Uncoupled ductile fracture initiation model for 5052 aluminum alloy with machine learning assisted identification of the material parameters

Abstract

Ductile fracture is the predominant failure mode in plate forming; analyzing and predicting this fracture phenomenon is essential for enhancing the forming process and improving product quality. In this paper, the plastic model (modified Bai-Wierzbicki model) and the uncoupled ductile fracture criterion (Lou-Huh criterion) associated with two stress state parameters were used to construct an uncoupled model to predict the ductile fracture initiation of 5052 aluminum alloy, and a new method of machine learning assisted identification of the material parameters of the Lou-Huh criterion was proposed. This method overcame the difficulties of the traditional optimal fitting method, which requires a large amount of stress state information, and had a simple and easy operation procedure. The study results show that the uncoupled ductile fracture model can accurately predict the ductile fracture initiation of 5052 aluminum alloy, and the machine learning assisted calibration method can obtain more accurate material parameters than the optimal fitting method.