An A300-600 uncommanded yaw event

The 03 Dec 02 AA Flight 068 Event

AC 068 AA PIR.84 SYS 2701

MDIS.03DEC/JFK T-0647 EMPL 089838

DURING DEPARTURE CLIMB, AIRCRAFT ENCOUNTERED

TURBULENCE AT ABOUT 300 FEET UP TO ABOUT 1200 FEET,

AND THE AIRCRAFT YAWED SEVERAL TIMES. AIRCRAFT WAS

BEING HAND FLOWN, WITH SLATS EXTENDED WITH TAKE OFF

FLAPS SET. NO RUDDER TRAVEL FAULT INDICATED. NO YAW

DAMPER DISCONNECTS. AIRCRAFT FLEW NORMALLY ON

AUTOPILOT AFTER THE EVENT. AIRCRAFT FLEW NORMALLY

BEING HAND FLOWN. THE AIRCRAFT FELT AS IF IT WAS

PURPOSEFUL YAW INPUTS. NO RUDDER PEDAL MOVEMENT WAS

NOTED. AIRSPEEDS WERE ABOUT 200 TO 225 KTS IAS.

0682700E7/1849.03DEC.JFK KQ NBR 3

AC 068 AA TFI. TWD 03DEC02 DTG 0 SYS 2720

PRI 1

MDIS.03DEC/TUL **TECH LIST** EMPL 159064

CREW REPORTS UNCOMMANDED RUDDER MOVEMENT.

ATBT.03DEC/TUL NAME R. B. KNOX EMPL 159064

1 PERFORM ENGINEERING BILL OF WORK. IF UNAVAILABLE

2 CONTACT TECH. SERVICES FOR A COPY.

3 BE SURE TO CONTACT ENGINEER DAVID SERATT AT

4 918-641-5180 AND REPORT RESULTS.

0682700E6/1732.03DEC.TUL KQ NBR 4

AC 068 AA TII. TWD 13484 CTG 933 SYS 5700

PRI 4

MDIS.07NOV/TUL EMPL 437846

*******1000 CYCLE REINSPECT*******

RH INBOARD FLAP RIB 25A CRACKED. CRACK STOP DRILLED

PER AARD 57-00-00-6. DATE 11/02/02.

ATBT.07NOV/TUL NAME A. E. BENSON EMPL 437846

1 ACCOMPLISH AN EXTERNAL VISUAL INSPECTION OF THE

2 STEEL RUB STRIP FOR CRACKS EVERY 4TH B CHECK.

3 NOTIFY TECH SERVICE AND ENGINEERING IF CRACKS ARE

4 FOUND.ACCOMPLISH AN INTERNAL INSPECTION OF RIB

5 25A REAR FOR CRACK PROGRESSION AT 1000 FLT CYCLE

6 INTERVALS. REPLACEMENT OF THE RIB TO BE DONE AT

7 NEXT MAIN BASE VISIT.

06857009F/1601.07NOV.TUL...DFRD- 26 DAYS KQ NBR 5

Any thoughts?

-----Original Message-----
From:
Sent: Friday, December 06, 2002 5:04 PM
To: T W
Cc: David
Subject: Re: #68 today

Thanks,

Strange to say, sounds like 'autopilot intrusion' of some kind... if such an animal is possible. Brief autopilot activation even though not engaged, signalling turn/rudder deflection not from pedals but from autopilot computer down the line or servomechanism itself.

1) Autopilot software fault or

2) arc-faulting event that caused spurious signals? HAS ANY UNCOMMANDED RUDDER A/C HAD A THOROUGH WIRING CHECK? (Visual and intrusive bundle inspection... a/c out of service for x time, "expensive" -- not really when you consider all airlines' current overcapacity. Mechanics' and avionics technicians' time, yes, but there seem to be potential time bombs and one thorough check might uncover something).

Turbulence seems to be a possible catalyzing factor--as perhaps with 587.

Just a thought.

Lee

----- Original Message -----

From: IASA

To: 'David

Cc: anthony.brownrc.ca

Sent: Tuesday, December 10, 2002 8:19 AM

Subject: RE: #68 today (two files attached)
Analysis

David

"NO RUDDER PEDAL MOVEMENT WAS NOTED."

"Turbulence seems to be a possible catalyzing factor--as perhaps with 587."

