified form of FSW

treatment is called as Friction Stir Processing (FSP) [1]. The two mentioned methods are character-

ized by completely different ways of thermal activation of material bonding processes or modifica-

tion of component surfaces when compared to traditional methods. In the FSW and FSP technolo-

gies, microstructural changes in materials do not take place for the reason that temperatures reached

in friction nodes are significantly lower. The effectiveness of the mentioned methods depends on the

selection of treatment parameters such as tool pressure force applied on a component surface, the

tool's rotational and linear speeds as well as the tool's inclination angle to the surfaces being bonded

or the components being modified. The shape and type of a welding tool have a considerable effect

on the method's effectiveness as well [2].

On an industrial scale, FSW is used mainly as an advanced technology for welding of light alloys

of Al, Mg, Cu and Ti. Welding of the mentioned materials with the use of conventional methods is

significantly limited considering the formation of undesirable structural changes in the material as

well as defects as a result of supplying large amounts of heat to the welding area. In aircraft struc-

tures, the quality of bonding between sheets created by means of the FSW method as well as the

fatigue strength of a joint are of priority significance. Comprehensive research conducted by Airbus

[3] and Bombardier [4] has shown a significant improvement of static and fatigue properties of

joints between components made with the use of the FSW method in relation to traditional rivet

bonding.

The use of friction technology to bond steel sheets meets huge difficulties. Considering high

plasticization temperatures of steel, creating a FSW joint requires using high-strength materials for

welding tools and machines (turning lathes and milling machines) that can produce high pressure

204 Fatigue Failure and Fracture Mechanics

forces and high rotational speeds of the tool in a friction node. When the said difficulties concerning

steels are overcome, instant implementation of the FSW method may be expected in various engi-

neering applications (building engineering, power engineering, bridges, ship building, etc.).

Good weld quality and the structure's grain size reduction in the area of the weld were obtained

for low and medium carbon steels [5-7]. The joint's mechanical properties slightly improved in rela-

tion to the base material. At the same time, the presence of complex structural phases in the weld

was observed. The joint's microhardness increased significantly, namely to 900 µHV, in relation to

the parent material's microhardness, which was 250 µHV. Its profile depended considerably on the

tool's rotational speed. Repeatable quality of friction welded joints is closely connected with provid-

ing invariable conditions of the welding process. To this end, primary process parameters are con-

stantly monitored.

Apart from the main application of the mentioned methods to create lasting metallic joints and

multilayer structures [8], it is of particular interest to develop technologies concerning thin anticor-

rosive layers with properly modified structures on structural carrier materials [9]. The findings of

numerous studies that have been published confirm the prospects of the said technologies from the

point of view of studies that are being conducted on new methods for increasing the fatigue strength

and corrosion resistance of structural steels.

The study presents the results of fatigue tests on crack propagation in air of two structural carbon

steels (S235JR and S355J2), which were surface treated by means of FSP method at asymmetric

tension (R= -0.2). Crack growth rates in the said specimens were compared with crack growth rates

of the specimens in the delivery state. The examined steels are used in building structures, bridges

and ships.

Two structural carbon steels of ordinary (S235JR) and of higher strength (S355J2) were selected

for fatigue tests. In shipbuilding industry, they are used as hull steels. These ones are characterized

by good welding properties and cold crack resistance. The steels' chemical compositions and me-

chanical properties are presented in Table 1. Tension tests on fatigue crack growth rate in the air of

FSP treated and untreated specimens at asymmetric tension (R= -0.2) were performed for flat dou-

ble edge-cracked specimens (DET). The specimens were cut out from 12-mm thick S235JR and

S355J2 steel sheets. Specimen geometry is illustrated in Figure 1. In the initial state, both steels

(S235JR and S355J2) show a ferritic-pearlitic structure with average grain size of 15-25 µm, how-

ever, in S235JR steel, the ferritic structure prevails.

Table 1. Chemical composition and mechanical properties of S235JR and S355J2 steels

C Mn Si P S Cr Ni σY σUT A5

[%] [%] [%] [%] [%] [%] [%] MPa MPa [%]

S235JR 0.08 0.85 0.4 0.006 0.010 0.02 0.02 262 396 36.1

S355J2 0.19 1.67 0.48 0.013 0.006 0.02 0.01 360 534 18.5

Friction-mechanical treatment of both steel specimens was conducted on two surfaces with the

use of a tool in the form of a cylindrical disc, which rotated at the speed of 50m/sec. Plastic defor-

mation of the steel surface layers was obtained after the tool had run two times along the specimens’

axes at linear speeds that were 0.5 m/min and 10 m/min, respectively. The measure of the tool's

pressure force to specimen surfaces was the depth of the tool's penetration into the material. The

depth was 0.25 mm when the tool run along the specimen for the first time and 0.1 mm for the se-

cond time. Surface treatment was performed at the Physico-Mechanical Institute in Lviv, Ukraine.

Dariusz Skibicki 205

It should be emphasized that the physical aspect of friction-mechanical treatment differs from

treatment with the use of a stirring mandrel in FSW and FSP methods. As a result of contact fric-

tion, the component's surface layers heat up to a temperature that is higher than the temperature of

the material's phase transformations (approximately 1100-1300°C). Simultaneously, there take place

plastic deformation of the component surface, intense cooling in the air of the near-to-surface layers

of the tool as well as the component and the tribological medium at the speed of the order of 103-104

degrees C/sec. Such temperature and force conditions occur during forced friction between a rotat-

ing metal disc, which plays the role as a strengthening tool, and a component being treated. High

local heat-up temperature in the metallic micro-area, high contact pressure in a friction node and

dynamic tool action lead to the hardening of the surface and grain refining to nano-sizes in subsur-

face layers. A metallographic analysis of specimen cross-section structures after friction treatment

revealed grain size reduction in steel to the size of 20 – 50 nm. As a result of friction treatment, in

steel components there was formed a plastically deformed surface layer with size-reduced grains of

a total thickness up to 150-250 µm. It was confirmed by microhardness tests conducted on longitu-

dinal and transverse cross-sections of steel specimens that were etched in 4% nital. Vickers micro-

hardness in the near-to-surface layer of plastically deformed S235 steel varied from 300 to 170

µHV0.1 in the parent material, whereas for S355J2 steel, it varied from 360 to 180 µHV0.1 in the par-

ent material. Figures 2 and 3 show the images of specimen cross-sections and the structures of the

surface layers of S235JR and S355J2 steels, respectively.

a) b) c) d)

Figure 2. Plastically deformed surface layer in S235JR steel observed on a longitudinal cross-

section (a), on a transverse cross-section (b), the martensitic-austenitic structure of the second layer

(c), surface layer with size-reduced grains (d) at higher magnification

The surface layer of both treated steels consists of three layers. The first one, which is a near-to-

specimen surface, is a white layer (the colour is a result of etching in nital) of a thickness from sev-

eral to a dozen or so micrometers. It was formed from remelted metal, which solidified again very

quickly. The said layer is characterized by a very fine-grained structure of a hardness that is slightly

higher than the parent material hardness. The second layer, which shows a martensitic-austenitic

structure, is a strongly heat affected layer. Its hardness is considerably higher than the parent materi-

al hardness. An increase hardness in this layer results, among other things, from the diffusion of

carbon into steel. The carbon comes from the decomposition of the tribologic liquid in the friction

node, a very high speed of steel cooling in micro-areas as well as the tool and the tribologic liquid.

206 Fatigue Failure and Fracture Mechanics

The structure of the third layer is a result of heat effect as well. Heat absorbed by the steel causes the

tempering of the layer. However, the speed of cooling in this layer is too low to lead to its re-

hardening. The hardness of this layer, which is called a layer of tempered material, is significantly

lower than the hardness of the second layer and may be even slightly lower than the parent material

hardness.

a) b) c)

Figure 3. Plastically deformed surface layer in S355J2 steel seen on a longitudinal section (a near-

to-surface layer at the bottom of the image) (a), on a transverse section with size-reduced grains in

the surface layer (b), at higher mag. (c)

Results of an analysis of the mechanical properties of steels in the delivery state and after fric-

tion- mechanical treatment revealed slight changes (10-15%) in primary material quantities for both

steels, namely yield stress σY and ultimate tensile strength σUT after friction treatment. Moreover,

the plasticity of S235 steel decreased by 10%, i.e. σUT - σY interval, after surface treatment. In the

case of S355J2 steel, plasticity decreased by about 4%. The results of the above-described tests

prove that the surface treatment changed the static properties of steels to a little degree. However, it

led to a decrease in the plasticity of both steels to a greater degree. Internal stress measurements

with the use of the method of successive etched material layers were performed on both surfaces, A

and B, of treated steels. Direction "x" of the component of normal stress tensor, σxx, agreed with the

direction of the specimen longitudinal axis and, at the same time, with the direction of the tool ac-

tion along specimen surfaces. Direction "y" and normal stress measurement, σyy, corresponded with

transverse stresses in specimens. By means of the method of successive etched material layers, the

level of stresses in the surface layer was assessed to the depth of 0.1 mm. The results of the men-

tioned tests prove the operation of high longitudinal tensile stresses σxx (200 MPa) and transverse

tensile stresses σyy (400 MPa) on surface A in the S235 steel specimen. Whereas on surface B in the

same specimen, there were present compressive stresses σxx (170 MPa) and tensile stresses σyy (160

MPa). For the S355 steel specimen, on both surfaces, there were measured compressive longitudinal

σxx and tensile transverse stresses σyy, which reached 120-300 MPa and 300 MPa, respectively. At

the depth of 0.1 mm, there was observed the presence of compressive internal stresses σxx and σyy of

the order of 160-180 MPa. Such a heterogeneous distribution of internal stresses as well as residual

stresses left in sheets after the manufacturing process had a significant effect on the process of fa-

tigue cracking of steel specimens.

In order to assess the effects of surface friction treatment of S235 and S355 steels on their fatigue

behavior, there were conducted crack propagation tests in DET untreated and surface treated speci-

mens at asymmetric tension (R= -0.2). Crack length measurements were performed optically with

an accuracy of 0.01 mm on a stand equipped with a telescope. At each stress level, two specimens

were tested. For both investigated steels, a characteristic feature of crack propagation in either un-

treated or treated specimens was different crack propagation from edge notches. In the initial stage,

cracks propagated identically from both notches. Then, the crack propagating from one of the

notches arrested and did not grow until an ultimate specimen fracture. From the other notch in the

specimen, the crack propagated during the whole period of the specimen's life. This latter crack con-

Dariusz Skibicki 207

tributed to specimen fracture. Sample surface fractures in S235 and S355 steel specimens with no-

ticeable non-identical crack propagations from notches are illustrated in Figures 4a and 4b, respec-

tively. In the Figures, an arrow denoted as PSO indicates a range of specimen cracking in a plane

strain state, whereas an arrow denoted as PSN indicates cracking in a plane stress state.

a) b)

Figure 4. S355 fracture surface (σmax = 115 MPa) (a) and S235 fracture surface (σmax = 85 MPa)

(b), that indicate fatigue crack propagation mainly from one side of the notch specimen

The above pictures prove that, in the cracking zone of the plates, a substantial area of the fracture

surface corresponds with a stable cracking in plane strain state. After a transition to cracking in

plane stress state, the crack growth rate increases. It is a range of subcritical crack growth until an

ultimate plate fracture. Sample curves of fatigue crack growth rates in the air of untreated specimens

(the blue curve) and treated ones (the red curve) of investigated steels at the stress of σmax= 85 MPa

(R= -0.2) are presented in Figures 5a and 5b, respectively.

a) b)

Figure 5. Experimental curves of crack growth rates in the air for S235 steel (a) and S355 steel (b)

that were surface treated and untreated at σmax= 85 MPa

Experimental results indicate a higher crack growth rate in the air of treated steel than of steel in

the delivery state, particularly in the case of cracking in a plane strain state. The said stage of speci-

men cracking constituted a major part of their fatigue lives. Cracking in a plane stress state was

faster in untreated specimens. Similar behavior of crack growth rates in specimens was also ob-

served at other stress levels. Such behavior of specimens with treated surfaces was a result of less

stable, in thermodynamic sense, surface after friction treatment. However, the fatigue life of speci-

mens that were subjected to friction treatment was higher by 10-15% for both steels in relation to

the fatigue life of specimens in the delivery state.

It follows from the analysis of fracture surfaces of S235JR and S355J2 steel DET specimens that

for the most part of their fatigue lives, the cracking proceeded stably in a plane strain state. Cracking

in a plane stress state proceeded fast and led to an ultimate specimen failure.

