|
sfrom prior “C” Check
Under Wreckage and Impact Information , on AAR page 7, paragraph 3; the Board’s description of the orientation of the debris disagrees with data presented in the report from the manufacturer, page A-10 Figure 4. The Board must carefully describe the distance and direction between aircraft components found on the ground. This trail of debris is the best evidence available in this accident and can be used to reveal the sequence of some failures sustained by the accident aircraft.
The NTSB report should have shown the trail of debris properly mapped with accurate scale distance and direction between components. In the later sections of their report the Board should fully analyze such a map of the trail of debris and explain the sequence of failures.
Under Wreckage and Impact Information, on AAR page 7, the last paragraph should clearly state that investigators did not check the annunciation for the #7 slat retract locking mechanism. The current wording of the paragraph may mislead the reader to think that the annunciation for the #7 slat retract lock was found to function properly. (See admission on page 24 of the NTSB’s AAR, mid-page: “the #7 actuating and indicating system could not be checked . . . .”)
Under Wreckage and Impact Information , on AAR page 8, the second paragraph misleads the reader to think that investigators had examined the #7 Slat Retract Lock Indicating Switch. Records in the NTSB docket show that post accident examination of the remnant of the #7 slat actuator focused on the fracture to the outer cylinder barrel of the actuator. None of the records in the NTSB docket reflect any functional test nor tear down analysis of the Retract Lock Indicating Switch (illustrated on page 17 of the NTSB Report).
In the docket a letter dated February 20, 1981 from L. D. Kampschror, NTSB Investigator in Charge, to Capt. J. A. McIntyre, ALPA, stated:
The actuator that was on N840TW had no spring in its switch mechanism which imposed a load or force on the locking keys. [4]
In a letter dated March 5, 1981, Capt. McIntyre responded:
If this statement means the specific actuator remnant for #7 slat on N840TW was found to have no spring force in its switch mechanism, you are revealing a defect in this actuator not hitherto disclosed. . . . Switches measured in TWA’s shop indicated a nominal force of 13 pounds required to actuate the switch.
In their report the Board should have included the information provided in the quotations above. Such a Slat Retract Lock Indicating Switch, lacking any internal spring force, would be unable to activate the amber slat in-transit annunciation in the cockpit. This amber annunciation would have been the only clue to the pilots that the mechanical slat actuator retract lock had not engaged after slat retraction. Under normal conditions with the slats retracted, the hydraulic pressure would act against the retract side of the actuator piston. However, various forces could pull the #7 slat from the retracted position following a failure of Hydraulic System “A”. This extension would result if the mechanical slat actuator lock were not engaged and the slat was subjected to a combination of tensile loads (G forces, vibration, and aerodynamic forces).
Section 1.16.1 -- Under BOEING COMPANY TESTS ,on AAR page 9; the Boeing OMB 75-7[5]is referenced. This paragraph should also state that the OMB 75-7 outlined specific conditions which would permit a leading edge slat to be pulled from the retracted position (Mach _ .8M, with failure of hydraulics to the slat actuator).
Section 1.16.2 -- Under FLIGHT SIMULATOR TESTS ,on AAR pages 9 and 10, the NTSB’s report writers included erroneous information provided by the manufacturer. The first paragraph underFlight Simulator Tests, on pages 9 states that testing was conducted in a B-727-200 simulator. In the footnote (note 4) the NTSB repeated false information found on page C-4 in the Boeing Report:
The only difference aerodynamically between the two airplanes is a 120_ difference in body length.
First the Boeing Report, and then the NTSB Report, incorrectly stated the magnitude of the difference in body length between the (short body) accident aircraft and the B-727-200 modeled in the simulator tests. In fact, the difference in body length is twice that magnitude (body length of the 727-100 is 116 ft 2 in, the 727-200 is 136 ft 2 in; as published in the manufacturer’s document D6-1420).
Furthermore, the false information provided in the Boeing Report misled the NTSB to neglect other major “stability and control” differences between the B727-100 and -200 models. Such differences include “tail volume,” yawing moments due to rudder, sideslip generated by rudder, yawing moment due to sideslip, yawing moment of inertia, and pitching moment of inertia. The manufacturer’s attempt at post-accident simulation thus suffered from inaccurate modeling, having neglected such differences between their simulator and the accident aircraft. The 118 simulation trials also neglected the effects of the up-float of the right outboard aileron. Their B727-200 simulator could not duplicate the tightly coupled high altitude control characteristics of the accident aircraft. Limitations of the simulator prevented the investigators from exploring possible faults of the lateral and directional control systems that interacted to induce the dynamic yawing, rolling, and pitching upset maneuver of the accident aircraft.
In the first paragraph underFLIGHT SIMULATOR TESTS , on AAR page 9, the Board summarized the simulator testing. The Board should state that the investigators prearranged test conditions, intending that test results would support their assumption that the extension of the #7 slat was a cause of the upset. This paragraph should restate an important admission included in the manufacturer’s report, on page C-1:
It was also determined that the extension of slat #7 at the flight condition of N840TW prior to the incident can be easily controlled with prompt corrective action by the pilot.[Emphasis added.]
That statement suggests that the initial upset of the accident aircraft was NOT induced by an extended slat.
The second paragraph underFlight Simulator Tests , on AAR page 9, should state clearly the test conditions used to simulate uncommanded rudder deflection. The manufacturer, on page C-6 of the Boeing Report, described the method used for simulating yaw damper hardovers during the piloted simulator trials:
. . . yaw damper hardovers were not directly simulated . . .[Emphasis added.]
Regarding yaw upsets, the Board should refer to the results from Project Race, and the CAB’s investigation of the AA Flight One accident.[6]
Project Race was a program of flight tests originated by the FAA to shed light on the cause of the AA Flight One accident. NASA, AA, Boeing and the CAB participated in the tests. The tests measured the response of the airplane to the effects of slips and skids, and malfunctions of the rudder control system. In their report of January 1963, the CAB included a message meant for future accident investigators:
Project Race provided the Board with much information that will prove helpful in future accident investigations. [7]
During the investigation of the AA Flight One accident, the CAB concluded that the manufacturer’s testing of rudder upsets was inadequate.
