The SpaceX “Test Failure”

This isn’t new — I wrote it on Saturday at Ricochet, but it’s behind the paywall, so I thought I’d repost it here:

So the big news yesterday for people in the space business was that SpaceX finally lost an experimental test vehicle in its program to make its vehicles reusable (crucial to dramatically reducing costs to the point necessary to achieve its corporate goal of opening up the solar system). Some criticized it as a “failure” of the company. This is nonsense.

People need to understand that the purpose of an engineering test is to learn something. As I said on Twitter last night, the only “failed” test is one in which you didn’t get the information you were seeking. Losing hardware in a test is not a “test failure,” per se:

For example, consider the crash testing of cars, in which a successful test results in a wrecked car, but tells you what its weak points are so that you can improve the design, and the only test “failure” you can have is if the car fails to hit the barrier. In SpaceX’s case, the goal of the test wasn’t to destroy the vehicle per se, but they were fully aware that this could be an outcome. In fact, Gwynne Shotwell, the company president, said last year that she was a little disappointed that they retired the first test vehicle, Grasshopper, because the fact that they didn’t lose it in a test meant that they weren’t pushing the envelope hard enough.

Had it failed to deliver a payload of a paying customer to its designated destination, that could have rightfully been called a “failure” and the company justly criticized for it. But when an experimental vehicle crashes during a flight test, that’s called “flight test.”

SpaceX probably knows, but it hasn’t yet been reported what the cause was. The most common cause of failure in rockets is failure of stage separation, which doesn’t apply in this case, of course, since it is a single-stage test vehicle. Also, it could be an engine failure, but they have a lot of experience with their engines and hardware in general, so that’s an unlikely cause.

For this kind of vehicle, it’s really a test of the flight-control system, which is not only the computers, and sensors, and software, but the actuators that steer it. It’s possible that they had an actuator or engine-gimbal hardware failure, but they’ve had lots of test flights and never run into that problem. My guess (and it’s only that), based on viewing the video, is that they were pushing the vehicle beyond its capabilities to do something (perhaps translate, i.e., go sideways, while also descending or changing attitude) that they’d never attempted before, and it lost control (like an aircraft in a tailspin) without ability to regain it.

Once you lose control the decision to terminate flight comes pretty quickly, because bad things can happen very quickly after that. If they hadn’t been able to do the flight termination, and if it had resulted in unexpected damage on the ground, that would have been grounds for criticism, but the vehicle was safed exactly as planned, under FAA guidance and supervision.

Other than losing the vehicle, this flight was indeed a great success by the criteria of providing the information desired. At least two people from SpaceX, including Lars Blackmore, the lead of their entry, recovery and landing team, tweeted last night that they got “lots” of data.

Presumably in this case, if my theory is correct, they now understand the limits of the flight-control system. It may be that they will be able to ground simulate the failure, and tweak the software to avoid it in the future.

Was this a setback for SpaceX? Someone on Fox referred to the test last night with “A small rung on a long ladder to Mars broke on Friday, when a rocket test in Texas ended in a midair ball of fire.”

Jeff Foust called it that in his piece at the NewSpace Journal, and Jeff is a very smart guy, but I think he’s wrong, or at least, it’s not obvious that it is. In fact, when I asked him, Lars tweeted that he didn’t necessarily consider it one:

I would consider something a setback if it actually results in a delay of a critical program milestone. I think they have another test vehicle (that they’ll be flying out of New Mexico soon to do higher-altitude testing), and if they need yet another for McGregor, given their production capacity, they could probably pull one off the line and modify it pretty quickly. They’ve found something to fix in the next test vehicle (and possibly, though not necessarily, depending on what caused it) in an operational one. Also, in a sense, they’re no longer test-flight virgins, and may even be more bold going forward.

It’s certainly not going to affect their future launches (most importantly, next week’s), since it’s a side experimental program on which none of their current customers are dependent. So no, I don’t think it was much of a setback, if any.

On the other hand, I think that Blue Origin’s loss of its test vehicle three years ago may have been a setback, because they haven’t flown anything since (as far as I know). Unlike yesterday’s event, it may have been a totally unexpected, “back to the drawing board” thing. But I have no inside knowledge.

In addition to the general point of the difference between a hardware loss in a test and failure in operations, there is another point to consider here. While you expect problems in flight test of any new vehicle, VTVL (vertical take-off, vertical landing) types are particularly susceptible, not having wings to come home on if there’s a failure (though some use chutes as backup). I don’t think there is any serious VTVL company that hasn’t lost a vehicle in flight test, from Blue Origin, to Masten, to Armadillo, to Unreasonable Rocket. As Elon Musk tweeted last night, rockets are tricky:

I’d say that losing a VTVL vehicle in flight test is inevitable, almost a rite of passage, and that SpaceX just finally joined the club.

