Wavelet-packet-transform-based identification of motor systems

Abstract

Accurate system modeling is the key to high-performance motor control. However, there usually exist nonlinear position-dependent cogging torque disturbances in motor systems, which makes classical linear system identification algorithms unavailable to such systems. Therefore, in this paper, a wavelet-packet-transform (WPT)-based preprocessing method is proposed to extract and attenuate the cogging torque disturbances of the identification data, so as to obtain an accurate linear model of the motor system by applying a classical identification algorithm without compensating the cogging torque in advance. Both the simulation and experiments show that the proposed method is a simple and efficient way to obtain a linear model for a motor with cogging torque disturbances.