The tests are obviously planned maneuvers under which conditions the pilot is not confronted with the necessity of analyzing the malfunction, deciding what corrective action he will take, and experimenting to produce the desired results. . . . It is unreasonable to assume that . . . the pilot, confronted with an unexpected roll, would start corrective action as soon and to the extent characteristic of planned flight tests.
The above is borne out by recorded instances of yaw damper malfunction or mismanagement. In all instances the crew was late in recognizing the yaw damper as being the source of the problem and were slow in initiating corrective action. In some cases, even after initiation of corrective action the dangerously steep banked attitudes increased and persisted well beyond flight test values before recovery was effected. In some instances of yaw damper mismanagement the crew never properly analyzed the difficulty and the flights were completed after application of additional lateral control . . . There are some instances wherein the crew took advantage of additional lateral control capabilities, recovered to level flight, analyzed the difficulty, and then disengaged the offending yaw damper. [8]
The NTSB should reiterate those findings of the CAB. The CAB found that pilots had difficulty in accurately identifying the source of an upset induced by a malfunctioning yaw damper. The CAB focused on a B707 type of design: a single rudder, with only one yaw damper. It is much more difficult to identify the source of a rolling motion, caused by yaw, when operating the B727-100 type aircraft at high cruise altitude and high gross weight combinations.
The third paragraph underFLIGHT SIMULATOR TESTS , on AAR page 9, should mention that investigators did not consider the effects of ratcheting of the aileron control wheel. Such ratcheting would reduce the lateral control margin.
The Board was unaware of the results of the 1977 flight tests done at TWA. That testing resulted in an incident characterized by near loss of control. That incident occurred aboard aircraft N840TW. The 1977 test conditions explored the complex problem of an autopilot false disconnect and the ratcheting of the aileron control wheel. In a revised NTSB Report the Board should adopt two proposed new sub sections. One proposed new sub section, 1.16.5, should describe the TWA tests aboard N840TW. The Board should analyze the 1977 TWA test and compare the results with the accident. We include an analysis of this lateral control anomaly in a proposed new sub section2.5.3,titled “Autopilot Roll Channel.”
Section 1.16.3 -- HEADING GYRO TESTS ,on AAR page 11. The combined effects of gyro gimbal error and recorder stylus shifts (due to worn recorder mechanisms), made the FDR heading trace unreliable at crucial moments. The Board should limit their use of the specious heading data and acknowledge the weakness of conclusions based on that heading data.
Section 1.16.4 -- See the first paragraph underFLIGHT TESTS , AAR page 12. The Board should honestly state the nature of the test conditions for the October 2, 1980 flight test. The nature of the flight test was recorded in the transcript of the post-flight conference:
The purpose of the test . . . is to provide a signature on a flight data recorder for several trailing edge flap and leading edge slat configurations. [9]
An assumption guided investigators as they planned the flight test. The intent of the investigators was to design test conditions that would yield useful results. The Board assumed that the extension of trailing edge flaps, and leading edge slats, had occurred immediately prior to the upset of the accident aircraft. Consequently, the flight test results are of little value in identifying any alternative failure sequence.
Section 1.16.4 -- Under FLIGHT TESTS , second to the last paragraph on AAR page 12, the statement regarding airspeed decrease is misleading. Following extension of flaps to 2 degrees, the Flight Test Instrumentation (FTI) airspeed trace showed a steady decrease, at a slow rate. The test crew sustained the condition for 39.2 seconds during which the decay of airspeed continued. Had the condition been allowed to continue, the rate of deceleration may have increased as the autopilot altitude hold function commanded increased nose up pitch
Section 1.16.4 -- Under FLIGHT TESTS , see the last sentence in the last paragraph on AAR page 12. The data did not substantiate the assertion that the airspeed increased as a result of flap retraction only. On the contrary, the test director stated during the debrief, “It’s so terribly painful at 39000 feet to accelerate back to the test speed.” (See page 83-1-D7, of D6-44357.) The investigators had pre-defined no test condition with a stabilized interval at TE flaps 5, followed by retraction of TE flaps to 2. An increase in thrust most likely yielded the increase in airspeed. The test aircraft, E209, was equipped with JT8D-15 engines. [10]These engines provided the test aircraft with a thrust advantage over the accident aircraft, which had lower powered JT8D-7 engines.
Section 1.16.4 -- Under FLIGHT TESTS ,the three paragraphs on AAR page 13 related to a single condition from the flight test. During that test condition, the test crew extended leading edge slats 2-3 and 6-7 while at FL390. The Board accentuated data from that one prearranged test condition. The Board needed to support their contention that, just prior to the upset, the accident aircraft had experienced such a slat extension.
The first paragraph on AAR page 13 described the test condition and results. The Board listed the rate of airspeed decrease as 0.50 Kts/sec following the extension of slats 2-3 and 6-7. However, on examination of the test data included in the NTSB docket, the reader finds that the rates of airspeed decay, following extension of slats 2-3 and 6-7, varied over time. Data from the two test recorders also differed. The more accurate FTI airspeed trace decayed at 0.3 Kts/sec.
This comparison of airspeed decay is discussed later in the proposed new analysis section, 2.8 -- “Investigative Errors Sparked Erroneous Findings.”
The investigators focused on the results of that single condition from the flight test. In its endeavor to link that single condition from the test flight with the initial upset interval of the accident aircraft, the Board proposed an inconclusive comparison. In the second paragraph on AAR page 13, the Board cited a similarity of the recorded airspeed decay and a similarity of the recorded frequency of vibration.
A human fingerprint is unique. However, in flight dynamics the airspeed is affected by several variables; one rate of airspeed decay might be a common response to several conditions involving various flight control deflections. In the case of the accident aircraft, the airspeed decay was most likely due to drag increments caused by other factors such as sideslip and extension of flight spoilers on the left wing as the aileron-spoiler mixer responded to autopilot roll commands.
