Upload
a-ahmadi
View
217
Download
3
Embed Size (px)
Citation preview
Lifetime simulation under multiaxial random loading
with regard to the microcrack growth
A. Ahmadi a, H. Zenner b,*
a Volkswagen AG, Wolfsburg, Germanyb Institute for Plant Engineering and Fatigue Analysis, Clausthal University of Technology, Leibnizstr. 32, D-38678 Clausthal-Zellerfeld, Germany.
Received 1 October 2004; received in revised form 8 August 2005; accepted 12 September 2005
Available online 5 January 2006
Abstract
Generally, most machine parts are loaded with a combination of different variable forces and moments which often causes a state of multiaxial
stress in the fatigue critical areas of the parts. In the most cases, a non-proportional cyclic multiaxial state of stress occurs. Compared to the in-
phase loading, a multiaxial loading with a phase shift between the stress components and a load ratio of ta/saz0.5 between tension/compression
and torsion leads to a significant influence on the fatigue lifetime. The reason is the changing direction and rotation of the principal stresses during
one cycle. In this paper, a model designed to simulate the damage process based on the growth of microcracks under the influence of cyclic loading
is presented. The crack growth is initially dominated by shear stresses leading to microstructurally short cracks (stage I) and continues to grow
under the influence of normal stresses (physically short cracks). The results of the lifetime estimation generated by means of the new concept on
the basis of microcrack growth are compared and verified with those experiences obtained from multiaxial fatigue testing.
q 2005 Elsevier Ltd. All rights reserved.
Keywords: Microcrack growth; Crack growth simulation; Multiaxial stress state; In-phase; Out-of-phase; Lifetime
1. Introduction
Many parts of structures and engineering components are
subjected to multiaxial loading which often cause a multiaxial
stress in the fatigue critical areas. The influence of multiaxial,
in-phase or out-of-phase stresses and strains on the fatigue life
becomes more and more important for fatigue analysis using
numerical concepts [1,2]. This investigation was aimed to build
an experimental database for the verification of a new concept
of lifetime estimation with the application of the simulation of
microcrack growth [3–7]. This fatigue damage model requires
to be expressed in terms of cracks.
In the following paper, a damage accumulation model to
describe physically the material damage mechanisms will be
presented. The rate of fatigue damage accumulation should be
physically expressed in terms of fatigue crack growth rates [8].
0142-1123/$ - see front matter q 2005 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ijfatigue.2005.09.015
* Corresponding author. Address. Tel.: C49 5323 72 2201; fax: C49 5323
72 3264.
E-mail address: [email protected] (H. Zenner).
2. Calculation of fatigue life
2.1. The simulation software mCracksim
The program mCracksim designed with Visualbasic 6.0,
Visual CCC and Delphi can be operated with Windows and
was developed to simulate the material damage process caused
by microcrack growth. Our first aim was to predict the
microcrack distribution as a function of their orientation as
well as the simulated crack growth behaviour up to the defined
final crack length of about 500 mm during a fatigue test.
Therefore, for the verification with experimental data an
adequate parameter identification is necessary. Fig. 1 shows the
flow chart of the simulation software mCracksim.
2.2. Model concept and microcrack simulation
The model must be capable to describe microcrack growth
in metallic materials subjected to cyclic loading. The
polycrystalline material is modelled as a structure of two-
dimensional hexagonal elements. The program mCracksim
offers to simulate also the possibility irregular grains up to
real crystalline structures. Individual slip systems are active in
each grain with a randomised crystallographic orientation u,
Fig. 2. Metallographic slip planes beyond that are not
International Journal of Fatigue 28 (2006) 954–962
www.elsevier.com/locate/ijfatigue
Fig. 1. Flow chart of the simulation software mCracksim.
Fig. 2. Microstructure, stress state and crack growth of simulation.
A. Ahmadi, H. Zenner / International Journal of Fatigue 28 (2006) 954–962 955
considered. The stress state in the slip plane of each grain is
dependent on its orientation and the externally applied loading.
By the following simulations in this paper only the material
surface with its plane stress conditions is considered, where the
crack grows at the material surface as shear and tensile crack.
