The compensation of speed ripple caused by the angle-measuring error of inductosyn using neural network

This paper tackles the problem of the compensation of speed ripple of brushless DC motors by means of neural networks. The speed ripple is caused by the error of inductosyn angle-measurement. This error is equal to add a disturbance to the input of the speed control system. A parallel three-layer feedforward artificial neural network (FANN) has been proposed to compensate this error in order to improve the speed performance of the DC motor. The compensation is based on the separating technique of the angle-measuring error of the inductosyn. An online learning algorithm for the FANN has also been proposed. Practical results illustrate that the speed performance of the motor is greatly improved by the proposed compensation scheme.