=== Step C: drilling-factor sweep on in-house DKT (site=baseline (typical CONUS)) ===

  building corner network (70 panels, dedup_tol=300.0 mm)...
    89 unique mid-plane corners (3.15 panels per corner).
  subdivision depth = 3 (mean edge 973 mm; target 200.0 mm; 4^d sub-triangles per base triangle)
  mid-surface mesh built: 4639 nodes, 8960 triangles (128 per panel).
  boundary nodes (y < 200 mm): 338
  wrote VTU -> reports/full_dome_midsurface.vtu
Mesh: 4639 nodes / 8960 tris
Load: wind_cc_peak p = -3831 Pa (p_outward = 0.00383 MPa)

Reference solver outputs (from prior runs):
  OpenSees ShellDKGT  u_max=52.42 mm  sigma_b=1.7450 MPa
  CalculiX S3         u_max=12.80 mm  (membrane VM=0.385 MPa)

In-house default drilling_factor = 1e-2 (current)

 drilling_factor    u_max [mm]  sigma_b_max [MPa]   sigma_m_max     VM_max  ratio_u_to_OPS  ratio_u_to_CCX
         0.0e+00        30.330            23.9853        0.8402    27.1016           0.579           2.370  (6.1s)
         1.0e-09        29.908             0.7763        0.8167     2.4094           0.571           2.337  (5.8s)
         1.0e-07        29.908             0.7763        0.8167     2.4094           0.571           2.337  (5.6s)
         1.0e-05        29.904             0.7761        0.8166     2.4091           0.570           2.336  (5.6s)
         1.0e-03        29.790             0.7562        0.8146     2.3888           0.568           2.327  (5.7s)
         1.0e-02        29.469             0.6450        0.8087     2.2597           0.562           2.302  (5.7s)
         1.0e-01        28.321             0.6586        0.7769     1.8192           0.540           2.213  (5.8s)
         1.0e+00        23.879             0.6291        0.6451     1.4018           0.456           1.866  (5.8s)

Interpretation:
  - If u_max grows substantially as drilling_factor goes to 0,
    drilling stiffness is artificially reducing in-house deflection
    => OpenSees may have weaker / no drilling stabilisation.
  - If u_max is roughly constant across the sweep, drilling is NOT
    the cause of the OpenSees disagreement.