In the case of biaxial loads s1Zs2 cracks start to grow into the
material, for this opportunity provide the simulation model
some assumption for microcrack growth simulation [6].
The nucleation point of microcracks is determined by a
random generator. Note also that the starting point of cracks is
anywhere in the grain. The shape of the microcrack seed is a
point with no spatial extension, denoting an initial crack length
of zero. It is assumed that the points of crack nucleation are
given at the beginning of the simulation and that the crack
growth starts with the first load cycle.
The growth of microstructurally short cracks (MSC, stage I)
is simulated starting from the points of microcrack nucleation
using the equations for crack growth rate developed by Miller
[9]. Further assumptions consider the additional effect of the
normal stress on the shear crack stage I as well as the
coalescence of neighbouring microcracks. Finally, after a
specific crack length has been attained a diversion of the crack
takes place from stages I to II–physically small cracks (PSC).
The initiating cause for crack growth is the stress state in the
direction of the appropriate slip plane. It is assumed that the
microcrack growth can be divided into phases of stages I and II
crack growth. During the stage-I crack growth, the cracks are
driven by the cyclic shear stress which occur in the slip planes
of the polycrystalline material. The crack growth rate depends
on the magnitude of shear stress amplitude and on the distance
(dKa) between the crack tips and the dominant microstructural
barriers, in this case the grain boundary. The microcrack
growth equation is
da
dN
� �I
ZAðDtuÞaðdKaÞ mm=cycle (1)
where Dtu is the shear stress amplitude, d is the grain size and
a the length of the microcrack. (dKa) is the crack tip distance
to the next barrier, and A and a are material parameters. This
relationship states that the crack growth rate decreases with
A. Ahmadi, H. Zenner / International Journal of Fatigue 28 (2006) 954–962956
increasing crack length (decreasing distance between crack tip
and barrier). Eq. (1) assumes that the shear cracks propagate
(stage I). In the current model, the grain boundary is regarded
to be the dominant material barrier. When the stage I-crack is
sufficiently long to permit an opening of the crack front, the
development of stage II (tensile) crack occurs. At this point, the
influence of the microstructure is limited, and crack growth can
be described by continuum mechanics. Stage II crack growth
can be described by the equation proposed by Hobson, Brown
and de los Rios [10]
da
dN
� �II
ZBðDsuÞbacKC mm=cycle (2)
where Dsu represents the tensile stress perpendicular to the
crack plane, and b, c, B and C are experimentally determined
material parameters. In the transition zone, the crack growth is
calculated by using the higher value between Eqs. (1) and (2).
In addition to this continuous crack growth, there is a sudden
discontinuous crack growth through crack coalescence [6]. In
the following simulations the final crack length is 500 mm.
The simulations in this paper do not consider the crack
growth in the depth direction of the material. Furthermore, the
deformation behaviour of microstructure, the cyclic hardening
and softening of the material, specially the out of phase
hardening and the crack opening effects have to be taken into
account in further improvement steps.
The major advantage of the presented microcrack concept
over other possible multiaxial models is to have a simplified
model with only five material parameters, which could be
delivered by the parameter identification tool in mCracksim
from uniaxial/multiaxial S–N curves of the material or
component. The verification of the simulations and exper-
iments specially for multiaxial load situations in Section 3
confirms the suitability of these damage accumulation model.
The state of damage is described as the length of maximum
micro crack.
2.3. The parameter identification tool in mCracksim
The parameter identification tool is capable of identifying
material parameters for the simulation of microcracks to
predict the number of cycles to failure (final crack length of
aiZ500 mm). As a result of the parameter identification, the
simulated S–N curves performed for uniaxial loading show a
close match with the experimental results [4].
The essential reason to integrate a parameter identification
tool in the simulation software is to describe the effects of
various loading sequences, multiaxial stress states as well as
overloads on the fatigue lifetime.
In the case of multiaxial loading, material parameters must
be determined for tension/compression loading as well as for
torsional loading. A successful identification of parameters
allows for a good comparability between the simulations and
the experiments.
The cyclic material parameters, uniaxial S–N curve data of
the investigated material and also synthetic S–N curve data can
be used as entry data for the identification tool. Note that for the
following evaluations, the adaption of the material constants
was only determined in the case of uniaxial loading and
afterward the simulation of fatigue crack growth is applied for a
random load sequence in the case of variable amplitude loading
and for constant amplitudes.
