Looks like we’ll have some answers in about a week:
Expect to reach preliminary conclusions regarding last flight by end of week. Will brief key customers & FAA, then post on our website.
Anybody have a clue as to the reliability of the V2? Thousands of those were launched in combat , so at least some data should be available.
I had some small involvement (in a review capacity) on the original configuration of the Super Strypi vehicle (Castor 4XL, Orbus 7S, Star 30BP) before the configuration changed to the current Aerojet-supplied LEO motors. I assume your payload is launching on SPARK, which is the inagural launch of this configuration out of PMRF. As an aero- (Stage 1) and spin-stabilizied (Stage 2 & 3) vehicle a lot of things that can go wrong in flight–specificially, with TVC system and recovery of control authority after staging–are eliminated by the simplicity of the vehicle operation. However, there are, as with all launch vehicles, plenty of other things that can go wrong and that may not be apparent until experience is gained in the launch environment. I know that’s a harsh thing to say to a payloader, but it is the unfortunate truth, and you just have to kind of grin and bear it.
The application of an R=0.50 prection for success for the first flight is purely a naive prior distribution which assumes nothing whatsoever about the vehicle design or experience of the operators. Realistically, the LEO familly of motors is built upon commercial experience with similar technology and barring a basic design or manufacturing error in the in case, insulation, or nozzle, some unlikely malfunction of initiation or ordnance seperation systems, or unanticipated aggressive natural environments (high wind shear, lightning, et cetera) I would anticipate realized reliability to be much higher (the Scout family–a somewhat similar concept suborbital and space launch vehicle–has a realized reliability of around 80% to 85%, and the Black Brant family of sounding rockets has about a 95% realized reliability overall).
However, even starting with an initial prior of R = 0.80 still gives reliability (for no failures) that is within a 0.01 variance with the R=0.50 distribution after ten flights, and is essentially the same after 25 flights, although it does improve estimates (perhaps artificially) in the case of failures. Given the good correlation to realized data using the naive prior trying to juke the prediction with assumption of higher reliablity is probablhy not justifiable without a more rigorous methodology, but it doesn’t literally mean that success of the first launch is literally the same as flipping a coin; it is more of a statement about the dearth of knowledge at that point to make a more accurate prediction on an empirical basis.
Astronautix.com lists a success rate of ~78%. The Germans built somewhere around 5000 units, about 3000 of which were flown operationally and the rest either test flights (the German A4 test program or the later US Project Hermes after WWII using V-2s appropriated from the Mittlewerks and engineers brought over under Project Paperclip) or were non-functional models examined by the Americans, British, and Russians in order to develop their own native rocket designs.
The V-2 was the first operational ballistic missile of its class and size with a number of then-novel innovations that continue to be applied in rocket propulsion systems to this day, and was also built under wartime conditions with poor material and process control by a largely forced labor workforce, so it is unsurprising that the reliability is poor. It was also more of a propaganda effort–part of the Vergeltungswaffe (“Retribution Weapon”) program–than an effective weapon of war; although it resulting in the deaths of around nine thousand people, mostly in Antwerp and London, more people actually died working on the rocket than were killed by it, and each one cost as much or more than a strategic bomber while delivering only a 1 metric ton explosive payload.
Stranger
Hey I came in to this thread to be confused by technical specs and physics, not statistical analysis!
: 
Neat! The more I learn about the space industry, the more I realize that it’s a pretty small world. So I’m pleasantly surprised but not shocked that you would have had some degree of involvement.
The project as a whole has been an odd one and at some point I’d like to piece together the history.
Yep. I can’t give too much detail until the launch, but we’ll be a subpayload on one of the larger cubesats. Nothing too special–just some cameras, a radio, and basic sensors–but the hope is that we’ll exceed the results of our previous attempt (we achieved contact but didn’t get any pictures back).
That’s for sure! I don’t think we’re under any illusions on our side. Still, it is encouraging that the Super Strypi is built from a reliable family of motors.
Sorry, I might be a bit slow today.
Is your position that his primary motivation is profit, evidenced by the fact that he’s getting all that government money, or that his primary motivation is not profit, evidenced by the fact that he can’t seem to turn one at all without government subsidy?
Or are you maybe just annoyed that he’s getting government money?