208 Fatigue Failure and Fracture Mechanics

Nevertheless, the cracking process in the surface treated S235 and S355 steels develops in tensile

internal stress field σS (secondary stresses). Either in the white layer or in heat-affected layer for

both steels there was found an intergranular brittle cracking mechanism up to the depth of about 150

µm (Figures 6a and 6b). In accordance with the FITNET procedures [10] the influence of primary

stress σp and secondary stress σS fields on specimen’ cracking is expressed in the following way:

K I ,eff (σ , a ) = K I σ p , a + K I σ s , a + ρ (a ) (1)

where: Keff is an effective stress concentration factor, a - crack length, σP and σS mark respective-

ly primary (external ) and secondary (internal ) stresses, ρ(a) is the correction factor to account en-

larged plasticity of the material under the influence of tensile internal stresses.

a) b)

Figure 6. Mechanism of intergranular brittle cracking observed in the surface layer of the S235 (a)

and the S355 (b) FSP treated steels.

The distribution of secondary stresses σP in the depth g of a sample with tensile internal stresses

acted in near-to-surface layer is described by Equation (2). In this equation the denotation c means

the depth of a specimen at which the internal stresses change their sign. Hence, the stress intensity

factor K Is (σ , a ) suitable a flat DET sample is defined by Equation (3) [10]:

[( )(

σ s = σ max ⋅ 1 − ( g / c )2 / 1 + ( g / c )2 )] (2)

( ) ( )

1/ 2

K IS (σ , a ) = σ max ⋅ π ⋅ a ⋅ 1 + (a c )4 − (a c )2 / 1 + (a c )4 (3)

and the correction factor ρ(x) is estimated by formula (4):

ρ (x ) = 0.1 ⋅ ( x )0.714 − 0.007 ⋅ ( x )2 + 3 ⋅ 10 −5 ⋅ (x )5 ; where x = Lr ⋅ K IS K IP (4)

In order to describe fatigue cracking in an elastic-plastic material, there was used the J-integral.

Compliance function Y(a/W) for a flat DET specimen of a width of 2W with a notch cut along

length a has the following form:

π ⋅ a / 2W

2 3 4

a a a a (5)

Y= ⋅ 1.122 − 0.561⋅ ⋅ −0.205 ⋅ + 0.471⋅ + 0.190 ⋅

1 − a / W W W W W

For a stable crack growth in a plane strain state the J-integral range ∆J has the form:

π ⋅Y 2(a /W) ⋅ ∆σ 2 ⋅ a H⋅ ∆σ (n+1) ⋅ a

; H = W − a ⋅ Y a ; F = K ' ( )n '

∆J = ∆Je +∆J pl =

( ) + (6)

E / 1−ν 2 F W W

and

( (

∆J e = ∆K I2, eff (σ , a ) E / 1 − ν 2 ))

where: a is the sum of a physical crack length and the depth of a notch, n’ is hardening exponent

determined from a cyclic stress-strain curve of the material; K’ is the fatigue strength coefficient. K‘

and n’ values were determined experimentally.

Dariusz Skibicki 209

The condition for the occurrence of cracking and a transition from a plane strain state to a plane

stress state is a situation where ∆J=JR , where JR is the critical value of the plastic strain energy that

is necessary to initiate cracking by tearing in a component:

( n +1)

H ⋅ σ cr ⋅a E ⋅R

JR = where σ cr = (7)

F Y ⋅π ⋅a

2

Material cracking resistance R is related to the energy release rate G by the following relation:

( )

R = G IC ⋅ 1 − ν 2 = (

π ⋅ a ⋅ σ cr2

E

)

⋅ 1 −ν 2 (8)

It is well known, that in the case of plasticization of a short distance, JR and R quantities are iden-

tical; therefore, it is enough to use one of the above parameters for cracking process description.

However, in order to allow for material plasticization of a longer distance, both parameters are nec-

essary to characterize this process properly. With the use of the above relations (5-8), there were de-

veloped diagrams of crack growth rates da/dN=f(∆J), which are presented in Figure 7.

In calculations of estimated crack growth rates in steels, the following values were adopted for

the quantities that appear in the abovementioned equations:

S235 untreated: n’ = 0.1473, K’= 581.34 MPa and for treated steel n’ = 0.129, K’ = 501.5 MPa,

S355 untreated: n’ = 0.156, K’= 798.2 MPa and for treated steel n’ = 0.136, K’ = 608.2 MPa

a) b)

Figure 7. Crack growth rate curves against the J-integral and the JR-integral in the air for S235 steel

(a) and S355 steel (b) that were untreated and surface FSP treated at σmax= 85 MPa

In Figures 7a and 7b that present crack growth rate curves in examined steels in the air in the

function of the J-integral and the JR-integral, there were marked the stages of specimen cracking.

Stage I is related to cracking in a plane strain state. Stage II is related to cracking plane change and a

transition from a plane strain state to a plane stress state. Stage III is related to the critical crack

growth. It was described by the JR-integral, taking internal tensile stresses operating in the surface

layers of analyzed specimens into account.

5. Conclusions

The study discussed characteristic features of fatigue cracking in the air of two series of DET

specimens made of S235JR and S355J2 carbon steels under asymmetric tension (R= -0.2). One se-

ries was used in tests on untreated steel specimens, whereas the other series was used in tests on sur-

face treated by FSP treatment specimens. The results of the fatigue tests indicated faster cracking of

treated specimens in a plane strain state in the phase of stable crack growth than in specimens in the

delivery state. It is connected with a thermodynamically less stable surface layer in specimens after

friction treatment, which can also be proved by the presence of high internal tensile stresses on

specimen surfaces and a heterogeneous structure of surface layers. However, cracking of the said

specimens in a plane stress state was slower than in untreated specimens. As a result, the fatigue life

of treated specimens was higher by 10-15% compared to the specimens in the delivery state. It

210 Fatigue Failure and Fracture Mechanics

should be stressed that the surface modification of a 12-mm thick steel by means of friction treat-

ment was carried out on a non-professional stand. From an analysis of world-wide literature, it fol-

lows that few studies in this field concern modification of steel surfaces with the use of the FSP

method.

The research was financed by The Polish Ministry of Science and Higher Education within the

scope of the project N N501 0097 33.

References

[1] R. S. Mishra, Friction Stir Processing technologies, Advanced Mater. Processes, October 2003.

[2] H. Fuji, L. Cui, M. Maeda, K. Nogi, Effect of tool shape on mechanical properties and micro-

structure of friction stir welded aluminium alloys, Mater. Sci. Eng. A 419, 2006, pp. 25-31.

[3] H. J. Schmidt, B. Schmidt-Brandecker, Advanced materials and manufacturing technologies for

aircraft application. 2nd Int. Conf. Material and component performance under variable ampli-

tude loading. 2009, Germany, pp. 67-87.

[4] L.J.J. Kok, K. Poston, G. Moore, Bombardier aerospace FSW demonstrator. ICAF 2011 Struc-

tural Integrity: Influence of Efficiency and Green Imperatives. Springer, New York, 2011,

Proc. 26th ICAF Symposium, Canada, 2011, pp.73-81.

[5] L.Cui, H. Fuji, N. Tsuji, K. Nogi, Friction stir welding of a high carbon steel, Scripta Mater.,

Vol. 56, 2007, pp. 637-640.

[6] Cho Hoon-Hwe, Heung Nam Han, Sung-Tae Hong et al. Microstructural analysis of friction stir

welded ferritic stainless steel. Mater. Sci. Eng. A 528 (211) pp. 2889-2894

[7] A.P. Reynolds, Wei Tang, T. Gnaupel-Herpld, H. Prask, Structure, properties and residual

stress of 304L stainless steel friction stir welds, Scripta Mater. 48 (2003), pp. 1289-1294

[8] J. Gandra, R. Miranda, P. Vilica, J.P. Teixeira, Functionally graded materials produced by fric-

tion stir processing. J. Mater. Proc. Tech. 211 (2011), pp. 1659-1668.

[9] D. Kocańda, A. Górka, Formation of a metal coating by means of friction stir processing, ICAF

2011 Structural Integrity: Influence of Efficiency and Green Imperatives. Springer, New York,

2011, Proc. 26th ICAF Symposium, Canada, 2011, pp.167-178.

[10] A. Neimitz, Fracture mechanics, PWN SA, Warsaw, 1998 (in Polish).

© (2012) Trans Tech Publications, Switzerland

doi:10.4028/www.scientific.net/MSF.726.211

fuselage riveted lap joints

Adam Korbel 1,d

1

AGH University of Science and Technology, Faculty of Mechanical Engineering and Robotics,

A. Mickiewicza Av. 30, 30-059 Kraków, Poland

a

askorupa@agh.edu.pl, bmskorupa@agh.edu.pl, cmachniew@agh.edu.pl, dkorbel@agh.edu.pl

Keywords: riveted joints, fatigue tests, squeeze force, thickness effect, fatigue life.

Abstract. Effects of variables related to design and production of riveted lap joints representative of

longitudinal sheet connections for a pressurized transport aircraft fuselage were experimentally

investigated. The specimens from an aircraft Al alloy D16 Alclad sheets of three different

thicknesses (1.9, 1.2 and 0.8 mm) were assembled under load control using round head rivets and

rivets with the compensator from a P24 Al alloy. For the joints from 1.9 mm thick sheets fatigue

tests indicated a dependency of the crack initiation site and crack path on the squeeze force level

and on the rivet type. At the same time, increasing the squeeze force led to improved fatigue

properties of the joints, specimens assembled using the rivets with the compensator showing fatigue

lives consistently longer than joints with the round head rivets. All observed trends have been

explained based on hole expansion and load transfer measurements. For thin sheets connected using

the round head rivets, local deformations and indentations under the driven rivet head promoted

crack initiation and failure in the adjacent sheet. Fatigue test results indicated that the detrimental

effect of this type imperfections could outweigh the benefits associated with a decrease in secondary

bending due to thinning the sheets. The rivets with the compensator were observed to cause

significant local imperfections beneath the manufactured head, which adversely affected the joint

fatigue performance.

Introduction

Riveting remains a preferred method for connecting elements of an aircraft structure, though

adhesive-bonded and riveted-bonded joints are also applied. A typical design solution for joining

sheets of a pressurized transport aircraft fuselage in the longitudinal direction is a riveted lap joint,

usually comprising three rivet rows, as shown in Fig. 1. Due to eccentricities occurring in the

overlap region for this type of a joint, the so-called secondary bending is induced under nominally

axial loading on the sheets. The phenomenon of secondary bending can lead to considerably

elevated stresses in the sheets and affects the mode of failure of the joint [1].

The fatigue crack nucleation location, crack path geometry and fatigue properties of a riveted lap

joint depend on the integrated effect of a number of factors related to joint design and production as

well as loading conditions. This paper focuses on the influence of the squeeze force, sheet thickness

and rivet type.

Configuration of three-row riveted lap joint specimens used in the fatigue tests is shown in Fig. 2

and the specimens’ dimensions are specified in Table 1. The rivet row spacing s=5d (d – rivet

diameter) and the rivet pitch in row p=5d are typical for fuselage skin connections. The rivet holes

were drilled according to the process specification of the Polish aircraft industry. The total length L

of the specimens was chosen to eliminate the effect of specimen fixture in the fatigue machine on

stress conditions in the overlap region [2].

212 Fatigue Failure and Fracture Mechanics

Sheet thickness Rivet diameter Hole diameter Specimen length

t [mm] d [mm] do [mm] L [mm]

0.8 3.5 208

+0.12

1.2 4.0 (d+0.05) 0 260

1.9 5.0 345

The sheet material was a Russian Al alloy D16CzATWH in the Alclad condition. The

mechanical properties (0.2% yield stress = 291 MPa, ultimate strength = 433 MPa, elongation =

13%) and the fatigue crack growth behaviour of this material are similar to those of the western Al

2024-T3 alloy [3]. Two types of protruding head rivets differing in the manufactured head

geometry, namely with a round head and with the so-called compensator were used to assemble the

sheets, Fig. 3. The compensator, which is a small protrusion on the mushroom rivet head, causes

increased rivet hole expansion. The rivet material was the P24 Al alloy equivalent to the western

2117-T3 material used for the AD rivets. Force controlled riveting was applied using a squeezer

Dariusz Skibicki 213

mounted in the grips of a MTS 810 fatigue machine [4]. The same machine was utilized in the

fatigue tests carried out under constant amplitude loading at a stress ratio of 0.1. This type of

loading simulates variations of the hoop stress in the fuselage skin generated due to the cabin

pressurization. Crack growth on the sheet surface was monitored using a travelling microscope. The

testing equipment is shown in Fig. 4

Fig. 3. Rivet types used in experiments: (a) round head rivet; (b) rivet with the compensator

In production practice, the squeeze force is represented by a ratio of the rivet driven head

diameter (D) to the rivet shank diameter (d) which increases with the squeeze force level. The D/d-

value is, therefore, a first indicator of the riveting process quality. Typical D/d ratios range from 1.3

to 1.5, the latter value being considered as optimal [2]. The rivet installation causes rivet hole

expansion, which generates compressive residual tangential stresses in the hole vicinity. The higher

the squeeze force level, the larger the compressive tangential stress area, which affects the initiation

location and path of fatigue cracks at rivet holes and the joint fatigue life. Increasing squeeze force

yields also a higher residual clamping between the sheets beneath the rivet heads. This leads to

transmitting a portion of the applied load by friction, which again can influence a mode of joint

failure. As an example, Table 2 gives fatigue test results observed under an applied maximum cyclic

stress Smax=120 MPa for specimens from 1.9 mm thick sheets with the round head rivets installed

using four different squeeze force levels, resulting in four different D/d-values.

