US Air 5050
Accident Case Study
Reading the NTSB probable cause we are told the captain failed to "exercise command authority in a timely manner to reject the takeoff or take sufficient control to continue the takeoff, which was initiated by a mistrimmed rudder." Let's translate that into plain pilot-speak. The pilot got ample clues that something was wrong before V1 and should have aborted the takeoff. The need to abort was caused by his inadequate rudder inputs, placing more strain on the nosewheel tires than they could handle, causing one of those tires to separate from the rim. But the airplane was perfectly flyable even with the failed tire and mistrimmed rudder. He was a former Air Force pilot where the mantra beaten into our heads was "first, fly the airplane." Had this pilot done that, the aircraft would not have been destroyed and the two fatalities would have made it to their destination alive.
Figure: USAir 5050 N416US, from NTSB/AAR-90/03, Figure 5.
- Date: 20 September 1989
- Time: 23:21
- Type: Boeing 737-401
- Operator: USAir
- Registration: N416US
- Fatalities: 0 of 6 crew, 2 of 57 passengers
- Aircraft Fate: Destroyed
- Phase: Takeoff
- Airport: (Departure) New York-La Guardia Airport, NY (LGA/KLGA) United States of America
- Airport: (Destination) Charlotte-Douglass Airport, NC (CLT/KCLT), United States of America
- During the initial interview with the Safety Board, the flight crew described starting the engines and taxiing out to runway 31 as uneventful. They said six to eight airplanes were ahead of them on the taxiway awaiting takeoff clearance. Two minutes after push-back, the ground controller told the crew to hold short of taxiway GOLF GOLF. However, the captain failed to hold short of that taxiway and received modified taxi instructions from the ground controller at 2256. The captain then briefed takeoff speeds as V1: 125 knots, VR: 128 knots, and V2: 139 knots. The captain had flown the BWI-LGA segment, and the first officer was to be the flying pilot on the LGA-CLT segment. No company or Federal regulations govern flying pilot choices. As the flying pilot, the first officer's departure briefing consisted of his reciting to the captain his turn and altitude clearance and the LaGuardia3 departure clearance.
- About 2 minutes later, the first officer announced "stabilizer and trim" as part of the before-takeoff checklist. The captain responded with "set" and then corrected himself by saying: "Stabilizer trim, I forgot the answer. Set for takeoff." According to USAir's B-737-300/400 normal procedures checklist, "set for takeoff" was the correct response, although the captain's words "stabilizer trim" failed to restate the correct challenge. The captain said during the public hearing that he had no specific recollection of checking trim settings on the accident flight but that his normal procedure would be to do so. The first officer said during the hearing that he did not check the trim settings himself while he was running the checklist during taxi-out. USAir procedures did not require him to do so.
- The last item on the before-takeoff checklist was AUTO-BRAKE. When challenged on this item, the captain responded "is off," and the first officer called the checklist complete.
- Flight 5050 was cleared into position to hold at the end of the runway at 2318:26 and received takeoff clearance at 2320:05. The cockpit voice recorder (CVR) disclosed the sound of increasing engine noise, and shortly thereafter the first officer pressed the autothrottle disengage button instead of the takeoff/go-around (TO/GA) button. He later said that he then pressed the TO/GA button, but noted no throttle movement. He then advanced the throttles manually to a "rough" takeoff-power setting. The captain then said: "Okay, that's the wrong button pushed" and 9 seconds later said: "All right, I'll set your power." The captain later said he thought he had rearmed and reengaged the autothrottles and had advanced the throttles to the N1 target setting of 95 percent while depressing the TO/GA button. The first officer later explained that "I'll set your power" meant to him that the captain was "fine-tuning" the setting to takeoff power. Both crewmembers agreed that the airplane then began tracking to the left during the takeoff roll. About 18 seconds after beginning the roll, the CVR recorded a "bang" followed shortly by a loud rumble. The captain later said that during this time the airplane continued tracking to the left and that he was becoming concerned about the unidentified bang and rumble. The first officer later said he believed he had stopped the leftward tracking and the airplane "began to parallel the runway centerline."
- At 2320:53, the CVR recorded the captain saying "got the steering." The captain later testified that he had said, "You've got the steering." The first officer testified that he thought the captain had said: "I've got the steering." When the first officer heard the captain, he said "Watch it then" and began releasing force on the right rudder pedal but kept his hands on the yoke in anticipation of the V1 and rotation callouts.
