Maintaining the Activated Machine

Abstract

Fusion maintenance begins before the first component is drawn. Neutron activation and tritium contamination can prevent direct human access to the vessel, while blankets, divertors, diagnostics, pumps, and heating launchers remain life-limited. I define the maintenance system as an industrial flow from diagnosis through isolation, disconnection, removal, shielded transfer, hot-cell work, replacement, testing, and configuration closure. Module size, port geometry, connector count, tolerances, lifting paths, contamination boundaries, cask traffic, tooling recovery, spares, and waste routes determine outage duration. The reference plant uses replaceable sectors, standardized service interfaces, redundant remote-handling trains, full-scale mockups, and a hot cell sized by peak outage throughput rather than average waste volume.

Maintenance is part of the power core

Neutrons activate in-vessel materials and damage components. Tritium and dust contaminate surfaces. Even after shutdown, radiation and atmosphere can prevent people from entering the vessel. If blankets and divertors cannot be replaced remotely, the plant has no credible life beyond its first component set.

The energy consequence is direct. Annual delivered energy is

$$E_{\mathrm{annual}}=P_{\mathrm{net}}\times8760\ \mathrm{h}\times A,$$

where $A$ is availability. A $500\ \mathrm{MW}$ plant at $30\%$ availability delivers less energy than a $300\ \mathrm{MW}$ plant at $70\%$. Maintenance architecture can outweigh plasma rating.

Start with the outage sequence

I design the replacement process backward from a proven pressure, vacuum, electrical, cooling, tritium, and alignment test. Before a component can pass that test, its interfaces must be connected. Before connection, the module must be positioned. Before positioning, the old module and any obstruction must leave through a defined path.

The sequence is:

1. End plasma operation and reach a maintainable thermal state.
2. Characterize radiation, tritium, dust, residual energy, and component condition.
3. Isolate coolant, fuel, electrical, vacuum, and structural interfaces.
4. Drain, purge, decontaminate, and verify boundaries.
5. Deploy tooling, establish control, and disconnect services.
6. Release structure and capture the component.
7. Transfer through vessel, port, cask, and hot-cell boundaries.
8. Inspect the exposed bay and correct damage.
9. Install and connect the replacement.
10. Leak, pressure, electrical, dimensional, flow, and functional test.
11. Update the as-maintained configuration before operation.

Each step has a recovery state. A stuck fastener, severed cable, failed manipulator, dropped tool, distorted module, or contaminated seal cannot be allowed to trap the plant indefinitely.

Module size is an outage trade

Large sectors reduce the number of in-vessel joints and handling operations. They require massive casks, ports, cranes, storage stands, and hot-cell bays. Small modules reduce individual mass and selective waste but multiply disconnections, alignment tasks, leak tests, and failure opportunities.

If each of $N$ connections has independent success probability $p$, the probability that all succeed on the first attempt is

$$P_{\mathrm{first}}=p^N.$$

Even with $p=0.995$, one hundred connections yield only about $61\%$ first-pass success. Independence is optimistic because contamination, misalignment, or tool calibration can affect many interfaces together.

The module boundary should minimize total expected outage time, not connection count alone. I evaluate mass, dose, cask moves, parallelism, inspection access, common-cause failure, and replacement selectivity.

Interfaces must cooperate with machines

Remote operations need chamfers, guide features, captured fasteners, visible state, standardized tool engagement, torque reaction, protected seals, drainable volumes, and fiducials that remain readable after exposure. Connectors should tolerate small misalignment while preventing a false complete state.

Structural load paths are separated from fluid and electrical connectors. The module lands and is mechanically supported before delicate services engage. Double isolation and test spaces permit leak verification. Cut and weld operations remain possible where demountable joints cannot meet radiation, temperature, or leak requirements, but they need qualified remote preparation and inspection.

Designers must see through the robot’s camera, not through an unobstructed computer model. Lighting, contrast, occlusion, radiation noise, condensation, dust, and limited depth perception change routine work.

