Debate on SpaceX Starship development methodologies

[ChrML]

After this series of explosions and delays isn't it time to question the paceX Starship development methodologies appropriatness?

The SN4 explosion is for me (should be) is a turning point for SS developement programe, SX can no longer do as nothing did happen, they should make a dramatic decision by switching to a classical approach, with fewer but more elaborate prototypes, a switch to classical materials to approve the design while delaying the stainless steel for cost saving for later developements.

Stuborness lead to nowhere but failer.I'm getting more and more depressed with those explosions.

Classic approach, you mean like SLS time and budget? How is that more successful?

An exploded Starship only sets them back for 3 weeks.

[Nomadd]

Disagree! No company should be allowed to blow things around every few weeks.  It scares and even panics people around, causes environmental damage, and is a dangerous practice. Somebody should be hold responsible.

Nobody was scared or panicked, the only environmental damage I saw was the bush we use to stand near to take photos got charred, and nobody was ever in any danger. There wasn't a single dead animal or bird in sight and the bush was an invasive species.

[Lemurion]

Disagree! No company should be allowed to blow things around every few weeks.  It scares and even panics people around, causes environmental damage, and is a dangerous practice. Somebody should be hold responsible.

SpaceX has been held responsible; they are required to carry insurance and handle any necessary cleanup. They have also filed the required environmental impact reports and close the immediate area around the pad during testing. Also, there are very few people in the Boca Chica area and rather than being panicked they seem to be focusing on their new careers as photojournalists documenting Starship development.

As it is, the explosions allow all the SpaceX haters on various platforms a chance to write thousands of words that reflect their complete lack of understanding of what's actually going on and the people here endless fuel for speculation.

From the armchair space analyst's position the explosions are a godsend, and a far better entertainment value than cable TV.

[OTV Booster]

Uh oh, a quick-disconnect failed during a test of a quick-disconnect on a test stand! Quick, SpaceX, abandon stainless steel and your entire production system! /s

I’m getting thoroughly hacked off with the attitude of some people on this forum. Expressing an opinion that SpaceX may have gone too far in cutting engineering oversight is not concern trolling. Constantly deriding people who dare to dissent, many of whom are long standing fans, and appalling sarcasm like this is trolling in my book.

SpaceX does get things wrong. The fault that lost SN4 and the launch mount may well have been caught by an hour HAZOP study. There may be a more optimum approach between the paralysis by analysis of SLS and the crazy mistakes of Boca Chica. It is a valid topic for discussion, and this forum was created specifically to allow it.

It would not, because HAZOP studies don't push operational boundaries.
SpaceX does indeed occasionally things wrong. The thing is they learn from it.
We don't know what started failure cascade. What we know is how it ended. What I know that the events will rise a lot of questions most probably missed or postponed being "non important". It is always the case with the ruds, and it is always fantastic learning experience when treated right. your attitude is not.
P.S. "status" (being "long standing fans"= what is it actually?) as an argument? really? people write ridiculous things and get ridiculed as result.

I've been working up this thread and holding back on comments. One thing that has impressed me is how few food fights there have been on a touchy subject and how the participants have been self policing when a food fight does loom.

It's now time for a comment (3-4 pages yet to read). To quote the immortal Tom, "Just stop it!" Some if you are collectively becoming Tom.

Disagreeing with someone, especially in a topic that resolves down to opinion, is no excuse for disrespect. If you find yourself becoming frustrated, refine your arguments, don't dis. Even if you don't convince, refining your arguments should also refine your understanding.

Phil


Dondar, I'm not pointing my finger at you. I got this far and just couldn't take it any more.

[OTV Booster]

Here's my question for you: after how many SN explosions will you admit that their methodology needs to change? If your answer is "never", you are no different than Richard Shelby in your disconnection from reality and blind faith, you just have a different allegiance.

For me it will only take one test flight to space and back to change my mind.

