Torque and speed prediction of a brushless direct current motor using nonlinear autoregressive with exogenous inputs and neural network

Brushless DC (BLDC) motors are widely used in industrial applications due to their high efficiency and performance. However, accurately predicting key parameters such as torque and speed remains a challenge because of the motor's inherently nonlinear dynamics. This study presents a data-driven modeling approach using a Nonlinear Autoregressive Neural Network with Exogenous Inputs (NARX-NN) to predict the torque and speed of a BLDC motor. Input-Output data were obtained from a Simulink-based BLDC motor model under varying input voltages and load conditions. The proposed NARX-NN architecture was trained on this data, effectively learning the nonlinear Multi-Input Multi-Output (MIMO) system dynamics. The model achieved high prediction accuracy, with a Mean Squared Error (MSE) of 3.4162e-04 (training), 3.0296e-04 (validation) and 8.4225e-04 (testing) while R-values of 1 in each in case of speed. While the model also achieved high prediction accuracy, with a Mean Squared Error (MSE) of 0.0062 (training and validation), and 0.0065 (testing) while R-values of 0.9997 (training and validation) and 0.9998 (testing) in case of torque. These statistical results are compared with the work already carried out for prediction of speed of BLDC motor, dominating the superiority of the proposed approach. Hence, it confirms the model's robustness in capturing complex motor behavior. The proposed approach offers a reliable predictive tool suitable for integration into real-time control systems, enabling enhanced motor efficiency, early fault detection, and torque ripple mitigation.