History of the flight

Shortly after take-off from runway 34L at Sydney, the flight crew of the Boeing 747-400 aircraft received a forward cargo compartment fire warning on the Engine Indicating and Crew Alerting System (EICAS). On receiving the warning message the crew actioned the appropriate checklist, activated the fire suppression system and transmitted a MAYDAY. At the same time, flight attendants noticed a fine mist and the smell of smoke in the passenger cabin. The crew then returned the aircraft to Sydney, where an uneventful overweight landing was conducted.

Prior to landing, the EICAS fire warning message ceased. This indicated that the aircraft fire suppression system may have successfully extinguished any fire, however the cabin fumes were still evident. After landing, the flight crew stopped the aircraft on the runway where emergency services came to their assistance. After confirming with the flight crew that the fire warning message was no longer present, the emergency services assessed the aircraft from the ground, then allowed the passengers and cabin crew to disembark to a safe distance via mobile stairs positioned at the aircraft's front left door. Once the passengers and cabin crew were clear of the aircraft, the emergency services opened the forward cargo door.

A hot spot was detected on the left side of the forward cargo bay at body station STA900, where the side wall lining was found to be heat affected. Removal of the lining revealed burned insulation blanket material, discolouration of the aircraft skin and burned/broken electrical wires that powered the forward galley chiller boost fan situated in the area (see Fig 1). As the fire was no longer evident, ground engineers isolated the chiller boost fan electrical circuit and towed the aircraft clear of the runway.

FIGURE 1: Forward cargo bay with expanded view of chiller boost fan location

Aircraft structural damage

Non-destructive testing to check for cracking and conductivity of the aircraft skin adjacent to the affected area was carried out. No cracks were detected, however the conductivity test revealed three locations where the skin had been substantially affected by heat (see Fig 2). The most severely affected area required a temporary skin repair before the aircraft could be flown back to the operator's maintenance facility in the United Kingdom, where the heat-affected aircraft skin was replaced.

FIGURE 2: Heat affected areas

Sidewall lining and insulation blankets

The fibreglass sidewall lining between STA880 to STA900 was visibly heat damaged with discolouration observed on the side facing into the cargo compartment. Inspection of the reverse side revealed burned layers of fibreglass confined to a localised area approximately 30cm x 45cm (see Fig. 3). The insulation blankets that lined the aircraft skin were made of a fibreglass core with a metallised TedlarTM film on one side and a MylarTM film on the other and had been subjected to localised heat and fire (see Fig. 4).

Samples of the sidewall lining and insulation blanket were sent to the United States of America, Federal Aviation Administration (FAA) technical centre and the aircraft manufacturer for analysis and testing.

FIGURE 3: Sidewall lining	FIGURE 4: Insulation blanket

The examinations determined that both the sidewall lining and insulation blanket samples complied with the appropriate material specifications for aircraft use.

The flammability testing, conducted by the FAA, on samples of the insulation blanket included a vertical Bunsen burner test, which was mandated in Federal Aviation Regulation FAR 25.853 - Appendix F. The samples tested met the requirements, but due to their limited size, the result was not conclusive as to the integrity of the entire blanket.

The aircraft manufacturer's tests revealed contamination on the insulation blanket samples. This contamination consisted of environmental dust, fibres and corrosion inhibiting compound. These contaminants were consistent with general contamination found during evaluations of other in-service insulation blankets and were considered to be normal.

The aircraft manufacturer's 'flame propagation cotton swab tests' found areas on the blanket samples that were self-extinguishing while other areas showed "flame propagation uncharacteristic of that expected for new insulation blankets". It was unknown whether contamination, in-service ageing, or heat exposure, or a combination of these, altered the blanket's flame propagation characteristics.

Boost fan system

A galley chiller boost fan system was installed in the aircraft to provide forced air circulation over the forward galley chiller units increasing their cooling efficiency. The system incorporated a vaneaxial-type three-phase fan, powered by the aircraft's number 3 alternating current electrical system. Control power was supplied by the aircraft's direct current electrical system, with operation being automatic on selection of the galley chillers to ON. Circuit protection was provided by a 20 ampere circuit breaker and a cargo fire cutoff relay.

