Machine learning guided prediction of mechanical properties of TPMS structures based on finite element simulation for biomedical titanium
Jiwu Zhang, Jingxiao Zhao, Qiguo Rong, Weibin Yu, Xiucheng Li, R.D.K. Misra
Abstract
In the present study, we predict elastic modulus of triply periodic minimal surface (TPMS) structures for biomedical material, titanium, using three different machine learning (ML) methods (Random Forest, XGBoost and Adaboost). A dataset is generated from elastic finite element analysis, which model has large number of lattice-cells (4 × 4 × 4 lattice-cells). In terms of three manufacturing features including unit configuration and two structural parameters (k and C), the elastic moduli of TPMS structures are calculated. It was found that all methods have high R2 and low mean square error (MSE). The Adaboost performed best (R2 = 0.959, MSE = 0.532) and the RF performed worst (R2 = 0.929, MSE = 0.923). This shows that ML methods realise a leap from limited results of finite element analysis to theoretically infinite results with ML model, and computing efficiency has also been greatly improved.