What these documents change in the model
Nine sources — the CAA Hydrogen Challenge set (sandbox, refuelling bowtie, HSE regulatory review), Project NAPKIN, and two journal papers on liquefaction and methane accounting. Read together they do three things: they validate the rev-2 expansion of the data request, they supply concrete numbers we can adopt now, and they expose gaps the model does not currently cover — chiefly that our turnaround has no purge, leak-check or inerting phases, and that "boil-off" is not the same thing as "emitted".
The three findings that matter most
- Our turnaround model is missing safety-critical phases. The CAA bowtie lists purge verification, earthing/bonding, pre-refuel inspection, interlocks and leak detection as preventive controls. Every one of those consumes stand time. Our six phases (disembark, clean, pre-cool, fill, vent, board) contain no purge, no leak-check, no bonding step — so even the 16.1 min dual-hose case is optimistic against a CAA-compliant ConOps.
- "Loss" ≠ "emitted", and Qdot proves it. Cryo-compressed storage absorbs heat ingress as pressure rise instead of venting. Our model treats the 2% fill vent and Dewar boil-off as mass leaving the system; the atmospheric fate split is an assumption, not a result — which is exactly why data-request item 7 asks BRS what they can measure.
- NAPKIN already built the layers we are missing. Its "five As" whole-system approach with an airline behaviour model, operating-cost model and hydrogen demand model is precisely the Demand and Cost layers absent from brsupercool. We should reuse its structure rather than reinvent it.
Source documents — click to read in place
CAP 3284A — Bowtie Analysis: Aircraft Refuelling
Defines the standard hazard HC.LH2 1.3 (SP) and a single top event: loss of safe
operating conditions during hydrogen refuelling. Five threats, eight preventive controls, six
consequences, ten mitigative controls. The CAA intends this as the common reference framework — aligning
to it is effectively mandatory for a defensible trial ConOps.
CAP 3284 — Hydrogen Sandbox Summary Report
Thirteen projects including Heathrow P601 (DSEAR separation distances, ATEX Zone 1 above vents), Exeter Zero Carbon Turn (winter HyGPU trial), Qdot (cryo-compressed storage), Ultima Forma, ZeroAvia, Rolls-Royce. The Exeter data-yield lesson is important for our metering ask: of 33 HyGPU operations, only 15 produced usable data.
CAP 3284E — Regulations Relevant at Aerodromes Utilising Hydrogen
The full UK regulatory stack: COMAH 2015, DSEAR 2002, PESR 1999, PSSR 2000, PUWER, LOLER, COSHH, RIDDOR, CDM 2015, Carriage of Dangerous Goods, and the Planning (Hazardous Substances) Regs 2015 — plus collated codes and standards for cryogenic storage, transportable vessels, gaseous storage, fuelling, sampling and electrolysis. Our model currently represents exactly one of these (COMAH tier).
CAP 3284F — Regulations Infographic
Single-sheet map of which regulation bites where across the aerodrome hydrogen chain. Useful as the cover exhibit for the D1 assumptions register and for the PXLICE validation pack.
CAP 3284D — supporting material (HTML)
Companion web document in the CAP 3284 series, retained here so the evidence set is complete and self-contained for the validation workshop.
Project NAPKIN — Final Report
The UK whole-system study: "five As" (Aircraft, Airlines, Airports, Airspace, Air Passengers) with an airline behaviour model, operating-cost model and hydrogen demand model across 7–90 seat concepts. Finds ZEF viable on sub-regional routes from mid-decade, retrofits first, clean-sheet by 2035, full UK regional fleet replacement plausibly cost-effective by 2040. This is the template for our missing Demand and Cost layers.
Project NAPKIN — Report Pages / figure set
The figure and exhibit set accompanying the final report — route maps, fleet concepts, demand build-up and infrastructure schematics. Useful source material for the D5 report and ACT presentation.
