Jump to content

SpaceX Discussion Thread


Skylon

Recommended Posts

Ballutes!  Suspend the cargo inside. Slow down up high with lower weight penalty.  The driving design rule being to maximize blunt surface area using as little mass as possible.  Keep it pressurized enough above Q all the way down to retain shape.

If big enough to keep temps low enough maybe even silicone rubber would work.  Might take a 60m radius sphere of 5 mil silicone rubber for a 15 metric ton payload according to grok3, fwiw.  4 to 5 m/s terminal velocity (!).

It lands where it lands, lol

Color it piebald like a big spherical cow.  What’s not to like?

Link to comment
Share on other sites

Large language models generally don't understand anything about either mathematics or the concepts they sprout.  I wouldn't trust them to actually calculate anything more complex than a straightforward unit conversion.  Most of them can't even correctly multiply two large numbers together.  ChatGPT once said that one of the advantage of an underwater telescope would be to study the stars without atmospheric distortion.  Consider them good at tasks like "When the training data talks about topic X, this word tends to come next after this other word".  That means they can often parrot back information that is already in their training data, but makes them terribly unreliable once you get to anything that is outside their training data.  In general, they do not have enough understanding to actually calculate anything, they just parrot back whatever information was in their training data.

Link to comment
Share on other sites

  On 3/27/2025 at 2:48 AM, AVaughan said:

In general, they do not have enough understanding to actually calculate anything, they just parrot back whatever information was in their training data.

Expand  

You forgot the word 'garbled' in your description.

If you trained a LLM on flower names and color names, gave it the dataset 'Roses are red and violets are blue' then the odds of answering 'what color are violets?' would be ~50% 'red' and ~50% 'blue' because it was not trained on 'and' or 'are' in this context and could only look at proximity in the dataset for clues.

I cannot imagine that there are many LLM that have been trained on enough rocketry information to have any context for the rocket equation beyond linking it to the name(if that).

Leave LLMs to what they are good(ish) at: writing fiction in the style of a prolific author that is well represented in their dataset.(the stories may not be any good, but I have confidence that they word usage and sentence structure will match very well)

Link to comment
Share on other sites

I asked Grok 3 the following:

"On a non rotating planet, launching from the equator, how much larger must a rocket be to place a satellite into a polar orbit relative to an equatorial orbit?"

Which is obviously a trick question. If the planet is not rotating, there is no equatorial boost, and thus there is no difference in the size of the rocket required.

The relevant part of Grok's response:

  Quote

tfe7Y68.png

Expand  

While it does correctly calculate the Delta-V to make a plane change of 90 degrees once you are in orbit, it fails to grasp that a plane change is not needed, then makes up everything after that point, including what appears to be an arbitrary multiplication by the square root of two to correct an "overestimation" of cost which actually ends up 5 m/s higher.

To its credit it does use real formulas everywhere, correctly reasons the orbital velocity, correctly describes the difference between an equatorial and polar orbit, and then afterwards has a wonderful long section where it does successfully reason through the rocket equation to find the required mass ratios to meet its (incorrect) Delta-V numbers.

However it then divides these mass ratios by each other to come up with the answer. I guess that is true when the only extra thing you need to add is fuel and you only have a single stage. But it is a little misleading, though I could have phrased the question a little better.

 

I did then remind Grok that the launch azimuth is not restricted to a specific direction and then it proceeded to absolutely nail the answer (though I specifically restricted the answer to Delta-V only).

 

I would still hesitate to use it for complex things I know nothing about because I have no way of knowing it was correct.

 

 

But to be honest I had a big "Oh no" moment there. The last time I used AI it couldn't even add up Delta-V numbers correctly or comprehend that landing on an atmospheric world took far less Delta-V than taking off from it. Before I asked this question I asked it about how shockwaves influence hypersonic heat shield design, and I couldn't actually tell if it was right or not. It was outside my ability. When I asked it about this I saw my entire profession flash before my eyes as I saw correct formula after correct formula pop onto the screen.

