Tuesday, August 2, 2016

LCS Harpoon Failure

We have a report that a Harpoon test fire from the LCS USS Coronado failed during a RIMPAC 2016 SINXEX (1).  The missile apparently launched successfully but then disappeared.  The Navy is investigating. 

Let’s get this part out of the way.  I do not link the Harpoon failure to the LCS.  The Harpoon launched from the Mk141 rack launcher which is a self-contained, modular launch system that was, at one time, ubiquitous throughout the fleet.  This appears to be a pure Harpoon failure.  Plus, the Harpoon and launcher were not integrated into the LCS combat system, anyway, so the LCS seems an unlikely cause of the failure.  Still, the LCS program can’t seem to catch a break, can it?  But enough of that …

You’ll recall that the Navy’s inventory of Harpoons is old and most have been pulled from active duty in order to extend the life of the remaining stockpile.

This appears to be a simple failure associated with complex weapons.  When you consider all the components of a missile (motors, sensors, circuit boards, flight control surfaces, etc.) that must function perfectly for overall success, you immediately realize that each component has a failure probability and that all those probabilities added together (statistically, it’s actually the product of the individual probabilities that determine the overall probability) determine the overall probability of success.  Thus, statistically, there is a probability of failure for every missile (and every weapon system, for that matter).  In other words, a certain level of failure is inevitable.  Of course, the manufacturer and the Navy claim that the probability is almost non-existent while actual test data, such as this, demonstrate that the probability is significant.  For what it’s worth, my overall assessment of Tomahawk/Standard/Harpoon missiles is that 10%-15% will fail to guide.  That’s fine, as long as we have a realistic understanding of the failure rate and compensate for it by having and launching a few extra missiles.


Harpoon Launch Failure


So, what’s the point of this post?  It’s testing, of course.

It’s not enough to simply conduct tests when a weapon is first introduced.  That’s good, and madatory, but we must continue to test weapons throughout their service life, especially as the weapons reach the end of their shelf lives.  I’m guessing that the Harpoon failure rate has doubled by now, compared to when the missile was first introduced.  Again, that’s okay as long as we test and know the failure rate as time passes and the rate changes.

Has anyone heard of reliability testing of Harpoon in the last 5-10 years?  I haven’t. 

I’m just speculating about the failure rates.  The Harpoon failure rate may be 70%, for all I know!!!

We’ve seen how hard it is to get the Navy to conduct proper testing when a weapon is new; how much harder and rarer is it to get the Navy to conduct proper testing of established weapons?  And yet, these are the weapons we’ll go to war with.  It’s vital to know how they’ll perform and how they are holding up over time.

Solitary live fire tests such as this one demonstrate nothing.  That failure could be the unlucky one in ten thousand failure or it could be the common one in three failure.  Without a statistical test, we don’t know – and that’s both the danger and the point of this post.

Someone is sure to bring up the cost of testing and that’s just idiotic.  Say we test two dozen Harpoons.  That’s $24M or so worth of missiles.  That’s a vanishingly small drop in the budget bucket and if it saves a single ship in combat by having a better idea of how our weapons perform, it’s completely worthwhile.


________________________________

(1) Popular Mechanics website, “Navy's Harpoon Missile Misses Target During Test Fire”, Kyle Mizokami, 21-Jul-2016,


34 comments:

  1. " each component has a failure probability and that all those probabilities added together (statistically, it’s actually the product of the individual probabilities that determine the overall probability) determine the overall probability of success."

    Statistically, it is actually the product of the probabilities of non-failure of the individual components that gives you the overall probability of success. The probability of failure is, of course, 1 - P(success).

    Ah, the pleasures of sweet pedantic accuracy.

    ReplyDelete
    Replies
    1. Don't be a pedantic pinhead.

      Delete
    2. Can't help being pedantic (= getting the details right). Microcephalic, I'm not.

      Seriously, I agree with everything else you said, and it needed saying.

      If statistical pedantry had been applied to the RN's testing of armour piercing (or more accurately, sometimes piercing) shells pre-1916, Jutland might have gone very differently.

