Composite Adaptive Internal Model Control: Theory and Applications to Engine Control

Abstract

To meet customer demands for vehicle performance and to satisfy increasingly stringent emission standard, powertrain control strategies have become more complex and sophisticated. As a result, controller development and calibration have presented a time-consuming and costly challenge to the automotive industry. This thesis aims to develop new control methodologies with reduced calibration effort. Internal model control (IMC) lends itself to automotive applications for its intuitive control structure with simple tuning philosophy. A few applications of IMC to the boost-pressure control problem have been reported, however, none offered an implementable and easy-to-calibrate solution. Motivated by the need to develop robust and easily calibratable control technologies for boost-pressure control of turbocharged gasoline engines, this thesis developed new control design methodologies in the IMC framework. Two directions are pursued: adaptive IMC (AIMC) and nonlinear IMC.

A plant model and a plant inverse are explicit components of IMC. In the presence of plant-model uncertainty, combining the IMC structure with parameter identification through the certainty equivalence principle leads to adaptive IMC (AIMC), where the plant model is identified and the plant inverse is derived by inverting the model. We propose the composite AIMC (CAIMC), which identifies the model and the inverse in parallel, and reduces the tracking error through the online identification. ``Composite" refers to the simultaneous identifications. The constraint imposed by the stability of an n-th order model is nonconvex, and it is re-parameterized as a linear matrix inequality. The parameter identification problem with the stability constraint is reformulated as a convex programming problem. Stability proof and asymptotic performance are established for CAIMC of a general n-th order plant.

CAIMC is applied to the boost-pressure control problem of a turbocharged gasoline engine. It is first validated on a physics-based high-order and nonlinear proprietary turbocharged gasoline engine Simulink model, and then validated on a turbocharged 2L four-cylinder gasoline engine on a Ford Explorer EcoBoost. Both simulations and experiments show that CAIMC is not only effective, but also drastically reduces the calibration effort compared to the traditional PI controller with feedforward.

Nonlinear IMC is presented in the context of the boost-pressure control of a turbocharged gasoline engine. To leverage the available tools for linear IMC design, the quasi-linear parameter varying (quasi-LPV) models are explored. A new approach for nonlinear inversion, referred to as the structured quasi-LPV model inverse, is developed and validated. A fourth-order nonlinear model which sufficiently describes the dynamic behavior of the turbocharged engine is used as the design model, and the IMC controller is derived based on the structured quasi-LPV model inverse. The nonlinear IMC is applicable when the nonlinear system has a special structural property and has not been generalized yet. Simulations on a high-fidelity turbocharged engine model are carried out to show the feasibility of the proposed nonlinear IMC.