Crack propagation life prediction of a single lap shear joint: a linear elastic fracture mechanics based machine learning approach

Abstract

Estimation of fatigue life in the aerospace industry plays a key role in terms of safety, certification, and inspection. Individual aircraft tracking programs aim to calculate the fatigue index for each individual aircraft based on its own load history. Therefore, accurate estimation of fatigue life on the basis of unique load spectra for each aircraft plays a crucial role. The purpose of this study is to predict the crack growth life of a single lap shear joint for an aircraft based on different spectral loads using random forest regression and k-nearest neighbors regression. A finite element model was built using the Huth fastener flexibility method to obtain fastener loads and stress values on the structure under maneuver loads. Then, finite element results were compared to analytical calculations. Based on Fighter Aircraft Loading Standard for Fatigue and Fracture (FALSTAFF) spectra, 90 different spectrums were developed to be used for the calculation of crack growth life. The AFGROW software was used to calculate the crack growth life based on the load spectra and the findings of the finite element model. Analytical calculations of crack growth lives and features of individual load spectrums were fed to machine learning models. Finally, the crack growth life predictions of the machine learning models were compared with analytical calculations. According to the findings, a good correlation was observed between the analytical and predicted crack growth lives.