Chiller boost fan

An inspection of the boost fan revealed a burn hole and sooting on its casing adjacent to the electrical terminal (see Fig. 5). The electrical wiring to the fan was found to have four of its seven wires broken, with all of the wires displaying sooting discolouration (see Fig's. 6 and 7). The soot marks corresponded to those on the fan casing and when positioned together, revealed that the wires had separated at a point adjacent to the corner of the electrical terminal. The failure of the wires produced electrical arcing, which melted the casing, resulting in the burn hole observed.

Further inspection found that all of the fan impeller blades had failed just above their roots (see Fig. 8). Neither the impeller nor the fan shroud showed signs of hard body impact damage.

FIGURE 5: Electrical terminal	FIGURE 6: Broken wires
FIGURE 7: Sooting evident	FIGURE 8: Fan impeller blades failed

Technical examination of the fan found that the impeller was made from a moulded resin material. There was no evidence of any pre-existing defects or cracking found on the blade fracture surfaces. However, a number of blades showed breakage of a curved lip of material from their forward corner. This condition was consistent with overload fatigue possibly due to the blade tips contacting the fan shroud. Such a condition may have occurred prior to the blades total failure. The inside surface of the impeller was coated with a brown powder, determined to be primarily iron oxide.

The aluminium alloy shroud contained several circumferential wear marks that were adjacent to the impeller blade path. Although there were random scratches, no evidence of gouging or penetration of the shroud skin was found.

Further disassembly of the fan revealed wear on the electrical motor stator, indicating that it had been subjected to armature rubbing. The armature did not display similar wear patterns. Rubbing of this nature usually occurs as a result of bearing failure or excessive wear, leading to armature oscillations. For the full technical report see Attachment 'A'

Chiller boost fan service history

The chiller boost fan entered service in 1994, with the last overhaul being in June 2000, after removal from service because of electrical failure. The maintenance records for that overhaul stated: "Unit noisy due worn bearings, all other parameters ok. Reported defect not confirmed. Disassembled, cleaned and inspected, bearings renewed, unit reassembled and tested to spec". The fan was then fitted to the incident aircraft on 2 August 2000. No subsequent maintenance was recorded.

Chiller boost fan circuit breaker and electrical relay

The installed circuit breaker was a 20 ampere three-phase, push-pull high performance, trip free type, designed for aircraft installations. It's design allowed for increased amperage through the circuit for a specific time before tripping (breaking the circuit) and was used in large motor load applications where the inrush current would trip a standard circuit breaker. The length of time taken to cause the circuit breaker to trip varied according to the current it received. The aircraft manufacturer advised that "At 385 per cent or 400 per cent [load rating], this breaker will trip between 2.3 to 10 seconds". This prevented aircraft electrical power surges from "nuisance" tripping of the circuit breaker and rendering the boost fan inoperative.

A number of tests were conducted on the circuit breaker, including a 'load withstanding test'. This required the controlled increase in current through the circuit breaker, with time to trip recorded. This test was conducted at 105 per cent, 140 per cent and 200 per cent values, as per the manufacturers test procedures.

The installed relay was a 25 ampere, electromagnetic, three pole, single throw, normally open type. This was also subjected to a number of tests including 'Coil resistance', 'Coil hold and drop voltage' and 'Voltage drop and switching test across all three phases'.

These tests were performed under the supervision of the United Kingdom Air Accident Investigation Branch. Both components were found to comply with their operational specifications, with no adverse mechanical or operational functions found during the testing. As a result, both components were considered to be serviceable.

Quick Access Recorder

The aircraft's Quick Access Recorder (QAR) data was analysed by the Australian Transport Safety Bureau with the following information retrieved.

During climb the number 3 alternating current system showed a momentary increase in load from a nominal 31 per cent to 54 per cent, which equated to an increase in current draw of 57 amperes.

