Feature extraction method of cavitation acoustic emission signals of hydraulic turbines based on improved EMD

Abstract

The cavitation of hydraulic turbine is one of the most difficult problems in the world. Empirical mode decomposition (EMD) is capable of decomposing the cavitation signals in an adaptive way. Because the traditional EMD has a special end effect, an improved EMD method was erected and applied to extract the features of cavitation acoustic emission (AE) signals of hydraulic turbines in this paper. Comparison of the mirror extension method and the average extremum extension method was carried out on the base of their ability to reduce the end effect, and the better mirror extension was adopted to deal with the cavitation AE signals. The measured AE signals generating from the cavitation were processed and the features were extracted by the improved EMD. Spectrogram can be drawn from cavitation inception to cavitation intensifies. In combination with the energy ratio of the intrinsic mode function (IMF) and the cavitation coefficient, it is found that the absolute energy ratio varies sharply under the cavitation conditions. The results show clearly that the improved EMD method can solve the defects of the end effect and is effective in the feature extraction of the cavitation acoustic emission signals of hydraulic turbines. The findings can be used as a technical basis for the identification of the cavitation conditions of Francis turbines.