Bae et al. (2021) — Multi-objective optimisation of H₂ liquefaction integrated with LNG
Mixed-refrigerant precooling 298→80 K followed by four-stage hydrogen Joule–Brayton cascade 80→20 K, with ortho–para conversion handled explicitly across the 80–20 K stages and LNG cold integration as the efficiency lever. Directly relevant to task T3.7 and a genuine external check on our 11 kWh/kg baseline and 6.03 kWh/kg TPMS point.
Alvarez et al. (2012) — Greater focus needed on methane leakage (SI)
Fuel-cycle methane leakage accounting and the technology-warming-potential framing — the methodological basis for treating a light, leaky gas as a climate agent over time rather than a single GWP100 number. Underpins both our blue-hydrogen upstream figure (3.9 kg CO₂e/kg) and the treatment of H₂ itself as an indirect greenhouse gas in the loss-fate work.
Extracted assumption register — 24 items against the model
| Assumption / requirement | Source | Status | What it means for brsupercool |
|---|---|---|---|
| Standard hazard + top event: HC.LH2 1.3 (SP); "loss of safe operating conditions during H₂ refuelling" | CAP 3284A | gap | Adopt the CAA hazard statement verbatim in the D1 register so our ConOps is comparable to every other UK trial. No model object represents it today. |
| Preventive controls consume stand time: pre-refuel inspection, earthing/bonding, interlocks, purge verification, leak detection | CAP 3284A §3.28 | gap | Highest-priority model change. Add bond, leak_check and purge phases to the DES. The 16.1 min dual-hose result is optimistic until they exist. |
| Helium purge of hose and tank before transfer, plus post-transfer purge | FlyZero p24 · CAP 3284A | gap | A per-turn helium consumable, a cost line, and a phase. Currently absent from both the turnaround and the cold chain. |
| Mitigative controls: ESD, fixed+portable H₂ detection, UV/IR flame detection, RFFS procedures, stand evacuation, separation distances | CAP 3284A §3.37 | partial | We model exactly one of ten — separation distances (the 20 m / 8 m circles). The rest are infrastructure cost and ConOps items for D4. |
| DSEAR separation, Heathrow P601 precedent: ≥8 m from buildings, ≥5 m from ignition sources, ATEX Zone 1 above vents, no overhead obstruction, marked parking area | CAP 3284 §4 | partial | Concrete, adoptable numbers — but note these are gaseous-truck figures. Use as the bowser-parking envelope; do not conflate with the LH₂ transfer zones. |
| Full regulatory stack beyond COMAH: DSEAR, PESR, PSSR, PUWER, LOLER, COSHH, RIDDOR, CDM 2015, CDG, Planning (Hazardous Substances) | CAP 3284E | partial | We model COMAH tiering only. DSEAR/ATEX zoning is the one that actually decides the parallel-activity question in data-request item 3. |
| COMAH thresholds 5 t lower / 50 t upper; MAPP, safety report, emergency plans | CAP 3284E · FlyZero p54 | in model | comah_tier(); 14 t site inventory → lower tier, reported in every scenario brief. |
| Safety zones 20 m connect/disconnect, 8 m during flow (2030+ basis) | FlyZero p50 (HSE) | in model | exclusion_connect_m / exclusion_flow_m, drawn on the apron view and phase-aware in playback. |
| Fill rates: 6" hose @ 5 m/s ≈ 5 kg/s; dual-hose for narrowbody | FlyZero Table 12 | in model | ops_presets — flyzero_single_6in / flyzero_dual_6in. |
| Cryo-compressed H₂ absorbs heat ingress as pressure rise instead of venting; reduced boil-off; refuellable from LH₂ or GH₂ | CAP 3284 §5 (Qdot/NCC) | gap | Directly challenges "boil-off ⇒ vented". Our boiloff_frac_per_day should route through a fate split, not straight to atmosphere. |
| Loss fate must be measured, not assumed; H₂ is an indirect GHG | Alvarez 2012 · HyFlux LCA | gap | Add vented / recovered / re-liquefied split with uncertainty. This is data-request item 7's whole purpose. |
