You are on page 1of 9

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/257336182

An equivalent stress process for fatigue life estimation under multiaxial


loadings based on a new non linear damage model

Article in Materials Science and Engineering A · March 2012


DOI: 10.1016/j.msea.2011.12.105

CITATIONS READS

20 406

6 authors, including:

A. Aid Bendouba Mostefa


Mascara Universit, Algeria University Mustapha Stambouli of Mascara
49 PUBLICATIONS 189 CITATIONS 25 PUBLICATIONS 66 CITATIONS

SEE PROFILE SEE PROFILE

Laid Aminallah Amrouche Abdelwaheb


University Mustapha Stambouli of Mascara Université d'Artois
37 PUBLICATIONS 149 CITATIONS 64 PUBLICATIONS 599 CITATIONS

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Intenational publication View project

Three-dimensional finite element analysis of fracture behaviour of adhesively bonded single lap-joints View project

All content following this page was uploaded by A. Aid on 03 October 2018.

The user has requested enhancement of the downloaded file.


Materials Science and Engineering A 538 (2012) 20–27

Contents lists available at SciVerse ScienceDirect

Materials Science and Engineering A


journal homepage: www.elsevier.com/locate/msea

An equivalent stress process for fatigue life estimation under multiaxial loadings
based on a new non linear damage model
A. Aid a,b,∗ , M. Bendouba a , L. Aminallah a , A. Amrouche c , N. Benseddiq b , M. Benguediab d
a
Laboratoire LPQ3M, BP305, Université de Mascara, Algeria
b
Laboratoire de Mécanique de Lille, Université Lille1 Nord de France, Cité scientifique, 59655 Villeneuve d’Ascq, France
c
Laboratoire de Génie Civil et géo-Environnement LGCgE, EA 4515, Faculté des Sciences Appliquées FSA Béthune, Université d’Artois, France
d
Laboratoire LMPM, Département de Mécanique, Université de Sidi Bel Abbes, BP89, Cité Ben M’hidi, Sidi Bel Abbes 22000, Algeria

a r t i c l e i n f o a b s t r a c t

Article history: Fatigue life estimation or fatigue damage evaluation of mechanical components under a multiaxial state
Received 4 October 2011 of stress time-history has an important role in virtual design phases, but this evaluation is a real problem
Received in revised form to resolve, because this goal is not reached by classical fatigue criteria. In this context, the equivalent
23 December 2011
uniaxial stress has been chosen as the approach useful for this purpose by combining it with a strength
Accepted 24 December 2011
curve for the material (i.e. S–N curve) and damage evaluation (Damaged Stress Model (DMS) and Miner’s
Available online 20 January 2012
rule). So the authors have taken equivalent stresses based on a multiaxial criterion that can be used as a
uniaxial stress time history for the application of damage evaluation methods to predict the fatigue life,
Keywords:
Fatigue life
under multiaxial random loading. The cycles were counted with the Rain Flow algorithm, using equivalent
Multiaxial loading stress as a variable counting. In the case of “asymmetric” histories, the algorithm of transformation of the
Damaged Stress Model stress history (because of its global expected value different from zero) is required. In the case of 10HNAP
Random loading steel, stress amplitude transformation according to the Goodman relationship gives good results. The
proposed algorithm was verified during fatigue tests of cruciform specimens made of 10HNAP steel,
subjected to biaxial non-proportional random tension-compression.
© 2012 Elsevier B.V. All rights reserved.