A forensic test should, as a matter of common sense, satisfy three criteria: the underlying scientific theory must be considered valid by the scientific community; the technique itself must be known to be reliable; and the technique must be shown to have been properly applied in the particular case. [11]
The Board’s comparison of the normal acceleration traces (in the second paragraph on AAR page 13) did not meet the above criteria, and did not qualify as sound forensic technique. There is a limited frequency response of the acceleration channel of such flight data recorders. Thus, unrelated vibrations of unequal frequency beyond the response limit of the recorders would appear on recorder foils as the same vibration frequency. The FDR foil G-trace from the accident aircraft indicated vibration onset at the start of the upset. The increasing amplitude of that vibration signature suggests that spoiler buzz and the flutter of the outboard right aileron were main vibration components. The spoiler buzz resulted from the rising flight spoilers on the left wing. Flutter of the right outboard aileron may have resulted from the dynamic loads on that free floating aileron panel. However, numerous other conditions involving buffet or flutter could have produced vibration of a frequency beyond the response limit of the recorder. The accident FDR would have recorded such vibration at that frequency response limit.
The NTSB’s comparison between recorded data from the flight test and recorded data from the accident aircraft proved inconclusive.
In an effort to explain these weaknesses, the Board included false information in their report. The third paragraph on page 13 of the NTSB report stated:
It was determined that during the flight tests . . . a test switch . . . in the DADC . . . had been left in the test (HOLD) position.
The source document for that information was a letter (dated May 11, 1981, included in the NTSB docket) from Boeing’s H.P. Hogue to NTSB’s Kampschror. A close examination of that letter shows that the manufacturer never actually observed the switch in that HOLD position. Prior to the flight test, proper verification of configuration of crucial test components should have been planned by the test engineers. Normally, the test records would have included specific documentation of component configuration variables. However, the manufacturer presented no such documentation of the position of that DADC test switch for the October 1980 flight test.
During the flight test there proved to be a periodic, undamped oscillation, of the FDR g-trace that developed after leading edge slats 2-3 and 6-7 extended. Such an oscillation was not seen on the FDR g-trace of the accident aircraft. The manufacturer was unable to explain this dissimilarity. Then in an effort to correlate this dissimilar result from the test, the manufacturer made an assumption about the configuration of the test aircraft. These were the actual words reported by the manufacturer to the NTSB:
Therecan be little doubt that this switch was in the test position the day E209 was flown. [12]
The manufacturer’s use of such an assumption in test analysis should have provoked the skepticism of the NTSB investigators. Suspicion should have prompted the Board to ask: What other variables were mismanaged during this one crucial test condition? In the statement shown in their official Aircraft Accident Report, the NTSB altered an assumption made by the manufacturer and improperly reported it as fact.
The three paragraphs on page 13 of the NTSB’s AAR are not pertinent to the accident. Analysis shows that the #7 slat on the accident aircraft remained in the more secure retracted position until late in the dive. Furthermore, the eye witnesses, the pilots, testified that they deployed neither trailing edge flaps nor slats prior to the upset at FL390. The passengers aboard the accident aircraft didNOThear the growl of the flap motors prior to the upset. There were numerous passengers seated in the cabin just above those flap motors. Passengers would have felt and heard several seconds of disturbance while the flap motors operated. The T.E. flaps must extend until near the “two degrees” position before they actuate any slats. Direct evidence from the accident refuted the NTSB assumption regarding slat extension prior to the upset. The NTSB ignored this direct evidence. The results from the flight test ranked only as circumstantial evidence, created after the accident at the manufacturer’s facilities. The NTSB erred in its use of such circumstantial evidence to support an erroneous hypothesis.
Section 1.16.4 -- UnderFLIGHT TESTS ,the top paragraph on AAR page 14 improperly described a flight test condition in which the pilot pushed the right rudder pedal to full travel. The exact words printed in the NTSB report are:
the pilot deflected the rudder fully to the right . . . .
In fact, during the test condition, the FTI shows the exact rudder displacement to have been: upper -3.5 degrees, and lower -4.5 to -5.0 degrees. These angles represented the full deflection available from normal rudder actuators. An accident analyst needs to understand that at FL390 cruise conditions, due to the limitations of the hydraulic pressure acting against aerodynamic resistance, the maximum rudder deflection available from the pilots’ pedals is no greater than the deflection capability of a discrepant yaw damper signal.
This particular condition of the flight test may be more pertinent to this accident than any of the test conditions devoted to flap or slat extension. Had the investigators not fixated on the slat that separated during the accident, they might have more thoroughly tested several possible failures involving sideslip induced roll. Unfortunately, in the results from this most pertinent test condition the manufacturer clipped the FTI data plots. The manufacturer did not show the Board any information that described the set-up of the sideslip. That set-up interval would show the response of the aircraft (yaw, sideslip, bank, and pitch) to the increasing displacement of the rudder surfaces. The NTSB should have noticed this data editing and requested a more complete time-history plot. Working with the FDR plot included for the test Condition No. 1.22.013.025.1, an investigator would have found that the sideslip condition was first initiated about four minutes prior. Therefore, a careful investigator would have ordered a more complete time-history plot to include FTI (IRIG) coordinated time 13:38:30 through 13:44:0. Such a “data request” should have ordered time history plots of other parameters in addition to those parameters presented in the original data plots. Such additional parameters include: rudder pedal position (or rudder pedal force), yaw rate, and sideslip angle. These parameters were probably available from the flight test recording system but not plotted on the graphs provided by the manufacturer’s flight test section.