      I really wouldn't want to see the USN have a similar problem.

      Delete
    3. That is assuming that the mission kill failures are independent events when in reality Bayesian probability could be more appropriate.

      When we retire the harpoon or if we are about to deactivate specific harpoons at the end of their service life, it would cost almost nothing to test the failure rate.

      Delete
    4. For a practical use, the various failures can be considered multiple independent events. Could there be failure relationships? Of course. A circuit board may have a higher failure rate when subjected to acceleration, for example. However, for general discussion purposes, multiple independent events describes the situation well enough. Besides, no one has any data on failure relationships!

      Delete
    5. The USN needs to test a large enough sample size to get an acceptable confidence level on the reliability of Harpoons.

      They must be truly random tests too in conditions resembling combat.

      Delete
    6. "They must be truly random tests too in conditions resembling combat."

      Exactly right! A missile might launch and guide 100% of the time in benign test conditions but fail every time in an ECM environment. The US WWII torpedo example is instructive and yet we ignore the lesson.

      Delete
  2. I agree with you. But I wonder if this is a side effect of the cost of modern naval wonder weapons.

    While you correctly point out that testing should be made part of the price (even if it means you get fewer missiles) I wonder if the powers that be don't want to know, and don't want to deal with fewer weapons, because so much money is tied up in this.

    If testing showed that there is a 10% failure rate, that would be embarrasing, and run counter to the 'Everything is awesome' theme song that the Navy likes to play during acquisition.

    Further, the Naval accountants who regularly complain that there isn't enough money may quail at, literally, blowing up $24 million in goods.

    I also wonder if this is just limited to the US. Does the Brahmos get tested regularly? DF-21? These things are cool, but expensive.

    Its easy to line up a Leopard II or an Abrams and fire the gun all day long. But those are guns. Compared to a missile...

    Again, while you are correct, I imagine its much harder for people to convince those in charge, who regularly poormouth their budget, to fire off a few $3 million Standards.

    It does make me question everything.

    If we aren't testing Harpoons, Tomahawks, and Standards (!!) regularly, what are the chances that when we need them we end up with a WWII Torpedo situation.

    ReplyDelete
    Replies
    1. "If testing showed that there is a 10% failure rate, that would be embarrasing ..."

      It's embarrassing only if you lie about it. We all understand that electromechanical devices have a failure rate. TV remotes fail. The toaster fails. Our cars break down on occasion. The consumer doesn't expect a 100% perfection rate unless you lie and tell them to expect it.

      If the military would be honest and say that we only have a 90% successful guidance rate on missiles, no one would be upset especially if it's combined with a manufacturer's warranty so that failed launches get a free replacement missile.

      All too often the military sets itself up for failure and then blames the critics!

      Delete
    2. Somewhere is the specifications for the Harpoon missile the reliability of the missile is defined. The question then is if the supplier is meeting those requirements.

      There is also the possibility that the Navy didn't properly store and maintain the missile that failed. If that's the case, Boeing shouldn't be held liable.

      Delete
  3. Your failure estimate seems high.
    These aren't war stocks being shot, these are the retired from service missiles, that are being shot off in training, for training purposes, and are past their used by date.
    In all honesty, unless either Russia or China get well and truly rambunctious, i doubt we'll ever know the real failure rate of the US missile armoury. Im fairly sure the stuff being used in Afghan and Iraq are older supplies to...

    Eg of a western army using modern weapons and failure rates, would be 2006 Lebanon war, where in the final week an inordinate amount of precision weapons were used. The reason being they were all approaching or just past their used by date. It'd give you a more realistic fail rates of modern guided weapons,
    That is, if you could get the IDF to release its figures ;)

    ReplyDelete
    Replies
    1. The best data I've come across is Desert Storm. There are many books and lots of data available. A 10% failure to guide seems to be about the consensus for the Tomahawk missiles, which we used a lot of so the data is pretty solid. This is a pretty reliable estimate.

      My data on Standard missiles is very sketchy but what I've been able to glean supports a 10% failure rate. Ditto on Harpoons.

