Enhanced PMDC motor modeling using a neural network multi-model approach

This paper presents a novel modeling methodology for Permanent Magnet DC (PMDC) motors using a multi-model framework based on Neural Network (NN) Multi-Layer Perceptron (MLP) architecture. Departing from conventional single-model paradigms, the proposed approach employs an ensemble of specialized MLP networks, whose optimal configuration is guided by adaptive performance evaluation. Comprehensive experimental and simulation results reveal that this multi-model framework achieves substantial enhancements in modeling precision, increased resilience to operational disturbances, and superior overall performance when contrasted with single-network implementations. Laboratory bench tests verify the method's efficacy across varied operational scenarios, including transient states. Owing to its adaptable and scalable nature, this methodology emerges as a valuable asset for sophisticated control strategies and real-time diagnostic applications in PMDC motor systems.