OK - I'll put a stake in the ground.  I expect that they'll break the Karmen line with SN8 at the latest - probably SN7.  My pick for orbital is SN9, and I haven't seen any reason to change.
As jrhan48 an others have stated numerous times up thread - for me, they can RUD any number of SN's - as long as they aren't making the same mistake over and over.
don't know who first said it - but - Rockets are Hard!

Is there something to gain from a suborbital burn into the thermosphere that is not gained by a 20km hop aside from bragging rights?

If they boost back in they get heat shield and control knowledge. Flight duration is also a dimension worth exploring.

[Nomadd]

From the armchair space analyst's position the explosions are a godsend, and a far better entertainment value than cable TV.

Well, it would be nice to see all of the parts go in one direction eventually.

[Nevyn72]

From the armchair space analyst's position the explosions are a godsend, and a far better entertainment value than cable TV.

Well, it would be nice to see all of the parts go in one direction eventually.

I'll qualify this comment by adding the the phrase "..... in one piece" 

[OTV Booster]

...Can't we learn from successes or faileres is an imperative to learn?

If you succeed on the first try, then yes, you learned what works without needing to fail. But on a highly complex project with extremely small margins, the chance of succeeding on your first try is vanishingly, exceptionally small. Basically zero.

You can get around this by breaking up your highly complex project into small chunks, spending a lot of time and money planning and designing and testing these chunks separately, but you're still going to fail at some level.

This raises an interesting question .... would it even be desirable to to succeed right away with no failures?  The problem we've got is that getting to orbit is hard.  I know this is getting to be a bit of a cliche at this point, but its true .... if Earth was much more massive than it is, we wouldn't be going anywhere.  Not with chemical rockets anyway.  Earth-based aerospace is, almost by definition, an exercise in playing near the edges.  Destructive testing is how we find out where the edges are.  When Bopper 2 "failed" at 8.5 bar, that was no failure - it successfully ruptured right when they wanted it too.  If it had survived that would have been a failure, because it meant it was stronger, and thus heavier, than it needed to be.

Does that mean SN4 ended in a "destructive test"?  Nope, not on purpose anyway.  But if you never blow something up, you're probably not playing as close to the edge as you really want to be.  The important thing as to have lots of instrumentation and telemetry so that you learn why it blew up, and, as Elon might say, make sure the next one blows up for a different reason.  And above all, fail safely - no death, no injuries, no near misses - both because its the right thing to do, and because there is probably no more certain way to bog down you project than in a government safety review panel.  As for the infrastructure and other equipment?  Its just stuff ... it can be replaced.  Just try not to blow anything up that you can't afford to replace.

This making lemonade to some extent, but one lesson from SN4 is that the header tank is probably overbuilt. It faced up to a goodly detonation and kept its shape. Failure can give lesions where least expected.


Edit to add: lessons too.

[OTV Booster]

...Can't we learn from successes or faileres is an imperative to learn?

If you succeed on the first try, then yes, you learned what works without needing to fail. But on a highly complex project with extremely small margins, the chance of succeeding on your first try is vanishingly, exceptionally small. Basically zero.

You can get around this by breaking up your highly complex project into small chunks, spending a lot of time and money planning and designing and testing these chunks separately, but you're still going to fail at some level.

Sorry but we're still at the basic stages of the developement, so the complexity is at it lowest level, you make me fear the worse for next steps.

When you know the least, which is at the beginning of development, there is the most complexity. Which is why you experiment, to start narrowing down the areas of least knowledge and gain confidence that what you think you know is what is actually true.

This latest test failure showed SpaceX that their GSE setup is not as mature as they thought it was, meaning they uncovered a layer of complexity that needs to be investigated and dealt with in a different manner.

My background is in manufacturing and in new product development, and what SpaceX is doing is not that different from what I have experienced in the non-aerospace world. It is just that the openness they have, and the scale they are working at, is much greater. But the concepts are the same.

