How do you do maintenance on a fusion reactor? I assume the simplest way is to turn it off and open it up. But how dangerous are the dangerous parts when it's running? Could you send a drone inside?
(Disclaimer: all the information and images in this post are from open-access articles and press releases, and I am not speaking on behalf of any research group or company)
Depends. Are we talking about modern experimental devices, or future power plants? In the former case the answer is "send a human or robot in there," in the latter it's either "send in a robot" or "take the thing apart (with robots)."
Either way, you can't send anything inside while it's running. Everything needs to happen during downtime, every year or so during scheduled maintenance.
After long periods of operation, many experimental fusion devices end up somewhat radioactive, from nuclear transmutation of the vessel itself and small amounts of trapped tritium in graphite heat shield tiles. Usually this means they send people in wearing dosimeters to keep careful track of exposure, good PPE, and set a limited amount of time that any one individual can spend inside every year Here's a technician upgrading equipment inside DIII-D, a tokamak in San Diego:
Source: DIII-D National Fusion Program Completes Facility Upgrade
Don't touch the graphite with your bare hands, leave the vessel after you've reached a pre-approved radiation dose (usually something roughly equivalent to a medical x-ray), and you'll be quite safe.
There's an exception though. Most devices (like DIII-D above) use ordinary hydrogen or pure deuterium (hydrogen-2) as fuel, which is useful for experiments but is not intended to produce significant amounts of power. However, some now-decommissioned experiments – specifically, JET in the UK and TFTR at Princeton – used a 50/50 deuterium-tritium (hydrogen-3) fuel, which is prototypical of an actual power plant.
This produces several orders of magnitude higher neutron flux than deuterium alone (which leads to higher rates of nuclear transmutation in the metal structure) and leaves significant tritium contamination in the walls, to the point where it is unsafe for humans to enter the vessel at all.
So you send in remote handling robots!
At JET, they replaced thousands of tiles with robotic arms controlled by a guy. This works for an experiment, but is costly, time consuming, and requires extensive training.
Many future power plant concepts envision a combination of internal robotic maintenance, like JET, and some method to remotely split apart the reactor to send in something larger than a couple of robotic arms (cranes, forklifts, etc). You still can't send a human in there, but having the whole thing opened up makes access a lot easier.
Proxima Fusion has some nice figures in an open access paper on their Stellaris power plant concept: Stellaris: A high-field quasi-isodynamic stellarator for a prototypical fusion power plant. Their idea is to divide the stellarator into four sectors that can be slid out:
Commonwealth Fusion Systems intends a similar maintenance scheme for ARC, as published in their latest physics basis article: Overview of the physics basis for the ARC fusion power plant. There's no figure, but they describe splitting the reactor in half in a similar way to replace much of the structure all at once. (The original paper back in 2015 involved splitting it at the equator and lifting the top half, but they state here that their new concept is to split it on a pole-to-pole plane).
