Fuel-cell operators lose money when…
transient voltage swings shorten stack life, degradation is poorly measured, warranty disputes are hard to settle, and safety margins stay conservative because the internal state is uncertain.
The Continuum Fuel Cell Regulator (CFCR) is a deterministic 1–10 ms overlay supervisory layer that runs alongside an existing controller architecture for fuel-cell and electrolyzer systems — sensorless diagnostics in, hash-chained interval receipts out.
✓ Public-dataset Stage-1 validation with verifier-ready provenance.
Selected Stage-1 proof anchors — measured on public datasets, not live-OEM deployment claims.
CFCR is a real-time control, diagnostics, and evidence layer for hydrogen fuel cells and electrolyzers. It rides on top of your existing controller — it does not generate hydrogen, change your chemistry, or make any energy-balance claim.
CFCR can start in passive observer mode with zero actuator authority — no firmware replacement, no added specialty sensors, and no stack redesign — a lower-friction path for observer-mode evaluation before any safety-critical authority is considered, and taken on only when your own pilot criteria are met.
transient voltage swings shorten stack life, degradation is poorly measured, warranty disputes are hard to settle, and safety margins stay conservative because the internal state is uncertain.
hydrogen production records are hard to verify, compliance reporting is costly, system health is opaque, and efficiency drifts over time with no clear signal.
governing operation every few milliseconds, estimating hidden internal state without extra sensors, enforcing machine-verifiable safety envelopes, and generating tamper-evident operational records — all from signals the stack already produces.
CFCR runs alongside an existing fuel-cell or electrolyzer controller, computes sensorless diagnostic kernels from standard stack signals, and emits cryptographically hash-chained interval receipts — without changing actuation authority unless and until you authorize it.
| Conventional controller | CFCR |
|---|---|
| Reactive control from setpoints | Physics-aware control from estimated internal state |
| Limited diagnostics; needs added sensors | Sensorless state estimation from standard signals |
| Logs events after the fact | Cryptographically proves each decision in-loop |
| Safety assumed and margined conservatively | Safety-envelope compliance evaluated interval by interval |
| Warranty analysis is difficult and disputable | Auditable, tamper-evident operating history |
| Compliance handled by separate systems | Evidence generation built into the control loop |
Every row below is measured on independent public data (IEEE PHM 2014 FCLAB, CC BY 4.0; NREL DOI 10.7799/3007836; University of Michigan GDL Dec17). The reports, provenance carriers, SHA-256 manifests, and one-command verifier are available under NDA — the claim is not “believe our numbers,” it’s “recompute them yourself.”
| What | Result |
|---|---|
| Cumulative-stress durability | CFCR cumulative-stress index vs measured voltage degradation on the FCLAB 1000-h tests: r = 0.87 on FC1 (constant load) and r = 0.82 on FC2 (5 kHz ripple), across 271,232 samples. Outperforms four baseline models under Steiger correlated-correlation testing at p < 0.0001 on both stacks. Report 02 · IEEE PHM 2014 FCLAB |
| Envelope holdout discipline | FC1→FC2 train/holdout split. Under a binding 500 V/s bus limit, peak |dV/dt| reduced 901.5 → 512.4 V/s (43.2%) with zero envelope violations across five baseline ramp rates (500–10,000 A/s) on both training and holdout sets. FC1-derived limit + 10% margin (3.080 V/A) covers the FC2 holdout maximum (3.000 V/A). Report 07 · IEEE PHM 2014 FCLAB |
| No-humidity-probe hydration Claim 23 enablement | Standalone HYDRA v2 kernel computes λ_mem and H_virt from {V, I, T, P, flow} only — no humidity probe, RH sensor, dew-point sensor, or EIS module at any interval. Specification Table X matches reference outputs to 3 decimals across 7 rows; an independent notebook reimplementation diff-matches to 6 decimals at all 14 convergence checkpoints. Reports 09 + 10 |
| Deterministic timing & receipt chain | 10,800-interval receipt stream at a 10 ms target cadence, HMAC-SHA256 authenticated, full chain continuity. P99 solve 5.88 ms, zero overruns. ACCEPT/DERATE/REFUSE exercised (8,653 / 2,146 / 1). Report 06 · receipts_v2.jsonl |
| Electrolyzer production record | 12,006 one-second intervals across 8 NREL MW-scale runs. Frozen-rule deterministic methodology disclosed before processing: estimated H₂ 18.221 kg vs Coriolis-measured 18.261 kg → mass-balance relative error 0.22%, far inside the 20% tolerance. PASS Cryptographic chain integrity verified (zero breaks across 12,005 links). Report 08 · NREL DOI 10.7799/3007836 |
| Sensorless proxy behavior | FLOOD proxy (air-side pressure σ over 100-sample windows) correlates with the flooding-risk index at r = 0.987. PION degradation-proxy tracks the FC2 voltage-aging trajectory at r = 0.9989 using standard channels (no EIS). Presented as proxy tracking — not future-life prediction. Report 06 |
| Audit-grade verifiability | 11-artifact diligence pack with SHA-256 manifest and one-command verifier (verify_all.py): four gates — file integrity, report/provenance binding, sensorless structural posture, receipt-chain continuity. Current pack: OVERALL PASS · 10,834 receipts across embedded JSONL streams. Master Provenance Manifest · under NDA |
| Capability | Simulation | Public-data replay | Hardware-in-loop | Live stack |
|---|---|---|---|---|
| Deterministic timing / overrun firewall | ✓ | ✓ | software only | pending |
| Sensorless hydration (HYDRA) | ✓ | ✓ | pending | pending |
| Flooding / degradation proxy | ✓ | ✓ | pending | pending |
| Safety envelope & ACCEPT / DERATE / REFUSE | ✓ | ✓ | pending | pending |
| Cryptographic receipt chain | ✓ | ✓ | pending | pending |
| Electrolyzer production records | — | ✓ | pending | pending |
CFCR turns recurring operational problems into outcomes a non-engineering stakeholder can act on:
| Problem | Consequence | CFCR role |
|---|---|---|
| Transient stack stress | Reduced stack lifetime | Per-interval envelope enforcement |
| Warranty disputes | Slow, costly adjudication | Tamper-evident interval receipts |
| Compliance uncertainty | Delayed 45V / incentive claims | Verifiable production records |
| Conservative safety margins | Lost utilization | Better hidden-state visibility |
| Added diagnostic sensors | Higher BOM and failure points | Sensorless estimation — no EIS or humidity probe |
A deterministic 1–10 ms supervisory loop under the existing controller (Observer / Advisory / Full-Authority). Cryptographic per-interval evidence supports an ISO 26262 documentation pathway for bus, truck, and rail.
