Missile Defense Test Hits Target, But Not Off the Hook Yet

, senior scientist | June 25, 2014, 12:02 pm EST
Bookmark and Share

The Pentagon announced that the Ground Based Midcourse (GMD) test FTG-06b on Sunday was successful.

It is likely to take a few months to sort through all the data and to see in detail what went right and what went wrong.  However, at some point soon, the Obama administration will need to make some decisions about the GMD system, and one of them is what to do about the 14 new interceptors it has stated it plans to build. Raytheon is apparently standing at the ready to provide some more CE-II kill vehicles. According to a Reuters article, Wes Kremer, vice president of air and missile defense systems at Raytheon said, “There are no other hurdles that we’re aware of, so we expect that we will go into production shortly.”  Reuters also reports that the Defense Comptroller, the Pentagon’s top financial official, said that he “think[s] our plan now remains to buy the original 14 interceptors.”

However, when the Secretary of Defense Hagel announced the plan to buy 14 interceptors by 2017, he said this would be on the condition that the U.S. “fly before we buy,” requiring that “we have the complete confidence that we will need.” And Congress will have its say, as well.

This success doesn’t let the Missile Defense Agency off the hook—while Sunday’s success must have been a needed boost of confidence for the missile defense team, since they were able to identify and find a mitigation scheme for the inertial guidance system problem revealed in FTG-06a, it tells us very little about the reliability of the fleet of fielded interceptors or the CE-II kill vehicle. It certainly isn’t close to providing the “complete confidence” that Sec. Hagel called for. A 1 in 3 record might be acceptable for the developmental stage of a system, but it is not acceptable for a system on which one wants to depend.

Using multiple shots to improve confidence is only a good strategy if the effectiveness of each shot is uncorrelated. But if there is a problem common to all the kill vehicles—a “common-mode failure”—then if one fails due to that problem the others are likely to as well. In that case firing more interceptors doesn’t necessarily increase the kill probability. So far at least two of the failure modes of the interceptors have been common, such as the problem with the CE-IIs inertial guidance system in FTG-06 and the problems with the batteries uncovered in last summer’s FTG-07 CE-I test.

Posted in: Missiles and Missile Defense Tags:

Support from UCS members make work like this possible. Will you join us? Help UCS advance independent science for a healthy environment and a safer world.

Show Comments


Comment Policy

UCS welcomes comments that foster civil conversation and debate. To help maintain a healthy, respectful discussion, please focus comments on the issues, topics, and facts at hand, and refrain from personal attacks. Posts that are commercial, self-promotional, obscene, rude, or disruptive will be removed.

Please note that comments are open for two weeks following each blog post. UCS respects your privacy and will not display, lend, or sell your email address for any reason.

  • nikkkom

    > But if there is a problem common to all the kill vehicles—a “common-mode failure”—then if one fails due to that problem the others are likely to as well. In that case firing more interceptors doesn’t necessarily increase the kill probability.

    Wrong.
    The phrase “the others are likely to as well”, in particular word “likely” instead of “definitely” means that other interceptros are not guaranteed to fail. In other words, they also *can succeed*. That raises interception probability.

    In general, your constant attacks on missile defense programs, even when they succeed, show that you have a strong bias. You already made up your mind. No matter how well it works, it’s “not good” for you.

    Remember Reuven Pedatzur, the analyst who was absolutely sure that Iron Dome can’t possibly work. Incredibly, he maintained his position even after Iron Dome demonstrated hit probability in excess of 87% in real-world situation.

    • M J

      No, the good doctor was right. She did not say the interceptors were guaranteed to fail, only that common mode failures make it harder to compensate for the low kill probability by firing more. If the 1/3 ratio were due only to independent factors, then firing additional missiles would raise the probability of killing all incoming targets a LOT faster than it would in the presence of common mode failures.

      Nor do her “constant attacks” prove a bias. Rather, you have got it completely backwards: for forty years now, proponents of the systems have been the ones with the bias, consistently bamboozling people into believing the systems were far more effective then they really were.

      Nor does belief in their continued failure mean a bias. It could instead be, for example, a recognition of what a very difficult problem missile interception really is. Progress has been made over the last 40 years, but the failure rate is still distressingly high. Possibly not because the military contractors are so thoroughly corrupt, but simply because it is such a very difficult problem.

      • nikkkom

        > No, the good doctor was right. She did not say the interceptors were guaranteed to fail

        I fairly sure she did say exactly that:

        “In that case firing more interceptors doesn’t necessarily increase the kill probability”

        What part of “does not increase probability” is not clear?

        • M J

          I should be asking you that question. For it is clear enough to me that “does not increase probability” is not the same as “the interceptors are guaranteed to fail”.

          Besides: she did not even say “does not increase probability”: she said “doesn’t NECESSARILY increase the kill probability”. But that is not the same as “the interceptors are guaranteed to fail”, either.

          You are not batting 1.000 here.

          • nikkkom

            > For it is clear enough to me that “does not increase probability” is not the same as “the interceptors are guaranteed to fail”.

            Yes, it is the same. In order for probability to not increase, second interceptor must fail in 100% of cases. If it fails only in 99% of cases, then probability DOES increase. It’s a simple mathematical fact.

          • M J

            No, it is not a “simple mathematical fact”. If the first interceptor failed because of the same common mode failure, AND that common mode is such that another missile fired at nearly the same time will fail in exactly the same way, then the probability of a kill does NOT go up, the second interceptor COULD fail in 100% of cases when the first one fails.

            This could happen, for example, if the logic for the radar fails to lock quickly enough, so that both missiles get the same wrong data for the direction to go. Or if there is a arithmetic error in the radar data, so that it does not start giving correct data until the position changes enough that it no longer encounters the same arithmetic error (e.g. numericial instability).

            These are, of course hypothetical examples. There are many other ways for a common mode failure to happen. But as long as the common mode failure correlates the probability of error, then just as Dr. Grego said, firing another missile might not increase the kill probability at all. If the common mode failure means getting it into a ‘stuck’ state where error is certain (as described in my hypothetical examples), then the probability does not increase at all.

            What she perhaps should have added is that even if it does increase the kill probability, it will increase it substantially less than it would if the probabilities were uncorrelated (as they should be).

            In fact, when you say “simple mathematical fact”, what you are forgetting is that the probabilistic model for failure is only an approximation. The approximation breaks down particularly badly in the presence of state dependent common mode failures. For that means that you have don’t even have an identifiable probability measure on an identifiable event space. But then it is misleading to talk about “kill probability” at all. Now THAT is a simple mathematical fact.

          • Laura Grego

            nikkom, MJ is ably covering the waterfront, but pls note that I wrote more about this topic in a post last Tuesday.

    • J_kies

      Nikkkom; the good Dr. was actually too nice to the GBI –
      A common mode failure experienced as a design flaw should certainly fail consistently, such consistent behavior under stimulus is the basis for expecting your design processes/design commonality provide confidence in the event of success. If a software algorithm choses the wrong behavior based on a scene presented to the interceptor, then we would expect that wrong choice / wrong behavior to occur consistently for all vehicles shown that same scene.