Uncommanded autofeather and erroneous signals

Citation

(1998), "Uncommanded autofeather and erroneous signals", Aircraft Engineering and Aerospace Technology, Vol. 70 No. 4. https://doi.org/10.1108/aeat.1998.12770dab.023

Uncommanded autofeather and erroneous signals

Uncommanded autofeather and erroneous signals

Keywords Aircraft, Autofeather, Safety

A DH Canada DHC-8 series 311 aircraft took off from Aberdeen for Newcastle with the anti-ice systems selected on. There was a strong gusting crosswind and it was raining. The intial climb was normal and the flaps were retracted. The standard operating procedure is then to deselect the autofeather system and select climb power, however, the commander decided to retain take off power in order to expedite the climb through the layer of moderate to severe turbulence they were experiencing; he informed the first officer of his intention and asked for the bleed air to be switched on. The first officer actioned the after check take off from memory; however, he did not deselect the autofeather system because climb power had not yet been requested.

The commander selected climb power at about 2,200 feet amsl and asked the first officer to reduce the propellers to the climb setting. Both propellers were reduced to 900 propeller rpm (Np) and the torque increased momentarily to 100 per cent before being reselected to 80 per cent. The autofeather system was not deselected. No. 1 engine torque increased slowly from 80 per cent until stabilising at around 90 per cent. However, No. 2 engine torque increased rapidly to 221 per cent and then returned to 80 per cent over a period of 12 seconds. This was the first of many excursions of No. 2 engine torque; most were in the range 10 per cent to 20 per cent and continued for a further 100 seconds until the commander retarded No. 2 power level to about 60 per cent. ATC was told of the problem and the climb was continued with No. 1 engine torque at 90 per cent and No. 2 engine torque still at 62 per cent both propellers were at 900Np.

When the aircraft climbed out of icing conditions the commander selected the No. 2 engine intake bypass doors to closed. Shortly afterwards, as the aircraft climbed through about 7,400 feet amsl, No. 2 propeller suddenly feathered, reducing to 400Np and the torque on No. 2 engine rose by about 15 per cent. In this installation it is possible to feather the propeller blades without stopping the engine. Because the engine had not actually failed, fuel was still being scheduled to maintain the level demanded by the power lever and consequently No. 1 engine torque increased rapidly to a peak value of 120 per cent before being reduced to below 100 per cent. About 100 seconds later No. 1 engine torque was further reduced to 38 per cent, the flight idle value, and No. 2 engine torque was increased slightly to 85 per cent.

The Unscheduled Propeller Feathering drill was consulted and the drill calls for the shutdown of the engine associated with the feathered propeller. The flight deck indications were that No. 1 engine parameters were normal. Because he now thought they had two, apparently unrelated problems, the commander decided not to shut down No. 1 engine. An emergency was declared and the aircraft diverted to RAF Leuchars.

The first officer tried to unfeather No. 1 propeller using the alternate system but was unsuccessful. As the aircraft passed about 7,000 feet amsl and during descent to Leuchars, the commander selected the No. 1 engine bypass door to closed and reduced the No. 2 engine torque to about 58per cent.

This disarmed the autofeather system and No. 1 propeller suddenly unfeathered; Np increased rapidly to 1,200 and the torque fell to zero. There was a yaw to the right of around 2 to 2.5° as the propeller unfeathered and it appears that the commander interpreted this as a failure of the No. 2 engine because, about 10 seconds later, he retarded No. 2 condition lever to the starter/feather detent which caused the propeller blade angle to coarsen towards the feather position. The power lever remained as selected and No. 2 engine torque increased rapidly to 100 per cent before reducing to 55 per cent as the lever was subsequently retarded to flight idle. After five seconds No. 2 engine torque reduced to zero in response to the condition lever being moved fully back to fuel off.

The first officer immediately told the commander that he through he had taken the wrong action and the latter quickly moved the condition lever fully forward to the max governing. An uneventful landing was made at Leuchars.