214 Fatigue Failure and Fracture Mechanics

Table 2. Fatigue lives and crack behaviour for specimens riveted with different squeeze forces (1.9

mm thick sheets, round head rivets)

Fatigue life*

D/d Crack initiation site; Crack shape Crack path

kcycles

1.3 81.6 Under driven head; Quarter elliptical Net cross section

Above net section,

Under manufactured head; Quarter/semi- through rivet hole

1.5 235.5

elliptical

1.6 298.2 Under manufactured head; Semi-elliptical Outside rivet hole

*

Average from three tests

A trend of increasing the fatigue life with the squeeze force, demonstrated in Table 2, was also

exhibited at Smax of 100 and 80 MPa. No impact of the stress level on the location of crack nuclei

and crack path was found.

An illustration of the results from Table 2 are fractographic observation results shown in Fig. 5.

It is seen that fatigue cracks always initiate on the faying surface in one of end rivet rows, which

results from the influence of secondary bending [1]. For a limited squeeze force, the cracks initiate

at the edge of the rivet hole and propagate in the net cross section, Fig. 5a. A more intense

squeezing of the rivet leads to crack initiation outside the hole, but propagation through the hole,

usually shifted above the net cross section, Fig. 5b. For a relatively high squeeze force fatigue

cracks nucleation occurs above the hole, near the edge of the clamping area beneath the rivet head,

and the crack propagates outside the hole, Fig. 5c. The latter behaviour is partly contributed by

fretting [5].

Fig. 5. Effect of squeeze force on fatigue crack initiation and path (explanation in text)

It is seen in Table 2 that the specimens with D/d of 1.5 and 1.6 always failed in the sheet adjacent

to the rivet manufactured head, while in the case of D/d≤1.4 the crack nucleation and failure

occurred in the sheet under the driven head. The above behaviour can be explained based on rivet

hole expansion measurements shown in Fig. 6 and load transfer measurements shown in Fig. 7. In

Fig. 6a, hole expansion is defined as he=(de – do)/ do, where de is the expanded hole diameter. As

shown in Fig. 6a, for D/d of 1.5 and, especially, 1.6, he in the sheet next to the rivet driven head

considerably exceeds that in the sheet next to the manufactured head. At the same time, Fig. 7

demonstrates that loads transferred by the end rivet rows are almost equal. Consequently failure

occurs in the sheet with smaller hole expansion, i.e. under the manufactured head. For D/d≤1.4 he in

both sheets is relatively small and only slightly larger under the driven head (Fig. 6a). In that case,

the negative influence of a much higher transfer load in the sheet adjacent to the driven head (Fig. 7)

Dariusz Skibicki 215

dominates and determines the failure location. A more uniform load transmission distribution for

D/d of 1.5 compared to D/d of 1.3 shown in Fig. 7 stems from lower flexibility of rivets installed

with a higher squeeze force [1].

Fig. 6. Hole expansion measurement results for sheet thickness t=1.9 mm: (a) round head rivet;

(b) rivet with the compensator

Fig. 7. Load transfer distribution in riveted joint for two squeeze force values: round head rivet,

sheet thickness t=1.9 mm

Fig. 6b shows measurements results on he for the rivet with the compensator for two D/d-values.

From a comparisons with Fig. 6a is seen that due to the compensator he in the sheet next to the

manufactured head becomes considerably larger than for the standard, round head geometry. Fig. 6b

indicates that he below the manufactured head of the rivet with the compensator is larger than below

its driven head, which explains why in all fatigue tests on specimens assembled using this type

rivets fatigue failure occurred in the sheet adjacent to the rivet driven head. Similarly as in the case

of specimens with the round head rivets, a higher squeeze force yielded an increase in the fatigue

life. For a given D/d ratio, fatigue lives of specimens assembled using the rivets with the

compensator observed at Smax=120 and 100 MPa were by 40 to 90% higher than for specimens with

the round head rivets.

In order to assess the effect of sheet thickness on the mode of failure and fatigue properties of the

joint, specimens from 0.8 mm and 1.2 mm thick sheets were fatigue tested in addition to the

specimens from 1.9 thick sheet considered in the previous section. The sheets were connected using

the round head rivets applying two different squeeze force values leading to D/d of 1.3 and 1.5 for

either specimen series. The fatigue tests were carried out at three Smax stress values, namely 120, 100

216 Fatigue Failure and Fracture Mechanics

and 90 or 80 MPa. In the case of the D/d=1.3 specimens, the crack path for both sheet thicknesses

and at all load levels was through the rivet holes, slightly above the net section, Fig. 8a. With the

D/d=1.5 specimens, the cracks initiated and propagated above the rivet holes, Fig. 8b. In all cases

failure took place in the sheet adjacent to the rivet driven head. It can be concluded from

confronting the above observations with information in Table 2 that the mode of failure for joints

from the thin sheets (0.8 and 1.2 mm) is different than in the case of joints from the thicker sheets

(1.9 mm). The reason behind the above differences can be local deformations and indentations

under the rivet driven head that occur during the rivet installation in thin sheets due to their low

stiffness. Note that the driven head diameter is smaller than the manufactured head diameter (about

2d).

Fig. 8. Failure mode for specimens from thin sheets with round head rivets: (a) t=1.2 mm, D/d =1.3,

Smax=90 MPa; (b) t=0.8 mm, D/d =1.5, Smax=120 MPa

Results presented in Table 3 indicate that sheet thickness has an impact on the joint fatigue life.

Increasing sheet thickness should yield a lower fatigue life due to the effect of secondary bending.

For example, at Smax of 120 MPa the bending factor kb=Sb/Smax, where Sb is the nominal bending

stress computed according to Schijve’s model [6], equals 1.1 and 0.85 for t=1.9 and 0.8 mm

respectively. For Smax=80 MPa, somewhat higher kb factors of 1.25 and 0.9 are obtained for the

above t-values [1]. However, as seen in Table 3, the observed effect of thickness on the fatigue life

is not systematic, due to the addressed above imperfections inherent in the joints.

Applying the rivets with the compensator to connect thin sheets brings no benefits compared to the

round head rivets because, due to a specific shape of the manufactured head bottom surface (cf. Fig.

3b), significant local imperfections of the sheet beneath that rivet head precipitate failure. For the

above reason, fatigue cracks develop in the sheet under the manufactured head and can grow outside

the rivet hole, Fig. 9.

Table 3. Fatigue lives (kcycles) for specimens of round head rivets and different thicknesses

Smax, MPa 120 100 90

D/d 1.3 1.5 1.3 1.5 1.3 1.5

t=0.8 mm 288.5 322.2 483.0 1666.1 743.6 1665.0

t=1.2 mm 177.0 396.4 347.7 768.5 586.8 1135.4

t=1.9 mm 81.6 235.5 257.2 355.0 507.3* 1174.5*

*

Results for Smax=80 MPa

Fig. 9. Typical failure mode for specimens from thin sheets and rivets with the compensator: t=0.8

mm, D/d =1.4, Smax=120 MPa

Dariusz Skibicki 217

Conclusions

Experimental observations presented in the paper lead to the following conclusions:

1. The initiation and growth of fatigue cracks in riveted lap joints and the joint fatigue performance

depend on rivet hole expansion, and hence on the rivet type and rivet squeeze force, as well as on

the sheet thickness. Fatigue cracks initiate always on the faying surface of the sheets in one of the

outer rivet rows.

2. Essentially, joint from thicker sheets fail in a sheet with smaller hole expansion, but the

distribution of load transfer through the joint can also play a role. For the round head rivet

smaller hole expansion occurs in the sheet below the manufactured head, while for the rivet with

the compensator smaller expansion is observed in the sheet adjacent to the driven head. For

relatively low rivet squeeze forces the crack path is close to the net cross section along one of the

outer rivet rows. At high squeeze forces cracks can start and grow outside the rivet hole. The

fatigue life increases with the squeeze force value and is always longer for the rivets with the

compensator than for the round head rivets.

3. The above observations are not valid for joints from thin sheets. For round head rivets the

riveting process can locally introduce imperfections in the sheet adjacent to the rivet driven head,

which promotes crack nucleation at this location. In this case, no systematic dependency of the

joint fatigue life on the sheet thickness is exhibited. Rivets with the compensator are not suitable

for connecting thin sheets because significant local imperfections beneath the manufactured head

cause a premature failure at that location.

Acknowledgements

The financial support from the governmental research funds within the years 2009-2012 is

acknowledged.

References

[1] M. Skorupa, A. Skorupa, Load transmission and secondary bending in lap joints of aircraft

fuselage. Institute of Aviation Scientific Publications, Warsaw, 2010.

[2] R.P.G. Müller, An experimental and analytical investigation on the fatigue behaviour of fuselage

riveted lap joints. The significance of the rivet squeeze force, and a comparison of 2024-T3 and

Glare 3, PhD Dissertation, Delft University of Technology, The Netherlands, 1995.

[3] J. Schijve, M. Skorupa, A. Skorupa, T. Machniewicz, P. Gruszczyński, Fatigue crack growth in

the aluminium alloy D16 under constant and variable amplitude loading, Int. J. Fatigue, 26 (2004)

1–15.

[4] M. Skorupa, A. Skorupa, T. Machniewicz, A. Korbel, Effect of production variables on the

fatigue behaviour of riveted lap joints, Int. J. Fatigue, 32 (2010) 996–1003.

[5] A. Skorupa, M. Skorupa, Riveted lap joints in aircraft fuselage. Design, Analysis and properties,

Springer, Dordrecht, 2012.

[6] J. Schijve, Some elementary calculations on secondary bending in simple lap joints. Report NLR

TR 72036, National Aerospace Laboratory, Amsterdam, 1972.

© (2012) Trans Tech Publications, Switzerland

doi:10.4028/www.scientific.net/MSF.726.218

BOROŃSKI Dariusz 2,d, GIESKO Tomasz 3,e

1

University of Technology and Life Sciences, Faculty of Telecommunications and Electrical

Engineering, Kaliskiego 7, 85-789 Bydgoszcz, Poland

2

University of Technology and Life Sciences, Faculty of Mechanical Engineering, Kaliskiego 7,

85-789 Bydgoszcz, Poland

3

Institute for Sustainable Technologies - National Research Institute, Pułaskiego 6/10, 26-600

Radom, Poland

a b c

tomasz.marciniak@utp.edu.pl, zbigniew.lutowski@utp.edu.pl, slawomir.bujnowski@utp.edu.pl,

d e

dariusz.boronski@utp.edu.pl, tomasz.giesko@itee.radom.pl

Keywords: digital image correlation, crack growth analysis, multi-processor graphic cards, whole-

field displacement analysis

Abstract. In the paper method of displacement analysis in the cracking zone based on digital image

correlation and advanced multi-processor graphic cards procedures was presented. The basic

assumption for the discussed displacement and strain measurement method under time variable

loads was obtaining high measurement sensitivity by simultaneously minimizing the measurement

time consumption. The developed digital procedures for correlation of images has been used for an

example of displacement analysis in the crack propagation testing in airplane riveted joints.

Introduction

The possibility to apply digital image correlation method to analyse the displacement and strain

has been an exceptionally attractive choice in the mechanics of solids [1,2,3]. However, due to

technological limitations, its practical application is far from being satisfactory, especially for an

analysis of small strain values, typical for operating loads of structures and materials with high

stiffness, for example steel. Measurement of small displacements, which later provides basis for

determination of strains, requires the use of very high resolution digital video cameras and this

involves the necessity of transmission and processing of huge amount of data in a very short time.

The complexity of applying strain determination methods based on analysing the object’s surface

image appears to be high for time variable loads. Additionally long lasting cyclical loads involve a

significant increase in the amount of data to be processed with further reduction of the analysis time

and a considerably higher probability of the method de-correlation.

The remainder of this paper presents a method of displacement determination in the fatigue crack

zone, illustrated by example of tests carried out on sample from airplanes aluminium structure.

