Industries / Industrial equipment and services / Industrial services / Leak detection
Handheld surveys
Handheld surveys are rounds in which a technician, or now a robot, carries a portable instrument past each valve, flange, pump seal and fitting to find gas escaping, and sometimes to measure how fast it escapes.
7
Milestones achieved, from 1980 to 2025.
1
To watch for: announced, not yet confirmed outside.
2
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
1980
Screened first, then bagged
In the late 1970s the US Environmental Protection Agency (EPA) ran the Petroleum Refinery Assessment Study, in which Radian Corporation collected equipment leak data at 13 refineries. Crews first screened each component with a portable analyzer that reads the hydrocarbon concentration in the air in parts per million by volume (ppmv). Most components that read over 200 ppmv were then bagged: enclosed in a bag so that the gas leaking into it could be collected and the leak rate measured in mass per hour. The only process conditions that tracked the leak rates in a statistically significant way were whether the line carried gas, light liquid or heavy liquid, and how volatile the liquid was, so EPA's refinery leak estimates were split by type of service.
Credit
Radian Corporation for EPA, reports EPA-600/2-80-075c and 075d (April and July 1980), as summarized in EPA's 1995 equipment leak protocol
1996
A hose that catches everything
A high flow sampler draws air around a leaking part through a hose fast enough to catch all the escaping gas, measures the air flow and the methane in it, and multiplies the two to give a leak rate, without bagging the part. US patent 5,563,335, filed on 28 February 1995 and granted on 8 October 1996 to the Gas Research Institute (GRI), with Harry M. Howard as inventor, adds a second hose that samples air on the far side of the part so that gas from nearby leaks can be subtracted. Touché Howard says he developed the original Indaco high flow sampler in the early 1990s and that Bacharach's commercial Hi-Flow Sampler was based on it. A 2022 report for the California Air Resources Board (CARB) calls high flow sampling the predominant method for measuring single emission points.
Credit
Gas Research Institute; Indaco; Bacharach
2015
A sensor that failed quietly
The Bacharach Hi-Flow Sampler reads low concentrations with one sensor and high ones with another, and Touché Howard showed in 2015 that it can fail to switch to the high-range sensor and then report a large leak as a small one. He argued that a University of Texas campaign to measure methane at US gas production sites, used to improve the national inventory, had underestimated emissions this way: tracer measurements at sites with low-methane gas read three to five times higher than the sampler, and over 98% of the study's own inventory may have been affected. At an EPA workshop in 2017 he gave the example of a compressor valve leaking 20 standard cubic feet a minute (scfm) that a failed sampler reported as 0.2 scfm, 100 times too low, and said Bacharach had confirmed the failure in a 2015 revision of its manual.
LessonWhen a leak instrument goes wrong it usually reads low, and the largest leaks are the most affected, so measurement programs need independent checks.
Credit
Touché Howard, in Energy Science & Engineering, August 2015
2017
China's refineries start surveying
China's emission standard for petroleum refining, GB 31570-2015, required every existing refinery to run a leak detection and repair program by 1 July 2017, using methods set out in the technical guideline HJ 733-2014. Researchers at the Guangdong Provincial Academy of Environmental Science and South China University of Technology followed one Guangdong refinery's quarterly surveys with a flame ionization detector, a portable analyzer that burns the sampled air in a small hydrogen flame and counts the charged particles the vapor makes. Before the program, 0.67% of the refinery's sealing points were leaking under Guangdong's leak definition, yet they gave 54.9% of its equipment-leak emissions of volatile organic compounds (VOCs); the long-running program cut those emissions by about 40.5%. Across Guangdong's refineries they estimated that the programs cut 129.0 to 160.5 tonnes of VOCs a year from 2018 to 2022, though emissions began to rise again after 2020 as the refineries processed more crude.
LessonThe authors asked for strict checks on readings near the leak threshold, where a small error decides whether a part counts as leaking.
Credit
Xi-Bo Sun, Yong-Bo Zhang, Xiao-Ming Liang, Dai-Qi Ye and colleagues, in Petroleum Science, March 2026; China's environment ministry, for the standard
2022
Successors tested in the field
By 2022 the Bacharach sampler was no longer made. Colorado State University (CSU) and SLR International, working for CARB, published an open design for a high flow sampler built around an off-the-shelf single-board computer with swappable gas and flow sensors, and tested it beside the old Bacharach unit and three pre-production instruments from other makers at CSU's Methane Emissions Technology Evaluation Center (METEC), an outdoor site built like a gas field facility with controlled releases. Measurements of releases from realistic equipment showed substantially more variability and error than when the gas was fed straight into the instruments, which the team put down to user practices and site conditions, and it advised that these uncertainties be measured and best practices written.
LessonAn instrument that is accurate on the bench can still be far off in the field, because the way it is used adds error.
