The Wavelet Transform (WT) has been developed two decades ago, and has since then been put to use in an increasingly wide array of applications. The WT provides a time-scale analysis of a signal. Compared to the widely-popular Fourier Transform (FT), originally developed two hundred years ago, the WT provides the time-evolution of the signal at different scales. The Discrete Wavelet Transform (DWT) is a computationally efficient implementation of the WT, in which the time-scale analysis is performed on a dyadic scale.
The DWT is very suitable for knock detection systems, since it can provide the history of the knock signal at discrete scales within a crank angle window. It allows for the extraction of a multitude of features from the time-scale plane. Moreover, the DWT is suitable for real-time knock detection implementations on engine control units. Since the frequency spectrum of the knock-induced vibrations corresponds to a high-frequency band (typically 5 ∼ 18 kHz), the DWT decomposition need only be performed at the higher scales which covers this band, rendering the DWT analysis even more computationally attractive.
This paper gives an analysis of the knock signal of a four-cylinder spark-ignition engine using the discrete wavelet transform, identifying features from the time-scale decomposition which can be used for the detection of knock. Such an analysis is very important and can serve as a basis for the formulation of a DWT-based knock detection scheme.