Originally Posted by
theamrad
Ok then, AirRabbit, to go back to my use of the word moot - I think you misunderstood me - I'm not saying the report is relegating that point. I'm saying that there is enough information within the report itself to call into question some of the assertions which may be made. Specifically, the FDR acquisition rate and the boards own acknowledgment that the vertical simulator(due to it's own limitations) could not replicate the full range of motion of the accident airplane. This, in my opinion, places at least some doubt on the accelerations and therefore pilot perception of accelerations felt during those simulator tests.
Hey
theamrad
While I believe I understand your explanation about the use of the word “moot” when you described the PF “airmanship,” the determination as to whether or not the PF was exercising “good” or “not so good” airmanship comes down to whether or not we can determine what the PF did; and its almost irrelevant as to “why” he did what he did (and I DID say “almost”) – the only thing that counts is did he do it or not. I believe the FDR supplies abundant information in that regard. However, I believe you are saying that there may be additional information that is NOT available from the CVR and FDR, and that missing information may call into question the decision we make about the PF airmanship. I also believe you are saying this because of the Board’s report that says the following:
The VMS was not able to replicate the target longitudinal, lateral, and vertical accelerations for short periods of time during the accident sequence, and the entire sequence could not be replicated in a single series of motion because of limitations in the VMS motion system. Further, it is possible that small differences between the actual acceleration and the derived values (that is, the acceleration between data points) may exist because of the low sample rate of the pitch, roll, and heading data recorded on the FDR.
OK. I understand all that. Now, here’s the “but.” While I DO understand why you might feel this way, there are a couple of things about which I feel compelled to remind you.
The first is about motion systems on flight simulators. Yes, the VMS set-up at NASA Ames is the world’s largest motion-based aircraft simulator – and it IS impressive; and, yes, I have flown it. However, all simulators with motion systems, even the Ames VMS, are able to only provide “on-set” cueing to the pilots. While it is true that the VMS is able to provide a somewhat longer “on-set” cue than a traditional airline simulator with which you are no doubt familiar – it’s not as much as you might think. Getting a
sustained motion cue in a simulator would require at least something on the order of a centrifuge – but they have problems on the other end. It’s not very realistic to be “parked at the gate” and experience 1.3 g’s! The reason that motion-based simulators can only simulate “on-set” cues is that the acceleration (either linear or angular) must be dampened out and the system re-set to the neutral position to be ready for the next motion cue before the system reaches its maximum extension capabilities. This process must be done below the pilot’s threshold of recognition, or he will be “fooled” into thinking that the airplane is moving in the opposite direction. This is because the simulator IS, actually, slowing down first – and then it IS being moved in the
opposite direction. However, if it is done correctly, the pilot will
not be aware that this is happening. So, while the VMS is a monster of a machine – and it
can move a substantial distance in communicating the “on-set” cue – it really has its work cut out because it has to reset whatever distance it has moved and it has to be done quickly, quietly, and below the pilot’s recognition threshold.
The second is about the additive, or complimentary, effect that a visual system has in relation to a motion system. Airline training simulators have rather sophisticated visual systems – most of which provide something like a horizontal field of view that is 150 degrees and many go up to 210 to 220 degrees, and a vertical field of view that is at least 40 degrees and many go up to over 60 degrees. When the motion system provides an “on-set” cue that the pilot “feels” in his butt, and what he sees in the visual system out in front and along side his aircraft confirms this motion on-set, it is the visual system that continues to reinforce the “motion” cue – even when the motion system movement is slowing down and re-setting to neutral.
The next, and also a very important, factor that you must understand about the use of the VMS simulator cab for these tests is found also in the text of the Accident report. It says the following:
Regarding the second notable event experienced in the cab (that is, the second encounter with wake turbulence), the participants generally agreed that “very slight” cab motions were felt before the first movements of the control wheel and rudder pedal to the right. The cab motions were described as “barely perceptible” left lateral accelerations. Most participants did not experience any cab motion until less than 1 second before the first wheel motion. The first movements of the control wheel and rudder pedal to the right were “large and abrupt.” The participants did not observe a visual or (detect an) acceleration cue that would cause a pilot to apply the observed initial magnitude of wheel and pedal in response to the second notable event. After the first movements of the wheel and pedal to the right, large lateral accelerations were felt, and additional large, abrupt flight control movements in the yaw, pitch, and roll axes were observed. Although the participants felt lateral accelerations, they indicated that it was difficult to sense whether vertical and longitudinal accelerations were also present.
