Bayesian model averaging for probabilistic S-N curves with probability distribution model form uncertainty

•Both the model form uncertainty and model parameter uncertainty are considered.•Prediction validation of fatigue data with diverse known and unknown distributions.•The proposed method gives more robustness and reliability fatigue lifetime prediction.•The probability distribution of fatigue lifetime can be predicted. Reliability analysis of engineering components or structures heavily relies on accurately estimating the fatigue properties of materials. However, significant uncertainty exists regarding the distribution form and value in fatigue data, posing significant challenges in constructing a robust probability fatigue model. To address this challenge, we propose a Bayesian model averaging (BMA) method to incorporate model form uncertainty into the estimation of the probability density of fatigue life. The performance of BMA was verified through numerical experiments using both simulated and experimental data. The results highlight the robustness and reliability of BMA compared to individual models, as it effectively incorporates model form uncertainty. The proposed BMA model offers a general framework for developing probabilistic fatigue models with high robustness and accuracy in their predictions. This model contributes to advancing the field of reliability analysis by addressing the challenges posed by uncertainty and enhancing the understanding of fatigue properties for engineering components and structures.