Reproducibility package and public evidence for cross-solver finite-element mesh conversion and canonicalized validation

Overview This archive provides the public reproducibility package and evidence addendum accompanying the manuscript Cross-solver conversion and canonicalized validation of 3D solid finite-element mesh data using a neutral hierarchical intermediate representation. It supports inspection, integrity verification, and independent analysis of the publicly redistributable benchmark subset used to evaluate cross-solver conversion and canonicalized validation of 3D solid finite-element mesh data across Abaqus, ANSYS Mechanical/MAPDL, MSC NASTRAN, and Z88 formats. The archive includes public benchmark inputs, canonical reference and converted-route JSON artifacts, route-level result tables, raw stage-timing observations, parser/writer change reports, licensing and provenance documentation, and a complementary public evidence addendum. Archive structure Reproducibility-package/Primary public-subset reproducibility package for the Z88-pivot-canonicalized cross-solver benchmark. Evidence_addendum/Complementary public evidence covering direct FE-IR/JSON structural cross-checks, controlled fixtures, perturbation guardrails, and set-policy interpretation checks. Benchmark coverage and redistribution scope The full benchmark comprises 23 finite-element models and 92 evaluated model-level conversions across the four supported solver formats. Raw benchmark inputs are redistributed for the 10 models for which redistribution is permitted. Three Abaqus examples are available only to licensed users through the Abaqus Fetch utility and are therefore not redistributed. Proprietary or otherwise restricted raw decks, as well as complete derived-geometry or mesh artifacts based on those restricted inputs, are not included. Although public raw inputs are limited to the redistributable subset, the archive includes route-level summary and key-metric tables for all 92 evaluated conversions. Released contents The archive contains: 10 public raw inputs for redistributable benchmark models; 10 canonical reference model.json artifacts; 40 canonical converted-route model.json artifacts, covering 10 public models and four target formats; 315 parser/writer change reports documenting normalizations, target-compliance interventions, and released diagnostic information; route-level summary and key-metric tables for all 92 evaluated conversions; 5 released attention/diagnostic rows; 1,610 raw invocation-level stage-timing observations, including warm-up runs; publicly shareable converted solver decks where redistribution is permitted; controlled fixture oracles, perturbation guardrails, direct-path summary tables, and set-scope policy examples contained in the evidence addendum; and file-level manifests, SHA-256 checksums, provenance notes, and verification instructions. Terminology note The historical directory name json_roundtrip/ and compatibility-retained fields such as phase=roundtrip refer to Z88-pivot-canonicalized converted-route artifacts. Reproducibility and interpretation boundaries The released materials support inspection, checksum verification, and independent analysis of the public inputs, JSON artifacts, result tables, timing observations, and diagnostic reports. They do not constitute a complete end-to-end rerun package for the full 23-model benchmark. A complete rerun from public raw inputs alone is not possible because the converter implementation and restricted benchmark inputs are not included in this archive. The separately released evaluation software provides the public implementation of the primary structural-signature metrics and the reported three-stage validation-workflow runtime summaries. It is an evaluation package and not the mesh converter. Recomputing the metrics directly from the released JSON artifacts requires additional execution harnessing that is not included in this dataset archive. Integrity verification From the extracted archive root, verify the complete package with: sha256sum -c CHECKSUMS.sha256 The evidence addendum also provides a standard-library-only Python integrity checker: cd Evidence_addendum python tools/verify_addendum_integrity.py Release note This release updates reader-facing documentation, archive navigation, licence/provenance naming, and integrity metadata. The benchmark data, canonical JSON artifacts, result CSV payloads, metric values, and numerical results are unchanged. Earlier published versions remain accessible through the Zenodo version history. Related records Reproducibility dataset and evidence package: https://doi.org/10.5281/zenodo.21843668 Evaluation software, version 1.0.1: https://doi.org/10.5281/zenodo.21848108 Evaluation-software source release: https://github.com/EinmalmitProfis/cross-solver-fem-mesh-eval/tree/v1.0.1 Licensing and provenance Redistributed third-party inputs remain subject to their original licenses or documented redistribution permissions. Per-model provenance, applicable rights, and accompanying license files are documented in: Reproducibility-package/licences/THIRD_PARTY_NOTICES.md No broader rights are granted for third-party materials.