Position Sensorless Control For BLDCM Based On Neural Network

Brushless DC Motor (BLDCM) is a kind of mechatronics products, which not only reserves many excellent mechanical characteristics of DC Motor, but also has many good performances, such as high efficiency, high torque, small volume and inertia, long life-span and sparkless commutation. However, because of some disadvantage of the position sensor itself, the application of BLDCM is limited in some special occasions. Therefore, more and more attention is paid to position sensorless control technology of BLDCM.Firstly, this paper introduces the composition structure of brushless DC motor, then analyzes the working principles of brushless DC motor and establishes the mathematical model of brushless DC motor. Aiming at the features of multivariate, non-linear and time varying brushless DC motor servo system, this paper combines with the advantage of neural network which is function approach, and proposes using BPNN as control strategy in position sensorless control system.Secondly, position sensorless control for BLDCM can be regarded as a nonlinear system identification process. The relationship between between rotor’s position and phase voltages and current of BLDCM is analyzed. Because of those relationship, this paper proposes the system identification method base on BPNN. In order to solve the problem of slow partial convergence speed and local minimum of traditional BP algorithm, solution of calculate by optimization principle of grade decline with momentum method is deducted. The solution turns out to improve the performance of the network.At last, the method of control system based on BP for BLDCM without position sensor is confirmed by simulation. A governing system is built based on position sensorless control. The experimental results show that this method can provide an accurate commutating signals and realize position sensorless control for BLDCM. It is proved that the approach has good performance in control accuracy, adaptability and robustness. At the same time, this paper also complete the experiment based on DSP control system.