The autofeather system

The autofeather system is selected by a single switchlight on the lower right engine instrument panel; it is used only for take off. When the switchlight is pressed a green select caption illuminates to indicate that the system is selected. The system is armed when the power levers are advanced and both engine torques are above about 38 per cent; an amber armed caption illuminates in the bottom half of the switchlight.

The autofeather controller receives inputs from potentiometers on the power levers which provide power lever angle (PLA) information and also outside air temperature values from the digital air data computer. From this a discreet PLA "hi" or "lo"sgnal is output to a torque signal condition unit (TSCU) associated with each engine (PLA "hi" equivalent to a torque demand of 60 per cent or greater for a correctly rigged system). When the crew select Autofeather, each TSCU becomes enabled when it is in receipt of its own (local) PLA "hi".

Each TSCU also receives actual torque information from one of two sensing coils on its associated engine, Values of torque greater than 38 per cent result in a local torque "hi" discreet being generated. This is also fed to the remote TSCU and when a TSCU is in receipt of both local and remote torques "hi" signals, its autofeather status changes from "enabled" to "armed".

Transition to "armed and failed" occurs when local torque drops below 22 per cent on the correct standards of TSCU applicable to this specific aircraft. The affected TSCU will immediately send an uptrim signal to the remote engine electronic engine control (EEC) to uptrim its torque by about 10 per cent and a green uptrimadvisory light will illuminate on the engine control panel. If this condition lasts uninterrupted for 3 seconds, the TSCU transitions again to "not armed and feather" and the propeller is feathered. The TSCU is latched in this condition unless the Autofeather selector is switched off, or both PLA "hi" discreets are removed, ot the remote torque "hi" is removed.

Whilst one coil in each torque sensor provides information to the TSCU for autofeather purposes, the other is effectively a separate channel supplying torque data to the ARINC 429 data bus, where it is used by the cockpit instrument display and by the DFDR. In addition, it is used by the EEC for closed-loop power calculations for 60 seconds after uptrim occurs.

Under normal operation, the EEC propeller speed governing logic would adjust the engine fuel schedule to prevent propeller operation below the minimum speed of 785rpm. When a propeller feathers a cancel signal from the TSCU overrides this governor to prevent the possibility of propeller/engine overtorque as the propeller slows to below the minimum speed. However, should the propeller feather with the engine operating and the power lever at a high-power setting, an overtorque siuation will still occur. Should the TSCU cancel signal fail, a back up signal from the torque gauge is initiated when the overtorque limit of 115 per cent is reached.

An auxiliary feathering pump is energised and provides supplementary oil pressure for 18 seconds before automatically shutting off. Once the autofeather sequence has been initiated on one propeller, an interlock prevents autofeather of the other propeller.

The alternate leathering system provides a back up means of manually feathering/unfeathering a propeller if the condition lever fails to achieve full propeller feathering/unfeathering. Provided the power lever is at or above flight idle and the condition lever in start/feather, the alternate system can be used regardless of whether or not the engine is running. If the autofeather system is armed the alternate system will also initiate an uptrim.

The operator was in possession of two notifications from Bombardier concerning uncommended autofeather and opposite engine uptrim effects. The first of these was in 1990 and was in IN-Service Activity Report (ISAR) and mentioned that four cases of uncommended autofeather were being investigated by the manufacturer. In three of the four cases, there had been a loss of torque indication similar to that experienced on this specific aircraft (but on the autofeathered engine itself), with the display showing zero analogue and dashed digital display. This was explained as being due to an overtorque condition which, when it reached 135 per cent would be interpreted as an out-of-range situation by the indication system. The second notification, also an ISAR, was in 1995. This reported a case whereby an apparently faulty torque probe led to intermittent low torque signals to the TSCU and caused associated uptrims of the opposite engine. The ISAR presented information on the symptons to assist troubleshooting on both series 100 and 300 aircraft, again pointing out that a zero/dashes torque indication could result if the propeller actually does feather on engines programmed with early standard EEC software. For later standards of software the indication was said to read maximum on the analogue and 199 per cent on the digital display. Also mentioned was the fact that provision of separate torque channels on the series 300 aircraft could result in a normal torque indication to the crew on an engine which is experiencing a varying torque signal on the control channel.