The image correlation methods, developed in the 80s, involve comparing a sample image before

and after exposing the object to strain in the form of lighting with white light or laser light (spot

methods) [4,5,6,7]. Displacements of surface characteristic points allow determining the strain

values within the analysed area. The sensitivity of this method depends on the parameters obtained

using image observation methods such as dimensions of the observation field or the image

geometric resolution. Image’s surface points are recorded on a PC before and after exposing the

object to strains. This enables correlation of these images on the basis of intensity recorded for each

pixel using a CCD matrix or a other type of the image sensor.

Dariusz Skibicki 219

Development of the digital image correlation techniques is connected with the progress in

optoelectronics. Constantly increasing resolution of video cameras makes it possible to obtain more

and more accurate measurements. An increased image resolution and the usage the digital image

transmission systems involves the necessity of transmitting a huge amount of data in a very short

time. This significantly limits the possibilities of applying digital image correlation (DIC) methods

for time variable loads.

The basic assumption for the discussed displacement measurement method with time variable

loads was obtaining high measurement sensitivity by simultaneously minimizing the measurement

time. For this purpose special computing procedures based on multi-processor graphic cards (GPU)

were developed. Analysing displacements based on the assessment of the cross correlation

coefficient C is one of the basic solutions used in the image cross correlation methods. The cross

correlation coefficient is determined for a chosen point of the environment, P[x,y], as follows:

b d

I n ( x i , y j ) I m ( xi , y j )

C (u, v) = ∑∑ (1)

i = a j =c f ⋅g

∑∑ [I ] ∑∑ [I ]

b d b d

f = , g=

2 2

n ( xi , y j ) m ( xi , y j ) (2)

i =a j =c i = a j =c

where: a,b - initial and final values of the correlated image coordinate x index, c,d - initial and final

values of the correlated image coordinate y index, In(xi,yj) - the image intensity in the point with

coordinates x y, recorded in the phase of loading, Im(xi,yj) - the image intensity in the point with

coordinates x y, recorded in the phase of loading.

The value of the coefficient C can change in the range from 1 to 0. When C=1, the images of

point P in the environment, recorded in phases n and m, are entirely consistent and when C=0 they

are entirely different. In order to define a displacement of a chosen point P based on images

recorded in the two phases of loading n and m, sub-area O covering point P is separated from image

n and m is displaced in relation to the analogical area in image n. The value of cross correlation

coefficient C is calculated for each location of sub-area On.

Due to low efficiency of standard functions to determine the coefficient C, e.g. those available in

OpenCV library, an original algorithm of correlative search of the image similar areas has been

developed. The implementation was performed by means of CUDA library in version 3.2, facilitated

by NVidia Company.

Since there are two known ways of obtaining a resultant matrix of the pattern adjustment to the

searched image (correlation matrix), it is necessary to make a choice in the initial phase, between

the method based on Fourier transforms used by OpenCV library and the method to directly

determine the values of the correlation function. Therefore it was decided to carry out our own

implementation of the algorithm determining the correlation function according to the version

directly defined by equation (1).

For this version of the algorithm tests were carried out comparing the performance with its

possible maximum optimised equivalent, taken from OpenCV library. Thus, the developed

implementation, computing the functions of correlation on GPU was compared with a multi-thread

OpenCV implementation on CPU and GPU. The results of comparison tests are shown in Figure 1,

where "CPU" is the algorithm with simplest implementation performed on CPU (one thread),

"CPU-CV" represents an algorithm using OpenCV library in a multithread manager, "GPU-CV" is

utilization of OpenCV library in GPU version. "GPU" and “GPU II” are the tested implementation

of algorithms performed on GPU, where GPU II is an effect of additional GPU resources occupancy

optimisation. The authors’ implementation on GPU is more efficient than the other solutions for T

220 Fatigue Failure and Fracture Mechanics

pattern dimensions not exceeding 65x65 points, where T represents size of correlated part of

analysed images. For T pattern dimensions 30x30 points the computation speed achieved was ten

times faster (Fig.1 - “GPU”). Further optimisation of the algorithm’s speed improve efficiency of

image analysis even by 100% (Fig.1 - “GPU II”). The last phase of the algorithm optimisation

involved the use of textures memory for storing the pattern and searched images data. By using the

memory of textures the algorithm’s efficiency was improved by approximately 25%, regardless of

pattern area T.

1000

100

10

1

time, ms

CPU

0.1

CPU-CV

0.01 GPU-CV

GPU

0.001

GPU II

0.0001

0 10 20 30 40 50 60 70 80 90 100

pattern T dimension, points

Fig.1. The results of procedures efficiency comparison tests

The developed digital procedure for correlation of images has been used for an example of

displacement analysis in the method of fatigue crack propagation testing in airplane riveted joints.

A camera with an image resolution of 2448x2050 pixels and a set of lens with 50 mm area of

view was used for the sample image recording. Images was recorded on a PC equipped with GPU

card NVIDIA GTX480 by means of GigaEthernet card with a maximum frequency of 17Hz.

Measurement sensitivity s=0.0025 mm was obtained for the applied system configuration.

In Figure 2a, an images of specimen made of aluminium alloy 2024-T3 with a propagated crack,

which was exposed to constant amplitude of nominal stress and the stress ration R=0, has been

presented. Determined on their basis, distributions of displacements δ within the crack environment

are demonstrated in Figure 2d.

a) c) d)

b)

Fig.2. Crack images for minimal (a) and maximal (b) value of loading cycle. Crack image with

analysis pattern (c) and displacement map against the background of specimen image (d)

Dariusz Skibicki 221

(Fig.3b) in considered crack zone area were determined for tested specimen. On their basis crack

line and crack tip was determined. Further, the course of the crack length changes were determined

on the base of obtained displacement δ distributions in a function of the load cycle number, for the

discussed tests example.

a) b) c) -20 87 51 20

crack tip 1 px = 2.5 µm

δ, px

-25

-30

fatigue crack line

137

20 51 87 111 143 line -40

60 80 100 120

point140 160

number 180 200 220

124

Fig.3. Displacement distribution in crack zones: a) displacement map with cross-section lines,

b) displacement gradient distribution, c) displacement distributions along lines marked at Fig.3a

Summary

The considered method to analyse displacements with the use of digital image correlation

technique supported by GPU technology enables its application in displacement analysis for objects

exposed to time variable loads. By applying modern optoelectronic and IT solutions, the image

analysis algorithms developed for the equipment configuration used in this work was able to

determine the strain distribution in an 2448x2050 pixels image, with the size of a searched patterns

matrix of 30x30 elements and the pattern size of 20x20 pixels, in less than 0.2 sec. Further

reduction of the analysis time is possible for GPU cards with a higher number of cores. Presented

example show possibility of applying the developed method in fatigue crack length and growth

testing and analysis. It is possible use of elaborated displacement analysis method for crack tip

detection also.

References

[1] D. Giesko, D. Boroński, A. Zbrowski, P. Czajka, Detection and measurement of fatigue cracks

in solid rocket propellants, Maintenance Problems, 74/3 (2009) 75-84.

[2] A. Pilch, A. Mahajan, T. Chu, Measurement of whole-field surface displacements and strain

using a genetic algorithm based intelligent image correlation method, Journal of Dynamic Systems,

Measurement, and Control, v. 126, 3 (2004) 479-488.

[3] F. Lagattu, J. Brillaud, M.-C.Lafarie-Frenot, High strain gradient measurements by using digital

image correlation technique, Materials Characterization 53 (2004) 17-28.

[4] T. Schmidt, J. Tyson, Dynamic Strain Measurement Using Advanced 3D Photogrammetry,

Proceedings of IMAC XXI, 2003, Kissimmee.

[5] Information on http://www.correlatedsolutions.com

[6] Information on http://www.gom.com

[7] D. Boroński, Local material properties in fatigue analysis, Publishing House of ITeE-PIB,

Bydgoszcz-Radom (2009) (in polish).

© (2012) Trans Tech Publications, Switzerland

doi:10.4028/www.scientific.net/MSF.726.222

Dariusz Boroński4,d, Piotr Czajka5,e

1,2,3

University of Technology and Life Sciences, Faculty of Telecommunications and Electrical

Engineering, Kaliskiego 7, 85-796 Bydgoszcz, Poland

4

University of Technology and Life Sciences, Faculty of Mechanical Engineering, Kaliskiego 7,

85-796 Bydgoszcz, Poland

5

Institute for Sustainable Technologies - National Research Institute, Pułaskiego 6/10, 26-600

Radom, Poland

a b

tomasz.marciniak@utp.edu.pl, zbigniew.lutowski@utp.edu.pl,

c d e

slawomir.bujnowski@utp.edu.pl, dariusz.boronski@utp.edu.pl, piotr.czajka@itee.radom.pl

Abstract. The authors of this article have presented a hybrid method for inspection in the visible

and infrared bands. A model of a test stand equipped with a monochromatic CCD and thermovision

cameras which enable execution of tests in the field of active and passive thermography has been

presented. Application of two vision channels provides the possibility of observing the defects

caused by cracks and non-uniformity of a material structure, in the subsurface layer. An analysis of

experimental tests performed on selected objects has been presented. Images from both channels,

effects of the image overlapping and profiled charts along characteristic lines in thermograms have

been shown.

Introduction

Fast development of optoelectronics has contributed to a wider use of optical measurement

methods and rapid growth of a new scientific discipline, that is, machine vision [1]. The main

methods of this type employ visible range of the light. Since not all defects that occur in mechanical

structures can be seen in the visible band, these systems are often supplemented with a vision

channel operating in infrared. Thanks to this, it is possible to monitor the temperature of objects,

detect areas of local differences in temperature and zones of heat accumulation, and with the use of

active thermography, spot the defects in the subsurface layer [2]. Such possibilities are especially

useful for assessment of a material fatigue damage.

Images transformation

Application of two vision channels causes additional problems, one of which being

transformation of one image into the other. It results from the fact that the cameras working in

infrared and in the visible light are separate physical objects. Thus, it’s not possible to get images in

the visible light or in infrared recorded exactly from the same location. Due to the fact that cameras

have different optical axes the images do not match each other (they cannot be directly put on each

other). In order to make the two images overlap, one of them needs to undergo perspective

transformation which levels the angle of view from the other camera. If cameras viewpoints are

close, then most image changes can be modeled through a planar homography - linear mapping

relating two corresponding planar points in two views [4]. Planar homography between two views

can be determined by finding sufficient constraints to fix the (up to) 8 degrees of freedom of the

relation. It can be estimated from the matching of 4 points or lines in general positions in two views.

In this case the homography determining matrix H was estimated with the RANSAC algorithm use

[3].

Dariusz Skibicki 223

The system has two modes of homographic transformation determination. In the first mode key

points are indicated manually on the infrared and visual images respectively, so determined

transformation relates directly infrared and visual band images - Figure 1.

Fig.1. Planar homography (matrix H) relation between infrared (IR) and visual band camera pictures

(VIS) – manual mode

In the second mode the special pattern is placed instead of tested object, and obtained images of

the pattern are source of automatically searched key points. This makes possible to calculate two

matrixes H describing transformations between images and plane of the pattern.

The vision module consists of a thermovision camera and a visible band camera, placed on

separate rotary stools. The thermovision camera is equipped with a non-cooled microbolometrical

matrix with resolution 640x480 pixels. The camera sensor operates in a long-wave (7,5÷14µm)

range of infrared radiation. A vision system using a monochromatic 1626x1236 pixels resolution

camera was used to record images in the visible band. Two white diode panel illuminators were

applied to illuminate the observation area. A system with two infrared radiators, each with power of

500W, was used for heat impulse stimulation. LED illuminators and IR radiators were mounted on

articulated arms. Additionally a heating plate with 1000 W power was applied to simulate the heat

processes which occur during production and machining or simulation of operating conditions for

the tested objects. The heating plate temperature regulation range is 50÷300°C. This device makes it

possible to heat the samples through their direct contact with the plate surface. The use of this

heating method is limited mainly to highly heat conductive materials such as metal products.

Figure 2 shows a tests stand with its components.

Observation examples

The basic functions of the experimental inspection system, based on two vision channels are:

- detection and identification of defects of surface structures in the visible band;

- analysis of heat emission from the product surface enabling detection of surface and

subsurface defects and areas of heat accumulation.

Application of active and passive thermography provides the possibility to identify different

kinds of faults, including: fatigue cracks, surface cracks, flaking failures, voids in surface and

subsurface material, machining faults, discoloration of the coating and areas of elevated

temperature. In Figure 3, there are sample images of a faulty disc obtained in the visible band and

infrared active thermography, which has revealed cracks, invisible in the visible band.

