Research On Leakage Diagnosis For Intermittent Transport Pipeline

The development of pipeline transportation has greatly contributed to the development of chemical industry and oil and gas resource industry. At the same time, the laying of oil and gas pipeline increased year by year. The service term of pipeline is also growing. Due to construction defects, erosion, man-made destruction, aging equipment and the geography and climate change and other reasons, pipeline leakage accidents occur frequently, causing the waste of resources and the loss of a large number of property. Leakage diagnosis technology for pipeline is a comprehensive technology with complex knowledge and multi subject. For intermittent transport pipeline, the working condition of pipeline is very complex, and restricted by many factors. So far, there has not been a method for leakage diagnosis is general. Each method has its own adaptability. Therefore, to achieve a reliable diagnosis of intermittent transport pipeline leakage has important theoretical and practical value.In this paper, on the basis of the time domain statistical feature analysis and information entropy theory, a method for feature extraction of acoustic signal is proposed; through the study of the statistic features of acoustic signals, this paper extracts the time domain statistical features of acoustic signal to describe the "long tail" that exhibited in the frequency distribution histograms when the leakage occur; for the actual pipeline, acoustic signals is frequently disturbed by the sudden and persistent interference of transportation technology. This paper study the information entropy and its application in leakage detection for pipeline in-depth. A method for extraction of power spectrum entropy in frequency domain and time-frequency feature of wavelet packet space spectrum entropy is proposed to perfect a complete description of the acoustic signal. For the actual pipeline on stable working, it is difficult to obtain large amount of leakage samples. The leakage diagnosis model based on support vector data description with single value characteristics. With the training of normal samples, it avoids the influence on the generalization ability of the model due to diversified leakage forms.Field test shows that this method for leakage diagnosis proposed by this paper can realize the fast and accurate diagnosis of intermittent transport pipeline leakage effectively. This method has high reliability, stability and adaptability, and provides an effective method to solve problem of reliable diagnosis of intermittent transport pipeline leakage.