Latest SpaceX launch ends in failure

Tesla and Solar City both exist primarily because Musk had the political connections to get generous grant money from the EPA. Tesla, for all the hype it gets, is still not a profitable venture, and recently had to secure a half-billion dollars in credit to avoid running out of money. Solar City is only profitable because of the extensive tax breaks given to solar installations. The company itself has admitted this.

And of course, Space X would be launching model rockets in Musk’s backyard were it not for the NASA contracts.

His motivation is world domination, preferably directed from a orbiting HQ/underwater volcano lair. Here’s his current twitter profile pic for evidence https://pbs.twimg.com/profile_images/579207936620138496/o1LpNXMK.jpg

SpaceX and its fanboys have been constantly harping on about how the company is going to change spaceflight forever, how Musk is the Steve Jobs and Bill Gates of the space industry, how their brand new rocket is so much better then the competition, they made fun of Antares for using Russian made rocket engines.

When their actual accomplishments are fairly limited and not really impressive compared to the Government run programmes, and let’s face it, almost everything that has been built by NASA and the Sovs/Russians has actually been built by contractors and often enough the operation has been the contractor as well (astronauts being the obvious exception).
i

It’s hard to see what exactly SpaceX has done which is worth the hype.

It should also be said that the Atlas V (and previous iterations) use a Centaur upper stage. While the Centaur is a reliable stage now, that wasn’t always the case, and it experienced many failures over its 50+ year lifetime. The Falcon 9 upper stage is, relatively speaking, still brand new.

There’s some indication that the failure wasn’t due to the rocket but rather the payload (the Dragon capsule and its trunk cargo). Still total speculation at this point, but if true it would potentially mean its commercial customers aren’t affected (since they don’t use Dragon).

Tesla would be profitable if they weren’t rapidly expanding. Business take loans all the time; it’s not a sign that they’re on the brink of disaster. Tesla had many funding sources available and a loan made the most sense.

No, they made fun of Antares using Russian-made engines that were literally manufactured in the 60’s and stored in a warehouse somewhere.

They–like many others–cal into question the use of Russian RD-180 engines by the Atlas V. They do not doubt for a second the high quality and performance of these engines, but rather their political exposure to the whims of Putin, and the fact that we’re sending Russia billions of dollars for these engines.

They do it for half to a quarter the price as the others. And they’re this close to nailing first-stage recovery, which no one has done before with a liquid fueled rocket.

I don’t think I dispute any of that. Is this a response to the issue of whether Musk’s primary motive is profit?

I’ve tracked SpaceX pretty closely for a few years now, and sometimes I wonder whether Elon is intent on making himself a millionaire with this business. Yes, I know he is currently a billionaire.

But really, they have done very well so far. Some fanboys seem unwilling to admit that they have had problems - Elon’s businesses seem to have a pattern of promising that the whole world is going to change NEXT YEAR when something revolutionary hits the market - only to have delays pile up.

It’s an unfortunate event to have this mission fail, but I think everyone should recognize that it was pretty much inevitable to have a failure, and that SpaceX will recover and the industry is going to be better for it. Really, if it weren’t for SpaceX, nobody would be talking about the Vulcan. No way.

Yes. It’s pointing out that all three of his companies seemingly exist mainly to sop up government funding.

Case in point: Tesla has been promising that an “affordable” electric car (previously supposed to be $30,000, recently bumped up to $40,000) is coming along REALLY, REALLY SOON NOW, for at least the last couple of years. The way the electric car market currently is, if they ever actually do produce that car, it’ll drive them right into bankruptcy. It’s a telling point that all the major automakers produce electric cars in small quantities, and seem to treat them mainly as something they need to produce to make the government STFU.

My understanding is that there were 3 previous failures w/ Falcon. And the detonation of Grasshopper. Not to mention the seemingly routine delays for technical reasons. All of which adds up to money…

Heck, I’m a big fan of increased attention and resources towards space flight. To the extent Musk encourages that, I’m a fan. But my opinion is that he is an astoundingly canny businessman to the extent that he has gotten so many people to think of him as something other, or to think of his ventures as more successful and less subsidized than they are.

