Professional Documents
Culture Documents
00
Printed in Great Britain. All rights reserved Copyright 0 1988 Fatigue of Engineering Materials Ltd
Abstract-A non-linear cumulative fatigue damage model proposed previously is applied to different steels
and various loading situations, including two-level tests and block-programs. Its ability to describe all the
main features of fatigue damage and the ease of its practical use for engineers are discussed and several
examples cited. The relationship with other formulations are pointed out together with the main
advantages of the proposed model. A generalization is proposed for both cyclic temperature and
multiaxial loading conditions.
NOMENCLATURE
D = damage variable
D = damage variable consistant with Continuum Damage Mechanics
D, = damage value at micro-crack initiation
uM, 5 = maximum and mean tensile stresses
Acp = plastic strain range
q = plastic strain memory variable (model with interactions)
L =crack length
N = current number of cycles
NF.Ni = numbers of cycles to failure and to micro-initiation
n,= number of cycles in block “i” of a block program
a( ), M( ), u,( )=functions in the NLCD model
ulo= fatigue limit under reversed stress conditions
o, = ultimate tensile stress
b. M,,a, b = coefficients of the NLCD model
y, C = coefficients of the power fitting of S-N curves
p = exponent of non-linearity in the two-level test predictions.
-=
( ) = MacCauley bracket symbol. ( u ) = 0 if u 0, ( u ) = u if u > 0.
INTRODUCTION
The life of metallic structures is very often governed by the fatigue process caused by vibratory
loading conditions. A large part of the life is related to a sequence of processes during which slip
bands, localized strains, damage and microcracks initiate and develop until some macroscopic
crack initiates. After that stage crack propagation is much more rapid and is usually taken into
account through Fracture Mechanics analyses.
Non-linear cumulative fatigue damage theories were developed a long time ago and include the
works of Henry [I], Gatts [2], Marco and Starkey [3], Manson [4] and many others. They are based
either on the separation of fatigue life into two periods (initiation and propagation) [5] on the
progressive decrease of fatigue limit [6],or on remaining life and continuous damage concepts [7].
The present paper reviews the main features of a model proposed in 1974 [7] to describe
non-linear cumulative fatigue damage. It is called here the Non-Linear-Continuous-Damage
(NLCD) model. The formulation is supported by Continuum Damage Mechanics [7,8] but its use
1
2 and P. M. LESNE
J. L. CHABOCHE
under pure fatigue is mainly related to remaining life aspects. The model has also been used in high
temperature situations. Its form allows for the cumulative effect of creep and fatigue damaging
processes [9, 101.
The considered approach generalizes the model of Marco and Starkey [3] and the Damage curve
approach of Manson [l I]. The connections with other approaches will be pointed out together with
some detailed examples on different materials and for various loading conditions, including
two-level tests and block-programs. Also the cases of varying temperature conditions will be
considered.
Let us note that the model is developed as an engineering tool, no more difficult to use than
the simple linear rule. As a consequency, the physical microprocesses, involving nucleation,
initiation and short crack growth, are not considered in detail.
(g) Integrating equation (2) for constant dM and 5, between D = O and D = 1, leads to ( D = 1
for N = NF):
(h) Due to the dependence between a and the loading parameters the damage evolution curves
as a function of the life ratio N/NF,depend on oM and 5. As it is classical, this dependence leads
to the non-linear accumulation and allows the description of sequence effects. For instance,
considering two-level tests, one finds -by integrating (2) in two steps:
1 - a2
where NF,and NF,are the failure lives (on the S-N curve) for the two loading conditions, N , is
the number of cycles at the first level, N2 the remaining life at the second level.
(i) The function M ( E ) is chosen so as to describe the well-known linear dependency [14], between
mean stress and the fatigue limit (see below). Here, in terms of an equivalent amplitude, one writes:
M ( 5 ) = Mo( 1 - b5) (6)
where Mo and b, as well as exponent p in (2), are coefficients depending on the material.
where 0 , is the ultimate tensile strength, and y is the slope of the S-N curve (reversed stress
conditions) in the intermediate region:
NF = COG’ (8)
a, B, and C = a’,/a are coefficients. The correlationp = 0.55 y was satisfactory for many materials
[7]. Figure 1 gives the example of superalloy IN 100 used in turbine blades.
