Starship is Go...

SpaceNut · 2020-04-02 13:19:18

How so when the whispery winds of mars can sand blast things is problematic for retro propulsion for mars....

louis · 2020-04-02 13:49:05

Winds with 1/20th the force of winds on Earth...have you seen how breezy it is at Boca Chica? That's sand blasting!

SpaceNut wrote:

How so when the whispery winds of mars can sand blast things is problematic for retro propulsion for mars....

SpaceNut · 2020-04-02 14:18:17

Its the removal of sand from the legs pad which makes the landing on mars the problem....for retro propulsion since its so easily moved....
Even at 1/20th the engines are lots stronger than that for blasting the sand away from it landing.....

louis · 2020-04-02 16:49:03

No one is planning on landing in a sandy area.

SpaceNut wrote:

Its the removal of sand from the legs pad which makes the landing on mars the problem....for retro propulsion since its so easily moved....
Even at 1/20th the engines are lots stronger than that for blasting the sand away from it landing.....

louis · 2020-04-02 17:34:12

I accept whatever landing leg system they have for SN3 will not be good for a landing on Mars. Let's just take that as read. What will happen before even an unmanned cargo ship heads for Mars is that there will be multiple tests on ground on Earth similar to the proposed LZ on Mars. Furthermore the additional gravity on Earth will make this test robust.

Of course a cargo ship landing failure may well kill Mission One, for sure. But (a) there is no reason to assume that will be likely (I would say other aspects of the EDL are far better candidates for mission failure) and (b) it wouldn't mean the end of a Mission to Mars - well not as long as Elon Musk is in control.

GW Johnson wrote:

Louis:
The point is that it is unlikely in the extreme that a Starship can make a successful landing, whether as an unmanned cargo craft or not. It is far more likely to penetrate in unevenly, which forces it to lean, and very little leaning makes a tall thing topple over. It WILL explode when it does topple over.
If Spacex sends 2-3 unmanned cargo ships to Mars, and they all get destroyed upon touchdown, then what is the point of sending any more? It's the wrong design to land safely. That's a program killer, and you bloody well ought to understand that.
I've still got to look at Mark's "DrPhil" stuff.
GW
PS - initial look at "DrPhil" stuff is discouraging in that there may be another bad problem to solve, but encouraging in that at least somebody is looking at it. Meanwhile the surface bearing pressure/penetration issue and the static stability/topple-over issue that I keep raising still apply here.
The landing mass for Starship on Mars is as yet an unpublished unknown, but an educated guess says perhaps 120 m.ton inert structural mass, plus 100 ton cargo plus around 10 tons unused propellant, for about 230 metric tons. For the 6-legged design they currently show, that's about 38 tons per leg to support. At 0.38 gee, that's about 143 KN static force per leg.
Double it for the touchdown impact-effect dynamics to about 289 KN force per leg, which is typical design practice. Now guess a landing pad surface area as a circle about 0.5 meter diameter. That's something on the order of 0.196 sq.m bearing area per leg. the peak dynamic touchdown bearing pressure is leg force/pad area which is just a hair under 1.5 MN/sq.m = 1.5 MPa. The static pressure is near 0.75 MPa, which would be OK.
Typical fine sand here on Earth has a max allowable bearing pressure limitation of 1-2 MPa. You have to treat it as "the weakest link",
which is 1 MPa. That kind of surface is identical to 98+% of the surface of Mars! So landing the Starship is "iffy" at best. Once the landing dynamics are over, you are OK, but the risk here is "stabbing" the legs into the regolith during the transient.
If that happens, they WILL penetrate unevenly, and the ship will very likely topple over, precisely because it is tall and narrow. 5-6 degrees out-of-plumb is fatal to a geometry like that.
As for refueling on Mars for the return, the loaded weight is around 6+ times higher, so the bearing pressure is also 6+ times higher. Transient effects don't apply, so that is 6+ x 0.75 MPa = 4.5+ MPa compared to max allowable 1-2 MPa. It'll sink in unevenly and topple-over (and explode) as you try to refill it. Rather certainly.
I'm sorry to bust dreams here, but the numbers DO NOT lie!
Those same numbers are about 3 times higher here for Earth gravity. That increases the rough-field disaster risk. Which is why I keep saying Spacex is going to run smack into this design shortcoming, the first time they have to do a rough-field abort landing, which will hopefully be while they are still in flight test. And they will run into this. Murphy's Law says so.

louis · 2020-04-02 17:46:35

Another video with v. interesting detail on how Space X are pressurising tanks in flight...

https://www.youtube.com/watch?v=KhaBU87qVGY

Felix emphasises that Space X is moving at top speed in the development.

