SpaceX FH : Falcon Heavy Demo : Feb 6, 2018 : Discussion Thread 2

[wannamoonbase]

Assuming the launch goes perfectly, and the 3 boosters are recovered...
Wouldn't it make any sense to refurb the 3 boosters and do a fully expendable launch ?
Be it a fully commercial launch or the first large StarLink satellite load ?
Or is it actually cheaper to retire them ?
I think recovery makes most sense if you fly them again...

It is my understanding the side boosters are pre-flown block 3, a reusable booster, but not as reusable as block 4. They will be saved for historical purposes. No block 3 rockets are being saved at this point, only block 4 and later this year only block 5. The main center stage 1 booster is a new block 4, but it too will likely be on display somewhere, as a set, assuming all goes well.

The thought of all 3 boosters being reassembled for a display really makes me smile.

The Block 3 side boosters have been rebuilt in a way that no other boosters have been.  They are as likely as new as a booster can be.

I don't know anything, and its been suggested these won't fly again.  However, perhaps reuse will depend on loads experienced during flight and inspections after recovery.

If SpaceX has 3 healthy cores, why wouldn't they use them again and recover some of the huge amount of time and money invested in this beast of a rocket.

EDit:  Wow, we're 2 days away, I can't believe we are finally only 2 days away.

[Herb Schaltegger]

When I got my first bike, I fell 10 times. I then got the hang of it and rode 20 times without falling. What are the odds I'll fall on my next attempt?

100% if your older brother decides to jam a stick in the spokes as you take off. Painful first hand experience.

LOL, yeah, we may have had the same older brother.

I think if they successfully separate the 2 side boosters then I think the mission, including recovery, will be fully successful.

Ditto. Staging those boosters will be interesting given SpaceX's novel aversion to pyros and separation motors.

 I also think there's a slightly-higher risk to the core stage landing than with a "normal" F9 barge landing, simply due to the higher velocity at S2 staging compared to a standard F9, but as others have said, they'll have a better change to mitigate that on this flight due to the light payload and prop margins. A max-payload GTO missions might pose a higher risk.

I'm also just a bit concerned about thermal control issues on the central core due to the radiant heating of having 9 engines burning in close proximity to either side but I trust SpaceX will have done their due diligence modeling this extensively over the last couple years.

[Barrie]

If SpaceX has 3 healthy cores, why wouldn't they use them again and recover some of the huge amount of time and money invested in this beast of a rocket.

The centre core is new, and beefed-up compared to a standard F9, so there must be some temptation to use it again.  The Block 3 side-boosters, OTOH, will have had their single refurb and reuse...

[Lar]

If SpaceX has 3 healthy cores, why wouldn't they use them again and recover some of the huge amount of time and money invested in this beast of a rocket.

Because Gwynne said so.

EDit:  Wow, we're 2 days away, I can't believe we are finally only 2 days away.

IKR... Tuesday is going to be a very weird day at work.

[FutureSpaceTourist]

Excellent, John Kraus is accredited for the FH launch. Can’t wait to see what he shoots:

Official confirmation! I will be on-site under @observer at Kennedy Space Center for this week’s Falcon Heavy launch.

https://twitter.com/johnkrausphotos/status/960186396702068736

[wannamoonbase]

If SpaceX has 3 healthy cores, why wouldn't they use them again and recover some of the huge amount of time and money invested in this beast of a rocket.

Because Gwynne said so.

EDit:  Wow, we're 2 days away, I can't believe we are finally only 2 days away.

IKR... Tuesday is going to be a very weird day at work.

I know she said so, but if you parse her words there is window there.  One could read her description that this set of FH cores is the only version of this design to be built.

Also, the long promised Block 5 is taking longer than planned.  Reusing this hardware would give them time to crank out some F9 Block 5's before taking the time to build 3 cores for the next FH set.

I see a business case for reusing this hardware if it's capable of flying again.

That's all I'm thinking. Not trying to get into argument.

[marsbase]

If the reliability is 91%, then the odds of 21 straight successes are only 15%...

A better explanation is that whatever was wrong with those first two launches got fixed.

When I got my first bike, I fell 10 times. I then got the hang of it and rode 20 times without falling. What are the odds I'll fall on my next attempt?

No, the 91% represents the probability of success  of each future launch (generally referred to as reliability) based on the average of 23 past launches.  That does of course assume that all trials are interchangable.  As your bike analogy shows, that is not the case. 

[meekGee]

If the reliability is 91%, then the odds of 21 straight successes are only 15%...

A better explanation is that whatever was wrong with those first two launches got fixed.

When I got my first bike, I fell 10 times. I then got the hang of it and rode 20 times without falling. What are the odds I'll fall on my next attempt?

No, the 91% represents the probability of success  of each future launch (generally referred to as reliability) based on the average of 23 past launches.  That does of course assume that all trials are interchangable.  As your bike analogy shows, that is not the case.

Main point is that you can only multiply probabilities of statistically independent events.

A set of launches of a rocket under development, where a failure is followed by study and corrective action, are highly dependent and correlated, and so that 21/23 thinking is just fundamentally wrong.