- At 2320:58.1, the captain said: "Let's take it back then" which he later testified meant that he was aborting the takeoff. According to the captain, he rejected the takeoff because of the continuing left drift and the rumbling noise; he said that he used differential braking and nose wheel steering to return toward the centerline and stop. The sound of throttle levers hitting their idle stops was recorded at 2320:58.4. According to data from the digital flight data recorder (DFDR), indicated airspeed at that time was 130 knots. The sound of engine noise decreasing was recorded at 2321:OO.9. The first officer then told the tower about the rejected takeoff. In later testimony, the first officer said that he was unaware of the reason for the captain's decision to abort.
- Increasing engine sound indicating employment of reverse thrust was heard on the CVR almost 9 seconds after the abort maneuver began. The airplane did not stop on the runway but crossed the end of the runway at 34 knots ground speed. It came to rest in the water supported by the pier that holds runway 13's approach lights. The sound of impact was recorded at 2321:21.9.
- Both pilots agreed that the farthest the airplane tracked to the left during the rejected takeoff (RTO) was about halfway between the centerline and the left side of the runway; both said that during the RTO they thought the airplane could be stopped on the remaining runway. Neither pilot could recall noting the airspeed at initiation of the RTO, and the CVR recorded no standard airspeed callouts.
The crew had been delayed several times during the day and fatigue could have been a factor, but they were well within duty day limits. The NTSB report makes note of a few abrasive events between the captain and dispatchers, but those do not appear to be factors. A Pan Am pilot who visited the first officer in the cockpit might have bumped the rudder trim, but he denies that. In any case, while the misstrimmed rudder was a factor, it could have been easily overcome.
Pilots who make a point of articulating checklist challenges and responses exactly as given in the checklist are often accused of being too obsessed with minutiae but this mishap demonstrates the value of being detail oriented. A pilot who simply remembers this item as "stabilizer trim" can eventually forget there are more trim settings to check. The "and" in this item is crucial, especially given the fact the rudder trim on this airplane is easy to move unintentionally.
"Old school" pilots tend to mistrust "new school" technology and when autobrakes were first introduced many pilots feared they would be too aggressive when not needed. This mistrust was unfounded and had the autobrakes been used on this takeoff, the mishap may have been avoided.
The bang was one of the nosewheel tires separating from the rim and should have been a reason to abort.
There is no room for slang, jargon, or lazy terminology during the takeoff. As this mishap shows, there is too much room for interpretation that can lead to critical mistakes.
This was 5 knots above V1 but given the runway conditions the abort should have been successful.
[NTSB/AAR-90/03, ¶1.7.1] The Weather Service Contract Meteorological Observatory at LaGuardia issued the following observation: Time--2334; type--local ; ceiling--estimated 500 feet overcast; visibity--5 miles; weather--light rain and fog; temperature--73°F.; dewpoint--73°F.; wind--210 degrees at 4 knots; altimeter--30.20 inches; remarks--aircraft mishap.
Figure: USAir 5050 Rejected Takeoff Braking Levels, from NTSB/AAR-90/03, figure 2.
[NTSB/AAR-90/03, ¶1.10] Based on runway surface characteristics, crown, and texture depth values, the drainage analysis indicates that runway 31 had excellent water drainage capability (excluding the painted nongrooved blast pad [overrun] area) particularly for the rainfall rate occurring at the time of the accident (.008in./hr.). With the transverse grooving, and the good tread condition of the main gear tires, hydroplanning was not a significant factor. Except for approximately 1,200 feet near runway 4/22 intersection, the accident aircraft RTO track was established by the white tire erasure marks found on both the asphalt and concrete deck surfaces of the runway. In terms of tire friction performance, the Saab friction tester results indicate that the lateral displacement of the aircraft left of centerline actually provided better friction performance compared to that measured closer to runway centerline.
It appears the captain did not use all off the braking capability he had available.
Figure: USAir 5050 Cockpit aisle stand, from NTSB/AAR-90/03, figure 6.
[NTSB/AAR-90/03, ¶1.16.2] On-scene examination of the rudder trim system showed that the rudder trim control knob rotated freely in both directions with no evidence of binding or sticking. This control is located on the aft end of the center pedestal. The trim indicator, which is also on the center pedestal, showed an "off" flag, and there was no evidence of sticking. Extension of the rudder trim actuator rod was 10.5 inches, corresponding to a trim position of 16° left deflection of the rudder.
[NTSB/AAR-90/03, ¶18.104.22.168] Regardless of the trim setting the pilots' force on the rudder pedals can still move the rudder to its full deflection of 26° either side of center.
