Individual Cylinder Air-Fuel Ratio Estimation Based On Input Observer

Air-Fuel Ratio (AFR) is an important parameter in the engine control, and it affects the emissions, fuel economy and power performance of a car. Car manufacturers and consumers pursue lower emission, lower fuel consumption and higher power performance because of the increasingly strict emission regulations and rising fuel prices. Precise control of AFR can achieve the objects that are reducing vehicle emissions, improving fuel efficiency and enhancing the overall performance engine.Most AFR control methods take each cylinder as a system using the same struc-ture, without considering the difference among cylinders. Moreover, many control meth-ods focus on researches before the intake valve, which will influence the control accuracy to a certain extent. To get more accurate results, AFR needs to be controlled cylinder by cylinder, that is, individual cylinder AFR control. However, due to the special na-ture of high temperature and pressure in the cylinder, it is difficult to measure single cylinder AFR directly, which becomes a difficult part in individual cylinder AFR control. Therefore, how to estimate the AFR of each cylinder accurately through establishing the observer, it became a key issue of individual cylinder AFR control.Under the assumption that the amount of air into the cylinder is constant in each cycle, we focus on the fuel loop and establish a whole fuel path model of a SI engine. The fuel path model includes fuel wall-wetting model, exhaust gas mix model, exhaust gas transfer model and UEGO sensor model. The exhaust gas mixing process in the exhaust manifold is analyzed in a detailed because of the difference between the engine cycle and sampling periods of the UEGO sensor. The AFR at the exhaust confluence point sample formula is also derived. A discrete-time model of the fuel path is also set up using Matlab/Simulink library. The parameters of the model are obtained from enDYNA and other literature. Finally under the same working condition and with the same input, the fuel path model is checked out by the precise engine model of enDYNA.When analyzing the AFR estimation problem, the individual cylinder AFR is con-sidered as the system input and the measured AFR of the UEGO sensor as the system output, thus the individual cylinder AFR estimation problem belongs to the input esti-mation problem due to the unknown input of the system. Based on the fuel path model, a state/input observer for the individual cylinder AFR system is proposed and the AFR at the exhaust confluence point is estimated. The estimation value of the AFR in each cylinder is derived via analyzing the exhaust gas mixing process and the time sequence separation. Linear matrix inequality theory, to strike the input observer parameters and calculate the observer poles to verify the stability of the observer construction. LMI the-ory is used to calculate the parameters of the observer. The stability of the observer is validated because the poles for the observer is within the unite circle. Some offline simu-lation of the observer are implemented in enDYNA, the estimation results are compared with the true value which is obtained from enDYNA. The results of the experiments show that the designed input observer is rational and available.In order to verify availability of the observer algorithm in the real time environment, we also do the rapid prototyping experiment on the engine simulation platform, which is composed of dSPACE system and xPC-Target real time environment. Through a great deal of online simulation, it testifies that the designed observer is rational and feasible.In this paper, model-based observer design method is adopted. After detailed analysis and experimental verification of the discussion, the proposed individual cylinder AFR estimation algorithm achieved satisfactory results. But some problems still remain to be done. The individual cylinder AFR input observer is developed under the assumption that the air mass aspirated into the cylinder is constant in each cycle. In addition, some time delay, such as the exhaust gas transport delay and sensor response delay, is neglected. So there are some differences between the observer model and the actual system. The task of this article is just a preliminary study of our group in cylinder by cylinder AFR estimation and control. How to achieve more advanced method in AFR estimation and control needs further study and research.