[Pete]

The centre core is new, and beefed-up compared to a standard F9, so there must be some temptation to use it again.  The Block 3 side-boosters, OTOH, will have had their single refurb and reuse...

To me, this is an argument for recovery, but against reuse.
They will want to do a *thorough* post-mission stress analysis on that core booster, the sort of analysis that includes chopping it to bits and probing at the innards with destructive methods.

They need to know not only that it worked, but where they under- or over-estimated the material needs, so that a future design can be easier/cheaper/safer.

[Lar]

I know she said so, but if you parse her words there is window there.  One could read her description that this set of FH cores is the only version of this design to be built.

Also, the long promised Block 5 is taking longer than planned.  Reusing this hardware would give them time to crank out some F9 Block 5's before taking the time to build 3 cores for the next FH set.

I see a business case for reusing this hardware if it's capable of flying again.

Valid point but between wanting to move on, and wanting to examine the stages (probably destructively) I think the odds are against it. Non zero but less than 50/50...

[vanoord]

If the reliability is 91%, then the odds of 21 straight successes are only 15%...

A better explanation is that whatever was wrong with those first two launches got fixed.

When I got my first bike, I fell 10 times. I then got the hang of it and rode 20 times without falling. What are the odds I'll fall on my next attempt?

No, the 91% represents the probability of success  of each future launch (generally referred to as reliability) based on the average of 23 past launches.  That does of course assume that all trials are interchangable.  As your bike analogy shows, that is not the case.

Main point is that you can only multiply probabilities of statistically independent events.

A set of launches of a rocket under development, where a failure is followed by study and corrective action, are highly dependent and correlated, and so that 21/23 thinking is just fundamentally wrong.

Yes/No

Note my comment immediately after mentioning 21/23 = 91%

In reality, the probability of landing any given core now should be higher due to experience...

I then rounded up from 91% reliability to 95%.

That's a shift from a 1 in 11 failure rate to 1 in 20 - almost a doubling in reliability.

That 1 in 20 - I reckon - isn't far off where the landing reliability is at the moment (and bear in mind that Bulgariasat narrowly avoided being a RUD).

From that, the probability of getting all three cores back is just under 86% (I wrongly put 83% in my first post).

Given the experimental nature of the vehicle, I'm happy to stick with somewhere around 75% likelihood that all three cores will come back.

That is, of course, pretty much as scientific as holding a finger up in the wind and guessing.

[meekGee]

If the reliability is 91%, then the odds of 21 straight successes are only 15%...

A better explanation is that whatever was wrong with those first two launches got fixed.

When I got my first bike, I fell 10 times. I then got the hang of it and rode 20 times without falling. What are the odds I'll fall on my next attempt?

No, the 91% represents the probability of success  of each future launch (generally referred to as reliability) based on the average of 23 past launches.  That does of course assume that all trials are interchangable.  As your bike analogy shows, that is not the case.

Main point is that you can only multiply probabilities of statistically independent events.

A set of launches of a rocket under development, where a failure is followed by study and corrective action, are highly dependent and correlated, and so that 21/23 thinking is just fundamentally wrong.

Yes/No

Note my comment immediately after mentioning 21/23 = 91%

In reality, the probability of landing any given core now should be higher due to experience...

I then rounded up from 91% reliability to 95%.

That's a shift from a 1 in 11 failure rate to 1 in 20 - almost a doubling in reliability.

That 1 in 20 - I reckon - isn't far off where the landing reliability is at the moment (and bear in mind that Bulgariasat narrowly avoided being a RUD).

From that, the probability of getting all three cores back is just under 86% (I wrongly put 83% in my first post).

Given the experimental nature of the vehicle, I'm happy to stick with somewhere around 75% likelihood that all three cores will come back.

That is, of course, pretty much as scientific as holding a finger up in the wind and guessing.

Yup, but still..

Consider rocket "A" that failed 6 times in a row, and then flew 40 times straight.

Rocket B failed twice, and then flew 30 times straight.

Rocket C worked 15 times, failed once, worked 12 times, failed once, worked 3 more times.

Which rocket would you say has the best chance of working on the next launch?

[Semmel]

Consider rocket "A" that failed 6 times in a row, and then flew 40 times straight.

Rocket B failed twice, and then flew 30 times straight.

Rocket C worked 15 times, failed once, worked 12 times, failed once, worked 3 more times.

Which rocket would you say has the best chance of working on the next launch?

A guessing game is not statistics. Thats the crux with statistics, its hardly ever intuitive. If you want to do it mathematically correct, you cant take any of the standard ways. And if you dont want to do it correctly, you dont need statistics.

In which case I would choose Option B.

[meekGee]

Consider rocket "A" that failed 6 times in a row, and then flew 40 times straight.

Rocket B failed twice, and then flew 30 times straight.

Rocket C worked 15 times, failed once, worked 12 times, failed once, worked 3 more times.

Which rocket would you say has the best chance of working on the next launch?

A guessing game is not statistics. Thats the crux with statistics, its hardly ever intuitive. If you want to do it mathematically correct, you cant take any of the standard ways. And if you dont want to do it correctly, you dont need statistics.