Falcon 1 failed in its first three launches. Totally different vehicle from Falcon 9 (except for the first stage engine, which was an early version of the current Merlin 1D). Also, completely within the range of acceptable failure for a new rocket, with new engines, by a new company. Space is hard. Not just because rocket design is hard, but because it’s hard to put together an organization with the mindset needed to launch to orbit. Nobody has gotten to orbit without going through this phase first.

F9R (Falcon 9 Reusable–different from Grasshopper) did auto-terminate in flight. They have repeatedly said it was a test program, and that if they did not have an explosion or two, then they were not trying hard enough. The whole point was to explore the boundaries of the operational envelope, and the only way to do that is by losing a vehicle every so often.

Neither of these things have anything to do with the recent Falcon 9 disaster.

When he tries to get a first stage rocket to land vertically at sea to reduce space flight costs, I’m going to give him a wide latitude on his financing.

No, the NASA Crew Resupply Services missions, and government payloads in general do not carry launch or satellite/space operations insurance. Commercial satellite operators (typically just large space telecom companies) often purchase launch insurance but it doesn’t work like automobile or home insurance; instead of being underwritten by bank or company based upon an actuarial evaluation, launch insurance is basically high risk parimutuel betting for very wealthy people with little poor math skills. Insurance is provided by a contract agent like Lloyd’s of London and backed by a small bond (usually 3% to 5% of the maximum claim amount). Upon failure, the injured party generally has to sue the guaranteer for more than the bond amount, and such suits often last years and sometimes decades, with the lawyers getting most of the reimbursement. (I know as of at least a few years ago there was at least one suit that had been going on since a launch failure in the 'Eighties.)

Instead of buying insurance high value NASA and Air Force launch vehicles provide what is termed launch or mission assurance, in which an independent team (generally Aerospace Corporation and/or private contractors) reviews design information, vehicle and subsystem test and integration (T&I) data, material and component manufacturing certifications, and (depending on type of program support) may perform independent mission analysis, software IV&V, and detail performance analysis of components and systems. This has been the traditional mode of assuring launch vehicle reliability and can be statistically demonstrated to improve reliability by around 50% to 80% over launches conducted without mission assurance support. Unfortunately, doing all of this work (in parallel to what is already done by the launch vehicle integrator) is often viewed as redundant and unnecessary until a failure occurs due to an oversight in analysis or testing that should have been reviewed more thoroughly.

There is actually a significant difference in those numbers as can be seen in the table below (based upon a first level Bayesian estimate of probability of success). Before CRS-7 (the 14th flight of the Falcon 9v1.1 vehicle) it had a predicted P=0.93 (93% probability of successful flight). With one failure in 14 flights, it now has a P=0.88, which is considered marginal. (We generally like to see P>0.95 to be considered high reliability, which F9v1.1 hadn’t had enough flights to achieve even before this failure.) One failure in 25 flights gives a predicted reliability of P=0.93, and at one failure in 65 flights the predicted reliability is P=0.97. You can see how these compared to other launch vehicles at Ed Kyle’s Space Launch Report (scroll down to see the summary of all operating vehicles, and there are tables for previous years going back to 1998).


Probabilty of success p of flights for given number of flights with f failures
n\f 0    1    2    3
0   0.50
1   0.67 0.33
2   0.75 0.50 0.25
3   0.80 0.60 0.40 0.20
4   0.83 0.67 0.50 0.33
5   0.86 0.71 0.57 0.43
8   0.90 0.80 0.70 0.60
10  0.92 0.83 0.75 0.67
12  0.93 0.86 0.79 0.71
13  0.93 0.87 0.80 0.73
14  0.94 0.88 0.81 0.75
15  0.94 0.88 0.82 0.76
20  0.95 0.91 0.86 0.82
25  0.96 0.93 0.89 0.85
50  0.98 0.96 0.94 0.92
65  0.99 0.97 0.96 0.94
100 0.99 0.98 0.97 0.96
 
Probability p of k successes in n attempts
p = (k+1)/(n+2) = (n-f+1)/(n+2)
with f = n - k

