Improving Composite Materials with Machine Learning: A Predictive Approach

Abstract

Machine learning (ML) has become an important technology in the field of composite materials, offering efficientmethodsformaterialcharacterization, damageassessment, propertyprediction, anddesignoptimization. Conventional experimental techniques and physics-based simulations for predicting the complex behavior of composites can require considerable time, cost, and computational resources. In contrast, ML provides a data-driven approach capable of identifying hidden patterns, establishing relationships between variables, and generating accurate predictions. Algorithms such as Support Vector Regression (SVR), Random Forest Regression, and Linear Regression can process information related to material composition, filler characteristics, curing conditions, and environmental factors. This study examines the application of ML techniques to composite materials, particularly for predicting fracture toughness, characterizing damage, and optimizing mechanical properties. Correlation analysis reveals significant relationships between fracture toughness and important input parameters, demonstrating the usefulness of data-driven approaches in material development. Among the investigated techniques, Random Forest Regression demonstrates strong predictive capability and effectively represents complex material behavior. However, challenges including limited data availability, model interpretability, and generalization remain. Reliable ML performance depends on high-quality datasets, suitable feature selection, and effective model tuning. Overfitting can also affect prediction accuracy, making cross-validation and hyperparameter optimization essential. Integrating ML into composite material research can reduce experimental requirements, lower development costs, improve manufacturing efficiency, and accelerate the development of high-performance materials. Future research should emphasize larger datasets, explainable ML, and hybrid approaches combining ML with physics-based simulations.