Abstract: A probabilistic model is suggested for charactering the random fatigue crack growth. It is considered of the probabilistic effects of scattering regularity and amount of the test data. The model is established on an improved Paris equation, which is developed from a consideration of better-fit effects and more application of the test information. The parameters of model are measured by taking into account the stochastic characteristic of entire test data, which is connected with a maximum likelihood method and math programming. Availability and feasibility are indicated by an analysis of the 16MnR steel test data.

Key words: Confidence, Crack growth, Fatigue, Probabilistic analysis