1. Introduction One difficulty is that most of the existing multiaxial fatigue crite-
ria involve in their implementation for general complex multiaxial
Most engineering components and structures in service are sub- fatigue loadings.
jected to a two or three-dimensional stress state with complex Among current multiaxial fatigue criteria, the Sines [4] and the
loading. Crossland [5] criteria are very popular and easy to use for engi-
The origin of multiaxiality comes from various factors such neering design. They can provide good predictions for proportional
as multiaxial external loading, complex geometry of the com- loads with mean stress effects [6].
ponent and residual stresses. Such multiaxial stresses are often The real loadings are often more complex. Estimation of fatigue
non-proportional, that is, the corresponding principal directions life of a material subjected to random loading needs suitable calcu-
and/or principal stress ratios vary with time. lation algorithms. In such algorithms, cycle counting [7,8] according
The multiaxial fatigue criteria proposed in the literature may to a given method (for example, the Rain Flow method) and damage
be categorized in three groups: stress based methods, strain- accumulation according to the assumed hypothesis (for example
based methods, and energy-based approaches [1–3]. For high-cycle Palmgren–Miner, Haigh’s diagram) are the main operations.
fatigue, most of the fatigue criteria are stress-based methods. In this paper, our main aim is to extend our Damaged Stress
Although there are numerous multiaxial fatigue criteria in the Model (DSM) developed initially for the uniaxial loading, at the
literature, design engineers are often faced with difficulties in biaxial random loading on cruciform specimens. The equivalent
applying these criteria to modern engineering design. stress of von Mises, Sines and Crossland is applied and tested as
variables counting in Rain Flow approach. Several biaxial random
amplitude fatigue tests that were carried out at the Technical Uni-
versity of Opole allow the validation of the proposed approach with
both damage laws. The predictions by the approach adopted in this
∗ Corresponding author at: Laboratoire LPQ3M, BP305, Université de Mascara, paper for the prediction of life in multiaxial fatigue and the fol-
Algeria. Fax: +213 45804169. lowing models for multiaxial loading (see Smith–Watson–Topper
E-mail address: aid abdelkrim@yahoo.com (A. Aid). damage parameter used by Bannantine–Socie [9], Fatemi–Socie

0921-5093/$ – see front matter © 2012 Elsevier B.V. All rights reserved.
doi:10.1016/j.msea.2011.12.105
A. Aid et al. / Materials Science and Engineering A 538 (2012) 20–27 21

 orientating each material plane P→n at the point M on the surface of the specimen.
Fig. 1. Coordinates system used to define the unit normal vector n

[9,10], Socie for HCF [11], Wang–Brown [12,13] and Lagoda–Macha The equivalent stress of damage at the level (i + 1) is calculated
[14–19]) are analyzed and predictions are compared with experi- with the relation
ments. (i)d − i (i+1)d − i+1
The main advantage of our model is the ease of use which Di = = (4)
u − i u − i+1
request only the Wöhler curve (S–N). This model describes the non-
linear damage and takes into account the loading time history. where  (i+1)d , equivalent stress of damage at the level (i + 1) and
 i+1 applied stress at the level (i + 1).
2. Linear and non-linear damage laws So  ed is equal to  i to the first cycle, it means D = 0 and  ed is
equal at  u at the last cycle D = 1.
It should be noted that here two damage and accumulation rules For the application of DSM model in multiaxial fatigue, we need
have been used. The first one is the linear Miner’s rule [20], while an equivalent stress that takes into account the state of multiaxial
the second is the non-linear Damaged Stress Model (DSM), which loading. We propose the following criteria: von Mises, Sines and
has been developed in our laboratory. For more details, see Refs. Crossland. The equivalent stress is used as a “counting variable”
[21–25]. to extract the random cycles, the calculation of the elementary
damage and the accumulated damage.

2.1. Linear Miner’s rule


3. Critical plane approach and definition
The damage Di induced by a cycle extracted from the multiaxial
sequence is obtained from its fatigue life Ni as: As shown in Fig. 1, consider a material plane, denoted as Pn , pass-
ing through the point under consideration. The plane Pn is located
1 by its unit normal vector n  . This unit vector in turn is described by
Di = (1)
Ni its spherical angles (, ). All the fatigue life calculation methods
using a cycle counting algorithm can be summarized with the same
The damage accumulation is defined by the sum of the damage
methodology. First a cycle counting variable is chosen in order to
Di of all extracted cycles. The fatigue life N of the whole sequence
extract the cycles (with their amplitude and mean value) from the
corresponds to the number of repetitions of this stress history up
variable amplitude or random multiaxial loading signal (stresses
to crack initiation and is obtained by:
and/or strains). For the simulations presented hereafter the ASTM

D= Di (2) Rain Flow algorithm was used [7,8] (with 64 classes). All the random
loadings are considered as stationary for life calculation. Secondly,
The rupture occurs when D = 1. a damage parameter Dp is chosen; it depends on stress–strain quan-
tities. This damage parameter is computed on each material plane
Pn , orientated by the unit normal vector n  (, ) (Fig. 1), in order to
2.2. Non linear Damaged Stress Model determine the critical plane.
Finally, to quantify the damage generated by each cycle identi-
The Damaged Stress Model (DMS) [21] which is developed fur- fied with the counting algorithm it is necessary to use an equation
ther is regarded as having many advantages such as taking into relating the damage parameter and the number of cycles to failure
account the occurrences order of loading, considering cycles below Nf . According to the S–N curve of the tested material, it corresponds
and over the so-called fatigue limit in different manners, and pre- to stress amplitude smaller to 0.25 af in fully reversed tension [18].
senting a non-linear damage accumulation. This damage indicator
is easily connected cycle by cycle to the Wöhler’s curve.
The expression of the DSM indicator gives the increase of dam- 3.1. Fatemi–Socie model [FS]
age Di due to  i applied stress cycles defined by their amplitude  a
and their mean value  m , as follows: The Fatemi–Socie model [FS] [10] is widely applied for shear
damage model, which predicts the critical plane is the plane orien-
(i)d − i tation  with the maximum F–S damage parameter:
Di = (3)
u − i   
 nmax
where  (i)d , stress of damage;  i , applied stress; and  u , ultimate 1+k (5)
2 y
stress. max
22 A. Aid et al. / Materials Science and Engineering A 538 (2012) 20–27