The next step is the damage accumulation by using the
application of simulation of microcrack growth. Finally, the
fatigue life is given by a defined technical crack length of about
500 mm. Fig. 3 shows how the simulation software mCracksim
operates.
3. Simulation results and discussion
As an example, the simulated crack growth behaviour for
variable amplitude loading subjected to smooth specimens of
steel SAE 1015 (Ck 15) is illustrated in Fig. 4.
It can be observed that the microcracks occur in the zone of
maximum shear stress and change their growth direction at a
specific crack size in the direction perpendicular to the first
principal stress vector, as it can be seen in the experiments. The
simulated crack length a (mm) versus the number of load
cycles is plotted in Fig. 4.
3.1. Results of the fatigue tests with variable amplitude loading
The investigated materials in this study are a SAE 1015
(Ck 15) steel (314 MPa yield stress, 456 MPa tensile stress)
and a SAE 1042 (42CrMo4V) low alloyed steel (743 MPa yield
stress, 920 MPa tensile stress). The specimen geometries are
shown in Figs. 5 and 6.
The fatigue tests with hollow cylindrical specimens of steel
SAE 1015 were carried out strain controlled under combined
axial and torsional loading with a ratio of
ga=3aZ1:33ðta=saz0:5Þ. For this evaluation, a Gaussian
random sequence of amplitudes (normal distribution) [11]
with a sequence length of H0Z1!104, RZK1 and the
irregularity factor IZ0.99 was used. The failure criterion was
the macroscopic crack initiation of the specimen.
Fig. 7 presents the test results obtained for variable
amplitude in-phase loading (proportional loading) and the
simulated fatigue-life curve subjected to smooth specimens of
SAE 1015 (Ck 15) [6]. The comparison with the test results for
proportional random loading confirms the results of the
simulation.
Multiaxial random fatigue tests with notched specimens of
SAE 1042 (42CrMo4V), notch factor for bending, KtBZ2.0,
and torsion KtTZ1.6, were carried out load controlled under
combined bending, torsion and proportional loading with a
ratio of ta=saz0:5. A Gaussian random sequence of
amplitudes (normal distribution) with a sequence length of
H0Z1!106, RZK1 and the irregularity factor IZ0.99 was
used for the fatigue tests [12,13]. The failure criterion was
defined to be the failure of the specimens and the crack
initiation. Fig. 8 shows the results of simulation at RZK1 for
combined variable amplitude in-phase loading (proportional
Fig. 3. Simulation software mCracksim.
Fig. 4. Simulated crack growth for variable amplitude loading, SaZ450 MPa, steel SAE 1015 (Ck 15), smooth specimens, real grain structure (dz45 mm).
Fig. 5. Geometry of the smooth specimens of SAE 1015 (Ck 15) and the multiaxial MTS-deformation transducer.
A. Ahmadi, H. Zenner / International Journal of Fatigue 28 (2006) 954–962 957
Fig. 9. Fatigue test results for variable amplitudes under non-correlated loading
(non-proportional loading) and proportional loading, load controlled, notched
specimen, simulation and experiment.
Fig. 6. Geometry of the notched specimens of SAE 1042 (42CrMo4V) (notch
factor for bending, KtBZ2.0, and torsion KtTZ1.6).
Fig. 7. Fatigue test results under combined variable amplitude loading
(proportional loading), strain controlled, smooth specimen, simulation and
experiment.
A. Ahmadi, H. Zenner / International Journal of Fatigue 28 (2006) 954–962958
loading). The simulated fatigue-life curve shows similar results
to those of the experiments.
It can be observed that the accuracy of the lifetime
calculation in the case of proportional loading or loading
cases with small changing of first principal stress directions by
using the application of the simulation of microcrack growth
with the software mCracksim or other conventional hypotheses
is still satisfying.
Fig. 8. Fatigue test results under multiaxial proportional loading, load
controlled, notched specimen, simulation and experiment.
Fig. 9 shows a comparison of experimentally determined
and simulated results for tests with proportional and non-
correlated (non-proportional) loads. The simulated results
correspond to a large extent with the experimental data.
