Neural speed controller based on two state variables applied for a drive with elastic connection

Abstract

In this paper an adaptive speed control structure of electrical drive is proposed. Mechanical part of the system contains two machines connected with a long shaft. This construction can introduce additional disturbances appearing in transients of state variables. Tested controller is based on MLP neural network trained on-line. Two state variables are used as feedback signals: motor speed and shaft torque. Speed control error is minimized according to the backpropagation algorithm. Moreover output part of the neural network has additional input, where shaft torque is introduced. It means that at neural controller has the second feedback with adaptable coefficient. Obtained simulation results present precision of control and robustness against drive parameter changes. The proposed controller was implemented in digital signal processor of dSPACE1103 card and tested on laboratory benchmark.