Identification of the NLCD model is easier if the form of function a is specified in the following
way [9], which corresponds to the presently used formulation:
4 and P. M. LESNE
J. L. CHABOCHE
0 250 500
Maw stress aM MPa
Fig. 1. Values of a measured by the effective stress concept and by fitting various relationships for
IN100 at 900 and 1000°C.
where oi0is the fatigue limit for fully reversed conditions. o,(Z) is the fatigue limit for a non-zero
mean-stress. It is expressed in terms of the maximum stress. Symbol ( ), defined as ( u ) = 0 if
u c 0, ( u ) = u if u > 0, gives rise to o! = 1 when the maximum stress is lower than the fatigue limit
a,(Z). With function (9), the number of cycles to failure is calculated as:
o,,-nM [.M-Z]-’
NF =
a(bM - M(Z)
In the two-level test condition, the remaining life is predicted by ( 5 ) with the exponent:
-
p =-- - b M 2 - oi(Z2) Ou - OM,
I - xi C M i - ollfZl) Ou - oM2
The ultimate tensile stress o,,plays no role in the fatigue process under reversed and high cycle
conditions. It is used here as a normalizing parameter, which is able to reproduce the asymptotic
shape of the S-N curve in the regime of very short lives, especially under repeated conditions. The
relation (1 1) shows the two limiting cases of NF = 0 for oM= o,, and NF = 00 for 0, = q(Z). It
describes correctly S-N curves for different mean-stresses within a large range of cycles
10 c NF < lo’. Figure 2 gives a typical example. Cumulative effects are predicted by only specifying
the fatigue limit and the ultimate tensile strength q,.The symmetry between High-Low and
Low-High loading conditions, as shown by examples described later, should be especially noted.
Such symmetry is evident from the damage rate equation (1).
A 517
m
5 I A201
I
100
10 10’ 103 104 10s
and
In Ref. [ll], Manson uses 0.4 instead of exponent 0.45. The equivalence between the present
approach and the DC model is clear for loads higher than the fatigue limit, even though the DC
model introduces differently the damage growth equation [ll]. As pointed out in reference [14],
equation (14) presents some limitations. One advantage of the present approach is to consider the
damage produced by loading cycles below the fatigue limit (see later).
Let us mention also the method proposed by Subramanyan [15], defining a knee-point on the
S-N curve in the region of the fatigue limit. In the two-levels situation, this method predicts (5 = 0):
where NY corresponds to the “knee point” on the S-N curve, with various definitions [15]. These
expressions are equivalent for a S-N curve described by a power relationship. They give rise, for
the two-level condition, respectively to:
6 J. L. CHABOCHE
and P. M. LESNE
Steel (2.35
0
1
275 MPa
334 MPa
Fig. 3. Prediction ofthe two-level test results for the C-35 steel with different a functions.Data from Ref. [HI.
Figure 3 shows the comparison of the different predictions for two-level tests on a C-35 steel
[15] and demonstrates that the choice of equation (1 1) slightly overpredicts the non-linearity of the
accumulation.
Other non-linear cumulative rules are based on the decomposition of fatigue process into crack
initiation (Ni) and crack propagation stages. Reference [5] uses an exponential accumulation rule
in the two regimes but, due to the low values taken for Ni/NF(Ni/NF= 1.421. lop3NOF.42'), the model
behaves mainly in the propagation regime. In that case the remaining life can be calculated at the
second level by an equation similar to (1 5), with
log(&, 10g(q,/NF,)
P= (19)
lOg(Li,l&) W K , /NF,1
where Ni is the number of cycles to micro-crack initiation, Lithe corresponding crack length (which
depends on the life). Lf is the crack length at failure.
In Refs [4] and [19] the accumulation of damage is assumed linear during each of the two phases.
