A comparison of two methods of analysing staircase fatigue test data

Abstract

When the fatigue strength of components or material is tested by the staircase or up-and-down method, the data can be analysed in either the fatigue strength domain, which is the most popular approach, or the life domain, based on life regression models. In this paper we exploit a simple relationship between families of models in the two domains to obtain an estimate of the distribution of fatigue strength directly from a fitted life regression model. We use simulation to compare the two approaches, concluding that for estimating important features of the fatigue strength distribution the method of maximum likelihood is more effective if it is applied in the life domain. In particular, it is more robust to the choice of xh, the step height of the staircase, provided xh is in the range σX- 2σXx, where σX is the standard deviation of the log-strengths. Most of our results are insensitive to the value of the initial stress amplitude. Copyright © 2007 John Wiley & Sons, Ltd.