3.3. Socie’s proposal for HCF regime [So]

In the case of high cycle fatigue and ductile materials most of the
fatigue life is consumed by crack nucleation on the planes where
the shear stress is maximum. In this case Socie [11] proposes the
following stress based approach by using the linear combination
of the shear stress amplitude  a and the maximum normal stress
 nmax acting on the critical plane, both during the load cycle (Fig. 2)
as damage parameter: DP-So =  a + k2  nmax . The cycle counting algo-
rithm has to be applied on two variables: the shear stresses nx (t)
and ny (t). The equation linking the damage parameter and the
number of cycles to failure is given by the following relation:

DP-So = f (2Nf )b (8)

where f (2Nf )b is the elastic part of the strain-life curve. k2 is a


material parameter identified by fitting tension and torsion fatigue
Fig. 2. Definition of the mean and maximum normal stress during an extracted cycle data. The algorithm used to apply this method is illustrated [19].
(from t1 to t2 ) with the Rain Flow algorithm.

where /2 is the maximum shear strain amplitude on a plane , 3.4. Wang and Brown’s model [W–B]
 nmax is the maximum normal stress on that plane and  y is the
material yield strength. Wang and Brown put up their criterion first in 1993 in [12,13]
The critical plane is, for each of these two counting variables, the restricted to low (LCF) and medium (MCF) cycle fatigue according
plane experiencing the highest shear strain range. For the materi- to the assumption that fatigue crack growth is controlled by the
als where the fatigue crack initiation is dominated by plastic shear maximum shear strain. Brown and Wang introduced their specific
strains, [FS] recommend to use the following damage parameter: way of load history decomposition in [28]. The equivalent strain is
DP-FS =  a (1 + k nmax / y ) related to the Manson–Coffin curve in tor- a combination of the shear strain amplitude and efficient normal
sion by equation: strain range:

f f − 2nmean


DP-FS = (2Nf )b + f (2Nf )c (6) a + Sεn = (1 + e + S(1 − e )) (2Nf )b + (1 + p
G E
For each loading cycle extracted by the Rain Flow method,  a +S(1 −  c
p ))εf (2Nf ) (9)
is the shear strain amplitude,  nmax is the maximum of the normal
stress on the critical plane (Fig. 2). In this damage parameter, k
where e and p are respectively the elastic and plastic Poisson’s
is a material constant, which can be found by fitting fatigue data
ratio of the material.  nmean is the mean normal stress on the critical
from simple uniaxial tests to fatigue data from simple torsion tests,
plane during each extracted cycle (Fig. 2) of counting variable. The
k = 1.0 (for 42CrMo4 steel) and k = 0.4 (for CK45 steel) [26,27].
efficient normal strain range εn is the range between upper and
Eq. (6) is the description of the strain-life Manson–Coffin curve
lower values of normal strain in one shear strain half-cycle.
in torsion. When the strain-life Manson–Coffin torsion curve is not
The critical plane is the plane experiencing the highest shear
known, [FS] propose to approximate this curve from the tensile
strain range. The Rain Flow cycle counting method has to be applied
strain-life curve. The algorithm used to apply this fatigue life cal-
on two counting variables: the shear strains nx (t) and ny (t).
culation method is detailed in [18].
The algorithm used to compute the fatigue life according to the
Wang–Brown model is detailed in [19].
3.2. Bannantine and Socie model [Ba] Assuming that, during one cycle the normal strain εn plays an
important additional role, Wang and Brown propose the following
The Smith–Watson–Topper (SWT) parameter proposed by expression as damage parameter: DP-WB =  a + Sεn where  a is the
Smith et al. (1970) is a direct development of the strain-life shear strain amplitude and S is a material parameter identified by
equation which seeks to address the effect of mean stress on fitting tension against torsion fatigue data.
fatigue life. The SWT parameter is based on the combination of
the Coffin–Manson low cycle fatigue equation, Basquin’s high cycle
fatigue equation and consideration of the peak stress to account for 3.5. The strain energy density parameter [Eng]
the mean stress effect, as follows:
A change of strain energy density, applied in theory of plasticity,
  2f
DP-SWT,BS = εn,a nmax = (2Nf )2b + f εf (2Nf )b+c (7) has been proposed as a parameter to describe multiaxial fatigue
E [14–17]. The strain energy density versus time is given by [16,17]:
where f is the fatigue strength coefficient, E is Young’s modulus, Nf
is the number of cycles to crack initiation, b is the fatigue strength 1
W (t) = (t) · ε(t) (10)
exponent, εf is the fatigue ductility coefficient and c is the fatigue 2
ductility exponent. If the cyclic stresses and strains reach their maximum values  a and
Bannantine and Socie [9] recommend using the (SWT) damage εa , the maximum value of the energy parameter and its amplitude
parameter DP-SWT = εn,a  nmax . In Eq. (7) and for each load cycles are the same: Wa = Wmax .
extracted by the Rain Flow method, εn,a is the amplitude of the
normal strain and  nmax is the maximum normal stress during the 1
current cycle of the counting variable εn (t) (Fig. 2). Wa = a εa (11)
2
A. Aid et al. / Materials Science and Engineering A 538 (2012) 20–27 23

