Robust Model-Predictive Control for Inductively Coupled Plasma Generation With a Semiphysical Simulation

• Published in: IEEE transactions on industrial electronics (1982) Vol. 68; no. 4; pp. 3380 - 3389
• Main Authors: Wang, Junjie, Liu, Yanming, Liu, Xiaotao, Li, Xiaoping, Zhang, Haojie, Wang, Di
• Format: Journal Article
• Language: English
• Published: New York IEEE 01.04.2021; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Algorithms; Argon; Controllers; Field programmable gate arrays; Generators; Identification methods; Inductively coupled plasma; Inductively coupled plasma (ICP); Iterative closest point algorithm; Mathematical model; model uncertainty; model-predictive control (MPC); Plasmas; Predictive control; Process control; Robust control; robust tuning; Robustness; semiphysical simulation; Simulation; System identification; Tuning
• ISSN: 0278-0046; 1557-9948
• DOI: 10.1109/TIE.2020.2978720

In this article, a novel robust tuning approach based on model-predictive control (MPC) for an inductively coupled plasma (ICP) generation process is presented. Several key aspects related to this methodology are the modeling of the ICP generation process, the robust joint tuning (RJT) algorithm, and the semiphysical simulation of the MPC algorithm. The ICP generation process model is built by a system identification method based on steady and dynamic responses, followed by a general model-predictive controller. Then, an automatic RJT algorithm is developed for the model-predictive controller to quickly obtain the optimal weighting parameters. The proposed RJT algorithm can minimize the ICP generation control system's maximum total variations in the output with a specific dynamic performance requirement and model uncertainty, providing a tradeoff between the dynamic performance and robustness. Finally, a semiphysical simulation system based on dSPACE and a field-programmable gate array is designed to illustrate the effectiveness of the model-predictive controller for the ICP generation control process and the RJT algorithm. Experimental results indicate the validity and good performance of the proposed model-predictive controller and the RJT algorithm.