I was relieved when I read through it. But it couldn't do anything near this a year ago, and right now it made one bad assumption towards an admittedly poorly worded question. Maybe this is the practical limit, but if it makes the same level of improvement next year...

Link to comment
Share on other sites

  On 3/27/2025 at 3:42 AM, Ultimate Steve said:

I asked Grok 3 the following:

"On a non rotating planet, launching from the equator, how much larger must a rocket be to place a satellite into a polar orbit relative to an equatorial orbit?"

Which is obviously a trick question. If the planet is not rotating, there is no equatorial boost, and thus there is no difference in the size of the rocket required.

The relevant part of Grok's response:

While it does correctly calculate the Delta-V to make a plane change of 90 degrees once you are in orbit, it fails to grasp that a plane change is not needed, then makes up everything after that point, including what appears to be an arbitrary multiplication by the square root of two to correct an "overestimation" of cost which actually ends up 5 m/s higher.

To its credit it does use real formulas everywhere, correctly reasons the orbital velocity, correctly describes the difference between an equatorial and polar orbit, and then afterwards has a wonderful long section where it does successfully reason through the rocket equation to find the required mass ratios to meet its (incorrect) Delta-V numbers.

However it then divides these mass ratios by each other to come up with the answer. I guess that is true when the only extra thing you need to add is fuel and you only have a single stage. But it is a little misleading, though I could have phrased the question a little better.

 

I did then remind Grok that the launch azimuth is not restricted to a specific direction and then it proceeded to absolutely nail the answer (though I specifically restricted the answer to Delta-V only).

 

I would still hesitate to use it for complex things I know nothing about because I have no way of knowing it was correct.

 

 

But to be honest I had a big "Oh no" moment there. The last time I used AI it couldn't even add up Delta-V numbers correctly or comprehend that landing on an atmospheric world took far less Delta-V than taking off from it. Before I asked this question I asked it about how shockwaves influence hypersonic heat shield design, and I couldn't actually tell if it was right or not. It was outside my ability. When I asked it about this I saw my entire profession flash before my eyes as I saw correct formula after correct formula pop onto the screen.

I was relieved when I read through it. But it couldn't do anything near this a year ago, and right now it made one bad assumption towards an admittedly poorly worded question. Maybe this is the practical limit, but if it makes the same level of improvement next year...

Expand  

This tracks my experience.  Fun for ballparking something and fascinating to watch its process in “think” mode.

Grok 3 is currently the leader in LLM analytical math but it can error in surprising ways especially if the prompt leaves it room to do so.  Sometimes making the answer more complicated than necessary as your plane change experience showed.

 It is good at detecting problematic input.  I had specified kt instead of mt for weight units and it quickly decided that kilotons was ridiculous for the problem and assumed I meant metric tons after also quickly determining knots made no sense at all.  So that was interesting

  On 3/27/2025 at 3:22 AM, Terwin said:

cannot imagine that there are many LLM that have been trained on enough rocketry information to have any context for the rocket equation beyond linking it to the name(if that).

Expand  

Grok 3 in “think” mode is pretty far above that bar.

Link to comment
Share on other sites

  On 3/27/2025 at 1:11 AM, darthgently said:

Ballutes!  Suspend the cargo inside. Slow down up high with lower weight penalty.  The driving design rule being to maximize blunt surface area using as little mass as possible.  Keep it pressurized enough above Q all the way down to retain shape.

If big enough to keep temps low enough maybe even silicone rubber would work.  Might take a 60m radius sphere of 5 mil silicone rubber for a 15 metric ton payload according to grok3, fwiw.  4 to 5 m/s terminal velocity (!).

It lands where it lands, lol

Color it piebald like a big spherical cow.  What’s not to like?

