Abstract: An artificial neural network (ANN) method for high shape accuracy control is proposed. A multi layered neural network is trained off - line to identify the inverse model of the machining system. The inverse model is employed to tune the system. As a result, the output is obtained as close as possible to the desired input and machining error is minimized. It is shown, by experiments of turning the piston with middle-convex shape and variable ellipticity, that this control scheme can improve machining accuracy effectively.

Key words: Complex profile, Control, Inverse model, Neural network, Turning