Research on Multiscale Performance Prediction Method for Elastic Composite Materials Based on Data-Driven Approach

This study aims to develop a data-driven multiscale performance prediction method for elastic composite materials to improve the efficiency and accuracy of material design. By integrating advanced machine learning techniques with multiscale analysis concepts, an efficient and accurate prediction model is constructed. This model can extract grain size distribution and phase volume fraction from microstructure images as microscopic features, and take macroscopic processing conditions as macroscopic features, achieving the fusion of multiscale features through feature concatenation. Using neural networks as the prediction model, the fused multiscale features serve as input, while tensile strength is the output to train the model. The accuracy and generalization ability of the neural network are validated with an independent test dataset, followed by necessary adjustments and optimizations based on the evaluation results. Experimental outcomes indicate that the model excels under high-temperature and high-pressure processing conditions, and offers significant computational efficiency advantages over traditional physical experiments. This study’s multiscale performance prediction framework provides a robust tool for the design, optimization, and application of elastic composite materials, aiding in the achievement of customized material properties.