Transterrestrial Musings

Comments

Reliability may be a key issue for commercialization, but I hardly think the failure rates for current rockets are too extreme for public sector exploration. The entire point is for government spending to absorb the initial costs, take the initial risks, and accept the inevitable casualties of entering uncharted territory, because losing a spacecraft or five isn't going to put them in bankruptcy. Now, that isn't to say we shouldn't always strive for improvement, but insisting on having paved roads before you even build the town is kind of a strange way to approach a frontier. We didn't wait for the transcontinental railroad to send Lewis and Clark, and we don't need airline-level reliability in orbital launchers before going back to the Moon or setting foot on Mars. A LOT of people are going to have to die in space, on the Moon, and on Mars before any of it comes close to being safe, and I personally would consider it a real privilege to be able to take that risk. People die and lose fortunes every day for incredibly stupid reasons, so let's start doing it for good ones.

Posted by Brian Swiderski at July 27, 2006 03:54 PM

...we don't need airline-level reliability in orbital launchers before going back to the Moon or setting foot on Mars.

No one claimed that we do. There's a huge gap between airline-level reliability and current rockets, particularly when one considers how ridiculously expensive they are. This isn't a problem of technology, it's a problem of operational scale. If we were doing a lot more, unit costs would be much lower, and vehicles would be much more reliable. But NASA doesn't want to do much.

Posted by Rand Simberg at July 27, 2006 04:10 PM

I don't buy into blaming NASA. If they were given the money, and knew failures wouldn't lead to the budgetary guillotine, they would do everything we suggest and more. They've been hounded into abject humility by decades of cuts, abandonment from political leaders, and promising initiatives sent to the slaughterhouse in the middle of being realized. Now they have to deal with trying to get back to the Moon on the pittance Bush is giving them, and seem confident about it despite the widespread fear it will be cancelled the moment it goes a penny over budget a la X-33. NASA is like a once-beautiful woman who marries a wife-beater and turns into a cowering wraith; tragic, but not above salvation if the American people come around.

Posted by Brian Swiderski at July 27, 2006 04:54 PM

One hopes the reason for the failure will be found, and the appropriate fix made, before Bigelow commits more cargo to that design.

But it also speaks to what I said in the North Korea context about how ICBMs themselves are no more 'perfect' than defenses against them can be. Either one just has to be good enough to deter the opposition...

Posted by Frank Glover at July 27, 2006 06:13 PM

So Uncle Sam is a wife beater. No surprise there. What you may interpret as harsh criticism of NASA, is really just harsh criticism of any large government undertaking. Hence the focus on commercial space initiatives.

Posted by Orville at July 27, 2006 07:32 PM

Sorry if this off topic a bit, but I have a question about the CEV (OK, perhaps unreliable crew vehicles are on topic). I keep hearing that it is meant to get us to LEO/ISS, to Moon and to Mars. The first two I understand, but is anyone actually proposing the CEV will be sent to Mars? Perhaps to get the astronauts to the ship that will take them to Mars, but the CEV isn't leaving Earth orbit, is it? If not, it's being a bit oversold by NASA isn't it?

Posted by Mike Thompson at July 27, 2006 08:49 PM

...despite the widespread fear it will be cancelled the moment it goes a penny over budget a la X-33.

This isn't quite what I recall why the X-33 was canceled. I thought that cancelled because of many technical difficulties (fuel tank failures) that were leading to non-trival budget overruns. However, I do understand the gist of the point you were trying to make regarding NASA overall.

Posted by MIke Thompson at July 27, 2006 08:54 PM

...s anyone actually proposing the CEV will be sent to Mars?

Probably, because that's the party line. It may in fact be sent to Mars (as part of a much larger crewed vehicle), but only because that might be perceived as necessary to fulfill the current con job. But I suspect that if NASA ever does send a vehicle to Mars, the current CEV will play no part in that mission.

Of course, if it's based someplace like L1, it could be used to bring people home to earth, so it could be part of a Mars mission, but I consider this scenario unlikely as well.

