# Office package (ZS-2026-001) — erratum & updates, 2026-07-04

Issued alongside the studio package (ZS-2026-002). Four items surfaced while
building the studio pipeline apply to the office package. **The office
engineering verdict (PASS at FoS 2.5 on the refined-FE tiers) is
unchanged**; two headline framings should be revised at the next document
revision.

## 1. FoS basis alignment (pipeline artifacts regenerated)

The issued report quotes FoS = 2.5 (strength) / 3.0 (buckling) throughout,
but the pipeline's `DEFAULT_SAFETY_FACTORS` still carried an older
4.0/3.0/5.0 scheme, so the downloadable `acceptance-*.json` contradicted the
report's numbers (e.g. D/C 1.58 in the JSON vs 0.99 in the report for the
same check — same physics, different FoS). `pu_foam.py` now encodes the
issued scheme (the failing `test_allowables_apply_safety_factors` that
documented this intent now passes), and all office acceptance artifacts have
been regenerated on the aligned basis.

## 2. Hand-calc screening envelope on as-built geometry (headline change)

The published controlling figure — "most-conservative envelope D/C = 0.99,
PASS" — was computed from the nest-sheet DXF panel dimensions (type 1 edge
986.7 mm) and only for the largest-area panel type. Two corrections:

- **As-built dimensions.** The assembly OBJ (which the FE tiers meshed, and
  which the report's own Table 5 quotes: 1012 mm) fits type 1 at
  1012.3 mm — the DXFs are a ~2.5 % smaller revision. Bending demand scales
  with dimension², so the envelope on as-built dims reads **1.04**, not 0.99.
- **All panel types.** The envelope evaluated across every panel type (the
  studio pipeline now does this by default) is governed by the 1021.8 mm /
  69.7° type (×16), not type 1: **D/C = 1.12** at severe.

Corrected framing (as adopted in the studio report): the Timoshenko
inscribed-rectangle envelope is a *screening* tier; at the severe envelope it
is exceeded (1.04–1.12) and the converged refined analyses govern — filleted
full-dome FE **0.94** (R14, computed on as-built OBJ geometry, unaffected by
this erratum) and single-panel rhombic FE 0.54 (R15; ~0.57 when rescaled to
as-built dims). Baseline-site envelope values remain comfortably below 1.0
(worst 0.48 on as-built dims).

## 3. Buckling tier replaced (strengthens the office story)

The issued R4 buckling numbers came from `*BUCKLE` on the merged tet mesh,
whose lowest eigenvalues are non-deterministic (sliver-element ARPACK
pathology, NEXT_STEPS item 1 — recorded mode-1 factors 1.99/5.78/22.5
across runs of the same deck). The acceptance pipeline now uses `*BUCKLE`
on the S3 shell mid-surface (`tools/ccx_shell_buckle.py`), carried through
a genuine three-level mesh-refinement study (subdivision depths 3/4/5) and
extrapolated — the eigenvalues decrease from above with a stable geometric
ratio, so no single mesh is quoted:

| Case | Level sequence (mode 1) | Extrapolated | Criterion |
|---|---|---|---|
| Balanced snow, baseline | 31.7 → 23.1 → 19.4 | **16.6** | ≥ 3.0 — PASS |
| Balanced snow, severe | 9.51 → 6.93 → 5.82 | **4.98** | ≥ 3.0 — PASS |
| Wind uplift | ≥ 15.2 severe / ≥ 35.6 baseline at every level | n/a¹ | PASS |

¹ Suction puts the membrane in tension — uplift eigenvalues are local
crease-band values that do not extrapolate; not a credible buckling mode
(the strength checks govern the uplift case).

This replaces the marginal published "BLF 3.356 vs 3.0" with a robust,
repeatable margin (severe snow 4.98 extrapolated). (The studio, by
contrast, extrapolates to 2.12 at its severe snow — see the studio
package's rated-envelope disposition at 68 psf.) Records:
`reports/office/shell-buckling-convergence-{site}.json`.

## 4. Base-ring slope constant (immaterial)

`BASE_PANEL_SLOPE_DEG = 15°` was recorded for the base-ring membrane check;
the measured panel-plane tilt on the office OBJ is 42.8°. The check's
1/cos(θ) amplification rises 1.04→1.36×; office base-compression D/C moves
~0.09 → ~0.12 at severe — no verdict impact. The studio pipeline uses its
measured 40.3° from the start.

## 5. Wind-basis applicability & enclosure classification (external EOR feedback, 2026-07)

A reviewing structural engineer raised two site-independent method points
that apply to the office package as much as the studio (his full comments
and disposition: `reports/2026-07-04--eor-feedback-integration.md`):

- **Dome-analogy coefficients.** Both packages derive C&C pressures from
  ASCE 7's spherical-dome figures (external GC_p = −2.6 + enclosed
  GC_pi 0.18 → combined 2.78). No code figure covers the faceted
  zonohedron literally; the definitive product-line closure is an ASCE 7
  Ch. 31 wind-tunnel study or validated CFD. The studio report now
  quantifies coefficient headroom explicitly (§ V.4); the same treatment
  should be added to the office document at its next revision. Office
  headroom at severe on the refined R14 basis (0.94): the combined
  coefficient can rise only to ≈2.96 (+6 %) before unity — thin.
- **Enclosure classification.** Reclassifying as *partially enclosed*
  (GC_pi ±0.55) raises net C&C pressure ~13.5 %. The office refined
  severe check does **not** absorb this: 0.94 × 1.135 ≈ **1.07**. At the
  baseline envelope the office remains comfortable (0.38 → 0.43). Until
  the enclosure classification is confirmed from the door design, the
  office severe-envelope strength verdict is conditional on the
  *enclosed* classification — this should be stated in the office
  document's next revision. (The studio, with more refined-margin, passes
  the partially-enclosed case at 0.78.)

## Recommended actions for the office document's next revision

1. Re-frame the § I / R2 headline from "envelope passes at 0.99" to the
   screening-tier framing above (verdict: PASS on refined FE at 0.94).
2. Replace R4's tet-mesh buckling numbers with the S3 shell mid-surface
   results (this *improves* the office margin).
3. Note the per-type envelope table (the studio report's Table 7 format).
4. The regenerated `reports/office/*.json` artifacts already carry all of
   the above; the published HTML/PDF narrative is unchanged pending EOR
   review of this erratum.