Spectral transient-based multiple leakage identification in water pipelines: An efficient hybrid gradient-metaheuristic optimization

Alireza Keramat, Iman Ahmadianfar, Huan Feng Duan, Qingzhi Hou

Research output: Journal article publicationJournal articleAcademic researchpeer-review

2 Citations (Scopus)

Abstract

In transient-based leak detection (TBLD), the localization of multiple leaks from the measurement data of a few stations is challenging. Recent studies have made breakthroughs in offering novel maximum likelihood estimators, however, their objective function is multi-dimensional, non-convex, and nonlinear, for which a robust solution is hard to achieve. This study aims to combine the metaheuristic and gradient-based optimization (MGBO) techniques to find the optimum of the novel objective function and pinpoint the leak locations. In addition, two novel initialization algorithms are proposed and incorporated for single, two, and three leaks cases. The numerical results of the proposed method demonstrated a much greater efficiency than the classical gradient-based techniques: more than 96% of the three-leak localization cases had an RMSE of less than 15%, with an initialization population of less than 0.3% of the classical methods, thus implying a remarkable efficiency. The discussions on various leak realizations reveal that the proposed method is robust and efficient in localizing multiple leak cases, thus moving a step forward to enhance the TBLD techniques.

Original languageEnglish
Article number120021
JournalExpert Systems with Applications
Volume224
DOIs
Publication statusPublished - 15 Aug 2023

Keywords

  • Leak detection
  • Maximum likelihood estimation
  • Metaheuristic gradient based optimizer
  • Optimization
  • Signal processing
  • Water hammer

ASJC Scopus subject areas

  • General Engineering
  • Computer Science Applications
  • Artificial Intelligence

Fingerprint

Dive into the research topics of 'Spectral transient-based multiple leakage identification in water pipelines: An efficient hybrid gradient-metaheuristic optimization'. Together they form a unique fingerprint.

Cite this