Posted by Rand Simberg at July 27, 2006 09:18 PM

nothing wrong with unreliable cheap transport, so long as you dont put unique, irreplaceable and too valuable stuff on it ( which obviously was the case here )
For example, any roundtrip mission to lunar surface from LEO is 5/6ths propellant. Who cares how unreliably all this propellant was delivered to LEO, as long as its there cheaply ?

Posted by kert at July 28, 2006 12:16 AM

nothing wrong with unreliable cheap transport, so long as you dont put unique, irreplaceable and too valuable stuff on it ( which obviously was the case here )

No offense to the entities involved, but I don't think this cargo was particularly valuable.

Posted by Karl Hallowell at July 28, 2006 11:37 AM

If we were doing a lot more....vehicles would be much more reliable.

Rand, I don't understand why you say this. All mechanical devices I know of have a reasonably fixed failure rate -- so a rocket should have a fixed probability of failure per launch. The more you launch, the more failures.

The failure rate goes down with the care and quality of engineering and construction -- i.e. with an increase in the amount of money you spend per launch. Expensive rockets should have a lower failure rate than cheap. But I don't think you're saying we should be spending more per launch, that present launch systems try to get away with doing things too cheaply.

Failure rate also goes down with experience, as people get to know how to manufacture and use the machine optimally and avoid nonobvious pitfalls (e.g. launching in cold weather with O-rings that apparently get rather brittle when cold). Maybe this is what you mean? That if we launched much more often, we'd reach that happy land of experienced operations faster?

But...in that case, why criticize NASA for being very conservative about trying new technology? If the main goal is reducing failure rate through experience, maybe the right thing to do is to "evolve" something like the Shuttle -- which has 20 years and 120 launches to its credit -- rather than try something completely new. Completely new technology often translates into completely new problems. That's why it's the "bleeding edge," so to speak.

I'm not trying to be difficult, just trying to sort out what it is you're saying. I mean, I can see saying newfangled half-crazy ideas should be tried because, while most will prove expensive folly, some might prove to be breakthrough technology that will bring the price way down through tricks no one thought of before -- or I can see saying the main goal should be to just increase the launch rate to get experience (with whatever technology is in hand) and reduce costs per launch. But I don't see how you can argue both.

Posted by Carl Pham at July 28, 2006 05:03 PM

Rand Says

Until we fix the problem of reliability and affordability (issues that NASA's plans don't even attempt to address), it's pointless to plan lunar or Mars missions.

***

While existing vehicles are far from perfect the track record of many of them are pretty darn good and I would not expect improving the reliablity say by 50% is going to make any statistical difference in the ability to do lunar or even Mars missions.

Affordability is directly proportional to flight rate but by requiring greatly increased reliability before extensive exploration you preclude the very thing that gives you that reliability.

Dennis

Posted by Dennis Ray Wingo at July 28, 2006 06:05 PM

All mechanical devices I know of have a reasonably fixed failure rate -- so a rocket should have a fixed probability of failure per launch. The more you launch, the more failures.

No, items mass produced with statistical quality control are much more reliable than a few hand-built items (space vehicle manufacturing is currently more like a craft, or boutique). And ideally, what would be mass produced would be flights, rather than vehicles, so the infant mortality issue of flying a new vehicle every time would go away as well.

I wrote a column on this at Fox News a couple years ago (almost three, actually).

Posted by Rand Simberg at July 28, 2006 06:06 PM

I would not expect improving the reliablity say by 50% is going to make any statistical difference in the ability to do lunar or even Mars missions.

Well, if I thought we could do no better than that, then I'd agree. I think we can add nines.

Posted by Rand Simberg at July 28, 2006 07:08 PM

Rand

A couple of nines? You mean that you think, that with current art, that we can improve the reliability by two orders of magnitude?

It would take 5000 launches just to build the statistics to figure out if your design met that criteria.

Methinks you have not thought this one through.

