Controlling the fuel injection system of an internal combustion engine is a challenging and multifaceted problem. Current control algorithms rely heavily on lengthy experimentally-based calibration techniques. Even with these extensive calibration processes, suboptimal performance is often achieved because the selected control gains depend on calibrator experience and intuition instead of objective metrics. The cost and manpower required to calibrate a fueling controller can be staggering. Recent advances in air path actuation technologies such as the variable geometry turbocharger for diesel engines and variable cam timing for gasoline engines have expanded the dimensionality of the fueling control problem, further increasing the calibration burden of conventional controllers.
This dissertation presents model based alternatives to the current calibration-heavy fueling controllers used in production gasoline and diesel engines. The novelty of these controllers is derived from both the underlying plant models and the application of estimation/control theory to these models. One of the most unique aspects of these control problems is the time varying delay which characterizes the plant (i.e. the engine air path system). From a deep understanding of the physical phenomena involved, new types of air path models that directly capture the oxygen transport and mixing dynamics are developed. An experimental validation demonstrates that such a model can even account for cylinder-to-cylinder response variations caused by asymmetrical exhaust runner lengths.
Advanced model based estimation and control techniques are applied to these models to develop more effective methods of estimating the most important dynamic variables (i.e. in-cylinder oxygen concentration and cylinder imbalance) and controlling fuel delivery in diesel and gasoline engines. During the design processes, particular emphasis is placed on stability robustness, and comprehensive stability proofs are provided for the controllers and estimators developed. Because of the practical issues involved in dynamically measuring the in-cylinder oxygen concentration, the in-cylinder oxygen concentration estimator and the diesel fueling controller are validated in simulation using the engine modeling software GT-Power. Experimental validations demonstrate the capabilities of the cylinder imbalance estimator and comprehensively quantify the performance of the gasoline fueling controller through direct comparisons to the existing production controller. Above even their performance and robustness benefits, the most compelling reasons to adopt these estimators and controllers over their production counterparts are their modest calibration requirements.