They lead to different versions of a Double Linear Damage type of rule. As pointed out in Refs
A non-linear continuous fatigue damage model 7
0.5
Di
0
&Ni,/NF,-
Fig. 4. Schematic comparison of the NLCD and the DLDR approaches to the evolution of damage.
[ 1 11 and [20], there exists qualitative and quantitative correspondences between the two approaches.
As shown schematically in Fig. 4, in the DLDR rule the life ratio to crack initiation depends, for
example, on the total life, leading to sequence effects. In the continuous damage models, it is the
shape and the non-uniqueness of the damage evolution that dictates sequence effects. The
correspondence between the two types of approach is possible by considering a critical damage Di
(with a low constant value). On each continuous damage curve, Di corresponds to the micro-crack
initiation (see Fig. 4). From equation (3) one finds the following relation, where the function a(&)
is defined either by equations (7), (9), (13), (16), or (17):
Ni = NFD!-a(N~) (20)
The damage accumulation models based on the progressive decrease of fatigue limit [l, 2,6]
cannot be deduced explicitly from the present approach. Here, the decrease of fatigue limit (or its
complete annihilation) is only a consequence of the damage rate equation. Also the model proposed
by Kramer [21] cannot be written in a differential form similar to (1). In that model, there is no
symmetry between Low-High and High-Low two-level test conditions.
A :88MPa
Fig. 5. Prediction of two-level test results on 4340 steel using the NLCD rule and equation (16). Data
from Ref. [22].
importance for fatigue damage accumulation, even under complex block programs, as shown later.
If necessary the determination can be completed by using damage measurements such as:
(i) The number of cycles to micro-crack initiation N,,with an arbitrary associated damage
value Di, can yield an average value for coefficient Q, using together equations (19) and
(9).
(ii) The value of a can be obtained from microstructural damage measurements, using for
instance the surface crack length L, normalized by a reference value & as in Refs [23] and
[24], D = L / L f .
(iii) Measurements of damage can be obtained from the change in the stress-strain response
during the fatigue process, using the effective stress concept and Continuum Damage
Mechanics [8].
The measurements obtained from changes in the mechanical response suggested a different form
for the damage rate equation [7]. The damage variable associated to such measurements is called
D* and the rate equation writes:
Clearly, due to the limit values 0 and 1 for D* as well as for D, there is a one-to-one
correspondence between equations (2) and (21), with:
D = 1 - (1 - D*)B+I (22)
A non-linear continuous fatigue damage model 9
do
Fig. 6. Damage evolution curves as measured for IN 100 from the effective stress concept; prediction by
the NLCD model and equation (23).
Therefore the two theories give rise to exactly the same response under any fatigue loading. The
damage evolution under periodic loading becomes:
In several works [19,29,30], the short crack regime is described in terms of a crack growth rate
equation like
dL
- = BL"At,B
dN
where L is the crack length, Acp the plastic strain range. It is interesting to note the use of a global
plastic strain measure (not the strain at the crack tip) which can be justified only in the frame of
the Continuum Damage Mechanics (with multiple and homogeneously distributed micro cracks).
Moreover, the similarities between equations (24) and (2) are evident. Let us remark that, below
the fatigue limit the exponent a of equation (2) is equal to 1, which is the value for a often used
in equation (24).
APPLICATIONS
For two -level loading conditions
This is a direct application of equation (5). The non-linearity of damage accumulation is correctly
predicted, as shown in several materials [7, 11, 16, 171.
One advantage of the NLCD model is to allow the growth of damage below the initial fatigue
limit when the material is submitted to prior cycling above the fatigue limit. In the case of a
two-level loading, integration of the model (with a = 1 under the initial fatigue limit) leads to the
remaining life at the second level (below the fatigue limit):
Clearly the fatigue limit disappears, even for low initial damage. This effect is observed in the
experiments, as shown in Refs [7], [26] and [27].
Block programs
Fatigue under block programs very often shows life-ratio summations markedly lower than
unity. It is a common feature of the NLCD model and other models discussed previously to predict
correctly this situation. Figure 7 shows the example of Maraging steel with two blocks per sequence
(data from Ref. [4]).
