The Track Gate Anomaly: Does it also Affect CE-I Kill Vehicles? (August 15, 2014, revised 8/18/2014)

On Wednesday August 13, at the Space and Missile Defense Symposium, MDA Director Vice Admiral James Syring was one of the featured speakers. As Admiral Syring started to go through his slides, Aviation Week reporter Amy Butler began to photograph the slides. Even though the the slides were clearly marked “Approved for Public Release,” meeting organizers quickly stopped her, saying photographing of the slides was not permitted. However, before she was stopped she managed to photograph twelve of the slides and subsequently posted them on twitter (she tweets as @ABAviationweek). You can read her account about the slides and her (sucessful) attempt to ask Admiral Syring a question here:

The twelve slides that were posted are actually more detailed and interesting then those in your typical MDA briefing. Hopefully they will all soon be publicly available.

The slide that struck me as the most interesting was one titled “Track Gate Anomaly (TGA)” shown below:

TrackGateAnomaly

This slide reveals that the “Track Gate Anomaly” was the cause of the failed intercept test FTG-06a in December 2010. FTG-06a was the second consecutive failure of the new CE-II version of the Exo-atmospheric Kill Vehicle. This failure had serious consequences for the GMD system, forcing MDA to halt deliveries of new kill vehicles for about three and half years. This TGA problem has apparently been known to DoD insiders for many years, but this slide is the first time I have ever seen the TGA equated with the problem that caused the FTG-06a failure. [See the comments following this post for an explanation of what a “track gate” is.]

According to Syring’s slide, the TGA problem was first detected in intercept test IFT-06 in 2001, and it was seen in eight tests over nine years. In the period between IFT-6 in 2001 and FTG-06a in 2010, there were a total of nine GMD intercept tests where it would have been possible to observe this TGA (there were also three other tests in which the interceptor did not launch or the kill vehicle failed to separate from the booster rocket). Thus it appears that there was only one intercept test between 2001 and 2010 where the TGA problem did not occur. Syring’s slide also states that while MDA initially believed the TGA problem was caused by electromagnetic interference, by 2005 they had assessed it as a vibration-related problem. This was before the first CE-I intercept test, which was in September 2006 and more than four years before the first CE-II test.

It also now seems clear that this TGA is the “Pervasive, Enduring Problem ~ Not a ‘Glitch” cited by Philip Coyle, a former Director of Operational Testing and Evaluation, at a 2008 Congressional Hearing.[1] According to Coyle’s prepared statement:

“The GAO is describing an anomaly that has persisted in GMD flight intercept tests for seven years, since 2001. If not corrected this anomaly could cause the EKV to temporarily lock onto the “wrong” target” and miss the real target.”[2]

Note that Coyle’s statement was made in April 2008, more than a year and a half before the first flight test of the new CE-II kill vehicle.

Even earlier, in 2004, Senator Carl Levin similarly raised the issue of the TGA possibly causing the EKV to miss its target when he asked (in a written question):

“General Kadish, the DOT&E, in his Annual Report to Congress submitted in January of this year, noted that a ‘‘track gate anomaly’’ in the kill vehicle had existed in intercept tests IFT–7, –8, and –9. I also understand that repeated attempts to fix this problem have so far proved unsuccessful, and that this problem could cause the interceptor to miss its target by losing track. Why didn’t you continue intercept tests at least until this known ‘‘track gate anomaly’’ problem was proven to have been fixed?”

The occurrence of the TGA problem in tests FTG-6 through FTG-9 is particularly signifcant because the EKVs used in these tests used the same IMU as was used in the subsequently deployed CE-I equipped interceptors. After both intercept tests of the CE-II equipped interceptors failed in 2010, MDA officials had reassured questioners that the GMD system could still be relied on to defend us, because twenty of the thirty interceptors deployed use the older CE-I version of the EKV, which they said did not have this TGA problem. For example, as the MDA Director Lt. General Patrick O’Reilly told the Senate Appropriations Committee on April 18, 2012:

Senator COCHRAN. “Could you explain what contingency plans we may be developing to provide homeland defense if there are test failures?”

General O’REILLY. “Sir, the problems we’ve had in flight testing, and we’ve had two failures, were with the latest version of the front-end of the missile, the EKV. The older EKV is deployed today. It’s been successful in five tests (three intercept tests and two other flight tests). We have never seen any indication of a problem on the ground with the older EKV.”

