Tension identification of two-motor system based on neural network left-inverse

Abstract

Tension detection is a key to improve performance of two-motor system under sensorless operation. This paper presents a new identification method for two-motor system based on artificial neural network and the left-inverse theory. Considering that the system parameters are time-variant and the mathematic model of left-inverse identification is complex, BP neural network is used to build the left-inverse model in this method, which is easy to implement. A simulation model of a two-motor system is developed. The simulated results verify the proposed method. By using this control strategy, the tension can be identified quickly and accurately, in which satisfactory robustness is offered.