Engineering investigation and analysis

From the information provided by the crew and analysis of the FDR data, it was clear that the fluctuations in torque of No. 2 engine which precipitated the subsequent events were caused by uptrim commands due to erroneous low engine No. 1 torque signal being received by its EEC. These signals were not reflected either in the cockpit instruments or the data recorded on the FDR which are fed by a separate channel which did not receive the erroneous signals. It was also clear that the spurious signals were intermittent, since they did not initially result in a feather command to No. 1 propeller which requires the signal to be present for a minimum of three seconds. However, the uptrim command to No. 2 engine EEC was received immediately each time the low torque signal was sensed. It is for this reason that the first uptrim command exceeded the intended nominal 10 per cent increase by a factor of about three, since the first intermittent signals were received in such quick succession that the EEC "ramped" up the command each time until the limit of 107 per cent torque was reached. According to Pratt & Whitney Canada (PWC) the peak figure of 111 per cent recorded was due to overshoot. The subsequent excursions in torque were more in line with the nominal uptrim schedule.

The No. 1 and No. 2 torque sensing probes and No. 1 TSCU were despatched to PWC for testing and examination. No significant defects were discovered but the TSCU was found to be of an inappropriate type for use on the PW123 engines fitted to this specific aircraft, but repeated testing did not reveal any faults or anomalies with the unit. According to PWC, fitment of this incorrect model of TSCU to the particular modification standard was probably not significant in this incident since the principal difference in performance of the unit lay in the trigger torque for auto-feather of a "failed" engine being increased from 22 per cent to 29 per cent. Since the reason for the erroneous triggering signal probably lay in an intermittent open or degraded circuit, the TSCU may have been sensing effectively zero or very low torque and the variation in trigger point therefore was probably of minimal significance.

It was suspected at an early stage that the reason for the spurious autofeather signals probably lay in intermittent poor contact of the connections within the engine nacelle associated with the autofeather system. Such poor contacts would be difficult to trouble-shoot on the ground because they could be influenced by vibration and transient moisture contamination. Examination of the TSCU did show signs of light, greasy contamination of some connector receptacles, although it was not possible to state categorically that this was responsible for the malfunction. Bombardier have, however, pointed out that the AFM had been revised to require deselection of autofeather prior to reducing power after take off because the reduced propeller rpm results in a reduced voltage output from the torque sensor. Hence a torque signal circuit degraded by, say, contamination of a connector would be more vulnerable to generating a spurious low torque signal at this time. It is for the same reason that Service Bulletin 21456 was introduced to help overcome such problems.

Arising from previous incidents involving spurious low-torque signals, PWC has developed two modifications and a maintenance practice to minimise the problem and has advocated to the operator the following:

• •

  Service Bullein (SB) 21456. This modification introduced a revised torque sensor shim which optimises the air gap in the sensor, increasing its signal strength and hence reducing its sensitivity to noise caused by vibration. This modification was not embodied on the operator's fleet.

• •

  SB 21463. This modification introduced a new engine wiring harness having improved connector sockets which are less susceptible to fretting damage, have better locking retention and improved sealing against moisture ingress. This modification was also not embodied on the operator's fleet. In addition, the Dash 8 Maintenance Review Board had required periodic removal and cleaning of connectors and retorquing and re-taping of connectors.

Embodiment of SB 21463 obviated the above requirements, which the operator was not performing.

ReferenceAAIB Bulletin (1998), Vol. 2.