A Study on In-Cycle Combustion Control for Gasoline Controlled Autoignition 2016-01-0754
Gasoline Controlled Auto Ignition offers a high CO2 emission reduction potential, which is comparable to state-of-the-art, lean stratified operated gasoline engines. Contrary to the latter, GCAI low temperature combustion avoids NOX emissions, thereby trying to avoid extensive exhaust aftertreatment. The challenges remain in a restricted operation range due to combustion instabilities and a high sensitivity towards changing boundary conditions like ambient temperature, intake pressure or fuel properties. Once combustion shows instability, cyclic fluctuations are observed. These appear to have near-chaotic behavior but are characterized by a superposition of clearly deterministic and stochastic effects. Previous works show that the fluctuations can be predicted precisely when taking cycle-tocycle correlations into account. This work extends current approaches by focusing on additional dependencies within one single combustion cycle. A concept of an in-cycle combustion control algorithm is developed the shows the potential to improve the controllability and consequently increase combustion stability in stationary operation. It is implemented and tested on a single cylinder engine with an electromechanical valve train to control the demanded internal residual gas fraction. The control is based on real-time analysis of the cylinder pressure which is computed by a field programmable gate array module on a rapid prototyping engine control unit. Thereby, the subsequent combustion characteristics, characterized by the center of combustion (crank angle where 50 % heat release is achieved: α50 ) is predicted as a function of the thermodynamic state during recompression that includes influences of the previous cycles as well as the gas exchange of the current cycle. The intake valve timing is adapted following a linear prediction to minimize the fluctuations of α50. Applying the in-cycle control strategy at n = 1500 1/min and imep = 4 bar reveals an improvement of the standard deviations of imep and α50 by 40 % compared to open loop operation.