Case study: from architectural drawings to shop drawings, an entrance canopy end to end

From architectural drawings to shop drawings: an entrance canopy, end to end

An entrance canopy with a gatehouse underneath and an identical gatehouse beside it. Only architectural drawings existed — no structural system, no member sizes. I built the 3D detailing model first, derived the analysis model from it, designed the members and base plates, detailed the connections for site welding, numbered the model and produced the shop drawings, an IFC4 model and renders — all from the same model.

Case study

Cold-formed box sections, S235 · 481 parts · 254 analysis members · 14 base plates · 57 shop drawings · AISC 360-16 · EN 1991

The whole workflow in under two minutes (no sound).

Step by step

The three sheets, lines only.
The three sheets, lines only.

1 · The input: architectural drawings only

The architect’s DWG had three 1:50 sheets with 1,629 entities: a 14.9 × 4.0 m flat-roof canopy with a gravel roof and a parapet, and two identical gatehouses. No structural system, no member sizes. The file header declares millimetres, but the geometry is drawn in centimetres, and the levels in plan, section and elevation did not agree with each other. The engineer chose steel and box sections that fit the architecture, set the roof build-up and confirmed the levels.

Detailing model: canopy, gatehouse underneath and the guest gatehouse.
Detailing model: canopy, gatehouse underneath and the guest gatehouse.

2 · The 3D detailing model came first

I generated the detailing model from a geometry script and revised it four times: the parapet, the gatehouse levels the engineer measured on the elevations, then the grid. The grid lines and the gatehouse outline in the canopy plan sat inside blocks inserted far away from the sheets. My first reading did not look inside blocks and I placed the gatehouse in the wrong position; the engineer caught it from a screenshot. Since then every architectural drawing is read with block contents expanded in place. Purlins sit flush inside the beams so the roof finish reaches the architect’s level. Result: 225 parts.

Analysis model: canopy and gatehouse, unconnected.
Analysis model: canopy and gatehouse, unconnected.

3 · The analysis model, derived — not redrawn

The analysis model was derived from the detailing model. The canopy and the gatehouse under it are separate structures that never touch, and the general geometry step treated their nearby axes as one, shifted them and dropped purlins. For this project I replaced that step with a geometry builder that keeps each structure on its own axes, and compared every support position with the source. 154 nodes, 254 members, 10 fixed supports, 106 load panels, zero negative stiffness eigenvalues. The guest gatehouse is identical and stays out; the parapet members are not modelled, their loads are.

Applied wind load against the sum of support reactions, per component.
Applied wind load against the sum of support reactions, per component.

4 · Loads, each one balanced by hand

Gravel roof finish 1.10 kN/m², snow 0.88 kN/m², roof live load 1.0 kN/m², parapet loads on the edge beams. The canopy is a free-standing roof (EN 1991-1-4 §7.3): net pressure coefficients with blockage by the gatehouse, in both uplift and downward cases; the gatehouse is a closed building. For all eight wind cases the applied load equals the sum of support reactions. Along the way I found that the shared load step also put the gravel roof finish on the gatehouse walls; corrected here and logged for the shared code. 125 combinations.

Utilization of every analysis member.
Utilization of every analysis member.

5 · Member design — Direct Analysis Method

Members are checked to AISC 360-16 with the Direct Analysis Method, using the S235 grade the engineer chose for box sections. Highest utilization: 0.44 on a gatehouse post; the canopy beams reach 0.41 under the downward-wind combination, the canopy columns 0.27.

Every serviceability check as a ratio to its limit.
Every serviceability check as a ratio to its limit.

6 · Serviceability, span by span

The software’s own deflection check works per finite-element piece: on a beam split at every purlin, a 0.56 m piece gets a 2.3 mm limit — meaningless. I computed deflections span by span from node displacements: canopy beams L/1742, the worst cantilever L/1384. Wind drift of the gatehouse is H/457 against H/300; seismic drift ratios stay at or below 0.0058.

Canopy base plate and the governing checks of both plate types.
Canopy base plate and the governing checks of both plate types.

7 · Base plates in connection design software

From the analysis I picked 11 governing force sets per plate type — maximum anchor tension, uplift, compression, moment and shear, overstrength combinations included — and built the connection models automatically. The first import did not converge: the recommended welds were not created on import, so the welds are now written explicitly. The engineer then edited the model: concrete C30/37, shear through the anchors, weld size. Canopy plate: anchors 0.55, welds 0.97, concrete 0.26. Gatehouse plate: 0.25 / 0.68 / 0.22.

