4 ms·

You're missing nothing, it's pure PR fluff.

rbanffy 4y ago
Perhaps. We don't know if they sold the design and whether it was tested. The design itself is very interesting - in particular the complex heat exchange structures. On it being optimal, or even more efficient, as long as it's cheaper to build than a conventional engine, it's game.
- ImPostingOnHN 4y ago
  we know the engine wasn't tested, and we do not yet have enough evidence to reject the null hypothesis that the engine is unsold
- s1artibartfast 4y ago
  I think all of that is basically out of scope. I don't think they are claiming it is a good design. The impressive point is their software can make a design at all. It seems like they want to sell it to people that will use it to try to make designs that are good
  - ImPostingOnHN 4y ago
    algorithmic design on its own, with no further qualifiers (like functionality) isn't impressive though, I don't think that's the pitch
    - s1artibartfast 4y ago
      I think this level of algorithmic design is impressive, and as far as I know, novel. I certainly haven't seen anything this complex. It is like a computer displaying "hello world". The output isn't particularly exciting, I could do better with an index card. However, the fact that the a computer could take a set of inputs and display something at all shows exciting promise. The same is true here. For a real engine design, you would have to work extensively on the inputs, boundaries, and goals. The impressive part is that this software can handle this complex of a parameter set at all.
      - ImPostingOnHN 4y ago
        as far as we know, "this level of algorithmic design" could be randomly generated volumes which serve no purpose and render the device completely non-functional indeed, since it does't need to do anything but look cool, the inputs could be "weirdness factor" and "excessive internal wall ratio" and nothing else, and they thought "this kinda looks like an aerospike"
        s1artibartfast 4y ago
        >as far as we know... This is exactly why I am impressed. I have some basic understanding of rocket engine and aerospike design, and can recognize most of the features and the functions they would serve in this design. I have no clue how optimal they are, but it is clear that they are taking function into consideration. Moving from the outside of the engine in, you can see that support lattice has been added to the exterior and density adjusted on the exterior to reenforce high stress areas. Next, the coolant jacket around the combustion chambers has has baffles (verticals s curves) that seem to adjust cooling in specific areas. It looks like the spike cooling system (the dense horizontal lines) is trying to balance fluid resistance with how much the cryogenic oxygen. The one part I don't understand is what the Y shaped channels in the very center are supposed to do. This is the liquid oxygen flow to the tip of the spike before it returns along the outside of the spike as a heat exchanger (mentioned above). It looks like some sort of expansion chamber but I would have to ask a professional aerospike engineer what it is for. This is a long way of saying the structures show functional intent to converge on some solution. Some parts seems random, but most algorithmic design is random change and iterative testing; Essentially evolutionary code, where permutations are made, compared, killed off, then fed into future models. I wonder how they handle so many degrees of freedom, and how much of what we see in the output was pre-defined.
        ImPostingOnHN 4y ago
        Seems like a description of Homer Simpson's nuclear power plant yeah, it has a racing stripe, and fins, like some stuff which goes fast also has, but that doesn't mean that adding it to a power plant makes the power plant fast as another engineer that's worked in aerospace, none of those functional design aspects seem to be actually performing a function here other than "we saw this on something weird looking", as in not enough evidence that there's functional intent converging on something functional

Perhaps. We don't know if they sold the design and whether it was tested. The design itself is very interesting - in particular the complex heat exchange structures. On it being optimal, or even more efficient, as long as it's cheaper to build than a conventional engine, it's game.

ImPostingOnHN 4y ago
we know the engine wasn't tested, and we do not yet have enough evidence to reject the null hypothesis that the engine is unsold
s1artibartfast 4y ago
I think all of that is basically out of scope. I don't think they are claiming it is a good design. The impressive point is their software can make a design at all. It seems like they want to sell it to people that will use it to try to make designs that are good
- ImPostingOnHN 4y ago
  algorithmic design on its own, with no further qualifiers (like functionality) isn't impressive though, I don't think that's the pitch
  - s1artibartfast 4y ago
    I think this level of algorithmic design is impressive, and as far as I know, novel. I certainly haven't seen anything this complex. It is like a computer displaying "hello world". The output isn't particularly exciting, I could do better with an index card. However, the fact that the a computer could take a set of inputs and display something at all shows exciting promise. The same is true here. For a real engine design, you would have to work extensively on the inputs, boundaries, and goals. The impressive part is that this software can handle this complex of a parameter set at all.
    - ImPostingOnHN 4y ago
      as far as we know, "this level of algorithmic design" could be randomly generated volumes which serve no purpose and render the device completely non-functional indeed, since it does't need to do anything but look cool, the inputs could be "weirdness factor" and "excessive internal wall ratio" and nothing else, and they thought "this kinda looks like an aerospike"
      - s1artibartfast 4y ago
        >as far as we know... This is exactly why I am impressed. I have some basic understanding of rocket engine and aerospike design, and can recognize most of the features and the functions they would serve in this design. I have no clue how optimal they are, but it is clear that they are taking function into consideration. Moving from the outside of the engine in, you can see that support lattice has been added to the exterior and density adjusted on the exterior to reenforce high stress areas. Next, the coolant jacket around the combustion chambers has has baffles (verticals s curves) that seem to adjust cooling in specific areas. It looks like the spike cooling system (the dense horizontal lines) is trying to balance fluid resistance with how much the cryogenic oxygen. The one part I don't understand is what the Y shaped channels in the very center are supposed to do. This is the liquid oxygen flow to the tip of the spike before it returns along the outside of the spike as a heat exchanger (mentioned above). It looks like some sort of expansion chamber but I would have to ask a professional aerospike engineer what it is for. This is a long way of saying the structures show functional intent to converge on some solution. Some parts seems random, but most algorithmic design is random change and iterative testing; Essentially evolutionary code, where permutations are made, compared, killed off, then fed into future models. I wonder how they handle so many degrees of freedom, and how much of what we see in the output was pre-defined.
        ImPostingOnHN 4y ago
        Seems like a description of Homer Simpson's nuclear power plant yeah, it has a racing stripe, and fins, like some stuff which goes fast also has, but that doesn't mean that adding it to a power plant makes the power plant fast as another engineer that's worked in aerospace, none of those functional design aspects seem to be actually performing a function here other than "we saw this on something weird looking", as in not enough evidence that there's functional intent converging on something functional