Fusion and big science: cryogenics, magnets, new plants

Cooling transients threaten a quench of the magnets

What happens. Superconducting magnets stay superconducting only inside a narrow temperature band. When the cooling plant changes load, or a pump trips, the coolant temperature moves, and the margin to a quench shrinks.

What it costs. A quench stops the experiment for days: the magnet must be warmed, the fault understood, and the system cooled down again. The coolant lost may be hard to replace.

Why the usual fix fails. Wide alarm margins make the operators stop earlier than necessary. A better temperature sensor at the magnet shows the problem only when it has begun.

How a fixed-price remote diagnosis checks it. From plant logs and the cooling-circuit layout, I check how fast the temperatures react to load changes and where the margin goes. The fixed-price remote diagnosis tells you whether a predictive controller could keep the magnets inside the band through the transients, and what model it would need.

Helium cooling circuits have few sensors

What happens. Magnets cooled by superfluid or supercritical helium sit in long circuits with a handful of temperature and pressure sensors. What happens between the sensors, how the heat spreads and where the flow goes, is not observed.

What it costs. The operators run with margin because they cannot see the real state. That margin costs cooling power and availability, and a fault is noticed late.

Why the usual fix fails. Adding sensors in a cold, radiation-exposed, hard-to-reach environment is costly and each one is another leak path. Models used offline do not follow the plant when it changes.

How a fixed-price remote diagnosis checks it. I worked on predictive control and state estimation for the superfluid-helium cooling of the Large Hadron Collider at CERN, the European particle-physics laboratory. The fixed-price remote diagnosis looks at your sensor layout and logs and shows which unmeasured states can be computed and used for control.

Pulsed heat loads from the machine trip the cryoplant

What happens. The machine behind the cryoplant does not draw heat steadily. It pulses with the experiment or the accelerator cycle, and the plant sees steep load swings.

What it costs. If the plant cannot follow, pressures and temperatures move out of range and it trips. Each trip means a long recovery and lost experiment time.

Why the usual fix fails. Operating at a lower average load keeps margin but wastes capacity. A faster feedback loop in the usual way amplifies the swing instead of absorbing it.

How a fixed-price remote diagnosis checks it. From plant logs and the machine’s planned load cycle, I check how the plant reacts to the pulses. The fixed-price remote diagnosis tells you whether predictive control that knows the next pulse in advance could keep the plant running, and what forecast signal it would need.

Fusion start-ups have no control team for the plant systems

What happens. A fusion start-up is built by plasma physicists and mechanical designers. The cooling, power, vacuum and fuel systems of the plant need a control concept, and nobody in the team owns it.

What it costs. Without an owner, each subsystem supplier delivers its own controller and the interfaces are found during commissioning, when they are expensive to change.

Why the usual fix fails. A full control department takes years to build. Leaving it to the system suppliers produces controllers that work alone but not together.

How a fixed-price remote diagnosis checks it. I look at your plant description and the interface lists you have, and sketch where the control gaps are. The fixed-price remote diagnosis tells you what a control architecture for the plant would need, and what could be done by an interim lead until a team exists.

New fusion machines come with no operating experience

What happens. A stellarator or tokamak plant is the first of its kind. Nobody has run its cooling, power and vacuum systems together over a long period, and no one knows how they respond to each other.

What it costs. Commissioning takes longer when the first real experience is gathered on the installed machine. Rework on installed equipment is expensive.

Why the usual fix fails. Experience from other machines helps only partly, because the design is new. Static design calculations say nothing about how the plant behaves in transients.

How a fixed-price remote diagnosis checks it. From the plant’s design documents and the data of the subsystems already built, I check which dynamic behaviours decide the first operation. The fixed-price remote diagnosis tells you whether plant models and a control concept can be built early enough to find the problems before commissioning.

Heat loads in cryogenic circuits cannot be measured

What happens. The heat that reaches a cold circuit, from the magnet, the beam or radiation, cannot be metered directly. All the operators have are temperatures, pressures and valve positions.

What it costs. Without the heat load, the plant is operated blind: the cooling power is set with margin and a growing load is noticed only when a limit is hit.

Why the usual fix fails. A calorimetric test at a stand-still gives one value at one condition. A flow meter in a cold line is expensive and drifts.

How a fixed-price remote diagnosis checks it. From the temperature, pressure and valve data of your plant, I check whether the heat load can be computed with useful accuracy. The fixed-price remote diagnosis shows what a computed heat load would give the operators, and which extra sensor, if any, would make it reliable.

Vacuum pump-down times are long

What happens. Large vacuum systems take hours or days to pump down after opening. The pumps, valves and bake-out stages are run by a fixed procedure that was written for the worst case.

