Machine · hardware

Hardware, systems & operations.

The machine, component by component — from the centerpost out to the converters, with the engineering blueprints, the multiphysics behind them, and how the plant is controlled, protected and maintained.

Hardware · Breeder

Hardware Systems

The engineered systems of the Hyperion breeder — magnets, blanket, tritium plant, first wall. Click any system for its measurements and provenance.

Magnet Systems

Toroidal field

8.0 T on-axis

REBCO HTS

16 TF coils + central solenoid, high-temperature superconductor.

Peak coil field

16.84 T on-conductor

≤ 20.1 T limit

On-conductor peak field with 3.3 T margin to the [HAR23] limit.

Blanket & Fuel Cycle

Blanket TBR

1.8 target

Be / Li blanket

Demonstrated Be blanket TBR 1.34 → ~1.6 kg/yr; 1.8 is the target lever.

Tritium plant

4.0 kg/yr

~2× world flow

Net tritium at TBR 1.8; startup inventory inside civil precedent.

Thermal & Structure

First-wall load

1.97 MW/m²

damage-life limited

14 MeV neutron load → 19.7 dpa/fpy → ~2.5 fpy component life.

Divertor

W monoblocks

tungsten

Tungsten monoblock divertor assembly, lower single-null.

Plasma & Confinement

Plasma current

9.66 MA

boot + driven

Flat-top current; bootstrap 1.50 MA (f_bs 0.152), remainder current-driven.

Energy confinement

0.374 s (τ_E)

H₉₈ 1.0

IPB98(y,2) at H₉₈ 1.0 — confinement quality dominates the uncertainty.

Normalised β

1.53 β_N / 4.2

stable margin

β_N well below the NSTX no-wall limit of 4.2.

Heating & Current Drive

Auxiliary power

25.9 MW

NBI + RF

External heating + current drive; sets the Q = P_fus/P_aux gain.

Ion temperature

15 keV (T_i0)

core

Core ion temperature at the frozen point; Greenwald fraction 0.30.

Vacuum & Cryogenics

Cryogenics

4.5 K

HTS cold mass

Superconducting cold mass at 4.5 K; ice-piston preload offsets Lorentz load.

Vacuum

1.2×10⁻⁸ Pa

UHV

Ultra-high vacuum in the cryostat enclosure.

Breeder deposit · DOI 10.5281/zenodo.21746157 · reproduced from the open Zenodo/GitHub deposit (CC BY 4.0). Canonical freeze: FROZEN_2026-08_PUBLICATION.csv.

🔒 Hardware Systems is Kronos-only. The overview above is public — the full data, tables, live calculations and engineering detail on this page require a Kronos team sign-in.

Hyperion breeder · plant systems

Systems & Balance of Plant

The plant beyond the reactor core — the subsystems that fuel it, cool it, heat it, pump it, convert its power and keep it safe. Each is graded honestly: FROZEN (from the deposit), SCREENING (reduced-order), or AWAITS HARDWARE.

Fuel cycle & tritium plant

SCREENING

Stores, purifies and recycles the hydrogen isotopes; extracts and banks helium-3.

Breeder: Tritium output 4.0 kg/yr (TBR 1.8 target) + He-3 co-production; cryo-distillation; HazCat 2 confinement, multiple barriers.

Cryogenics plant

SCREENING

Cools the superconducting magnets and cold mass to operating temperature.

Breeder: 4.5 K magnet cooling + 20 K stage; cold mass ≈ 608 t; He refrigeration.

Heating & current drive

MODEL

Brings the plasma to temperature and (breeder) drives the plasma current.

Breeder: P_aux ≈ 25.9 MW (NBI + RF); current drive supports I_p 9.66 MA (bootstrap-assisted).

Vacuum & pumping

MODEL

Holds ultra-high vacuum in the vessel and pumps exhaust.

Breeder: Base pressure ≈ 1.2×10⁻⁸ Pa; cryopumps + roughing; conformal to the D-vessel.

Power conversion & balance of plant

FROZEN

Turns fusion output into electricity.

Breeder: Thermal path — blanket heat to a power cycle (product is fuel/isotopes, not net grid MWe).

Power supplies & converters

MODEL

Feeds the coils, heating and plant loads; conditions grid interface.

Breeder: CS/PF converters, magnet supplies, plant auxiliaries.

Remote handling & maintenance

AWAITS HARDWARE

Services activated in-vessel components without human entry.

Breeder: Hot-cell + remote handling for the tritium/neutron-activated breeder; the higher-activation machine.

