The complexity of a fusion reactor means the reliability of all systems and components must be very carefully considered. This issue is empathised in a presentation by Mohamed Abdou: Lessons Learned from 40 Years of Fusion Science and Technology Research . The availability is taken up in a sidebar in the Science article of Daniel Cleary : Running out of Gas .

Availability is defined in terms of rates of component failure as:

where MTBF is the Mean Time Between Failures and MTTR is the Mean Time To Repair. Abdou calls availability, the "Achilles' Heel of Fusion" and suggests it might be as low as 5%. The EU DEMO reactor has a target for availability >30%, defined as fraction of the time generation electricity. For US DEMO designs the availability goal is >50%. The ITER operational availability goal is 32%.

The sections below explore what availability could realistically be expected for a commercial fusion reactor. Device availability is reduced by two types of outages: scheduled outages (maintenance) and unscheduled outages(breakdowns).


Maintenance

A fusion reactor must employ a remote handling maintenance system to replace the components inside the vacuum vessel. The EU-DEMO design has 48 divertor cassettes and 80 breeding blankets module which must be replaced at regular intervals: approximately every 2 years for the divertor and every 4 years for the breeding blanket. Each divertor module will weight about 8 tonnes and each breeder blanket modules about 80 tonnes. The replacements take place through access ports shown in the figures below, divertor parts through the lower ports and breeder blanket modules through the upper ports. 

Access ports to exchange breeding blankets


Key in-vessel components to be exchanges and their entry-egress routes

The radiation dose rate, inside the torus, at the beginning of a maintenance operation, is expected to be 2000 Sv/hr from gamma rays, which can be compared to a the lethal dose of human beings of about 5 Sv. So all operations must be performed with remoted handling equipment very far from any people. A consequence of the radioactivation, is that once disconnected from the cooling pipes, the components warm up to temperatures exceeding 200°C.

The details above are taken from: EU DEMO Remote Maintenance System development during the Pre-Concept Design phase. While this paper has an optimistic conclusion it is clear that the construction of the needed equipment will be very difficult.


Breakdowns

The rate of component breakdown in very difficult to estimate and obviously depends on design details. Estimates for the US-DEMO design from Abdou et al. (2015) are given in the table below.

Results of component availability estimates for US-DEMO components.

A feature which is important in this table, is the estimate for a major coil failure at 0.4%/year, which with 24 items becomes 10% for one major coil failure per year. Here, in the US-DEMO design, the repair time is given at about one year. In the ITER/EU DEMO design, repairing a coil may be much longer or impossible.


Availability Estimates

No published estimates for total fusion power reactor availability exist and so in this section a guesstimate is made from information which is published, supplemented by private communication from a number of sources.

Scheduled downtime for component replacement will be a large contribution to total reactor downtime. With limited published details, an estimate is made as follows. The replacement of the divertor; limiter and breeder blanket modules will be the biggest contributions. It is likely, any other regular maintenance for equipment such as the cryogenic system; tritium system ; fueling system; vacuum system; etc., can be done simultaneously and so add no further downtime. Two large factors are common to any intervention inside the vacuum vessel: a cooldown time at the after machine stop, to allow for radioactive decay of activated components, and at the end, a vacuum pumping time to re-establish the vacuum quality necessary for tokamak operations. Three types of replacements will be assumed with different frequencies and difference durations beyond the two end times above.  The frequencies and durations are taken as: limiters every 1 year, taking 1 month; divertor every 2 years, taking 4 months and blankets every 4 years, taking 6 months. Replacements are done simultaneously and so short operations are masked by longer ones. These assumptions give 20 months of downtime every for every cycle of 48 months, so a scheduled average availability of 58%.

Unscheduled breakdowns due to component breakdown, make up the second contribution to the overall availability. This is taken from the, Abdou et al., component estimate in the table above at 61.5% availability. In addition there could be a rate of interventions for disruptions which might make a significant difference. As already stated in previous pages, this rate must be reduced dramatically from the experience of  previous tokamaks with lessons from ITER. Just taking one disruption per year, it is assumed entry into the vacuum is necessary to check for damage, scheduling with other operations would likely be difficult so this would reduce the yearly availability to 10/12 months, 83%.

Overall availability is by combination 0.58 x (0.615 X 0.83) = 0.30. If there are more disruptions per year this would become even lower. The handwaving estimate of availably made here, surprisingly, is not far from the DEMO and ITER goals mentioned above. Unfortunately, for a commercial reactor an availability factor of 30% seems economically unviable. Further, such a value would make tritium self-sufficiency difficult given the tritium stock decay rate of 5% per year.