Diceman doesn’t have a point at all. His characterization of Musk’s companies is absurd when put into context. The electric car subsidies Tesla receives are available to any manufacturer. The $465M loan that Tesla received from the DOE–which they’ve paid back–was from the exact same loan program (the ATVMLP) which provided $5.9B to Ford and $1.6B to Nissan (nevermind the tens of billions of bailout money that GM got). While NASA did provide SpaceX with development funds, most of the funding has been through ordinary launch contracts, and is no different from a company that sells pencils to the government. Contrast to the usual cost-plus contracts that are usual for government development.
And so on. Musk’s companies are far less dependent on government funds than their competition. They compete well despite being in an environment of regulatory capture, and while they have made some progress in dismantling anti-competitive laws (dealerships, etc.), it is still a challenge for them.
Should I even mention the massive externality costs of fossil fuels? That is, the trillions of dollars in health and environmental costs that should, by all rights, be included in the cost of every gas/diesel car and the price of electricity? All these clean energy subsidies are really just a way of leveling the playing field a tiny bit. If these costs were paid for, the subsidies wouldn’t be needed and we’d have switched away from fossil fuels long ago for most things.
Ho hum - another successful ULA launch.
I respect and appreciate what SpaceX is doing. But, MAN, it is something when ULA has this down to the point that it is as reliable as (a very expensive) bus service.
Preliminary failure investigation has reached The Tweeterz. Basically it seems that the helium tanks in the second stage (which I think may be located inside the liquid oxygen tank) broke free, probably due to failure of a single strut. I can’t quite follow the second-hand-live-tweeted discussion of the over/underpressure events… for some strange reason, Twitter may not be the best medium for communicating the details of a highly technical investigation.
Other relevant and potentially interesting details: if it is a strut failure, it’s straightforward (by aerospace engineering standards) to test and reinforce it. Also, the Dragon spacecraft was actually in good shape as it tumbled beyond the horizon. It might easily have survived if its parachute were armed and deployed.
Snapchat it is then.
yet another departure from standard practice-announcing the cause so quickly. Of course Space-X has an interest in showing everyone that they are on top of the situation. But announcing this so quickly, many months before the investigation is complete is a sharp difference from normal practice. Since initial conclusions are often wrong, it is likely there will be more to the story. But now that a cause has been identified, even if it is wrong, this will be the story that sticks with everyone. It will be interesting to see whether Space-X benefits or ultimately is hurt by doing this. I am going to assume Musk will benefit. Whether the company does will be interesting.
ArsTechnica article summarizing the preliminary findings.
I was waiting for an official statement from SpaceX before posting more, but… it takes a lot of beer to counteract this heat. Anyhow, with that in mind…
Basically it seems that a substantial cause of this failure was due to outsourced quality control. They used an outside supplier for the strut that holds each helium tank down, but it seems that the struts were not up to spec. The one that failed here (assuming it’s the cause) failed at only 20% the rated load. SpaceX had a stockpile of this particular part and tested all of them (“hundreds” have flown so far, and they had “thousands” on hand). A few failed at loads less than the loads they were specified to handle.
At this point we don’t know why this particular part failed, and we probably won’t know for sure. It could be simply because shit happens, even to experienced, reliable, high-cost suppliers. Or it could be because SpaceX cut one too many corners, and went with a low-cost supplier that promised higher performance than they could actually deliver.
Second hand, I do know that there are difficult engineering trade-offs between choosing low-cost suppliers and needing to employ lots of engineers to ensure that those low-cost suppliers are delivering parts that meet those specs. That’s kind of a different matter with, say, a car, where a certain level of major and minor failures are tolerable. A $500 warranty repair is a hell of a lot cheaper than a $100,000,000 lost payload…
Stranger, is there any possible way of estimating the probability of a supplier failing to meet the specifications on a critical part? As a function of part cost?
obligatory “moar struts!” statement goes here.
I don’t even think it’s a matter of a “low cost” supplier, rather a “simple” part that SpaceX didn’t feel they needed to micromanage. It’s not like NASA has never had a problem with an expensive, supplier-produced part (coughHubblehack)
honestly I think it’s commendable that SpaceX has been so open about its operations.
The impression that I got was that a bunch of them failed at less than rated loads, but only one failed at the same load as it would have had in flight.
They did say there was a grain structure problem with the strut material. So it seems like it should be traceable upstream to some degree.
