Short answer
Detecting fuel theft across distributed diesel tanks means separating abnormal drops from normal consumption at each tank. A fixed litre threshold produces false alarms because a generator tank, a site tank and a vehicle tank consume differently. Reliable detection compares every level change against that specific asset's own consumption behaviour and operating context.
Key takeaways
- A fixed litre threshold is the most common cause of false theft alarms across a mixed tank estate.
- Each asset needs its own consumption baseline: a running generator legitimately loses level, a parked vehicle does not.
- Level accuracy is a property of the installed system, not of the datasheet, calibration and mounting dominate.
- A capacitive probe on RS-485 gives a digital, addressable reading that resists the tampering analogue senders invite.
- Detection is only useful if the alert reaches a named owner with the context needed to act within minutes.
01What makes distributed tanks harder than a single tank?
A single tank is a closed accounting problem. You know what went in, you know what the equipment on that site should have burned, and anything left over is a question worth asking. Distributed tanks break that arithmetic. Ten tanks across four sites are ten different consumption patterns, ten different refuelling schedules and ten different sets of people with legitimate access.
The failure that follows is predictable. Teams apply one rule to all of them, alert when the level drops more than fifty litres in an hour, and the rule is simultaneously too sensitive for the generator that legitimately burns through its tank overnight and too blunt for the standby tank that should never move at all. The alerts that result are mostly noise, and an alert stream that is mostly noise gets muted within a fortnight.
The correct unit of analysis is therefore not the litre. It is the individual asset and what normal looks like for that asset, on that site, at that hour.
02How does level sensing actually work on a diesel tank?
Most fixed-tank monitoring uses a capacitive probe: a vertical element whose measured capacitance changes as diesel rises and falls between its electrodes. The probe is cut or extended to the tank depth and then calibrated against known volumes, because the relationship between height and litres depends entirely on tank geometry, a cylindrical tank lying on its side does not lose litres linearly as the level falls.
The Eurosens Dominator RS is a representative example. Its published specification gives a maximum error of ±1%, an operating range of −40 to +85 °C, a supply range of 10–60 V and IP67 ingress protection, with RS-232, RS-485 and a K-Line service interface conforming to ISO 9141. Under IEC 60529, that IP67 rating means dust-tight and able to withstand temporary immersion.
The interface matters as much as the probe. RS-485 is a differential, multi-drop bus, so several sensors can share one line to a gateway and each returns an addressed digital value rather than a voltage that anything in the loop could influence.
03Why do fixed thresholds produce false alarms?
A threshold expressed in litres per hour assumes every tank has the same right to lose fuel. It does not. A generator running through the night is supposed to empty its tank; a standby tank on the same site is supposed to sit still. Applying one number to both guarantees that one of them is monitored wrongly.
Temperature adds a second layer. Diesel expands and contracts with temperature, which is why the petroleum industry corrects measured volumes to a standard reference temperature under ASTM D1250 and API MPMS Chapter 11.1 rather than treating an observed volume as final. On a site where daytime and night-time tank temperatures differ substantially, an uncorrected reading will show the level moving when no fuel has left the tank.
The practical consequence is that a threshold tight enough to catch real siphoning also fires on thermal contraction, on sloshing in a partly filled tank, and on the sensor settling after a refill. Each is a false positive, and false positives teach a team to ignore the system.
04What does a per-asset consumption baseline look like?
A baseline is simply a description of what this asset normally does, built from its own history rather than from a global rule. For a generator that means the litres-per-hour it consumes while running, which is a function of load, so the baseline is conditional on the machine actually being on. For a site tank it means the shape of a normal working day: draw-offs clustered around shift changes, flat overnight.
Once the baseline exists, the question the system asks changes. Instead of "did the level fall by more than X," it asks "did the level fall in a way this asset does not normally fall." A generator losing forty litres over eight running hours matches its baseline. The same generator losing forty litres in twelve minutes while its run-hours counter is static does not, and that mismatch is the signal worth alerting on.
This is also what makes the difference between a monitoring system and a theft-detection system. Both know the level. Only one of them knows whether that level should have changed.
05How should the alert itself be designed?
Detection that nobody acts on is not detection. An alert has to arrive with enough context for the person receiving it to make a decision without opening three other screens: which asset, which site, how much, over what interval, and what the asset was doing at the time.
Severity should follow that context rather than the raw magnitude. A twenty-litre drop on a sealed standby tank is more interesting than a two-hundred-litre drop on a generator that ran all night, and an alerting system that ranks by litres alone will bury the first under the second.
