A Novel Approach of Wavelet Packet Transform for Knock Signal Analysis

The computation complexity dominates Continuous Wavelet Transform (CWT) and the low frequency resolution limit the Discrete Wavelet Transform (DWT). The degradation in frequency resolution in Discrete Wavelet Transform (DWT) is due to the fact, that, DWT decomposes only the approximation coefficients at each level. This detail is not enough to analyse the signal like knock in the engine. The acceleration sensor or knock sensor receives the signal from various part of the engine (i.e. the signal from various sensors will super impose) and a specific technique to analyses this signal is in need. The idea of decomposing both approximation coefficients and detail coefficients is achieved by Wavelet Packet Transform. This paper proposes a novel approach of analysing the knock signal by Wavelet Packet Transform. The signals are captured from the engine running at various RPM (Revolution Per Minute) (i.e. 1000, 2000, 3000, 4000 and 5000). The signals are sampled at a rate of 150, 200, 250, 300 and 350 per clock cycle. The samples are aggregated for a period of 10,000 clock cycles and analysed using the Wavelet Packet Transform in Matlab 2014. The results are promising in terms of error rate, when compared with the conventional Fourier and Wavelet Transform.