Integrated Engine States Estimation Using Extended Kalman Filter and Disturbance Observer 2019-01-2603

Accurate estimation of engine state(s) is vital for engine control systems to achieve their designated objectives. The fusion of sensors can significantly improve the estimation results in terms of accuracy and precision. This paper investigates using an Extended Kalman Filter (EKF) to estimate engine state(s) for Spark Ignited (SI) engines with the external EGR system. The EKF combines air path sensors with cylinder pressure feedback through a control-oriented engine cycle domain model. The model integrates air path dynamics, torque generation, exhaust gas temperature, and residual gas mass. The EKF generates a cycle-based estimation of engine state(s) for model-based control algorithms, which is not the focus of this paper. The sensor and noise dynamics are analyzed and integrated into the EKF formulation. To account for ‘non-white’ disturbances including modeling errors and sensor/actuator offset, the EKF engine state(s) observer is augmented with disturbance state(s) estimation. Case studies demonstrate that the disturbance augmented EKF can identify the sources of estimation errors and mitigates these errors automatically within several engine cycles. This paper concludes that the number of disturbance states is required to be the same as the number of sensors to ensure bias-free estimation and observability of the augmented system. Different disturbance state(s) selections are investigated in the case studies. Experimental validation of the proposed observer shows that it estimates unmeasurable engine states, suppresses sensor noise and detects disturbance sources. Successful implementation to rapid prototype engine controller also indicates high computational efficiency of the algorithm.