Industries / Energy, water and waste / Fuel jobbers and wholesale fuel
Tank monitoring
Tank monitoring is the work of knowing how much fuel is in each customer's tank, so that the truck arrives before the tank runs dry and losses show up early.
4
Milestones achieved, from 1993 to 2018.
2
To watch for: announced, not yet confirmed outside.
2
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
1993
Tank gauges pass outside tests
The US Environmental Protection Agency's (EPA) 1988 underground tank rule required monthly leak checks, phased in through 1993. An automatic tank gauge (ATG), a probe in the tank wired to a console, qualifies if it can detect a leak of 0.2 gallons per hour (gph) with a 95% probability of detection and a 5% probability of false alarm. An EPA Region 7 reference manual lists Veeder-Root's TLS-250 gauge as evaluated by a firm it names as MRI on 14 May 1993 and certified at 0.2 gph with 99% detection and 0.1% false alarms. The same console prints each tank's inventory and its deliveries, the readings a supplier needs to plan the next load.
federalregister.govlaw.cornell.edulaw.cornell.educti.uconn.edu
Credit
US EPA; Veeder-Root
1995
Leaks found in sales records
Statistical inventory reconciliation (SIR) finds leaks in the paperwork: a vendor runs statistical software over daily dipstick readings, deliveries and sales. EPA's 1995 booklet says a month of good data can find a loss of about 145 gallons a month, 95 times out of 100, using only the tank stick and tank chart most sites already have; vendors charge a monthly fee per tank. Each method must pass an evaluation, and a work group of state and federal tank regulators, the NWGLDE, keeps the list of methods that have.
LessonEPA warns that a FAIL may mean miscalibrated dispensers, wrongly metered deliveries or stolen fuel rather than a leak.
Credit
US EPA Office of Underground Storage Tanks; National Work Group on Leak Detection Evaluations
c. 2010
Monitors set the delivery day
Ed and Kelly Burke of Dennis K. Burke Inc., a US fuel distributor, wrote in August 2025 that the firm had monitored customers' fuel, lubricant and diesel exhaust fluid tanks remotely for over 15 years. Most tanks report twice a day, and dispatchers check them each morning and night. By the Burkes' figures, deliveries on a fixed calendar typically topped up tanks at about 50% of capacity, while monitored tanks average fills closer to 75%, with fewer trips and less fuel left on the truck.
Credit
Dennis K. Burke Inc
2018
Tank gauges tested every year
When EPA revised its underground tank rule in 2015, it reported that leak detection equipment was finding only about 50% of the releases it was designed to detect, largely through poor operation and maintenance, and that overfills caused about 15% of releases. From 13 October 2018 US owners must test their gauges' alarms, settings, backup batteries and probes every year and inspect overfill prevention every three years. The rule also named continuous in-tank leak detection, which combines the gauge's readings with dispenser meters, as a method, and allowed only SIR that reports a calculated leak rate.
Lessonequipment installed to meet the 1993 deadline found leaks only when someone kept checking that it worked.
Credit
US EPA
Watch for
Fewer deliveries per tank
FoxInsights GmbH, which sells tank sensors and delivery-planning software, says distributors using its system raised tank utilisation from about 53% to 70%, which it equates to around one fewer delivery per tank per year. It does not publish the data behind the figures.
Achieved whena distributor's delivery records, checked by an outside party, show the change against comparable tanks without sensors.
Claimed by
FoxInsights, April 2026
Leaks flagged as data arrive
Ruimin Chu and colleagues at RMIT University and the University of Adelaide report a method that checks a tank's fuel variance (the gap between measured and expected volume) every 30 minutes instead of waiting weeks for a statistical reconciliation. On 160 Australian tank records from 2019 to 2022 with simulated 0.2 gph leaks added, it scored an F2 of up to 0.61, a measure that weights catching leaks above avoiding false alarms. One author works for Titan Cloud Software, and the data are available only on request.
Achieved whenan outside evaluator tests it under EPA's procedure and the NWGLDE lists it, or another group reproduces the result on its own data.
Claimed by
Chu and colleagues, July 2025
Open
01
Test the tank monitors
A side-by-side test of at least five commercial remote tank monitors (ultrasonic, radar and pressure types) on water-filled tanks of common shapes, each compared with a reference reading from a scale under the tank.
The briefCredit · unclaimed02
An open leak-test set
A public set of tank inventory records, simulated and, where an owner agrees, real, with leaks of known rate added to a hidden half, plus a scoring script that applies the US standard of 0.2 gph or 150 gallons in 30 days at 95% detection and 5% false alarms.
The briefCredit · unclaimed
Open 01 · the brief
Test the tank monitors
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A side-by-side test of at least five commercial remote tank monitors (ultrasonic, radar and pressure types) on water-filled tanks of common shapes, each compared with a reference reading from a scale under the tank.
Why it is still open
Makers publish their own accuracy figures, such as ±0.78 inch between 4.7 and 118 inches of level for Tecalemit's TecSonic ultrasonic sensor, and radar makers say vapour in a tank throws off ultrasonic readings. Distributors set delivery days from these readings. The one independent comparison found is from 2011 and covered propane: Battelle tested 13 remote monitoring systems from nine makers for the Propane Education & Research Council, in a laboratory and on an outdoor propane tank, with each transmitter reading the tank's own gauge. This session found no independent test of the ultrasonic and radar monitors fitted to diesel and heating oil tanks.
Built with what exists
Monitors bought at retail; a 275-gallon steel tank or a plastic tote filled with water; a load-cell platform as the reference; a Raspberry Pi to log readings; Python.
First step
Mount two monitors of different types on one water tank and log them against a scale for a week.
Done looks like
At least five monitors on two tank shapes for 30 days, with water pumped out daily to mimic use; error reported for each tenth of the tank's height and at the hottest and coldest hours; all readings published. Water stands in for fuel only for level measurement, so pressure sensors are set for water's density and vapour effects are left untested.
Open 02 · the brief
An open leak-test set
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A public set of tank inventory records, simulated and, where an owner agrees, real, with leaks of known rate added to a hidden half, plus a scoring script that applies the US standard of 0.2 gph or 150 gallons in 30 days at 95% detection and 5% false alarms.
Why it is still open
Chu and colleagues tested their method on 160 real tank records with simulated leaks, but their data are available only on request. The rule sets the bar, and each SIR or gauge method is evaluated on its own, with no shared public data. This session found no public dataset of fuel tank inventory records with known leaks.
Built with what exists
A Python simulator with tank charts, meter error and temperature swings; EPA's criteria in 40 CFR 280.43; open change-point tools such as the ruptures library; simple baselines such as a 30-day variance test and CUSUM.
First step
Simulate one 10,000-gallon tank for 90 days with daily sales and two deliveries a week, add a 0.2 gph leak on day 45, and see whether a 30-day variance test flags it.
Done looks like
At least 1,000 simulated tank-months with leaks of 0.05, 0.1 and 0.2 gph; detection and false-alarm rates published for at least three methods; a second group reproduces the scores.