Study on state and parameter estimation of three-way catalyst based on nonlinear kalman filter

Abstract

Air pollution became an issue due to the rapid industrialization and the development of car industry. Emissions form automobile cause destroying ozone layer, global warming, photochemical smog to environment and cause cancer, declined nerve center, respiratory diseases to humankind. From 1960s, society has concerned about the environment problem, and since California Air Resources Board(CARB-1967) and Environmental Protection Agency(EPA-1970) were established, they tighten regulation of exhaust gases from automobile. Since 1980s, three-way catalyst has been used to purify the car exhaust fumes. Characteristics of three-way catalyst are oxidize CO and HC when the air fuel ratio is rich, deoxidize NOx when the air fuel ratio is lean. When air fuel ratio is nearly to stoichiometric value, the efficiency of catalyst is almost 100 percent. It is desirable to adjust air fuel ratio nearly stoichiometric value to reduce harmful gases. As TWC have the different characteristics and parameter changed by reason of durability, temperature and environment, the model has to corrected accurately. Gradient descent method is widely used to estimate parameters, but in this thesis we used nonlinear Kalman filter such as the extended Kalman filter(EKF) and the unscented Kalman filter(UKF) to estimate parameters and oxygen storage state. While EKF uses first order of Taylor expansion to approximate nonlinear system, UKF performs a stochastic linearization by using a weighted statistical linear regression process based on the unscented transform. The advantage of Kalman filter is it can estimate well instead of noise in system or sensor measurement. Bias compensator using polynomial regression function is also proposed to reduce bias error that occurs due to nonequilibrium effect. The purpose of this thesis is to find the best estimation method by applying several nonlinear Kalman filters with bias compensator. Finally, innovation based adaptive Kalman filter is proposed for ...