Investigation of deformation behavior of Mg-2Ho alloy based on hyperbolic sine constitutive equation and PSO-BP neural network

Abstract

In this study, uniaxial hot compression experiments were conducted on as-cast Mg-2Ho binary alloy that had undergone heat treatment. The hot deformation behavior of the alloy and the influence of strain rate (0.001–1 s−1) and deformation temperature (300–450 °C) on the mechanical properties were systematically studied. The microstructure and dynamic recrystallization (DRX) mechanisms of the alloy were analyzed in detail using scanning electron microscopy (SEM) and electron backscatter diffraction (EBSD) techniques. A strain-compensated Arrhenius constitutive equation based on Zener-Hollomon parameter and a particle swarm optimization-backpropagation artificial neural network (PSO-BP ANN) based on machine learning (ML) were established to study the flow behavior of the alloy. The experimental results demonstrate that the flow stress decreases with increasing temperature and/or decreasing strain rate. The elevated temperature promotes the occurrence of DRX, resulting in a softening behavior observed in the stress-strain curve. The DRX mechanisms for alloys primarily involve continuous dynamic recrystallization (CDRX) and discontinuous dynamic recrystallization (DDRX). Further evaluation was conducted by using the correlation coefficient (R2) and mean squared error (MSE) to assess the constitutive equation and PSO-BP ANN. The prediction accuracy of the constitutive equation is poor (R2 =0.924), whereas the measured flow stress and predicted flow stress by PSO-BP ANN exhibit good consistency (R2 =0.99948), making it widely applicable for analyzing and predicting the hot deformation behavior of Mg-2Ho alloy under different experimental conditions.