Aviation Investigation Report A98H0003
Supporting technical information (STI)
Aircraft behaviour and track after the recorders stopped
The final 5 1/2 minutes of the SR 111 flight were not captured by the flight recorders, as fire-related system damage caused the recorders to stop prematurely. However, secondary radar data from several sources were available for the entire flight, including the diversion to Halifax and most of the final 5 1/2 minutes.Footnote 1 Data recovered from the FADEC NVM from engines 2 and 3 also provided useful information for the final minutes of the flight.
To characterize the final minutes of the flight, the NTSB Vehicle Performance Division made an initial radar data study using the Halifax radar data. Although several radar sites captured parts of the latter portion of the flight, the Halifax radar gave the most complete data, with primary returns down to 600 feet above the ocean. The altitude floor of radar depends on line-of-sight, and the Halifax radar site is only about 38 nm from the crash site. At about 53 nm distant, the Greenwood military site was the next closest site, but data from that location was of a lower resolution and was not used in the analysis of radar data.
The Halifax data showed that transponder Mode C altitude information was temporarily lost at 0125:11, approximately 30 seconds before the flight recorder stopped.Footnote 2 Given the 4.82-second sweep rate, it is possible that the loss of Mode C occurred as much as 4.82 seconds earlier. The Mode C altitude information was regained between 0125:45 and 0125:50, and indicated a constant pressure altitude of 9 700 feet over the next four samples. The Mode C was subsequently lost for the final time between 0126:04 and 0126:09. Primary radar returns were recorded for the next five minutes after the final Mode C loss, with the final primary radar return recorded at 0131:08.Footnote 3 The impact with the ocean occurred 10 seconds later, at approximately 0131:18, based on seismic information.
The radar data was studied with the DANTE computer program, developed at the NTSB. This program takes inputs of radar position data, wind information, and aircraft weight and aerodynamic coefficients and calculates several flight and performance parameters, including pitch, roll, and yaw angles, altitude, ground speed, true airspeed, track angle, drift angle, flight path angle, angle of attack, and load factor. The wind information used in this analysis was based on the 0000 UTC weather balloon data from Yarmouth, Nova Scotia, recorded on 3 September 1998. An aircraft gross weight of approximately 230 tonnes was assumed for the study. Radar inaccuracies in the measurement of time, range, and azimuth information typically result in unrealistic oscillations or discontinuities in the ground track, and this characteristic is generally even more apparent with primary radar data. Thus, the derivatives calculated from radar position tend to be erratic as well. For this reason, the radar data were smoothed to attenuate the oscillations in the calculated derivatives. Since the process of smoothing can eliminate realistic peaks in the data, the calculated performance parameters may be suspect in certain areas. For example, the calculated bank angle ranged between 0 and 40 degrees during the final right-hand descending turn; however, the calculation shows an increase to approximately 50 degrees of right bank for the final few primary radar returns. It cannot be confirmed whether this calculation is accurate or reflects excessive smoothing of the radar returns, which artificially increased the radius of curvature of the ground track in that area of the flight.
The results of the radar study showed that there was generally good agreement between the radar-derived parameters and FDR data, up to the time the flight recorder stopped. For example, speed differences were found to be within 3%. The study also suggested that there were no sudden changes in the aircraft's behaviour or trajectory that would indicate a dramatic event between the time of the last transponder Mode C return at 9 700 feet and the last primary radar return as low as 600 feet above the ocean. The performance calculations suggested that the angle of attack remained well below the stall angle, and bank angles remained below 40 degrees.
The precise altitude profile during the final descent from 9 700 feet was not known. Based on the elapsed time between the last transponder Mode C return at 9 700 feet and the crash, it was possible to calculate an average rate of descent consistent with the spacing of the radar returns, the site of the impact, and the estimated time of impact. Based on the above assumptions, the radar study showed that it was possible for the aircraft to have flown from 9 700 feet to the point of impact at an average rate of descent of approximately 1 800 fpm. The calculations also showed a ground speed varying between 230 and 370 knots, and bank angles ranging between 0 and 40 degrees right. The average ground speed for the last 5 1/2 minutes of flight was calculated to be approximately 285 knots. Based on the assumed altitude profile, the calculated pitch angle varied between -3 and 3 degrees during the descending right turns. This suggests that the aircraft was in relatively stable flight in a right descending turn down to about 600 feet.
Fault information normally used for engine maintenance purposes was recorded by each engine's FADEC. Each recorded fault had an associated time delay between fault occurrence and the time it was written in NVM. Each recorded fault also contained additional information, including N2, Pamb, M, and pressure altitude, sampled at the time the fault was written into NVM. Since the FADEC data was not synchronized with any standard time references, the fault timing had to be synchronized with FDR data, using recorded N2 and Pamb.
