See

discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/287747709

Theory and Applications of FEMFAT - A FE-

Postprocessing Tool for Fatigue Analysis

Conference Paper · June 1999

CITATIONS READS

11 104

3 authors, including:

Christian Gaier
Engineering Center Steyr GmbH & Co KG, Magna Powertrain
52 PUBLICATIONS 137 CITATIONS

SEE PROFILE

All content following this page was uploaded by Christian Gaier on 22 December 2015.

The user has requested enhancement of the downloaded file.

THEORY AND APPLICATIONS OF FEMFAT - A FE-POSTPROCESSING
TOOL FOR FATIGUE ANALYSIS

FATIGUE CONFERENCE 1999 BEIJING

Ch. Gaier*, B. Unger* and J. Vogler+

An overview about the theoretical background of the Finite Element

postprocessing software tool FEMFAT for the lifetime prediction
of components is given. The theory applied in FEMFAT combines
methods of the nominal stress concept and of the local one in a
specific way, which can be characterized by the term ”Influence
Parameter Method”. This approach has been applied successfully
for several years now and was extended recently for the case of
multiaxial loadings. The main arising problems are considered and
possible ways for their solution or diminution, respectively, are
presented. An example demonstrate the validity of these methods.

INTRODUCTION

In today’s automotive industry mechanical engineers are forced to develop light

weighted vehicles to reduce the consumption of energy. At the same time the safety
standards, which become more and more rigorous, must be fulfilled. For the
successful combination of these two contradictory requirements a lot of costly tests
have to be done. Numerical analysis of the component’s lifetime and preoptimization
before prototyping can reduce the number of testing loops and therefore time and
costs dramatically. FEMFAT is a software package for the prediction of the lifetime
and the crack initiation point of vehicle components like axles, gear housings, engine
blocks, crankshafts, connecting rods, truck frames, car bodies, etc. It exists for about
fifteen years now, is widely applied for extensive problems and its development is
permanently continued since that time (Eichlseder et al (1-5)). The assessment of
uniaxially and nowadays multiaxially loaded components including welded seams is
possible.

* Engineering Center Steyr, Steyr-Daimler-Puch AG, Schönauerstr. 5, A-4400 Steyr,

Austria
+
Audi AG, D-85045 Ingolstadt, Germany
THE INFLUENCE PARAMETER CONCEPT

1
The loadings, obtained e.g. from measures of a test-drive, and the structural stress
distribution, obtained from a linear static finite element analysis, are needed as basic
input data for FEMFAT. Further the knowledge of the specimen S/N-curve of the
used material is required, which can be approximately represented by two straight
lines. The user can choose between three types of S/N-curves: Palmgren-Miner
original, elementary and modified by Haibach (6). These curves differ by the slope
of the line representing the high-cycle-fatigue-domain. For each considered influence
parameter (e.g. stress gradient, mean-stress influence, surface roughness,
temperature, etc.) the slope of the S/N-curve in the low-cycle-fatigue-domain, the
cycle limit of endurance and the stress limit of endurance are modified in a specific
way in advance, before the actual damage calculation by linear accumulation is
started.

Methods of modifications are taken on the one hand from German guidelines as
FKM and the old Eastern German standard TGL to consider influences like surface
roughness, surface treatments (shot penn, flame hardening, ...), temperature,
technological size, etc. On the other hand own methods have been developed (1-5),
e.g. to take into account the important notch effect, which denotes the well known
fact, that the ratio of the lifetimes of specimens without and with notch is less than
the so-called ”notch-factor”, which denotes the increase of the local notch stress in
comparison to the nominal stress. For an arbitrarily shaped component it is generally
not possible to specify a nominal stress. Instead, the relative stress gradient can be
used for the quantification of the notch effect.

Another important factor is the influence of the mean stress, which is

determined by means of a Haigh-diagram as a method taken from the nominal stress
concept. Otherwise, as a part of the local one, mean-stress-rearrangements caused by
plastic deformations are treated by the Neuber-method (Neuber-hyperbola), making
use of the cyclically stabilized stress/strain-curve.

