Data Driven Learning Based Robust Control for Nonlinear Electromechanical Systems: A Case Study on SRV02

Abstract

In this paper, a data-driven neural network-based iterative learning sliding mode control (NN-ILSMC) scheme is proposed for precise position and velocity control of the SRV02 motor-driven system. Controlling electromechanical systems like the SRV02, which exhibit second-order nonlinear behavior under significant parametric uncertainties, remains challenging in industrial applications, particularly when model inaccuracies degrade control performance. Traditional model-based ILSMC requires accurate system identification, which limits its effectiveness under high levels of uncertainty. This paper develops and validates a novel NN-ILSMC strategy that leverages neural network approximation to enhance learning from tracking errors, eliminating the need for explicit mathematical models. The proposed control approach is tested against model-based ILSMC using a Quanser SRV02 DC motor system subject to 30% parametric uncertainty, Coulomb friction, and time-varying external disturbances. Monte Carlo simulations with 20 trials and 50 learning iterations are conducted to assess performance using position and velocity RMSE metrics. Results demonstrate that the proposed NN-ILSMC achieves an 86.3% reduction in position RMSE and an 87.6% reduction in velocity RMSE compared to the model-based ILSMC. The NN-based data-driven approach provides superior accuracy, robustness, and practical scalability, making it highly suitable for real-time industrial automation systems where dynamic uncertainties and modeling challenges are prevalent.