Water Flow Detection Device Based on Sound Data Analysis and Machine Learning to Detect Water Leakage

2025-01-19

AI Summary

Detecting small-scale leaks (≥100 mL/min) in urban water distribution networks remains challenging due to the trade-off between cost and detection accuracy. To address this, we propose a novel leak detection paradigm integrating passive mechanical acoustic amplification with lightweight deep learning. A passive mechanical acoustic amplifier enhances weak leak-induced acoustic signals; these are then captured via low-cost audio hardware, digitized, and transformed into robust time-frequency domain features. An optimized lightweight deep neural network processes these features to classify leak presence. The system eliminates the need for complex wiring or power-hungry sensors, supports rapid adaptation across pipe diameters, and achieves >95% classification accuracy for 100 mL/min leaks in experimental validation. This work introduces the first “physics-based amplification + lightweight AI” co-design architecture for leak detection, delivering a scalable, cost-effective, and high-performance sensing unit for smart water infrastructure.

Technology Category

• Machine Learning: Hardware-aware ML
• Search and Optimization: Learning to Search
• Natural Language Processing: Learning & Optimization for NLP

Application Category

• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

In this paper, we introduce a novel mechanism that uses machine learning techniques to detect water leaks in pipes. The proposed simple and low-cost mechanism is designed that can be easily installed on building pipes with various sizes. The system works based on gathering and amplifying water flow signals using a mechanical sound amplifier. Then sounds are recorded and converted to digital signals in order to be analyzed. After feature extraction and selection, deep neural networks are used to discriminate between with and without leak pipes. The experimental results show that this device can detect at least 100 milliliters per minute (mL/min) of water flow in a pipe so that it can be used as a core of a water leakage detection system.

Problem

• Acoustic Leak Detection
• Supervised Machine Learning
• Cost-effective Universal Device

Innovation

• Acoustic detection
• Supercritical computing
• Water leak detection