Cryptographically authenticated per-interval receipts that support DNV / Lloyd’s / ABS audit-trail workflows for marine-class and stationary deployments.
Hash-chained receipts plus CHI / SHI fleet indices give tamper-evident operational evidence for warranty adjudication, usage-based insurance, and residual-value underwriting — verifiable without exposing controller internals.
Sub-hourly production records with accumulator root, energy-attribute pointer, and completeness metadata map onto MRV / 45V-style workflows and EnergyTag-style granular certificates — subject to independent program eligibility review.
No dedicated EIS / impedance module required — a direct per-stack BOM reduction and one fewer class of drift-prone sensors. (Per-stack savings modeled in Report 06.)
Overlay-first: runs as a virtual observer alongside the production controller, with no change to actuation authority until pilot criteria are met.
CFCR’s deterministic control loop and safety-envelope enforcement operate locally; fleet analytics and evidence aggregation are optional layers and do not make safety dependent on cloud availability.
Chemistry coverage. Chemistry-agnostic via a parameter vector θ_chem: PFSA ionomers (Nafion), hydrocarbon sulfonated membranes, high-temperature PEM (HT-PEM / PBI, hydration interpreted as an acid-saturation index), and PGM-free cathodes (Fe-N-C), each with chemistry-specific degradation coefficients and acceptance thresholds.
We separate what is measured on public data from what is modeled under stated assumptions. This discipline is deliberate — it is what makes the evidence above worth re-checking.
Because the CFCR application is already filed and technically developed, we are approaching a limited number of technically sophisticated hydrogen companies before patent issuance for preliminary roadmap-fit and technical-diligence discussions. We are not seeking an immediate binding acquisition, license, or commercial commitment at this stage. The purpose is to determine whether CFCR is relevant to your organization’s fuel-cell or electrolyzer roadmap and, if so, to proceed through the appropriate NDA-gated technical review process.
The public page shows the architecture, selected proof anchors, deployment posture, and current validation status. Under NDA, the recipient receives a reproducible evaluation package: implementation-level reports, provenance carriers, verifier tooling, detailed claim language, and integration-level technical materials. The purpose is not to ask for trust; it is to let a qualified technical team recompute the evidence and evaluate CFCR against its own fuel-cell, electrolyzer, fleet, warranty, or compliance workflow.
If CFCR looks relevant to your roadmap, the next step is a short fit call to determine whether the architecture maps to your fuel-cell, electrolyzer, fleet, warranty, or compliance workflow.
On a signed NDA, the technical review opens a reproducible observer-mode evaluation package: deterministic replay tooling, provenance carriers, verifier materials, detailed reports, the complete specification and claim set, and integration-level technical disclosures. One proposed next step after review is a 2–4 week observer-mode evaluation using historical or simulated stack data, with no controller changes and no actuator authority.
Designed for fast proof/disproof — run first as observer-mode analysis on existing telemetry, before any control authority is considered.
Download the non-confidential Stage-1 Credibility Pack: public proof anchors, validation maturity, observer-mode evaluation workflow, NDA review inventory, a small runnable receipt-chain demonstration, and a public-data mass-balance recompute script, and a non-confidential 45V evidence-engine pilot proposal. The full Diligence Fortress Review, detailed reports, verifier tooling, provenance carriers, claim language, and implementation-level materials are available during NDA-gated technical review.
Download Stage-1 Credibility PackReview path: public Stage-1 materials are available now. If there is qualified interest after a fit call, a Diligence Lite package may be provided under a signed NDA or counsel-approved controlled disclosure. Full technical diligence — including the complete Diligence Fortress Review, full reports, verifier tooling, provenance carriers, claim language, and implementation-level materials — is reserved for full NDA review.
To proceed, simply reply to the email that accompanied this page. The full specification, complete filed claim set, and verifier-ready provenance pack are provided under NDA.