BRENNEN

Industries / Industrial equipment and services / Industrial services

Leak detection (LDAR)

Leak detection and repair (LDAR) is the work of finding gas and vapor escaping from the valves, flanges, pump seals and other equipment of refineries, chemical plants and oil and gas sites, fixing the leaks, checking that the fixes hold and reporting the results.

1978

In June 1978 the US Environmental Protection Agency (EPA) issued its guideline for controlling leaks from refinery equipment, which told refiners to check pumps, valves and seals with a portable vapor detector, to treat a reading above 10,000 parts per million (ppm) as a leak and to repair it within 15 days.

nepis.epa.gov

41

Milestones achieved across 6 domains.

9

To watch for: announced, not yet confirmed outside.

12

Open deliverables. Nobody has shipped them.

In one minute

Where it started

By EPA's 1978 account, a joint project about twenty years earlier had measured the emissions of the refineries in the Los Angeles air basin, and its emission factors were still in use. EPA's guideline of that year set the method most plants still follow: walk past every pump, valve and seal with a portable detector, call a high reading a leak and fix it within days, because a small share of the parts gave about 90% of the vapor. A very large refinery, the guideline noted, has more than 100,000 places that can leak.

Where it stands

Leak surveys are now required at refineries, chemical plants and oil and gas sites in the United States, Canada, the European Union and China, and the handheld probe has been joined by infrared gas cameras, fixed sensors, aircraft, drones and satellites. Blind tests show how much each one misses: a gas camera at 6 m saw a leak of about 20 grams an hour half the time, aircraft reliably measured leaks above 10 kg an hour, and the free Sentinel-2 satellites see only leaks of about 3 tonnes an hour or more. Rules are moving both ways: Canada will set repair deadlines by leak size from 2028, while EPA has put back its 2024 deadlines and proposed to end its Super Emitter Program and most greenhouse gas reporting.

What is open

Most testing so far has scored instruments, and far less has scored the people, robots, software and records around them. No public test has had several people measure the same hidden leak, scored a robot's rounds against a person's, or set AI against certified surveyors on the same video, and the leak reports that Colorado and Alberta publish have not been gathered into tables that show whether inspections and repairs work. The open jobs here need public files, bench rigs, non-flammable test gas or test sites run by others, not live plants.

Domains

AchievedWatch forOpen

Open in leak detection (ldar)

3 of 12 open deliverables

Open · Handheld surveys

Ten users, one leak

A public reproducibility test of high flow sampling, in which ten trained users measure the same hidden leaks on a mock equipment skid with the same instruments, so that the share of the error that comes from the person, not the instrument, is known.

Built with what exists

A commercial high flow sampler such as the SEMTECH HI-FLOW 2 or the AddGlobe GFM, or CSU's open-source design from the CARB report; cylinders of a non-flammable mixture of 2.5% methane in nitrogen; mass flow controllers, such as Alicat's, to set each leak rate; a skid of unpressurized scrap valves, flanges and threaded fittings with hidden release tubes; repeatability and reproducibility statistics as set out in ISO 5725-2.

Credit · unclaimed

Open · Gas imaging cameras

AI and surveyors, same clips

A public reading test in which certified OGI surveyors and AI leak detectors watch the same blind video clips, with and without leaks of known size, so that people and software can be compared directly.

Built with what exists

GasVid, the Stanford team's labeled METEC video set, if its authors share it, or new clips recorded at a controlled-release test site; the published GasNet and VideoGasNet models and open plume-segmentation code such as Gasformer on GitHub; a web page that plays clips in random order and logs each answer; ten surveyors trained to Appendix K.

Credit · unclaimed

Open · Fixed sensors

Find the units behind peaks

A public analysis, refinery by refinery, that takes every two-week benzene result from every fence-line tube, pairs it with the hourly winds over the same two weeks, and estimates which part of the refinery the benzene most likely came from. For refineries that exceeded the action level, it sets that estimate beside the cause the refinery gave, where the cause is public.

Built with what exists

Monitor-level results from EPA's WebFIRE database and Benzene Fenceline Monitoring Dashboard; hourly winds from NOAA's Integrated Surface Database for the nearest airport; Python with pandas for wind-sector weighting; aerial photographs from the USDA's National Agriculture Imagery Program to mark process units and tank farms; pdfplumber to pull stated causes from published consent decrees and state reports.

Credit · unclaimed

Where it sits