The misset trim should not have been an issue. The pilot's job is to position the aircraft where it is supposed to be.
[NTSB/AAR-90/03, ¶1.17.5] The Safety Board collected about 90 reports of rudder trim anomalies for the Boeing 737-300/400 series airplane. [ . . . ] Many reports described the inadvertent setting of rudder trim by the foot of a jumpseat occupant behind the captain's seat. The reports imply that casual visitors to the cockpit did not strap in, sat sideways and used the end of the center pedestal as a footrest for their right foot. This allowed their shoe sole to push the trim knob counterclockwise and set left rudder trim. The Pan American captain who had visited the cockpit before departure of flight 5050 said that he did not rest his foot on the center pedestal at any time. [ . . . ] Other reports show that placing objects on the center pedestal can inadvertently turn the rudder trim knob. While the Pan American captain said that he believed the first officer placed his chart holder on the pedestal during prestart activities, the first officer did not recall having done so.
[NTSB/AAR-90/03, ¶2.2] Data from the DFDR showed that rudder trim on N416US was neutral after arrival at the USAir gate at LGA. The DFDR subsequently shut down. When repowered after engine start for flight 5050, the DFDR showed that the rudder had moved to the full left trim position. Electrical and hydraulic power from the Auxiliary Power Unit (APU) were available during the intervening period to change the rudder trim position, if commanded. The time to run trim from neutral to full left is about 30 seconds, so momentary knob rotation would not have produced full left trim.
- The captain could have noticed the misset rudder trim almost immediately after engine start, even before he began taxiing away from the gate, because the rudder pedals were offset from each other by 4 1/4 inches. He did not mention the offset to investigators 2 days after the accident. However, after the DFDR evidence indicated the trim anomaly, he said that he had noticed the offset pedals, adding that the offset did not bother him because he was used to taxiing with offset pedals in the C-130.
- Although the captain had more than 2,600 flight hours in the B-737, he had only about 140 hours in the left seat. Since taxiing is performed only from the left seat, his taxi experience was somewhat limited. However, the full rudder trim would also have turned the nose wheel about 4° left.
- According to Boeing's Chief B-737 Test Pilot, the nosewheel steering is effective only until the rudder gains aerodynamic authority. Afterward, the nosewheel tires cannot produce sufficient cornering force to redirect the airplane in opposition to the aerodynamic force resulting from rudder deflection. Using Flight 5050's gross weight, an engineering simulator showed that, at more than 81 knots on a dry runway and 64 knots on a wet runway, nosewheel steering alone could not turn the airplane in opposition to 16° of rudder deflection.
- The captain testified that he tried to halt the leftward track of the airplane by using both rudder pedal and the nosewheel steering tiller prior to rejecting the takeoff. But the DFDR data refutes such an occurrence because it indicates a maximum rudder deflection of only 1 degree right during the 4 l/2 seconds from "got the steering" to the captain's signaling his rejection of the takeoff by saying "let's take it back."
- Although 1° nose-right rudder required about 58 lbs on the right rudder pedal, neither pilot applied the 71 lbs. of force needed for full right rudder.
We often think about VMCG (Minimum control speed - ground) as the speed where we no longer need nosewheel steering to keep the airplane on the runway following an engine failure. Another way to look at it is the speed where the rudder will do the job for us. Something we don't often think about is at what point is the nosewheel steering no longer effective on its own. But underlying it all, is the fact we should be using aerodynamic controls as soon as they are effective.
The captain was applying significant pressure on the nosewheel to turn the airplane right just as airplane's large vertical fin was pushing it in the other direction. The rudder would have been a much more effective tool.
- The captain of flight 5050 had several indications of a problem during the takeoff roll before the airplane reached 100 knots. First, he must have been aware of the first officer's difficulty in maintaining runway heading as more and more nosewheel steering commands were applied by the tiller; second, the sound of the "bang" occurred at 62 knots and the subsequent rumble was heard at 91 knots. The Safety Board believes that the captain should have decided to reject the takeoff immediately. Having failed to do so, he must have been aware that the airplane was accelerating and rapidly approaching the V1 speed, even though he failed to make the 80 knots and V1 callouts. With such awareness, the captain should have given his total attention to control of the airplane with the rudder pedals and continued the takeoff.
- Either pilot was physically capable of but did not use substantial right rudder to maintain directional control. The Safety Board concludes that because the pilots had full rudder authority, a safe takeoff was possible, and that the pilots could have corrected the mistrim condition after lifting off. Successful takeoffs have been accomplished four times with full rudder trim and five times with partial rudder trim, according to reports from B-737-300/400 pilots located during this investigation.