Depending on the material type (ductile or brittle, e.g. steel versus grey cast
iron) an equivalent stress is generated (von Mises and maximum principal stress,
respectively), which is compared to the tensile strength of a cylindrical test
specimen. But it must be stated, that this procedure is applicable only for
proportional loads, i.e. loads which scale the magnitude of the multiaxial stress state,
but which do not change the directions of the principal stress axes.

THE TREATMENT OF MULTIAXIALLY LOADED COMPONENTS

For multiaxially loaded components the classical damage-hypotheses are no longer

valid. A convenient method to consider situations with changing principal stress
directions is the ”Critical Plane Approach” (7). This hypothesis was used to develop
a ”multiaxial-damage-module” for FEMFAT (See Fig. 1, left side). The basic idea
of the critical plane approach is that cracking starts in the cutting plane with

2
”maximum damage”. This method can be well applied for each combination of
external loads.

Firstly, the history of an equivalent stress is generated by transforming the

channel stress tensors into several cutting planes and summing them up all together to
a resulting stress vector. Depending on the material and the load situation,
respectively, the equivalent stress can be equated with the normal or shear
component, or it can be defined as a function which is composed of both of them.
After rainflow-counting and the projection onto an amplitude-mean-stress-rainflow-
matrix, a damage calculation is performed in the actual cutting plane by following a
similar way as for the assessment of uniaxially loaded components, based on the
influence parameter method.

Similar to the von Mises-stress, but supplied with a sign, a convenient

definition of an equivalent stress seems to be (Sanetra and Zenner (8))

2
σ  2
σ e = sign(σ n ) ⋅ σ +  a
2
 τ (1)
τa
n


However, it must be stated, that (1) represents a discontinuous function, which can
deliver wrong results for some special situations. For the case of a large but steady
shear component, and for a small normal component which changes its direction
periodically, too large amplitudes are obtained from (1). Similar problems arise if
the sign is taken from the larger of the two terms below the root. Therefore a
plausibility control should to be done in any case, when using (1).

To reduce the computation effort, some restrictions have to be done. Firstly,

only nodes lying on the component’s surface are evaluated, thereby assuming, that the
stress tensors are plain and parallel to the surface. This is true if there are no
external loads and restraints acting upon the domain of interest. Secondly two
different cutting plane filters have been implemented. The first one is based on a
scheme proposed by Chu et al (9). The status of the plain stress tensor is marked by
points in the stress space, spanned by two normal and one shear component, for each
time step. The aspect ratio of the least square ellipsoid, spanned by this cloud of
points, is a measure for the degree of multiaxiality. A straight line through the origin
indicates the special case of uniaxial loading. The directions of the ellipsoid’s first
principal axis refer to the position of the critical plane. This method is recommended
for short load histories.

As a second method the upper limits of the channel equivalent stresses are
considered for each cutting plane. The maximum indicate planes, which may be
critical. But a secure prediction is not possible, because it is not known in advance,

3
how the channels will be superimposed. Otherwise this method behaves to be a very
fast one, because no load histories need to be processed in advance.

EXAMPLE

A wheel carrier from a rear axle of a passenger car was calculated by two different
ways. Firstly, three forces acting on the component (in longitudinal direction for
acceleration/rough road, in lateral direction for curve driving and in vertical
direction) have been measured on a servo-hydraulic test bench for a period of about
half an hour. These load histories were used as input for FEMFAT in common with
the corresponding stress data, which were obtained from a finite element analysis.
The finite element mesh consists of ~35000 tetrahedron elements (with ~60000
nodes equivalent to 180000 degrees of freedom) and ~1000 beam/rod-elements,
which were used for the modelling of the screws and the elastic bearings. The finite
element mesh and the stress distribution for the second load case (lateral direction),
which is the dominating one, can be seen in Fig. 2 at the left side.

