Data and code for: Probabilistic reliability assessment and robust design of PET fiber mortar
Description
This dataset supports the manuscript “Probabilistic reliability assessment and robust design of PET fiber mortar.” It contains the original 27-run Box–Behnken experiment and 21 author-reported historical independent mixing batches representing 12 formulations, together with documented analysis code, frozen numerical inputs and calculation outputs aligned with manuscript revision v9. The older 12-record historical table is a subset of the 21 records. Repeated formulations F01/F02/F03 contain 3/6/3 batches. The calculations cover the reconstructed model pool and common-error screening, continuous-window numerical integration, retained interval and covariance diagnostics, source-specific external confirmation analyses, response/dose sensitivity and conditional cost comparisons. This is a transparent reconstruction where original statistical source code was unavailable; original measured values are preserved. Historical records are retrospective development evidence, not a newly conducted prospective blind test. Predictions, simulated scenarios and numerical draws are distinct from measured batches. Numerical uncertainty intervals do not include all model or physical-process uncertainty. The archive provides CSV and workbook discovery files, a data dictionary, module-specific reproduction instructions, source attribution and licensing notices, and a file inventory. It excludes the unpublished main manuscript, reviewer correspondence, private annotations, factory photographs/video/audio and third-party full articles. Original data/documentation use CC BY 4.0 and original code uses MIT; literature-derived material retains its source rights. See THIRD_PARTY_NOTICES.md for scope.
Files
Steps to reproduce
1. Download RINENG_data_code_v9.zip and extract it with its directory structure intact. Read README.md, DATA_DICTIONARY.md, THIRD_PARTY_NOTICES.md and analysis/README.md. The separately listed workbook and documentation are identical discovery copies, not extra observations. 2. Use Python 3.12. Run python verify_archive.py from the extracted archive root to check SHA256 integrity. Install each module's documented requirements. Use a copy of the archive and new output directories when recomputing scientific results. 3. Follow analysis/README.md for the model-pool fit/search/report workflow, continuous-window integration and its audit, external confirmation reproduction, cost arithmetic, restored-figure checks and retained batch/interval/bounded-metering analyses. The original 27 rows are the PET fitting set; the 21 retrospective batches represent 12 formulations. R3=0.0171 means 0.0171%. 4. Consult analysis/analysis_packaging_validation.json for computations actually repeated during packaging and previous validation sources. Full model training, complete domain search and complete continuous integration were not repeated during packaging. Some retained chains start from documented frozen caches; transparent reconstruction is distinct from recovering every unavailable original implementation. 5. Keep observations, published source predictions, model-based scenarios and numerical draws separate. Literature source DOIs and transcription locations are documented. Source full texts and factory media are excluded. Numerical/conditional results are not independent certification of production reliability.
Institutions
- Northwest Minzu UniversityGansu, Lanzhou