Machine learning-based crack prediction modeling of laser cladding coatings

Abstract

Because of its flexibility, precision, and strong generalization capacity, machine learning offers a whole new research viewpoint to the fields of materials science and engineering when compared to traditional experimental and computer simulation methods. This paper adopts laser cladding cracking research as the application background. A prediction model for small datasets is established using well-developed prediction algorithms, and a crack prediction model with superior generalizability, accuracy, and efficiency for cladding is proposed. The findings indicate that the tendency to crack increases with scanning speed, but it decreases with laser power; the random forest-based crack density prediction model has an accuracy of 90.1% and a coefficient of determination of R 2 = 0.874, which can better realize the prediction of the density and enhance some engineering practice guidelines.