Prediction of mechanical properties of ZL702A based on neural network and regression analysis

Abstract

The quantile regression neural network (QRNN) has shown high potential for predicting the mechanical properties of the alloy. The QRNN model and the regression model were developed to predict the mechanical properties of the low-pressure cast aluminum alloy ZL702A using the mechanical properties, the temperature, and the microstructure data, and the prediction accuracies of the two prediction models were compared in this article. The regression model predicted better for the screened data, while the QRNN model predicted better for the unscreened data. Finally, the evolution characteristics of the microstructure with temperature are analyzed, and it is found that the changes of SDAS and composition with temperature are the main reasons for the changes of material properties with temperature. After the analysis and comparison, it is determined that the QRNN model predicts the mechanical properties more concisely and accurately.