Of course, these are summary statistics based upon just launch volume and number of failures without consideration to sequence or type of failures. It is a valid point that the “infant mortality” of systems often leads to early failures as essential design flaws or operational limitations are discovered, and thirteen successful flights of the Falcon 9v1.1 (which is a very different vehicle from the predecessor Falcon 9) is actually a pretty phenomenal record that certainly beats the odds versus most other families for the first ten flights, in which there is almost inevitably one failure or major mission anomaly; in fact, I can only think of two other vehicle families with a comparable number of flights which have achieved that record of initial success, which can only be attributed to the hard work and expertise of the engineers working on the system.

However, SpaceX also had a number of substantial anomalies on many of those flights and has always operated under the notion that the trade of risk to increase throughput and reduce cost is worthwhile, and to the extent that their customers are cognizant of and in agreement with the risk, it is a reasonable way to approach reduction of launch cost. The high cost of conventional launch systems is, in part, because of the low risk tolerance and desire to do every possible thing to mitigate launch failures. (Unfortunately, most launch failures that do occur, at least in the last few decades, like the O-ring blowby on Challenger and failure of the RCC leading edge TPS on Columbia, are due to known problems that the launch operator has come to view as low risk despite the potential high criticality of a failure or lack of understanding about the nature of the possible failure modes.) Unfortunately, applying the kind of “80/20” approach still leaves the potential for low probability but high criticality failures, and rocket launch vehicles have thousands of components which have to perform critical functions in a specified sequence within a narrow range of parameters against which only very limited redundancy or resiliency against failure can be provided. You’d like to believe that you’ll pick up design problems in functional and margin testing and T&I problems with good people and processes, but in many aspects there is nothing short of an actual flight environment and conditions to rigorously test a system, and even at that no single flight will stress every aspect of the system, so like any other complex system, a large amount of experience is required to achieve true realized (instead of just predicted) reliability.

Rendering a spastic, off the cuff judgement about the future of SpaceX or commercial spaceflight in general based upon a couple of recent failures is just not warranted or justifiable at this point. This business has always been–and will remain for the foreseeable future of propulsion technology–one of flying at high risk of catastrophic loss with small margins and substantial uncertainty in the small population data for realized reliability; the notion that any conventional space launch system is going to become “commoditized” akin to automobiles or commercial air travel without substantial advances in materials, propulsion systems, and avionics, (and a high volume of flight to identify design flaws and reliability problems) is a fantastic and unreasonable expectation. We don’t know what caused the CRS-7 failure yet and may not know for weeks, the delay not being that SpaceX doesn’t have adequate telemetry to evaluate the vehicle operation (the Falcon 9 is one of the most highly instrumented launch vehicles I’ve seen) but rather they have an enormous amount of data to look through and no immediately apparent root cause. Even if the cause is some fundamental flaw in the system (which somehow wasn’t stressed for the first thirteen flights) it is probably fixable by redesign, and it isn’t as if SpaceX is adverse to making substantial changes in the vehicle almost literally on the fly. If it is an incidental failure (e.g. some kind of procedural misstep or bad flight code) it may be possible to flight the same design indefinitely with just minor changes.

Regardless, I have little doubt that SpaceX will continue to fly; it’s just a question of what it takes to regain confidence in payloaders versus flying on the substantially more expensive EELV or Ariane (or other) vehicles. The larger question is ability of SpaceX to come anywhere close to their advertised launch costs (which have been incrementing upward and only represent basic manifesting without the additional payload services or access to test and integration data to perform assurance activities) while clearing a sustainable profit. I remain dubious that even if SpaceX does recover the first stage (which I give odds that they’ll be successful within the next three to five attempts) that they’ll be able to substantially reduce costs below what they’re currently charging as the majority cost is in the T&I effort rather than propulsion system and stage manufacture. Even at the current price point, however, SpaceX is forcing the United Launch Alliance (the current provider of Delta IV and Atlas V vehicles) to both reduce costs and increase potential launch volume, as well as providing an alternative vehicle, all of which are valuable and worth weathering a single (or even several) launch anomalies.

