Pipelines are one of the most reliable and environmentally benign ways of transporting liquid and gas commodities. However, pipelines are complex systems, with assets of varying ages, diameters, thicknesses, and materials. Furthermore, pipelines are tasked with carrying batches of product, each having different compositions, viscosities, and other properties. This variability makes pipeline evaluation and assessment challenging. Regardless of substantial investments in pipelines detection systems, as well as stringent regulations to the energy transportation industry, the annual average of pipeline leak incidents in the last ten years is 125 incidents (Transportation Safety Board of Canada, 2019). The characteristics of leak detection systems are widely different, resulting in limitations of practicality and reliability. For instance, deploying sensor suites can be prohibitively expensive because of their retrofitting costs, restricting their implementation to local high-risk areas or new pipelines. On the other hand, inaccurate and insufficient sensor data hinders accurate modeling complex of real-world pipeline systems, leading to uncertainties which often result in false-positive and false-negative alarms.

The present research aims to optimize the sensor suites of pipeline systems to determine the state of pipelines in situ. The proposal of this research is the design, development, and preliminary characterization of a testbed for laboratory-scale comparative assessment of leak detection methods. Additionally, the proposed testbed generates data that potentially can be used in the design of computational pipeline monitoring systems based on artificial intelligence.

A unique laboratory-scale apparatus which combines a wide variety of leak detection methods employed by the energy pipeline industry has been built and presented to address the shortcomings in existing leak detection techniques. Technologies tested are dielectric permittivity probes, temperature probes, thermal imaging, pipe vibrations, pressure profiles, and pressure transients. They were evaluated with the technological challenges proper of real-world pipeline systems.

Experimental results are varied because the relationship between factors (input) and responding variables is complex and multifactorial. However, experimental analysis elucidated some of the complexities that leak detection methods face. For instance, soil permeability substantially influences how the oil releases from a leak because it defines the leaked-liquids and soil relationship and consequentially the chances of the commodities reaching the probes. Soil moisture content also influences the probes’ performance. For example, the dielectric permittivity probes perform inefficiently in saturated soils because the soil cannot absorb any more liquids. Thermal imaging is insensitive to soil permeability because its technology relies on rendering infrared radiation. Pressure transients are useful to detect leaks as proven through experiments. These findings help to select the most suitable leak detection method to improve the reliability of leak detection systems. More experimental conditions are suggested to evaluate additional uncontrolled variables that affect leak detection systems.