No faults were recorded until the aircraft had descended to 10 000 feet. While some of the faults were recorded when the FDR was still operational, many were written to NVM after the flight recorders had stopped. Consequently, the precise time of most of the faults is not known. The manufacturer of the engine analyzed the data and found that engines 2 and 3 lost FCC-1 inputs between 0125:05 and 0125:07; this is consistent with the FDR information. No faults were recorded by the Engine 3 FADEC below 10 000 feet. Engine 2 recorded some faults, possibly between 10 000 and 2 000 feet. Engine 2 also recorded a fault corresponding to a FADEC reset, at a pressure altitude of 1 782 feet. The reset was consistent with an Engine 2 shutdown via the fuel condition switch, with subsequent faults indicating a decay in N2 down to windmilling speed. Given the tolerance of ± 470 feet on the recorded pressure altitude, and assuming a sea level pressure of 29.80 inches, the reset occurred between approximately 1 300 and 2 300 feet above the ocean. A rough estimate of the time of the Engine 2 shutdown, using the altitude profile developed for the flight animation, was approximately 45 seconds to one minute before impact, based on an average descent rate of 1 800 fpm.
The debris field was small. If the aircraft had broken up while skipping across the surface of the water, the debris field would have been larger. The concentrated location of the wreckage and the severity of the damage suggest that the aircraft did not enter the water at a glancing attitude. In a crash into water at a sufficient speed to shatter an aircraft, as in this case, the heaviest parts travel farthest along the wreckage path. Heavy items originally on the fuselage reference line tend to stay on that line; light items tend to drift with the current.
During examination of the wreckage, it was often noted that the impact force appeared to have been at an angle of 15 degrees to the right of the fuselage centerline. This is indicative of descent into the water in an uncoordinated right bank. The damage to the right horizontal stabilizer is consistent with the impact forces having been more severe on the right than on the left. In addition, both wing engine mounts failed instantaneously in overstress. The Engine 3 mount experienced a significant clockwise torque loading; that is, the cross-beam that slips over the lug tended to twist clockwise when viewed aft to forward. By contrast, the Engine 1 mount lug broke off with less indication of a clockwise torque load. The principal loading to cause the separation would likely be upward bending of the lug.
ADG fuselage frame and door
The door through which the ADG drops is between fuselage stations STA 555 and STA 575. The opening is about 20 inches by 11 inches, with semicircular ends 11 inches in diameter. The long edges parallel the fuselage centerline. The door hinge is on the outboard long edge. The door is swung open by its linkage to the ADG. The ADG is mechanically released from the cockpit, which allows a pressurized snubber to force it into the deployed position.
Wreckage of the ADG door frame was recovered. From this wreckage it appears that the door frame was "dished in" from the bottom and crushed aft to create major folds at an angle of 15 to 25 degrees to the longitudinal axis of the door frame and, therefore, to the fuselage. This is consistent with the impact having been at 15 degrees to the right of the fuselage centerline.
Avionics bay door
The trap door from the cockpit into the avionics bay was crushed into buckles, starting at its right front corner and oriented 15 degrees to the fuselage centreline.
The throttle quadrant pedestal was crushed and twisted about 15 degrees clockwise about its vertical axis.
The standby compass has an almost perfect witness-mark crease in its upper right corner, which matches the left edge of the centre bar of the cockpit windshield frame at its upper end where it starts to curve. This crease also suggests an impact angle 15 degrees to the right of the fuselage centerline.
The standby gyro horizon, which is almost on the centerline of the aircraft, was crushed in at the upper right corner, and has buckles oriented 15 to 20 degrees to the fuselage centreline.
The right cockpit window was slightly more damaged than the left window. This is consistent with it facing more to the front, with an impact at 15 degrees to the right of the fuselage centerline.
Angle of Attack Vane
The angle of attack vane on the left side of the aircraft was bent 22 degrees downward relative to the fuselage reference line (and now has 32 degrees of anhedral). Taken in isolation, this tends to support a relatively low angle of impact with the water and the view that the aircraft did not crash inverted. The premise for this theory is that the aircraft, on entering the water at an acute angle, tends to plough, as it meets greater resistance to motion downward than forward. The angle of attack vane then streamlines under the water-exerted force, moving the vane arm to a positive angle-of-attack position and bending the vane wingtip down into anhedral. Had the aircraft entered the water inverted at an acute angle, the vane wingtip might now be displaying dihedral. It is not possible to say what effects shielding had on the water flow in the area and, therefore, whether this view is accurate.
