Spacing is a safety constraint before it is a fact
How far apart the vessels sit decides what can be worked on at once and what has to wait. Reading it off the plant rather than off a drawing is the difference between planning and assuming.
The problem
The plan of record is a drawing, and the plant is not.
A process plant is a dense site where almost nothing can be measured by walking up to it. Vessels are large, hot, fenced or live; cranes come and go; and the questions people need answered are about sizes, spacing and what has changed. The drawing answers what was built, which is not the same question.
Nobody can measure a live vessel with a tapeso sizes and spacings come off a drawing, and the drawing was right once.
Temporary plant is invisible to the recorda crawler crane is on site for weeks and in no document, and its reach is the thing that decides what can happen around it.
Condition is looked at, not trackedsomebody notices staining on a walk-round, or they do not, and either way there is no dated record of it.
The film
Thirteen spheres, two cranes, one screen.
Why this gets commissioned
Everything on a plant is decided by distance.
Clearance between vessels, the reach of a crane, where a hazard zone ends, how far a muster point is. These are the questions that govern what work can happen and when, and all of them are distances. On a site where you cannot stand next to most of the things you are asking about, that is a hard way to run.
Lifting plant turns up and the drawing does not know
A crane’s reach against the nearest vessel is the whole of the question for a lift plan, and it is a measurement that exists for a few weeks and then stops existing.
A walk-round is a sample, not a survey
Coating and corrosion are judged by whoever walked past most recently. There is rarely a dated picture behind the judgement and almost never one of every vessel.
The expensive part is access, not reading
Getting a person to the right place on a live plant is permits and time. A drone orbit is neither, which is what makes doing this repeatedly thinkable.
What it changes
What it changes.
Sizes and spacings read off the plant
Thirteen vessels measured on the 3D, with the gap between them, rather than taken from a document.
Temporary plant put on the record
Both crawler cranes placed in 3D with their boom tips and their reach drawn on the ground, from the same footage.
Condition with a date and a picture behind it
Every sphere’s coating read across six views, with the one worth looking at flagged and then checked by a person.
A base to repeat against
Nothing about this pass is bespoke. Fly it again and the second pass is a comparison rather than a fresh start.
What it is used to decide
The questions it is there to answer.
None of these is decided by the model. Each is a person deciding, with a measurement in front of them instead of a drawing and a memory.
Will this lift reach, and what is underneath it?Each crane’s boom tip is placed in 3D and its reach is drawn on the ground against the nearest sphere, so the lift plan starts from the plant as it stands this month.
Which vessel should an inspector look at first?The coating on all thirteen was read across six views and one was flagged for staining and rust streaks. That is a shortlist to go and check, not a verdict.
Is the spacing what the drawing says?Ten of the spheres sit 10.6 m apart shell to shell, measured on the 3D. Where that disagrees with the record you hold, you know before it matters.
What has changed since the last pass?Everything is dated and the method repeats, so a second flight answers the question the first one could not.
The readings
What was measured, and what kind of figure each one is.
Every figure on this page says what kind of figure it is. That is the whole of why it can be used.
13storage vessels measured on the 3DMeasured. Ten of 22.7 to 23.0 m across and three of 11.4 to 14.7 m, with their volumes.
10.6 mbetween the spheres, shell to shellMeasured on the 3D, to about seven per cent. Scale comes from truck cabs of known size, not from ground control.
41 m and 99 mthe two crane boom tipsMeasured. Each found in two frames, with the sight lines meeting within 0.1 m.
1sphere flagged for staining and rust streaksRead by AI across six views, then checked by a person. A prompt for an inspector, not a condition grade.
9 clipsand 412 frames, all placed in one 3DMeasured. One hour thirty-seven minutes from footage in to the finished 3D, on a single desktop graphics card.
Honestly
What this is not.
Where a surveyed figure exists, it wins
Sizes carry about seven per cent, which is what reading scale off the footage rather than off ground control costs. Good enough to compare with and plan with, not good enough to settle anything.
Not a condition assessment
The coating readings are AI, raised for a person to confirm. Nothing here grades a vessel or clears one.
The operations layer is sample data
Every reading in it was invented to show the shape of the thing. None of it is a real plant’s.
No saving is claimed, because none was measured
Nothing about time or cost on a real plant was measured here, so there is no figure for it and we are not inventing one.
Built on stock footage, as our own demonstration. Nobody commissioned it: it exists to show what the method does on a working industrial site rather than on a tidy one. The site and its operator are not named here and are not ours to name.
Questions we get asked
Before you ask.
Is this a real plant?
Yes, a working petrochemical plant, and the footage is stock. Nobody commissioned this and the operator has no involvement in it. The operator’s name is painted on one sphere and is blurred in the films and flattened in the 3D.
How accurate are the sizes?
About seven per cent either way. Scale comes from truck cabs of a known size in the footage rather than from ground control, which is what that figure is the price of. It is good enough to compare vessels and to plan with, and not good enough to settle anything.
Did AI decide the sphere was corroded?
AI read the coating on all thirteen across six views and raised one. A person then looked at it. The output is a shortlist for an inspector and the page says so on the picture, because a condition verdict from a drone pass is not a thing we can honestly sell.
What is in the operations view?
Hazard zones, escape routes to muster points, safety equipment, a gas-detection map and an X-ray of fill levels and buried lines. All of it is sample data. It shows the shape of what an operations layer would do on your plant with your data in it, and it is not any real plant’s readings.
How long did it take?
One hour thirty-seven minutes from footage in to the finished 3D, on one desktop graphics card. That is the machine time, not the reading and the checking round it.
What did you get wrong?
The large white vessels were first fitted as upright tanks. Looked at again they are spheres, and they were measured as spheres. The first patch over the operator’s name also covered the back of a sphere the drone never saw; it now covers only the side it did.
Can we see it running?
Not as an open demo. We show it on a call, because what is worth showing is how it answers your questions rather than this plant’s.
What would it need from our site?
Drone orbits at a couple of heights, flown by pilots who already fly them, and something of known size in frame for scale. If you can give us ground control, the seven per cent comes down.
Got a site where nothing can be measured up close?
Tell us what it is and what you need to know about it. We will say whether this would answer it, including when it would not.