A novel auto-tuning PID control mechanism for nonlinear systems. (September 2015)

• Record Type: Journal Article
• Authors: Cetin, Meric; Iplikci, Serdar
• Is Part Of: ISA transactions. Volume 58(2015:Sep.)
• Page Start: 292
• Page End: 308
• Publication Date: 2015-09
• Subjects: Model-based predictive control -- Auto-tuning -- PID controller -- MIMO PID controller design -- Real-time control; Engineering instruments -- Periodicals; Engineering instruments; Periodicals; Electronic journals; 629.805
• DOI: 10.1016/j.isatra.2015.05.017
• Languages: English
• ISSNs: 0019-0578

Abstract

Abstract: In this paper, a novel Runge–Kutta (RK) discretization-based model-predictive auto-tuning proportional-integral-derivative controller (RK-PID) is introduced for the control of continuous-time nonlinear systems. The parameters of the PID controller are tuned using RK model of the system through prediction error-square minimization where the predicted information of tracking error provides an enhanced tuning of the parameters. Based on the model-predictive control (MPC) approach, the proposed mechanism provides necessary PID parameter adaptations while generating additive correction terms to assist the initially inadequate PID controller. Efficiency of the proposed mechanism has been tested on two experimental real-time systems: an unstable single-input single-output (SISO) nonlinear magnetic-levitation system and a nonlinear multi-input multi-output (MIMO) liquid-level system. RK-PID has been compared to standard PID, standard nonlinear MPC (NMPC), RK-MPC and conventional sliding-mode control (SMC) methods in terms of control performance, robustness, computational complexity and design issue. The proposed mechanism exhibits acceptable tuning and control performance with very small steady-state tracking errors, and provides very short settling time for parameter convergence. Abstract : Highlights: A novel adaptive PID mechanism is proposed for nonlinear continuous-time systems. It provides small steady-state errors and short settling times for convergence. It can adapt to the changes in the reference inputs. It can compensate the interactions between the state variables of the processes. It can estimate varying parameters on-line and thus tolerates their effects.