Accelerated Fatigue Tests for reliability estimation of chassis parts

Abstract

In order to assess the fatigue strength, the StairCase method is extensively applied, thanks to its independence from physical parameters and because of it provides reliable results using few parts, thus involving low testing time. Nevertheless, the Dixon &Mood (D&M) estimation of StairCase results does not always permit to obtain a reliable estimation of the scatter parameter for the fatigue strength (section 3.1), which is an essential feature for the overall failure risk management. Moreover, we even search for a more reliable mean estimation. The likelihood estimation, used in the Maximum Likelihood Estimation of the fatigue limit (MLE) (.3.2) and in Accelerated Life Testing estimation of the number of cycles to failure (ALT) (3.3), includes hypothesis on the mechanical acceleration model but leads to obtain good estimations of each parameter included in the method. We saw that for large sample we found a good estimation with all the methods: both D&M and MLE approaches give good estimations of the fatigue strength distribution, and the ALT approach gives a good estimation of the number of cycles to failure distribution. For small samples, i.e. the most common situation in the industry, we have found, on average, a good estimation of the mean, both on the fatigue strength's mean and on the number of cycles to failure's mean. But best estimations are found by MLE or ALT method. For the standard deviation estimation it is clear that the D&M estimation cannot be applied to small samples. For MLE estimation we have many outliers which may be removed by analyzing special cases giving those values.