Machine learning modeling for spectral transient-based leak detection

Vahid Asghari, Mohammad Hossein Kazemi, Huan Feng Duan, Shu Chien Hsu, Alireza Keramat

Research output: Journal article publicationJournal articleAcademic researchpeer-review

Abstract

This paper describes Machine Learning (ML)-based framework to detect leaks in pipes using transient waves. The so-called Transient-Based Leak Detection (TBLD) technology is a non-convex multi-dimensional optimization problem, usually solved by inefficient Metaheuristic Optimization algorithms (MOAs). This paper proposes an efficient ML approach to address this drawback. At the core of this methodology, an ensemble of CatBoost models was trained on >3.8 million data records for classifying the leaky sections and predicting the leak sizes. Results showed that the ML models could detect leaks with 97% accuracy and an F1-score of 0.86, implying a significant superiority compared to the MOAs. Substituting the cumbersome optimizations with an ML-based approach, this paper opens a new line of research in the TBLD, which seems more welcome for complex pipe networks.

Original languageEnglish
Article number104686
JournalAutomation in Construction
Volume146
DOIs
Publication statusPublished - Feb 2023

Keywords

  • CatBoost
  • Hydraulic systems
  • Machine learning
  • Transient-based leak detection
  • Water hammer

ASJC Scopus subject areas

  • Control and Systems Engineering
  • Civil and Structural Engineering
  • Building and Construction

Fingerprint

Dive into the research topics of 'Machine learning modeling for spectral transient-based leak detection'. Together they form a unique fingerprint.

Cite this