And assuring future reduced-cost access to space for critical space services and industry, with the eventual goal of having a sustainable presence beyond Earth orbit by accessing and utilizing space-based resources for raw materials (at least propellants and other consumables) is key to more wide scale exploration and exploitation of the vast untapped resources in the rest of the solar system, which is an avenue that has been poorly supported by existing governmental space programs which are primarily focused on national prestige and surveillance/military activities with a smattering of pure and applied science missions thrown in to keep the poindexters happy. If not SpaceX, then some commercial enterprise needs to achieve this threshold because even after five decades of space “exploration” by NASA, ESA, the former Soviet Union, JAXA, and now China and India (some of it, such as the Voyager program, the Soviet Venera probes, the Hubble Telescope, Hayabusa-2, and Cassini-Huygens, not to mention the ongoing New Horizons and Dawn missions, having provided great scientific insight) we still can’t do so much as fill a canteen with water from a space-based source, much less make use of commodity and rare metals or utilize fissionable materials and hydrocarbons that are readily available without contributing pollution to our planet’s ecosystem.

Stranger

Nice to see you back, Stranger.

Care to speculate at all on the causes of the current incident? The video clearly shows problems with the upper-stage LOX tank, and Musk’s tweets confirm this. Some obvious possibilities are something to do with the pressurization system–god knows they’ve had problems with their helium systems before–but the problem sounds a bit more mysterious than that. On the other hand, perhaps they’re cagey because assigning blame early may make it look like they’re pointing fingers at an outside supplier, and that wouldn’t be cool if it turned out to be something else.

Another speculation I’ve seen is that it had something to do with the payload; in particular, the International Docking Adapter mounted in the Dragon trunk. Possibly it came loose and impacted the LOX tank, causing a pressure spike and ultimately a puncture. The only real thing this speculation has going for it is that it could explain why this flight failed and not previous ones. The environmental conditions on this launch (winds, etc.) were about as ideal as could be hoped for and so it doesn’t seem like those could be a factor.

I’m sure we’ll find out soon enough, but in the meantime it’s fun to speculate.

It would be irresponsible of me to speculate (although I’m not involved with this or any future F9 mission I do have some proprietary knowledge of the launch vehicle) and also premature in absence of publicly available data other than the flight video. I have confidence that SpaceX will come to a root cause or at least a credible most likely cause, and because it is a NASA program I expect LSP will at least perform an assessment if not independent review which will be released into the public record even if SpaceX doesn’t relay it publicly, though I expect them to be reasonably candid for the purpose of regaining customer confidence that they’ve addressed and resolved the issue. I will say that I suspect it unlikely to be a fundamental design flaw not previously experienced or apparent in flight data, and more likely to be some issue in processing or integration, or some other incidental failure that can be rectified without making major changes to the vehicle.

With a few exceptions, most of the major components and systems on the Falcon 9v1.1 are built in house or assembled from commercial sources (with the intent that this ‘vertical integration’ would avoid the common problems with aerospace vendors of price fluctuations and delivery schedules) so it is less likely that SpaceX is trying to protect a vendor than just doing a thorough job of investigating the possible failure modes before releasing speculative information. This is standard (and good) practice in an investigation, and whenever contractor fails to do so it generally results in a lot of errant and incorrect speculation by enthusiasts that later has to be explicitly addressed. There are still people walking around claiming that Challenger was forced to launch because of Reagan’s State of the Union speech and that the SRB ‘exploded’ due to inappropriate and unguarded statements by some people involved with the investigation, and those misstatements are directly contradicted by both the internal NASA and independent Presidential (‘Rogers’) Commission reports as well as individual accounts of members of the Rogers Commission.

Stranger

Understood. I think SpaceX probably realizes that they have to be a little more judicious than usual in the release of tentative information for the reasons you mention, but that they will be candid once they understand the problem.