So to summarize, at the beginning of development, when you know the least, there is the most complexity.

The concepts of latent complexity and recognized complexity might be applicable here. This seems to be two points of view that are in violent agreement but fumbling over semantics. Don't bother googling it. I just made it up.

[meekGee]

The fault that lost SN4 and the launch mount may well have been caught by an hour HAZOP study.

Ok, let's discuss that. Provide one shred of evidence to support that opinion. Are you suggesting that SpaceX is doing no safety assessment at all? Based on what?

SN1 - QA inspectors ignored
SN3 - Test sequence mis-configuration
SN4 - Minor supplemental test has unforeseen consequences.

I could ask you in turn for a shred of evidence that BC has a functioning engineering assurance process, but that would be a cheap shot - they surely have something in place.

The problem is that all this is happening not in complex vehicle design space where rapid fly/fail/fix cycles can save years of analysis paralysis, but in boringly ordinary chemical engineering space where days of analysis could save months of rebuilding.

If you aren’t failing you aren’t trying hard enough - I get that and fully agree with it, but four losses in four flight vehicles prior to flight is at least suggestive that maybe they are trying slightly too hard.

You're looking at it all wrong.

Since this is not some incremental design iteration, there are bound to be component failures, like that QD a couple of days ago.

When dealing with rockets as large as houses, such failures can cause a bunch of damage.

I challenge you to show me how you can avoid that without taking year upon years to produce a prototype....

The parameter you CAN control, however, is how much you're impacted by the inevitable failure..


Ooooh - too long.  I'll paraphrase Douglas Adams instead:

The difference between a development program that has no failures and a development program that does is that when the development program that has no failures experiences a failure, it takes them a lot longer to recover.

[OTV Booster]

Disagree! No company should be allowed to blow things around every few weeks.  It scares and even panics people around, causes environmental damage, and is a dangerous practice. Somebody should be hold responsible.

What people does it panic?  Please provide evidence that even one person panics when a Starship blows up.

SpaceX purposely does this far away from people.  The few people near enough to even know when one blows up are very well aware of what is going on.  They might be sad when one blows but that's only because they want SpaceX to be successful.

As to environmental damage: no.  No more than burning that same amount of natural gas in any other way.  A small amount of land directly around the vehicle will be burned.  It's no different than many other uses humans make of land.

Nomadd said something about cursing in every language he knows between the RUD and the shockwave. If that doesn't qualify as environmental impact I don't know what does.

[GregTheGrumpy]

Disagree! No company should be allowed to blow things around every few weeks.  It scares and even panics people around, causes environmental damage, and is a dangerous practice. Somebody should be hold responsible.

I read this and thought, hum, forgot the /s or tongue-in-cheek smiley.

You can watch a video posted few pages ago, with a woman narrating it scared quite a bit. That's just oneexample.

Then I read this and I'm flabbergasted.  Seriously?

I'm reminded of one television show that asked at a HS graduation, 'How many moons does the earth have?" - One answer was: "I don't know, I didn't take astronomy."

There will always be Nervous Nellies and Chicken Littles and the grossly incompetent.  We cannot stop all progress because some-one (as in 1) gets the piss scared out of them.  Move on, nothing to see there.

[ZChris13]

I'm reminded of one television show that asked at a HS graduation, 'How many moons does the earth have?" - One answer was: "I don't know, I didn't take astronomy."

That's a perfectly legitimate response, there have been multiple loosely bound NEO that have become temporarily captured by the Earth-Moon system. Hearing about this through pop-sci articles one could easily be confused about the Earth's number of moons according to the astronomy community.

[cdebuhr]

...Can't we learn from successes or faileres is an imperative to learn?

If you succeed on the first try, then yes, you learned what works without needing to fail. But on a highly complex project with extremely small margins, the chance of succeeding on your first try is vanishingly, exceptionally small. Basically zero.

