Predicting temperature of semi-solid 7075 aluminum alloy slurry prepared by a novel rheocasting route via neural network

Abstract

The rapid preparation of desired semi-solid slurry with a specific average temperature and a high temperature uniformity is of essential importance for the rheocasting of high-strength aluminum (Al) alloys. In this study, a novel semi-solid slurry preparation process, namely Enthalpy Control Process (ECP), has been proposed for rapid preparation of 7075 Al alloy slurry. The machine learning approach from the aspect of neural network was utilized to predict the average temperature and maximum temperature difference of the Al slurry prepared by ECP. The results indicated that the temperature values predicted by Back Propagation Neural Network optimized by Genetic Algorithm (GA-BPNN) models exhibit an excellent fit with the experimental results. The average relative error and the coefficient of determination for average temperature of were 0.9988 and 0.027%, while the values for the maximum temperature difference were 0.9910 and 3.4%, respectively. The GA-BPNN has a great latent capacity in predicting the temperature of the semi-solid slurry.