# Interactive powertrains / model scope

These are reduced-order browser models grounded in the articles' component connections. They do not reproduce the complete authors' MATLAB/Simulink solver or calibrated Vanguard performance map. The original figures are viewable from each workbench.

## Cruise ship
Figure 1 and §2.2: LH2 tank → cold-recovery heat exchanger → PEMFC → exhaust exchanger → residual-hydrogen boiler; a separate coolant recovery loop. The electrical companion follows §2.2: DC bus, propulsion, hotel loads, battery. Source: https://doi.org/10.3390/jmse14050431

Input mass flow m in kg/s; LHV = 120 MJ/kg. Chemical input Q = 120m MW. Reacted chemical energy is min(Q×utilization, 76.1/ηFC). Gross electric output is ηFC times reacted energy. Net output is 97% of gross (browser assumption). Recoverable FC heat is the user-selected fraction of reacted energy minus gross electricity, allocated 35% exhaust and 65% coolant (browser assumptions). Boiler useful heat = min(4 MW, 0.95 × residual chemical energy), following the §2.2 rating; excess residual chemical energy is explicitly unconverted, not declared recovered heat. This selection matters because different prose/results in the paper are not fully consistent about boiler peak power.

Primary-energy ledger: net electric + recovered FC heat + boiler heat + unconverted fuel + FC/converter/boiler losses = chemical input. Motor efficiency is 96% (assumption). Unserved electrical and thermal demand is shown. Cold recovery is a separate duty, m × 3.07 MJ/kg × 0.70 (paper Eq. 17), not an absorption chiller or an extra primary-energy term.

The user's 45 GJ store is 12.5 MWh; the battery is 0.3 MWh. Thermal transfer is limited to ±24 MW and battery transfer to ±0.2 MW (browser operating-envelope assumptions). Stores start at 50%, clamp at empty/full, and evolve using ΔE[MWh] = −P[MW]Δt[s]/3600. Positive storage power means discharge. Excess output and unmet demand are reported. No temperatures, PCM kinetics, thermal loss, dynamic FC ramp or DC-bus voltage solution is claimed.

## Vanguard
Figure 2: compressed H2, two fuel-cell modules, DC conversion, HV bus, battery bank, propulsion inverter/motor, HV auxiliaries, HV/12V converter and cooling loop. Source: https://doi.org/10.1016/j.ijhydene.2026.156680

Paper values: 23 kWh battery, 9.7 kg at 700 bar, 250 kW propulsion. The table lists 90 kW modules; prose describes 85 kW automotive ratings and marine operation up to 100 kW per module. The sandbox uses independent manual 0/off or 30–100 kW commands, not the real staged controller. It is not a hardware prescription.

Browser assumptions, editable where shown: FC efficiency 50% LHV; stack DC/DC efficiency 97%; stack auxiliary load 8 kW when running; other HV load 3 kW; delivered 12 V load 1 kW through a 92%-efficient converter; motor/inverter efficiency 92%; battery ±100 kW, 95% one-way efficiency and 10–90% SOC window. Initial SOC 60%, initial tank 9.7 kg. No invented map is presented as measured telemetry.

Battery bus power equals requested bus load minus FC DC output, bounded by power, SOC and timestep energy availability. Positive is discharge, negative charge. Available electrical power serves auxiliaries before motor demand. The motor delivers up to requested shaft power; deficits and curtailment remain explicit. H2 flow = gross FC kW/(ηFC×120000 kJ/kg). Finite tank inventory limits fuel-cell output. Every step maintains the DC-bus balance and clamps energy states at limits.

Continuous workbenches default to 1× wall-clock time, with optional labeled 10×/30× playback; there is no required manual time step. Flow animation illustrates direction, not measured velocity. Preview shortfalls block integration. Reset returns cruise stores to 50% and Vanguard battery to 60% charge.

## Home and other visuals
Home is a research narrative across fission, electricity-to-hydrogen, fuel cells, ATLAS deposition, marine applications and ammonia. These are related research domains, not one demonstrated plant. Electrolysis is shown between electricity and H2; H2 is a feedstock for ammonia. ATLAS fabrication is a change of perspective, not a downstream product of an operating fuel cell.

The boiling artwork retains the previous constant-property water/Qu–Mudawar/Davis–Anderson teaching model; moving bubbles are stylized rather than CFD. The original reactor interfaces are restored from commit 247e815be469ea53b687e0887260fc6a3d00f624: point-kinetics training replay and delayed-neutron/lineage field. Their previous limitations remain visible. No live reactor controls or operating instructions are implied. Records is a brief WebReg hackathon story plus decorative traffic, not telemetry.


## Live Instruments interaction update
The equations above are retained. The workbenches now run continuously at 1× real time by default, with optional explicitly accelerated 10×/30× playback. Ten-second stepping and the earlier 15-minute playback stop are retired. Before committing a step, **every** requested duty must be met. Otherwise red preview shortfalls appear and the state/time freeze until the controls are rebalanced. This user-requested teaching gate is not an operating controller. Details: LIVE_MODELS.md.

## V3 superseding cruise model
The active cruise workbench now uses `voyage-model.js`, not the archived `PowertrainModels.cruise`. Its chemical conservation ledger is HHV; FC efficiency remains LHV. It includes a finite shared H₂ tank, voyage clock and delivered-power distance accounting. The revised exhaust routing, catalytic-burner basis, storage directions and assumptions are fully documented in [SCIENTIFIC_REVIEW.md](SCIENTIFIC_REVIEW.md). The Vanguard pure function remains unchanged; its diagram now resolves both fuel-cell/DC-DC branches and the battery connection explicitly.


## Home-film audit: net installed rating
The active `VoyageModel.cruise` net rating is now 76.1 MW, matching §2.4.4. With the retained assumed 97% conversion efficiency, reacted mass is limited by `76.1 / (0.97 × 120 × etaFC)`. The archived `PowertrainModels.cruise` description above does not govern the active workbench. Matching this rating and algebraic ledgers is not independent validation of off-design outputs.