Hydraulic oil gas tank sound identification method based on subdivision abstract features

The invention discloses a compressed oil gas tank sound identification method based on subdivision abstract features. The method comprises the following steps of: acquiring sound data and performing subdivision and re-processing by utilizing a high sampling rate of a sound sensor; equipment sound abstract feature extraction is carried out by using a matrix auto-encoder; building a convolution-circulation parallel neural network, and carrying out training through subdividing abstract coding features; and obtaining a sound recognition result of the compressed oil gas tank according to the trained convolution-circulation parallel neural network. The method effectively solves the problems of low identification speed, low reliability and the like of the current power plant auxiliary equipment state detection technology, reduces data redundancy through extracting and subdividing abstract features, effectively avoids the possibility that gas leakage features are hidden, and improves the detection efficiency. Accordi