A Multi-Leak Identification Scheme Using Multi-Classification for Water Distribution Infrastructure

Research output: Journal Publications and ReviewsRGC 21 - Publication in refereed journalpeer-review

4 Scopus Citations
View graph of relations

Related Research Unit(s)


Original languageEnglish
Article number2128
Journal / PublicationApplied Sciences (Switzerland)
Issue number4
Online published18 Feb 2022
Publication statusPublished - Feb 2022



Water distribution infrastructure (WDI) is well-established and significantly improves living quality. Nonetheless, aging WDI has posed an awkward worldwide problem, wasting natural resources and leading to direct and indirect economic losses. The total losses due to leaks are valued at USD 7 billion per year. In this paper, a multi-classification multi-leak identification (MC-MLI) scheme is developed to combat the captioned problem. In the MC-MLI, a novel adaptive kernel (AK) scheme is developed to adapt to different WDI scenarios. The AK improves the overall identification capability by customizing a weighting vector into the extracted feature vector. Afterwards, a multi-classification (MC) scheme is designed to facilitate efficient adaptation to potentially hostile inhomogeneous WDI scenarios. The MC comprises multiple classifiers for customizing to different pipelines. Each classifier is characterized by the feature vector and corresponding weighting vector and weighting vector pertinent to system requirements, thus rendering the developed scheme strongly adaptive to everchanging operating environments. Hence, the MC scheme facilitates low-cost, efficient, and accurate water leak detection and provides high practical value to the commercial market. Additionally, graph theory is utilized to model the realistic WDIs, and the experimental results verify that the developed MC-MLI achieves 96% accuracy, 96% sensitivity, and 95% specificity. The average detection time is about 5 s.

Research Area(s)

  • Adaptive kernel, Graph theory, Hydraulic model, Multi-criteria decision-making, Multiple classifiers

Download Statistics

No data available