In which case I would choose Option B.

And that's the point - you'd be wrong because you're treating all flights as independent events, where in fact they are not.

You're not "mathematically correct even though it is counter intuitive".  You're simply applying math to an incorrect model of reality, and then GIGO applies.

The problem with statistics is very often in the transition between the real world and the mathematical model. 

Same as it is, actually, in engineering in general...

[meekGee]

Here's a simpler example.

Rocket A flew 100 times and failed 10 times, spread roughly evenly throughout that period.

Rocket B had 10 straight failures followed by 90 successes.

According to your model, they are both 90/100...

But one rocket clearly has an ongoing problem that they can't find and fix, and the other clearly had a teething problem that's been fixed.

It's not that the math is wrong... It's that the statistical model was over-simplified.

Your model should be as simple as possible, but no simpler.

[JamesH65]

I look at it like this.

Lets say the first 5 flights failed, on flight one they had a problem that they fixed. In flight two they had a problem that they fixed, and so on. After every failure they found the problem and fixed it.

So actually the chance of a failure is unrelated to the first five flights, because all the causes of those failure have been fixed. So the failure MUST come from another unforseen source.

And what are the chances of another failure? Hmm, well, is it the same as the chance in the five flights? Because almost be definiton they were unforeseen failure modes. BUT, people learn from their mistakes, and after each failure, not only are the failures specifically fixed, but proceses to prevent unforeseen failures are also improved.

This is the point where it should now become apparent that using a straight average is completely useless.

[AncientU]

There is a subtle distinction between launch/rocket failures and the 'failed' landing attempts...  the former are generally planned to be successful and sold as such.  The landing attempts that 'failed' were all tests as SpaceX empirically determined what combination of hardware and software could be used to get the booster back in tact.  They were all experimental landings... the booster was going to be lost anyway, so why not run a test to collect data on this landing approach.  To include them statistically as 'failures' -- equivalent to the CRS (true) failure -- is missing the point, even if statistics are correctly applied.

[lrk]

If SpaceX has 3 healthy cores, why wouldn't they use them again and recover some of the huge amount of time and money invested in this beast of a rocket.

Because Gwynne said so.

EDit:  Wow, we're 2 days away, I can't believe we are finally only 2 days away.

IKR... Tuesday is going to be a very weird day at work.

I know she said so, but if you parse her words there is window there.  One could read her description that this set of FH cores is the only version of this design to be built.

Also, the long promised Block 5 is taking longer than planned.  Reusing this hardware would give them time to crank out some F9 Block 5's before taking the time to build 3 cores for the next FH set.

I see a business case for reusing this hardware if it's capable of flying again.

That's all I'm thinking. Not trying to get into argument.

The only way that I can see any of the cores flying again is if there is a failure severe enough to require them to redo the test flight, but somehow some of the cores land undamaged.  But I don't see many at all likely ways for this to happen - e.g. a major underperformance or something goes wrong after booster sep in a way specific to Falcon Heavy. 

[vanoord]

There is a subtle distinction between launch/rocket failures and the 'failed' landing attempts...  the former are generally planned to be successful and sold as such.  The landing attempts that 'failed' were all tests as SpaceX empirically determined what combination of hardware and software could be used to get the booster back in tact.  They were all experimental landings... the booster was going to be lost anyway, so why not run a test to collect data on this landing approach.  To include them statistically as 'failures' -- equivalent to the CRS (true) failure -- is missing the point, even if statistics are correctly applied.

FWIW, the overall statistic is 21 successes from 26 attempts - that includes F9 v1.1 but doesn't include GovSat1 or any of the other 'water landings'.

21/23 is for F9 v1.2 landing attempts.

Whether or not landings were considered to be experimental or not doesn't affect the figures - but if you put the cut-off in different places, the stats are different.

For example, 5 of 5 landing attempts for Block 4 F9s have been successful: does that mean that the next Block 4 to attempt a landing has a 100% chance of success?

Similarly, if you use just the statistics, you'd discover that an F9 core landing on OCISLY would have a greater chance of a RUD than if it were landing on  JRTI - or indeed that an F9 flying from Pad 39A has a 100% chance of landing.

The problem is several-fold, including:
- the dataset is too small (come back after 1,000 landing attempts if you want accurate figures)
- the vehicle is being developed, so it's not reasonable to consider it a single, simple dataset
- there may be an effect caused by whether cores are new or re-used, but the dataset will probably be too small for that to become apparent
- there are other variables that are going to come into play, eg weather conditions, human errors
- the dataset is too small

Basically, calculating probabilities is meaningless - but using past performance as a guide and modifying the numbers gives a decent enough guide to back up what is ultimately guesswork.

[Lar]

(mod) To a certain extent, detailed analysis of rocket reliability is off topic. I think we have a thread for it. Maybe someone could crosslink to it after you all come to rest on what the success probability is if you want to delve in deeper.

(fan)My estimate of the landing success probabilities....  is "indeterminate, not enough data and models to predict it, as it is unclear what things are the same and are significant vs not the same and are significant"