Page 54 - Biaxial Multiaxial Fatigue and Fracture

Evaluation of Fatigue of Fillet Welded Joints in Vehicle Components Under Multiaxial Service Loads 39

Validation of calculated fatigue results

Calculated fatigue lives or damage sums are influenced by several factors which are worth to
be discussed in the following:
As mentioned previously, the local stress approach is applied in practice in connection with
the submodelling technique. Within the latter, the usual procedure is to take displacements
(and rotations) from a coarse mesh as boundary conditions for the finely meshed submodel.
Attention should be paid to the requirement that the stiffnesses of the models do not differ too
much. For example, a very stiff submodel will yield much higher stresses under applied
displacements than a compliant submodel would. In the current investigation we found
differences in resulting bending and torsion moments in the tube in the order of 5%. The
higher stresses occurred for the stiffer new design variant. This is the reason for the lower
predicted lives for this design. Some inherent uncertainties are obviously linked with the
submodelling'technique itself.
Calculated results also depend on the multiaxial fatigue criterion used. In general
nonproportional loading cases, criteria based on conventional equivalent stresses (Tresca, von
Mises) are inappropriate. In special cases with locally proportional stress situation, they might
work to a limited extent as well. The hypothesis of the effective equivalent stress introduced
by Sonsino and Kuppers [21] showed good predictions for welded flange-tube joints from
fine-grained steel FeE 460 under bending and torsion with constant and variable amplitudes.
As mentioned above, these cases - dominating uniaxial stresses for example - have some
practical relevance. In case of doubts on local stress states, critical plane criteria should be
preferred: Critical plane - normal stress (mode I) or critical plane - shear stress (mode II+m).
The first one should normally be used when normal stresses are dominant. The second
criterion is appropriate in the case of dominating shear stresses.
Scatter of fatigue lives is an unavoidable matter of fact. It should be taken into account
when comparing calculated and experimentally determined lives. The number of tested
components here is quite low; thus, even mean values of lives are subject to uncertainties. On
the other hand, the baseline stress-life curves for prediction, Figs. 7 and 15, are only mean
curves in a scatter band. Within the local stress approach, the ratio T for probabilities of
survival of 10% to 90% (in stresses) is 1.5 for normal stresses and 1.39 for shear stresses,
respectively [ 19,201. Thus, a factor of 2 to 3 in lives can easily arise from this fact and can be
qualified as minor inaccuracies.
At last, the real fatigue life mainly depends on the geometrical form and quality of a single
weld. It is possible to model the real welded form or geometry from a drawing. In this report
the geometry of already existing welds is modelled. However, the plane surface of the weld
and the notch radius lmm are idealisations. But geometry and stress distribution depend on
each other. Therefore, the element with maximum damage sum is not the only one to be
looked at. Other elements with similar damage sums should also be regarded as failure-critical
as well.
These failure-critical locations can be determined quite accurately using the local stress
approach. Using the hot spot stress approach, only one critical location on the tube has been
detected. This location is verified and one more location has been detected with the local
stress approach. The corresponding finite element results for the older geometry verified the
weld root as failure-critical. This is in good agreement to the experimental results. After
modification of the tube to the new geometry the weld root is no longer failure-critical.
Figure 23 shows a tested component with new geometry. A further effect is shown: The
crack initiation does not start from the weld at all. The notch at the tapering of the tube far
away from the weld undercut is failure-critical. Finally, this result prevented the presentation
of experimental fatigue lives for the new design (for the weld undercut). On the one hand, it is
possible to calculate lives for this location based on local stresses or strains and on cyclic