Control methodologies for fast and low impact electromagnetic actuators for engine valves.

Abstract

In an effort to improve the performance of the standard internal combustion (IC) engine, electromagnetic valve actuators (EVAs) have been proposed as a solution to achieve variable valve timing (VVT). By replacing the camshaft commonly found in most automotive engines, EVAs decouple the motion of the engine valves from the crankshaft. In doing so, they allow for VVT which research has shown can significantly improve torque, fuel economy, and emissions. Unfortunately, EVAs suffer from excessively loud impacts between their moving components that prevent them from begin commercially viable. Precise control of the EVA motion is hindered by non-negligible electrical dynamics and nonlinearities in the magnetic subsystems. To ensure fast transition times, stiff springs are required to increase the bandwidth of the mechanical components of the system. To counteract the stiff springs, numerous turns are required in the magnetic coils resulting in a large inductance and thus a relatively low bandwidth electrical subsystem. Therefore the common simplifying assumption of current control is no longer valid. The non-negligible electrical dynamics introduce further complications due to nonlinearities in the magnetic force and gap reluctance. At large gaps the magnetic force is very weak and at small gaps back-EMF effects become significant resulting in poor control authority. This thesis introduces novel nonlinear control techniques designed to minimize the impacts associated with the operation of EVAs. Through a combination of Lyapunov based control and extremum seeking control the impacts are reduced from approximately 1 m/s to 0.1 m/s. Based on bandwidth considerations and non-minimum phase-zeros, control of the EVA at small air gaps using magnetic flux and position is shown to be more advantageous than using current and position. To stabilize the system and account for the nonlinearities present in the dynamics, the universal stabilizing feedback proposed by E. D. Sontag is used. To enhance the performance of the system and facilitate the controller gain tuning, the Lyapunov function used to construct Sontag's feedback is determined based on the solution to an algebraic Riccati equation. A discrete extremum seeking controller is then used to exploit the repetitive nature of the system for self-tuning. As the valves open/close several thousand times per minute, the extremum seeking control uses information from previous valve events to select a Lyapunov function on-line to minimize the impacts from one valve event to the next.