The aircraft was cruising at flight level (FL) 350 with the autopilot engaged and 35,000 ft pre-selected in the altitude display on the autopilot mode control panel (MCP). The crew observed the aircraft make an uncommanded climb from FL350 to FL351, where it remained for approximately 20 seconds until the autopilot returned the aircraft to FL350. The crew then noticed that the `Mach Trim Fail' warning light had illuminated. The `Mach Trim Fail' light extinguished for a short time, re-illuminated, then extinguished again. During this sequence of events, the crew reduced the aircraft speed from Mach 0.76 to Mach 0.73 in accordance with the appropriate Quick Reference Handbook (QRH) procedure. Shortly after, the `Stab Out Of Trim' warning light illuminated. The `Stab Out Of Trim' warning extinguished after descent was commenced.
The descent was initiated using the autopilot and was normal until approximately FL250 when the aircraft pitched down steeply and the airspeed increased rapidly. The autopilot was disconnected, and the crew flew the aircraft manually, noticing that a considerable amount of back trim was required to stabilise the descent at 300 kts. The aircraft was flown manually for the remainder of the descent, and the aircraft was landed safely without further incident.
After landing, the pilot in command notified maintenance support of the in-flight problems that had been encountered. However, he did not lodge an incident report with the operator until 2 days after the occurrence. Upon receipt of the incident report, the operator took action to recover the aircraft's digital flight data recorder (DFDR). However, the DFDR was found to be faulty, and information from the flight was not available. The operator tested the DFDR and reported that the recorder would intermittently return the recording head to a random section of tape, resulting in existing data being overwritten. In order to obtain data about the aircraft flight path, air traffic control radar information was examined during the investigation.
The operator also reported that during the subsequent maintenance investigation, the K11 relay in the autopilot accessory unit was found to have an intermittent fault when it was subjected to initial testing. However, further testing was unable to reproduce the fault that was experienced in flight, and both the K11 and K6 relays were replaced as a precaution. The "Mach Trim Actuator" was also tested by the operator. The tests were conducted under various temperature conditions but they revealed no indication of defect. The manufacturer was also unable to fault the actuator. The aircraft manufacturer suggested that the operator verify sections of the aircraft wiring and perform the Mach trim and stabiliser rigging checks specified in the Boeing 737-300 Maintenance Manual.
The automatic flight system (AFS) of Boeing 737-376 aircraft comprises the autopilot flight director system (AFDS) and the autothrottle (A/T). Management of the vertical and lateral navigation of the aircraft is controlled by pilot selection of appropriate AFDS MCP settings in conjunction with the flight management computer (FMC). Normally, once an aircraft is airborne and the AFS engaged by the crew, the AFDS and A/T are controlled automatically by the FMC to fly an optimised lateral and vertical flight path throughout the climb, cruise and descent phases of a flight.
Data entry into the FMC is made by the crew on the flight management system control display unit (CDU) fitted to the cockpit centre stand. The CDU displays various "pages" of information that are required for the management of the flight. The crew will enter relevant data into these pages, including the intended cruise level.
Once an aircraft reaches the planned cruise level, it accelerates from the climb speed to the cruise speed. During acceleration, rearward movement of the wing aerodynamic centre of pressure results in a tendency for the aircraft to pitch nose down in a phenomenon known as "Mach Tuck". The Boeing 737-376 is equipped with a Mach Trim System (MTS) to provide improved stability at the high altitudes and airspeeds typically encountered during the cruise phase of flight. The MTS functions automatically at speeds in excess of Mach 0.615 and causes elevator adjustment as speed increases. Failure of the MTS is indicated by the illumination (amber) of the Mach Trim Fail light, and the crew response is to limit the aircraft speed to below Mach 0.74.
During cruise and with the autopilot engaged, the AFS causes the horizontal stabiliser to be trimmed to a position which will ensure that the pitch of the aircraft maintains the required cruise level. The Stabiliser Out Of Trim (Stab Out Of Trim) light illuminates amber to alert the crew if the autopilot is not trimming the stabiliser properly and that the elevator position exceeds a certain value in relation to the stabiliser and/or that the elevator is positioned too far away from the neutral position. The crew response for a "Stab Out Of Trim" amber warning is to disengage the autopilot and re-trim the stabiliser. The autopilot may then be re-engaged as required.
The investigation was unable to determine the exact point at which the K11 relay failed. However, the most likely point of failure of the K11 relay was just prior to commencement of the uncommanded climb from FL350 to FL351. With the aircraft cruising at Mach 0.76, the MTS would have commanded approximately 2.8 degrees "up" elevator. The amount of "up" elevator commanded by the MTS is regulated by feedback of the airspeed signal from the air data computer. Failure of the K11 relay would result in open feedback to the MTS and, without airspeed signal feedback, the actuator would have commenced to drive the elevator up at approximately 0.4 degrees per second until the elevator reached 4.2 degrees "up". The 4.2 degrees "up" elevator corresponds to a computed speed of approximately Mach 0.83 while the K11 relay is operating without fault.