Assuming W(t) as the fatigue damage parameter according to 3.6.2. Crossland


criterion
Eq. (10), we can rescale the standard characteristics of cyclic fatigue oct,amax + ˛Cr · ≤ ˇCr (18)
( a − Nf ) and (εa − Nf ) and obtain a new one, (Wa − Nf ) by: Hmax

The coefficients ˛ and ˇ in both equations can be set through


(f )2 1 evaluation of the formulas at fatigue limits in torsion and tension.
Wa = (2Nf ) 2b
+ εf f (2Nf )b+c (12)
2E 2 The Sines formula does not allow alternating fatigue limit in tension
to be used due to a singular solution. Thus another load condition –
The fatigue limit is determined from Eq. (11). We obtain:
a repeated tension – has to be used. The appropriate value for Sines
2
af is:
Waf = (13)
2E √  √ 2
−1
˛Si = 6 − 2 , ˇSi = −1 (19)
Fatigue characteristic Eq. (12) for high cycle fatigue takes the 0 3
form
And for Crossland:
(f )2 √ 
Wa = (2Nf ) 2b
(14) −1
√ 2
2E ˛Cr = 6 − 2 , ˇCr = −1 (20)
−1 3
When the energy density history at the given plane had
been determined, the energy cycles were counted with the  a is the square root of J2,a , 
that is the alternate component of the
Rain Flow method; next damage was accumulated according to deviatoric second invariant, Hmean is the mean value of the hydro-
Palmgren–Miner hypothesis [20] taking into account energy cycle static stress and Hmax is the maximum value of the hydrostatic
amplitudes below the fatigue limit [14,15,17]. The life time, TRF is stress.
calculated from cycles and half-cycles in the time observation T0 of If the two S–N curves (the stress amplitude corresponding to fail-
stress history (t), ure at N cycles), i.e. the uniaxial repeated bending fatigue strength
 0 (N) at N cycles and the reversed torsion fatigue strength  −1 (N) at
T0 T0 N cycles are available, then the Sines formulation Eq. (17) becomes:
TRF,W = = J for Wai ≥ a · Waf
S(T0 ) n /(N0 (Waf /Wai )
i=1 i
m
) 
oct,amax + ˛Si (N) · ≤ ˇSi (N) (21)
Hmean
S(T0 ) = 0 for Wai < a · Waf , a = 0.25 (15)
where
where S(T0 ) is material damage up to time T0 ; j is number of class √  (N) √ 2
−1
intervals of the histogram of the amplitudes of the strain energy ˛Si = 6 − 2 , ˇSi = −1 (N) (22)
0 (N) 3
density; Waf is the fatigue limit expressed by strain energy den-
sity; a is a coefficient allowing to include amplitudes below Waf in Eq. (21) is an extension of the Sines fatigue limit formulation to
the damage accumulation process; m is the slope of fatigue curve high-cycle fatigue.
expressed by energy: If the two S–N curves, the uniaxial reversed bending fatigue
strength  −1 (N) at N cycles and the reversed torsion fatigue
log(Nf ) = A − m log(Wa ) (16)
strength  −1 (N) at N cycles are available, then the Crossland for-
N0 is a number of cycles corresponding to the fatigue limit Waf ; mulation, Eq. (18) becomes:
ni is a number of cycles with amplitude Wai (two the same half- 
cycles from one cycle). The algorithm used to apply this method is oct,amax + ˛Cr (N) · ≤ ˇCr (N) (23)
Hmax
illustrated in [15].
where

