Data-driven leakage diagnosis methods across pipeline and energy transportation system

A review recently published in the Journal of Pipeline Science and Engineering presents leakage diagnosis methods ranging from single pipelines to Energy Transmission Systems (ETS) — marking the first systematic attempt to connect single pipeline analysis with ETS. Its focus lies in summarizing leakage detection techniques under complex environmental conditions.

In leakage detection, despite theoretical advances, model-based methods face challenges in pipeline applications due to modeling inaccuracies and high computational costs. In contrast, data-driven approaches, especially deep learning models, show good potential by virtue of their strong capabilities in nonlinear mapping and spatiotemporal feature extraction, effectively addressing key ETS challenges such as highly coupled signals, background noise, and false alarms under multiple operating conditions.

Current research to improve detection in complex scenarios mainly proceeds in two directions: advanced signal processing and multi-modal fusion to enhance signal quality, SNR, and feature discriminability; and generative networks and transfer learning to solve few-shot or zero-shot learning problems for reliable detection with insufficient samples.

For leakage localization, the TDOA method remains fundamental due to its maturity, with research focusing on improving time-delay estimation via advanced signal processing and cross-correlation (CC) algorithms. To address weak signal attenuation in long-distance pipelines, novel localization methods based on attenuation model matching and adaptive dynamic programming (ADP) have been developed, redefining localization as a model parameter optimization problem. It paves a new avenue for the accurate localization of minor leakages in complex pipeline environments.

Notably, while data-driven methods have made significant progress, they still have limitations, including inherent constraints of pure data-driven models, weak self-learning ability, field deployment difficulties, and preventive maintenance issues.

Future research directions include data-physics fusion approaches for pipeline leakage diagnosis, self-learning pipeline leakage diagnosis method, large scale model-based leakage diagnostics, lightweight deployment of leakage detection models, locating multiple-point leakages in pipelines, pipeline leakage warning mechanism.

Contact author: 

Dazhong Ma, School of Information Science and Engineering, Northeastern University, Shenyang, China, madazhong@ise.neu.edu.cn

Funder: 

This work was supported in part by the National Natural Science Foundation of China under Grant U22A20221; in part by the Natural Science Foundation of Liaoning Province under Grant 2022-KF-11-02; in part by the GuangdongBasic and Applied Basic Research Foundation under Grant 2023A1515240040.

Conflict of interest:

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

See the article:

Dai Zhenxu, Wang Tianbiao, Hu Xuguang, Ma Dazhong, A review of data-driven leakage diagnosis methods across pipeline and energy transportation system, Journal of Pipeline Science and Engineering, Available online 6 February 2026, 100459, https://doi.org/10.1016/j.jpse.2026.100459.