Study On Pattern Recognition Of Storage Tank Bottom Corrosion Signal Based On Acoustic Emission

?Nowadays, the sustained high-speed national economic development makes the demand of energy supplies especially the oil and gas resource increasingly urgent. Qua the important facility of petrifaction area, storage tank is special for using wildly and incline to occur accident. The security of using and efficiency of testing is storage tank's two important issues.Storage tank works under changing environments, for a great many disadvantage factors they are inevitably subject to damage, especially the perforation, crackle and breach raised by electrochemistry corrosion are always bringing serious disaster and environment pollution. But routine testing methods require stopping the tanks'working, clearing them, removing preservation and so on. These processes cause oil loses and are very expensive, time-consuming also. So acoustic emission technology which can realize tank's online testing receive broad investigation.After consulting lots of acoustic emission investigation literature, the author devised and built a set of tank bottom corrosion experiment devices to obtain AE signals. And do deep researches on AE signal processing method to analyze corrosion, crackle and oil flow signals. By a mass of data testing, it proved to be feasible that to integrate signal feature extraction method based on the wavelet package decomposition with artificial neural network to analyze AE signals.The major study of this dissertation covers the following aspects:1. It elaborates the reason and principle of tank bottom corrosion and devises a set of tank bottom corrosion experiment devices based on electrochemistry corrosion principle to explore controllable signal. Besides it collects artificial crackle and oil flow signals as comparing signals.2. It applies feature extraction method based on the wavelet decomposition and wavelet package decomposition to extract feature vectors of tank bottom AE signals. And analyze their effects.3. It applies BP and RBF neural network approach to recognizing the three types of AE signals. Besides it construct a small-sample training instance to compare the validity of two neural networks.4. Through a great many field experimental dates, a more rational way to do some deep research on AE test of tank bottom corrosion was turned on.