Leveraging Hydraulic Cyber-Monitoring Data to Support Primitive Condition Assessment of Water Mains

Abstract

Buried water pipelines deteriorate in response to several variables such as pressure transients, corrosion, and pipeline material degradation, among others, that are dynamic processes, and it is therefore difficult to predict the pipeline condition without employing expensive sophisticated technologies. Such technologies are ad hoc in nature and may be worthwhile only for those pipelines that are known to be deteriorated and critical for the reliability of the water distribution network (WDN). Adopting cyber-monitoring methods for pipeline condition assessment, this paper presents and demonstrates a data-driven condition assessment platform that can serve as a primitive indicator of water pipeline conditions. Flow, pressure, and water consumption data collected in a synchronous manner are employed to predict pipeline roughness coefficients and effective internal diameters through a combination of hydraulic modeling, evolutionary algorithms, and neural networks utilizing two popular benchmark WDNs. The accuracy of pipeline condition prediction, measured using mean absolute percentage error (MAPE), ranged between 4.12% and 17.6% based on numerous scenarios in this study. Effective internal diameters were found to be more accurately predictable than pipeline roughness coefficients, and it was also found that pressure monitoring alone can suffice the requirements of the proposed framework in order to produce accurate pipeline condition prediction. It is recommended that future research be conducted over the robustness of this platform for other dynamic parameters such as leakages.