3.6. Sines, Crossland and von Mises criterion


√  (N) √ 2
−1
˛Cr = 6 − 2 , ˇCr = −1 (N) (24)
−1 (N) 3
These methods depend on the invariants of the stress tensor.
They have a less onerous computational effort than the critical Eq. (24) is an extension of the Crossland fatigue limit formulation
plane methods (they do not need of extensive projections on phys- to high-cycle fatigue.
ical planes), but its efficiency is not good as for critical plane For Eqs. (21) and (23), the key step is the evaluation of the
methods. Indeed stress-invariants criteria work better only if prin- equivalent
 shear stress amplitude oct,amax and hydrostatic stresses
cipal axes variation is large. Hmax , throughout a general multiaxial loading.

The numerical approach for evaluating the effective shear stress


amplitude throughout a loading cycle makes it possible to extend 3.6.3. Compute the hydrostatic and deviatoric stresses
the Sines and the Crossland multiaxial fatigue criteria for finite Split the stress tensor (t) into its deviatoric and spherical parts
fatigue life prediction under general multiaxial loadings. 1
Both Crossland [29] and Sines [30,31] published their works (t) =   (t) + tr((t))I (25)
3
throughout the fifties of the last century. Their criteria are very
where tr((t)) is the first stress invariant given by
much alike, using the amplitude of second invariant of stress ten-
sor deviator (which corresponds to the von Mises stress) as the tr((t)) = xx (t) + yy (t) + zz (t) (26)
basis. Another term is added to the equation in order to cope with
The stress deviator  is expressed as:
the mean stress effect – while Sines prefers the mean value of first
invariant of stress tensor (i.e. hydrostatic stress): ⎛ 2xx − yy − zz ⎞
xy xz
⎜ 3
2yy − xx − zz ⎟

3.6.1. Sines criterion  = ⎜ yz yz ⎟
⎝ 3 ⎠
oct,amax + ˛Si · ≤ ˇSi (17) 2zz − xx − yy
Hmean zx zy
3
Crossland recommends use of its maximum value. (27)
24 A. Aid et al. / Materials Science and Engineering A 538 (2012) 20–27

Fig. 3. Cruciform specimen descriptions.

 3.6.4. von Mises criterion


The hydrostatic stress H (t) is calculated as: The method presented here is based on a generalization of the
 1
theory of plasticity generally applied in the case of static loads.
(t) = tr((t)) (28) According to this conjecture, damage under multiaxial loading can
H 3 be estimated by determining a uniaxial equivalent stress or strain
For a cycle loading, the mean value of stress is computed as: [32]. According to the literature, we note that the amplitude of
 T the equivalent von Mises stress is usually used as a first approx-
1 imation. When dealing the study of a vehicle component, Heyes
[]mean = [(t)] dt (29)
T 0 et al. [33] have used this approach for the components subjected to
complex multiaxial loadings. It should be noted that the von Mises
the tensor of amplitude value of stress as:
equivalent stress is calculated by the formula:
[a (t)] = [(t)] − []mean (30)

and the shear octahedral stress as:


1

oct,a = (a11 (t) − a22 (t))2 + (a22 (t) − a33 (t))2 + (a33 (t) − a11 (t))2 + 6(a23 (t)2 + a13 (t)2 + a12 (t)2 ) (31)
3
For a loading cycle, the mean value of hydrostatic stress, the 1