The last requirement is ownership. Every alert needs a named recipient and an expected response window, because a fuel-theft signal has a short useful life: once the vehicle has left the site, the difference between an alert at two minutes and an alert at two hours is the difference between an intervention and an incident report.
06How do you tell a refill from a theft?
Both events are large, fast level changes, and a system that only watches magnitude will treat them as the same class of anomaly. The obvious separator is direction, a refill goes up, but that alone is not enough, because the interesting case is a partial delivery: the tanker records a full delivery and the tank shows less than was invoiced.
Reading the shape of the change is what separates them. A genuine delivery is a sustained rise over several minutes, ending at a level that matches the delivery note. Siphoning is a fall at a rate the equipment cannot physically consume. A short-delivered load is a rise that stops early, which is invisible to any rule that only looks for falls.
This is why delivery reconciliation belongs in the same system as theft detection. The question is not only whether fuel left, but whether all the fuel paid for arrived. On a distributed estate the same telemetry answers both, and splitting them across two systems is what lets a short delivery go unnoticed for months.
07What about slow, unexplained drops?
Not every loss is dramatic. Fuel drawn a few litres at a time, repeatedly, never trips a threshold designed to catch a siphoning hose, and neither does a seeping fitting. Both show up only as a slow divergence between what the equipment should have consumed and what the tank actually holds.
Catching that requires comparing cumulative consumption against cumulative running time over days rather than examining individual events. If a generator has run two hundred hours this month and its fuel use implies three hundred, the gap is real whether the cause is theft, a leak or an unrecorded transfer. The system does not need to know which; it needs to surface the discrepancy to someone who can go and look.
A leak also carries an environmental and safety dimension that theft does not, which is a reason to distinguish them once a site visit happens. A persistent slow loss on a tank whose equipment is idle is a very different investigation from one on a tank in heavy use.
08What does installing this on an existing tank involve?
The probe is mounted vertically through the top of the tank, cut to a length that reaches close to the bottom without touching it, and sealed at the entry point. The mounting position matters more than most teams expect: a probe placed where fuel sloshes hardest, or where a return line discharges, produces a noisy reading that no amount of software smoothing fully repairs.
Calibration follows installation and cannot be skipped. Because a horizontal cylindrical tank does not lose litres linearly with height, the system needs a mapping between measured height and actual volume, built by recording readings at known fill levels. A probe installed without that mapping reports a confident number that is wrong at every level except the ones it happened to be set against.
Power and wiring are the third constraint. The Eurosens Dominator RS accepts 10–60 V and carries reverse polarity protection with external voltage protection up to 170 V, which matters on sites where the supply comes off the same unstable bus as the equipment being monitored.
09How do you know the system is actually working?
A monitoring system that has raised no alerts is indistinguishable from one that is broken, and the difference matters. The first thing to track is not theft events but data health: is every sensor reporting, at the expected interval, with values in a plausible range? A tank that has gone silent is the most common real failure, and it is silent in exactly the way a quiet tank is.
The second measure is what happens to the alerts that do fire. If most are dismissed without action, the thresholds are wrong and the system is training its users to ignore it. A falling dismissal rate over the first weeks is the clearest sign that the baselines have settled.
The third is reconciliation over a period. Comparing fuel purchased against fuel consumed plus fuel still in tanks, across a month and the whole estate, closes the loop that individual alerts cannot. It is also the number a finance team will ask for, and being able to produce it is often what keeps the programme funded.
10What should a Saudi operator plan for specifically?
Ambient heat is the first consideration, and it acts on the measurement rather than only on the hardware. Diesel expands and contracts with temperature, which is why petroleum volumes are corrected to a standard reference temperature under ASTM D1250 and API MPMS Chapter 11.1. A tank standing in full sun and the same tank before dawn are not reporting the same volume even when no fuel has moved.
The hardware itself has to be specified for that environment rather than assumed to cope. A probe rated to +85 °C and IP67 under IEC 60529 is specified for dust and temporary immersion, which is the relevant combination for an exposed yard rather than a workshop.
Connectivity is the third. Remote sites do not have continuous coverage, so the device has to buffer readings and resend them in order once the link returns. A system that silently drops what it could not send during an outage will show a flat line exactly across the window an investigator most wants to see.
11How ENTEK.AI approaches distributed tank monitoring
ENTEK.AI treats a level reading as one input to an asset record rather than as the finished answer. A tank is an asset with its own device, its own telemetry history and its own alert rules, so the comparison that raises an alert is made against that asset rather than against a platform-wide constant.