Compared to the experimental data for a non-correlated
multiaxial loading with a load ratio of ta/saZ0.5, the
simulation results lead to an increase of the fatigue life (factor
of 2 compared with proportional loads). From these results, it
can be stated that due to the out of phase loading, the effective
shear stress amplitude in Eq. (1) is reduced compared to the in-
phase loading. For the uncorrelated load case most multiaxial
damage concepts lead to an overestimation of the fatigue life.
The results of the simulation for all test series are plotted in
Fig. 10 and compared with the experimental data. The
simulated results show a good correlation with the experi-
mental data.
In the case of proportional loading, the effective shear stress
amplitude tu in the slip planes is increased compared to the
non-proportional loading. On the other hand, due to out of
phase loading, different deformation processes because of out
of phase hardening effects and different material behaviour
lead to a reduction or increasing of fatigue life depend on
Fig. 10. Comparison of fatigue life, simulation mCracksim and experiments,
steel SAE 1042 (42CrMo4V), notched specimens, D is damage sum.
Table 1
Material behaviour under in-phase and out-phase loading, smooth specimens (dZ908 phase shift between shear and normal stress/strain)
Material Experimental procedure NdZ908/NdZ08 Reference
saZconst 3aZconst
0.39% C steel saZconst 0.1 Tipton
StE 460 saZconst 3aZconst 2–3 0.3 Sonsino
SAE-1045 saZconst 0.4 Sonsino
SAE-1045 saZconst 3aZconst 5 0.6 Pan
25 CrMo 4 saZconst 0.6 Grun
30 CrNiMo 8 saZconst 3aZconst 1.0 0.4 Sonsino
X6CrNiTi 1810 saZconst 3aZconst 3.8 0.5 Hug
X10CrNiTi 189 saZconst 3aZconst 1.0 0.5 Sonsino
AlMg4-5Mn 3aZconst 0.8 Vormwald
Al2O36061 Al T6 3aZconst 0.2 Xia
AlMgSi 1 3aZconst 0.5 Zenner
AZ 91 3aZconst 3–4 Zenner
A. Ahmadi, H. Zenner / International Journal of Fatigue 28 (2006) 954–962 959
different experimental procedure [14]. This effect can be shown
clearly in Table 1.
According to the results taken from Table 1, load controlled
tests with smooth specimens of ductile materials, for example
Fig. 11. Modification of the Palmgren-Miner rule for taking into account
damage below the endurance limit.
Fig. 12. Statistical analysis of the cumulative damage sum and comparison of
different modification of Miner’s rule with the simulation of microcrack
growth.
of steel SAE 1045 (Ck45) under out of phase loading, reveal an
increase of fatigue life (factor of about 5), but strain controlled
a decrease of fatigue life (factor of about 0.6). Components
have always notches and notches behave strain controlled state
at the critical area, only results obtained under strain controlled
smooth specimens are relevant and realistic for the develop-
ment of practically applicable damage models. For future
developments fatigue damage models which ignore these
different material behaviour of fatigue life should not be used
for fatigue life calculations.
Note that another misconception by the fatigue cycle
counting causes an unsuccessful calculation of the fatigue
life. It is necessary to analyse the different deformation
processes during one load cycle and to quantify the failure
process [15,16]. In addition to that due to the changing of first
principal stress directions in the case of out of phase loading
more crack plane with regard to microcrack growth are active.
More active planes means higher crack density in the case of
changing first principal stress directions. As a result, a
microcrack experiences more stress cycles during one load
cycle.
Fig. 13. Comparison of fatigue life, simulation mCracksim and experiment,
notched specimens, variable amplitude loading.
Table 2
Mean values and scatter ranges of damage sums for simulation of microcrack
growth and nominal stress concept with different modifications of Miner’s rule,
notched specimens, variable amplitude loading
69 test results (DABEF) Deff TD
Simulation mcracksim 0.99 3.2
Miner elementary 0.49 6.1
Miner modified (Haibach) 0.32 5.4
Miner consistent 0.32 6.0
Miner modified (Liu-Zenner) 0.85 2.9
Fig. 14. Frequency distribution of microcracks stage I versus c
A. Ahmadi, H. Zenner / International Journal of Fatigue 28 (2006) 954–962960
Please note that for notched specimens the notch region is
not modelled. But the simulation takes into account of different
strain/stress gradients around the notch by identifying the
material parameters from the component S–N curves, see
above Section 2.3.
