Fatigue life prediction using a new moving window regression method

A simple method to assess the safety of metallic structures in the presence of crack-like defects is considered. Linear regression techniques are used to fit fatigue-crack growth (FCG) laws of exponential type with respect to the number of cycles, in FCG experimental results. A moving window regression method is used to accurately predict whether or not the critical crack length region is reached. The probability of fast failure at any time (number of cycles) is formulated in terms of a characteristics defect size exceeding a critical value. The concept of conditional probability of failure is also employed here in order to evaluate the failure rate used in reliability practice. As an illustrative example, an analytical closed-form solution for the failure probability of structures is presented and evaluated using experimental FCG data.