Dennis

Posted by Dennis Ray Wingo at July 28, 2006 07:33 PM

You mean that you think, that with current art, that we can improve the reliability by two orders of magnitude?

Yes.

It would take 5000 launches just to build the statistics to figure out if your design met that criteria.

I said that we could achieve it, not demonstrate it empirically. But if you can get marginal costs that are a small multiple of the propellant costs, you can get launches that cost on the order of a million dollars per launch, so you're only talking about five billion dollars. Compare that to what NASA is proposing to spend on CEV alone.

Posted by Rand Simberg at July 28, 2006 07:42 PM

**********

Rand

How can you achieve something without demonstrating it? If I just take the number of critical path parts, apply the error function to them, you will not get the type of reliability that you seek. The reason that jet travel is as safe as it is today is that we have the huge statistical database of what works and what does not.

In the Apollo program we had over 30 launches of the system with no failures and according to an Aviation Week article that I have from 1966 the majority of the expenditures in the development of the Saturn V was dedicated to testing the parts. I see no similar approach by any one out there, including Elon. The old timers here in Huntsville talk about huge sums of money that was spent just to get reliable propulsion valves.

The only way that we are ever going to get to that level of reliability in launch is through climbing the experience curve. No one is a smart enough engineer or engineering team to design something like this without climbing an operational curve of far more launches than what we have today. The best that is out there today is on the Atlas V EELV team and they are not even to on nine much less three.

Dennis

Posted by Dennis Ray Wingo at July 28, 2006 09:36 PM

How can you achieve something without demonstrating it?

There is no correlation between actual reliability and demonstrated reliability, for numbers smaller than those necessary to demonstrate it. Reliability is what it is, regardless of the empirical data.

If I just take the number of critical path parts, apply the error function to them, you will not get the type of reliability that you seek.

A well-designed space transport should have no more critical-path parts than a modern airliner.

The reason that jet travel is as safe as it is today is that we have the huge statistical database of what works and what does not.

That is only one reason. But we'll develop such a database for space transports as well.

No one is a smart enough engineer or engineering team to design something like this without climbing an operational curve of far more launches than what we have today.

I agree. We will.

But not with NASA's approach.

The best that is out there today is on the Atlas V EELV team and they are not even to one nine much less three.

They have trivial test flight rates, because their launches are far too costly. There's no reason to think that we will ever get highly reliable ELVs. There's abundant reason to think that we'll get highly reliable space transports.

Posted by Rand Simberg at July 28, 2006 09:48 PM

There's abundant reason to think that we'll get highly reliable space transports.

There's also abudant reason to think we probably will not.

Posted by Dave at July 29, 2006 08:44 AM

"abundant" - sorry pressed post too soon and I know how offended you are by spelling.

Posted by Dave at July 29, 2006 08:45 AM

I know how offended you are by spelling.

As is often the case, you "know" something that isn't true.

Posted by Rand Simberg at July 29, 2006 09:00 AM

Rand Says

***********

This is the same logic that NASA used for their completely insane reliability numbers on the Shuttle before Challanger and Columbia.

____________

Rand Says

That is only one reason. But we'll develop such a database for space transports as well.

****************

Without an experience base reliability numbers are fantasies. This is true, no matter what you are building, whether it is a computer or a launch vehicle. In the computer industry we almost can ignore the reliability of individual parts because there has been a 30 year program of reliability improvements that have increased components reliability from 1 in a thousand to one in a hundred million.

Your argument would imply that in 1975 we could have designed a 7400 series IC with a reliability in the 1 in 100 million range which is patently absurd. NASA and the USAF during and after Apollo tried this and this is where class B and Class S components came from. Jan King proved in 1990 with the launch of the AMSAT Microsats that the reliability of commercial components, based upon statistical reliability improvement methods derived from examining billions of components and improving them by destructive testing over time, resulted in a standard of reliability that exceeded class S by 1990 and the gap has increased by orders of magnitude since then.

