Modeling and control of homogeneous charge compression ignition engines with high dilution

Homogeneous charge compression ignition (HCCI) engines provide the benefits such as low NOx emission and high fuel efficiency. Control of the HCCI engine, however, is difficult since its ignition cannot be directly actuated. Controlled autoignition requires regulation of the charge properties, especially charge temperature, as observed by many experimental results and substantiated in this thesis. To facilitate the control analysis and development, this thesis first introduce a physics-based cycle-to-cycle model of' a gasoline HCCI engine with an internal feedback loop constituted by the recirculation of exhaust gas. One important assumption in our modeling work that propagates down to the multiplicity analysis and controller synthesis is that the charge composition does not affect combustion (autoignition timing, burn duration, combustion efficiency, etc). This assumption is validated by the dominance of thermal dynamics revealed by sensitivity analysis and balanced realization of plant linearizations. Good agreement with experimental data suggests that the charge composition effects probably are of secondary importance, at least, for the regulation of the crank angle of 50% fuel burned (theta CA50). Based on the balanced truncation of the linearized model, a decentralized feedback and cancellation feedforward controller is designed. This linear controller has minimal complexity and care be tuned using classical PID design rules or real-time identification. The impact of the: nonlinear temperature dynamics in the internal feedback loop is then investigated by using the returning maps. The returning maps reveal the existence of stable and unstable equilibria of the cycle-to-cycle thermal dynamics. In the end, a nonlinear observer-based feedback controller is developed to stabilize the cycle-to-cycle temperature dynamics and regulate the combustion timing during large load transitions within the HCCI operating range. Simulations and estimates of the region of attraction show that the designed controller is robust to uncertainties such as the manifold filling dynamics, exhaust runner heat transfer, the cycle-to-cycle variation of theta CA50 and the uncertainty in the nonlinearity of the temperature dynamics.