Pre-filter design for high speed contouring machines

Abstract

High-speed and high-precision tracking control is required for future contouring machine tools. It is known that, by pre-filtering the command trajectory in an appropriate way, the contouring error can be effectively reduced. This paper presents the result of a hybrid approach applied to pre-filter design, which incorporates a MLP (multilayer perceptron) neural network to enhance the performance of a recently-proposed adaptive calibrating controller. Due to the presence of nonlinearities, such as friction and signal quantization, the proposed method further reduces the contouring error by taking advantage of the nonlinear and learning nature of the neural network. Simulation results on a dynamic model of a commercial laser profiling machine are demonstrated.