The motions described here are well within the “on-set” cueing capability that a motion system is designed to provide. But, here, as you can surely read, the test subjects all indicate that the simulator cab motions were “barely perceptible” and didn’t start until less than 1 second before the first wheel motion (aileron control).
This warrants a discussion about the “accuracy” of the data used in the simulation as compared against the data that was recorded on the FDR. I can, if you desire, provide you with information about the accuracy of the data with which simulators are programmed – but, I suspect you probably don’t need that reassurance. The data collected by the FDR is certainly not in the same “league” as far as the frequency of data samples taken for things like aircraft certification is concerned. The FDR on this accident airplane took data at either 1 or 2 hertz (Hz), that is once or twice a second, depending on the data. Cockpit controller deflection information, I understand, was retrofitted in accordance with a recent FAA regulation change to be collected at 1Hz, but the remaining information was collected at 2Hz. This particular rate of data sampling is not very fast when compared against data collection during aircraft certification, for instance, where data sampling rates can be seen anywhere from 50Hz to 200Hz. The higher the rate of data sample collection, the more accurate a mathematical model may be that represents that particular function. The more data points in a given time, the smoother the curve is the result.
However, having said this, it is important to note that the “accuracy” of any individual data point – regardless of the frequency with which the preceding or succeeding data points were collected – is not in question. In this particular case, the issue has to do with the movement of the airplane. It would have been nice to collect that data at 50Hz, fifty times a second. That would have allowed a relatively smooth curve to be drawn between two specific times on a graph. But the issue here is whether or not it is likely that the airplane may have moved (or been moved) and then returned (or been returned) to essentially the same state between data samples taken by the FDR. The NTSB report clearly states that the time between the two data points (either 1Hz or 2Hz) is not accounted for – but that is one or one-half of a second. We’re discussing a rather large airplane. Large in mass. Large in momentum. It would take a rather serious force to move that mass any distance in one-half, or one-quarter, of a second and then another sizable force, operating exactly opposite to the first one, in exactly the opposite direction, also in one-half or one-quarter of a second, to bring the aircraft back in line with the other data points that were continually being taken at once or twice a second; and the distance moved would have to be quite small. Not only is this highly unlikely, it borders on the impossible. That is the reason that the Board reached the conclusion that:
The cab motions were described as “barely perceptible” left lateral accelerations. Most participants did not experience any cab motion until less than 1 second before the first wheel motion. The first movements of the control wheel and rudder pedal to the right were “large and abrupt.” The participants did not observe a visual or (detect an) acceleration cue that would cause a pilot to apply the observed initial magnitude of wheel and pedal in response to the second notable event.
AND
The external moments and wind gusts fully accounted for the effect that the wake turbulence had on the airplane, as recorded by the FDR. Any additional effects of the wake turbulence that were not recorded on the FDR (because of limited sampling rates, for example) would have had a negligible impact on the airplane motion.
Originally Posted by
theamrad
(W)e already know from the board's own tests (actually airbus's ground testing) that "test subjects" had difficulty making less than full rudder deflections when asked to do so. And also, IMO crucially, that "most" of the boards own "testees" used some rudder with aileron as their own preferred method of recovery in simulator tests. This particular point was also made during the public hearings, and not only by AA pilot sources. I think that should have been foremost in the thinking of the board when drawing their conclusions. I don't believe there was inadequate follow-up in the course of the investigation, personally I can't see a problem there. But what I'm baffled about, are some of the conclusions drawn by the board, and the order in which they were presented in the way of cause.