Because tanks, vehicles and containers sit on the same map, a drop can be read against what else was on site, such as which vehicle was there at the time, and that context helps separate a refuelling event from a loss. The Fuel Tank solution also flags refuel and drop events automatically and raises Critical alerts below 20% and Low alerts below 35%. Fuel monitoring, fleet tracking and smart-lock container security are modules of one system rather than three integrations.
The practical benefit is narrower than a marketing claim and more useful: a smaller number of alerts, each of which a site supervisor can act on, instead of a continuous stream that gets silenced.
| Behaviour | Fixed litre threshold | Per-asset baseline |
|---|---|---|
| Generator burning fuel overnight | Alerts repeatedly, legitimate consumption | No alert, matches its running baseline |
| Standby tank losing 20 L quietly | No alert, below the threshold | Alerts, this tank should not move |
| Level shift from day/night temperature | Can alert, reading not corrected | Suppressed, expected for the conditions |
| Refill event | Often alerts as an anomaly | Recognised as a refill, not a loss |
| Tuning effort as tanks are added | Re-tuned by hand per tank | Baseline forms from each asset’s own history |
Figures and sources
- Eurosens Dominator RS published maximum measurement error: ±1% (Source: Wialon sensor catalogue, Eurosens Dominator RS, as of 2026-08-30)
- Eurosens Dominator RS operating temperature range: −40 to +85 °C (Source: Wialon sensor catalogue, Eurosens Dominator RS, as of 2026-08-30)
- Eurosens Dominator RS supply voltage range: 10–60 V (Source: Wialon sensor catalogue, Eurosens Dominator RS, as of 2026-08-30)
- Eurosens Dominator RS ingress protection rating: IP67 (IP54 on the Dominator RS-I variant) (Source: Wialon sensor catalogue, Eurosens Dominator RS, as of 2026-08-30)
- Eurosens Dominator RS external power voltage protection: up to 170 V, with reverse polarity protection (Source: Wialon sensor catalogue, Eurosens Dominator RS, as of 2026-08-30)
- Eurosens Dominator RS digital interfaces: RS-232, RS-485, K-Line (Source: Wialon sensor catalogue, Eurosens Dominator RS, as of 2026-08-30)
- Standard governing the Dominator K-Line service interface: ISO 9141 (Source: Wialon sensor catalogue, Eurosens Dominator RS, as of 2026-08-30)
- IP67, second digit: protection against temporary immersion: immersion up to 1 m for 30 minutes (Source: IEC 60529, Degrees of protection provided by enclosures (IP Code), as of 2026-08-30)
- IP67, first digit: protection against solid foreign objects: dust-tight (no ingress of dust) (Source: IEC 60529, Degrees of protection provided by enclosures (IP Code), as of 2026-08-30)
- Petroleum volumes are corrected to a standard reference temperature under: ASTM D1250 / API MPMS Chapter 11.1 (Source: ASTM D1250, Temperature and Pressure Volume Correction Factors, as of 2026-08-30)
- RS-485 signalling type, which is what allows multi-drop sensor wiring: differential, multi-drop balanced bus (TIA/EIA-485) (Source: TIA-485-A, Electrical Characteristics of Generators and Receivers for Balanced Digital Multipoint Systems, as of 2026-08-30)
- SOJI LIGO Pro fuel level sensor, listed on the ENTEK marketplace: available as an alternative capacitive probe (Source: ENTEK Marketplace product listing, as of 2026-08-30)
Questions
Can fuel level sensors alone prove theft?
No. A sensor establishes that fuel left the tank and when. Proving it was theft rather than consumption, a refill correction or a measurement artefact requires the surrounding context, whether the equipment was running, who had access, and whether a vehicle was on site at the time.
How accurate does a fuel level sensor need to be for theft detection?
Consistency matters more than the headline figure. A probe with a published maximum error of ±1%, like the Eurosens Dominator RS, is well within what theft detection needs, provided it is calibrated to the tank and mounted so the reading is stable. An accurate sensor installed badly is worse than a modest one installed properly.
Does temperature really change the reading?
Yes. Diesel changes volume with temperature, which is why the petroleum industry corrects measured volumes to a standard reference temperature under ASTM D1250 and API MPMS Chapter 11.1. On sites with a wide day-to-night swing, an uncorrected reading can move without any fuel leaving the tank.
Where should monitoring start on a large estate?
Start with the tanks whose normal behaviour is easiest to describe, standby and storage tanks that should not move outside scheduled refills. They give clean baselines quickly and build confidence in the alert stream before the harder, continuously consuming assets are added.