Four seconds later, the load was again recorded and had returned to the nominal 30 percentage range, where it remained for the rest of the flight.

Approximately 1 minute later the QAR recorded a forward cargo fire.

Approximately 3 minutes later, the first cargo fire bottle low quantity message appeared, indicating that extinguishant had been discharged successfully.

Other recorded data received from the aircraft's central maintenance computer (CMC) confirmed the arming of the fire bottles approximately 2 minutes after the fire warning and the discharging of the last two fire bottles after the aircraft landed.

Cargo fire detection/extinguishing system

The aircraft incorporated two dual loop smoke detectors in each cargo compartment. Air from throughout the compartment was drawn through the detectors and sampled. In normal operation, both loops must sense smoke for a fire warning to be activated. If the system detects a loop fault during self-test at aircraft power on, it would reconfigure to a single loop operation.

The cargo fire module located on the overhead instrument panel in the flight deck incorporated forward and aft compartment ARM buttons and a DISCH discharge button. On sensing smoke, the relevant ARM button, along with an EICAS message would be illuminated, alerting the crew of the fire. The crew must then push the ARM button in. This action disables electrical power to a number of circuits, including the galley chiller fan circuit. Extinguishing is then achieved by pressing the DISCH button (see Fig. 9).

FIGURE 9: Overhead instrument panel with expanded view of cargo fire panel

Four fire extinguisher bottles (A, B, C and D) service the cargo compartments, each having discharge lines to both the forward or aft compartments. On depression of the DISCH switch, bottles A and B discharge flooding the selected compartment with extinguishing agent. Bottles C and D are not discharged until 30 minutes later. If the aircraft reaches the ground before the 30 minutes are up, the bottles will discharge on touch down. The system was designed to give up to 180 minutes of discharge time.

The Saab SF340B aircraft was being operated on a scheduled passenger flight from Dubbo, NSW to Sydney. The crew reported that, during the descent into Sydney and while passing flight level 110, they selected the auto-coarsen system ON. After passing through 7,000ft AMSL, the aircraft suddenly yawed to the left and the left propeller feathered. The crew reported that they immediately noticed illumination of master warning and caution annunciations consistent with a left engine failure. A scan of the engine instruments confirmed that the left engine appeared to have shut down, with the engine torque and fuel flow indications at zero. At the time of the apparent engine shutdown the aircraft was under autopilot control. The crew reported that they were not operating any engine controls and that the power levers were positioned below the 64 degree position. The crew reported that the aircraft's auto-ignition system, which was armed when the ignition system was selected to NORM at engine start, did not operate throughout the incident.

The left engine was secured, and the crew informed the flight attendant and passengers of the situation. The crew declared a PAN to air traffic control, reporting an engine shutdown, and a local standby was declared. The crew elected not to attempt an engine restart due to the aircraft's proximity to the airport, and an uneventful one-engine inoperative landing was conducted.

Immediately following the incident, the operator's maintenance personnel carried out operational checks of the left engine. During those checks the engine and its auto-ignition system were found to operate normally.

An investigation into the incident, carried out by the operator, resulted in the removal of the left engine's hydromechanical unit (HMU) and the digital electronic control unit (DECU) for examination. The aircraft's fuel system was also extensively inspected with no defects or anomalies detected. The aircraft was then returned to service and the problem had not recurred in the 15 months following the incident.

The operator forwarded the removed HMU and the DECU to the component repair vendor. That examination found no problems with either component that would have contributed to the incident and both components were returned to the operator as serviceable items. The DECU was subsequently fitted to another aircraft in the operator's fleet. A short period after fitment, the unit was again removed following engine `torque fluctuation' problems on that aircraft and the component was again returned to the component repair vendor for examination. During that examination, a fault with `stressed' solder joints on a central processing unit board within the DECU was discovered and corrected. Several internal microcircuits were also replaced.

