Predictive Modelling of the Surface Microgeometry of Laser-Modified Biocompatible Ti-TiB2 Powder Metallurgy-Based Composites

This paper deals with the surface functionalization of a biocompatible powder metallurgy-based Ti-TiB2 composite prepared by spark plasma sintering, utilizing a nanosecond fibre laser operating at a wavelength of 1064 nm in an ambient atmosphere. The effects of laser pulse energy (Ep), longitudinal laser beam spot overlap (OL), and transverse spot overlap (OT) on the integrity of laser-treated surfaces were investigated using a Box–Behnken experimental design to identify how these process parameters govern the amplitude and spatial roughness metrics strongly correlated with surface osseointegration. Statistical learning-driven analyses were performed to evaluate the predictability and overall functional dependency of the observed data, including Ordinary Least-Squares and regularized linear regression analyses, analysis of variance, and model evaluation using two approaches: first, by training and testing on the entire dataset to assess the in-sample fit, and second, via Leave-One-Out Cross-Validation to evaluate the model’s generalization performance. An underlying linear dependency was demonstrated for Rsm with predictors Ep and OL, and no statistical significance was found for the predictor OT. For Rsm, the model retained moderate predictive performance under LOOCV, despite the limited number of independent experimental observations.