2 2 2
eq = √ (xx − yy ) + (yy − zz ) + (xx − zz ) + 6(xy
2
+ yz
2
+ xz
2
) (35)
maximum values of hydrostatic and alternating octahedral shear 2
stress are calculated by:
  t1 +T  4. Validation with biaxial variable amplitude sequences
1
= (t) dt (32)
Hmean T t1
H The validation of the method has been obtained by the use of
biaxial tension-compression variable amplitude tests results. Seven
  biaxial random stress histories are considered. They are composed
= max (t)
Hmax t H of 177.000 up to 190.413 events. These tests were carried out in the
laboratory of Professor E. Macha (Bedkowski [34]) in Opole (Poland)
oct,amax = maxoct,a (t) kindly provided by Professor T. Palin-Luc (Laboratoire LAMEFIP,
t Bordeaux, France). The specimens have a cruciform shape as shown
The Sines and Crossland equivalent stresses are calculated by in Fig. 3. All test specimens were made of low carbon steel 10HNAP.
the equations: Both the full chemical composition and the mechanical properties
of the material used in this investigation are given in Tables 1 and 2
tr((t)) = xx (t) + yy (t) + zz (t) (33) [15,19,25] respectively.
  The required material fatigue data are three S–N curves  −1 (N),
eq Cr = Cr oct,amax + ˛Cr · (34)  −1 (N) and  0 (N). These give the material fatigue strengths versus
Hmax
the number N of cycles for a reversed tensile test (R = −1), a reversed
Si and Cr are the correction coefficients of the equivalent stress, torsion test (R = −1) and a zero to maximum tensile test (R = 0)
which can use the S–N curve with alternating tension. respectively. Bedkowski and Macha obtained the following S–N
A. Aid et al. / Materials Science and Engineering A 538 (2012) 20–27 25

Table 1
Chemical composition of the steel tested in (%).

Steel C Mn Si P S Cr Cu Ni Fe

10HNAP 0.115 0.71 0.41 0.082 0.028 0.81 0.30 0.50 Rest

Table 2
Mechanical properties of 10HNAP steel.

Steel Young modulus E (GPa] Poisson ratio ( ) Yield stress Re (MPa) Tensile strength Rm (MPa) Z (%) A10 (%)

10HNAP 215 0.29 414 566 60 32

curves that are shown in Fig. 4. The mechanical property used by


the Damaged Stress Model for this material is  u = Rm = 566 MPa.
According to [3,25], the applied loads are stationary and ergodic
with normal centred distributions and zero mean value. These loads
are generated by a random signal generator.
Seven tests have been provided with each temporal multiple
records (3–7 records) local deformations (εxx , εyy , εxy ), measure-
ments are taken with a frequency of 480 Hz. The total number of
measurement points is 59,420 points per sequence, each sequence
with duration of 123.8 s. Hook’s equations is used for calculated
the stress ( xx ,  yy ,  xy ) [25]. Note that  xx and  yy are out of phase
(ϕ ≈ /2). The shear stress  xy are virtually low values (between
−29.3 and 8.4 MPa) compared with others values (between −225
and 244.6 MPa for  xx and between −287.9 and 277.3 MPa for  yy ).
Applied loading are determined by the equivalent stresses for-
mulas (Eqs. (33)–(35)) of the chosen criteria for this study. The
characteristics of the equivalent stress are given in Table 3.
We can see in Table 3 that the von Mises value and the amplitude
are more important than Crossland and Sines ones. Consequently,
the von Mises equivalent stress is more damaging than Crossland
and Sines equivalent stresses.
Fig. 5. Assessed lives by ([Ba], [FS], [W–B], [So] and [Eng]) criterion against (VM[Mi] ,
Table 4 summarizes experimental and numerical results.
VM[DSM] ) ones for biaxial random loading.
Experimental lives of the biaxial sequences are compared with
theoretical ones (assessed using Miner and DSM model). They are
expressed as the number of predictions up to crack initiation. For all Figs. 5–7, it is noticed that for lives ranging between 104
The results of Table 4 are plotted in Figs. 5–7. These figures show and 106 , Socie’s method [So] is very conservative. Fatemi–Socie [FS]
the comparison of the calculated fatigue life by ([Ba], [FS], [W–B], and Bannantine [Ba] methods are closest to the experiment.
[So] and [Eng] criterion), experimental data and predicted life by Wang–Brown’s [W–B], method leads to non-conservative pre-
(VM-Mi, VM-DSM, Cr-M, Cr-Dsm, Si-Mi and Si-Dsm) by biaxial ran- dictions on the whole of these simulations. Application of the
dom loading for 10HNAP steel.

Fig. 4. Reversed tension-compression (S–N) curve  −1 (N), reversed torsion (S–N) Fig. 6. Assessed lives by ([Ba], [FS], [W–B], [So] and [Eng]) criterion against (Cr[Mi] ,
curve  −1 (N) and zero maximum tensile curve  0 (N) [35]. Cr[DSM]) ones for biaxial random loading.
26 A. Aid et al. / Materials Science and Engineering A 538 (2012) 20–27

Table 3
Equivalent stress characteristics.

