Axial flux machine performance enhancement using recurrent neural network controller

Abstract

Traditional control methods often face limitations in optimizing the performance of these motors, especially in complex industrial and automotive applications where precision, stability, and energy efficiency are paramount. By exploring advanced control strategies such as multi-level inverters and neural network controllers, this study aims to overcome these limitations and unlock the full potential of dual rotor axial flux induction motors. The integration of multi-level inverters enables finer control of motor operation and enhances power quality, while neural network controllers offer adaptive and intelligent control capabilities, enabling the system to learn and optimize performance in real-time. The study investigates novel approaches to enhance the performance and efficiency of electric motor control systems. The study aims to address the challenges associated with traditional control methods and optimize the operation of dual rotor axial flux induction motors. The research evaluates various performance metrics associated with the speed control system, including error histograms, training performance, regression accuracy, rotor speed dynamics, rotor torque characteristics, time series analysis, and training state assessment. The study achieves significant milestones in optimizing system performance, as evidenced by key findings such as a low mean squared error (MSE) of 0.00011396 achieved during training, strong correlation in regression analysis with an R-value of 0.99718, and effective training dynamics indicated by a gradient value of 0.0091742 and a learning rate (Mu) of 0.0001. These results underscore the effectiveness and reliability of the proposed control strategies in improving motor performance, efficiency, and reliability while reducing energy consumption and operational costs. The proposed method is implemented using MATLAB.