Automated Fatigue Data Fitting Using Structural Shift Detection

Abstract

A set of data fitting algorithms are presented for determining the cyclic material properties used in the local strain approach to fatigue life prediction. These properties are obtained by fitting power-law curves to the elastic and plastic components of uniaxial strain-life fatigue data obtained under constant amplitude testing. It is well known that these power-law relationships may be expected to be valid only within a limited range of strain amplitude; it is not uncommon, however, to have significant amounts of test data outside of this range. The algorithms presented here address this problem by applying a structural shift detection scheme to the elastic and plastic component life data sets to identify the limits of the valid range. The curve fits obtained are then the fits of only the data lying within this range. The fitting approach is illustrated for two sets of SAE 1045 steel data. In the examples shown, proper fitting of the curves is seen to improve the quality of life prediction by as much as a factor three.