SI Engine Knock Detection Method Robust to Resonance Frequency Changes

Knocking combustions in SI gasoline engines manifest in characteristic structure-borne vibrations at specific resonance frequencies. Knock detection is usually based on some way of processing of this vibration signal, focusing on one or more spectral peaks. The optimal method of detection requires exact knowledge of the peak positions, which is not available in practice due to different variations. An adapted detector is thus necessary in order to keep a good detection performance even when the resonance frequencies are only known approximately. We present a detection technique that is robust with respect to a mismatch between the presumed resonance frequency values and those actually present in the knock sensor signal. Results are illustrated by theoretical performance evaluation as well as by real data application.