Maximum value (MPa) Minimum value (MPa) Amplitude (MPa) Mean value (MPa)

Von Mises (Eq. (35)) 535.85 15.32 520.53 222.69


Crossland (Eq. (34)) 439.47 9.67 429.8 186.62
Sines (Eq. (33)) 459.2 −3.26 462.46 173.84

Table 4
Experimental and predicted lives ([Ba]: Bannantine, [F–S]: Fatemi–Socie [W–B]: Wang and Brown’s, [So]: Socie, [Eng]: strain energy density parameter, [Exp]: experimental).

No. of specimen T[Ba] T[F–S] T[W–B] T[So] T[Eng] T[exp] VM[Mi] VM[DSM] Cr[Mi] Cr[DSM] Si[Mi] Si[DSM]

B10H03 61,820 54,863 104,800 13,344 32,185 70,500 1630 1128 39,753 36,034 25,238 22,541
B10H04 70,560 74,397 133,012 20,598 14,948 58,800 3080 2600 53,749 49,818 36,479 33,450
B10H05 163,972 119,073 245,475 41,209 28,617 102,200 12,348 10,950 228,830 214,336 168,342 157,534
B10H06 217,375 135,921 307,989 51,694 34,001 188,000 19,059 17,053 392,813 371,030 329,850 293,846
B10H07 669,208 246,094 515,362 120,274 105,520 141,800 82,717 76,568 1,065,149 1,014,596 728,253 692,829
B10H08 354,745 226,009 529,573 102,807 43,574 309,600 35,857 31,966 487,364 458,272 409,648 383,984
B10H09 317,088 178,257 372,954 73,036 53,498 298,800 69,174 62,793 1,311,443 1,242,472 1,078,017 1,032,698

suitability of the proposed methods. The ratio between experimen-


tal lives and expected ones fall in a scatter band of a factor of 5.
The methods of Fatemi–Socie [FS] and Socie [So] and must be
used with caution. They are often very conservative. They can lead
to over sizing.
Von Mises equivalent stress is very conservative and its pre-
dictions are far from experimental results. Crossland and Sines
equivalent stress are in good agreement with experimental results.
We note that the equivalent stress calculated according to the
Sines criterion gives good results with the Damaged Stress Model
(DSM), for calculation of life and the accumulation of damage.

References

[1] Y.S. Garud, J. Test. Eval. 9 (1981) 165–178.


[2] I.V. Papadopoulos, in: K. Dang Van, I.V. Papadopoulos (Eds.), Advanced Course
on High-Cycle Metal Fatigue, ICMS, Udine, Italy, 1997.
[3] T. Lagoda, E. Macha, Proceedings of Fracture from Defects, ECF12, Sheffield,
1998, pp. 73–79.
[4] G. Sines, in: G. Sines, J.L. Waisman (Eds.), Metal Fatigue, McGraw Hill, New York,
1959, pp. 145–169.
[5] B. Crossland, Proceedings of the International Conference on Fatigue of Metals,
London, 1956, pp. 138–149.
Fig. 7. Assessed lives by ([Ba], [FS], [W–B], [So] and [Eng]) criterion against (Si[Mi] , [6] M. De Freitas, B. Li, J.L.T. Santos, in: P.J. Kalluri, Bonacuse (Eds.), Multiaxial
Si[DSM]) ones for biaxial random loading. Fatigue and Deformation: Testing and Prediction, ASTM STP 1387, American
Society for Testing and Materials, West Conshohocken, PA, USA, 2000, pp.
139–156.
[7] C. Amzallag, J.P. Gerey, J.L. Robert, J. Bahuaud, Int. J. Fatigue 16 (1994) 287–293.
[8] S. Dowing, D. Socie, Int. J. Fatigue 1 (1982) 31–40.
energy parameter [Eng] shows that the calculation results for
[9] J.A. Bannantine, D. Socie, in: K. Kussmaul, D. McDiarmid, D. Socie (Eds.), A Vari-
10HNAP steel fall in a scatter band of a factor of 3 [15]. able Amplitude Multiaxial Fatigue Life Prediction Method, ESIS 10, 1991, pp.
The majority of predictions using the equivalent stress of von 35–51.
Mises combined with linear or nonlinear damage is very conserva- [10] A. Fatemi, D.F. Socie, Fatigue Fract. Eng. Mater. Struct. 11 (3) (1988) 149–165.
[11] D. Socie, in: D.L. McDowell, R. Ellis (Eds.), Critical Plane Approaches for Multi-
tive and far from experimental reality. axial Fatigue damage Assessment, ASTM STP 1191, 1993, pp. 7–36.
Using the Crossland and Sines equivalent stress (assessed by [12] C.H. Wang, M.W. Brown, Fatigue Fract. Eng. Mater. Struct. 16 (12) (1993)
using linear and non-linear damage models) gives predictions clos- 1285–1298.
[13] C.H. Wang, M.W. Brown, J. Eng. Mater. Technol. 118 (1996) 367–370.

