An Approach to Fault Diagnosis Considering False Alarm and Middle State

Abstract

In order to monitor the false alarm and fault middle state in the process of fault diagnosis simultaneously, a mixed Hidden Markov Model(HMM) considering the false alarm and fault middle state was proposed to reduce the false alarm and monitor the middle state of the system. First of all, the state classification in the process of the complex system state monitoring was studied. The system states were divided into normal state, intermittent failure, random failure, intermediate state and failure state, and then the mixed HMM was established to reflect the true state of the system. Secondly, the effective feature vectors were extracted based on wavelet packet decomposition method to eliminate redundancy and high dimension of original characteristic signal. Then all kinds of the HMM states were obtained through the model training, and fault diagnosis was implemented through inputting feature vector. At last, the feasibility and effectiveness of the method was verified by an example.