Engine pylon attachment fittings
The fittings that attach the engine pylons to the front spars were both bent aft and twisted clockwise (looking down). The fitting associated with the left engine was bent aft to a greater extent, but the structure adjacent to the right fitting was more damaged.
The last three primary radar hits occurred at the following UTC times and aircraft coordinates:
|0130:58.2||44° 25′ 41.9″ N||63° 57′ 55.9″ W|
|0131:03.0||44° 25′ 21.6″ N||63° 57′ 53.8″ W|
|0131:07.8||44° 24′ 58.4″ N||63° 57′ 54.7″ W|
The scan rate for the radar was one revolution every 4.82 seconds. According to seismographic information, the aircraft hit the water at 0131:17.6. The impact time of 0131:17.6 ± 0.5 seconds was based on a wreckage position given as
- 44° 24.561′ N 63°58.425′ W
(which becomes 44° 24′ 33.6″ N 63° 58′ 25.5″ W)
The aircraft wreckage field was centred about 2 511 feet south and 2 228 feet west of the last radar hit.
The configuration at impact was determined to be wheels up, flaps 15 degrees, and slats retracted. The stall speeds chart from the AOM indicates, for this configuration, the following stall speeds:
|Aircraft weight||215 t||230 t|
|Level flight stall speed||160 kt||166 kt|
|60° banked turn stall speed||226 kt||235 kt|
An analysis of the aircraft's acceleration capabilities suggests that it probably did not overspeed to a structurally damaging extent. The average velocity over the aircraft's last seven radar hits was 264 knots. (The radar tabular print-out reports an airspeed of 240 knots for the last three primary radar hits, and 250 knots for the 10 hits before that; however, these speeds do not correlate particularly well with the changes in radar-established position.) Even in a vertical dive (which did not happen) from 1 100 feet at 264 knots there would be less than 2.5 seconds in which to accelerate before impact. This would lead to a lower impact velocity than an accelerated slant descent. Engine 2 was shut down within about one minute before the end of the flight, according to analysis of the Engine 2 FADEC. Engine 3 appears to have gone to near-idle. Conservatively assuming a take-off thrust for the remaining engine and a slant descent from 1 100 feet beginning at 264 knots at the last radar hit, the structural limit speed for the aircraft (350 knots IAS) would not have been exceeded.
The aircraft did not accelerate to a structurally damaging airspeed before it hit the water. Calculations suggest that it would be aerodynamically feasible for the aircraft to turn from its final radar heading to the crash site in the time available. The turn would have been steep.
A high-impact speed is suggested by the following factors:
- the extent of structural break-up, with many small pieces from the nose area and only a few moderately large pieces from well aft;
- the completeness of the breakup, with nothing escaping severe damage;
- the absence of any intact seats, windshield frames, or door frames; and
- the severe damage to all undercarriage, including the splitting of the nose gear cylinder by the compression of its oleo leg.
Some structural evidence points to an impact of about 15 degrees to the right of the aircraft centerline. A descending right turn would account for this and would explain the disparities in damage between the right and left engines and engine mounts.
Theoretical emergency descent profile
The theoretical emergency descent profile was developed using data from a simulation carried out by the aircraft manufacturer. The simulator profile was flown in a motion-based engineering simulator using the FAA-approved emergency descent checklist procedures for the MD-11. The profile did not take into account wind or any changes in aircraft weight or C of G that may have resulted from fuel dumping. The initial conditions were as follows:
|Gross weight||507 700 lb|
|C of G||.313% MAC|
|Flaps||retracted to 0 degrees|
|Speed brakes||fully extended|
|IAS||increased to maximum operating speedFootnote 4|
The simulated profile included a speed reduction from maximum operating speed to 270 KIAS while levelling off at approximately 10 000 feet to extend the landing gear. The descent was then re-initiated with a further reduction in speed. In descent through approximately 6 000 feet, the speed brakes were retracted and Flap 15 was selected. This was followed by a further levelling off at 3 000 feet with subsequent Flap 35 selection. Final descent was initiated with approach speed maintained at approximately 170 KIAS.
Actual wind information from the FDR was combined with engineering simulator data to calculate the aircraft's ground speed during the emergency descent and the earliest possible landing time. This ground speed was then mathematically integrated to derive displacement (or distance travelled over time). The upper winds were recorded once every 64 seconds based on inertial reference system information. These winds were derived to a higher resolution of once per second using recorded ground speed, drift, computed airspeed, and magnetic heading. The winds were then derived as a function of pressure altitude to determine the ground speed along the emergency descent vertical profile. Given the recorded winds, the ground speed attained was dependent on the track flown toward the Halifax airport, which in turn depended on where along the flight path the emergency descent profile would be initiated.
