Curie Temperature Prediction Of Heusler Alloy Based On Svr

Due to the potentially large application value in the fields of energy and medical,the development of high-performance magnetocaloric materials has become an important direction for scientists and engineers to explore in recent decades.Currently known magnetocaloric materials with large magnetic entropy changes,including Gd and its alloys,Mn-based alloys,lanthanide compounds,etc.,have some limitations in practical applications.Different from the traditional "trial and error" and "cooking" timeconsuming and laborious materials research,this topic uses the machine learning method to plan to find a new type of magnetocaloric material with low cost and environmentally friendly Curie temperature around room temperature.Heusler type magnetocaloric materials have attracted extensive attention due to their excellent physical properties such as electrical,magnetic and mechanical properties.In this paper,the data set for machine learning is obtained by collecting the data of the Curie temperature of the Heusler alloy measured in the literature and the physical parameters of the properties of the different elements constituting the Heusler alloy.Through the data set training,the four models of ridge regression,elastic network,regression tree and support vector regression are trained to find the relationship between material performance and structure.Feature-based weight analysis and screening to optimize the four models,the most reliable model-support vector regression model.Finally,based on the support vector regression model,the Curie temperature of all possible Heusler alloys is predicted.In this experiment,the Curie temperature of 66 different compositions of Heusler alloy was collected,and 38 kinds of NiMn-based Heulser alloy data were selected to form a new data set.The reliability of the model was improved by comparing the training results of the two data sets.At the same time,39 features were collected for each Heusler alloy.Fourteen characteristics related to Curie temperature were selected by weight analysis of the features.Using the optimized model,we predicted the Curie temperature of 1078 different compositions of Heusler alloy,and obtained the Curie temperature of 50 different components of Heusler alloy around room temperature.Taking into account the cost and toxicity of the material,33 potential promising Heusler alloys were finally obtained.