Quality & trust
Methods, sources and AI use
- R1_Committee_Brief.docx
- R2-R6_Research_Briefs.docx
- R7-R12_Research_Briefs.docx
- R13-R18_Research_Briefs.docx
- Spacecraft_Fire_Safety_Research_Summaries_Combined.docx
scripts/extract-findings.mjs reads the documents; nothing is typed in by hand.
pnpm verify fails the build if any value or citation is not found in the briefs.
Proxy vs real: the evidence legend
Fire burned in actual microgravity: Skylab, Mir, ISS, Cygnus (Saffire), drop towers, parabolic flights.
Ground tests in 1g that imitate low gravity — low pressure, forced flow, narrow channels. Useful but not the real thing.
Numerical simulation (FDS, lumped-capacity models). Only as good as the data that validates it.
Plans, proposals and summaries that cite other results.
Ranking rubric
All four criteria are computed from the briefs; weights are yours to change on the Rankings page.
| Safety relevance | Share of the paper's findings about extinguishing, detection, toxic gas, limits or crew hazard. |
| Realism | Evidence type (real µg > overview > ground proxy > model) and test platform scale (Cygnus > ISS > Mir > … > model). |
| Recency | Publication year, 1993 → 2025. |
| Data richness | Number of findings that report a measured value with units. |
| Evidence weight | real 1 · overview 0.55 · proxy 0.45 · model 0.3 |
| Platform scale | Cygnus 1ISS 0.9Mir 0.8Skylab 0.75Parabolic 0.5Drop tower 0.45Ground 0.35Model 0.25 |
| Safety keywords | extinguish|suppress|detect|smoke|\bCO\b|CO2|CO₂|crew|hazard|safety|habitab|vent|margin|toxic|danger|limit |
| Measured value | \d+(?:\.\d+)?\s?(?:%|cm\/s|mm\/s|kPa|psi|W\b|kW|s\b|sec|ppm|mmHg|°C|K\b|kJ|m³|cm\b|mm\b) |
Confidence meter
Per finding: real microgravity + a measured value. Per insight: several independent real-µg sources agree.
Real-µg qualitative statement, or a measured ground-proxy value. Insights from one team or short-duration tests.
Model, plan/overview, or qualitative ground statement.
Conflicts between papers
Both are true — the deciding variable is airflow. Still air starves µg flames; ventilation flows of a few cm/s make them more dangerous than on Earth.
Shutoff works for thin fuels in air; in enriched oxygen or on hot thick fuels it may not, and a small residual flow can help cool the material.
CO₂ behaves differently once radiation dominates. Helium deserves study for exploration vehicles.
Use low-pressure tests to screen and to estimate spread rate, not to predict flame colour, soot or toxic products.
Likely two different measurement methods (cabin vs fuel-consumption based). The dashboard shows both values as reported and uses 3217 W for the headline.
Missing-data warnings
Values that exist only inside a paper's plots, or are not stated at all.
AI usage disclosure
Required by the organisers: every AI tool used, what for, and where.
| Tool | Used for | Where | Not used for |
|---|---|---|---|
| Claude Code (Anthropic, model Claude Opus 5.5) | Building the dashboard: writing the Next.js/GSAP code, the data-extraction and verification scripts, chart and page layout, styling and animation, and the short one-line card summaries written from the team's briefs. | All files under src/, scripts/ and the project setup. | Choosing the research topic, selecting and reading the papers, writing the R1–R18 briefs, the dashboard feature list or the Flame Lab idea — these came from the team. No AI runs inside the dashboard. |