Optimized Closed Loop Control of Controlled Auto Ignition CAI
For the near future gasoline fueled combustion engines will be the main energy source for public transport. Controlled Auto Ignition (CAI) is a very promising combustion process with a high efficiency and low pollutant emissions. This can be achieved by a high amount of residual gas leading to a low peak combustion temperature and hence to less NOx-Emissions. The combustion self ignites simultaneously in several spots which leads to a spontaneous reaction and therefore to a high efficiency and less pollutants. A very good tool to do research on this combustion is a single cylinder engine with a fully variable electromechanical valve train and piezzo direct injection. The drawback of this process is the limited operating range in the engine map. The limits can be extended towards higher loads by boosting, but nevertheless the combustion is sensitive to disturbances especially with dynamic changes in the load. The valid range for the actuating variables is limited. The location of the maximum pressure is related with the load. The higher the load the earlier the combustion´s maximum pressure is located. Additionally the combustion´s efficiency shows a dependency on the location of the maximum pressure. This location should always be kept at an optimum position dependent on the load. Obviously the two control tasks load and efficiency are coupled. The spontaneous combustion can lead to a steep pressure rise from which inadmissible noise results and mechanical stress of the piston can result. Therefore a controller is needed to stabilize the combustion which satisfies all mentioned demands. A Model based Predictive Controller is able to regard a limit on the pressure rise as well as on the actuator´s operating range. Additionally this controller is capable of optimizing the transient behavior during dynamic load profiles. The desired load can be reached and the most efficient possible location for the maximum pressure can be realized. Obviously this controller has a high potential for the described task. The paper presents the research engine-setup and the valve timing strategies for the recirculation of the exhaust gas from an automatic control point of view. The possibilities to reduce the degrees of freedom of the valve train are discussed. For the development of a Model based Predictive Controller a model is needed. A group of researchers at the RWTH Aachen University is developing a physical model for the use with a Model based Predictive Controller. As long as this model is not available a neural network is used for modeling the process. Based on a special form of neural net a new nonlinear observer is created which also estimates disturbances. This nonlinear observer is used in combination with a nonlinear Model based Predictive Controller to simultaneously control load and location of the maximum pressure as a measure for efficiency. This innovative concept allows formulating boundaries on the pressure rise for the closed loop. The controller will be evaluated in a closed loop simulation with and without measurement noise added. An outlook will present further research steps.
Session: HCCI Combustion