Re turbulence. That's agreed, which brings me to "hysteresis" - the lagging of an effect behind its cause.

I continue to suspect that there is a flight control system hysteresis set up in the rudder circuit whenever there is an external and extraneous event that brings the yaw damper into very active play. A yaw damper normally constrains inappropriate yaw (such as high-level Dutch Roll coupling) from ever getting started in the first place. In sudden LOC/stall, wake-induced or atmospheric turbulence situations involving sudden large-displacement yawed flight conditions, it (the YD) is entering the theatre halfway through Act One - and that may be a large part of the problem. I'm not sure that when they set out to design a yaw-damper, that it was ever a design requirement that it should intervene capably in the midst of a mini LOC. Indeed I'm not sure if that would be written into the flight-test program or indeed, could be set up according to any laid-down test parameters. In particular I am always led back to the fact that it is a 2 x YD dual-actuating affair acting through a singular duplex valve - upon one of the system's three rudder actuators. The hand-over/take-over ("you're malfunctioning" - "No I'm not, you are") routine (and initiators) between the two yaw-dampers may also be a wild-card in this equation.

That SIA A340 event had simultaneous airspeed fluctuations - and so might this incident (have had). It's likely that the flight-crew probably (and routinely) dismiss those fluctuating airspeed indications as being symptomatic of the turbulence and therefore not really reportable or indeed, significant. Sharp pressure spikes in the Air Data Computer's input data may be causing lagged (and inappropriate) corrective rudder inputs and leading to a system induced oscillation (SIO - aka an erratic hunting for equilibrium in yaw). With the dubious DFDR filtering, it also becomes debatable as to whether any such ADC high frequency pressure flux (or high-rate airspeed fluctuation output) would be identifiably recorded. As a matter of urgency they first up need to fit more informative flight-test variety DFDR's into a reasonable number of these aircraft. However I doubt that they'd do that while they are in passenger-carrying service - on liability grounds..... so maybe just fit them to the freighters.

The "cause and effect" key to the A300 tail-wagging is in the train of events kicked off by the external initiator, apparently whether or not the autopilot is in the circuit. The common denominator would seem to be the pressure spikes that are present during LOC/stall/incipient autorotation, wake turbulence, atmospheric turbulence - and also (once initiated) by the continuing, undamped yawing condition itself. Any yaw could amplify static port errors as the ports (one on each side of the aircraft) are individually and alternatively airflow-blanked and then subjected to an error-inducing inflow (as the face of a port on one side becomes unblanked during a yaw cycle - and then presented obliquely to the airflow). Any discrepancy in the sensed static (caused by the pro and adverse flow of any water trapped in the static lines for instance) has a disproportionate effect upon the airspeed (you will recall). But on top of any such error will be the p1+ and p2- flux induced by the port-face exposure during yawing and a further superimposed error due to the actual pressure spikes present in a vortex rotor. The net effect may well be that the sensed airspeed (momentarily inputting to both the rudder limiter and yaw-damper) will be, at any given moment, significantly in error and/or trending away from the real indicated airspeed (i.e. CAS or calibrated). The undesired outcome is likely to be a software integration issue. IAS and CAS do not normally differ by significant amounts in a static situation - however dynamic motion in the yawing plane (or side-gusts) can induce much larger transient pressure errors, depending upon the particular installation and its system's damping.

Any sudden yaw will invite a yaw-damper correction to the rudder and that rudder deflection has to be an appropriate and timely augment to the sizeable stabilizing moment simultaneously acting (naturally) due to that large vertical fin - or it will drastically overshoot the null position (instead of just slightly overshooting it in a rapidly-damped, decreasing amplitude phugoid). If, due to ADC-generated airspeed confusion, it's too large a yaw-damper input, the yaw correction will overshoot, a further correction will be required of the system and consequently the aircraft is going to cycle in an uncommanded yaw i.e. describe a continuing undamped sinusoidal phugoid around the yaw axis. Unlike a Pilot Induced Oscillation (PIO) in the pitch axis, development of any PIO (or SIO) in the yaw axis is limited by that large restorative moment generated by the vertical fin..... and cannot become divergent (i.e. due to yaw stability). However what can happen is that the rudder's inputs can become so out-of-phase and inappropriately deflected, that a fin overstress can result after a very few asynchronous cycles of its hysteresis condition (per AA587).

