Abstract
Pipelines are one of the most widely used means for oil/gas and water transportation worldwide. These pipelines are often subject to failures like erosion, sabotage and theft, causing high financial, environmental and health risks. Therefore, detecting leakages, estimating its size and location is very important. Current pipeline monitoring systems needs to be more automated, efficient and accurate methods for continuous inspection/reporting about faults. For this purpose, several pattern recognition and data mining techniques have been brought into the research community. In light of the issues of low efficiency and high false alarm rates in traditional pipeline condition monitoring, in this paper, we have used negative pressure wave (NPW) coupled with intelligent machine learning techniques integrated in distributed wireless sensor network (WSN) to identify specific events beased on raw data gathered by individual sensor nodes. This collaborative approach reduces communication overhead to minimum by processing raw data on sensor nodes directly and reporting the detected events only. We apply the methods of support vector machine (SVM), K-nearest neighbor (KNN) and Gaussian mixture model (GMM) in multi-dimensional feature space. The suggested technique is validated using a serial publication of experimentation on a field deployed test bed, with regard to performance of detection of leakages in pipelines.