Fatigue Life Prediction in Service - A Statistical Approach

Doctoral thesis, 1996

This thesis treats the problem of fatigue life prediction. The emphasis is on the development of simplified methods, intended to be used in engineering design and the starting point is the established methods for fatigue prediction in service, namely the Wöhler curve, the Paris´ law and the Palmgren-Miner law of cumulative damage. The thesis can be separated into three parts: a) the investigation of the level crossing model, b) the generation of pseudo-random processes for laboratory tests, and 3) an overall study of the prediction problem from a statistical point of view. a) A new model for fatigue life at variable amplitude, based on level crossing properties, is investigated and modified in order to take sequential effects into account. Comparisons are made between the modified level crossing model and the traditional model based on the rain flow count. b) A method is developed for the generation of a discrete time auto-regressive process, that fulfil specifications regarding the level crossing properties and the irregularity factor. c) The overall uncertainty in prediction of fatigue life in service is separated by uncertainties originating from different sources, and the result is used as a base for discussion about the choice of models in engineering design.