The aircraft was equipped with an engine ignition system that, when selected to NORM, automatically provided `flameout protection', or auto-ignition. The DECU sensed the rate of engine deceleration, comparing it to an internally programmed `rate of change' or `flameout schedule', and at a predetermined figure energised the ignition system for 7.5 seconds. This was indicated by illumination of a white ignition light on the flight status panel. The ignition was shut off when the gas generator speed decreased below 62% to prevent a `sub-idle relight'. The pilot in command reported that on the previous sector, the auto-ignition light had illuminated for a period of about 4 seconds during the descent, for no apparent reason.

The auto-coarsen system was designed to automatically feather a propeller in order to achieve a rapid reduction in propeller windmilling drag in event of an engine failure during takeoff, approach and go-around. The auto-coarsen system operated in a high or low mode depending on whether the power levers were above 64 degrees (high) or below 64 degrees (low). The active mode was indicated by the illumination of an annunciator positioned on the flight status panel. There were no markings on the power lever control quadrant to indicate the 64 degree position to the crew. The auto-coarsen system monitored a number of parameters including power lever angle and when in high mode, engine torque. The engine torque signal was provided by the DECU. During an auto-coarsen event, where an engine failure is not the reason for the event, the engine can continue to operate at low power with corresponding fuel flow, inter-turbine temperature (ITT), oil pressure and temperature indications.

The engine manufacturer reported to the ATSB that they had received several field reports that identified where an internal failure within the DECU had resulted in an intermittent loss of the torque signal output from an engine. The engine manufacturer further reported that, when allied with conditions of power lever position and engine parameters that are pre-programmed into the auto-coarsen computer, an intermittent loss of torque signal could result in the auto-coarsen computer mistakenly detecting an engine failure, triggering an inadvertent propeller auto-coarsen event.

The engine manufacturer reported that they had recently become aware that their standard acceptance test procedures (ATP), performed on DECU's following maintenance, had not always been successful in isolating intermittent loss of torque signal faults. They advised that they had improved the ATP procedures for DECU's that had been identified as `loss of torque signal' units. The occurrence DECU had not undergone the improved ATP inspection when it had been returned for examination following the occurrence.

The standard operating procedures to be followed by flight crews were detailed in the operator's Aircraft Operations Manual. These procedures contained the flight checklists to be followed in normal, abnormal and emergency situations. The aircraft manufacturer issued Revision 32 of the Saab 340B Aircraft Operations Manual in February 2001. That revision modified the normal checklist, deferring selection of auto-coarsen from the transition checklist to the landing checklist in order to minimise the time with auto-coarsen on, thus reducing the probability for an inadvertent auto-coarsen event. At the time of the occurrence the operator's checklists specified that auto-coarsen be selected to ON at FL110, as part of the transition check. The reason the operator's checklist was not revised to reflect the manufacturer's data was not available due to subsequent organisational changes.

The Flight Data Recorder (FDR) information from the aircraft was analysed by the ATSB to assess the operating parameters of the engine during the incident. That analysis revealed that approximately 37 minutes after take-off, with the aircraft in cruise flight, the flight data recorder had begun to record invalid information. This resulted in no useful data being available from the FDR for the remainder of the incident flight, a period of about 24 minutes. An examination of the aircraft's flight recorder system carried out by the operator during post incident system checks found no reason for the malfunction.

The co-pilot of a Boeing 737-800 (B737) en route from Perth to Brisbane on air route T91 reported maintaining flight level (FL) 390. He later reported maintaining FL370 at APOMA, located 140 NM north-northeast of Broken Hill, at 0501 EST. The Bourke sector controller in the Melbourne Air Traffic Control Centre requested the co-pilot to confirm the aircraft's level. The co-pilot replied that the aircraft was at FL370. A Boeing 747 (B747) maintaining FL370 was on a crossing route and estimated APOMA at 0503, two minutes after the B737. To be separated, aircraft at the same level on intersecting tracks required 15 minutes between their respective intersection estimates. The controller issued traffic information to the co-pilot of the B737 and to the crew of the B747.

