Hybrid Prediction Model for Mechanical Properties of Low Alloy Steel Based on SVR-MLP

As research into alloyed materials continues to advance there are different types of low-alloy steel with different chemical compositions and organisations, so there is a need for more efficient methods of predicting the mechanical properties of low-alloy steels compared to experimental studies. In this paper, three data-driven regression models are developed to predict the mechanical properties of different types of low-alloyed steels. The first model, Support Vector Regression (SVR), achieves good results for the prediction of nominal yield strength and tensile strength, but not for elongation and cross-sectional shrinkage; the second model is a multilayer perceptron regression model, which is similar to but worse than SVR for the prediction of the four mechanical properties; finally, it is a hybrid model that combines SVR and Multilayer Perceptron (MLP) through a linear model, and its prediction accuracy is very The final model is a hybrid model fusing SVR and MLP with high prediction accuracy, which is significantly better than the single prediction model, confirming the usefulness of the model fusion strategy in predicting the mechanical properties of low-alloy steel.