9 thoughts on “Machine Learning”

  1. I haven’t heard that much about how machine learning has helped in fusion research. I’d say better magnets, better models, and faster computers are the main reasons fusion seems closer to reality now. Along with nontrivial amounts of venture capital.

    1. When I was working in a plasma physics lab the joke was everybody said it was 30 years away because that’s how far away the people working on it were from retirement. Those people retired a decade ago.

  3. Space propulsion.

    Unless you are talking “torch ships” and interstellar travel, what does fusion do for you that good-old-fission doesn’t do?

    Are your talking nuclear thermal? Yeah, I know that fission reactors have an upper temperature limit, but would that be higher for fusion-thermal?

    Are you talking nuclear electric? This means you need to generate electric power, a lot of it, which means at some level, you need to run a power cycle, and how to you intend to reject heat at the cold end?

    The problem of an effective space radiator is probably where effort needs to be concentrated. Getting fusion going won’t help that problem.

    1. Fusion can give you less residual radiation shortly after shutdown. With some of the more difficult fuels you might get less of the hard to shield radiation while the reactor is operating.

  4. The answer to the question is almost certainly “no”. Even if confinement were a perfectly solved problem, fusion would face grave barriers to practicality.

    Tri-Alpha (now TAE) is mentioned in the article. They were told 20+ years ago that the colliding beam idea would not work, and where the error in their analysis was that led them to think otherwise.

    https://www.researchgate.net/publication/235032059_Comments_on_the_Colliding_Beam_Fusion_Reactor_Proposed_by_Rostoker_Binderbauer_and_Monkhorst_for_Use_with_the_p-11B_Fusion_Reaction

    For DT fusion schemes, the skeptical question to ask is “what is the power density of your reactor?” For ITER, it’s 0.05 MW(th)/m^3. For MIT’s ARC and Lockheed’s concept, it’s about 0.5 MW(th)/m^3.

    For an existing commercial PWR fission reactor, the power density of the primary reactor vessel is 20 MW(th)/m^3.

    So, if you build a DT fusion reactor instead of a PWR, you’re multiplying the size of the reactor by more than an order of magnitude (and this is not just a design detail, but fundamental to the physics of the reactors), and the cost of the reactor likely goes up even more, as it is much more complex. How is this supposed to be competitive?

    Fusion energy is a perennial bad idea, just as air-breathing launchers are. You have to work much harder to very little benefit.

    1. For comparison, one of the large passenger steam locomotives could burn 20,000 lbs/coal per hour on a 100 sq ft grate.

      At 14,000 BTU/lb coal and 3400 BTU/kWHr and making some assumption about the primary firebox volume, the locomotive, if I haven’t missed a decimal, 4 MW thermal/m^3?

      This level is seriously “forcing” the combustion rate, and I am told the stationary power plants do not operate anywhere near that level? So is .4 MW/m^3 a reasonable figure for a hydrocarbon fuel-fired boiler?

      At 20 MW/m^3, a fission reactor core is putting out serious heat, I dare say. What is the power density of a natural-uranium fueled CANDU? Of some of the, what are they up to, Gen 4 proposals? Isn’t the idea to lower the power density for safety of operation, for better capability of natural convection cooling in the case of loss of backup power? That the high power density of the PWR is a carryover from Admiral Rickover wanting to stuff a reactor into a submarine?

      OK, OK, a magnetic confinement fusion reactor is a whole lot more resource intensive that a hydrocarbon fueled boiler, but even so, maybe .5 MW/m^3 is getting in the realm of workable?

      1. Fission reactors would be safer at lower power density. At the power density of a fusion reactor, the time scale for accidents would be greatly extended, due to the much larger thermal inertia of the core.

        So one might ask: why aren’t fission reactors designed with low power density? A big part of that is that it wouldn’t be economical. Fission’s problem is not that it’s unsafe or produces waste, but that it’s too costly.

        I cannot find a reference for the power density of a pulverized coal fired boiler. What ever that is, it doesn’t change the argument of fusion compared to fission. Unless the fusion reactor can dispense with major cost-driving elements of a fission reactor, the larger core will make it more costly. And this disregards the additional effect of RAMI (reliability availability maintainability inspectability) issues of those larger more complex fusion reactors.

        One driver of high power density fission cores is that small things tend to be more reliable than large things, as they have fewer parts and joins. Compared to a fission core, a fusion reactor is nightmare of complexity with many components for which redundancy is impossible. 2% of the fuel elements in a fission reactor can leak and it can still operate; a single leak in a cooled component of the first wall will shut down a fusion reactor.

        If you want to look at the champions of high power density, look at the burners of combustion turbines (or the thrust chambers of rocket engines). Not coincidentally, combustion turbine based plants are the last fossil fuel plants standing. Their cost per kW of capacity is a small fraction of that of a fission power plant.

  5. Machine learning needs examples of “good” vs. “bad”, or whatever other categories you want the machine to recognize. For a fusion reactor that would seem to imply a good model of the system on which parameters can be adjusted. Last I heard we didn’t have models that good.

Comments are closed.