The remote fleet needs diverse bodies

No universal robot should become a single point of plant availability. Heavy transporters carry modules. Articulated manipulators operate tools. In-vessel carriers position along rails or temporary fixtures. Inspection crawlers reach tight regions. Casks provide shielded transfer and confinement. Recovery devices retrieve disabled machines.

Common control, coordinate frames, connectors, software interfaces, tooling, and training let the fleet cooperate. Critical motions include local autonomy for collision avoidance and force limiting, while operators retain supervisory control and verified hold points.

The fleet undergoes routine proof testing away from radiation. Spare manipulators and control cabinets remain available. A failed robot inside the vessel is a planned scenario with towing points, manual releases, redundant brakes, and a second access path.

Layer	Primary job	Recovery requirement
Transport cask	Shielded component movement	Tow, isolate, and dock from alternate drive
Heavy transporter	Carry and align sectors	Safe support after loss of power or actuator
Manipulator	Disconnect and reconnect services	Tool release and retrieval by another arm
Inspection robot	Map condition and verify work	Recover through tether or capture fixture
Metrology network	Establish common coordinates	Independent survey after fiducial loss

The hot cell is a production facility

ITER’s hot-cell system illustrates the breadth of work: receive activated components, decontaminate, inspect, maintain, test, separate waste, and support remote equipment. A commercial plant adds schedule pressure and repeated throughput.

The hot cell needs parallel bays for incoming casks, component cooling, decontamination, nondestructive examination, cutting, repair, assembly, testing, waste packaging, tool maintenance, and storage. Airflow and drainage preserve confinement. Shielding permits adjacent human work. Cranes and manipulators have overlapping reach so one failure does not block a bay.

Capacity is sized for the outage peak. If ten large sectors arrive within a planned shutdown, an average annual waste-flow calculation is irrelevant. Congestion in the hot cell can extend the critical path even when vessel work is ahead of schedule.

Mockups are operational infrastructure

Every high-risk task is rehearsed in a geometrically and functionally representative mockup. The mockup verifies reach, force, visibility, tool change, collision clearance, cable management, procedure timing, and recovery from faults. As-built scan data updates it.

Digital simulation is useful for sequence design but cannot reproduce flexible hoses, worn connectors, surface contamination, backlash, imperfect lighting, or an unexpected burr. Physical trials generate the time distributions used in availability models.

Configuration control closes the job

Remote maintenance creates hidden change. A jumper takes a different route. A replacement sensor has a new calibration. A shim changes alignment. A temporary cover remains installed. The digital plant record changes before the work package closes.

Return-to-service evidence includes component identity, material and manufacturing records, interface torque or weld inspection, leak rate, pressure test, electrical insulation, flow, valve state, alignment, foreign-material exclusion, software configuration, and unresolved concessions.

PrepareDiagnose, cool, isolate, and purge

Convert an operating nuclear system into a bounded work state.

MoveDisconnect, capture, and transfer

Use recoverable machines and shielded paths sized for real modules.

ProcessInspect, repair, store, or package

Operate the hot cell as a parallel production and evidence facility.

RestoreInstall, test, and close configuration

Return only after every boundary and function is physically verified.

Availability comes from distributions

Outage estimates should not add best-case task durations. I model uncertain task time, resource contention, radiation cooling, inspection findings, rework, tool failure, and spares. Monte Carlo or discrete-event models reveal which resources govern the tail of the outage distribution.

PlanSimulate and rehearse exact work ExecuteRun parallel bounded work fronts RecoverResolve faults without losing access LearnUpdate design, tools, spares, and timing

Remote maintenance controls occupational exposure and economic survival. It also interacts with tritium, activated dust, coolant energy, magnets, and waste. The next report collects those hazards into a safety architecture built from bounded inventories and independent confinement rather than broad comparisons with other nuclear technologies.