1 Z(n/NF)
t
A Block-program
o Two-levels
Miner's rule -Calculation
base block :
0.5-
0
2 blocks
44 4 1 block 1300 88OOO
'3 ' I
0
I I
1 2 3
n, /1300 = n2/88M)O
4
I
5
Fraction of the "base block"
I-
Fig. 7. Fatigue under block loading programmes and the prediction of life summation using the NLCD
model; Maraging steel, data from Ref. [4].
A non-linear continuous fatigue damage model 11
Let us note that predicting fatigue under block-program is very easy through sequential
calculations [7, 111. Such calculations are manageable with a pocket computer, using some variable
change:
y = D!-% (26)
where Di is the value of damage at the end of the ith block in the program, with the associated
value ai for a. In that block ni cycles are applied. In the case of loadings above the fatigue limit,
integration of (2) easily leads to:
For a block 'T' under the fatigue limit integration is a little more difficult. With tl = 1, one finds:
Di = Di_lexp(ni/N*) (28)
where NT denotes a fictitious reference number of cycles to failure:
After some manipulations, damage accumulation for the next block above the fatigue limit gives:
1 1
J. L. CHABOCHE
and P. M. LESNE
ol,::‘[
Tests
Calculations
x
+
(I-
RU=
Re=-1
R,=O
-- 1
- 0.485
In such cases, the damage rate equation (1) is insufficient. An additional parameter plays role
and one can write:
dD =f (oM,5,Dq) d N (31)
where q is for instance a measure of prior hardening. An explicit model was proposed in Ref. [34]
and applied to the 316 stainless steel, where q is a memory of maximum plastic strain range:
AcpM
q=-
2
Figure 8 shows the possibilities of the model for experimental repeated conditions under stress
control. Let us note the marked difference by comparing to other steels in Fig. 2. The two level
test conditions, with an improved life at the second level was also correctly predicted due to the
higher hardening during the first level [34].
In the common Low-Cycle Fatigue domain, equation (31) corresponds to a certain extent to a
model with a mixed parametrization, stress oM and strain range A$, as in other approaches [35,36].
The model taking into account the interaction with hardening is more complex to determine and
to use, due to the need of some constitutive equations (taking into account complex hardening
memorization). In fact, it predicts situations similar to the ones described by damage models “with
interaction” [21,37], where the damage evolution at the second level is modified by the initial
loading. Figure 9 is a schematic of the behaviour of equations of type (3 1) for a two-level loading:
the internal state at the beginning of the second level depends not only on the present damage but
also on the hardening state, which is different after a high level. Let us note that models with
interaction such as in Refs [21] or 1371 need also many additional experimental data.
t"
of 316 L stainless steel, taken from reference [38] was treated with:
with using t, and t , as empirical parameters (here t , = 0.15%, cu = 3%). Figure 10(a) shows the
two-level test conditions and Fig. 10(b) reports both predictions with the linear summation and
with the non-linear one for block-programs (in that case the level change was controlled in order
to maintain approximately a zero mean-stress for each level). The Marco-starkey model, used in
Ref [38], and the present model [Fig. 10(b)] improve the prediction by a factor 5 as compared with
the linear rule.
The situation under stress concentration is of a large practical interest. In that case the local
stresses (maximum and mean stresses) can be obtained through the use of cyclic constitutive
equations and a cycle by cycle analysis following the simplified Neuber's rule [39,40]. Due to stress
redistributions, the non-linearity of damage accumulation is less pronounced 1381, but the
predictions made with the NLCD model agree fairly well with experiments.
T* is defined as the temperature (constant) for which the present strain range At produces the same
number of cycles of failure N $ as the one calculated by relation (33). In the second member, NF
is defined, for each temperature at each instant t, from conventional isothermal data.
F F . E . M S 1111 I
14 J. L. CHABOCHE
and P. M. LFXNE
1 % '0.3%
0 1% 0.2%
+
'-\
Fig. 10. Predictions of fatigue life under two-level tests on 316 stainless steel using the modified NLCD
model. (a) Comparison with a linear rule. (b) Fractional life plots; data from Ref. [34].
