Valve internal leakage rate detection method based on acoustic emission

The invention discloses a valve inner leakage rate detection method based on acoustic emission, which comprises the following steps: firstly, collecting acoustic emission signals output when a valve operates at different inner leakage rates, extracting feature information of a time domain and a frequency domain, and combining features extracted by an EMD (Empirical Mode Decomposition) algorithm to form a feature vector before screening; performing feature screening by adopting a PCA (principal component analysis) algorithm, and taking a feature vector group formed by the screened features as a basis for model training; the method comprises the following steps: 1, determining a proper hidden layer number by using a reconstruction error algorithm, optimizing a hidden layer node number and a learning rate by using a particle swarm algorithm, finally determining a DBN network structure, and training model parameters; and finally, the trained deep belief network is used for predicting the inner leakage rate of an