It now appears that this statement, while it may be technically correct, is misleading. In particular, the claim that MDA has never seen “any indication of a problem on the ground” with CE-I EKVs may be true, but that is only because the ground-based testing equipment at that time was incapable of reproducing the conditions under which the TGA occurred. In April 2011, MDA Director Patrick O’Reilly told Congress about the FTG-06a failure:

“In our latest GMD test, we did have a failure mode that could not be replicated on the earth. And that’s why I’m going to request an additional test to verify we fixed it.”[3]

But Syring’s slide makes it clear that MDA were in fact seeing this problem repeatedly with older EKVs (CE-Is and earlier prototypes) in actual test flights.

It may the case that the newer CE-II kill vehicle is more vulnerable to the TGA because its inertial measurement unit is more sensitive than the one in the CE-I EKV. But Coyle’s and Levin’s statements above makes it clear that years before the problem was identified and fixed, there was concern that a TGA could cause earlier versions of the EKV to fail. There does not seem to be any indication in the public record that MDA has done anything to correct this problem in the twenty deployed CE-I interceptors.   A fuller accounting of this problem seems to be needed if MDA is going to continue to argue that the CE-I EKVs are not also vulnerable to this problem.

I’ll follow up this post with another one containing a few more details in a few days.

[1] National Security and Foreign Affairs Subcommittee, House Committee on Oversight and Government Reform, April 16, 2008.

[2] The GAO Report Coyle is citing is GAO-07-387, p.25.

[3] Lt. General Patrick O’Reilly, Strategic Forces Subcommittee, Senate Armed Services Committee, April 13, 2011.

 

 

Advertisements
Leave a comment

5 Comments

  1. “And that’s why I’m going to request an additional *test* to verify we fixed it.” (I checked.)

    Any clues to what “track gate anomaly” might mean, technically?

    Reply
  2. Thanks, Mark. I have fixed the quote.

    Here’s an attempt to interpret what “track gate anomaly” might mean (actual details may differ):

    As the kill vehicle views the scene early in homing process, typically it will see several objects in its field of view. The EKV then places a “gate” around each object. A gate can be a single pixel or multiple pixels. For example, it might assign object A to a gate consisting of the pixels that are in rows 38-40 and in columns 82-83 on the EKV’s 256×256 infrared detector array. The gate should be large enough so that in the next image frame Object A will still be within the same gate, taking into account changes to the gate position based on measurements of the EKV’s acceleration or rotation by its inertial measurement unit (IMU). Thus using these gates allows the EKV to correctly associate objects from one frame to the next. In particular, once the EKV has designated one of the objects as its target, it can use the gate around the target to maintain its track on it as it homes in.

    (If you look at intercept video on MDA’s website for either GMD or Aegis intercept tests, you will frequently see video from the EKV that shows several objects, each with a box of various color around them. These boxes presumable represent the gates.)

    My guess is that the “track gate anomaly” apparently occurs when the IMU incorrectly interprets vibrations from the EKVs divert thrusters as showing, for example, a rotation of the EKV. Such a rotation would change the apparent location of the objects, so the EKV adjusts the location of the each objects’ gate accordingly for the next frame of data. However, because this adjustment is based on false data, the gates are moved to the wrong places. If this error is large enough, then in the next frame when the EKV looks in the target object’s gate, the gate is empty, causing the EKV to lose lock on the designated target. It then has to go through the target designation process again or perhaps even needs to begin a new search if some of the objects are now outside of its detector’s field of view. If this anomaly occurs early enough in the homing process, the EKV may still have time to reacquire the target and hit it. This is likely what happened in the CE-I and earlier intercept attempts. But if the anomaly occurs late in the homing process and/or is unusually large, it can be fatal.
    The CE-II EKV apparently incorporated a new, more sensitive IMU, which apparently increased the vulnerability of the EKV to this anomaly (for example, a more sensitive IMU might lead the EKV to use smaller gates).

    This issue could be much more problematic in a countermeasures environment where there are numerous closely-spaced objects. In this case such an anomaly might end up shifting gate on the target onto another object.