This most pertinent condition of the test plan (test Condition No. 1.22.013.025.1, NTSB #7) began with the aircraft in a nose-right sideslip (a negative sideslip angle). Then the pilot disengaged the autopilot, with the rudder pedal still displaced. The aircraft rolled from 5 degrees left bank through 33 degrees of right bank in 4.4 seconds. The test pilots terminated the condition after 4.4 seconds by releasing the rudder pedal and using left wheel throw, while the aircraft continued to roll right to 41 degrees right bank. The maximum roll rate recorded on FTI was 13 degrees per second; averaging about 10 degrees per second. The FTI heading trace went from 349 degrees, through 352.5 degrees, during the 4.4 second test condition; with the heading stabilizing at 355 degrees 1.5 seconds later. This FTI heading data had been generated by an Inertial Navigation System aboard the test aircraft, and was independent of any alleged gimbal error that may have affected the #1 DG (FDR Heading trace) of the accident aircraft. Following the official test condition, the pilots used 35 to 45 degrees of left (CCW) throw of the aileron wheel to bring the left wing down to level. This generated a peak roll rate of 23 degrees per second. After the pilot released the rudder, the test aircraft rolled through wings level (zero bank angle) after another 4.1 seconds.
The manufacturer clipped the data presentation for the period during which the pilot established the sideslip (test data was unavailable in the docket). Following the 4.4 second test condition, the data is of little use in the analysis because the test crew rolled the aircraft into a sustained left bank while climbing.
The FTI sampled the positions of the upper and lower rudders during the flight test. The manufacturer presented some test data, with the deflection angle of each rudder plotted in units of degrees. These plots show that during the sustained sideslip, the upper rudder was at -3.5 degrees (the minus sign signifies TE right) and the lower rudder was -5.0 degrees. At the start of the test condition, just after the pilot disengaged the autopilot, the lower rudder deflection diminished about 10% to -4.5 degrees, as the aircraft began a yaw and rolling motion to the right. Sideslip remained roughly constant, though fluctuating, during this brief interval as the lower rudder deflection diminished. The lower yaw damper may have commanded the 10% decrease in rudder deflection in response to the onset of right yaw. In contrast, the upper rudder deflection remained at -3.5 degrees. Because the plots, provided by the manufacturer, did not include a time history of rudder pedal force, the analyst can not be certain of the exact inputs made by the pilot. However, from the traces of rudder surface position it would appear that yaw damper was active at FTI time 13:43:10 through 13:43:19. These apparent yaw damper inputs consisted of a 2 degree right rudder deflection, followed two seconds later by a 0.5 degree left deflection, followed 1.5 seconds later by a 1.5 degree right rudder deflection. The manufacturer did not include instantaneous yaw rate on the data plot. Examination of the INS heading trace shows that these rudder (yaw damper) deflections occurred at slope changes of the heading trace.
The test instrumentation for the upper rudder may not have been accurate. A comparison of the upper and lower rudder traces after time 13:43:25 suggests that the FTI recorded neutral position for the upper rudder as about 0.3 degrees left deflection. Thus, an analyst should make a correction of -0.3 degrees to the indicated values for upper rudder.
The sideslip trace, plotted as differential pressure in units of inches of water, varied from 6.5_ during the sustained left sideslip, to 1.5_ twenty seconds after the rudder release. Following the test condition, after the release of the rudder pedal, there are several peaks and dips shown on the sideslip trace. The analyst can compare these local maxima and minima with those shown on the rudder position traces. The slope reversals on sideslip trace lag those of rudder (yaw damper) trace by 1 to 2 seconds. Had the NTSB ordered a more thorough “data request,” the time history of sideslip angle might have been a more useful parameter.
The Board included the flight test data discussed above in the NTSB docket. The data was part of the Addendum to Performance Group Report, dated January 7, 1981.
Proposed new subSection 1.16.5, TWA TESTS ABOARD N840TW . The TWA flight tests of 1977 included an incident that involved aircraft controllability. The incident occurred after a suspected autopilot false disconnect. The pilots encountered heavy stick forces, uncommanded control inputs, and the uncentering (ratcheting) of the aileron control wheel. The Board did not previously consider this anomaly of the B727 flight control system. It constitutes new evidence. The Board should include in the docket the enclosed affidavit from the eye witness of that 1977 incident. The Board should also require that airlines forward to the NTSB details of past B727 control incidents, or incidents of high altitude instability. A proposed new sub section, 2.5.3 (titled Autopilot Roll Channel) includes further analysis relating the 1977 TWA test results to the accident of TWA 841. The enclosed affidavit describes that 1977 incident.
The affidavit describes a flight controls incident that had occurred previously aboard the accident aircraft.
|
|
During a crew debrief of the flight controls incident that occurred on 17Dec89 aboard TWA Flight 1070, I, Capt. P. T Williams (TWA), was prompted to recall a problem with the B727 flight control systems that had occurred years earlier.
I, Capt. P. T . Williams (TWA), witnessed an incident that occurred during a flight aboard the B727-100, plane N840TW, on Flight 5562, from MCI, on May 23, 1977. At the time, I was a TWA Check Airman. I was flying the aircraft from the left seat (but not as PIC) while receiving a Proficiency Check Ride from Capt Sal Fallucco (who was Pilot-in-Command). Oddly, also aboard that day, as observers, were several TWA employees involved in engineering, and pilot training. During the flight I was directed to forcefully overpower the roll channel of the engaged Autopilot, so that the observers could evaluate the response of the aircraft - autopilot systems. As a result of such repeated test conditions, an abnormal flight control incident developed.
I was directed to disconnect the autopilot during these “overpower” conditions. During the first test condition the autopilot cleanly disconnected. The difficulties began when during the second test condition, the autopilot apparently disconnected (the A/P Paddles dropped to the Disengaged Position) but the flight controls became extremely difficult to manipulate in roll, and there was no annunciation provided to indicate the source of the problem. After much effort on the part of the pilots the aircraft was returned to MCI. With emergency equipment standing by, a precautionary landing (at flaps 5 degrees) was made by Capt. Wayne Disch -- Manager of Training at that time. During the landing rollout at MCI, when aileron inputs were no longer needed, Capt. Disch released the control wheel, which then rotated full travel. Something then snapped beneath the floor of the cockpit, and the control wheel returned suddenly to a normal position. Thereafter the control wheel was free, and could be rotated normally in either direction.
The foregoing is true to the best of my knowledge and belief.