Secondly, the load histories of the three components of forces acting at three
different locations were generated by a multi-body-simulation with ADAMS. Fig. 1,
right side, shows the load histories at the center of the wheel. The resulting stress
distribution for the lateral force can be seen in Fig. 2 at the right side and is almost
the same as calculated only with FEM. But there is a spot, where the stress,
calculated only by FEM, is too large, whereas the combined FE/ADAMS-method
delivers a correct result. This method considers the nonlinearities of the bushes and
kinematics and is therefore the preferable way. The damage distributions for both
variants (three and nine channels, respectively), calculated by FEMFAT, are shown
in Fig. 3. The component broke at three spots, which coincide exactly with the
predicted ones of maximum damage. But the predicted lifetime was too conservative
(about factor 10). Several reasons can be quoted to face this fact: All critical spots
are lying near the screws, which were modelled coarsely by linear beams. Therefore
the calculated stresses around the screws may be too large. A detailed FE-model of
the screws can overcome this problem. Further a linear model of the wheel bearing
was used. An extensive nonlinear analysis will deliver more accurate results.
Thirdly no mass forces have been taken into consideration.

CONCLUSION

FEMFAT is a powerful software tool for the prediction of critical spots and the
lifetime of arbitrarily shaped components, based on precalculated linear finite
element stress results. It can deal with uniaxial and multiaxial loadings and also the
assessment of welded seams is possible. As an example the results of a wheel
carrier have been presented and compared with test results.

4
 SYMBOLS USED

σe = equivalent stress
σn = normal stress component
τ = shear stress component
σa = tension/compression alternating endurance stress limit
τa = torsional alternating endurance stress limit

REFERENCES

(1) W. Eichlseder, ”Rechnerische Lebensdaueranalyse von Nutzfahrzeugkomponenten

mit der Finite Elemente Methode”, Dissertation, TU Graz, Austria October 1989.
(2) W. Eichlseder and B. Unger, ”Prediction of the Fatigue Life with the Finite Element
Method”, SAE Technical Paper Series 940245, U.S.A. 1994.
(3) W. Eichlseder and B. Unger, ”Assessment of Welded Seams with the Finite Element
Method”, SAE Technical Paper Series 950712, U.S.A. 1995.
(4) B. Unger, W. Eichlseder and G. Raab ”Numerical Simulation of Fatigue - is it more
than a prelude to tests?”, FATIGUE ’96, Berlin, Germany May 1996.
(5) G. Steinwender, B. Unger and W. Eichlseder, ”Fatigue Simulation on Finite Element
Structures”, Computational Structures Technology Conference, Hungary, Budapest
August 1996.
(6) E. Haibach, ”Betriebsfestigkeit – Verfahren und Daten zur Bauteilberechnung”, VDI-
Verlag, Germany, Düsseldorf 1989.
(7) H. Zenner, R. Heidenreich and I. Richter, ”Fatigue Strength under Nonsynchronous
Multiaxial Stresses”, Z. Werkstofftech. 16, pp. 101-112, Germany 1985.
(8) C. Sanetra and H. Zenner, ”Betriebsfestigkeit bei mehrachsiger Beanspruchung unter
Biegung und Torsion”, Konstruktion 43, Springer-Verlag, Germany 1991.
(9) C. C. Chu, F. A. Conle and A. Hübner, ”An Integrated Uniaxial and Multiaxial Fatigue
Life Prediction Method”, VDI Berichte Nr. 1283, Germany 1996, 337-348.
MEASUREMENTS FE-MODEL
CHANNEL-HISTORY
t
Channel 1

t
Channel 2
.
.
. MBS - SYSTEM
t
Channel N UNIT-STRESS
DISTRIBUTION
RFAC FOR EACH
TE LOAD CASE 1...N
ES
IN

SPECIMEN MATERIAL DATA

S/N

DAMAGE DISTRIBUTION
CALCULATION
REPORT

Fig. 1. Overview multiaxial fatigue lifetime simulation (left), load histories at the
center of the wheel (right)

5
 Fig. 2. Finite element mesh and stress results calculated with
FEM (left) and combined FE/ADAMS-method (right)

Fig. 3. Damage distribution calculated by FEMFAT using load-

histories from tests (left) and generated by ADAMS (right)

View publication stats