Stabilization of the GCAI combustion process by in-cycle correlations

点击翻译按钮获取中文摘要

The need for mobility is constantly increasing. At the same time, the increasing release of anthropogenic CO2 from fossil sources and local air pollution are increasingly coming into the spotlight. Research on energy-efficient and low-emission propulsion systems can make an important contribution in this respect. Gasoline Controlled Auto Ignition (GCAI) is a promising approach to both increase efficiency and reduce pollutant emissions. However, the technical use of GCAI still faces unresolved challenges, such as the characteristic process characteristics with autoregressive cycle coupling and the dependence of combustion stability on the thermodynamic boundary conditions. Furthermore, the restriction of the engine operation map must be pointed out. In order to enable transient operation as well as to extend the possible operation range, the research is focused on concepts based on control engineering. Within this research unit, a novel optimization-based multi-scale control approach is followed to stabilize the GCAI combustion process by using in-cycle correlations. In the first project phase the potential of in-cycle control has already been successfully demonstrated.For model-based control, the prediction accuracy for stochastically occurring outlier cycles and the associated cycle coupling is of essential relevance. A main focus of the experimental work is the methodology for determining the autoregressive character of the combustion process by transient excitations. Due to the characterization as Markov process it is assumed that machine learning (e.g. reinforcement learning) has substantial advantages which cannot be achieved by conventional methods. Tailor-made algorithms for measuring the GCAI process will be developed to generate a broad database with transient data for system identification. A focus of the process modelling is on the emissions in order to be able to integrate them later into the cost function of nonlinear model-predictive control.The low-temperature kinetics is strongly dependent on the thermodynamic boundary conditions, which cannot be measured directly. Novel sensor concepts such as analysis of the ion current have the potential to determine the chemical/thermodynamic state in the cylinder more precisely and subsequently improve modelling and control. In cooperation with TP6 the ion current signal will be analyzed to use it as an additional input variable for the controller. Beyond the current state of research, information from the ion current sensor will be used to improve the models. Afterwards it will be investigated whether the ion current is suitable as an additional sensor quantity for integration into the controller. Finally, the control system will be validated in MiL and HiL operation.