With 4.2 degrees "up" elevator resulting from the failed K11 relay, the autopilot would have been unable to maintain the preselected level of FL350, and the uncommanded climb to FL351 would have resulted from the incorrect "up" elevator setting. It is probable that the Mach Trim Fail light illuminated at this point due to the elevator's position relative to the stabiliser. At the same time, it appears likely that the autopilot commenced to trim the stabiliser "down" to a position that would counter the failed Mach trim position and also to regain FL350. The Mach Trim Fail light would have extinguished once the stabiliser was in a "neutral" position to the out of trim elevator. It is likely that the Stab Out Of Trim light would also have been illuminated during the period that the stabiliser was out of position with the elevator.
When FL350 had been regained, the crew responded to the illumination of the Mach Trim Fail light by reducing the aircraft's speed to Mach 0.73. As speed reduced, the autopilot would have trimmed the stabiliser "up" to maintain FL350. The retrimmed position of the stabiliser for Mach 0.73 would have resulted in it no longer being in a "neutral" position with respect to the out-of-trim elevator. This would have caused the Mach Trim Fail light to re-illuminate in conjunction with the Stab Out Of Trim light at this time.
The aircraft continued the cruise segment with the elevator in an incorrect "up" position as a result of the failed K11 relay, opposed by the stabiliser also in an incorrect position, thus nulling the two conflicting control inputs.
Descent commenced by the AFDS commanding the A/T to close and the autopilot maintaining the reduced speed of Mach 0.73. The crew reported that at approximately FL250, the aircraft pitched down steeply and the airspeed rapidly increased. Examination of air traffic control radar information revealed that the aircraft's vertical profile appeared normal until approximately FL270. From that level, the aircraft's ground speed gradually increased from 386 kts to 399 kts at FL240, then reduced to 394 kts at FL 210. At the same time, the rate of descent increased from approximately 3,000 ft/min to a peak of 5,143 ft/min between FL240 and FL235, then stabilised again at 3,000 ft/min. This information corresponds with the crew's recollection of the events that took place.
The likely explanation for the pitch down was that the K11 relay resumed normal operation at about this point, permitting airspeed feedback to be re-established to the MTS. This would have resulted in a reduction of the 4.2 degrees "up" elevator to the correct position corresponding with the aircraft's actual speed of Mach 0.73. With the elevator re-established in the correct position for the aircraft speed, the incorrect position of the stabiliser resulted in the pitch-down event.
Because the fault in the autopilot accessory unit did not recur after replacement of the K11 and K6 relays, the operator did not consider that the Mach trim and stabiliser rigging checks as called out in the Boeing 737-300 Maintenance Manual were warranted.
The incident occurred on 24 April. However, the pilot in command did not report the incident to the operator until 26 April, and the Lockheed Model 209 DFDR fitted to the aircraft was not removed for examination until 27 April. Upon its removal, the DFDR was found to be faulty. The Lockheed Model 209 is an early 1970's design electro-mechanical flight data recorder. It records data from the last 25 hours of aircraft operation onto a magnetic tape; the tape contains six tracks, and data is recorded on one track at a time. Each track contains approximately 4 hours and 10 minutes of data. Track switching is sequential to ensure that the last 25 hours of data are continuously recorded. When power is removed at the end of a flight, the DFDR "remembers" which track was being used for recording. When power is re-applied at the start of the next flight, the DFDR resumes recording on the correct track.
On readout, data recorded during the incident flight was not observed in the recorded data while some data from 22 April was observed. It was apparent that the recorded data was not continuous. A strip-report on the DFDR revealed that the DFDR did not "remember" which track was active and "jumped" to track 3 when power was re-applied at the start of some flights. Consequently, while the data for the current flight was correctly recorded, data from earlier flights (but within the 25-hour nominal recording duration) could be overwritten. This is a known problem with the Lockheed Model 209 DFDR.
The K11 relay in the flight control accessory box experienced an intermittent failure.
Local safety action
As a result of this occurrence, the operator advised it had issued a Flight Standing Order to pilots, outlining required reporting procedures following an air safety incident.
|Date:||24 April 1999||Investigation status:||Completed|
|Time:||1830 hours CST|
|Location:||167 km ENE Adelaide, (VOR)|
|Release date:||12 January 2000||Occurrence category:||Incident|
|Report status:||Final||Highest injury level:||None|
|Aircraft manufacturer||The Boeing Company|
|Type of operation||Air Transport High Capacity|
|Damage to aircraft||Nil|
|Departure point||Sydney, NSW|
|Departure time||1730 hours CST|