Expand  

 There are several variations of this inflatable heat shield idea. The most researched one is a conical inflatable heat shield. It’s being investigated for example as a heat shield to make the Cygnus cargo capsule reusable:

NaDe6r8iN44sVNnGCwfFma.jpg

 

 Here’s a research article on it:

HEART FLIGHT TEST OVERVIEW
9th INTERNATIONAL PLANETARY PROBE WORKSHOP 16-22 JUNE 2012, TOULOUSE
https://websites.isae-supaero.fr/IMG/pdf/137-heart-ippw-9_v04-tpsas.pdf

 As discussed there the parameter used to measure the capability of a particular shape to slow down descent is not wing loading, weight divided by wing area, but the ballistic coefficient, (mass)/(drag coefficient*drag area), β = m/CDA, given in metric units. This takes into account the fact different shapes are more effective in slowing down the spacecraft by including the coefficient of drag CD as well as being more general than just looking at wings for the decelerator.

 In this report, the mass used for their analysis is 5,000 kg and the diameter of their conical decelerator is 8.3 meters. There is thermal protection applied but I gather less of it is needed since the conical aeroshell is just made of silicone rubber. 

 If it is just ballistic coefficient determining this then for a spacecraft or stage about 9 times heavier, say, 40,000+ kg, then the area needs to be 9 times more, that is, a conical shell about 25 meters in diameter.

 This is useful for just drag decelerators, but is incomplete for winged reentry because it does not include the effects of lift. For instance if wings with high lift/drag ratio at hypersonic speeds were used the descent rate would be decreased even further. The hypersonic aerodynamics of the Space Shuttle have been described as falling “like a brick.” Then wings with high L/D ratio could greatly improve on this.

  Bob Clark

Edited by Exoscientist
Typo
Link to comment
Share on other sites

 New high temperature ceramic heat shield materials promoted to SpaceX:

AN OPEN LETTER TO ELON MUSK.
Dr. Ed Pope
https://www.linkedin.com/feed/update/urn:li:activity:7306136414910300160

Published 3/14/2025
MATECH, Cal Nano aim to commercialize UHT composites.
The partnership’s combined technological advances and manufacturing prowess will target the scale-up and industrialization of FAST SPS for high-temp and UHT composites serving aviation, defense.
https://www.compositesworld.com/news/matech-cal-nano-aim-to-commercialize-uht-composites

  Bob Clark

Edited by Exoscientist
Link to comment
Share on other sites

  On 3/27/2025 at 8:33 PM, Exoscientist said:

New high temperature ceramic heat shield materials promoted to SpaceX:

Expand  

Normally I read your posts with my "Oh, this should be good" glasses, but no, this is very interesting. 

As Dr. Ed Pope says "{T}he current SpaceX approach is to make incremental improvements to outdated materials technology in the hope that its many deficiencies can be overcome. It’s 'throwing good money after bad.' Even worse, it’s wasting precious time."

For those who didn't read the link, the technology being promoted is basically to make a reusable, ablatorless, tough carbon material called C/ZrOC that can be made into a thin aeroshell heat shield. C/ZrOC can be made large and in complex shapes, with insulating foam or blankets behind.

Link to comment
Share on other sites

  On 3/28/2025 at 8:13 PM, Meecrob said:

For those who didn't read the link, the technology being promoted is basically to make a reusable, ablatorless, tough carbon material called C/ZrOC that can be made into a thin aeroshell heat shield. C/ZrOC can be made large and in complex shapes, with insulating foam or blankets behind.

Expand  

Grok3 search and analysis of the MATECH claims and related info about the material.  Will be interesting to see what pans out:

(I verified all the key citation links. The second to last citation link is the only one that 404s and this looks like a news feed that likely purges old news and grok referred to an old cache perhaps.  Or hallucinated)

  Reveal hidden contents
  •  
Link to comment
Share on other sites

  On 3/28/2025 at 8:13 PM, Meecrob said:

Normally I read your posts with my "Oh, this should be good" glasses, but no, this is very interesting. 

As Dr. Ed Pope says "{T}he current SpaceX approach is to make incremental improvements to outdated materials technology in the hope that its many deficiencies can be overcome. It’s 'throwing good money after bad.' Even worse, it’s wasting precious time."