- The Boeing calculations showed that a B-737-400 should have stopped after an RTO initiated at a 125 knot V1 speed using 4,050 feet on a dry runway and 5,670 feet on the "l/2 dry" runway, both without consideration for reverse thrust. That flight 5050 failed to stop on a 7,000 foot runway is of concern to the Safety Board.
- The spoilers serve two purposes: to increase drag for deceleration and to place weight on the tires for braking by reducing wing lift. If the pilot does not extend the spoilers, automatic deployment comes with reverse thrust. Flight 5050's DFDR data shows that the thrust reversers unlocked 5 l/2 seconds after the rejected takeoff started and about 4,800 feet from the beginning of the runway. The captain testified that he could not remember extending the spoilers or if the selection of reverse thrust automatically extended them. Spoiler position was not recorded on the DFDR. Thus, the Safety Board could not determine whether late deployment of the spoilers delayed the attainment of full braking force.
- . . . the DFDR data showed that the maximum deceleration was not achieved until 5 l/2 seconds after the initial RTO action was taken, whereas the AFM data assumed an increment of only 1 second from brake application to achieve maximum deceleration. In analyzing the captain's performance, the Board assumed a reaction time of 2 l/2 seconds from brake application to achievement of maximum deceleration as being reasonable. The Safety Board's study indicated that the additional 3 seconds of delay added 786 feet to the theoretical stopping distance required.
- The captain chose not to use the RTO feature of the autobrake system during the accident takeoff run, despite recommendations from both Boeing and USAir. [ . . . ] The Board has heard the argument that autobrakes may throw passengers against seatbacks during low-speed aborts. This logic is spurious since the B-737-400's autobrakes are not active until ground speed reaches 90 knots. Moreover, serious injuries and deaths during unsuccessful high-speed aborts far outweigh the few, if any, minor injuries that might occur during low-speed rejected takeoffs.
The lackadaisical manner in which the abort was called, "Let's take it back then," seems to have carried on to the manner in which the abort was executed. The reversers, brakes, and spoiler steps are executed simultaneously. The only variation permitted is if the spoilers are not automatic, in which case they can be extended by the other pilot or a split second before the reversers. But a 5 1/2 second delay in unacceptable.
Autobrakes may have saved this crew, had they been used. The captain seemed unprepared to apply maximum braking, the electrons would not have been so unprepared.
- The tire marks on the runway suggest that the captain's continued attempt to steer using the nosewheel caused the "bang" and rumble noises that prompted the RTO. Rumbling began when Flight 5050 reached 95 knots ground speed 1,736 feet from the start of runway 31. The "bang" was most likely caused by the left nosewheel tire suddenly coming off the rim allowing the air to escape violently.
- The CVR shows that 23 seconds elapsed after the takeoff started before the rumble started. During this period, the rudder was deflected to the left much of the time. The Safety Board believes that Flight 5050 stayed on the runway, instead of running off the left side of the runway, because the captain was overpowering the rudder by commanding the nosewheel to steer right with the tiller. Erasure marks on the runway and damage to the nose tires confirm this.
If true, the nosewheel was not only countering the leftward wind-pushed vertical fin, but leftward rudder pressures as well.
[NTSB/AAR-90/03, ¶2.10] The Safety Board views the absence of a comprehensive departure briefing, the absence of airspeed callouts, the failure of the first officer to clearly communicate his directional control problem, and the non-assertive manner in which the captain communicated his intent to reject the takeoff as indications of poor cockpit coordination. That the pilots of flight 5050 were ineffective as a team is probably the result, in part, of their lack of any formal training on cockpit resource management (CRM). Both of the flight crewmembers were hired and trained by Piedmont Airlines before the merger with USAir had been completed. Piedmont Airlines did not provide formal CRM management training to either pilot involved in this accident.
[NTSB/AAR-90/03, ¶3.2] The National Transportation Safety Board determines that the probable cause of this accident was the captain's failure to exercise his command authority in a timely manner to reject the takeoff or take sufficient control to continue the takeoff, which was initiated with a mistrimmed rudder. Also causal was the captain's failure to detect the mistrimmed rudder before the takeoff was attempted.
NTSB Aircraft Accident Report, AAR-90/03 Aircraft Accident Report, USAir, Inc., Boeing 737-400, La Guardia Airport, Flushing, New York, September 20, 1989