For the present non-linear damage model the above equation (33) is still applicable, considering
linear accumulation only for the step of defining the effective temperature T*:
NF is defined for isothermal conditions by equation (1 1). After determining T* from the reciprocal
of equation (34), the damage rate equation can be applied simply as:
dD =f(aM, 8, D,T*) dN (35)
This method has been used systematically for the turbine blade superalloy IN 100, taking into
account both fatigue damage under varying temperature and creep damage [9,10,42].
Multiaxial conditions
The Non-Linear-Continuous-Damage model was written in its uniaxial form, based on
A non-linear continuous fatigue damage model 15
maximum and mean stresses. Under multiaxial loading conditions, three successive problems are
encountered:
(i) The definition of the multiaxial stress parameters associated to the maximum and mean
values, for instance under proportional loading.
(ii) The checking of these parameters and of the damage rate equation for non proportional
loading conditions, considering a scalar damage variable (isotropic damage).
(iii) The anisotropic fatigue damage, that is its directionality, which can influence the damage
growth under multilevel conditions with changing the loading direction. The out-of-phase
loading conditions is one of the problems.
The generalization of the NLCD model to multiaxial conditions has been proposed in references
[43] and [lo], and some partial experimental checking has been done in [36]. The theory is based
on equation (2) and a judicious choice for stress invariants corresponding to the two limiting cases
(fatigue limit criterion N + co,and monotonic rupture N + 0). This generalization is not discussed
in the present paper.
CONCLUSIONS
The considered non-linear continuous fatigue damage model (NLCD) describes adequately the
various phases of the deterioration processes, including, in a continuous form, the microinitiation
and micropropagation stages. It is developed for design purposes and predicts correctly:
(i) The one level or two-level stress controlled fatigue tests on smooth specimens.
(ii) The influence of mean stress for the whole domain of number of cycles.
(iii) The remaining life, even under the initial fatigue limit.
(iv) The block program loading conditions, where Miner’s rule always overpredicts the life.
(v) Strain-controlled fatigue tests (even in the Low-Cycle-Fatiguerange), including two-levels
and block program tests.
Several similarities with other rules have been pointed-out but the developed model presents
some specific advantages; namely:
It describes the non-linear accumulation, even when changing the mean stress.
There are no rate-discontinuities as in the models based on the separation between
initiation and propagation.
It describes the whole S-N curve (10 to lo’ cycles) with a unique set of coefficients,
including the effect of mean-stress.
The damage growth below the initial fatigue limit after prior damage, which is not the case
with some other models.
Material parameters are determined easily from the S-N curve and the model can be used
for complex programs without difficulty.
Interaction effects can be taken into account, using an additional hardening variable.
Fatigue damage can be combined in a natural way with creep damage, under constant or
varying temperature.
A multiaxial generalization can be formulated, combining multiaxial fatigue-limit criteria
to the Von-Mises equivalent stress.
Concerning the physical basis of the model, some deficiencies have to be pointed out:
(i) The definition of macro-crack initiation is conventional and some correction to the total
rupture life has to be done in order to substract the final propagation, phase.
16 J. L. CHABOCHE
and P. M. LESNE
(ii) There is no distinction between the rate equation for the initiation regime and the rate
equation in the case of micropropagation.
(iii) The fatigue limit disappears after an amount of initial damage produced by higher stress
levels.
(iv) The behaviour of short cracks below the initial fatigue limit is described only in an
approximate global way, but the classical equations used in a Fracture-Mechanics-based
approach have the same form.
(v) The Continuum Damage Mechanics approach which supports the above mentioned global
measure of damage, presents two kinds of deficiencies: the surface character of fatigue
damage and the fact that during the propagation period the number of large defects is
small which is inconsistent with a continuum approach.
However, due to the above mentioned correct description of many experimental results in terms
of cumulative fatigue damage, the NLCD model constitutes a good and simple approach. For
practical engineering applications, it is in fact as simple to use as the linear Palgreem-Miner’s rule
is.
