Machine learning-based investigations of the effect of surface texture geometry on the wear behaviour of UHMWPE bearings in hip joint implants

Abstract

The purpose of this research is to develop data-driven machine learning (ML) models capable of estimating the specific wear rate of ultra-high molecular weight polyethylene (UHMWPE) used in hip replacement implants. The results of the data-driven models are demonstrating a high level of consistency with the experimental findings acquired from the pin-on-disk (POD) trials. With a performance evaluation of 0.06 mean absolute error (MAE), 0.17 Root Mean Square Error (RMSE), and 0.96 R², the Random Forest Regression is found to be the best model. Another machine learning model, called Gradient Boosting Regression, is also found to possess satisfactory predictive performance by having an MAE of 0.09, RMSE of 0.24, and R² of 0.96. According to the findings of a parametric analysis that made use of an ML model, the surface texture geometry has a substantial dependence on the wear behaviour of UHMWPE bearings that are used in hip replacement implants. This strategy has the potential to enhance experiment design and lessen the necessity for time-consuming POD trials for the purpose of assessing the wear of hip replacement implants.