I’m really curious about how outside suppliers are handled in general. SpaceX didn’t individually test every strut (they say they are making a process change to test every one), but I’m sure they did at least randomized testing. Still, that doesn’t help you if there are rare, wildly out of spec parts. If you assume a normal distribution, you might look at the tested part variation and conclude that the chances of being outside the load limit is a one in a billion chance. But if the actual distribution has a “long tail”, the junk parts might be far more frequent, like one in a thousand. You wouldn’t necessarily be able to determine this by just sampling a couple hundred parts.
Not as a generic function of part cost, but there are clearly differencies between components qualified to MIL standards or commercial “space rated” (or equivalent) and commodity “off-the-shelf” components. The difference isn’t so much design or construction as it is that vendors of qualified components perform the required qualification and lot acceptance testing and provide certification data that shows that their processes are consistent compared to the qualified baseline and their subtier vendors are providing certified components and material. All of this costs money and creates the off-decried “blizzard of paperwork” to be reviewed as part of a mission or flight readiness certification, but assures that the launch or space vehicle contractor is getting what they expect. Failures of qualified components are very rare and are almost always found to be due to either exceeding the qualification baseline loads in some way (excessive cycling, temperature, mechanical stress, whatever) or because it turned out to be a counterfeit component (forged or misattributed certs). The probability of failing genuinely qualified components is usually predicted to be somewhere between three and six standard deviations from the mean based upon statistical sampling of the unit population and assumptions about the confluence of load extrema and component robustness, but also goes toward increasing the costs of qualified components to a rough order of magnitude greater than comparable commercial off the shelf components. BTW, the automotive industry has gone almost exclusively to using certified components (similar to qualified, and arguably more dillgent) for critical components to obtain modern reliability, and has managed to keep cost down by a combination of market pressure and efficiencies in documentation (signed digital rather than paper certifications and automated acceptance and proof testing) so it is possible to get good reliability with sufficient volume, although the degree of reliability in automotive is still significantly less than typical realized reliability of MIL spec or space industry standard components, and the criticality of a failure in a launch vehicle is almost certainly more catastrophic (in terms of fiscal loss and hazard) than any failure in an automobile.
Now, most qualified components are either avionics, energetics (explosive or combustion powered actuators and separation mechanisms), or pressurized system components. For the most part, except for very standardized components like fasteners, structural components are not qualified by a vendor and their suitability is determined by application specific testing (ultimate tensile/compressive/shear strength and proof strength testing). However, on every vehicle I’ve ever worked on, every single attachment point, click-bond stud, bracket, and bond was tested by a standardized pull test to the rated load plus some margin (typically the statistical maximum load x 2, or higher for critical structure). While the COPV brackets are not primary load paths for vehicle loads, they are certainly critical structure in retaining the system ullage pressurization, spin-up, and ACS system, which is certainly mission and vehicle critical. I find it difficult, to the point of utter improbability, that they don’t perform a pull test on each bracket. While tedious, it really isn’t that time consuming and it gives assurance structural failures won’t occur under the worst case flight loads.
I would also caution against the assumption that because a weak strut was detected in inventory, that this was the cause of the vehicle failure. There is a tendency, when performing an anomaly investigation, to attribute the first design weakness or component problem to the failure, but before pulling out the Jump to Conclusions Mat it is wise to consider whether a) the phenomena seen during the failure could really be attiributed as the root failure of this specific component, b) whether the are alternative failure mechanisms that could be root or contributory in cause, and c) how probable the supposed failure mode really is. If you found one defective component out of 10000 random parts across all lots, the odds are pretty low. If you found three or four defective items out of 1000 in the same production lot as your suspect component, on the other hand, your odds of concurrance have done up by a couple orders of magnitude to the point that it it is the next thing to a smoking gun (statistically speaking). Other failure modes are certainly possible, and this wouldn’t be the first time SpaceX has had problems with their tank pressurization system (previous replacement of a COPV on an integrated vehicle).
SpaceX is under pressure to get back to launch status because their valuation (last I saw was ~US$10B) is almost entirely based upon their manifest; significant slips may result in loss of confidence by commercial customers and demanifesting of planned payloads. But it is important to determine the actual root cause (or at least most probable) so the failure doesn’t occur again. To have a failure on a launch vehicle is expected, almost to the point of being inevitable at some point, but to have the same failure repeatedly would significantly undermine confidence in good engineering and manufacturing practices that underly practical reliability. This is why being too open about an investigation, and inviting speculation and Monday morning quarterbacking (largely by self-described “space nerds” who have little or no engineering background) is not the best practice. It isn’t about transparency but in going through the root cause and corrective action process thoroughly before making a diagnosis.
Stranger