Optimal Calibration and Transient Control of High Degree of Freedom Internal Combustion Engines.

Increasing engine system complexity for achieving better engine performance and fuel economy induces intricate engine calibration and transient engine control problems. The classical experiment based procedure cannot deal with the exponential increase in size of the calibration problem for the high degree-of-freedom (DOF) engine. The increased number of independent variables leads to complex inter-relationships, and characterizing them by means of traditional experimental sweeps of individual variables is simply not possible. In addition, increased number of actuators creates a new challenge under rapid engine transients. Various devices might have different response times, thus leading to significant excursions of operating parameters during dynamic changes of load and speed. The higher the DOF in the system, the more probability that the engine may deviate from optimum during a transient. Since transients are very frequent during normal driving, the sub-optimum engine behavior during these events can cause significant performance and emission penalties. Thus, developing transient control methodologies is an indispensable complement to optimal steady-state calibration if we aim to realize the full potential of the modern engine with variable devices and sub-systems. This dissertation covers the entire procedures for achieving the optimal feed-forward steady-state control strategy and transient control of a high degree-of-freedom engine based on performance, combustion stability and emissions goals. Contributions critical for achieving the overall objective are: (1)Improved high-fidelity simulation tools as alternative to experiments; (2)Virtual sensing methodologies using artificial neural networks (ANNs); (3)Characterization of the combustion stability for the real time estimation; (4)Simulation based optimization framework for determining optimal actuator set-points in a high DOF engine considering a multi-objective cost function; (5)Nonlinear model predictive control (NMPC) of engine transients. The NMPC development is enabled by using a proposed control oriented model (COM) and applying a receding horizon concept.PhDMechanical EngineeringUniversity of Michigan, Horace H. Rackham School of Graduate Studieshttp://deepblue.lib.umich.edu/bitstream/2027.42/62420/1/kteel_1.pd