Sensorless control of brushless DC motors at low speed using neural networks

A sensorless control method of brushless DC (BLDC) motors is presented in this paper. Differently from the previous ones, this method is based on the zero crossing of line back EMFs, which is proofed corresponding to the commutation points and more reliable at low speeds. For the accuracy and robust of the method, an artificial neural network with a section dealing with the outliers is also employed to estimate the line back EMFs by only using measurements of the stator line voltages and currents. The method provides more wide-speed range, and no neutral voltage, phase shifted or silent phase required. A simulation result of the proposed sensorless method is shown. Also, the performance of the proposed method is verified by real-time experiments using the digital signal processor (DSP) TMS320LF2407A.