Starship User Guide has been published. Felix has high praise for the fairing design (a flexible system that allows for multiple uses).

kbd512 · 2020-04-02 18:22:05

Louis,

The overriding point some of the rest of us are trying to make is that we want that first mission to be a success, which means the landing gear for Starship needs a redesign. That's another way of saying SpaceX needs to iterate Starship's design again to make it work here on Earth. If it's supposed to have rough field capabilities, which is a requirement for both the moon and Mars, then that means it needs to be part of the vehicle design from the word "go".

louis · 2020-04-02 18:29:44

I am not sure "redesign" is the right word since I don't believe Musk or the Space X team think SN3 is the last word on leg design. A lot of things have to be sorted before the team focusses on leg design for a Mars landing. But that time will come and then it will be extensively tested on Mars-like landscapes on Earth.

It will need to be a robust system but I think it will be landing on a less than 5% gradient and there won't be any large boulders in the area. We'll probably be talking about rocks and stones less than 25 cms across.

kbd512 wrote:

Louis,
The overriding point some of the rest of us are trying to make is that we want that first mission to be a success, which means the landing gear for Starship needs a redesign. That's another way of saying SpaceX needs to iterate Starship's design again to make it work here on Earth. If it's supposed to have rough field capabilities, which is a requirement for both the moon and Mars, then that means it needs to be part of the vehicle design from the word "go".

kbd512 · 2020-04-02 18:42:56

Louis,

Whatever. You can design it to land in rough terrain or watch it topple over and explode. I don't know why anyone would be obtuse about that point.

SpaceNut · 2020-04-02 18:44:58

The Apollo Lem had detailed images and we almost lost one due to a boulder in the field that was to be flat if it was not for turning off the autolanding and doing a manual control since man saw what machines had not.
Starship will have no such landing capability to see the ground as it lands and will need to be able to avert an automated missed field of landing area. There is no landing Ai being developed for obstacle avoidance.

tahanson43206 · 2020-04-02 18:50:37

For SpaceNut re #585

Why do you think this:

There is no landing Ai being developed for obstacle avoidance.

I expect that every space faring nation has teams hard at work on that problem, and the Chinese may have already successfully deployed it on the back side of the Moon.

It is an essential component of any reasonable plan to land an automated vehicle on any site.

Perhaps you were expecting someone would tell you about their work on such a proprietary component of their landing system?

(th)

SpaceNut · 2020-04-02 19:32:55

The true sense of AI is that it makes choices like you and can from informational reasoning without any set of instruction tree to follow. Ai can carry on a conversation and ask questions not respond to the question.
Automation command that has an instruction fault tree of inputs is not AI.

tahanson43206 · 2020-04-03 04:19:42

For SpaceNut re #587

Thanks for your clarification of what you meant by an Artificial Intelligence landing system.

Let's mark this date, April of 2020, as one in which you have asserted that no AI systems are in development in research laboratories on Earth.

I consider AI to be working right now in many specialized applications where computer systems are far superior to humans. The ability to have a conversation is not required for AI to be useful, effective and far superior to human capability.

Perhaps we are dealing here with a matter of definition.

I consider the achievement of the SpaceX vehicle landing software engineering team to be a demonstrable form of AI. However, I don't ** know ** that the SpaceX team has used neural networks for their solution, so I am holding back from offering certainty on the point.

From my perspective, an AI exists if the software is capable of learning, which is certainly the case with the chess solution by Big Blue, or the Jeopardy solution by Watson.

It seems to me premature to decide that the Chinese landing solution has learning ability, or if it is just traditional clever human programming.

Eventually (it seems likely) we will find out.

(th)

SpaceNut · 2020-04-03 08:31:04

The computer did not start with no coding, it did not start with no definition of what each piece or what it can do for a move, it did not start with no decision making for a move, it did not start with these are all loop parts to a programming tree. It does not store moves made to make a winning game it only plays the moves presented for probability to gain a statistic to winning outcome. The computer if its stores winning moves has another loop which does not allow it to enter a failing move. Does it make all the moves while none are made from the opponent..so there is another programming loop for a wait decision...This is all done at the computers clock speed, slow the clock down and it plays to a timeout for each move.

