Optimization and bioapplication of deep learning algorithm in the prediction of mechanical properties of metal matrix composites

Abstract

Listen

Translate article

This study addresses the optimization and bioapplications of a deep learning algorithm for predicting the mechanical properties of metal matrix composites (MMCs), a critical task for efficient material design. And it is also beneficial for deploring more bioapplications of MMCs. Leveraging a comprehensive experimental dataset from multiple research institutions, we employ a Convolutional Neural Network (CNN) for feature extraction and the Recurrent Neural Network (RNN) for sequence analysis. The dataset encompasses mechanical properties such as tensile strength, elastic modulus, and yield strength for diverse MMCs with varying compositions and processing conditions. The research methodology involves rigorous data preprocessing, feature selection, model development, and performance evaluation using metrics like R2 score, Root Mean Squared Error (RMSE), Mean Absolute Error (MAE), precision, and recall. Addressing the challenge of model robustness and generalizability, we utilize k-fold cross-validation for training and validation. Optimal hyperparameter settings are identified to enhance predictive accuracy. Our results reveal high predictive performance, with R2 scores ranging from 0.89 to 0.92 for different mechanical properties, thereby demonstrating the model’s efficacy in facilitating material design and optimization processes for MMCs.