Integrating deep learning modeling with implicit Lyapunov function-based super-twisting sliding mode control for robust performance of a chain conveyor

Abstract

This paper presents a novel control strategy, an implicit Lyapunov function-based super-twisting sliding mode controller (ILSSMC) with deep model compensation, to address the challenges of low positioning accuracy and slow positioning speed in a chain conveyor. Initially, a collaborative model for the chain conveyor is developed by integrating data-driven techniques and mechanism cognition. This model leverages deep neural networks to capture time-varying parameters and unmodeled dynamics, thereby mitigating the effects of system uncertainties. Subsequently, a novel adaptive super-twisting sliding mode controller grounded in the implicit Lyapunov function method is proposed to further suppress residual uncertainties. In this approach, control gains are adjusted online based on the Lyapunov function to ensure robust performance. Notably, the proposed ILSSMC requires fewer control parameters, which enhances its practicality in real-world applications. The experimental results demonstrate that the ILSSMC with deep model compensation reduces the positioning time by 25% (from 2.114 to 1.590 s) while maintaining high positioning precision across various conditions, thus validating the effectiveness of the proposed control strategy.