Application Research Of Machine Learning In Predicting The Hardness Of High-entropy Alloys

High-entropy alloys are new metallic materials with multiple elements,high miscibility and homogeneous distribution.This highly miscible composition and homogeneous distribution make high entropy alloys have excellent properties such as thermal stability,oxidation resistance,high strength,corrosion resistance,wear resistance and other properties.The traditional way to explore and design new high-entropy alloy materials is the inefficient"trial and error"method.Using machine learning techniques,useful information can be extracted from large amounts of material data to help researchers quickly predict the properties of new high-entropy alloys.Machine learning is an efficient tool in accelerating the discovery of new materials and the study of the properties of high-entropy alloys.Hardness is one of the important physical properties of high-entropy alloy materials.Therefore,this paper investigates the application of machine learning in hardness prediction of high entropy alloys.The main work of this paper is as follows:(1)Based on the composition of Al_xCo_yCr_zCu_uFe_vNi_w system high-entropy alloys,this paper predicts the hardness of high-entropy alloy materials using algorithms such as support vector regression,ridge regression,MLP neural network regression,and random forest regression,and compares and analyzes the results.The results show that the high-entropy alloy hardness prediction models based on support vector,ridge regression,MLP neural network regression,and random forest all have high prediction accuracy,with the MLP neural network regression model having the best prediction performance.(2)This paper constructs a model based on the double-layer Stacking integrated regression algorithm,using support vector regression,ridge regression,neural network regression,and random forest regression models as the first layer base learners,and using multiple linear regression models for integration in the second layer to predict the hardness of high-entropy alloys.The results show that compared to single base learners,the double-layer Stacking integrated regression model has better prediction performance and higher generalization ability.This paper provides a valuable reference for the study of high entropy alloys.In the future,more machine learning-based methods can be used to explore other performance prediction issues,such as the strength and plasticity of materials,to further improve the research level and application value of high-entropy materials.