You simply cannot design a system and claim x reliability without a huge experience base to rely on. This is at the root of the problem with the first launch of the Falcon 1 and with Deke Slayton's EER launch vehicle as well as with the first Delta IV launch. You simply cannot anticipate in the design phase every possible failure mode without that experience base. This is also at the core of the huge problems ahead with the CEV and CLV. The people doing the work don't have the experience to understand where the problems are and how to deal with them.

Dennis

Posted by Dennis Ray Wingo at July 29, 2006 11:38 AM

Rand says:
>

***********

This is the same logic that NASA u>sed for their completely insane reliability numbers on the Shuttle before Challanger and Columbia.

No. The statement that I made is true. The fact that NASA chose to use it as an excuse to delude themselves about the reliability of the Shuttle doesn't make it not true.

Without an experience base reliability numbers are fantasies.

That doesn't mean that we can't build much more reliable vehicles than current ones. It merely means that we can't verify and quantify their reliability until we fly them a lot.

Your argument would imply that in 1975 we could have designed a 7400 series IC with a reliability in the 1 in 100 million range which is patently absurd.

As has been pointed out in another post, this is a foolish analogy. That really was a technology problem.

There is no reason that a well-designed space transport can't be built to be much more reliable (yes, by at least two orders of magnitude) than any current ELV, nor is there any reason that we are incapable of designing such a vehicle with today's technology base. The biggest reason that we haven't successfully done so is that we've never made a serious attempt. But the key is incremental testing, and full reusability (something that none of your examples provide).

Posted by Rand Simberg at July 29, 2006 12:13 PM

******************

Rand

I am sorrry but if you don't see the point then I submit that you are part of the problem. Engineering is a learn as you go art and in another post where you are skeptical of Elon, you miss the point there as well. Elon is learning, the prerequesite for moving forward and that of building a reliable system.

The 7400 series reliability was not a technology problem but a process problem. Do you really know how reliability is improved in the semiconductor world?

The Saturn V was what it was because there was several hundred man-decades of engineering expertise behind it and an incredibily expensive program of component-to-systems testing to instill a measure of reliability in a system that was very very new. It is funny that a great deal of time is being spent by the engineering teams at Thiokol and other contractors to go back to the Saturn data and the hardware itself to figure out what to do. If the CLV gets built at all the upper stage is going to look almost exactly like an SIVB.

Even with all that was accomplished in the Saturn 1 and Saturn V design we dang near lost the first one, survived a lightning strike on another one, and almost lost skylab because the pogo problem had still not been fixed. The Delta IV first launch went down because pogo was ignored and even afterwards Boeing did everything in their power to call it "c$avitation" rather than pogo.

There is NO substitute for experience and the most reliable design in the universe is going to have operational failures until such time as as all of the unknowns at the time of the design are fixed.

I am really astonished that you don't understand this.

Dennis

Posted by Dennis Ray Wingo at July 29, 2006 02:17 PM

There is NO substitute for experience and the most reliable design in the universe is going to have operational failures until such time as as all of the unknowns at the time of the design are fixed.

I completely agree.

I am really astonished that you don't understand this.

And I'm even more astonished that you seem to have no comprehension of my point, as exemplified by your reference to Saturn.

Posted by Rand Simberg at July 29, 2006 06:13 PM

Rand

Then, clearly and concisely explain what your position is and how you can build a reliable system that is two orders of magnitude better than current art without the learning curve applied to every single large scale system designed and built in the past.

Dennis

Posted by Dennis Ray Wingo at July 29, 2006 06:37 PM

It is very simple, Dennis. We will build vehicles that are not thrown away with every use, and we will fly them repeatedly, wringing the bugs out of them, and making design changes as necessary.

What do you think that the reliability of airplanes would be today if we always threw them away after each flight?

Posted by Rand Simberg at July 30, 2006 08:53 AM

Rand

Fine, and I agree, but what you said is that we can design such systems today that are orders of magnitude more reliable than current art which is poppycock.

