Simulation studies of wavelet neural networks for intelligent CNC turning

Abstract

Wavelet neural networks are investigated for learning a multi dimensional input-output complex nonlinear function. The objective is to predict accurately the error likely to be produced on parts during CNC machining and use it further to correct the process for improving the part accuracy. Exhaustive analysis is carried out by varying the number of wavelons in the wavelet neural network for determining the minimum number of wavelons. Artificial neural networks are also investigated for function learning and a comparison is drawn with the performance of wavelet neural networks. It is found that wavelet neural networks perform better as a function approximation tool as compared to artificial neural networks.