FRP Fatigue Literature Database (S-N)

Data are provided as normalized CSV tables and as a nested JSON representation. The release also includes the record schema (v0.21; v0.16 is retained unchanged for compatibility), a view schema (v5), field dictionary, coding manuals, provenance and quality documentation, known limitations, and Python scripts for validation, loading, basic S-N fitting, recomputing the tables of the accompanying data paper, and redrawing its figures.

The supplemental materials of that paper are deposited here as supplemental_materials.pdf, holding the two figures and five tables it cites as Figs. S1 and S2 and Tables S1 to S5.</p><p dir="ltr">The confirmed S-N dataset covers multiple FRP material systems, including GFRP and CFRP, and coupon-, bond-, reinforced-concrete- and strengthened-structure-related fatigue tests. Variable-amplitude data that could not be represented within the constant-amplitude S-N structure are excluded and documented separately.</p><p dir="ltr">Two accuracy measurements are reported.

A post-review random re-check of 50 data points yielded a value accuracy of 94.0% (95% CI: 83.5-98.7%). The judgement columns, which state what the source does and does not report, were re-extracted blind and adjudicated with the two readings hidden from the adjudicator; the residual rate of claims the source does not support is 1.56%, with no cell resting on an inferred or model-predicted basis. All released files are subject to automated checks for table counts, foreign-key integrity, CSV-JSON consistency, schema conformity, field consistency, exclusion evidence, agreement between a cell's value and the words of its basis, and release-manifest integrity.

Known limitations, including the four series columns that carry no per-cell record of what the source said, are listed in docs/known_limitations.md.</p><p dir="ltr">The dataset is intended for fatigue-data synthesis, model development and validation, comparative studies of FRP fatigue behavior, and data-driven applications. Users should consult the documentation and the known limitations before combining data across stress measures, units, material systems, or test configurations.</p><p dir="ltr"><br></p><p dir="ltr">This version adds one file, supplemental_materials.pdf, described above.

The figure and table numbers that scripts/make_figures.py and scripts/make_tables.py write were brought into line with the data paper after part of its material moved to the supplement: the completeness figure is now Fig. 1, the acquisition funnel Fig. S1 and the publication-year-by-application-form figure Fig. S2. Every label in the submission figures is 8 pt and the TIFFs are written without an alpha channel. Four statements about what release_manifest.md covers were corrected, in the manifest itself, in REPRODUCE.md and in README.md: the manifest lists every file except itself, which cannot carry its own checksum.

No data value changed in this version and every count is unchanged.</p><p dir="ltr">This version is delivered as a single ZIP archive of 58 files; release_manifest.md lists the other 57 with their size and checksum. Unpacking the archive reproduces the published layout, so scripts/validate.py runs from the unpacked folder without any path change. sha256 of the archive: 6be7b25861e92d0d410e966eee6e6de452cfe79c6c29a683ae53e710218954b0.

Data and documentation are under CC BY 4.0; the code directories scripts/ and audit/scripts/ are under the MIT licence, as stated in LICENSE-DATA.txt.</p>