A space x AI landing has no chess board or markings, its not able to identify the pieces of the game, no meanings to what move it can make, ect... its a computer program series of loops with probability for what choice can be made that improves probability of success...

tahanson43206 · 2020-04-03 12:18:25

For SpaceNut re Deep Learning vs Control Programming

This topic came up on Quora, and several people who came across to me as knowledgeable took a stab at answering the question.

https://www.quora.com/Does-SpaceX-use-a … nd-rockets

There seemed to be general support for your prediction, that SpaceX may be using traditional programmed control subsystems to achieve successful tail-first rocket vehicle landings.

However, several contributors explained Deep Learning for the Quora audience, and one offered the opinion that if SpaceX ** is ** using Deep Learning to control any part of it's vehicle flight, it would be proprietary and ** extremely ** valuable to keep secret for as long as possible.

What I came away with is the understanding that NO ONE (outside SpaceX) has any idea what SpaceX is doing, beyond what they reveal in their patents.

And most CERTAINLY, no one outside China has any idea what computer programs are running in their probe on the back side of the Moon.

Artificial Intelligence research has been going on full throttle for decades. The field has explored some ultimately disappointing branches, but I get the impression the rate of progress, and the capability of the resulting non-programmed systems are increasing dramatically.

I am pretty sure the US Air Force is pursuing Artificial Intelligence with vigor, so found this report on initiatives from 2019:
https://emerj.com/ai-sector-overviews/a … air-force/

The WHOLE point of using AI is NOT to code in Cobol, or Basic, or FORTRAN , or any of the step-by-step languages that have evolved over the years.

Edit#1: It appears I owe you a clarification of my own understanding of the term "artificial intelligence"

The meaning of that term has (apparently) evolved .... What you (correctly) identified as traditional programming (Deep Blue Chess program) was characterized in popular reporting as "artificial intelligence".

At this link: https://en.wikipedia.org/wiki/Deep_Blue … _computer)

Design
Deep Blue employed custom VLSI chips to execute the alpha-beta search algorithm in parallel,[14] an example of GOFAI (Good Old-Fashioned Artificial Intelligence) rather than of deep learning which would come a decade later. It was a brute-force search approach, and one of its developers even denied that it was artificial intelligence at all.[15]

I'm picking up on the argument in the last line, that brute force is deprecating as a method for "artificial intelligence".

That program and ones like it certainly are superior to human beings, but so is a calculator, within the small subset of intellect it represents.

My understanding of "Artificial Intelligence" (the term) has been evolving without my realizing it, so that now I consider it to apply ONLY to self-learning systems, and I agree (if that is what you were saying) that there are no published reports of use of self-learning systems to operate space vehicles.

However, the problem of landing on the Moon or Mars or anywhere else (including the Earth) would appear to be an excellent problem to be addressed by self-learning computer systems.

(th)

Last edited by tahanson43206 (2020-04-03 13:11:39)

SpaceNut · 2020-04-03 12:45:54

One of the ways which most think of is to have the initial stuff in flash chips that can be removed with a new pointer to the learner code in another area for startup access for when power is reapplied. The ability to learn starts after it can not find the means to execute the logic to which its starting with. Depending on the remaining code it must start to create its own language to program its self with from where it starts over with. Sensor must have direct interupts to poke hidden from view of starting code that it runs to make meaning and its that repeated inputs that bring it to a thinking state by filling in what it did not have to start with.
Basically an electronic baby....

louis · 2020-04-03 18:52:26

SN3 implodes...

https://www.youtube.com/watch?v=D-RP4eQi2pk

Sad...but now the focus is on SN4.

That's how it goes with Space X!

kbd512 · 2020-04-03 23:52:24

Louis,

When SN4 meets the same fate, is there a SN5 under construction?

Maybe it's time to slow down just a little to analyze what's happening and also to start using proper tooling?

Mark Friedenbach · 2020-04-04 04:11:47

kbd512 wrote:

Louis,
When SN4 meets the same fate, is there a SN5 under construction?
Maybe it's time to slow down just a little to analyze what's happening and also to start using proper tooling?

It's an assembly line. There's at least an SN4, SN5, and SN6 already under construction. They're re-tooling the assembly line based on the tests. But we can expect a new SNx basically one every other week (eventually every week) until it works or Musk runs out of money.

louis · 2020-04-04 04:48:09

I think I heard somewhere that SN5 or 6 were sheduled for orbital flight...they might well be missing that target now if SN4 goes kaput.

