The identify method of boiler leakage acoustic emission signal based on AR and HMM

Acoustic emission signals exist in the boiler leakage accident widely.It can reduce the harm to boiler if classify acoustic emission signal's type when the accident occurred effectively.It put forward a classification method based on the autoregressive model and hidden Markov model.Firstly we gathered tapping,sand paper and broken lead three kind of simulated leakage acoustic emission signals,and adopted the method of framing to the acoustic emission signals.Then we extracted characteristic value using the autoregressive model (AR) for each frame signal.Finally we used hidden Markov model (HMM) to train and test characteristic parameters.The test results showed that the recognition accuracy is 91.1%.It identifies that the method based on AR and HMM can identify three types of simulated acoustic emission signals effectively.It provided the basis for visual diagnosis of boiler leakage fault.