|
Acoustic emission on-line detection technology has the characteristics of continuoustesting, less demanding to the environment and the shape of the component, a wide rangeof detection. In recent years, it draws the attention of the public. Kaiser, Tatro, Green’sresearch achievements are the milestone in the history of the acoustic emission. Acousticemission technology began in the nineteen fifties, introduced to the field of non-destructivetesting in the early sixties, gradually come to the pressure vessel research. Recently, as anew method for the detection of tank bottom, it has become more and more valued.Acoustic emission signal is a typical non-stationary signal, having the transient andpolymorphism. The traditional identification methods such as fuzzy recognition, artificialneural network, and statistical recognition are difficult to establish the membershipfunction, the classification is not much, not from the overall point of view to identify, sothey can not response characteristics of the acoustic emission signal very well.Autoregressive Model (AR) has good time-frequency resolution, can be sensitive responsesignal characteristic, this paper uses Autoregressive Model for feature extraction method,with acoustic emission experiment platform to adopt the plate crack cracking signal,corrosion of steel in week load deformation signal and corrosion of the oxide from thestripping signal, the method each signal segments respectively extract the AR eigenvaluecan overcome the defects that traditional AR model parameter extraction method used inlong time series. Hidden Markov Model is a kind of markov chain, because of itsadvantage of statistical learning and probabilistic reasoning ability, this paper choose it as atool to analysis poor reproducibility, large amount of information acoustic emission signal.Q235steel corrosion experiment, folding pressure crack experiment and oxide stripping experiment show the following conclusions:(1) AR analysis method from the constraints of sampling points, the most importantfeature is the coefficient with time-varying, so it is very effective to analysis non-stationarysignal, the autoregressive parameters is sensitive to state change, can accurately reflect thecharacteristics of acoustic emission signal.(2) Combination with autoregressive model and Hidden Markov Model caneffectively identify corrosion acoustic emission signal, flaw acoustic emission signal andoxide stripping acoustic emission signal.
|