est to the experimental reality. For all simulations, 
the predictions [14] T. Łagoda, E. Macha, in: S. Kalluri, P.J. Bonacuse (Eds.), Multiaxial Fatigue and
are generally all contained in an interval 0, 2Texp , 5Texp . Deformation: Testing and Prediction, ASTM STP 1387, 2000, pp. 173–190.
[15] T. Łagoda, E. Macha, W. Bedkowski, Int J. Fatigue 21 (5) (1999) 431–443.
[16] T. Łagoda, Int. J. Fatigue 23 (6) (2001) 467–480.
[17] T. Łagoda, Int. J. Fatigue 23 (6) (2001) 481–489.
5. Conclusions [18] A. Banvillet, T. Łagoda, E. Macha, A. Niesłony, T. Palin-Luc, J.F. Vittori, Int. J.
Fatigue 26 (4) (2004) 349–363.
A fatigue life prediction method has been proposed for multiax- [19] A. Banvillet, Prévision de durée de vie en fatigue multiaxiale sous chargements
réels: vers des essais accélérés. PhD thesis, ENSAM CER de Bordeaux, France,
ial random loading stress histories. It is based on a counting variable
2001, p. 274.
representative of the stress states and of the evolution versus in [20] M.A. Miner, J. Appl. Mech. 12 (1945) 159–164.
order to identify and extract multiaxial cycles. [21] G. Mesmacque, S. Garcia, A. Amrouche, C. Rubio-Gonzalez, Int. J. Fatigue 27 (5)
(2005) 461–467.
The counting variable used here is an equivalent stress defined
[22] S. Garcia, A. Amrouche, G. Mesmacque, X. Decoopman, C. Rubio, Int. J. Fatigue
by the multiaxial criteria selected for this study to solve the prob- 27 (10–12) (2005) 1347–1353.
lem of damage accumulation for the nonlinear Damaged Stress [23] A. Aid, J. Chalet, A. Amrouche, G. Mesmacque, Fatigue Des. (2005).
Model (DSM). Both models DSM and Miner’s model are usable for [24] A. Aid, A. Amrouche, B. Bachir Bouiadjra, M. Benguediab, G. Mesmacque, Mater.
Des. 32 (1) (2011) 183–191.
the fatigue assessment for biaxial random stress histories issued [25] A. Aid, Cumul d’endommagement en fatigue multiaxiale sous sollicitations
from test carried out on cruciform specimens allow validation the variables. PhD thesis, Universitè de Sidi-Bel Abbes, Algérie, 2006, p. 195.
A. Aid et al. / Materials Science and Engineering A 538 (2012) 20–27 27

[26] L. Li., B. Reis, M. Leite, M. Freitas, Fatigue Fract. Eng. Mater. Struct. 28 (2005) [31] G. Sines, in: G. Sines, J.L. Waisman (Eds.), Metal Fatigue, McGraw Hill, 1959, pp.
445–454. 145–169.
[27] L. Li., B. Reis, M. Leite, M. Freitas, Fatigue Fract. Eng. Mater. Struct. 27 (2004) [32] X. Pitoiset, A. Preumont, Int. J. Fatigue 22 (7) (2000) 541–550.
775–784. [33] P.J. Al Heyes, 4th Int. Conf. Biaxial/Multiaxial Fatigue, Saint-Germaint, France,
[28] C.H. Wang, M.W. Brown, J. Eng. Mater. Technol.-Trans. ASME 118 (1996) 1994.
367–374. [34] W. Bedkowski, 4th Int. Conf. Biaxial/Multiaxial Fatigue, ESIS, France, 1994, pp.
[29] B. Crossland, Proc. Int. Conf. Fatigue of Metals, Ins. Mech. Eng., London, 1956, 435–447.
pp. 138–149. [35] B. weber, G. Vialaton, A. Carmet, J.L. Robert, 5th Int. Conf. Biax-
[30] G. Sines, Failure of Materials Under Combined Repeated Stresses with Super- ial/Multiaxial Fatigue & Fracture Cracow, ESIS, Poland, 1997,
imposed Static Stresses. NACA-TN-3495, NACA, Washington, 1955. pp. 218–231.

View publication stats

You might also like