Probabilistic fatigue parameter estimation in crack growth models using solely S-N data

Abstract

This study developed a K-S minimization-based probabilistic method (KSMP) for estimating fatigue parameters in the Linear Elastic Fracture Mechanics (LEFM)-based crack growth models using solely stress-life (S-N) data without precise crack length measurement. The proposed method was applied to both Paris' law and the Walker model, considering the effects of various stress ratios (R). The probability distribution of fatigue parameters was estimated by minimizing the Kolmogorov-Smirnov (K-S) statistic between the probability density of the number of cycles to failure (Nf) obtained from fatigue tests and that predicted via Monte Carlo simulation. Fatigue experiments were conducted on 62 single edge notch tension (SENT) specimens made of S355 steel under R = 0.5. The fatigue parameters were estimated from 50 specimens and validated against the remaining 12 specimens using conventional estimation method (CE), single-point optimization method (SPO), and KSMP. The predicted fatigue parameters by the KSMP method were comparable to those from the CE method involving precise crack length measurements. The SPO method also followed the general trend of crack propagation but showed a risk of non-conservative predictions. Overall, the proposed KSMP method offered a robust and less labor-intensive method for remote fatigue life assessment, accurately predicting fatigue life for structural members.