A method for extract and classifying vibration mix features of pumped storage unit

A method for extracting and classifying mix characteristics of cavitation vibration signal of pumped storage unit features that a set empirical mode decomposition method is used to extract characteristic of collected original signal to obtain a series of eigenmode functions, and energy characteristics and singular value characteristics of IMF components are extracted. At the same time, a variety of typical time-domain and frequency-domain features of the original signal are extracted manually. Then, the mixed eigenvector of the original signal is composed of the time-domain, frequency-domain,energy and singular value features, which is used as the input of RBF neural network to classify and identify the cavitation signals of pumped storage units under different working conditions. The invention can effectively solve the problem of accurate diagnosis of cavitation vibration signals under different operating conditions of the pumped storage unit by extracting and classifying the characteristics of cavitation vi