Inside a tank there is no satellite signal, no ambient light and nowhere to land. What a drone removes from that job is real and substantial. What it does not remove is the reason most tank inspections exist.
Flying Where There Is Nothing to Navigate By
Outdoor flight rests on satellite positioning, a magnetometer that finds north, and a horizon. Inside a steel tank, hold, caisson or ballast space, all three are gone. The structure blocks satellite signal entirely and the steel distorts the magnetic field, so heading reference is unreliable even where a compass reads.
What remains is onboard sensing. The aircraft builds its own position estimate from visual and inertial odometry, sometimes with ranging sensors, which means it is tracking its movement relative to surfaces it can see rather than to a fixed external reference. That works, and it works well enough for useful inspection, but it carries two consequences worth stating.
The first is drift. An estimate built from accumulated movement accumulates error, and in a large featureless compartment, where every stretch of plate looks like every other, there is little for the system to correct against. The second follows from it: position reported by the aircraft is referenced to its own estimate. Research on aerial contact inspection has reported centimetre-level position accuracy, while noting explicitly that accuracy was measured against the vehicle’s own odometry rather than an external ground truth.
The aircraft knows where it thinks it is. Tying a finding to a specific weld or frame number still depends on a human recognising the location, which is why the footage matters more than the coordinates.
Lighting compounds it. The aircraft brings its own light, so the inspection sees what the beam reaches and nothing else. Reflective wet surfaces, dark coatings and corners all behave differently under a moving light source than under the diffuse lighting that staged inspection provides, and defect appearance changes with the angle the light arrives from.
Collision tolerance is the design response that makes the rest viable. A caged aircraft can touch structure without ending the flight, which is what allows it to work close to plate in confined geometry rather than maintaining standoff it has no room for.
What It Genuinely Removes
The case for the approach is strong and does not need overstating, because the costs it displaces are the large ones.
Staging and access. Scaffolding a tank internally is often the single largest line in the inspection, and it is erected and dismantled whether or not anything is found.
Confined space man-hours. Every hour a person spends inside is an hour of entry exposure, attendant cover, rescue standby and atmospheric monitoring. Removing the hours removes all of it.
Time to first information. An aerial survey can establish whether a compartment needs detailed attention before committing to the access that detailed attention requires.
Screening at scale. Where a vessel has many similar compartments, surveying all of them to decide which few need entry is a different proposition from entering all of them.
A durable record. Footage can be reviewed later, compared against a previous survey and shown to a third party, which a person’s notes cannot.
Read together, the strongest use is triage rather than replacement. Survey everything from the air, then send people only where the survey says people are needed, and send them knowing what they are going to look at.
The Limit Buyers Are Not Told About
Here is the part that gets left out of the proposal. A camera establishes condition. It does not establish thickness, and thickness is why a great many tank inspections happen at all.
Ultrasonic thickness measurement requires physical contact between a probe and the steel, and the requirements are unforgiving. Air gaps significantly degrade ultrasonic signal transmission, so the probe has to be firmly coupled to the surface. Couplant gel has to be present, dispensed once contact is established. Contact force has to sit in a usable band, since too little gives poor coupling and too much squeezes couplant out from under the transducer. Surface condition and probe geometry both affect whether a reading can be obtained at all. And a valid result is not a single number but a sequence of stable readings.
A drone that flies beautifully through a tank has told you what the steel looks like. Whether the steel is thick enough is a different measurement, and it requires something to press against the plate.
Aerial contact inspection does exist and has been demonstrated. Published research using a caged quadrotor with a magnetically held probe on a galvanised steel duct obtained good, stable readings across repeated autonomous inspections, measuring wall thickness at 3 millimetres with a spread of 0.04 millimetres and describing the system as showing high repeatability, at a reference contact force of around 2 newtons.