      Have a little faith. I don't just make these numbers up. If you're going to read this blog, do so with confidence that my various numbers and data are reasonably solid. I generally cite my sources or I make clear when I'm just speculating. If you can find better data, I'm completely open to it. On the other hand, an unsupported statement like "Your failure estimate seems high." carries no weight whatsoever unless you back it up with data or qualify it as unsupported speculation (nothing wrong with that!).

      Up your game!

      Delete
    2. I think Russia's Syrian cruise missile strike had a failure rate of a little over 10%

      And that just counts the missiles that were found to have ditched in neighbouring lands.

      Delete
    3. Sure thing CNO

      IDF sources sighted in the 2014 gaza conflict, that Iron Dome Tamir interceptors were at or above 94% hit rate.
      So, Israeli missiles, sure, but, more realistically, these are the result of US missile defence design and manufacturing.

      So, there, a western missile system, recently tested, using new built missiles, punching at well above 90%.

      Delete
    4. I've seen the Israeli data and I have severe doubts about it. I've seen reports that suggest a much lower success rate (40%-70%). The Israeli public data, I suspect, is a public relations point to reassure the populace.

      The Israeli data also seems to be highly qualified meaning that it is only considering the missiles that actually guided (a "dud" that failed to even launch wouldn't be counted in the data). Further, it seems to be grouped meaning that if three missiles were fired at a given object and only one hit, that would be considered a one "shot" success (100%) rather than a one out of three (33%).

      I presume you're not so naive as to believe government/military claims at face value? Every claim the US military makes is "spun" to look good. Israel is no different. I know that you, as an astute observer and keen analyst, dig deeper into military claims and would not simply parrot public relations claims in an effort to try to prove me wrong about something, right?

      You'll recall that immediately after Desert Storm, the military was aglow with claims of 95+% accuracy with laser guided bombs and Tomahawks? Later, more careful and objective analysis pegged the LGB performance at numbers around 70% and Tomahawks at 85% or so, depending on what source you want to use. Similarly, Israel is almost certainly putting out public relations claims to justify their investment and reassure the public. It's up to us, the objective observers, to recognize it for what it is and attempt to analyze it rather than blindly accept it.

      So, lacking a detailed analysis of the Israeli data, I'm simply discounting it until better data is provided (probably never will be - they're not quite as free with their military data as we are).

      Of course, it doesn't matter what Israeli performance is, good or bad. For us, the only thing that matters is US Tomahawk, Standard, and Harpoon missile performance.

      Delete
    5. thats kind of my point bud,
      If you read a little deeper, its a volley of 2 missiles per target when 1 inbounds detected, and an over shoot of 15% interceptors to targets.
      What we see is what we're fed.not what is, if you noticed in my first post, i said, if you could manage to get IDF to release its figures.

      Anecdotally, i can tell you that TOW2's we fired in training, (many dozens) had a little over 80% success rate.
      the dozen war shots were 100% successful.
      Again, not dogmatic, as, one soldiers experience, and cheaper simpler missiles, but you see my point i think?

      Delete
    6. A TOW is almost a bullet, in terms of complexity. It's not even remotely in the realm of the Tomahawk, Standard, and Harpoon.

      I've explained why I discount the IDF data.

      So, I'm left with my original data and conclusions and nothing I've seen yet changes that.

      Also, you appear to be looking at hit rates which is not at all what I'm concerned with. The post was about failure to guide - those types of failure that prevented the missile from guiding and having a chance to hit something. Hit rates are another issue and are dependent on targeting data quality, ECM, enemy maneuvering, defensive measures, etc. The missile can perform its job perfectly and still be shot down, for example. I'm looking at a 10-15% fail to enter guide mode, as stated in the post.

      Delete
  4. So it launched successfully but then disappeared. To be honest, thats kinda vague and leaves multiple possibilities to be looked at including operator error or sensor failure.