Applicability of the Non-Linear Continuous Damage model to fatigue life prediction for stress
concentration problems is discussed in Ref. [40].Further researches will consider both the checking
of the multiaxial form and its generalization for out-of-phase loading conditions.
REFERENCES
1. Henry D. L. (1955) A theory of fatigue damage accumulation in steel. Trans. ASME 77, 913-918.
2. Gatts R. R. (1961) Application of a cumulative damage concept to fatigue. J. Basic Engng, Trans. ASME
83. 529-540.
3. Marco S. M. and Starkey W. L. (1954) A concept of fatigue damage. Trans. ASME 76(4), 627-632.
4. Manson S. S., Freche J. C. and Ensign C. R. (1967) Application of a double linear damage rule to
cumulative fatigue. NASA TN D 3839.
5. Miller K. J. and Zachariah K. P. (1977) Cumulative damage laws for fatigue crack initiation and stage
1 propagation. J . Strain Analysis 12(4), 262-270.
6. Bui Quoc T., Dubuc J., Bazergui A. and Biron A. (1971) Cumulative fatigue damage under stress-
controlled conditions. J. Basic Engng, Trans. ASME 93, 691-698.
7. Chaboche J. L. (1974) Une loi diffkrentielle d’endommagement de fatigue avec cumulation non-lineaire.
Rev. Fr. M&. No. 5&51, pp. 71-82. English trans]. in Muter. Building Res., Ann de I’ITBTP 39 (1977),
pp. 117-124.
8. Chaboche J. L. (1981) Continuous damage mechanics. A tool to describe phenomena before crack
initiation. Nucl. Engng Design 233-247.
9. Lemaitre J. and Chaboche J. L. (1978) Aspect phenomtnologique de la rupture par endommagement. J.
M ~ c Appl.
. 2(3), 317-365.
10. Lemaitre J. and Chaboche J. L. (1985) La MPcanique des MatPriaux Solides. Dunod, Paris.
11. Manson S. S. and Halford G . R. (1981) Practical implementation of the double linear damage rule and
damage curve approach for treating cumulative fatigue damage. Znt. J . Fract. 17(2), 169-192.
12. Levaillant C. (1984) Approche metallographique de I’endommagement d‘aciers inoxydables austenitiques
sollicites en fatigue plastique ou en fluage: description et interprktation physique des interactions
fatigue-fluage-oxydation. Thkse d’etat, UTC.
13. Ostergren W. J. and Krempl E. (1979) An uniaxial damage accumulation law for time-varying loading
including creep-fatigue interaction. J. Press. Vess. Technol. Trans. ASME 101, 1 18.
14. Miller K. J. and Ibrahim M. F. E. (1981) Damage accumulation during initiation and short crack growth
regimes. Fatigue Engng Muter. Struct. 4(3), 263-278.
15. Subramanyan S. (1976) A cumulative damage rule based on the knee-point of the S-N curve. Trans.
ASME, J . Muter. Technol. 316-321.
A non-linear continuous fatigue damage model 17
16. Hashin Z. and Rotem A. (1978) A cumulative damage theory of fatigue failure. Mater. Sci. Engng 34,
147- 1 60.
17. Hashin Z. and Laird C. (1980) Cumulative damage under two-level cycling. Fatigue Engng Mater. Struct.
2, 345-350.
18. Kujawski D. and Ellyin F. (1984) A cumulative damage theory-fatigue crack initiation and propagation.
Int. J. Fatigue. 6(2), 83-88.
19. Ibrahim M. F. E. and Miller K. J. (1980) Determination of fatigue crack initiation life. Fatigue Engng.
Mater. Struct. 2, 351-360.
20. Cailletaud G. and Levaillant C. (1984) Creep-fatigue life prediction: what about initiation? Nucl. Etzgng
Design No. 83, 279-292.
21. Kramer I. R. (1976) Prediction of fatigue damage. Proc. 2nd Int. Con$ on the Mechanical Behaiiior of
Materials, Boston, ASTM, pp. 8 12-8 16.
22. Erickson W. H. and Clyde E. W. (1961) A study of the accumulation of fatigue damage in steel. 64th
Annual Meeting of A S T M .