In fact, this isn’t actually the first experimental vehicle they’ve lost attempting to land it. It’s just the first on land. In a very real sense, every previous attempt to do an ocean recovery of the first stage, after it had completed its primary mission, was a flight test, and a success in that they got great data from each one to build on the next, and “failure” only in the sense that they didn’t succeed in actually recovering them. The company plans one more of these water “recoveries” this fall. Based on history, they have low expectations of getting the vehicle back this time as well, but obviously expect to get critical data needed to start to land actual first stages on land (though the first attempt or two will be on a barge at sea before they have demonstrated the control required for the FAA and the range to allow a flight back to the launch site).

But with each test, regardless of whether they get the vehicle back, they continue on their risky quest, with their own money, to achieve a long-time dream of the space industry (though one that NASA abandoned after the Shuttle), of an end to the wasteful and costly practice of throwing vehicles away. They should be encouraged to continue in their boldness. As I note in my recent book, such boldness, not caution or timidity, is crucial in opening up the harshest frontier humanity has ever faced.

[Sunday-morning update]

OK, not exactly a “setback,” but SpaceX has announced that they will delay Tuesday’s planned AsiaSat 6 satellite launch one day, to Wednesday, to allow them time to review the test results to ensure that the vehicle loss wasn’t caused by something that could affect the flight. “Mission assurance above all.”

[Wednesday-morning update]

They announced yesterday that they’re delaying the launch for several days now, but it’s unclear if it’s related to the vehicle loss on Friday.

9 thoughts on “The SpaceX “Test Failure””

  1. It’s a shame that the Falcon 1 rocket was such a disaster that SpaceX could never be a successful company.

  2. I’ve read a bit about the early history of space flight and compared to the Redstone/Atlas era, SpaceX isn’t doing too badly. They’re trying to do something (make a stage of the rocket reusable) that hasn’t been done before. That requires pushing things to the breaking point and finding out why they broke. The test flight generated valuable data, therefore it was a success.

  3. What the heck is a “tail spin”?

    (NB I’m a flight instructor and know what a regular “spin” is)

      1. I think “tailspin” is a figure of speech. See http://www.av8n.com/how/htm/spins.html#sec-spin-types for a pretty good, not too technical explanation of spins. Denker is a CFI plus a PhD in physics, and so far I haven’t found any big holes in his aerodynamics and flight dynamics knowledge.

        We had our discussion in the other thread; I still think you and many others are underplaying this (having myself been in the same shoes they are in), but I agree that it isn’t a showstopper. I’m still thinking that actuation may be the culprit, though not in an “it failed” sense: if they were pushing the capabilities of the system, the actuation may have rate limited unexpectedly and put the vehicle into the equivalent of a PIO. I also wonder if they got the tail-wags-dog effect right; again, that would tend to show up at more extreme maneuvering. I don’t know directly anybody in their GNC group, so I don’t have a good idea of the experience level of their GNC team.

        I was also very surprised about the Blue Origin failure, especially what it was – made me think that they didn’t do a good job in the wind tunnel. But given how secretive they are, they may have been doing stuff under the radar. I notice they are still advertising for GNC and Aero types too…

      2. Wiki: In the early years of flight, a spin was frequently referred to as a “tailspin”

        Looks like it’s just a dated popular term.

  4. Great post as always. Can we get others to read you (with comprehension) before spouting nonsense? (Don’t groan, I’m… what’s the word?… special!)

    Automatic termination could be as simple as a single parameter going beyond preset bounds.

    The engine geometry might be another issue but how would that impact actual launches which use a different configuration?

  5. There was an article over at Spaceflight Now posted this afternoon saying that the launch is possibly delayed a couple of weeks. Appears the investigation of the test flight uncovered a wiring issue:

    Musk said the problem during the Falcon 9R test flight was with a “blocked sensor port.” If the same issue occurred during an operational Falcon 9 satellite launch, “it would have been outvoted by several other sensors. That voting system was not present on the test vehicle.”

    “What we want to do is triple-check is whether even highly improbable corner case scenarios have the optimal fault detection and recovery logic. This has already been reviewed by SpaceX and multiple outside agencies, so the most likely outcome is no change,” Musk said.

    So it appears the delay is a result of the outcome of the test flight.

    1. Yes. It sounds as though SpaceX is going to do a code walkthrough of the entire software corpus of the Falcon 9 and “Grasshopper” avionics suites.

Comments are closed.