3.2. Reliability of fatigue life estimation
For validation of the simulation concept, 11 various steel
materials with 69 experimental horizons taken from DABEF
database [17] (failure criterion fracture) were evaluated.
rack orientation in experiment and simulation, SAE 1015.
A. Ahmadi, H. Zenner / International Journal of Fatigue 28 (2006) 954–962 961
All test series were carried out with different notched
specimens and a Gaussian random sequence of amplitudes
with the irregularity factor IZ0.99 was used for these
investigations. The determination of effective damage sums
and scatter band TD (log. normal distribution) are given by the
following equations:
Deff ZNexperiment; 50%
Nsimulation; 50%(3)
and
TD ZD90%
D10%(4)
Fig. 12 illustrates the statistical analysis of the cumulative
damage sum for the simulation of microcrack growth and
comparison with the nominal stress concept by different
modifications of Miner’s rule, Miner elementary, Miner
modified (Haibach), Miner consistent and Miner modified by
Liu-Zenner [18]. In the modification of Miner consistent is
assumed that the endurance limit decreases continuously by
increasing the damage [20]. Fig. 11 indicates clearly the
different modifications of Miner’s rule based on the nominal
stress concept. The major object of the modifications is the
assumption that how the amplitudes below the endurance limit
are damaging.
It can be shown that the simulation software mcracksim gives
a more accurate assessment of fatigue life (TDsimZ3.18) than
the nominal stress approaches with the known modifications of
Miner’s rule, with the exception of the modification Liu-
Zenner. As a reason for that it can be supposed that the damage
accumulation with regard to the microcrack growth is not
linear and the simulation model is capable to describe the
effects of various loading sequences, multiaxial stress states as
well as overloads on the fatigue lifetime quantitatively good
(Fig. 13).
Table 2 gives a overview about the mean values and scatter
ranges of damage sums for the simulation of microcrack
growth with the software mCracksim and for different
modifications of Miner’s rule.
3.3. Comparison between simulation and experimental results
The orientation, length and density of microcracks depend
on the magnitude and type of loading. The density and
orientation of microcracks taken from experimental results
with steel SAE 1015 (Ck 15) compared to the density and
orientation of microcracks obtained from simulation results are
shown in Fig. 14 for uniaxial and multiaxial loading cases.
These observations are based on surface replica studies [19].
In addition to the crack density, the variation of normal and
shear stress amplitudes acting on the crack plane are also
plotted, respectively. In all loading cases, the maximum crack
density appears in the direction of maximum shear planes. In
contrast, the experiment yields an irregular distribution of
microcracks affected by anisotropy of the material structure.
The anisotropy might be caused by the rolling process. Despite
that, the results of the simulation are in a qualitatively good
agreement with experimental results.
4. Conclusions
A microcrack simulation software is presented in this paper.
A useful feature of the software is the ability to determine the
material parameters for the simulation of microcrack growth
from uniaxial tests with constant amplitudes and to predict the
results of experimental tests subjected to uniaxial and
multiaxial loading with constant and variable amplitudes
with a reasonable accuracy. From the study described in this
paper, reliable statements are obtained about the fatigue life
estimation by using the simulation software mcracksim. In the
future, the simulation software should be extended for
achieving higher accuracy and a field for wider application in
the assessment of fatigue life for structural components.
Acknowledgements
The authors would like to thank the German Research
Council ‘Deutsche Forschungsgemeinschaft DFG’ for the
financial support of the research program.
References
[1] Sonsino CM. Influence of load and deformation-controlled multiaxial
tests on fatigue life to crack initiation. Int J Fatigue 2001;23:159–67.
[2] Varvani-Farahani A. A new energy-critical plane parameter for fatigue
life assessment of various metallic materials subjected to in-phase and
out-of-phase multiaxial fatigue loading conditions. Int J Fatigue 2000;22:
295–305.