What it costs. Every pump-down is time the machine is not available for experiments. Long pump-downs also mean more people on shift to babysit the process.

Why the usual fix fails. Adding pumping speed costs money and may not help when outgassing is the limit. A fixed procedure cannot adapt to the actual state of the surfaces.

How a fixed-price remote diagnosis checks it. From pump-down logs (pressures, pump currents, temperatures, valve times) I model how the pressure falls and what limits it. The fixed-price remote diagnosis tells you whether a model-based procedure could shorten the pump-down on your system, and which stage gains most.

Cooling-water temperature must stay stable for magnets

What happens. Magnets, power converters and detectors are cooled with water that must stay at a given temperature to a fraction of a degree. Changes in heat load, ambient temperature and flow disturb it.

What it costs. Temperature swings change the magnet field and the beam. The result is lost beam quality and time spent re-tuning.

Why the usual fix fails. A tighter standard controller on the cooling loop reacts after the disturbance has arrived. Larger buffer tanks smooth the swing but cost space and money.

How a fixed-price remote diagnosis checks it. From logs of the cooling loop and the heat-load schedule, I check how the temperature responds to known disturbances. The fixed-price remote diagnosis tells you whether predictive temperature control, using the planned load, would hold the water temperature tighter than your present controller.

Large magnet power converters have to be precise

What happens. The power converters that feed big magnets must hold a current to a tiny error, even while the grid voltage dips and the load changes. A small current error is a field error.

What it costs. Poor current accuracy costs beam quality and measurement precision, and the converters run with filters and margins that cost efficiency.

Why the usual fix fails. Tuning the standard controller of a converter for one operating point leaves it worse at others. Faster loops are limited by delays and the filters.

How a fixed-price remote diagnosis checks it. I review the converter’s model and logs of current and voltage, and ask how much of the error is predictable. The fixed-price remote diagnosis tells you whether model-based feedforward could reduce the current error, and what knowledge of the magnet and grid it would need.

Heat deposition on target systems varies

What happens. Targets, beam dumps and plasma-facing components absorb a heat load that changes with the experiment and the beam. What each part receives is estimated from a few thermocouples.

What it costs. If the load goes above the design value, a part is damaged, and replacing it can stop the machine for a long time. So operators keep the experiment well below the limit.

Why the usual fix fails. More thermocouples in a radiation field are fragile. Rules based on a worst-case load restrict the experiment more than the hardware demands.

How a fixed-price remote diagnosis checks it. From temperature records and the geometry of the component, I check how well the heat deposition can be computed from the signals you have. The fixed-price remote diagnosis tells you whether a computed load, with safety limits attached, could replace part of the worst-case margin.

Operator screens for complex plants overload the operators

What happens. A research plant has thousands of signals and many subsystems, each delivered with its own screen. Operators in the control room must combine them to understand what is happening.

What it costs. Alarm floods in a disturbance hide the cause. Errors by operators in an unfamiliar situation can cost a quench or a trip.

Why the usual fix fails. Adding more screens adds more to watch. Alarm limits set one by one by each subsystem group do not know about the others.

How a fixed-price remote diagnosis checks it. I look at your alarm logs and screen structure from past disturbances and ask which signals told the story and which only added noise. The fixed-price remote diagnosis tells you what a supervisory layer would need, and where a computed value would help the operators most.

Controls are specified late in new facilities

What happens. In a large new facility the buildings, magnets and cryogenics are designed first. The control requirements are written when the hardware is already ordered.

What it costs. Late requirements mean interfaces that do not fit, sensors that are missing, and rework on equipment that is already installed.

Why the usual fix fails. A late control system integrator can only connect what was delivered. The people who know what operation will really need are in the physics group, not in the construction team.

How a fixed-price remote diagnosis checks it. I review your requirements and architecture drafts and ask what operation, start-up and fault handling will demand of the control system. The fixed-price remote diagnosis gives you a list of requirements that are better fixed now, with reasons, before more equipment is ordered.

Faults in large plants are hard to trace

What happens. When a large plant trips, the first hours go into finding what started it. Many subsystems report alarms at once, and the first alarm in the log is not always the first cause.

What it costs. Each hour of searching is an hour of downtime. Repeated faults whose cause was never found come back.

Why the usual fix fails. Reading logs by hand needs the people who know the plant best, and they are the ones also needed for the restart. Fixed alarm rules catch known faults, not new combinations.

