# Accuracy in Fatigue Life Analysis

### What is a fatigue life analysis?

With a fatigue life analysis, a prediction (or better, an estimation) is made of the fatigue life of a component. This is done by comparing the component to be analysed with something that has been tested that resembles as much as possible with that component with respect to geometry, material, loading, environment, surface conditions, etc. If the resemblance is exact, the fatigue life of the component to be analysed should be the same as the fatigue life of that tested something (scatter left out of consideration for this moment). The less good the resemblance is, the more assumptions and estimations are to be made and the less good the accuracy of the fatigue life estimation will be.

The best resemblance, or similarity, is achieved with a full-scale test with the same geometry, the same material & material conditions, service simulation loading and service environment, see the figure below.

Very often, the similarity is not so good, see next figure. If that is the case, better modelling of the component and the available data towards each other is required. That implies modelling of geometry, material, loads, surface conditions and environment. The fatigue performance of a structure always depends on these items. Modelling in general implies simplification and making assumptions and estimatons. It must be realized that fatigue is a structure related problem, not just a material problem.

### Accuracy in fatigue life analysis

The accuracy of a fatigue life analysis is affected by several aspects, such as:

• Large amount of scatter.
• Accuracy of loads: small changes in stress level results in significant differences in fatigue life.
• Accuracy of S-N data: S-N Data from notched specimens better than from un-notched specimens.
• Accuracy of correction factors.
• In case of FEA: small changes in meshing may lead to significant changes in fatigue life.

The lower the accuracy of data and factors and the more these data and factors are estimated, the lower the accuracy of the prediction will be.

The diagram below gives an indication of the effect of inaccuracy of the fatigue limit on the fatigue life. Suppose a stress level 30% above the fatigue limit and S5·N=C. If the actual fatigue limit is 20% lower than assumed in the analysis, the actual fatigue life is reduced with a factor 3. Similar effects are found if an incorrect S-N curve gradient is used.