Abstract
Pipelines constitute one of the most important ways to transport large amounts of water through long distance. However, existing damage detection methods do not well in monitoring pipelines due to the harsh environmental condition. for this reason, current systems need to be more automated, efficient and accurate for continuous supervising about damage. For this purpose, Wireless Sensors Networks (WSN) have been brought into the research community in the context of monitoring water pipeline. In this paper, we have discussed the task amounts to provide low-cost and real-time damage detection technique. Our proposed technique is based on Fisher discriminant analysis (FDA) coupled with Support Vector Machine (SVM). We aim to recognize data as normal or outlier to identify specific events based on WSN implemented in a water pipeline. Moreover, we have compared our adopted approach with other four classifiers including Bayesian Network, Neural Network, K-Nearest Neighbors and Decision Tree. Thus, the suggested technique is validated in terms of accuracy and training time.