You can get around this by breaking up your highly complex project into small chunks, spending a lot of time and money planning and designing and testing these chunks separately, but you're still going to fail at some level.

This raises an interesting question .... would it even be desirable to to succeed right away with no failures?  The problem we've got is that getting to orbit is hard.  I know this is getting to be a bit of a cliche at this point, but its true .... if Earth was much more massive than it is, we wouldn't be going anywhere.  Not with chemical rockets anyway.  Earth-based aerospace is, almost by definition, an exercise in playing near the edges.  Destructive testing is how we find out where the edges are.  When Bopper 2 "failed" at 8.5 bar, that was no failure - it successfully ruptured right when they wanted it too.  If it had survived that would have been a failure, because it meant it was stronger, and thus heavier, than it needed to be.

Does that mean SN4 ended in a "destructive test"?  Nope, not on purpose anyway.  But if you never blow something up, you're probably not playing as close to the edge as you really want to be.  The important thing as to have lots of instrumentation and telemetry so that you learn why it blew up, and, as Elon might say, make sure the next one blows up for a different reason.  And above all, fail safely - no death, no injuries, no near misses - both because its the right thing to do, and because there is probably no more certain way to bog down you project than in a government safety review panel.  As for the infrastructure and other equipment?  Its just stuff ... it can be replaced.  Just try not to blow anything up that you can't afford to replace.

This making lemonade to some extent, but one lesson from SN4 is that the header tank is probably overbuilt. It faced up to a goodly detonation and kept its shape. Failure can give lesions where least expected.


Edit to add: lessons too.

Perhaps, but whats wrong with that?  Lemonade is delicious.

Seriously though, isn't this sort of lemonade-making really at the heart of this hardware-rich development?  I mean, obviously we want something that works and works well at the end of the process, but while were getting there, we need to find out where the problems lie, and its hard to learn that from a "successful" test.  As an example, consider the Dragon Crew explosion.  Was that a test failure?  Oh yeah, but damned lucky one.  NASA was fine with the design, but I'm glad they never tried to abort with it.  If there was an actual need for crewed launch abort, and instead ... the capsule exploded?

I cant remember who said it (Elon?), but its true (and I might be paraphrasing a bit): One good, well documented failure is worth more than a dozen "successful" tests.

Oh and by the way, yes the CH4 header may well be a bit overbuilt, no tank should be expected to take that sort of abuse! (I've since come around and accept that there was possibly a secondary detonation i nthe LOX tank - I just wish SX would put us out our misery with actual data!)

[spacexfanatic]

...Can't we learn from successes or faileres is an imperative to learn?

If you succeed on the first try, then yes, you learned what works without needing to fail. But on a highly complex project with extremely small margins, the chance of succeeding on your first try is vanishingly, exceptionally small. Basically zero.

You can get around this by breaking up your highly complex project into small chunks, spending a lot of time and money planning and designing and testing these chunks separately, but you're still going to fail at some level.

Sorry but we're still at the basic stages of the developement, so the complexity is at it lowest level, you make me fear the worse for next steps.

Why? What's the worst that can happen in next steps? More test failures?

I don't think you understand how all-up testing works.

I think with mounting level of complexity risk of failure is going to be higher, a lot of parameters are going to be added , they will also be inter-reactions between those parameters, for exemple how multiple engines are going to inter-react?  how landing system is going to work? how separation system is going to react?  how the heat shield will sustain the heat/stress?, how aerodynamic will react?

They will be more electric wires and softwares, more stress, more heat, more gravity...................in the end more complexity which is more failure prone.

[ChrisWilson68]

...Can't we learn from successes or faileres is an imperative to learn?

If you succeed on the first try, then yes, you learned what works without needing to fail. But on a highly complex project with extremely small margins, the chance of succeeding on your first try is vanishingly, exceptionally small. Basically zero.

You can get around this by breaking up your highly complex project into small chunks, spending a lot of time and money planning and designing and testing these chunks separately, but you're still going to fail at some level.

