Comment on Test firing of a 3D-printed rocket engine designed through computational modelComments−InDubioProRubio2yWhere are the performance spec sheets?−0xfae2yThe second line of the article: "The engine with 5 kN (500 kg / 1124 lbf) of thrust, generated the expected 20,000 horsepower"−echoangle2ySpecific impulse would also be nice, thrust itself isn’t very impressive−jameshart2yI think the impressive thing here is more that they asked the computer to design an engine which produced a desired amount of thrust, and it did so.The thing being demonstrated isn’t ‘we made this engine, look how impressive its specs are’, it’s ’our computer can make working engines to spec’−greesil2yI suspect they're happy it didn't just blow up.−InDubioProRubio2yWhy downvote? Its a great design, by an ai, it should compete with human designs in all the specs?−dgacmu2yI didn't downvote you, but it seems like a pretty nitpicky question to ask at the "we demonstrated that the thing worked and didn't melt" stage.It shouldn't have to compete - yet. Your question doesn't get at the more interesting one, which is, "will it eventually?"−InDubioProRubio2yOn contrair el capitan: It is trained on human designs and design processes. It has knowledge on physics and flow mechanics + propulsion simulations baked into it. It should outcompete a human, who gains that intuition by try and terror.−dgacmu2yThere's a saying from Rod Brooks that I quite like: "Simulations are doomed to succeed."The hard part is making sure the simulation captures enough of the relevant factors and that the physics/etc simulation is accurate enough.
Comments
Where are the performance spec sheets?
The second line of the article: "The engine with 5 kN (500 kg / 1124 lbf) of thrust, generated the expected 20,000 horsepower"
Specific impulse would also be nice, thrust itself isn’t very impressive
I think the impressive thing here is more that they asked the computer to design an engine which produced a desired amount of thrust, and it did so.
The thing being demonstrated isn’t ‘we made this engine, look how impressive its specs are’, it’s ’our computer can make working engines to spec’
I suspect they're happy it didn't just blow up.
Why downvote? Its a great design, by an ai, it should compete with human designs in all the specs?
I didn't downvote you, but it seems like a pretty nitpicky question to ask at the "we demonstrated that the thing worked and didn't melt" stage.
It shouldn't have to compete - yet. Your question doesn't get at the more interesting one, which is, "will it eventually?"
On contrair el capitan: It is trained on human designs and design processes. It has knowledge on physics and flow mechanics + propulsion simulations baked into it. It should outcompete a human, who gains that intuition by try and terror.
There's a saying from Rod Brooks that I quite like: "Simulations are doomed to succeed."
The hard part is making sure the simulation captures enough of the relevant factors and that the physics/etc simulation is accurate enough.