They are pretty much building a production line at Boca Chica. I think the only thing that can stop further development is if Tesla go belly up in the self-inflicted mega-recession coming up.

kbd512 wrote:

Louis,
When SN4 meets the same fate, is there a SN5 under construction?
Maybe it's time to slow down just a little to analyze what's happening and also to start using proper tooling?

kbd512 · 2020-04-04 07:00:53

Louis,

Well, at least the sheet steel is cheap to play with. I really do think this can work, but only if they put some serious engineering behind it and stop trying to crank out another one as fast as humanly possible. I know that they're smart and capable, but this is increasingly looking like tossing darts over your shoulder at a wall and hoping something sticks. Stainless is a very old and well-proven technology, insofar as upper stages are concerned, but everything we're seeing looks like inexperience with using it and/or very poor quality control. I'd like to see a successful pressure test, successful landing, and stricter QC program, but I'll chalk up what I've seen thus far to "new technology" (for them). ULA makes stainless work with boring reliability, despite the extreme measures they've taken to reduce weight.

SpaceNut · 2020-04-04 07:46:22

Contract with a vendor to purchase a tank of the right size and get past the problem.

kbd512 · 2020-04-04 18:45:46

SpaceNut,

Maybe we should start "Tanks R Us" or "Tanks You Very Much"?

I like working with sheet steel. This will be fun.

Any other takers?

SpaceNut · 2020-04-04 19:17:04

I think that was taken by the ULA when it combined Boeing and Lockheed to make them more competitive...they should have plenty of discontinued tanks for cheap...

Mark Friedenbach · 2020-04-04 20:42:53

kbd512 wrote:

Louis,
Well, at least the sheet steel is cheap to play with. I really do think this can work, but only if they put some serious engineering behind it and stop trying to crank out another one as fast as humanly possible. I know that they're smart and capable, but this is increasingly looking like tossing darts over your shoulder at a wall and hoping something sticks. Stainless is a very old and well-proven technology, insofar as upper stages are concerned, but everything we're seeing looks like inexperience with using it and/or very poor quality control. I'd like to see a successful pressure test, successful landing, and stricter QC program, but I'll chalk up what I've seen thus far to "new technology" (for them). ULA makes stainless work with boring reliability, despite the extreme measures they've taken to reduce weight.

You are looking at it incorrectly. You are evaluating the performance of individual rockets coming off the assembly line. They are looking at it in terms of performance of the assembly line itself. There’s time to correct the engineering of the craft, but they will do so within the constraints of the assembly line.

It’s no coincidence that this strategy follows Musk’s do-or-die ramp up of Tesla’s automotive production line. You can’t just design anything and expect to be able to build it at scale, as Elon learned. You have to design for mass manufacturing. That is what SpaceX is doing right now.

Last edited by Mark Friedenbach (2020-04-06 05:01:49)

New Mars Forums

Announcement

#576 2020-04-02 13:19:18

Re: Starship is Go...

#577 2020-04-02 13:49:05

Re: Starship is Go...

#578 2020-04-02 14:18:17

Re: Starship is Go...

#579 2020-04-02 16:49:03

Re: Starship is Go...

#580 2020-04-02 17:34:12

Re: Starship is Go...

#581 2020-04-02 17:46:35

Re: Starship is Go...

#582 2020-04-02 18:22:05

Re: Starship is Go...

#583 2020-04-02 18:29:44

Re: Starship is Go...

#584 2020-04-02 18:42:56

Re: Starship is Go...

#585 2020-04-02 18:44:58

Re: Starship is Go...

#586 2020-04-02 18:50:37

Re: Starship is Go...

#587 2020-04-02 19:32:55

Re: Starship is Go...

#588 2020-04-03 04:19:42

Re: Starship is Go...

#589 2020-04-03 08:31:04

Re: Starship is Go...

#590 2020-04-03 12:18:25

Re: Starship is Go...

#591 2020-04-03 12:45:54

Re: Starship is Go...

#592 2020-04-03 18:52:26

Re: Starship is Go...

#593 2020-04-03 23:52:24

Re: Starship is Go...

#594 2020-04-04 04:11:47

Re: Starship is Go...

#595 2020-04-04 04:48:09

Re: Starship is Go...

#596 2020-04-04 07:00:53

Re: Starship is Go...

#597 2020-04-04 07:46:22

Re: Starship is Go...

#598 2020-04-04 18:45:46

Re: Starship is Go...

#599 2020-04-04 19:17:04

Re: Starship is Go...

#600 2020-04-04 20:42:53

Re: Starship is Go...

Board footer