Subscribed and sworn to before me this24thday ofMay, 1990.
Notary Public--State of Missouri
Section 1.17.1 -- Under B-727 FLAP SYSTEM , on page 14 of the NTSB AAR, the first sentence was erroneous. It incorrectly stated that there are four LE Flaps. In fact, there are six LE Flaps, three on each wing of a B727.
Section 1.17.2 -- HISTORY OF B727 LEADING EDGE SLAT PROBLEMS ,NTSB AAR page 18. The NTSB failed to mention that B727 aircraft had suffered numerous cases of slat separation. Prior to 1979 SDR’s showed a total of sixteen documented cases of loss of a slat surface. The documented cases indicated that the crew usually noted the loss of “A” Hydraulic System immediately after the loss of a slat. (In eight other cases poor documentation proved insufficient to determine whether the extended slat had actually separated.) These cases involved uncommanded extension of the slat prior to separation. Data included in the SDR’s and incident reports verified the manufacturer’s calculations regarding the limited load bearing strength of an extended slat. We discuss this subject in a proposed analysis section (2.4.3),titled “Limited Strength of an Extended Slat.”
Section 1.17.3 -- Under AIRCRAFT PERFORMANCE ,on pages 18 through 20 of the NTSB AAR, the Board based their analysis on the false assumption that a LE Slat extended at FL390. Paragraphs that discussed the calculated rolling moment, due to an extended slat, were not pertinent. A reader could mistake these calculations for fact. The Board should delete these paragraphs from the NTSB report (or at least remove the paragraphs from the factual section of the report, and insert them into the analysis section). The last paragraph on AAR page 19, and the next three paragraphs on page 20, discussed such calculations and are not pertinent to the upset sustained by the accident aircraft.
A graph was displayed on AAR Page 21 that depicted the rolling moment due to an extended #7 slat. The graph is not pertinent to the upset of the accident aircraft and should be deleted from the NTSB Report.
Section 1.17.4 -- titled NO. 7 LEADING EDGE SLAT OPERATION , was on pages 20 through 22 of the NTSB AAR. The two existing paragraphs that form this subsection of the NTSB report described the loads on a retracted slat under the initial upset conditions (FL390 cruise). The discussion was incomplete because it failed to mention the factors that would interact to pull a slat from the retracted position following loss of Hydraulic System “A” pressure. We present these factors in a proposed new (sub) Section 1.17.5 titled “Loads On A Slat.”
The Board should use this Section 1.17.4 to better describe certain aspects of the operation of the leading edge slat.
On page 16 of the NTSB AAR, there was an illustration that showed a side-view of a slat. The illustration showed the slat in the retracted position, and in a mid-extension in-transit position. Unfortunately that illustration was not thorough enough. The illustrator failed to label several components important to the proper operation of a slat.
Engineers presented a better drawing in the manufacturer’s report on page B-19. The Board should include this drawing in the NTSB report and refer the reader to it in Section 1.17.4. This drawing has a more complete depiction of the slat components, and a more thorough identification of those components. This drawing offers the reader a view of an extended slat and the interaction of important components.
The Board should describe the function of theSLAT ACTUATOR RETRACT LOCK INDICATING SWITCH. The Board must tell the reader that a fault in this indicating switch could leave the pilots with no means of identifying an unlocked slat, following slat retraction after takeoff. Normal hydraulic pressure would hold an unlocked slat actuator piston in the retracted position. However, any subsequent loss of hydraulic pressure could result in the slat being pulled from the retracted position by various forces acting on that wing section.
The inboard track of the # 7 slat suffered a structural failure of itsEXTEND STOP.[13]The Board should describe and illustrate the function of the components of that slat track and the slat track carriage. The Board should describe the impact overload forces encountered by the inboardSLAT TRACK EXTEND STOPwhen it contacted theTRACK CARRIAGE STRUCTURAL STOP.
The Board should describe the actuation of theSLAT EXTEND POSITION INDICATING SWITCHas the slat track moved to the extended position. The reader should understand that even after the #7 slat separated from the aircraft, the annunciation failed to alert pilots that the #7 Slat was anyway abnormal. Lacking any input from the faultedSLAT ACTUATOR RETRACT LOCK INDICATING SWITCH, the amber “LE Flap” annunciation would fail to illuminate when the #7 slat unlocked. The green “LE Flap” annunciation would have indicated that #7 slat extended after repositioning the flap handle during the Alternate Flap Extension Checklist.
Proposed new subsectionSection 1.17.5, LOADS ON A SLAT .This revised subsection should discuss both tensile and compressive loads on the actuator of a retracted slat, and the load limit of an extended slat. The Board should describe how such factors as high speed, high G’s, vibration, and sideslip, influence loads on the actuator of a retracted slat. Then the Board could describe the effects of those factors on the structural components of a slat subjected to sudden over extension following loss of hydraulic pressure.
Loads on an Extended Slat
The Board should discuss the design load limit of an extended slat. Consider this hypothetical question: Had the #7 slat extended at FL390, at what point during the uncontrolled dive would it have failed and ripped away from the wing? Using design coefficients, substantiated by flight loads survey data, the manufacturer calculated the “most probable” moment of slat separation to occur at a drag load of 5.11 psi. (The manufacturer presented that data as a graph on page B-12 of their report to the NTSB dated Sept. 24, 1979.) The manufacturer determined that the accident aircraft encountered this most probable failure load at 363 KIAS. By referring to that point in the FDR tabular data, a rate-time interpolation between altitude data points shows that the accident aircraft exceeded that airspeed while diving through 31812 feet FDR altitude. (After correcting for the known FDR altitude error, the slat could have stayed intact until an actual altitude of about 31500 feet.) The manufacturer stated, on Page B-5 of their report, that the probable time that an extended #7 slat would have separated due to overload was 24 minutes, 1.5 seconds (Flight Data Recorder time). Had an extended slat caused the upset and the uncontrollable rolling moment, the slat would have separated during the first third of the dive.