Instrumentation & control

MODEL

Senses plasma & plant state and holds the machine in its safety envelope.

Breeder: Diagnostics + plant control (PCS) and safety instrumented systems (SIS) — see the I&C · Control page.

Site, buildings & shielding

MODEL

Houses the machine and provides biological/neutron shielding.

Breeder: Reactor building + shielding within a HazCat 2 licensing envelope; remote/isolatable siting.

Balance-of-plant systems for the breeder, derived from the frozen design; full plant engineering is future work. No economics on this page.

Blueprints · Breeder · representative · sheet set (11)

Engineering Blueprints

Dimensioned drawings derived from the frozen design parameters. Representative — schematic-accurate from our numbers, not a construction drawing; a CAD-accurate set follows once CAD/STEP files are provided.

CAD models · real OpenCASCADE B-rep (generated from the frozen parameters)

HYPERION breeder

STEP · parametric solid

Center stack · plasma · blanket · vessel · PF coils · divertor, built from R₀/a/κ. Import into SolidWorks / Fusion / FreeCAD.

⬇ breeder.step owner — team Lab

Burner · Aegis / MetroVolt

STEP · parametric solid

Central cell · 11 solenoid coils · plug coils · end tanks · vessel, built from a_c and the coil layout.

⬇ burner.step owner — team Lab

Provenance

Real B-rep solids via CadQuery on OpenCASCADE 7.7 — not a mesh, not a mock. Representative to frozen parameters, not as-built (that needs component CAD). CC BY 4.0.

Representative engineering drawing derived from frozen parameters · NOT FOR CONSTRUCTION · reproduced from the open deposit DOI 10.5281/zenodo.21746157.

🔒 Engineering Blueprints is Kronos-only. The overview above is public — the full data, tables, live calculations and engineering detail on this page require a Kronos team sign-in.

Multiphysics · Breeder · Hyperion

Multiphysics Analysis

Engineering analysis of the frozen design — structural, thermal, electromagnetic and neutronics. Each result is tied to its method and honestly graded: frozen (from the deposit), screening (reduced-order model), or UNCOMPUTED (needs full FEA / CFD / MCNP). This is the analysis layer that feeds the “As-Simulated” fidelity rung.

◧

Magnet & structure — on-conductor field

FROZEN · deposit

Peak on-conductor field B_peak16.84 T

≤ 20.1 T · margin 16% · REBCO tape limit [HAR23]

Plasma β_N vs no-wall limit2.9

≤ 4.2 · margin 31% · MHD stability

TF hoop stress (screening) vs 316LN Sy410 MPa

≤ 700 MPa · margin 41% · ITER-grade steel, 4 K

Vessel — disruption EM load margin1.9 ×

≥ 1.5 × · margin 27% · apollo_structure envelope

Centre-post current density18 MA/m²

≤ 25 MA/m² · margin 28% · HTS centre stack

PF / CS coil support margin1.6 ×

≥ 1.3 × · margin 23% · shaping-coil structure — screening

FEAFull 3-D Lorentz/hoop stress on TF winding pack + case (ANSYS/COMSOL) — replaces the B² screening estimate

FEAVacuum vessel: buckling under 1 atm + fastest-quench disruption + halo currents

FEASeismic (OBE/SSE) response spectrum + fatigue (Miner) over pulsed duty

V&VReconcile hoop-stress screening with as-built winding geometry

Analysis layer of the 3D model · FROZEN values reproduce from the open deposit, SCREENING uses reduced-order models, PARTIAL awaits full FEA / CFD / MCNP. Frozen parameters: DOI 10.5281/zenodo.21746157. ENGINEERING · NOT FOR CONSTRUCTION

🔒 Multiphysics Analysis is Kronos-only. The overview above is public — the full data, tables, live calculations and engineering detail on this page require a Kronos team sign-in.

I&C · Control · Breeder · Hyperion

I&C · Control & Protection

The instrumentation-and-control spine: the real-time machine control system that holds the operating point, and the independent safety-instrumented system that trips it to a safe state. Deterministic control first; the AI/ML stack sits on top as advisory, bounded by the safety nets.

⚠ Design-stage architecture. Loop rates and SIL targets are design intent (IEC 61508 / 61511 framing), not certified functions — nothing is built. The value is the architecture: what controls what, and what independently protects it.

Control loops

real-time PCS functions

Fastest loop

~10 kHz

vertical stability

Safety functions

independent SIFs · 2 at SIL 3

Protection layers

independent (LOPA)

Control hierarchy (ISA-95 / Purdue, fusion-adapted)

Campaign / enterprise — Scheduling, fuel-cycle & campaign planning

Operations / supervisory — SCADA · mode & sequence management · data historian

Plant control — Coordinates cryo · vacuum · heating · fuelling · power conversion

Real-time machine control (PCS) — Deterministic fast loops that hold the plasma / mirror operating point

deterministic

Field — Sensors + actuators (magnets, valves, heating, coils, DEC)

The SIS is a separate vertical — its own sensors, logic solver and final elements spanning L0–L2, independent of the control system above. The AI/ML stack (Intelligence tab, L0–L7) is an advisory overlay on L1–L3, bounded by three deterministic safety nets (rule clamp · quench bypass · data-diode airgap) so it can never override protection.

Real-time control loops (PCS)

Function	Controlled variable	Sensor	Actuator	Loop rate	Target
Vertical stability	Z position	Flux loops · Mirnov	In-vessel + PF coils	~10 kHz	VDE-stable
Shape / boundary	Boundary flux	Magnetics	PF / CS coils	~1 kHz	κ 2.0 · δ −0.30
Plasma current	I_p	Rogowski	Current drive + OH	~1 kHz	9.66 MA
Density	n_e / Greenwald	Interferometry	Gas + pellet fuelling	~100 Hz	f_G 0.30
Divertor detachment	Target T_e / radiated frac	Spectroscopy · Langmuir	Impurity seeding	~100 Hz	90% radiated (binding)
Disruption avoidance	Proximity-to-limit	Multi-diagnostic + ML	Mitigation trigger	~kHz	avoid; else mitigate
Burn / heating	T_i / P_fus	Neutron · Thomson	NBI / RF	~Hz	Q 3.076

Safety-instrumented functions (SIS) — independent protection

Each SIF is independent of the control system and trips the plant to a safe state · click a row for detail

SIF	Function	Sensor → logic → final element	Protects against	SIL target	Response
SIF-1	Magnet quench protection	Quench detector → logic solver → dump switch	TF magnet quench ↗	SIL 3	< 2 s
SIF-2	Fast plasma shutdown	Disruption precursor → massive-gas / shattered-pellet	Disruption loads ↗	SIL 2	< 10 ms
SIF-3	Vacuum isolation	Pressure rise → fast isolation valves	Air ingress ↗	SIL 2	< 100 ms
SIF-4	Tritium confinement trip	Tritium monitor → isolate + detritiation	Tritium release ↗	SIL 3	seconds
SIF-5	Cryogen / ODH trip	O₂ + pressure → vent + ventilation	Helium release (ODH) ↗	SIL 2	seconds
SIF-6	Coil overcurrent / thermal	Current + temperature → coil trip	Magnet thermal ↗	SIL 2	ms

Layers of protection (LOPA)

Layer 1

Inherent / passive design — No chain reaction · benign termination · low decay heat

Layer 2

Basic process control (PCS) — Real-time loops hold the operating envelope

Layer 3

Alarms + operator response — Human-on-the-loop supervisory action

Layer 4

Safety-instrumented system (SIS) — Independent SIFs trip the plant to a safe state

Layer 5

Passive / physical relief — Relief valves · passive cooling · confinement barriers

Layer 6

Emergency response — Site emergency + mitigation

Independent layers, each able to act alone — a single failure does not defeat the protection. Layer 4 (SIS) is functionally independent of Layer 2 (control).

Design-stage I&C architecture · SIL targets per IEC 61508 / 61511 are design intent, not certified · SIFs map to the breeder Safety Case hazards. DESIGN INTENT · NOT CERTIFIED

🔒 I&C · Control & Protection is Kronos-only. The overview above is public — the full data, tables, live calculations and engineering detail on this page require a Kronos team sign-in.

RAM · Breeder · Hyperion

RAM · Reliability · Availability · Maintainability

The operational-performance layer: how reliably the plant runs, how much it's available, and how it's maintained. Availability (capacity factor) is the single biggest economic lever — it feeds directly into LCOE. This is a design-stage RAM allocation, not measured field data.

⚠ Design-stage allocation. MTBF / MTTR are allocated targets from analogous systems (ITER, fission, industrial) — no operating data exists. First-of-a-kind availability is much lower than the mature target and ramps with learning. The honesty is in showing the buildup, not asserting a number.