Sorry but we're still at the basic stages of the developement, so the complexity is at it lowest level, you make me fear the worse for next steps.

Why? What's the worst that can happen in next steps? More test failures?

I don't think you understand how all-up testing works.

I think with mounting level of complexity risk of failers is going to be higher, a lot of parameters are going to be added they will also be inter-reactions between those parameters, for exemple how multiple engines are going to inter-react?

After Falcon 1 when SpaceX was working on Falcon 9, some people predicted it would be horribly difficult, or that they'd never make Falcon 9 work reliably at all because of all the extra complexity.

In fact, it turned out to be a non-issue.  No Falcon 9 mission has ever failed because of multiple engines.

Then when they were working on Falcon Heavy again some people said 27 engines was way too many.  They said SpaceX would never make it work.  Again, in reality the 27 engines were simply never a problem.

SpaceX has had all kinds of failures in their development history.  But they've never lost a vehicle to a problem related to multiple engines.

how landing system is going to work? how separation system is going to react?  how the heat shield will sustain the heat/stress?, how aerodynamic will react?

They will be more electric wires and softwares, more stress, more heat, more gravity...................in the end more complexity.

All those things are additional complexity.  But, for the most part, they're incremental complexity that isn't tightly coupled with the core rocket system.  What really matters is coupled complexity, where the details of one part are likely to interact with the details of another part.

These things that you're listing are mostly unlikely to be tightly coupled with the core system, or with each other.  There are pretty clear boundaries between them.

These are also mostly things that SpaceX has done successfully with similar sub-systems on other vehicles.  So they are less likely to have many failures because they break less new ground.  The exception is the heat shield and aerodynamics of entry.

The heat shield and re-entry aerodynamics are definitely a big issue, and there are likely to be more failures as they work through them.  But the issues aren't massively more complex than what they're already doing.

By the way, the word is spelled "failure".  You've been spelling it incorrectly over and over.

[spacexfanatic]

...Can't we learn from successes or faileres is an imperative to learn?

If you succeed on the first try, then yes, you learned what works without needing to fail. But on a highly complex project with extremely small margins, the chance of succeeding on your first try is vanishingly, exceptionally small. Basically zero.

You can get around this by breaking up your highly complex project into small chunks, spending a lot of time and money planning and designing and testing these chunks separately, but you're still going to fail at some level.

Sorry but we're still at the basic stages of the developement, so the complexity is at it lowest level, you make me fear the worse for next steps.

Why? What's the worst that can happen in next steps? More test failures?

I don't think you understand how all-up testing works.

I think with mounting level of complexity risk of failers is going to be higher, a lot of parameters are going to be added they will also be inter-reactions between those parameters, for exemple how multiple engines are going to inter-react?

After Falcon 1 when SpaceX was working on Falcon 9, some people predicted it would be horribly difficult, or that they'd never make Falcon 9 work reliably at all because of all the extra complexity.

In fact, it turned out to be a non-issue.  No Falcon 9 mission has ever failed because of multiple engines.

Then when they were working on Falcon Heavy again some people said 27 engines was way too many.  They said SpaceX would never make it work.  Again, in reality the 27 engines were simply never a problem.

SpaceX has had all kinds of failures in their development history.  But they've never lost a vehicle to a problem related to multiple engines.

how landing system is going to work? how separation system is going to react?  how the heat shield will sustain the heat/stress?, how aerodynamic will react?

They will be more electric wires and softwares, more stress, more heat, more gravity...................in the end more complexity.

All those things are additional complexity.  But, for the most part, they're incremental complexity that isn't tightly coupled with the core rocket system.  What really matters is coupled complexity, where the details of one part are likely to interact with the details of another part.

These things that you're listing are mostly unlikely to be tightly coupled with the core system, or with each other.  There are pretty clear boundaries between them.