| Methane leakage / technology-warming-potential accounting for fossil pathways | Alvarez 2012 (PNAS SI) | partial | Our blue pathway carries a flat 3.9 kg CO₂e/kg upstream. Should become leakage-rate driven with a time-dependent forcing basis. |
| Ortho–para conversion handled across 80→20 K stages in a four-stage Joule–Brayton cascade | Bae 2021 · Krasae-in | gap | Confirms task T3.7's placement. Conversion heat (~530 kJ/kg) exceeds latent heat, so it belongs inside the liquefier model, not as a post-hoc correction. |
| Cold integration (LNG or equivalent) as an efficiency lever; mixed-refrigerant precool 298→80 K | Bae 2021 | gap | A liquefier architecture we don't represent. Relevant if Avonmouth has any cold stream available — worth one line in data-request item 10. |
| Liquefier specific work benchmarks and multi-objective (energy vs cost) optimisation | Bae 2021 | in model | External check on our 11 kWh/kg baseline / 6.03 kWh/kg TPMS point and the ε figure of merit against the REFPROP-verified 3.495 kWh/kg ideal. |
| Whole-system "five As": Aircraft, Airlines, Airports, Airspace, Air Passengers | NAPKIN | partial | We cover Aircraft + Airports. Airlines (behaviour/scheduling) and Passengers are absent — the Demand layer of the rev-2 mapping. |
| Airline behaviour + operating-cost models as first-class components | NAPKIN §3–4 | gap | Reuse NAPKIN's structure for layer 6 rather than inventing one. Feeds data-request item 14. |
| Retrofit-first adoption path, clean-sheet by 2035, fleet replacement by 2040 | NAPKIN | gap | Gives a defensible adoption S-curve for the residual Jet-A pool (item 13) instead of an arbitrary share. |
| Operating cost not a barrier by the 2040s for the NAPKIN fleet | NAPKIN exec summary | gap | A citable anchor for the cost/feasibility layer — and a useful counterweight in the ACT narrative. |
| Airport infrastructure is not the binding constraint; national distribution network is | NAPKIN policy findings | partial | Supports the Avonmouth corridor framing: our sourcing-route result is the strategically interesting one, not the apron result. |
| Trial data yield is poor: 33 HyGPU operations → 15 usable after cleaning | CAP 3284 §3 (Exeter) | gap | Plan metering for ~2× the needed sample and specify logging rate up front in item 7. A validation-harness requirement, not a nice-to-have. |
| Dormancy affects performance more than ambient temperature (HyGPU winter trial) | CAP 3284 §3 (Exeter) | gap | Reinforces FlyZero's 10 h pre-flight / 3 h post-flight dormancy limits. Our model has no dormancy state at all. |
| Fuel-cell thermal management is bidirectional — overcooling causes permanent degradation | CAP 3284 §2 (CAeS) | gap | Cross-check for the cold-synergy claim: LH₂ as both fuel and coolant has a floor as well as a ceiling. Relevant to the coil/FC architecture assumptions. |
| Hybrid battery + fuel cell needed for CS-E 745(a)(3) acceleration and CS 23.77 baulked landing | CAP 3284 §2 (CAeS) | partial | Our architecture field offers fc_electric / fc_hybrid; the certification rationale for hybrid is now evidenced rather than assumed. |
What we will do with this — proposed backlog additions
- v0.5-A · CAA-compliant turnaround. Add bond / leak-check / purge phases and a helium consumable; re-run all three fill cases. Expect the 16.1 min figure to rise — better an honest number than a headline one.
- v0.5-B · Loss-fate split. Replace "boil-off ⇒ emitted" with vented / recovered / re-liquefied fractions carrying uncertainty, and report H₂ as an indirect GHG alongside CO₂e.
- v0.5-C · Ortho–para in the liquefier. Conversion heat inside the cycle model using ORTHOHYD/PARAHYD from REFPROP, benchmarked against Bae et al.'s cascade.
- v0.5-D · Demand layer from NAPKIN. Adoption S-curve + residual Jet-A pool, so hydrogen displaces activity rather than adding to it.
- D1 register. Adopt the CAA hazard statement and cite CAP 3284E as the regulatory basis — cheap, and it makes the PXLICE pack materially stronger.