93results about How to "Reduce missed diagnosis" patented technology

The invention discloses a circuit breaker voiceprint and vibration entropy feature fusion fault diagnosis method. The method comprises the steps of carrying out band-pass filtering on a sound signal,and obtaining the roughness, box dimension, direction degree, contrast ratio, linearity and regularity which reflect time-frequency texture features through generalized S transformation; constructingvoiceprint features based on the roughness, the box dimension, the direction degree, the contrast ratio, the linearity and the regularity which reflect the time-frequency texture features; selecting parameters for a vibration signal, and obtaining permutation entropy through variational mode decomposition; constructing vibration entropy features based on the permutation entropy; performing deep fusion based on the voiceprint features and the vibration entropy features to obtain a joint feature vector; performing learning training of an SVM model on the joint feature vector; and based on the GWO-SVM diagnosis model, optimizing a penalty factor and a mixing coefficient in the SVM model by adopting the GWO model. In the embodiment of the invention, the influence of sensor frequency response and environmental noise on the fault diagnosis of the circuit breaker is reduced.