Iterative learning based neural network sliding mode control for repetitive tasks: With application to a PMLSM with uncertainties and external disturbances. (1st June 2022)

• Record Type: Journal Article
• Authors: Liu, Weike; Shu, Feng; Xu, Yunlang; Ding, Runze; Yang, Xiaofeng; Li, Zhi; Liu, Yang
• Is Part Of: Mechanical systems and signal processing. Volume 172(2022)
• Publication Date: 2022-06-01
• Subjects: Precision motion control -- Neural network -- Sliding mode control -- Repetitive task -- Iterative learning -- PMLSM; Structural dynamics -- Periodicals; Vibration -- Periodicals; Constructions -- Dynamique -- Périodiques; Vibration -- Périodiques; Structural dynamics; Vibration; Periodicals; 621
• DOI: 10.1016/j.ymssp.2022.108950
• Languages: English
• ISSNs: 0888-3270

Abstract

Abstract: The uncertainties, force ripple, friction and other external disturbances in permanent magnet linear synchronous motors (PMLSMs) pose great challenges to the design of advanced controllers with high speed and high precision. Existing neural network based sliding mode control (NNSMC) methods typically adopt neural network as an online compensator for uncertainties and disturbances, meanwhile stability and robustness are assured owing to sliding mode control. However, PMLSMs usually need to perform repetitive motion tasks within a finite time interval in practical applications, where existing NNSMC methods ignore the previous repetitions' information, and consequently limit the motion performance improvement for trajectory tracking tasks. This paper proposes an iterative learning based neural network sliding mode control (ILNNSMC) scheme, in which the concept of iterative learning is incorporated into NNSMC methods for the first time. In ILNNSMC, the uncertainties and disturbances can be compensated by NN (updating in the time-domain) and ILC (learning in the iteration-domain) in a complementary way, and the sliding variable is driven to converge to zero for arbitrary time within the predesigned time interval with successive iterations. The motion performance for repetitive tasks can be improved substantively. Moreover, a non-linear dynamic learning gain (DLG) strategy is proposed and adopted in iterative learning law, which guarantees both convergence speed and steady-state accuracy in the learning process. Stability of the scheme is discussed in both time and iteration domains, and the control performance is verified by extensive experiments on a PMLSM. Highlights: Iterative learning is incorporated into neural network sliding mode control. Cross-domain compensation is realized by neural network and iterative learning. Dynamic learning gain trades off convergence speed and steady-state accuracy. Stability is proved based on Lyapunov direct method in time and iteration domains. Experimental results on a PMLSM verify the validity for repetitive tasks.