Overview
A multi-storey self-storage building was going up in Mississauga, a steel-framed structure with metal cladding on the outside and long-span floor slabs inside. The frame was erected, the slabs were poured, and the next trades were waiting. Then the fabricator’s model and the building stopped agreeing with each other. The fabricator had an as-designed IFC model. The contractor had the site. The installer had panels to hang off the girts and floors that needed to be level. Everyone had a version of where things were, and none of the versions matched. The project sat for weeks while the owner asked for a plan.
The steel fabricator brought us in. The ask, on paper, was a 2D as-built CAD of the structure. What the project actually needed was an as-built that sat on the same grid as the as-designed model, so the differences could be measured instead of argued about. This case study is about that gap: as-designed versus as-built, and what it takes to put the two side by side.
As-designed is a promise. As-built is what’s there.
The IFC model is what the building was supposed to be. It is the fabricator’s promise to the contractor, and it is what the installer’s shop drawings were made from. The point cloud is what was actually erected and poured, to a few millimetres. On a building that has gone up cleanly the two agree and nobody thinks about it. On this one they didn’t, and the first job was to say by how much, where, in a way three companies would accept.
That only works if the scan and the model share a coordinate system. The project had survey control and grid lines. We tied every scan to that control, and we tied the as-built to the same grid the model was drawn on. Without that, an overlay is two pictures on top of each other. With it, every offset is a number on a grid line that everybody on the project already uses.
Three instruments, one dataset
Site conditions were not good for a single method, so we used three. Mobile scanning with an RTK kit covered the site and the exterior quickly and put the whole dataset on real-world coordinates. A high-precision terrestrial scanner did the steel, the girts, the perimeter and the slab surfaces, where the accuracy actually mattered. The survey control tied both together. That is the point people miss about surveying and scanning: they are not competing, they are the same job done at different resolutions, and on a project like this you need all of them.
The edges got the most attention. The corners and the places where one part of the structure meets the next are where erection tolerance stacks up, and that is exactly where the as-designed and the as-built disagreed most.
Levelness
Slabs came into it as well. The installer needed to know whether the floors were level and where they weren’t, because everything above them assumed they were. Floor levelness came off the same control as the steel, so a high spot on the slab and a girt that was out of plane were two lines on the same drawing, not two separate arguments.
The deliverables, and the gap in the brief
The brief said 2D CAD. We delivered it, floor by floor, with the point cloud and a viewer so anyone could check a dimension the drawings didn’t show. Then the marked-up elevations with the as-built measurements written on them. Then, when the fabricator asked, plan views at each light-girt elevation so the installer could read the offset at every panel line.
Here is the gap. The fabricator had the as-designed model from the start. If that model, or a 3D CAD of it, had been in our hands on day one, we could have put the point cloud directly against it and reported the differences without drawing a 2D as-built from scratch. That is faster and it is exactly what an as-designed-versus-as-built check is for. Instead the model came to us near the end, we overlaid it against our data, and the fabricator built the compared plans from there. It worked. It would have worked sooner with the model up front.
The outcome
By the end of it the project had one as-built dataset, on survey coordinates and on the project grid, and the as-designed model overlaid on it. The fabricator could see which members were out and by how much, the contractor could put a rectification plan in front of the owner, and the installer could see where the panels and the floors actually were. The job finished, everyone was working from the same numbers, and everyone was happy with the result.
What we learned
If there is an as-designed model, ask for it on day one. Scanning against a model is faster than drawing an as-built from nothing, and it answers the question the project is actually asking.
Survey control is not optional on an alignment problem. Tie the scan to the project grid or the overlay means nothing.
Mobile, terrestrial and GNSS are one workflow, not three choices. Use the fast one for coverage and coordinates, the precise one where the millimetres matter.
Ask who else is reading the drawings. Three companies were. Knowing that up front would have put the plan views at the girt elevations in the first delivery, not the third.
Steel or slabs not matching the model? Have a look at our construction laser scanning and as-built documentation services, the difference between static and mobile LiDAR, or 3D laser scanning in Toronto.
