This content is not included in your SAE MOBILUS subscription, or you are not logged in.

Nonlinear Model Predictive Engine Airpath Control with Dual-Loop Exhaust Gas Recirculation and Variable Nozzle Turbocharger

Journal Article
03-16-05-0034
ISSN: 1946-3936, e-ISSN: 1946-3944
Published November 10, 2022 by SAE International in United States
Nonlinear Model Predictive Engine Airpath Control with Dual-Loop
                    Exhaust Gas Recirculation and Variable Nozzle Turbocharger
Sector:
Citation: Liu, Z., Dizqah, A., Herreros, J., Schaub, J. et al., "Nonlinear Model Predictive Engine Airpath Control with Dual-Loop Exhaust Gas Recirculation and Variable Nozzle Turbocharger," SAE Int. J. Engines 16(5):2023, https://doi.org/10.4271/03-16-05-0034.
Language: English

Abstract:

The control of the engine airpath is a constrained multi-objective tracking problem. Multiple control variables including Exhaust Gas Recirculation (EGR) and Variable Nozzle Turbocharger (VNT) valve positions are simultaneously adjusted to accommodate fast, slow, and coupled nonlinear airpath dynamics. This work proposes a Nonlinear Model Predictive Controller (NMPC) that exploits a convex and multi-rate prediction model for the real-time airpath control of a Compression Ignition engine equipped with dual-loop EGR and VNT. The benefits of the approach are verified using a simulation study against a EURO 6 production-line controller and Hardware-in-the-Loop (HiL) implementation using a 480 MHz processor that is comparable to nominal Engine Control Units. The NMPC demonstrates improved control performances including reduced tracking error for intake manifold pressure, oxygen concentration, and torque by 12.23%, 21.45%, and 26.99%, respectively, as well as a 0.98% fuel economy improvement compared to the production-line controller. These benefits hold even with simulated 5% and 10% sensor noises, under one set of objective weightings over the Worldwide harmonized Light vehicles Test Cycles (WLTC). The HiL implementation of the NMPC shows the average and maximum computational time of 1.80 ms and 2.94 ms, respectively, across the WLTC, which are below the required 10 ms control interval.