Using multi-rate filter banks to detect internal combustion engine knock

Artikel i vetenskaplig tidskrift, 1997

The wavelet transform is used in the analysis of the cylinder pressure trace and the ionic current trace of a knocking, single-cylinder, spark ignition engine. Using the wavelet transform offers a significant reduction of mathematical operations when compared with traditional filtering techniques based on the Fourier transform. It is shown that conventional knock analysis in terms of average energy in the time domain (AETD), corresponding to the signal's energy content, and maximum amplitude in the time domain (MATD), corresponding to the maximum amplitude of the bandpass filtered signal, can be applied to both the reconstructed filtered cylinder pressure and the wavelet coefficients. The use of the filter coefficients makes possible a significant additional reduction in calculation effort in comparison with filters based on the windowed Fourier transform. A knock classification system based on the duration of the knocking process, defined as the time during which the signal exceeds an energy threshold, is presented using different scales. An advantage of this method is that the relation of the duration between different frequency bands can be used to draw conclusions about the presence of multiple knock.