224 Fatigue Failure and Fracture Mechanics

Thermovision camera

Slide

Infrared

radiators

CCD camera

Frame

Panel

illuminators

a) b)

Hybrid analysis of images from both bands (the visible and infrared channels) provides more

testing possibilities. For this purpose the test stand has been equipped with advanced methods of

image treatment enabling planar homographic transformation, thanks to which it is possible to

match the images recorded at different points of space and with different resolutions.

a) b) c)

Fig.4. Image of a detail with a subsurface defect: a) in the visible band, b) in infrared, c) in the

visible band with a defect marked after matching the pictures

Dariusz Skibicki 225

Tests were performed on a sample made of polyamide 6 in which blind holes with diameters

1,2,4 and 6 mm were drilled, leaving 2 mm of material (from the plate head) in one case, and 2 mm

of material in the second sample. Samples were exposed to heating impulse with two infrared

radiators, each with power 500 W. The impulse lasted 5 seconds. In Figure 4 there is a temperature

profile obtained from the tested sample along two lines L1 and L2. Line L1 is a straight line running

through the holes centers for which there has been left 2 mm of material (from the hole bottom ),

whereas, for line L2 there has been left 1 mm.

Summary

Despite the widespread use of thermovision in the cracks detection systems (including dual

systems - e.g. with ultrasound or ultraviolet) there are still rarely seen thermovision solutions

combined with visible light images. The experiments performed at the above discussed test stand

prove, that application of a second vision channel in the infrared band is justified and enables

detection of a material subsurface defects. Efficiency of detection depends on the material type and

depth of the defect occurrence. However, it requires application of more complicated algorithms of

the image treatment, hence computers with higher computing capacity are needed.

Application of hybrid visual analysis of specimens during fatigue testing make possible detection

and tracking of fatigue crack growth process.

Research work carried out within the framework of the European Union Strategic Program called

"Innovative Systems of Technical support for Sustainable Development of Economy " in the

"Innovative Economy Operational Programme".

References

[1] T. Marciniak, D. Boroński, Z. Lutowski, S. Bujnowski, T. Giesko, Digital Image correlation -

universal tools versus custom solutions, Wydawnictwo Naukowe ITeE-PIB, 4 (2010) 19-28.

[2] W. Minkina, S. Dudzik: Infrared thermography - errors and uncertainties, John Wiley & Sons

Ltd, Chichester, 2009.

[3] M.A. Fischler, R.C. Bolles, Random sample consensus: A paradigm for model fitting with

applications to image analysis and automated cartography, Communications of the ACM, 24(6)

(1981) 381–395.

[4] R. Hartley, A. Zisserman, Multiple View Geometry in Computer Vision, Cambridge University

Press, 2000.

© (2012) Trans Tech Publications, Switzerland

doi:10.4028/www.scientific.net/MSF.726.226

Tomasz Giesko1, a

1

Institute for Sustainable Technologies – National Research Institute,

ul. Pułaskiego 6/10, 26-600 Radom, Poland

a

tomasz.giesko@itee.radom.pl

Abstract. The article presents a dual-camera vision system for fatigue monitoring composed of a

vision unit, a camera positioning set and a computer unit. Vision modules are mounted onto the four

degrees-of-freedom (4DOF) positioning sets, which allows for an easy determination of the position

of the camera in relation to the sample. The application of motorized measurement lenses with

changeable configuration, thanks to the alteration of the distance of observation and the vision

angle, enables the adaptation of the system to different scales of observation of the fatigue processes

in the specimen surface. Automatic focus setting is realised with the use of the implemented

algorithm. The software developed allows for the analysis of fatigue fracture for two two-

dimesional (2D) images or the three-dimensional (3D) stereovision image.

Introduction

Fatigue processes taking place in materials and structural elements are one of the most important

issues in the area of knowledge on the maintenance of technical objects. The monitoring of fatigue

is facilitated by the application of experimental methods with the use of optoelectronic technologies

that are applied both in fatigue diagnostics of materials and structures as well as local deformation

analyses. Among the aforementioned, the optical methods are more common. They employ sensor

cameras of high resolution, lenses ensuring proper vision angle, and mechatronic systems for the

positioning of the vision module in relation to the object. On the basis of images acquired, the crack

trajectory is analysed and the increase of the crack length in the function of load and number of

cycles is measured [1]. Great accuracy and high frequency of measurements, as well as the

possibility of automation of the measurement process are the basic requirements for the apparatus

enabling the conduction of the advanced fatigue tests. The primary restrictions to the development

of measurement apparatus for fatigue tests employing visual methods stemmed from the incapability

to meet these requirements properly. The more complex and advanced equipment solutions and

programming tools enabling the fatigue monitoring and the measurement of cracks are the results of

the intense development of optoelectronic technologies and image processing and analysis methods.

The literature review indicates that experimental systems equipped with single CCD camera have so

far been the major type of optoelectronic measurement apparatus used for fatigue monitoring and

tests. The optomechatronic systems for the monitoring of fatigue fracture in materials and structural

elements developed jointly by the Institute for Sustainable Technologies – National Research

Institute (ITeE-PIB) in Radom and the University of Technology and Life Sciences in Bydgoszcz

[2,8] constitute good examples of such solutions. These systems are equipped with single camera

vision module and mechatronic system for the positioning of camera in relation to the objects tested,

and they are adapted to work with fatigue testing machines.

The developing imaging systems employing high resolution CCD cameras are being widely used

in surface geometry measurements in micro and macro scale [3]. Making use of the possibility of

simultaneous monitoring of the surface from the two cameras, the assessment of deformations

employs the advanced methods of three-dimensional digital image correlation (3D-DIC), which

facilitates the development of experimental mechanics. The application of dual-camera vision

systems in the experimental tests for the fatigue monitoring in the specimen under various types of

Dariusz Skibicki 227

load is still a rare phenomenon. The experimental stereovision system for three-dimensional surface

deformation measurement and crack length measurement in the tension-torsion loading test is

presented in publications [5,7]. In the dual imaging system high speed CCD cameras are used. The

supporting structure of the cameras mounted on the frame of the fatigue testing machine enables the

positioning of the cameras in the range of 16÷80 [º] between optical axes. The leading commercial

solutions of this type are the systems by ARAMIS GOM GmbH [10] and Trilion Quality Systems

[11]. In these dual-camera systems, the position of the cameras in relation to the specimen is

regulated manually. A similar solution is offered by the system for measuring the shape,

displacement and strain of surfaces in three dimensions developed by the Correlated Solutions, Inc.

[12]. The system allows for the monitoring of objects within the distance of 1÷10 [mm]. In the

aforementioned systems DIC methods are used for the strain and displacement analysis. However,

the high price of these solutions impedes their wide application in research organizations. Thus,

tasks aiming at the development of alternative solutions of similar technological level but lower

price are being undertaken. Publication [6] presents the dual-camera system equipped with the

cameras of sensor resolution ca. 1 megapixel enabling the stereovision imaging, whereas

publication [4] deals with the topic of the stereovision system enabling 3D measurements at the

time of experiments conducted on the fatigue testing machine. In the systems applying analysis of

images taken from several cameras, the problem of the calibration of the cameras deciding on the

measurement accuracy is of great importance. The calibration processes employs different

calibration patterns, frequently developed by the authors themselves. On the basis of the

publications available, it can be stated that the scarce multi-camera vision systems that have so far

been developed are mainly of experimental character. No structures with vision and mechatronic

modules enabling precise positioning of optical paths with the measurement of their position for

each degree of freedom have been found.

The article presents a dual-camera vision system for the monitoring of fatigue wear in materials

and structural elements developed at the Institute for Sustainable Technologies – National Research

Institute in Radom. The software of the system was developed in close cooperation with the

University of Technology and Life Sciences in Bydgoszcz.

SYSTEM CONCEPT

The system concept consists in the application of a dual-camera system for the monitoring of the

surface of the object based on the epipolar geometry (Fig. 1). In the system presented, the optical

axes of both the cameras cross in one point on the surface of the object.

The imaging system in such a case ensures the achievement of a wide range of the dimensional

observation field, which was one of the basic requirements of the monitoring system. The scale of

the observation field is the result of the parameters of the lens used and the size of the sensor of the

camera (Fig. 2).

228 Fatigue Failure and Fracture Mechanics

a) b)

Fig. 2. View of the model of the vision module: a) degrees of freedom, b) basic geometrical

parameters

The optical resolution of the vision path is determined by the relation of the real dimensions of

the observation field to the pixel size of the camera sensor. The parameters of the optical system

were determined with the use of basic formulas in classic optics. As the measurement system was

intended to be applied for tests on fatigue testing machines, the functional requirements and

structural restrictions stemming from the assumed work conditions were also taken into account. In

the positioning system for each of the modules four degrees of freedom were assumed, which is

supposed to ensure the easiness of the regulation of the position of the vision path with reference to

the sample observed.

Within the analysis conducted, the specification of basic parameters of the designed stereovision

system was developed (Tab. 1).

Table 1. Selected parameters of the stereovision system

Parameter Value

80 [mm] ÷ 360

Working distance WD

[mm]

Field of view FOV (dimension H) 1 [mm] ÷ 50 [mm]

Maximum optical resolution 3 [µm]

Pan angle range between the

16 [º] ÷ 110 [º]

cameras α

Positioning range in the O-Y axes 150 [mm]

Positioning range in the O-Z axes 300 [mm]

Positioning range αx ±45 [º]

Positioning range αy ±30 [º]

The matrix of design solutions for the structure of the vision module was developed, which

constituted a starting point for the determination of its individual components. It is, however, worth

underlying that the design of the structure of the vision system is a process of selecting from the

limited availability products the following components: the camera, lens, filters, lighting set-up.

Using the parameters identified, the components meeting the predefined requirements were selected.

In order for the required measurement resolution to be ensured, the CCD camera with 2448x2050

pixel sensors was used. Due to the necessity to adjust optical parameters of the lens to the

observation scale, the motorised lens Zoom6000 by Navitar was selected [9], easy to reconfigure

with the use replaceable components. The simulation and the analysis of the dependency between

the working distance of observation (WD) of the specimen mounted onto the fatigue testing

machine and the (α) pan angle between optical axes of cameras are possible (Fig. 3).

Dariusz Skibicki 229

90 110

80 100

90

70

Pan angle between cameras α [°]

80

60

70

50 60

40 50

Lower limit

40

30 Upper limit (with grips)

30 Upper limit (without grips)

20 Lower limit

Upper limit (with grips) 20

10

0 0

100 105 110 115 120 210 230 250 270 290 310 330 350 370 390

Working distance WD [mm] Working distance WD [mm]

Fig. 3. Positioning ranges for the angle between optical axes with mounting grips considered

The grips in the fatigue testing machine limit the top range of vision angles for greater working

distances. The grips, however, do not limit the bottom ranges for the regulation of vision angles.

The optional configurations of lenses ensure that the required measurement ranges and optical

resolutions are obtained (Fig. 4 and 5). The realisation of the assumed configuration of the

Zoom6000 lens consists in the application of replaceable components (body tubes and accessory

optics) that cooperate with the main lens module.

100

90

80

Field of view FOV [mm]

70

60

Upper limit line

50

I

40

30

20

II

10 III

Lower limit line

0

0 50 100 150 200 250 300 350 400

Working distance WD [mm]

Fig. 4. FOV for configurations of lenses I, II, III with reference to the WD working distance

(possible solutions in the I, II, III areas between the Lower and Upper limit lines)

30

25

Optical resolution [µm]

I

20

15

10 II

5 III

0

0 50 100 150 200 250 300 350 400

Working distance WD [mm]

Fig. 5. Optical resolution with reference to the WD working distance

The Zoom6000 lens is equipped with miniature DC motors that adjust the zoom and focus,

which allows for the automation of the image scaling and calibration processes. The developed

mechanical structure of the monitoring system is presented in Fig. 6.

230 Fatigue Failure and Fracture Mechanics

Supporting frame

LED illuminator

Zoom6000 Lens

Specimen

CCD camera

Fig. 6. General view of the fatigue monitoring system mounted onto the fatigue testing machine

Rigid aluminium profiles ensuring high stiffness of the structure at a considerably low weight

were used. The frame mounted onto the fatigue testing machine is the main support element. The

vision modules are fixed on the 4DOF positioning sets, which allow the adjustment of the camera

position in relation to the specimen. The degrees of freedom of the vision module include the

following: the linear shift along the optical axis, rotation in the vertical and horizontal planes, as

well as rotation in relation to the optical axis.

In the lighting module for the scene, LED ring illuminators with the illuminance of ca. 18 [klux]

at the distance of ca. 300 [mm] from the sample were used (Fig. 7). The LED illuminators are

triggered in the frequency range of 1÷99 [Hz]. In the case of the imaging of high speed processes,

when a very high luminous flux density is needed, a halogen illuminator with the illuminance of ca.

500 [klux] at the distance of ca. 300 [mm] from the specimen can be applied.

a) b)

b) with the use of halogen illuminator

The developed software allows for the analysis of fatigue processes. The method of digital image

correlation DIC is used for the detection and identification of the areas characteristic for the images

recorded. In order to ensure the proper functioning of the programme modules responsible for the

realisation of digital image correlation in the 2x2D mode and the reconstruction of the surface

analysed in the stereo 3D mode, the camera calibration process is conducted. The automatic image

sharpness setting realised by means of the electronic lens control is also applied.