23. Socie D. F., Fash J. W. and Leckie F. A. (1983) A continuum damage model for fatigue analysis of cast
iron. A S M E Conf. “Advances in life Prediction”, Albany, N.Y.
24. Hua C. T. and Socie D. F. (1984) Fatigue damage in 1045 steel under constant amplitude biaxial loading.
Fatigue Engng Mater. Struct. 7(3), 165-179.
25. Plumtree A. and Nilsson J. 0. (1986) Damage Mechanics applied to high temperature fatigue. Int. Spring
Meeting “Fatigue at High Temperature ”, Paris.
26. De 10s Rios E. R., Tang Z. and Miller K. J. (1984) Short crack fatigue behaviour in a medium carbon
steel. Fatigue Engng Mater. Struct. 7(2), 97-108.
27. Lesne P. M. and Cailletaud G . (1987) Creep fatigue interaction under high frequency loading. ICM5,
Beijing, China, pp. 1053-1061.
28. Miller K. J. (1982) Elastic-plastic fracture mechanics of short fatigue cracks. “Mechanical and Thermal
Behaviour of Metallic Materials”, SOC.Italiana di Fisica, Bologne, Italy, pp. 176187.
29. Manson S . S . (1966) Interfaces between fatigue, creep and fracture. Int. J. Fract. Mech. 2(1), 327-363.
30. Tompkins B. (1968) Fatigue crack propagation-an analysis. Phil. Mag. 18(155), 1041.
31. Manson S. S. (1965) Fatigue: a complex subject-some simple approximations. Exp. Mech. 5, 193-226.
32. Manson S . S . , Nachtigall A. J., Ensign C. R. and Freche J. C. (1965) Further investigation of a relation
for cumulative fatigue damage in bending. J. Engng Ind., Trans. A S M E Ser. B 87, No. 1.
33. Chaboche J. L., Raine P. and Kaczmarek H. (1981) On the interaction of hardening and fatigue damage
in the 316 stainless steel. ICF5, Cannes.
34. Chaboche J. L., Kaczmarek H. and Raine P. (1981) Hardening and fatigue damage interaction in 316
L stainless steel. La Rech. Aerospatiale (English Edition) No. 1980-3, 35-55.
35. Jacquelin B., Hourlier F. and Pineau A. (1983) Crack initiation under low cycle multiaxial fatigue in type
316 L stainless steel. J . Press. Vess. Technol., Trans. A S M E 105, 138.
36. Cailletaud G . and Kaczmarek H. (1984) Prevision d’amorgage en fatigue oligocyclique sous sollicitations
multiaxiales. J. Int. Printemps, Paris.
37. Bui-Quoc T. (1981) An interaction effect consideration in cumulative damage on a mild steel under torsion
loading. 5th Inr. Con$ on Fracture, Cannes, pp. 2631-2725.
38. Baudry G . , Amzallag C . and Bernard J. L. (1984) Etude de cumul de dommage en fatigue sur un acier
austenitique. J. Int. Printemps, “Amorcage Fissures sous Sollicitations Complexes”, Paris.
39. Chaudonneret M. and Culit J. P. (1985) Adaptation de la theorie de Neuber au cas de concentration de
contrainte en viscoplasticite. La Rech. Aerospatiale, No. 1985-4.
40. Chaudonneret M. and Chaboche J. L. (1986) Fatigue life prediction of notched specimens. Int. Conf. on
Fatigue of Engineering Materials and Structures, Sheffield.
41. Taira S. (1962) Lifetime of structures subjected to varying load and temperature. Creep in Structures
(Edited by Hoff N. J.). Academic Press, London.
42. Lesne P. M. and Chaboche J. L. (1984) Prediction of crack initiation under thermal fatigue and creep.
2nd Int. Conf. on Fatigue Thresholds, Fatigue 84, Birmingham.
43. Chaboche J. L. (1978) Description thermodynamic et phenomenologique de la viscoplasticite cyclique
avec endommagement. These Univ. Paris VI. Publication ONERA No. 1978-3.