TL;DR
Laser scanning resolves deformation larger than its own registration uncertainty — ±5mm on ZEALOT building-scale work, with 2–4 mm RMS interior residuals. It is the right tool for full-field geometry, deviation against design, and before-and-after comparison across a large disturbance; instrumentation still wins for sub-millimetre trending. A defensible program locks permanent control, one capture method, and a written detection threshold before the second epoch.
Structural engineers ask a version of the same question every time a building starts behaving badly: is this moving, and how fast? Laser scanning gets pitched as the answer more often than it should be. It is a genuinely good tool for one half of that question and a bad tool for the other half, and the difference comes down to a single number — the accuracy of the registered dataset.
ZEALOT registers building-scale work to ±5mm. That figure is the floor of what a scan-based deformation study can resolve. Any movement smaller than the noise in the dataset is invisible, no matter how the deviation map is coloured.
A scan tells you the shape of a structure today with high confidence. It tells you the change since last quarter only if that change is larger than your registration error.
What deformation problems does scanning actually solve?
Full-field geometry. A total station gives twenty prism readings. A scan gives tens of millions of points across every surface, which means the deformation you did not think to instrument still shows up. On a warehouse floor, a plant rack, or a masonry facade, the failure is usually somewhere nobody placed a target.
Deviation against design intent. Comparing a registered cloud to the design model answers "is this column plumb, is this slab flat, is this wall bowed" in one pass. That comparison is absolute, not differential, so ±5mm is plenty when the tolerance being checked is a quarter inch.
Baseline capture before a disturbance. Adjacent excavation, underpinning, a heavy equipment install, a change of use that triples floor loading — capture before, capture after, compare. This is the highest-value scanning use in the deformation space because the comparison happens across a large expected change.
Documentation of an existing condition for a repair design. Once the structure has already deformed, the deformed geometry *is* the design input. Steel gets fabricated to what exists, not what the 1974 drawings claimed. Our as-built documentation work is exactly this: measure reality, hand the engineer numbers they can sign.
Where laser scanning is the wrong instrument
Sub-millimetre crack monitoring. Crack gauges, strain gauges, and tiltmeters read to a resolution scanning cannot approach. Do not replace them.
Short-interval movement trending. If a structure is moving a millimetre a month, a monthly scan campaign produces a graph of registration noise. Automated total stations with fixed prisms are the right tool at that scale.
Load testing with real-time feedback. Scanning is a snapshot, not a stream. Deflection under a live load test wants dial gauges or an automated monitoring system reporting continuously.
Anything requiring legal survey accuracy on a single point. Scanning is a distributed measurement. Where one point needs to be defensible to a surveyor's standard, set control and shoot it conventionally.
How a defensible scan-based monitoring program is set up
The programs that produce usable answers all share the same discipline. The ones that produce arguments skip step one.
- Establish permanent control first. Set monuments outside the zone of suspected movement, tie them to a project or state plane grid, and reoccupy the exact same control on every epoch. Absolute accuracy to control on our work runs 8–12 mm; without shared control, epoch-to-epoch comparison is meaningless because you are comparing two floating datasets.
- Fix the capture method and never change it. Same scanner class, same resolution, same setup locations, same registration workflow. A method change between epochs manufactures apparent movement. On the 220,000 sq ft plant retrofit we documented in our industrial plant case study, the pairing of VLX3 mobile capture with 38 RTC360 setups at tie-ins was locked for exactly this reason.
- Register tightly and report the residuals. Interior registration on our jobs targets 2–4 mm RMS, exterior 4–6 mm. Those residuals go in the deliverable. An engineer cannot assess a deviation map without knowing the uncertainty behind it.
- Compare cloud-to-cloud and cloud-to-mesh, and read them differently. Cloud-to-cloud picks up bulk translation well and rotation badly. Meshing one epoch and measuring the other against it handles curved and irregular surfaces far better — masonry, tank shells, sagging deck.
- Define the reporting threshold before the second epoch. State in writing that movement below the combined registration uncertainty is reported as "no detected change". This is the sentence that keeps a deviation map from being read as a crack report.
Reading a deviation map without being fooled
Colour ramps lie by default. The three failure modes that show up in review:
- Ramp scaled to the data, not the tolerance. Auto-scaling paints a perfectly acceptable structure in alarming red. Fix the ramp limits to the engineering tolerance and the picture becomes honest.
- Registration error read as deformation. A uniform tilt across an entire wing is almost always a registration artifact, not the building leaning. Real deformation is localised and has a mechanism behind it.
- Occlusion mistaken for movement. Points missing in one epoch and present in another produce edge artifacts. Compare only the overlapping coverage.
What this costs and how long it takes
Capture rate on open building interiors runs 80,000–120,000 sq ft per day. A repeat epoch on a previously scanned building is faster than the baseline because control is already set and setup locations are known. Scan-and-model work in general runs $0.25–$0.60/sq ft all-in; a monitoring epoch without modelling sits well below that, since the deliverable is a registered cloud and a comparison report rather than a Revit model.
The real cost comparison is not scan versus total station. It is measured evidence versus a repair designed off assumption. Our true cost of bad as-builts breakdown puts unverified existing conditions at 15–40 RFIs per project at $1,200–$2,500 each — a structural repair designed on guessed geometry lands at the top of that band.
Deliverables worth specifying
- Registered point cloud in E57 and RCP, one per epoch, on shared control
- Registration report with per-epoch RMS residuals
- Deviation map with ramp limits fixed to the engineering tolerance, not auto-scaled
- Tabulated deviations at named locations so the engineer can cite a value rather than a colour
- A written statement of the detection threshold
If the structure needs a model rather than a report, that is a separate scope — see Scan to BIM for how LOD gets set by system.
The short version
Use scanning when the question is *what shape is this structure in* or *what changed after a large disturbance*. Use instrumentation when the question is *how many millimetres did it move this month*. Most serious monitoring programs use both, and the scan is what tells the instrumentation where to go.
Structural teams working through a movement question in Ohio or nationwide can reach ZEALOT at 614-210-3679.