    I've seen missile failures (BGM-71, JAVE, & AT4 ) and most of those failures comes down to old age or the operator. I know that in the army, when we have a munition fail to fire or malfunction, that information gets record, and if possible the missile gets returned for inspection. Obviously, you can't recover ship based missiles easily, but there is a electric paper trail to follow along with the last recorded maintenance and inspection paperwork, to firgure out happened.

    ReplyDelete
  5. Anyone remember the MK14 torpedo failures of WW2. One could almost predict the same problems here due to a lack of testing

    ReplyDelete
  6. How many Harpoon missiles does the Navy plan to equip each LCS? Eight is typical, but the LCS is anything but typical.

    ReplyDelete
    Replies
    1. Bear in mind that the Harpoon has not yet been officially selected, as far as I know.

      Delete
    2. I thought they were looking at the Improved Harpoon? So, theoretically, all those would be new?

      Delete
    3. The Navy is looking at all options including the NSM and others.

      Delete
    4. Why not torps?

      Delete
    5. The standard RGM-84 load for an LCS is to be four missiles:

      "Once Coronado returns to Pearl Harbor, [it] will be outfitted with four Harpoon missiles prior to deploying to the U.S. 7th Fleet area of responsibility."

      http://thediplomat.com/2016/08/us-navy-to-install-harpoon-anti-ship-missile-on-littoral-combat-ship/

      That article also says:

      "The U.S. Navy intends to equip its fleet of LCSs with both the RGM-84D Harpoon Block 1C and the fifth-generation over-the-horizon Kongsberg Naval Strike Missile (NSM)."


      Another source also describes the limited nature of these installations. One ship gets Harpoon and one gets NSM.

      "Both the NSM and the Boeing-built Harpoon will be installed on deployed LCSs, the first of which is the Coronado. The LCS Freedom will be fitted with the NSM by the time it deploys this winter to the Western Pacific."

      http://www.defensenews.com/story/defense-news/2016/07/20/lcs-harpoon-missile-rimpac-coronado/87371686/

      Delete
    6. It's highly unlikely that the standard loadout of anti-ship missiles will be 4. That sounds like a developmental load. The standard Harpoon load on every US Navy ship that carries Harpoon is 8. There is no reason to believe the LCS will be different. From a salvo density perspective, 4 missiles is too little to do anything other sink a patrol boat. Again, 8 would be a minimum load, I would think.

      Having said all that, this is still the LCS program and they continue to make one idiotic decision after another, so who knows?

      Delete
  7. " For what it’s worth, my overall assessment of Tomahawk/Standard/Harpoon missiles is that 10%-15% will fail to guide. That’s fine, as long as we have a realistic understanding of the failure rate and compensate for it by having and launching a few extra missiles."

    "So, what’s the point of this post? It’s testing, of course."

    CNO, excuse me. But your two statements actually don't mesh. You're first statement already gave a solution (i.e. shoot 10-15% more missiles to cover that aspect of failure problem).

    Your 2nd statement means "10-15% failure is not acceptable and nor is shooting 10-15% more"; hence the testing. Your 2nd statement only makes sense, if your goal is to knock that 10-15% down to something below.

    Hence, you already provide one solution (with known extra expenditure). Your 2nd statement is to propose another solution with unknown 'scope of work and $$'. Different? yes. Better? don't know.

    ReplyDelete
    Replies
    1. Did you actually read the post? The purpose of testing was explained to be to understand the actual failure rate so that we could compensate and to monitor the change in failure rate over time as weapons and systems age and their failure rate presumably increases.

      Where in the post did I say that a given failure rate was unacceptable and that the point of testing was to lower the failure rate? The answer is nowhere, because I didn't say or even imply that. In fact, I explicitly said the opposite, that a given failure rate was fine as long as know what it is - HENCE THE TESTING.

      I'm not sure how much clearer I can make it.

      Feel free to comment but first read the post carefully.

      Delete
    2. My background is control engineering. A missile/plane/or you/me basically is a feedback (or other fancier control algorithm) loop of sensor/control/actuator. when I read your 'testing' I automatically break the whole thing down to decouple the modules, build testing/calibrating firmware around each module, pass-or-fail them, then reconnect them back together. With this type of testing, module-to-module interface is always the big deal. For example: if the missile sensor only has to interface to missile brain, it is easier than if the same sensor has to have two interfaces build in: to missile brain, and to engineer's bench (i.e. added electronics firmware, footprint, size & weight, and more places to fail). If you split every hair, you'll get a 100% success (from component to subsystem to whole missile), but how much do you want to pay?

