Assessment of the Performance of Neural Networks Models for the Prediction of Surface Roughness After Grinding of Steels

Abstract

Attaining acceptable levels of surface roughness is one of the primary goals of grinding. In order to achieve this goal, several approaches can be employed by varying the system parameters and attempting to find the conditions which ameliorate the outcome of the process. One way that can be rather assistive to the analysis of manufacturing processes is the use of soft computing methods such as artificial neural networks. In the current work, artificial neural network models of various categories, namely multi-layer perceptron and radial basis function neural networks, with various model parameters are developed and compared in order to determine the best performing model for the prediction of surface roughness during peripheral grinding of steel components. The results indicate that radial basis function networks outperform classical multi-layer perceptrons and constitute a promising alternative for the modeling of manufacturing processes.