Pipe Leak Detection System Employs Machine Learning to Limit Errors

San Antonio—Even as Donald Trump is executing orders to roll back environmental regulations, numerous technology researchers are developing new systems that may aid oil and gas businesses in better detecting leaks that lead to emissions—including one on the west side of San Antonio.

The Southwest Research Institute (SWRI) is developing a machine learning system that can quickly process image data from various cameras, which detects leaks in pipes that deliver various types of environmentally dangerous materials, from methane to oil. SWRI received a nearly $800,000 U.S. Department of Energy grant in October to develop the system for methane leaks, which account for a smaller percentage of greenhouse emissions than gasses like carbon dioxide, but are considered more threatening because methane absorbs heat more effectively, the institute says.

“It’s mitigation of emissions through early detection,” says Maria Araujo, manager of R&D in SWRI’s high reliability systems section. “You could have a valve or compressor that has a leak. You’re not going to know because methane has no smell. It has to be a specialized system.”

SWRI’s device would use thermal or infrared cameras equipped with infrared sensors to watch for the leaks—cameras that can see what the naked eye can’t. The organization may target the thermal cameras because, at about $10,000, they are about a fifth of the cost of the infrared cameras typically used to detect methane, Araujo says.

One of the primary goals of SWRI’s machine learning system is to remove the human element—cameras currently used need to have someone monitor the feeds to keep an eye out for the leaks. SWRI’s device would be autonomous, automatically alerting the user if something is wrong.

The data from the images recorded by a SWRI camera are processed by algorithms that SWRI is developing, which feed into a neural network that’s being trained to determine whether there’s a leak or not. SWRI plans to use Nvidia Tegra computing board to process everything.

SWRI isn’t alone in this kind of research. IBM (NYSE: IBM) is developing a small sensor chip with a laser that can detect when methane molecules pass through the air above the chip, according to Scientific American. The chips, which would be placed at various intervals along a pipeline like SWRI’s technology, would alert the owner of the pipeline about the leak, according to the magazine.

IBM told Scientific American it is targeting a price of $200 per chip—seemingly a lower cost than the SWRI option, though SWRI’s Araujo notes this system would require many chips to be placed throughout the pipeline while SWRI’s would only need one or two. IBM’s tool doesn’t yet have a machine learning component, though the company says in the article that it plans to incorporate one.

Araujo argues that the IBM system will be, if anything, complementary and she says it appears to lack certain benefits of SWRI’s technology. SWRI’s machine learning algorithms can prevent false alarms caused by changes in temperature or moisture near a pipeline, she says. The system could also be used in a drone, immediately processing images the drone records as it flies over. The research institute is feeding its machine learning algorithms inputs in various weather conditions other environmental factors to help it be able to adapt to any given environment or situation it might face.

“The more data, and the more variety of data, the better the algorithm will perform,” Araujo says.

SWRI’s technology could be used for various other efforts, too, such as detecting pipes leaking other liquids and oceanic oil spills.

David Holley is Xconomy's national correspondent based in Austin, TX. You can reach him at dholley@xconomy.com Follow @xconholley

Trending on Xconomy