Algorithm Comparison for Real Time Knock Detection

This paper addresses the implementation and comparison of algorithms for real-time knock detection. Knock is an unwanted abnormal combustion process that may damage engines and limit their efficiency. For series vehicles, knock detection is generally obtained from knock sensors that capture other noise sources, thus requiring robust algorithms. In order to estimate the performance of time-frequency and Kalman filter based algorithms, a knock signal model is proposed and the algorithms are tested under various noise conditions. Experiments on modelled and real signals show the superiority of the recently developed S-method with respect to the extended Kalman filtering.