Dennis

Posted by Dennis Ray Wingo at July 30, 2006 10:09 AM

...what you said is that we can design such systems today that are orders of magnitude more reliable than current art which is poppycock.

If you mean that literally, I guess it is, since Rome wasn't built in a day. I don't think I said that. My point is that there is nothing preventing us from embarking on such designs today. It will obviously take months and years to implement and debug them. I can't quite figure out just what it is you think that we can't do.

Posted by Rand Simberg at July 30, 2006 10:18 AM

Rand

Again, I agree that there is nothing preventing us from embarking on this quest (except a deep pocketed visionary investor). Even if that investor showed up today, It would take a couple of decades to get to the holy land of increased reliability. It takes that long just to build the statistics.

I do see rays of hope in Elon's first steps towards at least partial reusability and possibly the resurrection of Kistler. It will be VERY interesting to see who wins the COTS awards.

Dennis

Posted by Dennis Ray Wingo at July 30, 2006 11:17 AM

It would take a couple of decades to get to the holy land of increased reliability.

Nonsense. The very first generation of reusable vehicles will almost certainly be more reliable than anything currently flying. It makes no sense to talk about an amount of time to get to specified reliability unless you specify the reliability.

It takes that long just to build the statistics.

We don't need to "build statistics." We need to build increasingly reliable vehicles. The statistics will follow. You turn the order of things on their head. Statistics only tell you that you've done the job--they don't create reliability.

Posted by Rand Simberg at July 30, 2006 11:28 AM

Rand

************

There is absoulutely no basis for the first sentence. The second sentence only makes sense in the context of operational experience. NASA designed the orbiter to be 99.9 percent reliable. I have memo's from Deke Slayton telling the design and operations teams to ignore any failure modes with more than a 10e-3 probability of happening.

What has improved since 1970 that gives confidence that we can design a system with 99.99 or 99.999 percent reliability? Reliability numbers are a fantasy beyond a certain point until you have the flight numbers to verify them. I remember the vast amount of effort that NASA and the contractor teams put out to get to a failure rate on the Shuttle system of 1 in 256. This was a year before Columbia. They were off by a factor of five. These guys are not incompetent, nor were they actively trying to cut corners to kill crew. They were using their best knowledge and a bit of fudging to get those numbers, the same thing that any team does.

Dennis

Posted by Dennis Ray Wingo at July 30, 2006 12:33 PM

NASA designed the orbiter to be 99.9 percent reliable.

NASA also designed the Shuttle to be so costly to operate and with such a low intrinsic flight rate that it wasn't possible to get to a even a hundred flights until almost two decades into the program. A good flight test program would have revealed the Shuttle's flaws much sooner, and allowed them to be fixed (or more likely, led to a next generation).

What has improved since 1970 that gives confidence that we can design a system with 99.99 or 99.999 percent reliability?

Many things, including TPS, avionics, batteries, and lessons learned from the Shuttle. The design compromises forced on NASA by politics and penny pinching (e.g., SRBs, oversized orbiter that didn't allow interior propellant tank, etc.) were major contributors not just to its high cost, but to its poor safety.

You cannot draw general conclusions about reusable space transports from a single data point (any more than one can draw general conclusions about space stations from the disaster that is ISS). This is the fallacy of hasty generalization.

Posted by Rand Simberg at July 30, 2006 04:29 PM

************

Yep

I agree. This was after Cap Weinburger of OMB at the time said that NASA gets this much money and no more, deal with it. He was told that the redesigned system would have much higher operational costs than the original fully reusable two stage system and his response was, "that is for a future administration to worry about".

There are a lot of lessons to be learned from the Shuttle program. I am hearing that several improvements, the best of which are Electro Mechanical Actuators (EMA's) to replace all of the STS hydraulics, has the capability of dramatically reducing turn around time for a reusable system. Are they going to use this on the CLV? Nope, as the contractor for the hydraulic systems already has 60 flight sets of hardware built and they want to use them.