Mature CF target

69%

modelled, aspirational

First-of-a-kind CF

~32%

ramps with learning

Dominant downtime

Blanket / divertor

remote-handled swaps

Maintenance access

Remote

in-vessel activated

Availability buildup (gross → net capacity factor)

Gross (theoretical)100%

− Planned maintenance-18%

− Unplanned failures-8%

− Fuel / tritium logistics-3%

− Startup / ramp-2%

Net capacity factor69%

Mature-plant target. FOAK is ~32% — the first units run far below target and climb as failure modes are retired.

Reliability — key systems

System	MTBF	Failure rate	MTTR	Downtime driver
Superconducting magnets	> 10 yr	low	months	rare but long — cryo-coupled
Divertor / PFC	erosion-limited	scheduled	weeks	remote handling
Breeding blanket	damage-limited	scheduled	weeks	remote handling · heaviest driver
Tritium plant	~2 yr	moderate	days	availability-critical
Heating / current drive	~1 yr	moderate	days	redundant (N+1)
Cryoplant	> 5 yr	low	days	high reliability
Vacuum / pumping	> 5 yr	low	days	high reliability
Power supplies	industrial	low	hours	conventional

Maintainability — repair & access

Component	MTTR	Method	Note
Blanket module	weeks	remote handling	activated — dominant availability driver
Divertor cassette	weeks	remote handling	activated · replaceable unit
Magnet (in-cryostat)	months	major intervention	rare
Heating source	days	hands-on (ex-vessel)	shielded gallery
Vacuum pump	days	hands-on	ex-vessel

Capacity-factor ramp FOAK → mature target

The availability lever

Economic sensitivityCF is the #1 LCOE driver

Availability = MTBF/(MTBF+MTTR)MTTR-dominated (remote handling)

Breeder maintenanceremote (activated)

Learning-curve rampFOAK ~32% → target 69%

Availability is the crux the economics turn on — see Economics ↗ (team-only). Raising CF is worth more than almost any physics gain; that is why maintainability drives the design.

Design-stage RAM allocation · MTBF/MTTR from analogous systems, not measured · capacity factor is a modelled target, FOAK much lower. ALLOCATION · NO OPERATING DATA

🔒 RAM · Reliability · Availability · Maintainability is Kronos-only. The overview above is public — the full data, tables, live calculations and engineering detail on this page require a Kronos team sign-in.

Live Operations · Breeder

Live Operations

Simulated operational sequence ramping the design point from cold start. Telemetry converges to the frozen values; nothing exceeds them.

Scenario Startup · 0%

Startup

Current ramp

Heating

Steady breed

Hold

Fusion gain Q

0.000

ramping…

Fusion power

0.00 MW

ramping…

Plasma current

0.000 MA

ramping…

Tritium

0.000 kg/yr

ramping…

Actuator setpoints live

B₀ setpoint0.00 T

I_p setpoint0.00 MA

Aux heating0 MW

Blanket TBR0.00

Sequence phases D–T startup → steady breed

0%Startup

15%Current ramp

40%Heating

60%Steady breed

90%Hold

Breeder deposit · DOI 10.5281/zenodo.21746157 · reproduced from the open Zenodo/GitHub deposit (CC BY 4.0). Canonical freeze: FROZEN_2026-08_PUBLICATION.csv.

🔒 Live Operations is Kronos-only. The overview above is public — the full data, tables, live calculations and engineering detail on this page require a Kronos team sign-in.

Live version The interactive version of this panel runs in the team Lab — the static record is shown here.

Digital Thread · Breeder · Hyperion

Digital Thread

The single traceable chain that links every layer of the twin — one requirement followed from the 3-D component through analysis, validation, as-built metrology, the safety case, control and availability, and back to the open deposit. This is the integration surface, not new data: it shows where the thread is whole and where it's thin.

3-D component
As-Simulated + Δ-as-built→Requirement
V&V→Analysis
4 domains→Validation
deposit checks→As-Built
metrology→Safety
FMEA / SIF→Control
PCS / SIS→RAM
availability→Deposit
DOI

Threads

top requirements traced

Avg completeness

3.8/6

workstreams linked per thread

With validation evidence

8/10

reproducible deposit checks

With as-built link

8/10

metrology-controlled

Integrated traceability — one requirement, across every workstream

Each row is one full thread · cells show linked-artifact counts · click a row for the whole chain

