Machine Learning-Driven Prediction and Optimization of Mechanical Properties of Selective Laser Melting-Produced Ti-Grade 5 Implants


MERAL T., Yilmaz A. F., KORKMAZ M. E., Gunay M.

JOURNAL OF MATERIALS ENGINEERING AND PERFORMANCE, 2026 (SCI-Expanded, Scopus) identifier

Özet

The primary focus of this research is to investigate the influence of selective laser melting (SLM) process parameters-specifically laser power (LP), scanning speed (SS), and hatch spacing (HS)-on the mechanical quality of porous, wedge-shaped biomedical implants. These implants were fabricated utilizing Gyroid and Schwarz triply periodic minimal surface (TPMS) unit cell structures. A total of 54 implant specimens, developed from Ti-Grade 5 alloy, were prepared with a target 75% porosity. The experimental strategy was structured using the Taguchi L27 orthogonal array. All prepared samples underwent compression testing for the determination of Young's modulus and yield strength values. The findings indicate that mechanical behaviors are highly contingent upon the interaction among the SLM parameters, with scanning speed exhibiting the most pronounced effect based on ANOVA. The study identified the optimal processing conditions as a scanning speed of 700 mm/s, laser power of 210 W, and hatch spacing of 0.08 mm. Under these optimized parameters, maximum Young's modulus values reached 17232 MPa for Gyroid and 16009 MPa for Schwarz structures, while peak yield strengths were recorded at 148.79 MPa and 133.51 MPa, respectively. To facilitate predictive design, a Random Forest regression model was implemented, achieving exceptional fidelity with a coefficient of determination R2 >= 0.99 and relative root mean square error (RMSE) values consistently below 2% for both topologies. This combined experimental and machine learning methodology offers a robust foundation for the optimization of SLM parameters in the customization of implant mechanical properties. These results underscore the potential for integrating advanced design geometries with data-driven design models, suggesting that this integration can optimize performance and accelerate the development of patient-specific biomedical implants.