Tank level sensors on a construction site: what they are good for and how to use them

A diesel tank on a site compound is usually tracked twice: once on a clipboard or in a spreadsheet when fuel goes in or out, and once when someone climbs up with a dipstick. The two figures drift apart, and nobody notices until the tank runs dry on a pour day or a delivery will not fit. A level sensor on the tank does not replace the records. It gives you a second, independent figure every hour, so the drift shows up the same day.

This guide covers what a tank sensor measures, where it earns its keep, how to set one up so its readings can be trusted, and what it cannot tell you.

What a level sensor actually measures

Most tank sensors do not measure gallons. They measure something physical and leave the conversion to software:

• Distance. An ultrasonic or radar sensor on the tank roof measures the gap down to the liquid surface. The emptier the tank, the longer the gap.
• Pressure. A submersible sensor at the bottom measures the weight of liquid above it. More liquid, more pressure.
• Weight. Load cells under a tank, IBC or gas cylinder rack measure mass, and you subtract the empty weight.

To turn distance into gallons you need the tank's shape and size, and the distance from the sensor to the floor of the tank. A vertical cylinder is simple. A horizontal cylinder is not linear: the middle of the tank holds far more per inch than the top or bottom. Odd-shaped or partly buried tanks need the manufacturer's strapping table, which lists the volume at each depth.

That conversion is called calibration, and it is where most bad sensor data comes from. Get the diameter wrong by a few inches and every reading is wrong by the same proportion.

Where sensors pay off on a job

Sensors cost money to buy, fit and keep running, so they belong where a wrong figure costs more than the sensor:

• Fuel tanks and bowsers. Diesel is expensive, it is drawn many times a day, and running out stops equipment. A sensor catches a missed entry, a slow leak or fuel walking off site within the hour.
• Bulk liquids. Admixtures, curing compounds, form release and water treatment chemicals stored in tanks or IBCs, where a count means climbing up or guessing.
• Silos and hoppers. Cement and other bulk solids, usually measured by radar or by weight.
• Gas cylinders. Weighed on a scale rack, because a pressure gauge on a liquefied gas tells you little until it is nearly empty.
• Mobile tanks. A GPS tracker on a fuel trailer shows where it is, which matters more than its level when it can be anywhere on a large site.

Cans, tubes, bags and drums are better served by a label and a scan (see QR and NFC labels for site stock). A sensor on every drum is rarely worth it.

Keep the sensor's figure apart from the records

The most common mistake is letting the sensor overwrite the stock figure. It feels efficient, but it hides exactly what you wanted to see. If someone draws 40 gallons and does not record it, a sensor that silently updates the balance makes the gap disappear.

Keep two figures for each tank:

• Recorded: what the records say should be there, built from every delivery, issue and dispense someone entered.
• Measured: what the sensor, or the last physical count, says is there.

When the two disagree by more than a small tolerance, something happened that nobody wrote down. That difference is the useful information. It should open a question for a person to answer, not quietly change a number.

When you look at software for this, check that a sensor reading is stored as a measured figure beside the records, that a gap opens a review, and that accepting the measured figure adds an adjustment with a reason rather than editing history. See When the count and the records disagree.

An example: the hose fitting that wept

Take a 500 gallon diesel tank with a reorder point of 150 gallons. The foreman records fuelling the machines at 7:20, drawing 30 gallons. The hourly readings track the records within a gallon all morning.

At 1:00 pm the sensor reads 268 gallons. At 2:00 pm it reads 222. Nobody recorded a dispense. The 46 gallon gap is far beyond the tolerance, so a variance opens. The superintendent walks over, finds the dispenser hose fitting weeping into the bund, has it replaced and accepts the measured figure with a note. The adjustment is in the tank's history with who accepted it and why.

Two days later the recorded level falls below 150 gallons, the low stock alert goes up, and the fuel delivery the next morning clears it. Nobody dipped the tank once.

Readings you should not trust

Sensors fail in ordinary ways: a bird on an ultrasonic sensor, condensation, foam on the surface, a dying battery, a gateway that loses power over the weekend. Good software assumes some readings are wrong:

• Keep every raw reading. If the calibration was wrong, you can recalculate history instead of losing it.
• Quarantine the impossible. A negative level, a level above the tank's capacity, a jump that no pump could achieve, or a timestamp in the future should be held for review, not posted.
• Do not post every wiggle. A sensor that reports every few minutes produces noise. Posting only when the level moves past a small band, or once an hour, keeps the history readable.
• Treat silence as unknown. A sensor that stops reporting has told you nothing. Its last figure should age visibly and raise a stale-sensor alert.

A missing sensor signal is not confirmation that a tank is safe, full or empty. Emergency and safety screens should always show when each figure was last confirmed.

Set it up so the readings count

1. Pick the tank and the reason. One fuel tank with a known problem is a better start than ten tanks with no question to answer.
2. Check suitability first. Fuel and many chemicals give off flammable vapour. A battery sensor is not suitable for a hazardous location unless it is certified for that classification. Have a competent person assess the vessel, the substance and the zone before anything is fitted.
3. Survey coverage. LoRaWAN sensors need a gateway within range, and cellular ones need signal where the tank sits. Test on site before you buy in quantity.
4. Record the tank's geometry. Measure the tank, or use the strapping table, and the sensor's mounting height. Have someone approve the calibration before its figures count.
5. Tie the sensor to the container. The reading has to land on the right tank. A sensor moved to another tank without updating its assignment is worse than no sensor.
6. Agree who answers variances. A variance nobody reviews is just another ignored alert. Name the person, usually the superintendent or storekeeper.

What a sensor cannot do

A level sensor tells you how much is in a tank. It does not tell you whether the tank should be there, whether the product has a current safety data sheet, whether it is stored next to something incompatible, or whether the bund will hold a spill. Those stay with your register, your storage rules and the people who inspect the site. Treat sensor data as evidence that helps a person decide, not as the decision.

Start small

Fit one sensor to the tank that causes the most trouble, run it for a month alongside your usual records, and look at the variances it raised. If they caught real losses or saved a run-out, add the next tank. If they did not, you have learned that cheaply. Piloting new technology on a job has a fuller checklist.