External solver comparison

Verification methodSAME-MESH NUMERICAL CORRELATION · not physical validation

The 0.2.10 evidence includes bounded linear-static comparisons with Code_Aster 18.1 for named TET4, HEX8, TET10, and HEX20 cases.

01 Model and scope

A same-mesh comparison checks whether declared observables agree under matched geometry, material, loading and boundary conditions. It is a numerical correlation for those cases—not an independent physical experiment or a blanket solver ranking.

Route assumptions and calculation checks

Compare coordinates, connectivity, materials, constraints, loads, output locations, and sign conventions.
A small difference for one observable and mesh is not a universal accuracy statement.
The whole-repository G03 archive gate remains failed; WP14 remains on hold.

Useful foundations and prerequisites

Units, signs and error measures Interpolation, mapping and assembly

02 Essential equations and symbols

This sheet records a scope or execution boundary. Its detailed steps explain the relevant conditions.

Definitions and conventions

$u$: Compared displacement observable.
$\sigma{}$: Compared stress observable where recorded.
mesh/BC: Mesh and boundary-condition identity that must match.
$\Delta{}$: Difference measured against the declared tolerance.

01Freeze same model

02Run both solvers

03Compare observables

04Bound the claim

External solver comparison · method mapOnly after matching model data can solver outputs be compared as a bounded numerical correlation.

3 detailed steps

03 Step by step derivation

Follow the assumptions, equations and discrete operations in order. The full explanation stays visible.

1 · Make the comparison a same-problem comparison

Before comparing solutions, freeze coordinates, connectivity, material constants, supports, load resultants, units, solver assumptions, and output locations. Differences in any of these can dominate the solver discrepancy and make the comparison uninterpretable.

\mathcal M_{QF}=\mathcal M_{ref}\quad\text{for geometry, material, BCs, loads, and measured locations}

2 · Compare named observables with declared tolerances

Compare displacement, reaction, stress, or another explicitly recorded quantity at matched locations. Report the absolute and relative differences, the reference value, and the tolerance. Agreement in displacement does not imply agreement in stress recovery or every other output.

\Delta y=y_{QF}-y_{ref},\qquad e_{rel}=\frac{|\Delta y|}{\max(|y_{ref}|,s_0)}

3 · Keep numerical correlation separate from validation

The 0.2.10 Code_Aster 18.1 comparisons cover named linear-static TET4, HEX8, TET10, and HEX20 cases. They are not independent experiments or a universal accuracy ranking. The repository's failed G03 archive gate and WP14 hold remain part of the evidence record.