Advancing AI-Powered Underground Pipeline Leak Detection Toward Real-World Deployment and Safer Industrial Operations
■ Technology Introduction
▶ In response to the government’s strengthened requirements for underground pipeline
management, AI-based underground pipeline leak detection technologies are
being developed. These technologies integrate intelligent acoustic and vibration
anomaly recognition modules with dynamic satellite-based leak localization systems,
enabling enhanced safety management for both public utility and high-risk pipelines,
while reducing operation and maintenance costs and improving overall maintenance
efficiency.
▶ Through collaboration with a Taiwan Water Corporation, the system has been applied to
underground pipeline networks across Taiwan, achieving an average diagnostic accuracy of
83.07%. The demonstrated annual economic benefit reaches approximately NTD 19 million,
highlighting its strong commercial and operational value.
▶ The Smart City Pipeline Network Leak Detection System, developed in collaboration with
Taiwan Water Company, was awarded the Bronze Medal for the“Project Innovation
Award”at the 2024 World Water Congress & Exhibition(IWA-WWCE). The Smart City Pipeline
Network Leak Detection System, developed in collaboration with Taiwan Water Company,
was awarded the Bronze Medal for the“Project Innovation Award”at the 2024 World Water
Congress & Exhibition(IWA-WWCE)
▶ “Smart Aqua Leak Finder”was honored with the Gold Award at the 2025 Edison Awards.
companies.
■ Global and Domestic Deployment of an AI-Powered Smart Pipeline Leak Detection System
■ Industrial pipeline leak detection solutions have been promoted and deployed in domestic underground petrochemical pipeline networks to support operational management and enhance industrial safety
Remote Intelligent Visual Diagnostic and Early Warning System for Wind Turbine Blade Degradation
■ Application
This system enables remote audio-visual inspection and degradation diagnosis of wind turbine blades at distances up to 200 meters, achieving a recognition accuracy of 90%.
■ Intelligent Acoustic–Vibration Degradation Recognition
▶ Onshore wind farms(Enercon): Successfully identified anomalies in 33 wind turbines.
▶ Offshore wind farms(Taiwan Power): Successfully detected anomalies in 2 wind turbines.
▶ The high similarity of abnormal acoustic featuresdemonstrates the robustness and applicability of
the diagnostic model.
■ The remote audio enhancement module strengthens acoustic features, while the image super-resolution module improves image blurring issues
▶ Dynamic video-based visual recognition for detecting surface corrosion defects.
▶ Micro-displacement analysis to identify internal interlayer delamination.
▶ Intelligent audio-visual–assisted diagnostics enabling comprehensive detection of abnormal
conditions.
▶ Development of an acous t i c metamater ial phase database and a frequency-specific audio
enhancement module to effect ively ampl i fy acoustic signal intensity.
▶ Integration of the NAFNet deblurring model to establish an image super-resolution module,
significantly improving image blurring issues in wind turbine blade inspections.
■ Contact Us
Material and Chemical Research Laboratories
Dept. of Smart O&M and Engine(K900-1)
Yao-Long Tsai
Tel:03-5914178
E-mail:YLtsai@itri.org.tw