Determining Homogenization Parameters and Predicting 5182-Sc-Zr Alloy Properties by Artificial Neural Networks

Abstract

Artificial neural networks (ANNs) were established for the homogenization and recrystallization heat treatment processes of 5182-Sc-Zr alloy. Microhardness and conductivity testing were utilized to determine the precipitation state of Al3(ScxZr1−x) dispersoids during the homogenization treatment, while electron backscatter diffraction (EBSD) and transmission electron microscopy (TEM) were used to observe the microstructure evolution of the alloy. Tensile experiments were performed to test the mechanical properties of the alloy after recrystallization annealing. The two-stage homogenization parameters were determined by studying the changes in microhardness and electrical conductivity of 5182-Sc-Zr alloy after homogenization with the assistance of artificial neural networks: the first-stage homogenization at 275 °C for 20 h and the second-stage homogenization at 440 °C for 12 h. The dispersoids had entirely precipitated after homogenization, and the alloy segregation had improved. A high-accuracy prediction model, incorporating multiple influencing factors through artificial neural networks, was successfully established to predict the mechanical properties of the 5182-Sc-Zr alloy after annealing. Based on the atomic plane spacing in HRTEM, it was determined that the Al3(ScxZr1−x) dispersoids and the Al matrix maintained a good coherence relationship after annealing at 400 °C.