Feature extraction of machine sound using wavelet and its application in fault diagnosis

Abstract Machine sound always carries information about the working of the machine. But in many cases, the sound has a very low SNR. To obtain correct information, the background noise has to be removed or the sound must be purified. A de-noising method is given in this paper and is successfully used in feature sound extraction. We can easily diagnose a machine using the purified sound. This de-noising method is based on the wavelet technique and uses the Morlet wavelet as the mother wavelet, because its time–frequency resolution can be adjusted to adapt to the signal to be analyzed. The method is used for extracting the sound of some vehicle engines with different types of failure. The feature sound is extracted successfully.