Datasets:

---

sylvainHellin

/

ifc-bench

Like 22

Follow-up to discussion #9 and PRs #10/#11. Every changed value was checked against the STEP data with ifcopenshell.

• Encoding: repaired the remaining Mac Roman mojibake (e.g. TÖºr → Tür) in 54 rows, including Q515, Q541, Q905 and Q914. One question text (Q269) changes, encoding only.
• fzk_house areas: net and gross differ by more than the 2% threshold (Galerie 74.51 vs 107.16 m²), so Q512,and Q541 now label net/gross and give both where areas are asked. Dropped the storey GrossFloorArea from Q513 so it no longer conflicts with the space-area sums in Q510/Q511. Units unified to m².
• fzk_house counts: Q508 said 46 structural elements while Q544 says 50; 50 seems more correct (4 beams, 42 members, 4 slabs). Q554 counted every wall twice: 13 walls, so 82 elements with Umbau-Status = Bestand.
• Q539: HANDRAIL is on the two IfcRailing occurrences.
• Q917 (smiley_west): the Kalksandstein 268899148 175 layer is 0.175 m (IfcMaterialLayer #42752), as already fixed in Q923.
• Q356 (fantasy_hotel_1): removed the total of 205; the listed counts sum to 316.
• Removed details the questions don't ask for (Q536 area, Q547 slab areas/volumes, Q554 per-type phase lines).

69 answers change

Merged, thanks @lorenzo-mambretti ! Every changed value checked out against the IFC data (areas, Umbau-Status count, railing PredefinedType, the 0.175 m layer, the fantasy_hotel_1 counts). Released as v2.0.3.