Comparison of Fatigue Life Evaluation Methods

Abstract: The aim of this thesis is to investigate a selected set of fatigue life calculation methods and evaluate if they are suitable for fatigue life estimation of truck components at Scania. Failure due to fatigue can be cause by road induced vibrations, which is an inevitable phenomenon trucks are exposed to. By estimating when and where these components will fail, they can be designed to reduce the amount of failure per vehicle. Three types of fatigue life calculation methods, namely equivalent stress methods, critical plane methods and spectral methods, have been evaluated. These are methods for calculating fatigue life in both the time domain and the frequency domain. The chosen calculation methods have been evaluated based on their sensitivity to input parameters, their accuracy on predicting fatigue life and their ability to find the critical areas where the components are most likely to fail. The methods have also been compared to a method already implemented at Scania. To evaluate the methods, two different components were used. The first component was designed to give a multiaxial stress state and the other was a real truck component where fatigue data had been collected from a shake rig test at Scania. It was found that all investigated methods were successful in finding critical areas where failure will occur. However, the resulting estimated fatigue life had a very low accuracy. To draw any conclusions about the accuracy of the fatigue life estimations, a model that better reflects the dynamics of the real truck component is needed. Therefore, the conclusion is that the chosen methods can be used for finding critical areas in a component but not to determine the absolute time to failure for the model used. However, the method already implemented at Scania was equally successful in finding the critical areas and it has a much shorter computational time than the methods in the time domain. Since it is already implemented and used, the Scania method is recommended for the purpose of finding the critical areas of a component. A sensitivity study was conducted in order to investigate the influence of a variation of material parameters on the fatigue life calculated with the different methods. This study showed that the SN-curve parameters are important for the resulting fatigue life of methods that consider the endurance limit, and, therefore, that the choice of SN-curve is important. Since the road induced vibrations in this study caused load signals where the majority of the cycles were found below the endurance limit, methods that account for the endurance limit have to be used for calculations on components experiencing similar conditions. Furthermore, it was found that the resulting stress signal from the FE-analysis using input data from the shake rig test was non-Gaussian, this makes the results from all the chosen frequency domain methods invalid. To use these methods, they need to be extended to consider non-Gaussian signals.