Probability distribution of knock factor and knock-band method for knock detection

摘要

Knock Factor relatively indicates the knock energy during knocking combustion, and is widely used for knock detection of gasoline engines. Focusing on probability distribution of KF, mass data of combustion test were studied statistically by several approaches such as quantile graph, kurtosis and skewness check, as well as Lilliefors test. It is found that KF sample of test cycles without knocking combustion conforms to the law of normal distribution while KF sample of test cycles with knocking combustion does not. However, the KF sample embodying knock signal can be still normalized by removing abnormal individual which can be identified by the statistical method of step-phase check. Based on probability distribution characteristics of normalized KF sample, the confidence interval of μ+3σ was set as knock-band detector. Test cycle which KF locates within the confidence interval band was treated as a light knock cycle, test cycle which KF goes beyond the superior limit of the band was considered as a knock cycle. High-pass filter cylinder pressure signals of knock cycles and normal combustion cycle were investigated to test and verify the knock-band detection method.