Performance Enhancement of Switched Reluctance Motor Using Cascade Forward Neural Network and Bayesian Optimization

Switched Reluctance Motors (SRMs) have emerged as promising alternatives to permanent magnet machines for industrial and electric vehicle applications due to their simple structure, fault tolerance, and absence of rare-earth materials. However, their widespread adoption is limited by inherent torque ripple and nonlinear characteristics. This paper proposes an enhanced SRM control strategy based on a Cascade Forward Neural Network (CFNN) optimized using Bayesian regularization to achieve improved speed regulation and reduced torque ripple under dynamic operating conditions. The proposed controller exploits the learning capability of CFNN to model nonlinear motor behaviour, while Bayesian optimization improves generalization and robustness. The performance of the proposed method is evaluated in comparison with a conventional PI controller and a PSO-based Cuckoo Search Algorithm (CSA). To address practical applicability concerns, an implementation-oriented verification methodology is adopted, incorporating fixed-point realization and real-time feasibility analysis consistent with FPGA-based digital controllers. Detailed system specifications, robustness tests under load and parameter variations, and computational complexity analysis are presented. Simulation and real-time constrained results demonstrate that the proposed ANN-CFNN controller achieves superior torque ripple reduction, faster dynamic response, and improved robustness while maintaining real-time suitability. The findings confirm the potential of the proposed approach for energy-efficient and high-performance SRM drive applications.