A Novel Statistical Approach to Detect Leakages in Water Pipeline Systems

A Novel Statistical Approach to Detect Leakages in Water Pipeline Systems


Harshit Shukla, Clemson University, Clemson, SC United States
Kalyan R. Piratla, Clemson University

Abstract: Recent studies have established that a significant amount of treated potable water is lost through pipeline leakage in the United States and several other countries. Many of the current leakage detection techniques are expensive, human-dependent, and do not provide continuous monitoring capabilities. There is a growing interest in employing inexpensive wireless sensor networks (WSNs) to continuously monitor distribution pipeline networks and detect significant leakages in near real-time. The authors have previously developed and tested a vibration-based leakage detection technique (LDI) on a two-looped, pipeline setup comprising multiple bends, joints, and different burial conditions. In this paper, we aim to propose and demonstrate a novel statistical model to predict leak sizes and locations based on pipeline vibration measurements. Pipeline acceleration data measured at multiple locations along its length was processed and used to train and develop an artificial neural network model to correlate spatiotemporal acceleration data with leak sizes.

This trained model was then used to predict leak sizes and locations. Multiple leakage locations and multiple severities were configured on the two-looped real-size experimental pipeline testbed on Clemson campus to train and validate the proposed leakage prediction approach. Findings of this study show that the trained model was able to predict the leakage sizes with approximately 80% accuracy. The accuracy of the model depends upon the size and variation of training data. Predictive models such as this will advance the state-of-the-art knowledge of leakage detection, especially to bypass the physical/structural modeling of leakage-induced effects and their monitoring for leakage detection.

Publication Date: 2019

Presented at:
NASTT’s 2019 No-Dig Show Chicago, Illinois
March 17-21, 2019 | Paper Number: TM1-T3-04

Non-Member Price: $10.00

Discounted Member Price: $0.00

Your Price: $10.00