A study of multi-segment fatigue crack growth data analysis procedure for probabilistic crack growth prediction

Commonly used fatigue crack growth prediction models estimate life in a deterministic manner. Realizing that several variables influence fatigue crack growth, it is pertinent to assess crack growth using probabilistic models. In the present work, a probabilistic model is studied using continuous and segmented crack growth rate data models. It is observed that the prediction of life using Paris constants from continuous data model is accurate only in a finite region of the crack growth. The model based on segmented data provides more accurate life predictions with lesser variance with experimental data than the continuous data model.

► The present work is an improvement of an existing probabilistic crack growth model.
► Existing model does not capture the scatter at near-threshold region and rupture region.
► The problem is addressed by segmenting the data range.
► The overlapping of segments improved the prediction accuracy.
► The new model reduces the standard deviation of the errors.