Well, actually, that isn’t completely true. There were 2 tests conducted using simulators. The first was a “Control Force and Control Surface Displacement Ground Test.” This test was conducted at 4 specific airspeeds: 165, 190, 240, and 325 knots. It was during this test that the pilots, when asked to move the control wheel and rudder to 50 percent of their respective available ranges, and each of the three pilots, while they did use only 50 percent of the force required to get to 100 percent rudder pedal travel, they each moved the rudder pedals to their 100 percent travel limit. If that was all there was to that test, you would probably have a reasonable argument. However, as you would see, the control travel tests were required to be done with very specific
rates of control movements. Those rates were 0.25, 0.5 and 1Hz. That is, achieve the 50 percent control travel limit in one-quarter of a second, one-half a second, and one full second; at each of the prescribed airspeeds. Quickly moving the controls will usually provide for more control input being reached than intended. That is exactly what happened here. So, it wasn’t that the pilots could not determine where 50 percent control travel was, as much as they had trouble stopping the control travel at 50 percent when they were to move the controls to that point at the rates described. That’s a tough test.
The second test using a simulator was a simple “Training Simulator Study.” Here there were 6 trials using the A300B4-600 flight simulator. In the first trial, the simulator was inhibited in its response (to allow the airplane to reach a substantial bank angle before recovery began). While in a 20-degree left bank, at an altitude between 2000 and 2500 feet, and an airspeed of approximately 240 knots, the computer input a roll in one direction (the direction was computer-determined in accordance with an arbitrary determination routine) followed immediately by a substantial roll in the opposite direction. The pilots were instructed to recover in accordance with the ground school training. This procedure was repeated 5 more times for each pilot, except that the roll maneuver was initiated during level flight after the pilot indicated he was “ready.” Each pilot was instructed to follow one of 5 specific recovery methods: 1) partial wheel and no rudder; 2) full wheel and no rudder; 3) full wheel and partial rudder; 4) full wheel and full rudder; and 5) the pilot’s preference. It would be a waste of time and space to reprint the complete results here (you can see them in the official report at section 1.17.1.2.6 Training Simulator Study), but each pilot reported the worst method was partial wheel and no rudder. The last trial (pilot’s preference for recovery method), most of the pilots responded with nearly full control wheel and partial rudder pedal inputs. Slightly less input was made on both controls than during the first trial, and the pedal response was typically delayed by at least 1 second after the control wheel response.
I would submit that the results of these simulator tests do little but confirm the rest of the Board’s report and are a far cry from what you believe the test results show.
Originally Posted by
theamrad
The only further thing I can add is my opinion of what might have happened. Whether or not there was some adverse yaw due to wake turbulence is a matter which I or (I believe) the board cannot ascertain with absolute certainty. While many may say the use of rudder in nothing more than a wake turbulence encounter is wrong - it is a fact, established during simulator testing the most of the test subjects DID use rudder when recovering from and UPSET. I can only say that it's possible, through the slow FDR aquisition rate, that PF perceived a roll or yaw rate which was not picked up by the FDR, and which he thought warranted use of rudder as well as aileron. Whether right or wrong (and I think PROBABLY wrong) he got a whole lot more than he intended, thanks to the oversensitivity of the rudder, and ended up with a (severe) apc event. I can't say anymore about the failure to recognise this situation and unload the controls - but then that's a matter that's been discussed many times in other places.
Of course no one is going to say that anything is ever “ABSOLUTELY” determined one way or another. There is always going to be someone, somewhere who has doubts about the findings.
The rate of FDR data acquisition is irrelevant unless you believe that the airplane could have been doing something really “screwy” between FDR captured data points and return to the same state every second or every one-half second.
Yes, most of the pilots in the simulator tests DID use some rudder in recovering from the upset to which they were subjected. However, the simulator was inhibited for a while to ensure a really significant bank angle was achieved. The report says beyond 100 degrees on average. None of the pilots used full rudder except when they were asked to – and none of the pilots reversed the control inputs.
Did the PF on the accident airplane get “more than he intended” thanks to the “overly sensitive rudder?” Perhaps. But why was he using rudder in the first place? He didn’t use rudder input during the wake encounter merely 12 seconds earlier. Why? The two encounters were very close, if not identical, in effect on the airplane.
Did the PF get into a Pilot Induced Oscillation (PIO)? Yes, I think he did and he didn’t know how to get out of it.
In my estimation, on the basis of logic, reasonable doubt, the diligent and careful manner in which this particular accident was investigated, and the voluminous amounts of raw and refined data, I think the Board’s report on this one is pretty accurate.