Engine Knock Feature Extraction Based on Third-Order Cumulants and Filter

In order to detect knock in spark ignition engines, vibration signals are measured by knock sensor mounted on the head of engine block. The skewness of the vibration signal is analyzed, which is related to the probability distribution. Based on the analysis result, a method combine filter with third-order cumulants is proposed to extract knock feature. Experiments on modeled signals show that background noise is reduced and signal-to-noise ratio is improved. Estimated energy and period of knocking are appropriate. Knock features are also observed clearly from the results of real vibration signals processing.