Leak Detection, Localization and Size Prediction in Water Pipeline Systems
MetadataShow full metadata
Wireless Sensor Networks (WSNs) consist of wireless devices that are either installed above the ground or buried under dense soil or placed in any underground spaces. WSNs have an immense future to impact on diverse applications including leak detection in water, oil and gas pipelines. Any leak in the pipe can trigger significant financial losses and possible environmental damages. This thesis presents a novel method for detecting and locating a leak in a pipe and estimating its size using pressure sensors that can detect the slightest change of pressure. A laboratory-based test bench system has been designed and developed to collect real-world datasets from sensors using a wireless sensor network. Afterward, all datasets were preprocessed, and datasets containing leak information were separated. Next, exponential curve fitting with the leastsquare method was used to pinpoint leak location. However, leak size cannot be predicted using this method. Support Vector Machine (SVM) and Multi-layer Perceptron (MLP) neural network algorithms were then used to predict leak sizes. In our experiments, the MLP neural network showed higher accuracy over SVM in predicting leak sizes.