With the global expansion of unconventional gas plays and increased consumer demand, greater pipeline capacity to transport natural gas to market is a necessity. Old pipelines are being repurposed and new pipelines are being built. Regulators and watchdog groups are increasingly scrutinizing the safety and risk of these developments.

Leak detection, particularly computational leak detection, is a crucial component of a pipeliner's risk mitigation. The major concerns in a leak detection system include robustness, leak sensitivity, and location accuracy.

  • Robustness is primarily determined by the maximum variation expected in the system during normal running as a consequence of changing operations and instrument noise.

  • Leak sensitivity is typically determined by the scale of flow and pressure deviations observed in a leak scenario compared to normal operation.

  • Location accuracy depends on methodology.

In this paper we examine the effectiveness of gas pipeline leak detection by simulating several typical configurations. We estimate the maximum variation observed in normal running. We evaluate the pressure and flow deviations that occur when various leaks are imposed. Finally the accuracy of leak location from a transient computational pipeline model is examined.

The results will assist pipeliners, regulators, and concerned parties to develop realistic expectations about the efficacy of meter-driven leak detection systems in natural gas pipelines.


We focus here on the usefulness of computational leak detection and location using RTTM with the aim of demonstrating high performance on gas transmissions lines. For a survey of other leak detection methods, see [3]. [10] is a recent introduction to the hydraulic effect of leaks on a pipeline. The basic method for leak location used here is unchanged since [1]. For a discussion of selecting a leak detection method appropriate to your situation, see [11].

This content is only available via PDF.
You can access this article if you purchase or spend a download.