Likelihood Control of Knock Intensity

Control of knocking combustion in spark-ignition engines is challenging due to the random nature of the knock process. In this work, a new stochastic knock control system is proposed based on the likelihood ratio of the observed knock intensities to those expected under borderline knock conditions. Unlike most prior methods which implement some form of knock event-based control or quantify the error at a specific point of the knock intensity distribution, the proposed method is sensitive to any departures from the borderline knock reference distribution. The performance of the controller is analyzed using both individual time-history instances of the response and extensive Monte Carlo simulations both with and without disturbances. The results show that the new controller has a significantly improved regulatory response relative to conventional strategies and recovers from overly retarded conditions more quickly. The method is robust to disturbances and is able to operate closer to the knock limit without increased risk of engine damage.