Research On The Detection And Location Method Of Data-driven Long Oil Pipeline Leakage

Pipe which has a great advantage in conveying liquid items, especially in the field of petrochemical industry has been applied in a wide range. However, because of the development of pipeline network topology and the destruction of the environment and durable years, pipeline leakage has also gradually increased, which is a great threat to personal safety and the environment. Any manual inspection or maintenance must be inefficient and hardly be a real-time response to sudden accidents. Therefore, the design of pipe network leakage real-time detection system has a great significance.Previous pipeline leakage diagnosis and location methods are difficult to meet the requirements in judgment accuracy and real-time performance, especially serious lag phenomenon cannot deal with the accidents.In this thesis, we put forward a detection and location method of data-driven long oil pipeline leakage, which will greatly improve the real-time performance and robustness of the system.A low-pass threshold filtering method based on discrete wavelet decomposition is presented in this thesis, which mainly selects the low-pass filter as the way for making the decomposed high frequency into a quantitative threshold, and finally selects out the wavelet details including the local dynamic information for wavelet reconstruction. This method makes the final filtering effect to be of both high real-time and effective noise cancellation, preserving and sharpening the pressure waveform characteristics at the same time.Secondly, analyze the homology of the pipeline pressure signals, then judge whether a sampling point is abnormal, which is caused by pipeline leakage according to the time-varying threshold; finally, the continuous wavelet transform is applied to analyze the feature of the abnormal pressure points and to make a precise location of pipeline leakage points.Finally, this thesis goes for a research on the requirement analysis and design patterns for the pipeline leakage diagnosis system, and realizes a kind of improved operation department 10 architecture, which nearly keeps the efficiency of pressure data processing unchanged especially when the number of the host computers increases highly, just to find that the host computer is able to send the request by pipeline leakage to the operation department in real time,and then the operation department will swiftly respond to the request and deal with it.Compared with the traditional methods, the method in this thesis has high real-time performance, simple and effective, full of great accuracy. With respect to many diagnosis methods based on the theory of network of poor real-time performance, this method is more practical.