A novel position sensorless driving system of brushless DC motors based on neural networks

In this paper, a position sensorless driving method of brushless DC motors (BLDCMs) using neural networks has been proposed. Considering the nonlinear characteristics of BLDCM and the parameter errors in the modeling, neural networks can be considered as a powerful tool. Thus, we introduce a neural network to estimate the electromotive force (EMF). Instead of directly estimating the position information from EMF, we propose a new method to estimating the position errors and then using approximate algorithm to obtain the rotor position. The results of simulation and experiment using offline trained neural networks show that the BLDCM is controlled well under load conditions. The proposed method can be believed have high possibility in practical applications.