The aircraft would then have been controllable. Yet the accident aircraft continued to dive uncontrollably for another thirty seconds. An early separation of the slat, along with the associated fracture of the slat actuator, would have resulted in loss of pressure in the “A” Hydraulic System. Yet that hydraulic system was intact nearly thirty seconds later in the uncontrolled dive, when it powered the landing gear extension. The trail of debris found on the surface suggested that the #7 slat could not have separated until after gear extension. Correlation of the evidence suggests that an extended #7 slat could not have caused the sustained interval of uncontrollable flight.
The Board should carefully consider one further point. When estimating the most probable moment of failure and separation of an extended #7 slat, the above estimate of 363 KIAS is probably unrealistic because of the worn slat components involved in this case. The manufacturer arrived at that estimate by comparing air loads with slat design loads. The manufacturer’s report (on page B-4) alludes to this inadequacy,
Intact slats would have survived the initial seconds of the incident in an extended configuration. It is apparent from physical evidence, however, that slat #7 was not intact . . . due to pre-existing damage.
The manufacturer’s own metallurgist documented slat misalignment and fatigue progression in components. Had the #7 slat actually been in the extended position from the very beginning of the upset, the “most probable” moment of separation would occur prior to the instant calculated above. The worn slat would have ripped away even earlier within that first third of the uncontrollable dive segment.
LOADS ON A RETRACTED SLAT
The manufacturer’s report also provided the NTSB with a suggestion of how and when that #7 slat ripped away from the accident aircraft. The following excerpt (manufacturer’s report, Page B-2) addressed the conditions under which one or more slats could have extended;
(a) if the slats were unlocked, hydraulic system off, spoilers up and/or load factor less than 1.0; or if,
(b) inadvertent or intentional partial extension occurs under conditions of high mach number and low lift coefficient.
Each of the variables identified in the above excerpt contributed to thelocalized forces acting on a retracted #7 slat. Yet in this case, perhaps other variables contributed to the local aerodynamic forces. Various conditions affected the outboard section of the right wing. The aircraft was in a left sideslip (nose-right, left wing forward), which induced proverse roll to the right. [14]During most of the dive, the aircraft was experiencing high speed buffet. While diving through 460 KIAS, M.85, about 12,000 feet, the gear extension caused damage to hydraulic lines. Each slat actuator lost the hydraulic pressure that had forced its piston to the retract position. Air loads on the right outboard aileron forced it to float from its normally faired position. This up-float, over one inch upward, affected that retracted #7 slat as would a deflected spoiler. Flutter of the right outboard aileron may have excited the wing section into vibration. The evidence suggests that these conditions existed for a very short time during the sequence of failures. These conditions existed immediately after the right main landing gear suffered damage as it over extended due to the side loads. And the evidence suggests that it was at this point in the failure sequence that the #7 slat extended as the hydraulic pressure to the retract side of the slat actuator decayed.
Proposed new sub section ,Section 1.17.6 ,B727RUDDERCONTROLSYSTEMS.
The following information about the rudder control systems is provided to the pilots in the TWA Flight Handbook.
The upper rudder is powered by [hydraulic] system B and the lower rudder is powered by [hydraulic] system A. There is no manual operation of the rudders; however, the lower rudder can be powered by the standby hydraulic system through a separate actuator. . . . A cable system transmits rudder inputs from either pilot’s rudder pedals to the power units in the vertical stabilizer. [15]
The yaw damper system is designed to counteract yaw due to dutch roll. Independent systems on the upper and lower rudder sense yaw and position the rudders to stop it. The yaw dampers are designed to operate full time. [16]
At high altitudes and cruise mach numbers, the Dutch Roll characteristics of the 727, without the yaw dampers, is undamped and divergent. If not corrected it will deteriorate into a complete loss of control. [17]
Lacking the stability augmentation provided by the yaw dampers, the high altitude lateral dynamic instability characteristic of the longer B727-200 aircraft isNOTequal to the dynamic instability characteristic of the shorter B727-100 aircraft. The Dutch Roll characteristic of the shorter model is less damped. The fraction of critical damping decreases to zero as the altitude increases toward FL260. Above about FL260, for the B727-100, the fraction of critical damping changes sign and becomes more negative (divergent) at higher altitudes. (This data was disclosed in attachment D, of a letter from H. P. Hogue of Boeing, to R. VonHusen of the NTSB; dated December 19, 1980.)
The normal Power Control Unit (PCU) for each rudder serves acommon mode function . Each hydraulic powered rudder is positioned by a PCU which responds to commands from the pilot through his rudder pedals. Concurrently, each PCU is designed to respond to electrical signals input from its associated Yaw Damper (a subsystem composed of numerous parts, located in various sections of the aircraft, and accepting inputs from several other subsystems). The two input commands to each rudder PCU -- the rudder pedals and the yaw damper -- are completely independent. There is no feedback to the pilot’s pedals when a rudder is displaced by a yaw damper command.
The flight test of October 2, 1980, demonstrated the maximum deflection angle of the rudders for the cruise conditions at FL390. Full rudder pedal travel yielded a rudder surface deflection angle of:  ru = -3.5° for the upper rudder; and rL= -4.5° to -5° for the lower rudder. Instrumentation recorded those measurements after the aircraft was established in a steady nose right sideslip using full right rudder pedal travel. (The Board failed to determine the exact value of the sideslip angle during that important test condition.)