How a fixed-price remote diagnosis checks it. From the signal logs of past trips, I compare what a model says the plant should do with what it did. The fixed-price remote diagnosis shows whether model-based diagnosis would have pointed to the cause sooner, and what data would be needed to run it continuously.

Cryoplants consume megawatts of electricity

What happens. A large cryoplant draws a very large share of the electrical power of a research site, around the clock, because the cold must be kept. Compressors run near fixed points and the plant is not reconsidered when the load changes.

What it costs. Power cost, and with it the carbon footprint, is one of the largest operating costs of the facility, and tariffs vary during the day.

Why the usual fix fails. Efficient compressors help, but they are bought once. Hand-adjusted setpoints depend on the shift, not on the price or the current heat load.

How a fixed-price remote diagnosis checks it. From compressor, load and tariff data, I estimate how much power the plant uses beyond what its present heat load requires. The fixed-price remote diagnosis tells you whether economic operation, tracking load and electricity price, is feasible within the plant’s safety limits.

Detector cooling with two-phase carbon dioxide is hard to keep stable

What happens. Particle detectors are cooled with carbon dioxide that boils and condenses in the circuit. Pressure, flow and temperature depend on each other, and the circuit can oscillate or lose its liquid supply.

What it costs. Instabilities mean detector modules at the wrong temperature, noisy data, and shutdowns. Tuning the circuit takes expert time at every change of heat load.

Why the usual fix fails. Standard controllers set for one operating point cope poorly with the strongly non-linear behaviour. Larger buffers slow the response without removing the cause.

How a fixed-price remote diagnosis checks it. From operation logs of the cooling circuit, I check where the circuit is stable and where it is not. The fixed-price remote diagnosis tells you whether model-based control can keep the circuit stable over the working range, and which measurements it would need.

Magnet test stands have long cool-down and test cycles

What happens. Magnets are tested one after another on a test stand. Each needs a long cool-down to operating temperature, the test itself, and a warm-up before the next one is mounted.

What it costs. The stand is in use only a fraction of the time, and the cooling plant runs through it all. A large production run of magnets occupies the stand for a long time.

Why the usual fix fails. Cooling faster is limited by thermal stress on the magnet, so the schedule is set conservatively. Several magnets in parallel need more cryogenic plant.

How a fixed-price remote diagnosis checks it. From cool-down logs and the thermal limits of the magnets, I check how much of the time is margin and how much is physics. The fixed-price remote diagnosis tells you whether an optimised cool-down could shorten the cycle without exceeding the stress limits.

Tritium plant processes are new, with no practical experience

What happens. A fusion fuel cycle must handle, clean and recycle hydrogen isotopes in volumes no plant has handled before. The separation and storage steps exist on paper and in small experiments.

What it costs. Start-up of a first-of-its-kind plant with radioactive material is slow and cautious. Mistakes are costly in time and in licences.

Why the usual fix fails. Scaling up a laboratory process by hand-picked safety factors makes it large and costly. Without models, nobody can say in advance how the plant reacts to a disturbance.

How a fixed-price remote diagnosis checks it. From the process descriptions and whatever test data exist, I build up what a dynamic model of the fuel cycle needs. The fixed-price remote diagnosis tells you whether early process models and a control concept can reduce start-up risk, and which data to collect first.

Interactions between systems of the whole facility are unknown

What happens. Cryogenics, power, cooling water, vacuum and magnets are designed by different groups. Each group tests its own system, and how they act on each other is first seen when the facility is switched on.

What it costs. Surprises at that stage cost schedule and, in the worst case, damage. A disturbance in one system moves through the others in ways nobody planned for.

Why the usual fix fails. Subsystem tests do not show interaction. A full simulation built late is too slow to influence the design.

How a fixed-price remote diagnosis checks it. From the subsystem descriptions and available data, I check which couplings matter most. The fixed-price remote diagnosis tells you whether a plant-wide dynamic simulation is worth building, which interactions it should cover first, and what it would take to make it credible.

Helium is scarce and recovery systems are overwhelmed by quenches

What happens. Helium supply is tight and prices have risen after a disruption centred on a major producer (WNEM). Laboratories and hospitals with superconducting magnets are asked to save it, recycle it and plan for deliveries that do not arrive.

What it costs. A quench vents a large amount of helium at once. When recovery systems are not sized for that, the gas is lost and has to be replaced at a high price, if it can be had at all.

Why the usual fix fails. Buying more storage tanks postpones the problem. A recovery or reliquefaction plant that runs at fixed settings does not follow the changing flows.

How a fixed-price remote diagnosis checks it. The fixed-price remote diagnosis looks at your recovery and reliquefaction data and shows whether control can reduce losses, and how the system would behave when a quench hits.