Case study: a crane canopy, three support decisions and 104 shop drawings

A crane canopy: three support decisions, then 104 shop drawings

A pitched steel canopy over a concrete basin, with a 5 t bridge crane on column brackets. It started as a 2D steel drawing in a tender package. Twice, fixing one check moved the problem somewhere else before the design settled. Then the same model went to fabrication: checks, numbering, drawings, a revision and the delivery package — this is the sequence, with the numbers.

Case study

Hot-rolled I-sections, S275 · 5 t bridge crane · 35 load-bearing parts analysed · 104 shop drawings · AISC 360-16 · EN 1991-1-4

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

Step by step

The first model, as built from the tender drawing.
The first model, as built from the tender drawing.

1 · From a tender drawing to a 3D model

The input was the steel drawing of a tender package: 20,343 entities in one DWG. I took off the member list and built a 3D detailing model from line coordinates — 34 members: six columns, three portal frames, a crane beam under the ridge, edge beams, purlins and braces. The scale came from the grid: one drawing unit was 50 mm, although the file header said millimetres.

The 35 parts selected from the revised model.
The 35 parts selected from the revised model.

2 · The engineer reworked the model

The engineer then changed the model: the single ridge crane beam became two side crane rails on brackets, an eave extension got its own columns, and roof purlins and wall girts were added on all four sides. The lesson I keep: read the live model, never my own build script. The engineer selected the 35 load-bearing parts for analysis; purlins and girts carry the cladding through load panels. A name-based classifier would have dropped the crane rails, so this project got its own classification.

Before and after: sliver members circled.
Before and after: sliver members circled.

3 · Sliver members and silently lost panels

Work points in a detailing model are legitimately offset — braces 40 mm below the rafters, edge beams 26 mm above. The converter could not merge them and left 13 sliver members of 3–26 mm and asymmetric eaves. I snapped the axes, levelled the eaves and merged nodes: 52 → 40 nodes, 68 → 55 members. One more trap: after levelling, four corners of a gable panel fell on one node and the solver silently rejected 5 of 14 load panels. Corners are now de-duplicated, and panels are counted after import.

Bracing as detailed, and after the engineer's edit.
Bracing as detailed, and after the engineer's edit.

4 · First design run: the braces

Loads: exposure category 0, no snow, roof live load 1.0 kN/m², cladding 0.50 kN/m², seismic importance factor 1.5, crane loads from the crane data. The first design run flagged the chevron braces: CHS 88.9×4 with a slenderness near 190 buckles elastically. The engineer split them into X braces at the crossing, made them tension-only, extended the crane rail by 1 m and added hanger diagonals; I added a case with the crane at the cantilever tip. Then the trap: the new combinations were linear, so the ‘tension-only’ brace carried 55 kN in compression over its full 6.68 m — a false failure that disappears once the combinations are nonlinear.

Analysis model after the engineer's edits.
Analysis model after the engineer's edits.

5 · Direct Analysis, verified in the tables

All combinations run as P-Δ nonlinear cases. I confirmed from the property-modifier table — not from the member query, which does not show it — that the 0.8·EA and 0.8·EI reduction was applied to every member. Strength passed everywhere. Serviceability did not.

22.6 mmwind drift, pinned (limit 13.6)
3.9 mmwind drift, fixed
398 %anchors, fixed base

6 · Fixing the drift broke the anchors

With pinned bases, the transverse direction had only frame action: wind drift 22.6 mm against H/400 = 13.6 mm, seismic drift 150 mm against 65.1 mm. The engineer fixed the bases — drift dropped to 3.9 and 32.1 mm and every check passed. Then the base plate went into the connection software: a base moment of about 61 kNm sent the anchors to 398 % and the welds to 100 %. The plate could not grow: the concrete below it limits its size.

The three support states as ratios to their limits.
The three support states as ratios to their limits.

7 · Pinned again — with stiffer columns and the right stiffness

Final decision: pinned bases, columns HEA260 → HEB260. Per AISC 360-16 C2.3 the 0.8 reduction is for strength and stability; drift uses nominal stiffness: 14.4 mm against H/300 = 18.1 mm (with reduced stiffness it would be 17.9 mm and fail). My first ‘nominal’ run was wrong — I had unlocked the model before redesign, the modifiers stayed at 0.8 and I reported 17.9 as nominal. The correct order is preference → analysis → design → analysis, and I check the modifier table every time. Seismic drift 97.5 mm against 130.3 mm, with the limit for cladding detached from the frame. Anchors 12.7 %, welds 20.6 %.

