Decoupled disturbance rejection control for a turbocharged engine with a dual-loop exhaust gas recirculation system. (April 2018)

• Record Type: Journal Article
• Authors: Chen, Song; Yan, Fengjun
• Is Part Of: Proceedings of the Institution of Mechanical Engineers. Volume 232:Number 5(2018:May)
• Page Start: 599
• Page End: 615
• Publication Date: 2018-04
• Subjects: Dual-loop exhaust gas recirculation -- error analysis -- sliding-mode observer -- disturbance rejection control; Mechanical engineering -- Congresses; Transportation engineering -- Congresses; 629.2
• DOI: 10.1177/0954407017704780
• Languages: English
• ISSNs: 0954-4070

Abstract

Dual-loop exhaust gas recirculation with a variable-geometry turbocharger is an effective architecture for achieving desired intake manifold conditions, such as the temperature, the pressure and the oxygen concentration of the intake manifold, which have critical roles in advanced combustion mode control. However, the widely used control-oriented model is derived on the basis that the heat transfer between the pipes and the gas is negligible, which means that it suffers from non-trivial errors. Simulation results show that other error sources, including the volumetric efficiency and the orifice equation, are difficult to calibrate accurately and also cause significant errors in the system, particularly in transient situations. Modified active disturbance rejection control with an extended state observer is utilized to deal with the non-linear, multiple-input multiple-output system in this paper. It is demonstrated that the performance of active disturbance rejection control mainly depends on the performance of the extended state observer. In this paper, an extended state observer, which is based on the sliding-mode concept rather than the conventional linear observer, is introduced. By taking advantage of its strong robustness, the system is decoupled into three loops. For each loop, the internal errors and the external errors, including the modelling error and the coupling effects, are lumped into one term; they are then actively estimated and cancelled out by the control input in real time. The proposed method was validated using calibrated GT-Power model simulations.