Credit
Daniel Zimmerle, Timothy Vaughn and colleagues at CSU, with SLR International, for CARB
2025
Survey tools tested blind
CSU's METEC team tested 12 survey methods under a single-blind protocol between 2021 and 2023, meaning the teams being tested did not know where or how large the releases were: 4 handheld infrared gas cameras, 4 newer handheld systems and 4 surveys by car or drone. The handheld cameras did as well or better than the other groups on the smallest leak found 90% of the time, on missed and false detections, and on pointing to the right piece of equipment. The newer handheld systems matched them except for far more false detections, and the car and drone surveys were faster. Two of the three methods tested a second time improved.
LessonThe faster methods trade away small-leak detection, so the authors suggest using them for a first sweep followed by handheld checks.
Credit
Chiemezie Ilonze, Rachel Day, Dan Zimmerle and colleagues at CSU, in a 2025 preprint and Ilonze's 2025 CSU dissertation
2025
New handheld tools approved
Under its 2024 oil and gas rule (subpart OOOOb), EPA began approving other ways of doing the periodic leak survey. On 14 January 2025 it approved the Xplorobot Laser Gas Imager, a handheld tunable diode laser that reads methane along its beam at up to 50 m, in parts per million times meters (ppm-m); Physical Sciences Inc. and Heath Consultants had put the same kind of laser into a handheld detector for gas utility surveys by 2004. The Xplorobot method flags readings of 500 ppm-m or more, logs each component it inspects by computer vision, and found 156 grams an hour (g/h) of methane 90% of the time in blind testing at METEC, by the method document. On 17 November 2026 EPA approved the Distran Ultra Pro X, a handheld camera with 124 microphones that maps the ultrasound made by gas escaping under pressure, for screening at 1 kilogram an hour (kg/h), on pressurized components only.
Credit
EPA's Office of Air and Radiation; Exploration Robotics (Xplorobot); Distran of Zurich
Watch for
Legged robots on leak rounds
ANYbotics of Zurich says its four-legged ANYmal robot, fitted with gas detectors and a 360-degree acoustic imager, can find the source of a leak, estimate the loss and alert staff. The announcement gives no count of leaks found or missed.
Achieved whenan operator or a test center publishes the leaks the robot found and missed against a Method 21 or camera survey of the same equipment.
Claimed by
ANYbotics, June 2025
Open
01
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.
The briefCredit · unclaimed02
Robot rounds, scored blind
A blind tally of a ground robot doing leak rounds: a legged or wheeled robot carrying a methane sensor follows a fixed route through a mock equipment pad where small releases of a non-flammable methane mixture are hidden at random points, and its finds and false alarms are scored beside a person walking the same route with a handheld laser methane detector.
The briefCredit · unclaimed
Open 01 · the brief
Ten users, one leak
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
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.
Why it is still open
CSU's 2022 report for CARB found much more error on realistic equipment than when gas was fed straight into the instruments, blamed user practices and conditions, and advised that these uncertainties be characterized. CSU's current test of commercial samplers, described on a METEC page posted in May 2026, runs every device under the same conditions; it tests instruments, not users. METRONOME, a European metrology project whose summary was issued in July 2026, plans uncertainty procedures and a proficiency testing scheme for measuring leaks at single components, in support of standards for the EU methane regulation, but has published no results. Searches in November 2026 found instrument studies and Howard's accounts of operator mistakes, but no test in which several users measured the same leak.
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.
First step
Fit one release tube inside a scrap flange, set its rate with a mass flow controller, and have two people measure it 10 times each.
Done looks like
600 measurements (10 users, 4 kinds of component, 5 hidden rates from 5 to 50 g/h of methane, 3 repeats each); the spread between users and within each user as a percentage of the true rate; which practices, such as bag fit, hose distance and wind, explain the error; all data published.
Open 02 · the brief
Robot rounds, scored blind
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A blind tally of a ground robot doing leak rounds: a legged or wheeled robot carrying a methane sensor follows a fixed route through a mock equipment pad where small releases of a non-flammable methane mixture are hidden at random points, and its finds and false alarms are scored beside a person walking the same route with a handheld laser methane detector.
Why it is still open
ANYbotics' announcements of June 2025 and October 2026 describe robots on leak rounds but give no leaks found or missed. In the final of TotalEnergies' ARGOS Challenge at Lacq, France, in 2017, robots had to find a simulated ultrasonic gas leak among other tasks, and a jury judged the teams; TotalEnergies' account gives no tally of leaks found and missed. CSU's blind tests of survey methods covered handheld cameras and instruments, cars and drones, not robots. Searches in November 2026 found product announcements and patents, but no blind controlled-release tally for a ground robot.
Built with what exists
A Unitree Go2 legged robot or an AgileX Scout wheeled base; ROS 2 with the Nav2 navigation software for waypoint rounds; a handheld tunable diode laser methane detector mounted on the robot; cylinders of 2.5% methane in nitrogen with mass flow controllers; a phone camera on a pole to film each round.
First step
Build four release points from scrap fittings and drive the robot past them by remote control with the detector logging.
Done looks like
50 rounds of a 20-point pad, each with 0 to 3 releases of 5 to 50 g/h of methane placed by someone outside the robot team; for robot and person, the share of releases found within 1 m of the source, false alarms per round and minutes per round; video and logs published.