Abstract:
The main objective of this work was identification of automotive engine condition through its sound using exact wavelet analysis. Continuous wavelet transforms (CWTs) are widely recognized as effective tools for vibration-based machine fault diagnosis. In the present study, exact wavelet transform algorithm is used to identify fault signals of improper adjustment of valve clearance and defective nozzle. The design of exact wavelet transform was based on genetic algorithms (GAs).The algorithm uses three operators namely selection, crossover and mutation. At each selected timeframe, the algorithms generate an adaptive daughter wavelet. During optimization process, exact wavelet analysis considers both optimization of wavelet coefficients and satisfaction of the admissibility conditions of wavelet. The sound signals of the engine were captured in the isolated room, to control the interference of other sounds from the surrounding. Two scenarios were considered i.e. improper adjustment of valve clearance and defective nozzle. Exact wavelet transform transforms time-domain waveform in to time-frequency domain and estimates the signal in the time and frequency domains simultaneously. Consequently, time-frequency contour map introduced in fault identification of Toyota 3B diesel engine. The contour maps of a normal diesel engine scalogram revealed low (blue) energy distribution mostly from 37 Hz to 73 Hz. The sound signal from improperly adjusted valve clearance scalogram gave a sound signal from 28 Hz to 64 Hz with sound signal coefficient from low (blue) to high (red) energy distribution. The sound signal of faulty injection nozzle scalogram distributed from 37 Hz to 64 Hz. Maximum frequency for normal, improper adjustment of valve clearance and injection nozzle is 91 Hz, 82 Hz and 75 Hz respectively using EWT; maximum frequency for normal, improper adjustment of valve clearance and injection nozzle is 139 Hz, 138 Hz and 137 Hz using FFT. This shows a vital difference between EWT and FFT.