For this case of nose right sideslip, hinge moment calculations show much less rudder surface deflection available in the opposite direction, acting to oppose the sideslip. The negative sideslip angle, airplane nose right, acts to limit the magnitude of the available positive (left) deflection of a rudder surface. The rudder hinge moments increase proportionally with the square of the equivalent airspeed, thus also limiting the available rudder deflection
A malfunction of either or both rudder control systems could result in split rudders. The manufacturer (in their Document D6-8095 pages 4.3-1 through 4.3-6) considered such rudder anomalies:
Due to the several boost systems and the dual pressure available on the main systems, there is a multiplicity of failure modes. . . . It should be noted that there is a considerable difference in rudder hinge moments between normal operation (both segments deflected) and abnormal operation (segments deflected separately). [18]
It is difficult to model the rolling moments and yawing moments resulting from possible aberrant rudder deflections. Calculated control margins may ignore the additional moment increments that ensue from dynamic effects of roll rate and yaw rate. The manufacturer attempted some simulator testing following the accident. However, their simulator emulated the stretch model of the aircraft — a 727-200. The TWA 841 accident involved a 727-100 aircraft which has significantly different directional and lateral stability characteristics (especially at high cruise altitudes). The manufacturer consistently reassured the Board that the characteristics of the two models were similar. However, implicit in several communications from the manufacturer was the suggestion of critical differences. In a letter dated 19Dec80, from Boeing’s H.P. Hogue to the NTSB’s R. VonHusen, the manufacturer conceded that,
The 727-100 would have more sideslip due to rudder than would the 727-200 . . .
The manufacturer explicitly associated that information with its effects on the control margin of an aircraft with an extended #7 slat. The Board failed to recognize the real importance of that sideslip evidence. Sideslip and the rudder control systems were important variables, critical to the high altitude characteristics of the B727-100.
The manufacturer worked with the NTSB’s investigators and they consistently judged new evidence with such a bias. The investigators had constructed an interpretation of an ambiguous situation. Thereafter they routinely processed new information with a bias toward their interpretation. The NTSB’s investigators either adopted or abandoned evidence, based upon that bias. [19]The investigators’ erroneous assumption, that an extended slat had caused the upset, prevented them from properly documenting more pertinent evidence.
The controllability testing and related calculations accumulated during the NTSB’s investigation applied to an aircraft flying at FL390 with an extended slat. The NTSB’s records do not include sufficient information about faults in the lateral and directional control systems that may have contributed to spiral divergence. Lacking information, it is difficult for any analyst to verify the possible control margins during a fault in one or both of the rudder control systems. (The Board should study the investigations of the AA Flight One accident, and the MAC 59402 mishap, to learn appropriate investigative methods for such control system failures.)
The B727 Maintenance Manual (MM 22-00, Page 1) stated that yaw damper commands have an authority limit of ±5 degrees of rudder deflection.
The B727 Maintenance Manual provided the investigators with more information about the yaw damper commands, and the rudder system:
The yaw damper consists of two yaw damper couplers, two guarded yaw damper engage switches, a rudder trim and position indicator and two rudder position sensors. . . . Each of the two yaw damper couplers sends electrical signals to their respective rudder power control packages. . . . When the yaw damper is engaged, the yaw rate gyro senses any changes in the yaw axis, and the yaw damper provides the necessary airspeed compensated signals to the rudders to stabilize the airplane. [20]
Two yaw damper couplers are installed in the electronic equipment racks, one for the lower and the second for the upper rudder. Each coupler consists of a yaw rack assembly and several plug in modules. The plug in modules consist of the yaw rate gyro, yaw servo amplifier, yaw synchronizer, yaw calibrator . . . .
Each rate gyro is a non-hermetically sealed unit with simplified in-line construction. No slip rings or brushes are used for the microsyn stator, microsyn rotor, or the hysteresis motor stator. The gimbal assembly is supported by permanently lubricated ball bearings and a torsion bar whose movement is limited by a stop mechanism. A viscous damper is used to give an adequate damping ratio to the gimbal assembly.
The electronic plug in modules contain several electronic cards. . . . Each card contains the electronic or electro-mechanical components needed for amplification, synchronizing, or band pass filtering of the yaw damper signals. Each yaw rack assembly contains a 30 volt dc power supply and a transformer providing the various ac excitation voltages. The yaw calibrator contains various resistors to adjust the gains of the yaw damper coupler to a specific airplane configuration. [21]
The sensors for each yaw damper consist of the yaw rate gyro and the linear series actuator position transducer. The yaw dampers are engaged by separate Rudder Yaw Damper Engage switches. The yaw damper electronics are contained in the yaw damper coupler which comprises the following functional modules for signal coupling, shaping and amplification: yaw synchronizer, yaw servo-amplifier and gain calibrator. The air data sensor varies yaw damper gain as a function of airspeed (dynamic pressure Q-parameter control). The rudder servo system is composeed of a rudder power unit which contains the transfer valve, the yaw damper actuator, the actuator position transducer, the pilot input linkages and main control valve and actuator. [22]
The signal input circuitry, transducer position feedback, wipe-out circuit, filters, and signal summing circuits, are described in the MM 22-00, on page 19.
The Rudder and Elevator Position Indicator as installed on the center instrument panel of TWA B727 aircraft is different than that used on aircraft at some other airlines. There is a position pointer for each rudder, intended to display rudder deflection. The indicator also has a failure alert flag for each yaw damper. The materials provided to flight crews describe this fail flag as an annunciation for loss of ELECTRICAL power to the yaw damper, or the engage switch toggledOFF. As installed on TWA aircraft this failure flag signals loss of electrical power only. The flag does not annunciate the loss of hydraulic pressure to the rudder actuator. Such a failure alert is only displayed for the crew, and not recorded by the flight data recorder.
The lower Yaw Damper Fail Flag was observed by the flight crew of accident aircraft. They first noticed this failure alert after recovery from the uncontrollable flight segment. The NTSB never determined why the fail flag appeared. Investigators may have mistakenly assumed that the Yaw Damper Fail Flag was a consequence of the loss of hydraulic pressure to the lower rudder.
Section 1.18 -- titledUseful or Effective Investigative Techniques , on page 22 of the NTSB AAR. The Board adopted a deceptive investigative technique suggested by the manufacturer. The fourth sentence erroneously states,
This technique permitted the illustration of highly accurate g-trace frequencies . . . .
The data recorder mechanisms each had a limited frequency response. The data recorder mechanisms record all vibration of any frequency greater than about 6 cps at that limit for frequency response. This was an unprecedented and an unrecognized forensic technique. Furthermore, such a technique of vibration frequency comparison was inappropriate in this case due to the limited frequency response of the particular recorder mechanisms. The results from such a vibration comparison were therefore ambiguous. Yet the Safety Board applied this comparison technique in an attempt to correlate data from an accident with data from one prearranged test condition. By using this inappropriate comparison technique, investigators were able to find apparent vibration similarities between unrelated conditions.
