A Neural Network Solution for Comprehensive Contact Analysis in Hard-On-Hard Hip Implants

Abstract The contact mechanics of the cup–head tribo-pair in hip implants need to be studied to determine the effect of wear on the failure rates, which affect longevity. It is reported that the existing analytical models fail to comprehensively predict the contact conditions throughout a single ISO standard gait cycle. Understanding the limitations of analytical models, a novel data-driven neural network approach is performed to predict contact conditions in this research study. The model is trained with the dataset generated through the Finite Element Method (FEM). The developed artificial neural network model predicts the contact conditions with the same accuracy as FEM and less computational cost. From the Shapley additive explanation analysis, it is found that cup thickness has a significant contribution to affect the output contact conditions. Thus, it is recommended to consider cup thickness in the analytical model interpretation for hard-on-hard hip implants. Overall, the relationships obtained for the input and output contact variables would aid in developing a potential analytical contact mechanics model suitable for hard hip implant combinations.