At a subsequent position report the pilot in command (PIC) of the B737 overheard the co-pilot report the aircraft's level as FL370 and corrected the error. Later analysis of the flight data recorder of the B737 confirmed that the aircraft had maintained FL390. There was no infringement of separation standards.

The B737 operator had an approved procedure for the `controlled rest' of flight crew members while remaining on the flight deck. Controlled rest was recognised by the operator as `an effective method of improving levels of crew alertness for critical phases of flight'. There were a number of guidelines concerning the use of the procedure, including:

1. It was not to be used on sectors of less than two hours duration.
2. It was only to be used during the cruise phase of flight.
3. Periods of controlled rest were not to exceed 30 minutes per crew member per sector.
4. An additional 10 minutes was required after the period of controlled rest before a crew member resumed flight deck duties.

At 0213, the B737 departed Perth and was climbed initially to FL370. The flight plan advised an intention to change level to FL390 by BEZZA, a position about 340 NM west of Leigh Creek. At about 0405, the co-pilot requested and was approved by the PIC to take a controlled rest. At 0416, the PIC requested and was approved by the Melbourne Centre controller for the aircraft to climb to FL390. At 0421, the PIC reported maintaining FL390 to the Melbourne Centre controller. At about 0440, the co-pilot ceased the controlled rest and was briefed by the PIC in the next five minutes before assuming control of the aircraft. The briefing included the level of the aircraft. The PIC then commenced a period of controlled rest until about 0520. On initial contact with the Bourke sector controller at 0458, the co-pilot reported that the aircraft was maintaining FL390.

The roster for the 48-hour period preceding the incident required the crew to commence work at about 1900 two days before the incident and to operate a flight from Brisbane to Perth, arriving in Perth at approximately 0120. They spent the remainder of that day at their leisure prior to departing Perth for Brisbane on the incident flight at 0213 the following morning. That departure time required the crew to report for duty at approximately 0100.

On the morning of the Brisbane to Perth flight the co-pilot awoke at 0530 and had a 30-minute nap during that day. The co-pilot slept for approximately 6.5 hours after arriving in Perth, had a 90-minute nap late that afternoon and 30 minutes of controlled rest during the flight to Brisbane. Those periods of sleep accumulated to about nine hours during the 48-hour period. The co-pilot later reported that he felt rested prior to commencing the Perth to Brisbane flight.

Fatigue is recognised as a primary cause of transport accidents throughout the world as a result of reduced or impaired mental and/or physical performance following inadequate rest.

One component of fatigue relevant to the transport industry is sleep inertia. That phenomenon refers to the period of mental dullness or sluggishness immediately after awakening. During a period of sleep inertia people demonstrate the outward signs of being awake but are not mentally awake. Research suggests that it may take approximately 30 minutes after awakening for the effects of sleep inertia to fully dissipate.

The co-pilot was newly rated on the B737-800 and reported that he had found scanning the altitude indicator in that aircraft series took longer and required additional conscious effort. The co-pilot reported that when providing the aircraft's level, in response to the Bourke sector controller's query, he had relied on his memory of the information in the position report rather than re-checking the altitude indicator on the flight instrument display.

The Bourke sector controller twice requested confirmation of the flight level from the co-pilot of the B737 and was advised on both occasions that the level was FL370. At that stage, the controller passed traffic information to the co-pilot about the B747 at FL370 on the crossing route. The controller also passed traffic information on the B737 to the pilot of the B747. The co-pilot of the B737 later reported that at that time he was unsure why the controller had issued the traffic information as he could see the B747 displayed on the B737's Traffic Alert and Collision Avoidance System (TCAS) behind and 2,000 ft below his aircraft.

The Manual of Air Traffic Services (MATS) section covering safety alerts (Section 5.1.13) included the following:

'A safety alert shall be issued to an aircraft when a controller is aware the aircraft is in a situation which is considered to place it in unsafe proximity to terrain, obstructions, or other aircraft. The controller must remain vigilant for the development of such situations and issue a safety alert when the situation is recognised.

