Research on Performance Prediction Model of Metal Structural Materials Based on Machine Learning Algorithm

The traditional performance evaluation of metal materials depends on trial and error, which is expensive and time-consuming. Machine learning algorithm, especially deep learning, has brought revolutionary changes to this field because of its powerful ability in data processing and pattern recognition. In this paper, a Deep Random Forest Regression Network (DRFRN) model is proposed, which combines the feature extraction ability of the deep neural network with the high robustness and interpretability of the RF algorithm to realize the accurate prediction of the properties of metal materials. In the data preprocessing stage, abundant data of metal materials are obtained from databases such as Materials Project and NIST Inter-Laboratory Materials Database, and cleaned, standardized and principal component analysis are carried out to reduce the data dimension. In the model construction stage, the compound loss function is designed, and the strategy of phased training, early stop and dynamic learning rate adjustment is adopted to optimize the model performance. The experimental analysis shows that DRFRN model shows high prediction accuracy and good data fitting ability when predicting the key performance indexes of stainless steel, aluminum alloy and titanium alloy. Compared with support vector machine (SVM), decision tree (DT) and convolutional neural network (CNN) models, DRFRN model has obvious advantages in prediction accuracy, robustness and generalization ability. This study not only provides a new solution for the performance prediction of metal materials, but also provides a reference for the performance prediction in other fields.