The Community for Technology Leaders
Computer Science and Information Engineering, World Congress on (2009)
Los Angeles, California USA
Mar. 31, 2009 to Apr. 2, 2009
ISBN: 978-0-7695-3507-4
pp: 656-660
It is well known that the work condition of pipeline, the leak included, can be identified by a pressure signal analysis. Because of the high frequency data collection and always on-line pipeline leak detection, the pressure signal brings up massive data. A methodology for pipeline leak detection using data mining technology and work condition identification is presented here. Sixteen groups of raw data, which include each work condition, are selected from massive pressure data collected in this field. In order to analyze data conveniently, each group of raw data is normalized with mean zero. With wavelet transform, high-frequency noise is eliminated from pressure signal. The analysis on time-domain analysis proves that statistical value can describe pressure variation clearly and responsively. The paper extracts time-domain statistical value from de-noised pressure data as characteristic indexes for fuzzy clustering. According to the fuzzy clustering efficiency and accuracy, six time-domain parameters are regarded as the characteristic indexes. These parameters are root mean amplitude square, square root amplitude, skewness, fluctuation factor, variance and slope. Work condition of pipeline can be described completely by the Eigenvector, which is composed of six time-domain indexes. Clustering centers are found by fuzzy C-means algorithm with sixteen groups samples’ eigenvectors. Applied clustering centers, field work condition can be identified with calculating and comparing close degree. The result of field application showed that the work condition identification accuracy can be up to 95%.
Keywords: data mining, work condition, fuzzy C-means algorithm, model identify, leak detection

L. Wei, Z. Laibin and Y. Yingchun, "Data Mining Technology Based Leak Detection Method for Crude Oil Pipeline," 2009 WRI World Congress on Computer Science and Information Engineering, CSIE(CSIE), Los Angeles, CA, 2009, pp. 656-660.
89 ms
(Ver 3.3 (11022016))