Detection of misfire and knock in spark ignition engines by wavelet transform of engine block vibration signals

Abstract

The high-frequency components of engine block vibration signals during misfire and normal combustion show clear differences of duration and scale by wavelet analysis - continuous wavelet transform (CWT). A new factor, combustion noise intensity (`CNICWT'), has been defined, which has been derived from the CWT scalogram. It has been shown that CNICWT can detect misfire more accurately than fast Fourier transform power spectral density. Similarly, it has been found that CNICWT can also be applied to knock detection. The durations of both knock and misfire detection overlapped each other at the same wavelet scale 8; misfire, normal and knock cycles can be indicated on the same histogram by CNICWT. Moreover, knock and misfire detection can be combined to one CNICWT computation.