The actual denoising effect of wavelet denoising methed was studied through pipe flow test apparatus, which simulated the actual pipeline leakage condition. The reason why small leak was not easy to be found was explained from two expects through timefrequency analysis of the leakage signal attenuation process. And the disadvantages of the most commonly used wavelet de-noising method to deal with small leakage signal were also discussed. On this basis, a new threshold function was established to improve the signal reconstruction accuracy and the advantage was analyzed mathematically. Next, a blind source separation algorithm based on maximum signal-to-noise ratio (SNR) was proposed, which integrated wavelet transform with blind source separation. By separating the known structural signal, the applicability of this method was illustrated, and the practical denoising effect and industrial application value of the fusion algorithm were verified.

        Given the frequent accidents attributed to abnormal pressure fluctuation and the less effectiveness of preventive measures of complex pipe network, the FTA method of graph theory was applied to solve this problem. Induced mechanism of abnormal pressure fluctuations was interpreted and the FTA model of pressure abnormal fluctuations in complex pipeline network was firstly proposed and established in this paper. An imported crude oil pipe network was taken as a case study by using the Delphi method and fuzzy mathematics theory to solve the model. Results agree well with the actual statistics. Then K-means clustering method was carried on to analyze the three importance of typical basic events and the three-tier structure of the abnormal fluctuations is proposed using the DTA method, which lies a theoretical basis for the analysis on abnormal events as well as the optimization of the prevention and control measures. This paper mainly offers a new research idea in the field of operation security and integrity management of the large, complex pipe network especially when the historical data is far from complete.