For those who didn't read the link, the technology being promoted is basically to make a reusable, ablatorless, tough carbon material called C/ZrOC that can be made into a thin aeroshell heat shield. C/ZrOC can be made large and in complex shapes, with insulating foam or blankets behind.

Expand  

to be fair the current problem is not the tps, but it certainly is slowing down the testing of same.

Link to comment
Share on other sites

  On 3/29/2025 at 1:53 AM, Nuke said:

to be fair the current problem is not the tps, but it certainly is slowing down the testing of same.

Expand  

No excrement, lol, but you can work on more than one thing at once, right? The thermal is an aspect that isn't in line with the order-of-magnitude type changes on SS/SH

Link to comment
Share on other sites

  On 3/29/2025 at 2:33 AM, Meecrob said:

No excrement, lol, but you can work on more than one thing at once, right? The thermal is an aspect that isn't in line with the order-of-magnitude type changes on SS/SH

Expand  

i mean they tested their new tiles in the plasma wind tunnel, only thing left is a live test that you cant do if your ship blows up before it gets mostly to orbit.

Link to comment
Share on other sites

  On 3/28/2025 at 8:13 PM, Meecrob said:

Normally I read your posts with my "Oh, this should be good" glasses, but no, this is very interesting. 

As Dr. Ed Pope says "{T}he current SpaceX approach is to make incremental improvements to outdated materials technology in the hope that its many deficiencies can be overcome. It’s 'throwing good money after bad.' Even worse, it’s wasting precious time."

For those who didn't read the link, the technology being promoted is basically to make a reusable, ablatorless, tough carbon material called C/ZrOC that can be made into a thin aeroshell heat shield. C/ZrOC can be made large and in complex shapes, with insulating foam or blankets behind.

Expand  

 If this new high temperature material is able to solve the Starship thermal protection issues, I think SpaceX would accept this relatively mild criticism anyway.

  Bob Clark

Link to comment
Share on other sites

  On 3/29/2025 at 3:20 PM, Exoscientist said:

 If this new high temperature material is able to solve the Starship thermal protection issues, I think SpaceX would accept this relatively mild criticism anyway.

  Bob Clark

Expand  

I have a feeling SpaceX had a plan B with regards to thermal even before you brought this to my attention. SS/SH is shaping up to be a game-changer, but the one part that seemed off was the tiles. That's why I was seriously thinking that SpaceX was aiming to have the refuel ships without tiles. This changes everything though, as in if this technology performs as intended, its kinda like a cheat code almost.

  On 3/29/2025 at 5:05 AM, Nuke said:

i mean they tested their new tiles in the plasma wind tunnel, only thing left is a live test that you cant do if your ship blows up before it gets mostly to orbit.

Expand  

Oh, my apologies. I took the wrong meaning from your comment.

Edited by Meecrob
Link to comment
Share on other sites

I feel that Spacex wants to do everything inhouse so they probably won't be looking into that. I guess making everything inhouse makes it cheaper and there is no fear pf supplier going bust.

 

One option might be for Spacex to license that technology so they can build it by themselves for their use and not for sale.

Link to comment
Share on other sites

You talk like SpaceX should be embarrassed....Far from it, my man! SpaceX basically invented new Space.  They get to call the shots now....As much as I love NASA, the torch needs to be passed. Someone find that graphic of the amount of launches from countries around the world vs. SpaceX....and SpaceX is just breaking a sweat I imagine

  On 3/29/2025 at 3:20 PM, Exoscientist said:

 If this new high temperature material is able to solve the Starship thermal protection issues, I think SpaceX would accept this relatively mild criticism anyway.

  Bob Clark

Expand  

 

I still feel that Elon has cards up his sleeve that he is waiting to play.

Edited by Meecrob
Link to comment
Share on other sites

No official word, but the astronauts haven't gone to the pad yet and the weather's getting worse. The T-zero is still counting down, though.

Link to comment
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

×
×
  • Create New...