Statistical Models for Analysis of Fatigue Life of Long Elements

The paper discusses some of the existing models for the analysis of the fatigue life of long elements and points out possible sources of error when data from lab tests or short elements is extrapolated to predict behavior in large structural components. A discussion of the assumptions leading to conservative and nonconservative results is included. Two dependent models, namely a Marshall‐Olkin and a nonstationary Poisson model, are presented and their asymptotic behavior is analyzed. The nonstationary Poisson model is demonstrated to be the most general solution of an empirical model formulated by Bogdanoff and Kozin. An illustrative example of aplication is also given, and several alternative approaches are discussed. Finally, some practical conclusions are given.