Neural network adaptive control for machining process based on generalised entropy square error and wavelet analysis

To improve convergent speed of the neural network for machining process control, Generalised Entropy Square Error (GESE) function is defined and its availability is proved theoretically. Combining information entropy and wavelet analysis with neural network, a neural network adaptive control system is presented. Replacing the mean square error criterion of back propagation algorithm with the GESE criterion, an adaptive control algorithm is proposed. The proposed system is then applied to the online control of the cutting force by searching adaptively wavelet base function and self-adjusting scale parameter, translation parameter of the wavelet and weights of the network. The designed system is of fast response and less overshoot, and the suggested algorithm can tune adaptively the feed rate online till achieving a constant cutting force approaching the reference force in varied cutting conditions, thus raising machining efficiency and protecting tool. Finally, simulation and experiment examples are also given to demonstrate the effectiveness of the proposed system and algorithm.