Echtzeitfähiger Algorithmus zur automatisierten dynamischen Vermessung der ottomotorischen Selbstzündung
Wick, Maximilian Kurt; Andert, Jakob Lukas (Thesis advisor); Abel, Dirk (Thesis advisor)
Aachen (2019, 2020)
Dissertation / PhD Thesis
Homogeneous Charge Compression Ignition (HCCI) offers great potential for increasing efficiency of combustion engines while simultaneously reducing nitrogen oxide emissions. However, a broad application of HCCI fails mainly due to the challenges of low combustion stability at the edges of the operating range and a high sensitivity to changing boundary conditions. Stochastic outliers and incomplete combustion lead to unstable sequences which reduce efficiency and increase emissions. Closed loop control is necessary to solve the challenge of combustion stability. For stabilization, advanced control algorithms like model predictive control (MPC) use physical or data-driven models of HCCI combustion to achieve cycle-to-cycle control of the process. The models used for these model predictive controllers are normally based on measurements under stationary operating conditions and therefore contain only a few outliers, misfires or incomplete combustion. It is precisely the prediction of incomplete combustion or even misfire that has hardly been possible to date. Additionally transient states cannot be trained with these measurements. Especially the effects of single control interventions cannot been analyzed by steady state measurement. To improve the model quality, a dynamic measurement methodology of all control variables of the engine is necessary. At the same time, the condition of the last combustion must be taken into account, as the cylinder state of the last combustion, which cannot be directly controlled, has an enormous influence on the subsequent combustion due to the strong cycle to cycle coupling due to negative valve overlap. In this article, a measurement method is presented which makes it possible to automatically set up the transient limitations for HCCI combustion while maintaining limits of stability, maximum pressure gradient and others. At the same time a broad data base is created with this measurement method. The actuators of the engine control unit are changed dynamically on a cyclic and inner cyclic base in several dimensions. This new measurement method is then exemplary used to train artificial neural networks, which can be used to predict misfires during HCCI combustion. Finally, an feasibility study regarding the usability of the newly gained measurement data in a data driven control algorithm for the combustion process is carried out and validated on a single cylinder test bench.