Dariusz Skibicki 231

Discussion

The developed dual-camera vision system presents the alternative concept of the structure and

mounting of the system compared with ARAMIS GOM GmbH and Trilion Quality Systems. The

direct mounting onto the fatigue testing machine ensures the rigid and stable position of the vision

system in relation to the specimen which is a significant advantage in long-term tests. It should be

emphasized that every position change of the vision system (intended or unintended) requires the

calibration of the cameras to be repeated. Regulation ranges for the camera position in the 4DOF

system enable similar observation possibilities as aforementioned commercial solutions. The rare

experimental stereovision systems [5,7] are not equipped with the positioning mechanical modules.

These systems allow for a very limited positioning of cameras and are definitely inferior to the

system presented in the paper.

Summary

The dual-camera vision system was developed to meet the needs of the advanced fatigue tests in

particular. It allows for the observation of the surface of the specimen enabling the analysis of

created deformations, detection of the initiated fatigue crack and the monitoring of the crack

propagation. The wide range of possible applications is the main advantage of the system, as it

allows for the independent positioning of the cameras in the 4DOF systems and the reconfiguration

and modification of the components of the vision module, i.e. the camera, lens and illuminator. The

system can cooperate with the user’s individual software. Further development of the system can

include automation of the function of the positioning of the vision module in relation to the

specimen at the stage of calibration as well as during the monitoring of the fatigue process.

Scientific work created within the framework of the “Innovative Systems of Technical Support

for Sustainable Development of Economy” Strategic Programme within the Innovative Economy

Operational Programme.

References

[1] Boroński D., Szala J., Giesko T.: Automatic measurements of fatigue crack length and

trajectory. 19th Danubia-Adria Symposium on Experimental Methods in Solid Mechanics,

Polanica-Zdrój 2002 130-131.

[2] Giesko T., Boroński D., Zbrowski A., Czajka P., Detection and measurement of fatigue cracks

in solid rocket propellants, Maintenance Problems 3 (2009) 75-85.

[3] Hügli H., Mure-Dubois J., 3D vision methods and selected experiences in micro and macro

applications, Two and Three Dimensional Methods for Inspection and Metrology IV (proc.

SPIE) Vol. 6382 Iss. 10 (2006) 209-216.

[4] Karpour A., Zarrabi K., A Stereo Machine Vision System for measuring three-dimensional

crack-tip displacements when it is subjected to elastic-plastic deformation, E-Leader

Conference Singapore 2010, http://www.g-casa.com/E-Leader-Singapore_Program.htm.

[5] Sharpe W.N. (ed.), Springer Handbook of Experimental Solid Mechanics, Springer 2008 589-

597.

[6] Tang Z.Z., Liang J., Xiao Z.Z., Guo C., Hu H., Three-dimensional digital image correlation

system for deformation measurement in experimental mechanics, Optical Engineering 49 (10)

(2010) 103601-103609.

[7] Yan J.H., Sutton M.A., Deng X., Wei Z., Mixed-mode crack growth in ductile thin-sheet

materials under combined in-plane and out-of-plane loading, International Journal of Fracture,

vol. 14 no 4 (2009) 297-321.

[8] Zbrowski A., Samborski T., Giesko T., Boroński D.: Opto-mechatronic system for fatigue

crack monitoring of riveted joints. Maintenance Problems 4 (2010) 153-162.

232 Fatigue Failure and Fracture Mechanics

[10] Information on http://www.gom.com/metrology-systems/

[11] Information on http://www.trilion.com/

[12] Information on http://www.correlatedsolutions.com

© (2012) Trans Tech Publications, Switzerland

doi:10.4028/www.scientific.net/MSF.726.233

Jacek Jackiewicz

University of Technology and Life Sciences in Bydgoszcz, Faculty of Mechanical Engineering,

Department of Applied Mechanics, 7, Prof. S. Kaliski Street, PL 85-789 Bydgoszcz, POLAND

email: jacek.jackiewicz@utp.edu.pl

Keywords: Direct and iterative solvers, Fracture mechanics, BCC and FCC steels, Peierls stress,

Strain localization

Abstract. The paper discusses issues related to the damage accumulation and cracking in steels.

Special attention is paid to the selection of appropriate methods in the modeling of progressive

damage development. In special cases, the damage accumulation and crack propagation may lead to

the brutal destruction of machine parts. Hence, some attention is drawn to the conditions that can

lead to this brutal destruction. In order to model unstable crack propagation the applied solver for

solving systems of resulting equations should be as efficient as possible.

Generally, brittle solids contain small microcracks. Such cracks may propagate at a stress that is

lower than that required for any slip of dislocations. Brittle fracture in brittle solids can be defined

as occurring at very small plastic deformation when the average stress carried by material is less

than the yield stress. If preexisting cracks in materials are small, then the stress can reach the level

required to initiate slip or twinning. Brittle cracks nucleated by slip or twinning are not able to

propagate immediately. The ability of steel to plastic deformation in the vicinity of the crack tip

prevents the spread of brittle cracks. Substantial plastic strain or grain boundary sliding can cause

the nucleation of larger grain boundary cracks or cause preexisting cracks to grow in a stable

manner, until an increased length of one of these cracks coupled with higher stress due to work

hardening leads to unstable propagation of the cleavage crack.

During the ductile fracture of steel tensile strength is less than the stress required for crack

propagation. In this case, steel first deforms plastically, causing the nucleation and growth of voids.

Such cracking occurs along the grain boundaries and is called intergranular, when most of the

second phase particles, where voids are formed, is located on grain boundaries.

If the distribution of second phase particles is relatively uniform, the cracking occurs through

grains and is called transgranular.

The modeling of dislocations in crystals remains attractive due to reduction of computational

effort when compared to, for example, atomistic simulations. In steels, which have the body-

centered cubic (BCC) crystal lattice structure, such as ferritic steels, the Peierls stress for their slip

systems is predicted to increase with decreased temperature, thereby increasing the yield stress at

low temperatures. The Peierls stress (discovered by Rudolph Peierls and modified by Frank

Nabarro) represents the force necessary to move a dislocation within a plane of atoms in the unit

cell. The magnitude of the force needed to move the dislocation depends on the distance between

planes of atoms in the unit cell and also the size and width of the dislocation according to the

following equation [1, 2]:

= exp ‖ ‖

, (1)

where μ is the shear modulus, ν – Poisson's ratio, ζ ≝ 0.5d ⁄ 1 − ν" – the half-width of the

dislocation, d ≝ a⁄√h + k + l – the interplanar spacing between planes of atoms in the cubic

crystal with Miller indices h, k and l of cubic planes form a notation system in crystallography for

planes and directions in crystal (Bravais) lattices, a – the lattice constant (the edge of the cube), –

234 Fatigue Failure and Fracture Mechanics

the Burgers vector associated with a dislocation that represents the measure of a lattice distortion

caused by the presence of the line defect. The length of the Burgers vector is usually represented by

the equation:

‖ ‖ = 0.5a√h + k + l (2)

but its direction depends on the most likely plane for slip, which is usually the closest-packed

plane of the unit cell. In most metallic materials, the length of the Burgers vector for a dislocation is

equal to the spacing between atoms in the most closely packed direction on the most closely packed

plane. Because of the Peierls stress for the slip system of material is predicted to decrease with

increasing planar density of atomic packing (represented by d ⁄‖ ‖), any slip of dislocations is

preferred on closely packed planes.

The shortest interatomic distances, assessed for materials with face-centered cubic (FCC)

structures, are found to be smaller than that assessed for BCC ones. The yield stress dependence on

the temperature is low for FCC materials, therefore these materials can be used in cryogenic

conditions. However, BCC metals, of which the yield stress has strong dependence not only on the

temperature but also on the strain rate, are prone to brittle fracture.

Sequential multiscale modeling of crack growth requires the use of the incremental methodology

for the input equations of the structural model.

If an idealized problem of damage accumulation and crack growth in a steel part is solved by the

use of the finite element method (FEM), a system of algebraic equations with the coefficient matrix

is formulated. The coefficients of the system reflect relationships between grid nodes specified by

the spatial discretization for the assumed mathematical model.

Since crack propagation creates new surfaces, the number of finite elements should be increased

in the numerical model. If the number of finite elements is increased than it is obvious that the

number of system equations should be also increased. Very often, the coefficients of a system of

equations are written in the form of a matrix. There is a significant effect of the bandwidth of the

coefficient matrix on the time of calculations and the size of used memory.

In practice, it is difficult to determine optimal numbers of mesh nodes in the context of the

bandwidth reduction problem taking into consideration new surfaces of a propagating crack. For a

symmetric sparse matrix of the coefficients, which are determined for a system of equations, the

problem is to reduce the matrix bandwidth by permuting rows and columns such as to move all the

nonzero elements of the matrix in a band as close as possible to the matrix diagonal. The use of the

reverse Cuthill-McKee algorithm [3] for renumbering mesh nodes makes sense only for the initial

spatial discretization of the numerical model up to the start of crack propagation.

The bottleneck effect in simulations of crack propagation is generally associated with the

necessity of solving a large system of equations, which gets expanded every time not only by the

creation of new equations but also by the extension of old ones due to the occurrence of crack

growth.

The extended finite element method (XFEM) [4] may be used to model propagation of various

discontinuities: strong (cracks) and weak (material interfaces) without remeshing. A key advantage

of XFEM is that discontinuous basis functions are added to standard polynomial basis functions for

nodes that belong to elements intersected by a discontinuity. The discontinuous basis functions

provide a basis for discontinuity opening displacements using the concept of the partition of unity

method. In general, the partition of unity method is a powerful technique to model discontinuities

and singularities accomplished through local enrichment within a finite element setting.

Direct and iterative methods for solving of linear equations used to model crack growth

Modeling of damage accumulation and crack growth in steels by means of the FEM requires

solving systems of equations with sparse coefficient matrices, which contain relatively large

numbers of zero elements. Both direct as well as iterative solvers may be used to find a numerical

solution of a system of linear equations.

Dariusz Skibicki 235

The direct solvers allow to obtain a solution after a finite number of elementary operations. The

performance of elementary manipulations on coefficients and constant terms of equations

transforms the input original system of linear equations into an equivalent one by adding the two

equations from a system previously multiplied or divided by a nonzero number, by switching places

of two equations from a system or by multiplying both sides of one equation from a system by a

nonzero number. One of the direct methods for solving systems of linear equations, which is

equally simple, as well as accurate, is the method of Gaussian elimination.

The standard Gaussian elimination algorithm can be used to solve a general system of n linear

equations of the form: + - = , with n unknowns included in the vector - and n constant terms

included in the vector . The goal of elementary operations of the Gaussian elimination is to

transform or rather reduce the original matrix of coefficients, + = ./01 2 , into the upper

3×3

triangular matrix, 5 = .601 2 . Hence, the original system of equations, + - = , can be replaced

3×3

by an equivalent system, 5 - = 7, with the transformed vector of constant terms, 7.

In the first step of forward elimination, the first unknown 8 is eliminated from all n − 1

equations of the system by subtracting the multiplied first equation of the system from the 9-th

equation in order to zero the coefficient /0 assigned to 8 in the 9-th equation (for 9 = 2, … , n). The

subtraction of the relevant multiples of elements / 1 of the first row of + and the term = , which are

defined by quotients /0 ⁄/ , from the elements /01 of the 9-th row of + and term =0 can be

represented algebraically by the equations:

" >

/01 = /01 − > ?@ / 1 , (3a)

@@

" >

=0 = =0 − > ?@ = . (3b)

@@

In the second step, the second unknown 8 is eliminated from all n − 2 equations of the

transformed system, + " - " = "

, by means of the same type of manipulations as in the first

"

step. These manipulations are performed to elements /0 of the coefficient matrix + " (for

9 = 3, … , n), which are assigned to 8 .

Continuing this approach recursively the system of equations, 5 - = 7, with the triangular

coefficient matrix, 5, is obtained after n − 1 steps. A solution of the equation system, 5 - = 7, can

be determined by the inverse procedure:

" "

80 = >BC@ =0 3 − ∑31E0F /013 81 = G HI0 − ∑31E0F 601 81 J, (4)

?? ??

where: 9 = n, n − 1, … ,1.

The Gaussian elimination method is almost always effective if main diagonal elements /00 of +

are not equal to zero. Otherwise, its modifications are used with partial or full choice of pivot entry.

In the case of matrix computations, a pivot entry (i.e., an element of a matrix, which is selected first

by an algorithm) is usually required to be at least distinct from zero. The pivot element is the first

nonzero entry in the row, which does the elimination. The pivots are usually found along the

diagonal. Finding this element is called pivoting. Pivoting may be followed by an interchange of

rows or columns to bring the pivot to a fixed position and allow the algorithm to proceed

successfully, and possibly to reduce round-off error.