[3] Ahmadi A, Zenner H, Glatzer M, Schram A. Microcrack growth in steel
SAE1015 subjected to cyclic loading—material analysis and simulation.
In: DVM-Bericht, vol. 685; 2004. p. S.15–S.36 [in German].
[4] Ahmadi A, Zenner H. Simulation des Mikrorisswachstums bei zyklischer
Beanspruchung—Erstellen eines Parameteridentifikation-Tools. Materi-
alprufung, Jahrg 2003;45:S.189–S.193.
[5] Ahmadi A, Zenner H. Simulation of microcrack growth under multiaxial
random loading and comparison with experimental results. Fatigue
damage of materials 2003, First international fatigue congress, Toronto,
Canada; 2003.
[6] Ahmadi A. Lebensdauerabschatzung schwingend beanspruchter metal-
lischer Werkstoffe mittels Simulation des Mikrorisswachstums. PhD
Thesis, TU Clausthal, Germany; 2003.
[7] Ahmadi A, Zenner H. Multiaxial loading and fatigue lifetime in LCF—
comparison with simulation results. Materialprufung, Jahrg 2004;46:
S.163–S.167.
[8] Murakami Y, Miller KJ. What is fatigue damage? Cumulative fatigue
damage conference, Seville; 2003.
[9] Miller KJ. Metal-fatigue—past, current and future. Proc Inst Mech Eng
1991.
[10] Hobson PD, Brown MW, de los Rios ER. Two phases of short crack
growth in medium carbon steel. In: Miller KJ, de los Rios ER, editors. The
behaviour of short fatigue cracks. London: EGF Pub. 1; 1986. p. 441–59.
[11] Haibach E, Fischer R, Schutz W, Huck M. A standard random load
sequence of Gaussian type recommended for general aplication in fatigue
testing. In: Bathgae RG, edtior. Fatigue testing and design; 1976.
p. 29.1–29.21.
[12] Esderts A, Potter K, Zenner H. Fatigue of smooth and notched specimens
under multiaxial random loading—experimental results and prediction.
ICB5MF&F, fifth international conference on biaxial/multiaxial fatigue
and fracture, Cracow, Poland. ESIS; 1997. p. 609–20.
A. Ahmadi, H. Zenner / International Journal of Fatigue 28 (2006) 954–962962
[13] Potter K. Lebensdauerabschatzung ein- und mehrachsig schwingend
beanspruchter Bauteile. PhD Thesis, TU Clausthal, Germany; 2000.
[14] Ahmadi A, Zenner H. Influence of multiaxial loading on fatigue lifetime
in LCF. LCF5, fifth international conference on low cycle fatigue, Berlin,
Germany. DVM&FEMS; 2003. p. 253-58.
[15] Yousefi-Hashtyani F. Lifetime assessment under multiaxial random
loading—calculation and tests. ICBMFF7, seventh international con-
ference on biaxial/multiaxial fatigue and fracture conference, Berlin,
Germany; 2004. p. 343-48.
[16] Ahmadi A, Zenner H. Lifetime calculation under multiaxial random
loading with regard to the micocrack growth. ICBMFF7, seventh
international conference on biaxial/multiaxial fatigue and fracture
conference, Berlin, Germany; 2004. p. 415–20.
[17] Eulitz KG, Kotte KL. Datensammlung Betriebsfestigkeit Teil 1: Stahl-
und Eisenwerkstoffe, Teil 2: Aluminium- und Titanlegierungen.
Frankfurt/ Main: VDMA Verlag GmbH; 1999.
[18] Eulitz KG, Docke H, Kotte K, Liu J, Zenner H. Lebensdauervorhersage II,
Forschungskuratorium Maschinenbau e.V. FKM-Heft 227, Frankfurt/-
Main; 1997.
[19] Schram A, Glatzer M. Statistical analysis of microcrack growth in a 1015
SAE steel subjected to different types of cyclic loading. ICBMFF7,
seventh international conference on biaxial/multiaxial fatigue and
fracture conference, Berlin, Germany; 2004. p. 217–22.
[20] FKM-Guideline. Analytical strength assessment of components in
mechanical engineering. 5th revised ed. Frankfurt/ Main: VDMA Verlag
GmbH; 2003.