      Now back to the missile test. If it was done during the same test shooting at USS Thach, I believe, there were other Harpoon hits (3 success hit?), so if you added this failed one. What does it (or all 3-4 launches) tell me? Nothing much, because the sample size is too small. They can all hit, all miss, or half-and-half, I'd still fall back to your first statement of statistical result of '85-90% success rate'.

      As an engineer, I fully agree with you on testing. However, if I'm the admiral bean counter, I'd ask me (the engineer) $how much$. And the 4-star accountant is going to weight my $testing proposal$ vs. 'extra 10-15% missiles made" to solve essential the same issue: how to shoot enough missiles to complete the mission.

      That's all I'm saying.

      Delete
    3. Seriously, read the post! I addressed your "what does it tell me" question and "nothing much" answer when I said, "Solitary live fire tests such as this one demonstrate nothing." I then went on to say, "Without a statistical test, we don’t know – and that’s both the danger and the point of this post."

      I also addressed the cost issue in the last paragraph. Read it!

      Do not make another comment until you read the post and then compare your comment to the post to see if whatever you want to say has already been addressed. I expect better of my readers.

      Delete
    4. I'm not military. How does Navy test its missile? (As in firework destructive test on the whole package, or de-construct a missile to re-evaluate subsystems/components?)

      I would favor the 2nd approach: take several missiles out of each lot by year. Send them back to missile maker. Ask them to disassemble (to reasonable subsystem/component). Send to each part's OEM. Re-calibrate as is, and arrive the statistic on each's longevity.

      Delete
    5. Tim, I honestly don't know if you're being obtuse or you really don't grasp some of the basics, here. There's nothing wrong with disassembling and testing individual components but the only test the ultimately matters, and the one that must be done, is actual performance testing. The only test of an automobile that matters is to drive it and see if it works when it heats up and starts getting bounced around and stressed. The same is true of missiles. All the component and shelf tests in the world won't tell you whether the entire package will stand up to ignition, acceleration, g-force, heat, electromagnetic interference, etc. I'm pretty sure you understand this so I'm left to wonder what you're really asking and why?

      Delete
    6. 1. Car and missile are different. If there is problem with the car, I can pull over and have it diagnosed and fixed. A missile is gone, whether it works or not.

      2. And perhaps we interpret the word 'test' from different angle. I translate it as 'data gathering and analysis'; perhaps you translate it as 'validation'.

      3. Pretend I'm a missile mfg engineer, the admiral asks me: would it work near end of its shelf life? All I can say is: I got a 10yr old cal sheets says it would 97% of time new, and statistical prediction of 90% at its expiration. We shot the missile in an exercise (not war). If it worked, I won't learn anything new. If it didn't, it will be very difficult for me to find out why. Now, let's look at two different options for follow up

      3a. We take another 24 missiles. Shoot them all. And lo and behold, your nightmare of 70% failed. The sensible follow up after this is to take another 24 missiles, de-construct them, and find out what subcomponent(s) failed.

      3b. Instead of shooting off 24 missiles. I de-construct the first 24 and find out what (can) failed. I believe the diagnosis result, between 3a and 3b, will have correlated findings. You use 48 missiles to find out; I only use 24.

      4. As for 'whether the entire package will stand up to *****'. It is physics and non-destructive validation procedure. If the engineer do a good job of translating and de-constructing 'firework' to various QA procedures. Then the end-result should (and will) work, even without the firework. If it doesn't, then we go back to (3). Now, if the engineer is honest, and honestly real good, and he tells the admiral: there is that '10-15% failure' beyond my realm of knowledge and fix. Then, its up to the admiral how to solve that issue: buy more, or shoot more?


      Delete
    7. This is an unproductive discussion.

      Delete