It was assumed that the track flown to the airport was the simplest possible, with the least lateral manoeuvring. Consequently, the profile assumed direct tracking to the GOLF NDB from the point of initiation, followed by a straight-in segment from the beacon to the threshold of Runway 06. To simplify the calculations, the time and distance for the turns (less than 20-degree turns) were ignored. By mathematically integrating the ground speed for the straight-in segment from the beacon to the runway, and considering the known distance of 4.9 nm from the beacon to the threshold, a beacon crossing time was estimated for the emergency descent.
For any point along the flight path, the displacement based on the derived ground speed between the time of initiation and the time of beacon crossing could be compared against the distance to the GOLF beacon based on the recorded inertial position from the aircraft's on-board inertial reference system. It was presumed that there would be only one point along the aircraft's flight path where the inertial distance to the GOLF beacon would be equivalent to the distance travelled along the emergency descent profile over the specified period. Initiation of the emergency descent at this time would result in the earliest possible landing with the least amount of lateral manoeuvring. Several iterations of this calculation were required to determine the geographical position and associated time.
The emergency descent profile intersects the accident flight profile at about 0114:18. This represents the optimal point at which to initiate a descent for the earliest possible landing without having to manoeuvre to lose altitude. It is coincidental that the theoretical best time to initiate a descent to land in the shortest time is only a few seconds after the Pan Pan call was made. An earlier descent farther away from the airport would have been premature and would have taken longer than the ideal theoretical descent point. Descending later (closer to the airport), meant that manoeuvring would have been required to lose altitude. It is noteworthy that the optimal time for an emergency descent into Halifax was more than a minute prior to the decision to accept Halifax as the preferred airport. Decisions regarding the time and rate of descent would have been influenced by the cues available to the flight crew.
The simulated emergency descent performed by the manufacturer's simulator indicated that it would have taken approximately 13 minutes and 8 seconds to descend from FL330 down to a landing and complete stop. It should be noted that the time to descend was independent of the wind conditions, which have an effect only on the distance required to descend. The simulator database did not contain Halifax airport; therefore, a different airport was used–one that had a field elevation of 15 feet (462 feet lower than that of the Halifax airport). The simulated data were adjusted for use in the above calculation, to approximate descent to the Halifax airport. With this adjustment, the time from FL330 to a runway threshold crossing height of 50 feet agl was approximately 12 minutes. This time segment was used in the mathematical integration. The earliest estimated threshold crossing time was, therefore, 0126:17. This was considered to be an approximate time because of the limitations of the analysis, as described above.
The landing distance data must include correction factors for wind, aircraft landing weight, airport elevation, and runway surface conditions. The influence of temperature, barometric pressure, and runway slope on the landing distance of the MD-11 is not accounted for, since it is small enough to be covered by the operational reserve (40% of the available runway length remaining).
The elevation of Halifax International Airport is 477 feet. Runway 06 is asphalt-covered, with an average slope of less than 0.08%, down, and has an available runway length 8 800 feet. The runway was dry.
At 0121, the flight crew of SR 111 indicated that the aircraft weight was 230 tonnes. The recommended maximum landing weight of the MD-11 is 199.58 tonnes; however, an overweight landing up to 218.4 tonnes is permitted under certain conditions. The AOM states that "no overweight landings are authorized if one of the following conditions exist":
- There is a tire failure;
- The runway is contaminated;
- There is a crosswind of over 20 knots;
- The slats are inoperative;
- There are split flaps;
- The hydraulics system is unserviceable;
- There are flight control troubles;
- There is a jammed stabilizer;
- The anti-skid is unserviceable;
- The reverser is unserviceable; or
- Two engines are inoperative.
The AOM further states that "in case of a declared emergency the PIC may take any action deemed necessary, i.e., disregard the limitations/restrictions stipulated above."
Using runway and atmospheric conditions at 0135 at Halifax International Airport, the landing distance requirement for a serviceable MD-11 on Runway 06 (or Runway 24) was calculated as follows. (According to the regulations, an aircraft should be capable of stopping within 60% of the required runway length.)
|Aircraft weight||Flaps 35 degrees landing||Flaps 50 degrees landing|
|Runway 06||Runway 24||Runway 06||Runway 24|
|199.58 t||7 875 ft.||8 630 ft.||7 060 ft.||7 800 ft.|
|218.40 t||8 530 ft.||9 310 ft.||7 875 ft.||8 575 ft.|
|230.00 t||8 860 ft.||9 660 ft.||8 200 ft.||9 000 ft.|
- Date modified: