Machine learning-aided wear location identification and friction optimization of textured artificial joint surfaces

The friction in artificial joints accelerates wear and impacts service life, with wear locations that evolve over time. This study investigated Ti6Al4V surfaces with four texture shapes (circular, square, triangular, and hexagonal) to investigate the influence of texture size, density and depth under simulated body fluid (SBF) lubrication. Friction experiments revealed that square textures achieved the lowest coefficient of friction (COF) (46.5 % reduction), while hexagonal textures provided stable performance, characterized by the smallest standard deviation (0.018) and kurtosis (2.89) of COF, indicating minimal fluctuations and smoother friction behavior. The COF data was classified by machine learning to identify wear locations with an accuracy of 93.49 %, with the mean value played a dominant role for the classification. This study provides a basis for optimizing surface textures to reduce wear, extend joint lifespan, and improve prosthesis replacement timing through wear location identification.