Proposed new sub section 1.19,Efforts to Revise Investigative Errors . Newspaper articles and a documentary on network television -- about the mistakes in the NTSB’s investigation of the accident -- prompted an independent citizen to suggest an alternative hypothesis for the upset. In a paper offered to the Accident Investigation Department of the Air Line Pilots Association, Mr. Duane Yorke, of Massapequa, N.Y. described the results of his own analysis.
The strength of Mr. Yorke’s analysis was that it made sense of the evidence. Damage to various components of the accident aircraft fit with the observations offered by witnesses (the pilots) to suggest a failure sequence. Subsystem failures initiated the upset and sustained the uncontrollable flight segment. Then a final subsystem failure resulted in controllable flight. His inspiration for this analysis was a knowledge of previous yaw, roll-over, vertical dive incidents during the early days of swept wing aircraft.
Mr. Yorke has thirty years of experience in aeronautical engineering. He holds a BSAE from MIT. He served as Director of Supersonic Aircraft Development for Grumman Aircraft Corporation.
The Board should rearrange the Analysis portion of their report, so that they can present the analysis of the direct evidence in a logical manor. The analysis should begin with the trail of debris found on the ground north of Saginaw. The report should then “work backwards” from the recovery phase to the initial upset, correlating the direct evidence. The Board’s Report should correlate the specific damage sustained by the accident aircraft, with the information from the Flight Data Recorder, and the observations of the crew and passengers.
The direct evidence fits together to suggest a sequence of failures. The accident analysis should focus on direct evidence. The Board should carefully avoid the use of assumptions and circumstantial evidence. The Board should correct those errors in its analysis in which they ignored direct evidence. When the manufactured circumstantial evidence contradicted the direct evidence, the Board often employed the circumstantial evidence and rejected direct evidence. For example, the Board consistently rejected testimony of the crew in favor of conflicting circumstantial evidence.
The erroneous assumption that the aircraft experienced a slat extension while at FL390 tainted major portions of the Board’s Analysis Section. In order to correct the resulting mistakes in their report, the Board should now discard many whole paragraphs from their initial report.
Section 2.3 --The Aircraft , on pages 22 and 23 of the NTSB AAR. Two paragraphs described the accident aircraft as having been free of maintenance discrepancies, and stated that the flight crew had not noticed any malfunctions during the takeoff and climb-out. However, the Maintenance Records Group identified two discrepancies found during the previous “C” Check, which may have persisted and contributed to the later accident. The #7 slat actuator and the lower rudder actuator both may have been defective. During the previous “C” Check, maintenance inspectors documented hydraulic leakage near each of those actuators. Evidence suggests that each of those units experienced faults during the accident that occurred one month after the “C” Check.
The Board should add an additional paragraph discussing the lack of fault annunciation for many (normally passive) subsystems on that B727-100 model of aircraft. Most subsystems are poorly instrumented. In the design of such units the manufacturer provided no direct method to alert the pilots that a subsystem may be unreliable. Some examples of subsystems that lack fault annunciation in the cockpit are: the cockpit voice recorder; either yaw damper; either rudder actuator; a slat actuator retract locking mechanism and indicating switch. Thus, without any form of fault annunciation, a flight crew could only identify a fault in a subsystem after experiencing an “incident.” During such an event at least one crew member must first perceive that the faulted (normally passive) subsystem had failed to perform an intended function. Only after such a subtle discovery could a crew member attempt to identify the particular fault .
Proposed new analysis section 2.4,DIRECT EVIDENCE .
In the Board’s initial Report, Section 2.4 was titled “Extension of the No. 7 Leading Edge Slat.” That section had been a major part of their report because the Board regarded the slat extension as part of the initial upset. That section of their report included eighteen paragraphs, covering various topics, spread over pages 23 through 27 of the NTSB Report (AAR).
The Board should completely rearrange Section 2.4 of their report. The slat separation was only an effect of the upset, not a cause. The Board could utilized the Section 2.4 of their revised report to present analysis of the direct evidence.
There were numerous errors in the NTSB’s investigation. Some of the analysis was completely inappropriate, done in an attempt to support the Board’s erroneous assumption that a slat extended at FL390. Many of the paragraphs included in that Section 2.4 of the Board’s initial Report are discussed in the proposed new sections 2.6 and 2.7 and 2.8.
Proposed new sub section2.4.1,THE TRAIL OF DEBRIS .
The wreckage distribution chart is one of the most useful tools the investigator can use . . . failure patterns and failure sequences suggest themselves when the completed distribution chart is carefully studied. . . . the wreckage distribution serves as the only record of how the various pieces were located at the accident scene. The significance of later findings often depends upon reference to the original wreckage distribution chart; and if one had not been prepared, the investigation could be seriously hampered. [23]
Several aircraft parts ripped away from the aircraft in-flight. These aircraft parts were found impacted on the surface north of Saginaw. Included in the manufacturer’s report on page A-10 was Figure 4. (See the illustration reprinted on the next page). This figure showed that four parts ripped from the aircraft had impacted the ground along a path stretching five thousand feet, oriented to a northeast to southwest baseline. Most importantly, this figure revealed something not precisely stated in the final NTSB report: the separated flap track “canoe” fairing[24]impacted very near the inboard half of the #7 slat. This Figure 4 showed that the outboard half of the #7 slat impacted about one thousand feet from the previously mentioned parts, as measured along the reference baseline. About four thousand feet further to the southwest, a portion of the #6 flight spoiler panel6 flight spoiler panelimpacted.
Intentionally left blank.
Trajectory Analysis
|
at 230 KEAS, at FL390; with a malfunctioning lower rudder actuator inputting maximum right deflection. At normal hydraulic boost pressures the lower rudder could achieve approximately