    George

    Reply
    • Thanks, George. I understand it’s a best guess, although you may have heard or read some supporting bits. I’d bet on your interpretation of “track gate” as the window for tracking, or “area of uncertainty” as it’s called in other contexts. I wonder why in case of failure to find the target in the next frame the algorithm would not be to just progressively widen the gate until it’s found. It’s also not clear to me why vibration effects would be interpreted as a rotation, especially if there is an IMU that presumably tracks rotation – perhaps vibration effects on the IMU itself? Also, it isn’t clear in this scenario why the effects could not be observed in hover tests.

      On the other hand it is clear why the possibility of losing track on the target and shifting to a CM object can’t be observed in tests. Because what do you mean, “countermeasures”?

      Reply
  3. Mark,
    I only emphasized rotation over acceleration because a very small rotation can have a big effect on the observed position of objects. For example, if the seeker field of view is 1 degree, then a single pixel is just 0.004 degrees, so it wouldn’t take much rotation to produce a big shift in target location. On the other hand, in most circumstances, a small acceleration would not produce much of a shift in the target’s angular position. So that’s why I emphasized rotation.

    I agree that simply expanding the size of the gate might be the first thing you would try if you lost lock. This might not work in some situations, particularly if the shift in position is large. Bear in mind that in the eight cases that TGA was observed, the EKV recovered in time to hit the target in seven of them, so in many situations it can handle this problem. On the other hand, the problem was serious enough even in the earlier intercept tests, that people who should know were concerned that it could cause a miss.

    A key question here is whether they just got unlucky on FTG-06a, and got a unlikely situation the EKV couldn’t recover from (such as the TGA occurring very late), or if the CE-II kill vehicle was really much more vulnerable to this problem (with the higher closing speed in FTG-06a also possibly a contributing factor. No way to tell with evidence we have now.

    Don’t know why hover tests did not reveal problem. However, it is not obvious that hover tests would involve the kill vehicle using its maximum divert thrusts, given they are likely conducted in pretty confined spaces. Also if hover test results in expending the test EKV (I don’t know if they do), then they would get to be very expensive (I think a CE-II costs about $39 million).

    Yes, the problem is apparently directly due to vibration effects on the IMU itself.

    George

    Reply
  4. j_kies

     /  August 29, 2014

    Gentles, thanks to VADM Syring’s chart, TGA is now in the public record as to historic extent.

    In my opinion, per Mr. O’Reilly’s testimony on that date and topic; Congress would have been fully within its rights to hold him in contempt for knowingly lying to Congress under oath. (Alternatively, if it was unknowing falsehoods then he did not have technical or programmatic competence to execute his duties.)

    The MDA has speculated as to ‘root’ cause of TGA and come to different conclusions after every flight where it occurred. Some mitigation efforts were applied to address attributed causes in several cases. Others have disagreed with the MDA “failure review board” process and results that amounted to citing a failure hypothesis and then looking for evidence to confirm that mechanism. (The usual term of art is ‘selection effect’) Raytheon publically confirmed that process as their standard in their claims of discovering ‘new physics’.
    —-
    6/25/2014 Inside Missile Defense
    “Kremer offered a few reasons why the intercept tests of the CE-II variant were separated by about four years. “We essentially found a new physics phenomenon and so, in order to, one, prove that, took a lot of effort,” he said.

    Once Raytheon was able to prove the phenomenon, Kremer continued, “we then had to develop a ground test environment that would simulate that condition that we saw in outer space, and so it involved not only designing a materiel solution to address the problem, but it also entailed the design of an entirely new test sequence including test procedures and actual test equipment to be able to do that.”
    —-

    In these public statements, I see no evidence of either application of the scientific method or diligent failure review and replication of failure followed by repair with a sufficiency of instrumentation to validate design margins. Congressional testimony by both O’Reilly and Syring attributed shock / vibration from thruster firing as causing the IMU to miss-report. For Raytheon to ‘discover’ how to conduct thruster hot firings on the ground or to ‘design an entirely new’ means to measure structural transfer of shock and vibration environments to an IMU isn’t something that represents either new physics nor state of the art of missile engineering since the 1940s.

    Certain professionals have disagreed with MDAs claims in each case cited and the key question that should be put to the GMD team is “did you replicate the failure mode when re-applying the same environments as experienced in flight”?

    Reply

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: