Machine Learning-Based Investigation of the Compressive Behavior of TI6AL4V Lattice Structures

Abstract

This study investigates the prediction of the effective Young's modulus of Ti6Al4V lattice structures using four regression-based machine learning models, namely Linear Regression, Polynomial Regression, Support Vector Regression, and Gaussian Process Regression, developed in MATLAB from 25 sets of Finite Element Analysis simulation data. Ti6Al4V was selected for its biocompatibility and corrosion resistance, with accurate Young's modulus prediction being critical to minimising stress shielding between bone implants and surrounding bone tissue. Among the models evaluated, Polynomial Regression and Gaussian Process Regression demonstrated the strongest predictive performance, achieving the lowest RMSE and highest R² values in actual versus predicted comparisons. Learning curve analysis and 5-fold Cross Validation revealed that optimal model training required only 36–52% of the total dataset, highlighting a significant opportunity to reduce computational cost by limiting the number of Finite Element Analysis simulations needed. These findings demonstrate that machine learning regression models can efficiently and accurately predict the mechanical properties of lattice structures, supporting more precise control of implant porosity in biomedical engineering applications.