Two areas that I have not touched upon are also speed sensitive - i.e. pilot pedal input and any inappropriate rudder-limiter setting that might be caused by airspeed flux and IAS sampling rates. CROW instability effects that might occur in an along-track wake vortex encounter have been covered elsewhere and would neither add or detract from the possibilities of the above described theoretical phenomena. However, as a physical (and destructive) additive to the above phenomena, the CROW hammer (see the attachment) may well have been the final straw for AA587, particularly if the yaw onsets were so pronounced that exacerbating pilot pedal intervention was a psychologically natural outcome.

And, of course, an autopilot malfunction per the attached AD (99-NM-189-AD) may also have caused this incident below (see the box) - even though "AIRCRAFT WAS BEING HAND FLOWN"

After reading the analysis below (from the attached AD) one might either accept it at face value or perhaps see it as a convenient (yet potentially specious and/or spurious) explanation for an incident (for which they could find no other logical cause).

SUPPLEMENTARY INFORMATION: The Direction Générale de l’Aviation Civile (DGAC), which is the airworthiness authority for France, recently notified the FAA that an unsafe condition may exist on all Airbus Model A300, A310, and A300-600 series airplanes. One operator of an Airbus Model A300-600 reported high rudder forces and uncommanded rudder inputs during final approach. The uncommanded rudder inputs caused deflections of the rudder control surface resulting in yawing of the airplane. Investigation of the incident is ongoing, but preliminary results indicate that failure of both the main valve and the clutch valve of the autopilot yaw actuator can lead to the actuator generating uncommanded rudder deflections. The DGAC advises that the same autopilot actuator is used for roll and pitch control during autopilot operation, and this failure scenario can result in uncommanded deflections of the aileron and elevator control surfaces. (note: unknown as to whether these have actually occurred - or are just theorized to be another possible manifestation)

Preliminary results of the investigation of the incident airplane’s autopilot yaw actuator indicate that the electrical connectors between the actuator’s two main valves and the airplane’s two flight control computers (FCC) were crossed between side 1 and side 2. This hidden failure in combination with a failure of the clutch valve resulted in the autopilot yaw actuator remaining engaged when the crew disconnected the autopilot, allowing the actuator to remain hydraulically pressurized and provide inputs to the rudder and the rudder pedals.

This condition, if not corrected, could result in uncommanded deflections of the ailerons, elevator, and/or rudder, which could result in reduced controllability of the airplane.

from this link

IASA

From: David
Sent: Tuesday, December 10, 2002 6:33 AM
To: 'safety'
Subject: FW: #68 today

David

It gets murkier and murkier. I just don't believe the "bad lead" or the DFDR "bad sensor" (see below). All too facile.

I wonder if "a bad lead" is similar to an intermittent arc-fault (in AA maint parlance)?

Interesting that they are keeping DFDR readouts under ongoing review.

It is all narrowing down to being the yaw damper (as provoked by the wake encounter) (IMHO) - as chief AA587 miscreant.

After the December 3 incident, tail # 068 was released for flight two days after the event; and it was stopped minutes before takeoff when the Tulsa tech unit at AA discovered a .46 lateral acceleration on the DFDR from a previous flight. Subsequently, they have found a "bad lead" on one of the yaw dampers and the uncommanded sequential yaw movements on the Dec. 3 flight has been attributed to that. It has been speculated that, since the flight crew from the previous flight (the one with the .46 lateral acceleration, which was NOT the Dec. 3 flight) did not write up any problems, it was "just a bad sensor" on the DFDR. This leads to all sorts of questions about DFDR data and its reliability, and the yaw damper problem experienced by AA 587 on pre-flight at JFK.

Apparently, the F/O on the Dec. 3 flight has asked for a transfer off the A300 and the Chief pilot has said he will accommodate him.

Todd Wisisng

Flt AA903 - The West Palm Beach Incident

Panel 408VU and A300 Uncommanded Yaw Events

Some A300 Autopilot Discrepancies

The Rudder Trim Switch & Uncommanded Yaw

The Rudder Trim Knob AD

The CROW Hammer Wake Vortex Effect

Singapore and Lima (Peru) Events Also

Some other uncommanded Rudder Incidents

Air Safety Week 25 Nov 02