Fatigue life prediction for CFRP laminates using multi-mode Lamb wave velocity and Bayesian model selection

This paper proposed a new approach for fatigue life prediction of carbon fiber reinforced polymer (CFRP) laminates using Bayesian model selection and multi-mode Lamb wave velocity. It is aimed to overcome the difficulties of large dispersity observed when testing of the material, caused by both complex damage mechanisms and fluctuation of data measurement due to the high damping property of CFRP. Specifically, both S0 and A0 mode Lamb wave velocities were recorded during a controlled tensile fatigue experiment under various load severities. Considering the possible underlying damage mechanisms, two stiffness degradation models are employed to describe the evolution trend of the collected velocity data of the two modes, respectively, forming effectively four sub models. Then Bayesian evidence is calculated by nested sampling (NS) algorithm to evaluate the strength of each sub model, the sub model with the strongest evidence is selected, the significance of which against other unselected sub models is also evaluated by Jeffrey's scale. Based on this, the fatigue life is predicted using the selected sub model, which shows good accuracy and consistency, and the effectiveness of model selection with multi-mode Lamb wave velocity is validated through a comparison between the results. Finally, a 'most conservative' strategy is also tested and compared for the safety requirement in practical application.