Research On The Prediction Of Thermal Conductivity And Elastic Modulus Of Mechanical Engineering Materials Based On Machine Learning

In mechanical engineering practice,the thermal conductivity and elastic modulus of engineering materials play an important role in product design and quality.However,due to the changes in material properties caused by the material processing process,through a large number of experimental observation data,intuitively put forward hypotheses and verify hypotheses,the traditional methods based on theoretical research,experimental analysis and computational simulation have been unable to meet the intelligent development of modern manufacturing demand.Currently,the use of machine learning frameworks to build material research and design platforms to analyze and predict material big data resources has become an important means of developing new materials.Therefore,it is of great significance to explore new research methods to assist in the development of thermal conductivity and elastic modulus of mechanical engineering materials.In response to the above problems,this paper proposes a prediction method based on machine learning to predict the thermal conductivity and elastic modulus of mechanical engineering materials.The focus is on supplementing and excavating traditional materials disciplines based on experiments.The calculated data is used to conduct multi-level research and analysis of material properties from micro and macro scales to help researchers quickly screen out ideal mechanical engineering materials.The research content of this article is mainly reflected in the following two aspects:(1)Using Pymatgen to download data from the Materials Project database,a set of 232 thermal conductivity data and a set of 1108 elastic modulus data obtained as the original data of this experiment.Through feature engineering,the feature vectors of three molecular-based material characterization methods(Magpie,Atom2 Vec and One-hot)are selected as descriptor inputs,combined with commonly used machine learning algorithm models(KRR,SVR and RF)for material thermal conductivity and elasticity.(2)In order to reduce the prediction error and improve the model prediction ability,the 132-dimensional feature vector generated by Magpie is used as the feature input,and the activation function is selected to establish the FCNN prediction model.Draw a scatterplot,realize data visualization,and compare the prediction results of the FCNN model with KRR,SVR,and RF to verify the effectiveness of the method.