These are also mostly things that SpaceX has done successfully with similar sub-systems on other vehicles.  So they are less likely to have many failures because they break less new ground.  The exception is the heat shield and aerodynamics of entry.

The heat shield and re-entry aerodynamics are definitely a big issue, and there are likely to be more failures as they work through them.  But the issues aren't massively more complex than what they're already doing.

By the way, the word is spelled "failure".  You've been spelling it incorrectly over and over.

You're reminding here the succeful SX approach for F1 and F9 developement programe which is indeniablly true, but the issue of the thread is the diametrically opposit approach we are experiencing here with SS/SH developement.

-For the multiple engines interreaction, the issue is with central core stage which still never made it to land safely, it's useless to remind that heavy is 3x9 engines while SH is 31 or so engines which is more complexity.

-You did also mention the lack of tightly coupled interreactions between the system, but just imagine the results of a failing heatshield, or failing landing leg, or aerodynamics, the result is assuredly the destruction of the vehicle.

[Kabloona]

From the armchair space analyst's position the explosions are a godsend, and a far better entertainment value than cable TV.

Well, it would be nice to see all of the parts go in one direction eventually.

Hah. This reminds me of some wag's description of the DC-3 as a collection of parts flying in close formation.

https://www.dailymail.co.uk/news/article-518349/How-health-safety-rules-grounded-Dakota-war-workhorse.html

[steveleach]

They will be more electric wires and softwares, more stress, more heat, more gravity...................in the end more complexity which is more failure prone.

Yes, more complex is more failure-prone.

But more complex is also harder to do "traditional" up-front design for, so that doesn't favour one approach or the other.

Falcon, Starship, SLS - all of these are already well beyond the point where a single person can understand all the complexity. The next step up from there is the point at which a single small team can understand all the complexity and manage it with informal communication, and again, we're well passed that. I guess after that it would be the ability for a single organisation to cope, using formal communication methods, but given what we know about projects like Apollo, we're a way from that.

I don't see any particular threshold being crossed between F9 and SS.

[sebk]

You're reminding here the succeful SX approach for F1 and F9 developement programe which is indeniablly true, but the issue of the thread is the diametrically opposit approach we are experiencing here with SS/SH developement.

Nope. The approach is not diametrically oposite. It's pretty similar to their successful work on F1 and F9. The difference is we have visibility to things because they happen near a public road & beach, not in the middle of vast fenced area.

-For the multiple engines interreaction, the issue is with central core stage which still never made it to land safely, it's useless to remind that heavy is 3x9 engines while SH is 31 or so engines which is more complexity.

Central stage landed correctly once. It was lost to the sea afterwards because of incomplete ground (or rather sea) handling infrastructure.

Moreover the issues were not due to 27 engines but due to hotter reentry (central core stages later than F9 core, so it comes in very hot later).

-You did also mention the lack of tightly coupled interreactions between the system, but just imagine the results of a failing heatshield, or failing landing leg, or aerodynamics, the result is assuredly the destruction of the vehicle.

Yes, and?

If heatshield fails it's a heatshield failure. It's not coupled with other failures. The ship is destroyed, but it doesn't mean some complexity coupling.

They will be more electric wires and softwares, more stress, more heat, more gravity...................in the end more complexity which is more failure prone.

Yes, more complex is more failure-prone.

But more complex is also harder to do "traditional" up-front design for, so that doesn't favour one approach or the other.

Falcon, Starship, SLS - all of these are already well beyond the point where a single person can understand all the complexity. The next step up from there is the point at which a single small team can understand all the complexity and manage it with informal communication, and again, we're well passed that. I guess after that it would be the ability for a single organisation to cope, using formal communication methods, but given what we know about projects like Apollo, we're a way from that.

I don't see any particular threshold being crossed between F9 and SS.

More complex doesn't equal more failure prone. You can actually "use" complexity to improve the odds of success -- it's called redundancy.

Things should be as simple as possible, but not simpler.