Machine Learning-Based predictions of crack growth rates in an aeronautical aluminum alloy

Recent advancements in computational capabilities have made data -driven approaches increasingly valuable for accurately simulating complex engineering phenomena. These approaches provide mathematical approximations that serve as surrogate models, replacing complex explicit mathematical relationships with significantly reduced computational demands. In this study, we propose the development of a surrogate model to predict the crack growth behavior of aluminum 7075-T6, a commonly used material in the aviation industry. The crack growth behavior of this alloy is complex and challenging to describe with explicit mathematical equations due to multiple influencing factors. Notably, the "waviness" behavior observed in the linear Paris law region, coupled with the significant impact of the R -ratio and specimen thickness, complicates crack growth predictions. To address this complexity, four different machine learning regression algorithms are studied: Random Forest (RF), Gaussian Process Regression (GPR), K-Nearest Neighbors (KNN), and Kernel Regression (KR). Their performance is evaluated and the most promising algorithm is selected to predict a crack growth benchmark problem of a central crack in aluminum 7075-T6 plate under various fatigue stress spectra.