Analysis of the incipient cavitation noise signal characteristics of hydroturbine

In order to find the characters of incipient cavitation. Use the method of wavelet time-frequency analysis and wavelet packet decomposition to process the cavitation noise signals. Use the value of peak factor and slope of power spectral density curve as a threshold when incipient cavitation to judge whether the cavitation occurs. The results shows that the characteristics of incipient cavitation can be detected in the auditory frequency band. The wavelet time-frequency analysis of noise signals can distinguish the different operating conditions, also can discriminate between the phenomenon of incipient cavitation and the other state of cavitation by visual observation. The wavelet packet decomposition can obtain the feature frequency of cavitation signal is 10-13 kHz. The way of judge whether the incipient cavitation happens by threshold value can reach the accuracy rate up to 70% which meet the requirements of the detection for incipient cavitation.