The EMA's to replace the APU's on the Shuttle were proposed (and prototypes built) in the early 90's but were never integrated into the system. It is possible that if the orbiter had had this system that Columbia might have made it to a low enough altitude and speed as to allow for a bailout. If you really dig down into the data on Columbia you will see that the orbiter hung in there until the hydraulic system was compromised and the pressure dropped to zero.

Getting rid of the APU's, going to a non toxic OMS system as was proposed at one time, along with a better TPS system like the one Lockmart developed would have cut the orbiter turn around time to about 1-3 weeks.

Yes there are a lot of lessons to be learned, but that still does not give you two or three nines in reliability. For all the talk of reliability that Spacex touted, it was a 25 cent part that took the vehicle down.

Dennis

Posted by Dennis Ray Wingo at July 30, 2006 07:55 PM

For all the talk of reliability that Spacex touted, it was a 25 cent part that took the vehicle down.

Yes, a $0.25 part that would never have been used in a vehicle that had to operate through many cycles for many years...

The design philosophy between expendables and reusables is vastly different. Or at least it would be with a semi-intelligent designer.

Posted by Rand Simberg at July 30, 2006 07:59 PM

Uh Rand

The 25 cent part was on the resuable part of the vehicle.

Dennis

Posted by Dennis Ray Wingo at July 31, 2006 08:01 AM

I'll repeat. "Yes, a $0.25 part that would never have been used in a vehicle that had to operate through many cycles for many years..."

There's "reusable" (and rebuildable, which the Shuttle is) and then there are space transports. It was a poor design decision, and one didn't need to fly a lot of times to know that.

Posted by Rand Simberg at July 31, 2006 08:21 AM

Rand

Which brings me back to your original statement

*****

******

No one is perfect, neither engineers, engineering teams, or designs. Five billion dollars does not buy you the nines that you want because there are still humans involved. Look at all of the money that has been blown by private, semi private, and government sponsored teams. NONE of them have done perfect designs and until we have a much larger experience base, which means successive generations of improving design, we are not going to get to the reliability that you say we can today.

Dennis

Posted by Dennis Ray Wingo at July 31, 2006 11:56 AM

As is often the case, you "know" something that isn't true

Would you like a series of links to the volumous Rand Simberg rants about other people's spelling and grammar?

Or were you just pretending to waste time randing on those things?

Or is that you are just naturally argumentative and have to invent something if you run out of reasonable discourse?

No... don't answer that, it was a rhetorical queston.

Posted by Dave at August 2, 2006 09:38 AM

Would you like a series of links to the volumous Rand Simberg rants about other people's spelling and grammar?

I doubt if you'd find anything that would qualify as a "rant" except in your own mind. And even if you could, that doesn't mean I was offended by it. Words mean things (except, apparently, to you, to whom they mean whatever you want them to mean).

And I'll withhold comment on the irony of the misspelling in that sentence...

Posted by Rand Simberg at August 2, 2006 09:49 AM

Words mean things (except, apparently, to you, to whom they mean whatever you want them to mean).

No, words mean things, and occasionally, in the heat of responding to your reality challenged musings people get words incorrect, of course, within context, most people understand what is being said.

Except, you, for some reason, don't.

As I said, it was a rhetorical question. Getting you to admit to any position you've publically held is virtually impossible and has been over the 5 or 6 years I've noticed your ramblings on line.

Posted by Daveon at August 6, 2006 06:58 AM

If you remain curious, I've done some quick googling on this site for posts with references to Spelling. In the first 2 pages, you've 3 actual articles where you complain about spelling, including one where you say you, "see mispelled words" because they spring off the page at you.

There are also several comments from you which meet my view of a rant, although yours is quite obviously different, including a rather silly discource on the issue of "lose" and "loose" where the context of the post was clear to everyone, except, it would appear, you.

Heh. Plus ca change.

Posted by Daveon at August 6, 2006 07:15 AM

David, you really need to get a life.

Posted by Rand Simberg at August 6, 2006 08:23 AM

Post a comment