A digital twin is a computer replica of a real structure that's kept up to date with inspection data. Underwater it's built from video, photogrammetry, scanning and measurements, and the diver generates a good part of that raw material. What changes is less the task than how it's documented.
A term that wasn't there before has appeared in inspection specifications in recent years: digital twin. It shows up in client specifications, in contractor proposals and, increasingly, in the instructions a diving team gets before starting a project. It's worth understanding what it is, because it affects less what's done underwater and rather more how the work is requested and delivered.
What a digital twin is, and what it isn't
A digital twin is a computer representation of a physical asset that's kept alive. It holds the structure's geometry, information on its components and the history of what's been done to it, and it's updated every time new information comes in. Applied to a platform, a pipeline or a mooring system, it means having a copy of the asset where you can check the condition of an element without going back into the water.
The difference from a 3D model
This is the most common confusion. A 3D model is a photograph in volume: it describes what a structure looked like at a given moment. A digital twin incorporates that model, but adds two things that change everything: history and the connection to new data.
In other words, in a digital twin every flange, every anode and every support has an identity. You can check when it was last inspected, what was measured, what defect was noted and how it's progressed. That turns inspection into a time series instead of a pile of loose reports, and that's the whole advantage of the system.
What the operator wants it for
Fundamentally, to make decisions. Managing an installation's integrity means spreading a limited budget across many elements that age at different rates. With a well-organised history you can prioritise: what gets checked this year, what can wait, what justifies immediate intervention. Without that history, the decision is made by calendar and by gut feeling.
What data feeds it
This is where fieldwork comes in. A subsea digital twin isn't built with design engineering, or not only: it's built from what's collected in the water, project after project.
Inspection video and photography
It's the foundation. The video from a visual inspection provides the evidence of surface condition: corrosion, coating damage, deformation, biological fouling, detached elements. A good part of that material is still interpreted by eye, but more and more of it is processed automatically to detect changes between projects.
Photogrammetry and point clouds
Photogrammetry reconstructs a three-dimensional model from many images taken from different angles. The principle is simple: if the same point on the structure appears in enough frames and from different positions, the software can calculate where it sits in space. The result is a point cloud, a set of coordinates describing the object's surface.
That cloud can be converted into a mesh, measured against, compared with the design model, and overlaid with one from an earlier project to see what's moved or disappeared. Point clouds can also be generated with acoustic or laser scanners, with the advantage that acoustic systems cope with dirty water much better.
Measurements and testing
Ultrasonic thickness readings, cathodic protection potentials, defect dimensions, distances between flanges. All this numerical information is what gives the model its real value, because it's what allows margins and trends to be calculated. The work of underwater metrology fits in naturally here: its results are exactly the type of data the twin needs.
Position
A piece of data with no position is useless. To integrate any measurement into the model you need to know which element it belongs to and, in many cases, exactly where the equipment was when it was taken. That's why acoustic positioning systems and fixed references installed on the structure carry so much weight in these projects.
The diver as a data source
It's worth saying plainly: in a project aimed at feeding a digital twin, the diver is just another measuring instrument. His camera, his hands and his description are the sensor. And like any instrument, he produces good or bad data depending on how he's used.
That doesn't cheapen the trade — quite the opposite. An ROV can travel along a pipeline and record kilometres of video, but it can't clear fouling off a weld by hand, place a probe with the right pressure, or decide on the spot that it's worth looking behind a support. The part of the job that's still human is precisely the one that supplies the hard data.
Video quality
A photogrammetric model stands or falls on how it was filmed. The factors that matter are always the same: slow, continuous movement, with no sudden sweeps; wide overlap between passes so the software can find common points; stable lighting, avoiding glare and blown-out patches; and constant distance from the surface. A video shot in a hurry with the light too close can be completely useless for reconstruction even if it looks fine on the panel screen.
References and scale
Photogrammetry on its own reconstructs shape, not size. For the model to be usable it needs to be given scale, and that's done with known references: physical scale bars placed in the field of view, targets, measured distances between identifiable points. Placing those references correctly, and filming them so they're clearly visible, is a diver's task that determines the usefulness of everything else.
Metadata and record-keeping discipline
The third pillar is the least glamorous and the one that most often falls down. Every file has to carry with it what it is, where it was taken, when and under what conditions. Element identification according to the client's naming system, position reference, time, visibility, current conditions and any circumstance that explains an anomaly in the recording.
Without that, an excellent file becomes an orphan file. And an orphan file doesn't make it into the model: it stays in a folder nobody ever opens again.