14base plates in the detailing model
56anchor rods with washers
1 DWGgeneral arrangement, merged

8 · Back into the detailing model

The checked plates went back into the detailing model: 14 base plates, 56 anchor rods with washers, and the welds; the columns now start on top of the plates. Door and window frames were added (not structural, outside the analysis). General arrangement drawings with profile labels were produced, cleaned up and merged into a single DWG.

78pages in total
62calculation pages
2 × 8base plate reports appended

9 · The calculation report

The general report template produced sentences that belonged to other projects — a bracing system that does not exist here, a closed-building wind diagram for an open canopy. I wrote a project-specific report instead: 62 calculation pages plus the two base plate reports, 78 pages in total. Construction documents are reviewed and signed by a licensed partner engineer.

Middle column head: cap plate, continuous main beam, cross beam, purlins set flush.
Middle column head: cap plate, continuous main beam, cross beam, purlins set flush.

10 · Connections detailed for site welding

The detailing model still had the analysis geometry: every member ran to its neighbour’s axis. The engineer chose site welding. Columns become shop assemblies with base and cap plates (PL15 on the canopy, PL10 on the gatehouses); the main beams run continuously over the caps; cross beams, purlins, edge beams and the parapet are cut to the faces they meet. Three members were 14.6–14.8 m long, longer than 12 m stock, so each got one shop butt weld: on the main beams 1.40 m from the middle support, between two purlins; on the edge beam and parapet head between two parapet posts. Open tube ends that stay visible got 6 mm end caps. I let contacts between solids decide where welds go: 30 shop and 440 site welds. Afterwards every one of the 481 parts was welded to something, and steel-to-steel overlaps were zero.

An assembly drawing and a single-part drawing from the set.
An assembly drawing and a single-part drawing from the set.

11 · Numbered, drawn and checked

Assembly prefixes follow the member types; numbering gave 28 assembly marks and 43 part marks. The set has 57 drawings: 12 assembly, 43 single-part and 2 general arrangement sheets, all at a fixed 1:10 without broken views, so a length measured on the DWG is a real length times ten. The drawing check found one assembly where a part had no position dimension; it was fixed, and an erector simulation that reads only what is printed on the sheet could place 24 of 24 positions. Then I opened the single-part sheets myself: the template had drawn no dimensions at all on plain members — a square-cut tube or a plate without features showed only its outline. The automated check could not see this, because it looks for missing positions, not missing lengths. I added overall dimensions and hole chains to all 43 sheets from the part geometry, and checked them against the design lengths.

One camera: the finished entrance, and the steel inside it.
One camera: the finished entrance, and the steel inside it.

12 · IFC, a model for the crew, and renders

The same model went out in three more forms: an IFC4 file with all 481 parts, the anchor groups and the grid; an offline 3D model of about 1 MB that opens on a phone and shows every assembly mark; and renders. The cladding in the renders follows the architectural drawings — the fascia band, the 50 cm column casings, the gatehouse walls with openings taken from the window frames in the model. The cladding is not part of the engineering scope.

78report pages
57shop drawings
481parts in the IFC4 model

13 · The deliverable

Architectural drawings came in. Out went a 78-page calculation report, two base plate reports, 57 shop drawings, a merged general arrangement DWG, an IFC4 model and a 3D model for the erection crew, all from one model and all in the order the project actually went. Construction documents are reviewed and signed by a licensed partner engineer.

In numbers

1,629DWG entities read
254analysis members
0.44highest utilization
470welds detailed
57shop drawings
78report pages

What stays with people

Decisions

Steel and box sections, the roof build-up and levels, fixed bases, the S235 grade, one model for both structures, the load assumptions, the base plate edits, site welding with shop splices, the splice positions and the mark prefixes were the engineer’s decisions. I proposed, checked and documented them.

Responsibility

Foundations stay outside this scope, and so does the cladding shown in the renders. The calculation report and drawings going to construction are reviewed and signed by a licensed partner engineer.

Only have architectural drawings?

Send the drawings and a short description. The first step is a structural proposal for the engineer to review.

info@structomat.com