The same research is candid about how narrow that result is. Testing covered a single ferromagnetic material, with non-ferromagnetic surfaces, where magnetic coupling is reduced, left unevaluated. The evaluation was small: one site, three autonomous flights and one manual comparison flight. Accuracy was referenced to the vehicle’s own odometry rather than an external measurement. The approach still depended on manual probe orientation and manually identified inspection points. And low-level control and the judgement of measurement quality both rested on proprietary vendor firmware and software.
The manual piloting finding is the most operationally relevant of them. Pilots have only indirect cues about contact quality, propeller sound and aircraft pitch, so they cannot easily tell whether a reading is stable, and an intermediate-level pilot in that study lost measurement stability intermittently. Aerial UT is therefore not a skill a competent inspection pilot simply adds; it depends substantially on autonomy that is still developing.
How to Scope It Honestly
The practical approach is to be explicit about which question the flight answers and what happens next.
The fourth item is missed often enough to be worth repeating. Flying a drone into a tank does not de-classify the tank. If a person is at the manway operating the aircraft, with the space open, the entry controls that govern the compartment still apply to whoever is working at its boundary, and the atmosphere inside is unchanged by the fact that the thing inside it is not breathing.
Scoped as triage, internal aerial inspection is a strong proposition with a clear and defensible saving. Scoped as a replacement for thickness measurement, it promises something the physics does not currently support at scale. Our drone training services cover the operational planning that distinguishes the two, which is the part that determines whether a survey produces a decision or an argument.
Frequently Asked Questions
How does a drone navigate inside a tank with no GPS?
By onboard sensing, building a position estimate from visual and inertial odometry and sometimes ranging sensors, tracking movement relative to surfaces rather than to an external reference. Steel structure also distorts magnetic heading reference. The estimate drifts, particularly in large featureless compartments, and published research on aerial contact inspection notes that reported accuracy was measured against the vehicle’s own odometry rather than external ground truth.
Can a drone replace ultrasonic thickness measurement?
Not generally, at present. UT requires firm probe contact with couplant, since air gaps significantly degrade signal transmission, contact force within a usable band, suitable surface condition, and a sequence of stable readings rather than one value. Aerial contact UT has been demonstrated in research with good repeatability, but on a single ferromagnetic material, across a small number of flights at one site, with manual probe orientation and manually identified inspection points still required.
What does internal aerial inspection actually save?
Mainly staging and confined space man-hours. Internal scaffolding is often the largest line in a tank inspection and is erected whether or not anything is found, and every hour inside carries attendant cover, rescue standby and monitoring. Surveying from the air to decide which compartments need entry converts a whole-fleet access cost into a targeted one.
Does using a drone remove the confined space requirements?
No. The compartment’s atmosphere is unchanged by the absence of a person inside it, and the controls governing the space continue to apply to anyone working at its boundary, including the operator at the manway. Isolation, atmospheric monitoring and standby arrangements remain part of the job.
Sources: Autonomous Aerial Non-Destructive Testing: Ultrasound Inspection with a Commercial Quadrotor in an Unstructured Environment, published research on aerial contact ultrasonic measurement, covering the coupling requirements including the effect of air gaps on signal transmission, couplant dispensing after contact, the approximately 2 N reference contact force and the consequence of excessive force, magnetic probe retention on ferromagnetic surfaces, the requirement for a sequence of stable readings, reported wall thickness of 3 mm with a spread of 0.04 mm across three autonomous inspections, position accuracy referenced to the vehicle’s own odometry rather than external ground truth, the difficulty of manual piloting given only indirect contact cues, and the stated limitations covering single-material testing, the small evaluation scale, reliance on manual probe orientation and inspection point identification, and dependence on proprietary vendor firmware for low-level control and measurement quality assessment
MarineCraft Journal
Get the next briefing in your inbox
Safety protocols, offshore training, regulation and market analysis, written for Malaysia’s maritime and offshore industry. No sales pitches.
We use your address only to send MarineCraft Journal. Unsubscribe any time. See our Privacy Policy.