What changes in how the work is documented
The fundamental change is that the report stops being the end product. Before, the deliverable was a document with photographs and observations that someone read once and filed away. Now the deliverable is a set of structured data that's going to be integrated into a system and compared with previous years'.
That has three practical consequences for the team. The first is that naming matters: calling an element something different from what the client calls it breaks traceability. The second is that description has to be standardised, with defined categories instead of free-form prose. The third is that the record is made on the spot, not from memory at the end of the shift.
Shared vocabulary
The less obvious consequence is that ambiguity gets penalised. Writing that an area shows significant corrosion adds nothing to a system that needs to compare. Writing the approximate extent, the location referenced to a known point and the defect category according to the procedure does add something. Divers who regularly work in plataformas demanding integrity programmes already know this way of working; for someone coming from port works it means a change of habit.
The real limits of the technology
It's worth being honest about this, because sales presentations tend to always show the same flawless model reconstructed in clear Mediterranean water in summer.
The water
Optical photogrammetry depends on seeing. In water with suspended particles, with visibility down to a few centimetres, or with current that makes it impossible to hold a steady track, reconstruction is poor or flatly impossible. In those conditions you fall back on acoustic methods, which cut through murky water but offer much less detail. There's no solution that works every time.
Keeping it updated
A digital twin is only worth as much as its last update. Keeping it current costs money and projects, and it's common to find models built with enthusiasm at the start of a project that have gone years without a new piece of data. When that happens, the system becomes a nice-looking but misleading archive, because it invites you to trust stale information.
Engineering judgement
The model describes, it doesn't interpret. It can flag that a surface has changed between two projects, but deciding whether that change compromises an element's integrity is still a job for an engineer with judgement. And locating the actual defect behind a layer of marine growth is still a job for someone who goes down with a scraper.
What skills all this demands of the diver
It doesn't call for programming or modelling knowledge. It calls for other things, closer to the trade but demanding.
It calls for solid handling of camera and lighting underwater, understanding what a badly aimed light does. It calls for the patience to repeat a pass that's gone wrong instead of letting it pass. It calls for precision in naming and annotation. It calls for the ability to describe what's being seen over the comms in an orderly way, because in many cases that description gets transcribed straight into the record. And it calls for understanding the purpose of the project, because a diver who knows what his footage is for films it better.
These are skills built up with hours of real inspection work and with a equipment supervisor you know, not with a course. Anyone starting out will find the usual route described in the guide to commercial diver training, and the inspection qualifications that open the door to these projects on the page about certifications.
In closing
The digital twin is not a threat to commercial diving, nor a revolution that forces the trade to reinvent itself. It's a change in where the work ends up: what used to finish in a report now feeds a system that lives for the whole life of the asset. The dive is still just as physical, and the judgement involved is still just as human.
What does change, and clearly, is the relative value of record-keeping discipline. A diver who films well, measures carefully, identifies precisely and describes without ambiguity produces data that can be used for years. One who doesn't produces material that has to be redone. That difference always existed; now it shows up sooner, and it pays better.
Frequently asked questions
What exactly is a digital twin of a subsea asset?
It's a computer representation of a real structure, with its geometry and its technical history, kept up to date with data from each inspection. It's not just a 3D model: what defines it is that it's fed new information throughout the asset's life.
Does the digital twin replace the diver's work?
No. The digital twin doesn't generate data by itself: it consumes it. Somebody still has to go down, clean, measure, film and check. What changes is that this work is documented to a higher standard, because its destination is no longer just a paper report.
Do you need specific training to work on these projects?
There's no digital-twin qualification for divers. What's required is mastery of visual inspection, judgement for measuring, and record-keeping discipline. The specific capture procedures are set by the contractor and explained on the project.
What is underwater photogrammetry?
It's a technique that reconstructs a three-dimensional model from many photographs or video frames taken from different angles. It works well with good lighting and clean water, and degrades quickly when there's turbidity or suspended particles.
Why is there so much emphasis on metadata?
Because a video with no date, no position and no identification of the element inspected is worth very little to a model. The data has to be locatable in the structure and in time; otherwise it can't be compared with earlier inspections or integrated into the twin.
Published 4 September 2026 · By Arthur More, professional saturation diver
Photo: NOAA Photo Library · CC BY via Openverse
This article is part of our in-depth guide to how deep a commercial diver can actually go — the complete picture, with sources.
This article is part of our in-depth guide to how deep a commercial diver can actually go — the complete picture, with sources.
This article is part of our in-depth guide to how deep a commercial diver can actually go — the complete picture, with sources.
This article is part of our in-depth guide to how deep a commercial diver can actually go — the complete picture, with sources.