Prediction of wear in total knee replacement implants using artificial neural network

The current research work presents development of an artificial neural network (ANN)-based model for predicting linear wear depth using wearing parameters, non-dimensional contact-stresses, sliding distance, and cross-shear ratio in total-knee-replacement. The linear wear depths are computed from knee wear models available in literature. The values of linear wear depth from these models were used for training and testing of an ANN-based model. Multi-layered feed-forward neural-network is used for training and testing of the ANN model. Many architectures of neural-networks were tried and the 3-6-6-6-1 architecture was found optimum. The sigmoid activation function was chosen for input and hidden layers, the linear activation function was chosen for the output layer, Admax was used as optimiser function. The ANN model predicts the linear wear depth within reasonable accuracy. Therefore, the ANN modelling can be an alternative to total-knee-replacements implant testing over in-vitro studies relied on knee simulators to save substantial time and cost.