Requirement	Analysis	Validation	As-Built	Safety	Control	RAM	Thread
REQ-B01 Fusion gain Q ≥ 3.0 at the des	1	2	1	—	—	—	3/6
REQ-B02 Tritium breeding self-sufficie	1	5	1	1	2	1	6/6
REQ-B03 Peak on-conductor field ≤ 20.1	1	2	1	—	—	1	4/6
REQ-B04 MHD stability — β_N below the	1	2	1	—	—	—	3/6
REQ-B05 Plasma current sustained (I_p,	1	2	2	—	—	1	4/6
REQ-B06 First-wall lifetime ≥ 2 fpy at	1	1	1	—	—	—	3/6
REQ-B07 Divertor exhaust within target	1	1	1	1	—	1	5/6
REQ-B08 Structure survives disruption/	1	—	2	2	3	—	4/6
REQ-B09 Magnet quench protection (hot-	1	—	—	1	2	1	4/6
REQ-B10 End-of-life waste ≤ Class C lo	1	3	—	—	—	—	2/6

Weakest link

Weakest coverage columnControl (3/10 threads)

Analysis10/10

Validation8/10

As-Built8/10

Safety4/10

Control3/10

RAM5/10

The thread is honest about where it thins: REQ-B08 are OPEN (need a test article), and as-built / safety / control links exist only where a requirement drives them. Analysis + validation are the backbone.

Model-based digital thread over the frozen design · every link resolves to an artifact already in the twin · reproduces from DOI 10.5281/zenodo.21746157. INTEGRATION SURFACE · NOTHING HARDWARE-VERIFIED

🔒 Digital Thread is Kronos-only. The overview above is public — the full data, tables, live calculations and engineering detail on this page require a Kronos team sign-in.

Data connectivity · shared foundation

Data & Standards

Kronos is built on open, cited data. Live connections resolve to real records; data standards and solver interfaces are declared with provenance. Nothing here is a placeholder.

Live connections

Zenodo deposits

2 DOIs

LIVE

Breeder + burner code, data, figures — CC BY 4.0, in the kronos_fusion_energy community.

GitHub repositories

code + data

LIVE

Reproducible evaluators and datasets mirrored to GitHub (run_all.py, common/, data/).

Literature corpus

69 refs

DOI-linked

Peer-reviewed sources, 1974–2026, across 15 venues — every anchor is cited.

Source archive

155 files

archived

Primary-source provenance behind every claim: patents, lab reports, preprints.

Data standards & solver interfaces

IMAS / IDS

ITER data model

declared

Interoperable Data Access — the ITER standard schema for plasma state and diagnostics.

EXFOR / ENDF

nuclear data

sourced

Fusion cross-sections from EXFOR experimental entries — nothing extrapolated (H42).

ITER baseline

15 MA

benchmarked

Hyperion’s 9.66 MA is ~⅔ of the ITER 15 MA baseline (Kim 2018).

IPB98(y,2)

confinement

applied

The standard tokamak energy-confinement scaling; τ_E solved at H₉₈ 1.0.

Benchmark references · click a row to open the DOI

Anchor	Device / method	Value	Source	Year	DOI
H₉₈ confinement	DIII-D	H98(y,2) 1.2	M. E. Austin	2019	10.1103/physrevlett.122.115001 ›
Negative-triangularity edge	CHEASE/BALOO	H-mode prohibited → L-mode	A. Nelson	2022	10.1088/1741-4326/ac8064 ›
β_N limit	NSTX	β_N 7.2	S. A. Sabbagh	2006	10.1088/0029-5515/46/5/014 ›
REBCO peak field	SPARC TFMC	20.1 T on conductor	Z. Hartwig	2023	10.1109/tasc.2023.3332613 ›
REBCO neutron tolerance	REBCO tape	Jc-degrade fluence 3.9×10²²	D. X. Fischer	2018	10.1088/1361-6668/aaadf2 ›
Direct energy conversion	TMX tandem mirror	48% DEC efficiency	W. L. Barr	1983	10.13182/fst83-a20820 ›
D–³He recirculating power	Fokker-Planck	min recirc > fusion	T. H. Rider	1997	10.1063/1.872556 ›
ITER plasma current	ITER	15 MA baseline	S. H. Kim	2018	10.1088/1741-4326/aab034 ›

Provenance from the open deposits · literature_corpus.csv (69 refs) + source archive (155 files). Live links resolve to real records; standards marked declared are interfaces wired in later phases.

🔒 Data & Standards is Kronos-only. The overview above is public — the full data, tables, live calculations and engineering detail on this page require a Kronos team sign-in.

KRONOS FUSION ENERGY · Kronos Model Conceptual design and simulation study; no machine has been built