Prediction of the elastic properties of materials based on machine learning and visualization analysis

Abstract

The elastic modulus is an important performance parameter that measures the ability of materials to resist deformation and is critical for assessing their reliability and stability. Thus, the elastic modulus serves as a guide in engineering design and material selection, and finding materials with specific elastic properties is a hot issue in the field of novel materials synthesis. Predicting elasticity quickly and accurately is of great significance in engineering. It is not practical to measure the elastic properties of many materials using practical experiments because this requires a significant amount of time and cost. For many material samples, each sample needs to be tested and analyzed, which is a time-consuming and expensive task. Thus, screening material data through computer simulation, choosing candidate materials, and then confirming them through actual experiments is an ideal method for new material discovery. Currently, the main calculation methods for material performance prediction are first-principles high-throughput calculation, which is inefficient and difficult to efficiently complete the high-volume material screening. Machine learning prediction methods based on material statistics can rapidly predict material properties through big data mining, which has become a possible alternative to high-throughput computing. In this work, the feature selection method and machine learning model are integrated to choose the most effective combination scheme for elastic modulus prediction, and an interactive interface is developed to perform a visual analysis of the relationship between input features and elastic properties of materials. For the analysis of the prediction results, the root mean square error (Rmse) and R-Square (R2) are employed as evaluation indicators for the performance of the prediction model. The experiment shows that the Pearson/RFE/LASSO-GBDT combination model possesses the best performance. On the other hand, by visualization analysis, it is revealed that the energy of each atom, melting point, density, and other characteristics have a great effect on the prediction results. These important characteristics can be utilized to preliminarily predict the range of elastic moduli from the feature–target relationship, and the value of target attributes can be used for the estimation of important characteristics of materials. These findings can be applied to investigate the influencing factors of elasticity, predict the properties of large quantities of materials, and guide the synthesis of materials by visualization analysis. This work has certain significance for guiding the discovery of novel materials and exploring the influencing factors of material properties.

Keywords

• feature selection
• machine learning
• elastic property prediction
• visualization
• data mining