Fractal PI-PI μ D with Artificial Neural Network Speed Controller for BLDC Motor Drive

Busher Victor, Valerii Tytiuk, Oleksii P. Chornyi, Oksana V. Glazeva, Valeriy Druzhinin, Lubov Melnikova

2025

Closed access Metadata (CSL-JSON)
DOI 10.1109/khpiweek61436.2025.11288713
Data confidence 85 high
How this score is calculated

It is built from checkable properties of the record itself: whether the article has a DOI, how many independent sources confirmed its data, whether the required fields are filled in, and how recent the latest confirmation is. Disagreement between sources and a missing identifier lower it. A value of 80 or above means the record was published without further checking; 60 to 79 means it is published at reduced confidence, with a manual check advisable. A record verified by an editor by hand is never pushed below 90 by a recalculation. This scores our data about the article, not the article itself.

Abstract

The research is devoted to synthesize a controller with a fractional-order integral component to achieve the order of astaticism of the closed-loop speed BLDCM higher than 1 and with a neural network to improve the static and dynamic control accuracy. The control object is identified by a fractional-order differential equation. Four types of controllers are synthesized: a classic PID to provide settings close to the modular optimum (IntPID); a controller with a fractional-integral component that provides the order of astaticism of the loop 1.8 (FrPID), and – what is the scientific novelty of the work – hybrid controllers based on the first two, supplemented by neural networks of direct signal propagation NN–IntPID, NN–FrPID. FrPID compared to IntPID allowed to reduce the mean square error by almost 5 times, and the neural controller with a fractional integral component provided the best results: overshoot with a jump in the reference signal was reduced from 2.98% to 0.35%, the mean square error with a linearly changing reference signal was reduced by 11 times and the relative speed error decreases from 0.00801 to 0.00022, more than 35 times. The proposed algorithm for fast learning in each cycle of the control program also ensures the adaptation of the controller when the parameters of the control object change, what makes the possibility of using neural controllers with a fractional integral component for other objects, such as universal collector motors, electromagnetic brakes with adjustable torque.