Numerical analysis and neural-network modeling for predicting the thermo-viscoplastic behavior of Cu/AA2030 panel during the hot rolling process

Abstract

This study presents a three-dimensional thermo-mechanical finite element (3D-FEM) investigation of the hot rolling process for Cu/AA2030/Cu laminated composite panels, coupled with an artificial neural network (ANN) surrogate model for rapid prediction of process responses. The numerical model was developed to evaluate the effects of major process parameters, including friction coefficient, reduction ratio, roll speed, roll diameter, and initial copper layer thickness, on rolling force, torque, temperature distribution, and deformation behavior. Temperature- and strain rate-dependent material properties of Cu and AA2030 were incorporated into the finite element model using experimentally obtained hot compression data. Based on the numerical database generated by the FEM simulations, a feed-forward ANN was trained to predict the rolling force and torque within the investigated operating conditions. The developed ANN achieved correlation coefficients (R) of 0.9749 and 0.9742 for the training and testing datasets of rolling force, respectively, while corresponding values for rolling torque were 0.9811 and 0.9810. The maximum rolling force and torque obtained from the simulations were 32.5 kN and 0.445 kN·m, respectively. Increasing the friction coefficient from 0.1 to 0.5 increased the rolling force by approximately 33.8%, demonstrating the dominant influence of interfacial friction on the thermo-mechanical response of the laminate. The proposed FEM–ANN framework provides an efficient computational tool for predicting process responses and analyzing the influence of processing parameters within the investigated hot rolling conditions.