Neural Networks-Based 12-Sector Direct Torque Control of Asynchronous Machine Drives: Experimental Results

Abstract

Achieving optimal performance from industrial asynchronous motors requires precise control of both torque and magnetic flux. The conventional six-sector direct torque control (6-DTC) method is widely used for this purpose, but it suffers from several drawbacks, including significant torque and flux ripple and excessive switching activity in the power converter, which leads to energy losses. This research addresses these issues by introducing an enhanced control method that integrates neural networks (NNs) with a refined twelve-sector direct torque control (12-DTC) scheme. Unlike traditional approaches that depend on hysteresis comparators and fixed lookup tables, the proposed NN-based controller intelligently determines voltage vectors in real time, enabling smoother and more adaptive motor operation. Simulation studies verified the method’s effectiveness, showing substantial reductions in torque ripple, improved control tracking, and lower total harmonic distortion (THD). Experimental validation on a dSPACE DS1104 hardware platform confirmed these improvements in practice. The 12-sector NN-DTC achieved THD reductions of 30.87%, 42.22%, and 49.73% across different test conditions, along with 47% faster dynamic speed response. Overall, the proposed neural network–based 12-sector DTC represents a significant advancement over the traditional 6-DTC. It delivers smoother performance, more rapid response, higher efficiency, and robust real-world applicability—offering a more precise and energy-efficient solution for controlling industrial asynchronous machines.