Date of Award
Doctor of Philosophy (PhD)
Internal combustion engines require accurate cylinder charge estimation for determining engine torque, controlling air-to-fuel ratio (AFR), and ensuring high after-treatment efficiency. This is challenging due to the highly transient operating conditions that are common in automobile engines. The problem is further complicated by spark ignition (SI) engine technologies such as variable valve timing (VVT) and exhaust gas recirculation (EGR) which are applied to improve fuel economy and reduce pollutant emissions. With manifold filling/emptying/mixing phenomenon and different actuator response times, these technologies significantly increase the complexity of cylinder charge estimation.
Current cylinder charge estimation methodologies require a combination of sensors and empirical models to deal with the high degrees of control freedom existent on the engine. But these methods have the drawbacks of great dependency on accurate calibration and poor transient performance. Most importantly, the current methods isolate feed-forward cylinder charge estimation and feedback AFR control. When there is discrepancy between target lambda value and sensed lambda value at exhaust side, the current control/estimation method will trim the fuel injection amount no matter where the error source is. As a matter of fact, the error might come from the throttle flow estimation, the fuel injection flow estimation, EGR flow estimation, or any combination of these error sources. Increased air-path complexity and drawbacks of traditional methods drive the need for cost effective solutions that produce high air/EGR/fuel charge estimation accuracy with the ability to identify the error source while minimizing sensor cost, computational effort, and calibration time.
This research first evaluates the existing work on air charge estimation for SI engines with massive experimental tests covering various operating conditions, which are designed for the algorithm verification of this research. Then several estimation methods which utilize both Manifold Absolute Pressure (MAP) and Mass Air Flow (MAF) sensors are studied and analyzed. Reduction of calibration effort and improvement of accuracy are observed from the proposed cylinder air charge estimation methods. Following that, a model is built to study the engine gas path dynamics and characteristics and then simplified to provide system dynamic basis for the following estimation algorithm development. Using the developed model, a disturbance observer based cylinder charge estimation technique is developed based on a combination of sensors including MAF, MAP, and exhaust lambda sensors. This developed algorithm significantly improves engine states estimation accuracy compared to conventional Single-Input-Single-Output (SISO) methods. Also, the augmentation of disturbance observation is able to pin point the source of the estimation error. Through experimental validation, using the developed estimation method with proper parameters, the error source of estimation can be identified and rectified when disturbance is introduced to throttle flow model, EGR flow model, fuel injection flow model or any combination of these models. The structure of the proposed algorithm should adapt to most SI engine configurations. It can help the engine controller to mitigate modeling errors thus improve the performance of physics model based engine control especially AFR control.
Wang, Zhe, "Observer Based Cylinder Charge Estimation for Spark-ignition Engines" (2017). All Dissertations. 2323.