For the interested, there’s a slideshow of the tank burst here. It’s a little hard to be sure from the perspective, but to my eye it appears that the tank rips open along most of its length on both sides. You see two seams along the length brighten and then release a large volume of LOX. The Falcon 9 is partially pressure stabilized and the tank fails structurally at this point (whether or not it retained most of its inherent strength after the burst), since it is at quite a high acceleration at this phase of flight and it has to withstand the payload weight. It crumples like the aluminum can that it is. Later, you can see what is almost surely the remains of the Dragon capsule tumble off to one side.

To be precise, I was using “significant” in the frequentist sense, in that there is not a statistically significant difference between one failure in 19 and one failure in 65. (p = 0.40 using Fisher’s Exact Test).

The Bayesian approach is entirely reasonable, but I’m not sure you can really use it to confidently state that a rocket with a string of 14 successes is less reliable than another with a string of 25 consecutive successes and one failure. That is entirely dependent on a single rare event in a very small sample (if you’ll excuse me treating aerospace engineering like drawing balls from an urn…)

Quibbles aside, I greatly appreciate your posts on these topics!

Well, yes, using a frequentist approach there needs to be a population of at least a couple hundred launches before it is possible to make a credible estimation of reliability with any useful degree of confidence, hence the use of a Bayesian approach as a way of estimating reliability strictly from a single parameter (modified ratio of successes to attempts). In essence, it assumes the first attempt is a 50% chance of success and each success builds up reliability from there. As I noted above, it doesn’t address any changes to the vehicle or processing methods which may occur between launches, nor the sequence of failures. Realistically, the assessment should be weighted to give more prominence to a string of successes, and should penalize any changes to the design or application, but there is no rigorous methodology for doing so (although I have a proposal for a framework for this as a draft thesis proposal and have been intending to convert to an AIAA or SRE paper) and the first order model actually fits the as-flown data quite well as mapped against initial and subsequent even without any modifications to weighting (although this is less a consequence of the precision of the method than that it inherently represents a posterior modification of the prior distribution upon experiencing a failure, so you’d expect it to be reasonably accurate). How ‘good’ the initial estimates are at evaluating reliability for an individual flight is kind of a philosophical point, but when I’ve mapped probability of failure at any point after the first ten flights to the subsequent three or more flights, the actual incidence of failures of an a single class across the entire population of space launch vehicles happens to be right on the nose (again, not accounting for sequence of prior failures).

Now, the method for determining ‘actual’ mission reliability is quite different, and is based upon a combination of analytical system reliability analyses (in which applied reliability of components is multiplied to get subsystem reliability, and the subsystem R values are multiplied together to get vehicle reliability, and so forth) and qualification testing (which ostensibly provides an as-tested reliability value corresponding to the test margins), but these are widely regarded as being fairly bogus numbers just intended to meet a specified reliability requirement on paper. Getting demonstrated reliability on mechanical systems on launch vehicles from direct testing is too costly, labor intensive, or just physically infeasible to be practicable. With rocket launch vehicles, it is never really possible to do an end-to-end test that stresses the system with margin or test all of the subsystems in a genuinely flight-like manner and capture the dynamics of system interactions. And most actual failures or anomalies aren’t due to a component which failed to meet expected reliability, but rather either because of a previously known or suspected weakness (such as the SRB o-ring blowby on Challenger) or a previously unknown system interaction (POGO on Titan II and Saturn V stages) which weren’t or couldn’t be demonstrated in testing.

So actual (i.e. realized) reliability depends on the knowledge and experience of the engineers and technicians designing and integrating the system to build adequate robustness into the vehicle, be aware of all failure modes and consequences, and recognize when something isn’t operating correctly (“out-of-family”) in test and integration, even if it falls within test limits. And it is tough to do that every time, all the time, on a system that if it starts to fail in flight conditions there is pretty much nothing to be done about it. The history of “recoverable” failures (compared to catastrophic loss of vehicle or loss of mission) such as Apollo 12 and 13, or STS-51-F are notable for their infrequency; most failures result in a total loss of vehicle and mission.

Stranger

Don’t forget SpaceX CRS-1!

I don’t suppose you have any knowledge/involvement/etc. on the Super Strypi? I have a payload on their inaugural flight this October and will be visiting Kauai. Should I assume the chances of it exploding are 50-50 :)?