Modeling the wear of Polyamide 66 using artificial neural network

Recently, Artificial Neural Networks (ANNs) have emerged as a good candidates to mathematical wear models, due to their capabilities of nonlinear behavior, learning from experimental data and generalization. In the present paper the potential of using neural networks for the prediction of sliding wear properties of polymer was investigated using a measured dataset of 42 independent reciprocating sliding wear tests of Polyamide 66. The polymer was tested under various testing conditions such as applied load (constant and fluctuating), existence of surface crack and sliding media. Five different feed-forward (ff) neural network models were examined in order to find the optimum model that simulates the wear under such parameters. The optimized ANN was utilized to predict the wear rates of new input parameters, which were not tested. The quality of prediction was good when comparing the predicted and real test values. Finally, the proposed ANN was applied to four data sets adopted from previous works to evaluate its accuracy. The ANN showed good accuracy for the simulation and acceptable values of predicted wear rates. The results indicated that the well-trained neural network model is quite effective for prediction of wear response of materials within and beyond the experimental domain.