The direct method for solving systems of linear equations based on the Cholesky algorithm for

the decomposition of the matrix of coefficients, +, is often used. The Cholesky algorithm is a step-

by-step decomposition procedure for the symmetric positive-definite matrix, +, with real entries

into the product, + = K KL, of a lower triangular matrix K and its conjugate transpose KL .

The Cholesky decomposition is possible only if + is symmetric and positive definite. In this

case, the solution of the system of equations, + - = , can be obtained by first setting the vector M

by means of the formula: K M = , and then calculate the vector - using the system KL - = M.

236 Fatigue Failure and Fracture Mechanics

An alternative form of the Cholesky decomposition is the symmetric indefinite factorization [5]:

+ = KN O KLN . This form eliminates the need to take square roots. When + is positive definite the

elements of the diagonal matrix, O, are all positive. However, this factorization may be used for

most invertible symmetric matrices. An example of an invertible matrix, for which decomposition is

undefined, is one in that the first entry is zero.

The KN O KLN and K KL factorizations may be easily related by:

@ @ @ @ L @ @ L

+ = KN O KLN = KN OP OP KLN = KN OP OP KLN = KN OP KN OP = K KL. (5)

However, the main disadvantage of direct methods, which are used to solve a system of linear

equations with a sparse matrix, is the creation of new nonzero elements in this matrix during the

calculations. The new nonzero elements generated in individual items of the matrix + that have

been forced to zero before performing the arithmetic operations are called fill-in matrix elements.

The difference between direct and iterative types of algorithms for solving systems of linear

equations should be distinguished. Direct methods give the exact solution, - = + , after a finite

number of computational steps involving exact arithmetical operations. Iterative methods are based

on iterative improvements of an approximate solution until it reaches a satisfactory accuracy. The

processing performance of iterations begins with an initial approximation -Q .

For each iterative method, it is also necessary to define the criterion for terminating iterations,

usually in the form of the maximum permissible limit value for:

- the length of the line segment expressed by the norm: R- SF " − - S" R, between successive

approximations of a solution obtained in iterations S and S + 1, which, in Euclidean

geometry, allows to determine the distance between two points represented by the vectors - S"

i - SF " ,

- the vector norm given by the residual formula: RT SF " R = R − + - SF " R.

Iterative methods, which use exact arithmetic operations, does not give the exact solution after

a finite number of computational steps. It can be shown on the example of the iterative operator-

splitting method. The iterative operator-splitting method is based on the transformation of the

matrix + using mathematical operations defined by the formula: + = U + + − U". The system of

equations, + - = , can then be transformed as follows:

VU + + − U"W- = , (6a)

U - = U − +"- + , (6b)

- = U U − +"- + U . (6c)

Hence, the iterative method

"

- SF =U U−

XYYYZY +" - S" + U

YY[ ] = \- S" + ^ (7)

E\ E^

determines the series _- " , - " , - `" , … , - S" , … a, for which limS→e - S" = - = + .

Gauss may be counted as adherents of iterative methods for solving systems of linear equations

due to the fact that these methods are easier to perform than direct ones. In 1823 he wrote in his

native German-language a piece of text [6] that after translation into English sounds as:

The method of elimination is difficult to apply directly, at least when you have more than two

unknowns. Iterative methods can be carried out half asleep, or thinking about other things.

Large systems of equations, which are usually characterized by spare matrices of coefficients,

are solved mainly by using iterative methods. The required number of iterations can be reduced by

the use of the initial conditions called as preconditioning.

Assuming that if in Eq. 7 the positive-definite symmetric matrix, \, which approximates the

matrix +, is much easier to reverse than +, then instead of solving the system, + - = , it can be

considered the following system of equations: \ + - = \ . In particular, if \ = +, then the

Dariusz Skibicki 237

determination of a solution can be carried out only in one iteration because of + + = f. The

matrix \ is called a preconditioner for +. Selection of the matrix \ occurs as a result of

a compromise of the following requirements:

- preconditioner should be a sparse matrix of simple structure, in order to be easily reversed,

- preconditioner should have a structure as close as possible to the matrix +, in order to

accelerate the convergence of the algorithm as efficient as possible.

Very often there are used the following selection methods for the matrix \:

- diagonal scaling: g01 = h00 i01 (represented by the main diagonal of the matrix +) for 9 = j,

whereas g01 = 0 for 9 ≠ j,

- scaling method using the main diagonal and two adjacent minor diagonals: g01 = h01 for

|9 − j| ≤ 1, whereas g01 = 0 for |9 − j| > 1,

- incomplete Cholesky decomposition: \ = K∗ K∗L .

It should be noted that during the implementation of an incomplete Cholesky factorization as a

preconditioner \, the sparse matrix K∗ is only a certain approximation of a lower triangular matrix

K of the full Cholesky decomposition: + = K KL.

At present time, the most commonly used iterative method for symmetric positive definite

matrices is the conjugate gradient method. This method can be applied to solve systems of

equations with sparse matrices that may be too large to be handled by direct algorithms such as the

Cholesky decomposition. In addition, this method can also be used to solve optimization problems

without constraints, including determination of the minimum potential energy.

The method of conjugate gradients (CGs) was developed independently by Stiefel of the Institute

of Applied Mathematics at Zurich and Hestenes with the cooperation of Rosser, Forsythe and Paige

of the Institute for Numerical Analysis of the U.S. Bureau of Standards. The well-known form of

the algorithm of this method was developed jointly by Hestenes and Stiefel during a stay of the

second scientist at the U.S. Bureau of Standards [7].

The CG method can be best described by using the algorithm for minimizing of the following

square functional (which takes the quadratic form):

p -" = -L +- − L

-+7 (8)

for a particular set of vectors. If the matrix + is nonsingular, then the functional p -" can be written

as:

+ + 7. (9)

The functional p -" reaches a minimum only for the vector -, which is the solution of the system,

+- = . Hence, the designation of the minimum of the functional p -" is equivalent to solve the

system, +- = . If in Eq. 8 the matrix + is singular and its inverse + represents the generalized

inverse matrix, then, in consequence, the determination of the minimum of the functional p -" is

still equivalent to solve the system, +- = , for which minimization is performed in the orthogonal

complement of the null space of +.

For the algebraic residuum: q Q" = + - Q" − , which corresponds to the initial guess - Q" , the

CG method is carried out as the sequential line minimizations in a linear subspace of ℝ3 generated

by the images of q Q" under the first S powers of the square matrix + of degree n (starting from

+Q = s): q Q" , + q Q" , + q Q" , … , +S q Q" , i.e.,

tS H+, q Q" J = span_q Q" , + q Q" , + q Q" , … , +S q Q" a. (10)

The subspace determined by the formula (10) is known as the order- S Krylov subspace.

238 Fatigue Failure and Fracture Mechanics

The algorithm for finding a solution consists of looking for next approximate solutions along the

direction of the vector v S" according to the formula:

"

- SF = - S" + τS v S" , (11)

where for the S-th iteration the parameter τS is chosen in order to minimize the functional pH- S" +

τS v S" J given by Eq. 8, i.e.,

x x

v S" q S" v S" H+ - S" J

τS = S" x + v S"

= S" x + v S"

. (12)

v v

Moreover, the gradient of p in the point specified by the vector - SF " , which is equal to the

residuum q SF " , should be perpendicular to the direction of search v S" . This fact can be proven as

follows:

"

- SF = - S" + τS v S" ⇒ + - SF " − = + - S" − + τS + v S" ⇒

⇒ q SF " = + - SF " − = q S" + τS + v S" ⇒ (13)

S" L SF " S" L S" L

v q = v q S" + τS v + v S" = 0

During each iteration of the CG method the relatively simple procedure based on the product of

the matrix + by the vector v S" , which does not change the form of +, is used. In comparison to

direct methods, the CG method allows to reduce the computer memory requirements and to use

more effectively the specific structure of the sparse matrix. There are a number of variants of the

CG method. One of them is the biconjugate gradient method for asymmetric matrices [8].

Note that the conjugate residual (CR) method [9, 10] differs to the similar method of CGs

primarily in that the CR method is applicable for Hermitian matrices (i.e., symmetric matrices with

real coefficients), which should not be positively defined. This makes the CR method can be

implemented to deal with the numerical optimization problems using Lagrange multipliers in order

to find saddle-points of the objective function instead of minima, and during computer simulations

of the damage accumulation and cracking of strain-softening materials by mean of FEM.

Having regard to an initial (arbitrary) estimate of the solution of the system of equations, - Q" ,

the CR method may be characterized by the concise algorithm shown in Example 1.

Example 1. Program code of the CR method without preconditioning

- Q" = preliminary assessment of the solution (some initial guess)

T Q" : = − + - Q"

v Q" : = T Q"

9{ℯ6/{}, with S starting at 0:

S: = 0

6ℯ~ℯ•€

L

T S" + T S"

αS =

+ v S" "L + v S"

"

- SF = - S" + αS v S"

T SF " = T S" − αS + v S"

_convergence criterion for - SF " a

‰Š HT SF " is small enoughJ €Œℯ• exit loop ℯ•Ž ‰Š

"L "

T SF + T SF

βS = L

T S" + T S"

SF "

v = T SF " + βS v S"

+ v SF " = + T SF " + βS + v S"

S: = S + 1

ℯ•Ž •ℯ~ℯ•€

Dariusz Skibicki 239

Carrying out calculations by the CR method in comparison to the CG method requires the

introduction of an additional vector of the upgrading solution, and therefore the implementation of

2n more elementary arithmetic operations. The algorithm of the CR method with preconditioning to

improve the convergence of results can be written, as in Example 2.

Example 2. Program code of the CR method with preconditioning

- Q" = some initial guess

T Q" : = \ H − + - Q" J

v Q" : = T Q"

9{ℯ6/{}, with S starting at 0:

S: = 0

6ℯ~ℯ•€

L

T S" + T S"

αS =

+ v S" "L \ + v S"

"

- SF = - S" + αS v S"

T SF " = T S" − αS \ + v S"

_convergence criterion for - SF " a

‰Š HT SF " is small enoughJ €Œℯ• exit loop ℯ•Ž ‰Š

"L "

T SF + T SF

βS = L

T S" + T S"

SF "

v = T SF " + βS v S"

+ v SF " = + T SF " + βS + v S"

S: = S + 1

ℯ•Ž •ℯ~ℯ•€

The preconditioning matrix \ must be symmetrical.

Strain localization usually appears in narrow bands of material. It is identified with

heterogeneous forms of strain distribution in a place of the local loss of stability of material. The

mathematical description of this phenomenon is identified with ambiguity of a solution to the stress

equilibrium partial differential equations that results in a loss of ellipticity of these governing

differential equations. For strain localization problems, the acoustic tensor method can be applied to

only strain softening materials. If the phenomenon of strain localization is considered in fault zones

of materials exhibiting strain hardening and strain softening, then the vertex damage-coupled theory

developed by Stören and Rice [11] can be applied.

The elastoplastic theory for non-associated plastic strains (the yield surface does not coincide

with the plastic potential) shows that it is possible to fulfill the localization criterion (vanishing

value of the acoustic tensor determinant) before the plastic limit criterion (vanishing value of the

determinant of the elastoplastic matrix).

For certain stress states steels will exhibit narrow bands of intense deformation called shear

bands, since the deformation mode in these bands is usually shear. It is often claimed that shear

banding corresponds to a zero value of the determinant of the acoustic tensor. The eigenvalue

problem corresponds to the perturbed (i.e., linearized) equation of motion by means of the

governing equations for the continuum. Two general types of methods, namely transformation

methods and iterative methods, are available for solving eigenvalue problems. The transformation

methods, such as Jacobi, Givens and Householder schemes, are preferable when all eigenvalues and

eigenvectors are required. The iterative methods, such as power method, are preferable when few

and eigenvectors are required.

240 Fatigue Failure and Fracture Mechanics

Conclusions

1. Direct and iterative methods for solving systems of linear equations, which are intended

solely for the positive definite matrices of coefficients are unsuitable for modeling crack

growth in steels.

2. Parallel iterative solution method for large sparse linear equation systems should be preferred.

3. Checks for path stability need to be included in finite element programs for damage.

References

[1] J.P. Hirth, J. Lothe, Theory of Dislocations, 2nd Edition. Krieger Publishing Company, 1992.

[2] R.W. Hertzberg, R.P. Vinci, J.L. Hertzberg, Deformation and Fracture Mechanics of

Engineering Materials, 5th Edition. Wiley, 2012.

[3] E. Cuthill, J. McKee, Reducing the bandwidth of sparse symmetric matrices, in: Proc. 24th

ACM National Conference, Association for Computing Machinery, New York, 1969, 157–172.