0.54highest strength utilization
0.60deflection-type ratio, 305 mm bracket
62report pages

8 · A deflection ratio is not a strength ratio

The software’s summary listed 0.60 as the highest ratio — a deflection-type ratio on a 305 mm crane bracket under a serviceability combination. Strength governs at 0.54, on a brace. The 62-page report (54 calculation pages + the 8-page base plate report) states which is which, and notes that crane girder fatigue and web crippling are outside its scope.

Fabrication model (members shown; plates and bolts counted).
Fabrication model (members shown; plates and bolts counted).

9 · Back to the detailing model for fabrication

Connections, base plates and anchor rods were detailed in the model — 763 parts, 32 anchor rods. From here the work was fabrication: checks, numbering, drawings, then a revision.

404overlaps checked
8slotted plates, not errors
0 / 101wrong main parts

10 · Checks before a single drawing

Before any sheet is drawn the model goes through four checks: loose and unwelded parts, volumetric clashes, the main part of every assembly, and identical assemblies with different marks. The clash check found 404 overlaps; almost all were anchor hardware passing through plates, as designed. The rest were 8 roof-brace end plates, each overlapping its tube by about 360 cm³. The engineer confirmed the plates are slotted around the tube: the check sees the tube’s box, not the slot. No assembly had a wrong main part (0 of 101). Four groups of look-alike assemblies were listed for the engineer to review.

The 3D general arrangement sheet, with grid lines.
The 3D general arrangement sheet, with grid lines.

11 · Marks by member type — and why in one pass

Most assemblies were still in the default series. With the engineer’s prefix table I set the series of 46 assemblies by member type (columns, rafters, crane beam, edge beams, braces, anchors) and read every value back; numbering went from 15 to 36 assembly marks. Later the engineer asked for a part prefix on 156 secondary parts. Renumbering after that kept every group intact (36 stayed 36) but re-sorted 577 assembly positions — column 1 became column 5, and so on. Nothing was wrong, but any drawing made in between would have carried the old numbers. Since then every prefix change is done in one pass, before the first drawing.

An assembly drawing from the reissued set.
An assembly drawing from the reissued set.

12 · 112 drawings, then a revision

The first issue had 112 drawings, all at 1:10 with no broken views, so a length measured on the DWG is the real length times ten. Nine days later the engineer revised the model: a ninth column on a new grid line, with its anchors. I worked on a copy, renumbered, and reissued the set. During the reissue the engineer spotted roof-brace end plates rotated by a fraction of a degree; after the fix, the brace marks changed, so every affected sheet was deleted and drawn again rather than updated. The final set has 104 drawings: 30 assembly, 71 single-part and 3 general arrangement.

104drawings checked
159 / 165positions placeable
24 / 30assembly sheets complete

13 · Could an erector use the sheets?

The drawing check on the final set reports 41 findings: 3 single-member dimension chains left on purpose, 22 sheets the software marks as ‘up to date, may need checking’, one overlap of two plan views on a general arrangement sheet, and 15 notes. Then a stricter test: an erector simulation that uses only what is printed on each assembly sheet — marks and dimension lines — to place every part. 159 of 165 positions can be placed; 24 of 30 sheets are complete. The remaining six are one symmetric plate per crane column, located by symmetry rather than by a dimension; the engineer accepted that.

The steel model, rendered: crane rails on brackets, X-bracing, girts.
The steel model, rendered: crane rails on brackets, X-bracing, girts.

14 · The package — and the same model, rendered

The delivery package has 158 files (10.7 MB): 30 assembly and 71 single-part DWGs, 3 general arrangement DWGs plus one merged DWG, 42 plate DXFs with a manifest, 7 Excel lists, the model archive and an offline 3D model that opens on a phone and shows every assembly mark. The render is made from that same steel model; cladding and the basin below were not in our scope and are not shown.

In numbers

20,343drawing entities read
3support decisions
0.54highest strength utilization
62report pages
104shop drawings, reissued
159 / 165positions placeable

What stays with people

Decisions

The model rework, the X bracing, fixed and then pinned bases, HEB260 columns, the H/300 limit, the seismic drift limit for detached cladding, the mark prefixes, the model revision and the accepted open items on the drawings were the engineer’s decisions. I proposed, checked and documented them — including where my own first ‘nominal’ run was wrong.

Responsibility

Crane girder fatigue and web crippling are outside the report, and it says so. The calculation report going to construction is reviewed and signed by a licensed partner engineer.

Is one fix breaking another check?

Send the model or the drawings and the check that keeps failing. I will trace it and show the options with numbers.

info@structomat.com