Scanning a plan

The plan underlay asks you to do the reading: tap the two ends of a dimension, type its length, place the grid, edit the columns. Ten careful minutes, and free.

A scan does that reading for you. It finds a printed dimension to measure from, works out the scale, and marks the columns it can see — then hands the result back for you to check, with the dimension it used drawn on your own drawing. You edit it exactly as you would a grid you placed yourself, and the engine designs from what you approve.

It reads. It does not design. What comes back is inputs — a scale and a set of column positions, the same things you would otherwise type in. The calculation is unchanged: the same BS 8110 code, the same step-by-step calc sheets, the same PASS and FAIL on every check, the same ceilings. Nothing about a scanned building is decided by a model.

What it extracts, and what it refuses to

Two things come off the drawing, and one of them is far more dangerous to get wrong than the other.

• The scale, from a printed dimension string: which one it used, and the figure written on it.
• The columns: the centre of every one it can see, which then become setting-out lines.

Why it will not guess a scale

A wrong scale is the one error that hides. Every span comes out wrong by the same factor, so the plan still looks like a sensible building; the arithmetic downstream is perfect; the calc sheet is immaculate; and every bar on it is wrong. Nothing on the page gives it away.

So the scanner is told, at length, to say it could not read a dimension rather than estimate one — and Structura refuses a reading that has no dimension behind it, before it ever reaches you. Specifically it will not:

• Use the paper scale. 1:100 on the title block means nothing once a drawing has been photocopied, shrunk to fit or photographed — which is the same reason the underlay never asks you for it.
• Estimate from typical rooms. A bedroom is not a dimension.
• Measure off something short. An error in finding either end of a line is a fixed distance, so on a short one it is a large share of the answer. A dimension too short to measure from is refused rather than used.

If the drawing carries no readable dimension, the scan says so, and it costs you nothing. Set the scale yourself on the underlay.

The check is yours, and it is not a formality

When the reading comes back you are looking at a machine's opinion about a picture. You are the engineer of record for what happens next, so the app is built to make the check fast rather than to make it feel unnecessary.

Check the scale first

The dimension the scanner measured from is drawn back on your own drawing as a teal line, and named in words above the grid — "the 10.80 m overall dimension across the front". Two questions, and they take five seconds:

• Is that the line you would have picked?
• Does the length shown match the figure printed on the drawing?

If either is wrong, nothing else on the screen is worth reading. Start again, or set the scale by hand on the underlay.

Then check the columns

Tap a red marker to take a column out or put it back; tap open ground to add one; press and hold a column to drag it onto its line, with a ghost showing where it will land before you let go. Bay dimensions can be typed off the drawing, a bay repeated with its columns, and the layout mirrored. It is the same editing, the same gestures and the same code as a grid you fitted yourself — see the underlay tutorial for what removing a column actually does to the design.

The panel beside the plan keeps up as you edit, and says everything at once: in red what the engine would refuse, in amber what it will design happily but you should know. Both name the column by its grid reference — B/2 — and tapping a message moves the plan to the columns it names. The design button stays dead until the red is gone.

Undo has a redo here too, and it matters more on a scan than anywhere else: this is somebody else's reading, it gets edited more than a grid you placed yourself, and starting again costs credits. Stepping back and forward never rewrites what the scanner originally said.

A building that steps in

The scanner reads one plan, which is the ground storey. If the building sets back above it, that is yours to draw: pick the storey in the panel and take out the columns that stop below it, exactly as on the underlay. The storey count comes from what you were asked before the drawing was ever sent, so the setback lands on the level you meant. A stepped building is designed for gravity alone — up to 3 suspended floors — and the panel says so while the plan is still on screen.

The scanner is told not to tidy: not to complete a pattern, not to add a column because the grid "should" have one, not to nudge one to make the spacing regular. A stairwell where a column would be is the structure, not a mistake in the drawing. So expect gaps, and check that the gaps are real ones.

Read what it says it could not do

A scan reports its own confidence and lists what it could not make out, and both are shown next to the grid rather than tucked away. A low confidence is useful information, not a failure — it is the scan telling you which drawing needs a closer look.

What it costs, and what leaves your device

• Credits, on top of the design. A scan is charged separately from the project it feeds, because reading a drawing has a real cost behind it each time, unlike everything else in the app.
• Only when it works. A drawing it cannot read, a scan that fails, or a reading you reject costs nothing.
• The drawing is uploaded to be read. This is the one place in Structura where a plan leaves your device, it is the reason a scan is a separate feature rather than a button on the underlay, and the app asks before the first one.
• It is not kept. The plan is held only for as long as the reading takes. It is not stored on the server, not saved with your project, not in your backup, and never in analytics — see the privacy policy.
• Nothing else goes with it. Not your loads, soil bearing capacity, concrete grade, project name or client. Only the drawing and the number of floors you stated.
• One plan is one credit, and a building can be more than one plan. The scan reads the ground floor. If the floors above differ — a storey set back, a column that starts part way up on a transfer beam — you can scan those plans too, and Structura lines each one up against the floor below and tells you what differs before it changes anything. A plan it cannot line up confidently is refused rather than forced: a grid a bay out would turn every column into a setback.
• We do count how well it did. Whether a scan worked, how many columns it found, how many you kept, how long it took — counts and timings, with no account attached to them and nothing that could redraw your plan. It is how we can tell a scanner that is getting better from one that is getting worse.

If none of that appeals, the plan underlay does the same job with nothing leaving your device, and it is free. It just asks you to do the reading.

What a scan is not

• It is not a check on the architect. If a dimension is written wrongly on the drawing, the scan reads it wrongly too, exactly as you would.
• It is not a second opinion. It reads one drawing once. It does not know what the building is for, what is standing on it, or what the last engineer decided.
• It does not change what the engine will design. The same ceilings apply to a scanned grid as to one typed in by hand — see what it will not do.
• A PASS is still not approval. It means the gravity design checks out for the loads you entered. You are the reviewing engineer, and every output says so.

Next

• The plan underlay
• Load takedown
• Column design
• Pad footing design
• Approval drawings
• How Structura works
• What it will not do