Design of Efficient Adaptive Neuro-Fuzzy Controller Based on Supervisory Learning Capable for Speed and Torque Control of BLDC Motor

Brushless DC (BLDC) motors have been widely used in many field of drives for their high power/weight, high torque, high efficiency, long operating life, noiseless operation, high speed ranges and ease of control. In this paper, a Neuro-Fuzzy Controller (NFC) based on supervisory learning is presented for the speed and torque control of BLDC motors to enhance high control performance of the drive under transient and steady state conditions. This designed controller is combination of Neural Networks (NNs) and Fuzzy Logic (FL), therefore has parallel processing and learning abilities of NNs and inference capacity of FL. For improvement the performance of leaning algorithm and thereupon increase efficiency of drive, instead of usual Error Back Propagation (EBP) learning technique, a fuzzy based supervisory learning algorithm is employed. The proposed controller has simple structure and also due to its modest fuzzy rule in rule-base is relatively easy for implementation. This controller has high accuracy, suitable performance, high robustness and high tracking efficiency. In order to demonstrate the NFC ability to tracking reference speed and torque and also test of robustness and rejection ability of controller against undesired disturbances or suddenly changes in speed and torque, these designs are simulated with MATLAB/SIMULINK. In some cases, results are compared with that of a conventional PID controller and other designs.

---

W artykule zaprezentowane układ sterowania bezszczotkowym silnikiem DC z wykorzystaniem sterownika Neuro-Fuzzy. Dla poprawienia efektywności uczenia sieci zamiast wstecznej propagacji błędu zaproponowano algorytm wykorzystujący logikę rozmytą. Sterownik okazał się być dokładny, odporny i o dużej efektywności śledzenia zmian. Porównano możliwości kontrolera z konwencjonalnym sterownikiem PID.