[4] N. Moës, J. Dolbow, T. Belytschko, A finite element method for crack growth without

remeshing. Int. J. Numer. Methods Eng. 46 (1999) 131-150.

[5] J.A. George, J.W-H. Liu, Computer Solution of Large Sparse Positive Definite Systems,

Prentice-Hall, Englewood Cliffs NJ 1981.

[6] C.F. Gauss, Brief an Gerling vom 26 Dec. 1823, Werke, vol. 9, 278–281, in: Mathematical

Tables and Other Aids to Computation (a translation by G.E. Forsythe), vol. 5, 1950, 255–258.

[7] M.R. Hestenes, E. Stiefel, Methods of conjugate gradients for solving linear systems, Journal of

Research of the National Bureau of Standards 49 (1952) 409-436.

[8] W.H. Press, S.A. Teukolsky, W.T. Vetterling, B.P. Flannery, Section 2.7.6 Conjugate gradient

method for a sparse system, in: Numerical Recipes: The Art of Scientific Computing (3rd ed.),

Cambridge University Press, New York 2007, 87-92.

[9] Y. Saad, Iterative methods for sparse linear systems (2nd ed.), Society for Industrial and

Applied Mathematics, Philadelphia, PA 2003, 181–182.

[10] T. Sogabe, M. Sugihara, S.L. Zhang, An extension of the conjugate residual method to

nonsymmetric linear systems, Journal of Computational and Applied Mathematics 226 (2009)

103-113.

[11] S. Stören, J.R. Rice, Localized necking in thin sheets, Journal of the Mechanics and Physics of

Solids 23 (1975) 421-441.

Keywords Index

Fatigue Criteria 171, 189

A Fatigue Design 11

Fatigue Failure 118

Accelerated Method 11

Fatigue Life 27, 106, 171,

Acoustic Methods 55 181, 211

Ageing 3 Fatigue Life Estimation 189

Algorithm of Fatigue Calculations 17 Fatigue Life of Steel 17

Aluminum Alloy 63 Fatigue Monitoring 226

ARAMIS 110 Fatigue Strength 69

Fatigue Testing System 51

B Fatigue Tests 211

Bainite 55 FCC Steels 233

BCC Steel 233 Fem 100

Bearing Steels 55 Fictitions Radius 27

Block Loading 181 Fractography 171

Fracture Mechanics 233

C Fracture Toughness 195

Friction Stir Processing 203

Carbon Steel 203

Clad 106

Crack Growth Analysis 218 H

Cracks 222 Heat Affected Zone 110

Cyclic Fatigue 43 High-Chromium Cast Steel 3

Cyclic Hardening 63 High-Cycle Fatigue 11, 63

Cyclic Material Properties 51 Hybrid Method 222

Cyclic Properties 3, 150 Hysteresis Loop 143

D I

Diagrams of Fatigue Life 77 In-Plane Constraint 195

Digital Image Correlation 218 Infrared Thermography 156, 162

Direct Solvers 233 Initial Stiffness Modulus 84

Dual-Band System 222 Isothermal Heat Treatment 55

Dual-Camera System 226 Iterative Solver 233

E L

Efficient Material 118 Laser Weld 100

Explosive Cladding 106 Laser Welding 133

Explosive Welding 125 Linear Hypothesis of Fatigue 39

Damage Accumulation

Local Stress Approach 100

F

Low Cycle Fatigue 3, 77, 143,

Failure Probability Analysis 118 150

Fatigue 33, 51, 100, Low Cycles Fatigue Life 93

106, 125

Fatigue Characteristics 43, 143, 181

242 Fatigue Failure and Fracture Mechanics

M Two-Parametric Fatigue 69

Characteristics

Martensitic Cast Steel 150

Mean Stress 33

U

Microspecimen 51

Midrib 55 Undermaching 110

Mini Specimen 63, 156

Multi-Processor Graphic Cards 218 V

Multiaxial Fatigue 162, 171, 189 Vision System 226

Multilayer Pipe 133

W

N Weakest Link Concept 118

Non-Proportional Loading 189 Welded Joint 93, 118

Nonproportional Load 162 Welds 110

Whole-Field Displacement 218

O Analysis

Out-Of-Parallelism 181

Out-of-Plane Constraint 195

P

Peierls Stress 233

Pitting 55

Programmed Fatigue Testings 17

R

Random Loading 17

Ratcheting 125

Residual Stress 125

Riveted Joints 211

Rolling Contact Fatigue 55

S

S-N Curve 11

Section Method 33

Squeeze Force 211

Steel Sandwich Panel 100

Stepwise Loading 39, 84

Strain Localization 233

Strength 133

Strength Properties 156

Stress Ratio 33

T

Thermal Stress 125

Thermo Mechanical Fatigue 143

Thermography 222

Thickness Effect 211

Authors Index

B

Blacha, Ł. 118 M

Böhm, M. 33 Machniewicz, T. 211

Boroński, D. 51, 156, 218, Marciniak, T. 218, 222

222 Marcisz, E. 43

Bujnowski, S. 218, 222 Marecki, P. 93

Mazurkiewicz, A. 39, 84

C Mrozinski, S. 150

Cichański, A. 39, 84 Mroziński, S. 3, 133

Czajka, P. 222

N

G Neimitz, A. 195

Gałkiewicz, J. 110 Niesłony, A. 33, 43, 106

Giesko, T. 218, 226 Niklas, K. 100

Golański, G. 3 Nowicki, K. 39, 84

Goss, C. 93 Nykyforchyn, H. 203

H O

Hutsaylyuk, V. 203 Okrajni, J. 143

J P

Jackiewicz, J. 233 Pejkowski, Ł. 171, 189

Jelenkowski, J. 55 Piotrowski, M. 133

Junak, G. 143

R

K Ranachowski, Z. 55

Karolczuk, A. 118, 125 Robak, G. 27, 125

Kluger, K. 125 Rozniatowski, K. 55

Kocańda, D. 203

Korbel, A. 211 S

Kowalski, M. 125 Sempruch, J. 11, 63, 171

Kozak, J. 100 Skibicki, D. 162, 171, 189

Kurek, A. 106 Skocki, R. 150

Kurek, M. 181 Skorupa, A. 211

Kyryliv, V. 203 Skorupa, M. 211

Slezak, T. 203

L Strzelecki, P. 11

Łagoda, T. 27, 43, 181 Szala, G. 17, 69, 77

Ligaj, B. 17, 69, 77 Szymaniec, M. 27

Lipski, A. 156, 162

Lutowski, Z. 218, 222 T

Tomaszewski, T. 63

Topoliński, T. 39, 84

244 Fatigue Failure and Fracture Mechanics

Torzewski, J. 203

W

Werner, K. 3

Woźniak, T.Z. 55

Z

Żok, F. 125

- Creep and FatigueЗагружено:Ayush Shah
- Din743 Intro EnglishЗагружено:Uber_Konstruktor
- 2004-4052S1_12_MorrisЗагружено:Noor hossain
- 1979030Загружено:alinaauli
- LifeFrequency ManualЗагружено:Egwuatu Uchenna
- 2_1080wwwwwЗагружено:ankitsarvaiya
- Stress Strain StuffЗагружено:Adhithya Bhat
- Full TextЗагружено:irfanajai
- Cycling LoadingЗагружено:asmauzum
- Lecture 8 Fatigue 2 2011Загружено:Shantanu Mane
- Sample Paper Wind Eng gЗагружено:JsChahal
- 2 Introduction RCЗагружено:Hundeejireenya
- A 344374Загружено:Jesus Soares
- Introduction to the Mechanical Behavior of MetalsЗагружено:saborsita666
- 12 macuta 5Загружено:Sergiu Gheorghe
- Flaws Cg 2006Загружено:adiraju07
- 06552600.pdfЗагружено:Heather Smith
- cyclic_loading indefinite.pdfЗагружено:reky
- Lecture Fatigue.pdfЗагружено:riadmorina
- Efm 2011 Luke RwaЗагружено:jenmaxius
- DAPUSЗагружено:Aditya Nur Hidayat
- CH385Загружено:rsbd22
- ASME AcceptanceЗагружено:Steelera Engineers
- Fatigue Life Prediction of Pultruded E-glassPolyurЗагружено:Arturo Medina
- Helicopter component fatigue.pdfЗагружено:Etienne Gerber
- dokumen.tips_basic-drill-string-design.pdfЗагружено:reza kh
- FATIGUE Analysis NotesЗагружено:NITESH BORKAR
- 6bc007369b4e1933a000daa0226652197e45Загружено:preetham sai
- Aspects of Robust Computational Modeling for Plain and Reinforced ConcreteЗагружено:Marcelo Carvalho
- Paper Fatique CompositeЗагружено:Atul Gupta

- International Journal of Fatigue, Volume 113, August 2018, Pages 311-323Загружено:ikhsan febriyan
- 01_What is Engineering DesignЗагружено:ikhsan febriyan
- print uyЗагружено:ikhsan febriyan
- Fans and BlowerЗагружено:Mamoun1969
- 1-s2.0-S2214785318323411-mainЗагружено:ikhsan febriyan
- Thermodynamics Table - Cengel.pdfЗагружено:Fachransjah Aliunir
- Daftar Pustaka pkm.docxЗагружено:ikhsan febriyan
- (International Centre for Mechanical Sciences 275) H. P. Rossmanith (eds.) - Rock Fracture Mechanics-Springer-Verlag Wien (1983).pdfЗагружено:ikhsan febriyan
- Fatigue and Crack Initiation - George C. SihЗагружено:VAIBHAV AHER
- 1-s2.0-S0022460X18308587-mainЗагружено:ikhsan febriyan
- 1-s2.0-S0022460X15000905-mainЗагружено:ikhsan febriyan
- 1-s2.0-S0022460X15000905-main.pdfЗагружено:ikhsan febriyan
- (Solid Mechanics and Its Applications) George C. Sih, George C. Sih - Multiscale Fatigue Crack Initiation and Propagation of Engineering Materials_ Structural Integrity and Microstructural WorthinessЗагружено:ikhsan febriyan
- (Conference Proceedings of the Society for Experimental Mechanics Series) Beese, Alison M._ Berke, Ryan B._ Carroll, Jay_ Pataky, Garrett J._ Xia, Shuman - Fracture, Fatigue, Failure and Damage EvolutЗагружено:ikhsan febriyan
- VibrationsЗагружено:leanindustry
- 02 Method for Describing Sets of DataЗагружено:ikhsan febriyan
- DAFTAR PUSTAKAЗагружено:ikhsan febriyan
- present TM 17-2Загружено:ikhsan febriyan
- Mechanical Vibrations Tse Morse HinkleЗагружено:Harshit Dwivedi
- mod1.pdfЗагружено:Ahmad Budiman
- Lecture 34Загружено:shah_gen89
- Classification of Heat ExchangersЗагружено:4605566vivek
- PR1Загружено:ikhsan febriyan
- (Tugas 3 2017)ChopperЗагружено:ikhsan febriyan
- 4Загружено:ikhsan febriyan
- Chap_26Загружено:ikhsan febriyan

- What Makes a Biomedical EngineerЗагружено:Andreea Zimta
- F-86E&F Pilot's ChecklistЗагружено:huntting
- lesson plan 1 decemberЗагружено:api-270220688
- (Steven Steele) Inner Reflections InterpretationЗагружено:Steven Steele
- Mouth FreshnersЗагружено:King Nitin Agnihotri
- skema_engelsk(university qualification performa)Загружено:Meme Khan
- Elastic Bend Calculation-CR4Загружено:Sankar Cdm
- Why 0.6W.pdfЗагружено:Hyunkyoun Jin
- Cagayan de Oro CityЗагружено:Rosemarie Jacaban
- Data Dictionary MP2 SQL 6.0Загружено:Juan Solis
- STI 6931 Data SheetЗагружено:JMAC Supply
- gems international school al khail parent letter from peter milneЗагружено:api-237720707
- XYZs of Using a ScopeЗагружено:ee7ux
- 907Загружено:arunkanchu
- R5210303 Electrical EngineeringЗагружено:sivabharathamurthy
- Gravity and Circular Motion PracticeЗагружено:nanio_7
- Dino FlagellateЗагружено:x456456456x
- Value and Ethics in BusinessЗагружено:Zahir Kamal
- Opal 1000 Series Dtasht[1]Загружено:Armando Ramirez
- Steel Roof Truss TutorialЗагружено:ranto
- Supply Chain -ABCЗагружено:Yulia Tri Anggani
- Ats Routes ChinaЗагружено:Paco Estrada
- Getting Started Manuf Process MgmtЗагружено:irina
- ScampiЗагружено:shekharmajumdar

## Гораздо больше, чем просто документы.

Откройте для себя все, что может предложить Scribd, включая книги и аудиокниги от крупных издательств.

Отменить можно в любой момент.