Measurement that answers a question
Most fabrication problems are not discovered in the workshop. They are discovered at fit-up, when a spool, a skid or a structural section arrives on site and the holes do not line up. By that point the cost is no longer the cost of the error. It is the cost of the vessel on standby, the crane hired for the week, and the shift that cannot proceed.
Dimensional control is the discipline that exists to prevent that moment. It is not simply measuring things accurately. It is measuring them against a defined reference, to a defined tolerance, in order to answer a specific question before the item moves.
It starts with the question, not the scanner
A general survey captures what is there. A dimensional control survey sets out to confirm or disprove something — that two mating faces are parallel within tolerance, that a nozzle sits where the drawing says, that a structure has not moved between fabrication and load-out.
That distinction changes the whole job. The question determines where control is established, how dense the capture needs to be, what gets modelled and what gets ignored. Two surveys of the same skid, one for as-built records and one for dimensional control, look different on site and produce different deliverables.
It also determines what “accurate enough” means. Accuracy is not a property of the instrument. It is a property of the instrument, the control network, the number of setups and the environment together. A dimensional control survey that claims a tolerance the network cannot support is worse than no survey, because it will be believed.
Control is the part that matters
The scanner gets the attention. The control network does the work.
A control network is a set of reference points held common across every setup, so that each scan is tied to one consistent frame rather than to its neighbours. Without it you have a dataset that is internally tidy and externally meaningless — fine for visualisation, useless as evidence.
With it, you can compare the captured geometry against the design model, against a previous survey, or against the mating structure it has to join. That comparison is the deliverable. Everything upstream is just how you got there.
Where it pays for itself
The economics are straightforward, and they are almost always about avoided rework rather than the survey itself.
First-time fit is the clearest case. Where a fabricated item has to marry to an existing structure — a replacement section into a live plant, a new module against an existing deck, a spool into a run of pipework built years ago — the design model is a statement of intent and the as-built condition is the fact. Scanning the existing condition and fabricating to it, rather than to the drawing, removes the surprise.
Load-out and transport verification is the second. An item measured before it leaves the yard and again on arrival tells you whether transport did anything to it, and gives both parties a defensible record if it did.
Third, and most underrated, is the dispute that never happens. A measured record taken at handover is cheap. The argument about who was responsible for a 30 mm discrepancy, six months later, with no evidence on either side, is not.
What you should expect to receive
A point cloud on its own is not dimensional control. It is the raw material.
The deliverable is a report that states what was measured, against what reference, to what tolerance, and whether it passed. Deviation is usually shown two ways — as a colour heat map for the overall picture, and as specific measured values at the points that matter. Both are needed. The heat map shows you where to look; the numbers are what an engineer signs against.
Alongside that you would normally get the registered point cloud, the control network report, and, where the scope includes it, extracted as-built geometry or a CAD comparison. If a survey report gives you a colourful picture and no numbers, ask for the numbers.
When you do not need it
Not every job justifies it. If the item is being built to drawing with generous tolerance and nothing depends on it fitting something that already exists, a standard survey is enough. If the question is “what is there”, not “does it fit”, you want a point cloud survey rather than dimensional control, and it will cost less.
The test is whether anyone will make a decision on the numbers. If they will, the numbers need to be defensible, and that is what dimensional control buys.
Talk to us about a dimensional control scope
If you have a fit-up risk on a current project, the useful first conversation is about the question you need answered rather than the equipment. Tell us what has to fit what, and to what tolerance, and we can tell you whether a QA and dimensional control survey is the right route and what it would involve.