Conditions such as workload, traffic volume, the quality/limitations of the radar system, and the available lead time to react are factors in determining whether it is reasonable for the controller to observe and recognise such situations.

The issuance of a safety alert is a first priority.

When a controller is aware that an aircraft is in unsafe proximity to another aircraft, a safety alert shall be issued as follows:

"(Callsign) TRAFFIC ALERT (position of traffic if time permits), [SUGGEST] TURN LEFT / RIGHT (specific heading, if appropriate), and / or [SUGGEST] CLIMB / DESCEND (specific altitude if appropriate), IMMEDIATELY".

When a safety alert is directed to traffic not receiving a separation service, advice to turn or change level shall be prefixed with the word SUGGEST'.

As the Piper PA-28-161 (Cherokee) tracked from the 2RN inbound reporting point, which was 5.5 NM southwest of Bankstown Airport, via crosswind at 1,500 ft to runway 29, the instructor pilot saw a Beech Aircraft Corporation 76 (Duchess) pass close in front, tracking from right to left and on climb. The instructor pilot in the Cherokee turned the aircraft to the right to avoid the Duchess. Later analysis of the recorded radar information indicated that the two aircraft had passed about 150 m apart while at the same altitude.

Pilots of aircraft operating on Bankstown airport or within the control zone (CTR) were required to operate in accordance with General Aviation Airport Procedures (GAAP). The Visual Flight Rules (VFR) Flight Guide stated that GAAP catered for high-density operations in visual meteorological conditions (VMC). In VMC within a GAAP CTR, the pilot in command was primarily responsible for ensuring separation from other aircraft. Air Traffic Control (ATC) controlled runway operations with landing and take-off clearances and facilitated a high movement rate by providing traffic information and/or sequencing instructions.

The GAAP procedures were published in the Bankstown Visual Pilots Guide and the Aeronautical Information Publication (AIP) En Route Supplement Australia (ERSA). Bankstown procedures required pilots of aircraft to enter the CTR via specific reporting points, including 2RN, at 1,500 ft when runway 29 was the assigned runway (Figure 1). Pilots of aircraft operating out of the CTR in the runway 29 direction were required to depart via upwind and to maintain 1,000 ft until leaving the CTR. That procedure provided 500 ft vertical spacing between arriving and departing aircraft. Runway 29 Right was the nominated arrivals runway.

The Cherokee was on a VFR training flight with a flying instructor and student pilot. The student pilot was flying the aircraft as it tracked inbound to Bankstown Airport while the instructor briefed the student on geographical points. The instructor later reported that he did see the Duchess departing but it became obscured by the engine cowl of the Cherokee.

The Aeronautical Information Publication (AIP), used by pilots operating flights under the Instrument Flight Rules (IFR), stated that `arriving IFR aircraft which are visual outside the GAAP CTR, and can continue visually, must operate VFR within the CTR'. A pilot operating an IFR aircraft visually would only receive a traffic information and a sequencing service, and would not be separated from other traffic. The AIP further stated that `Departing IFR aircraft must operate VFR within the GAAP CTR until encountering Instrument Meteorological Conditions (IMC) or leaving the GAAP CTR, whichever is the sooner'. When aircraft are operating in conditions less than VMC, ATC will provide separation within the GAAP CTR. The weather at the time was VMC.

The Duchess was flown by a pilot on an instrument rating flight test monitored by a flying instructor. The pilot was conducting a Bankstown One departure on climb to 3,000 ft. The pilot departed from runway 29 Centre on a heading of 290 degrees M and had been instructed to report to the aerodrome controller (ADC) when the Duchess had passed 2,000 ft, the upper limit of the CTR. The instructor and the pilot in the Duchess did not see the Cherokee.

The instructor in the Cherokee monitored the ADC frequency as the aircraft approached 2RN. At 0214:00 the ADC issued departure and runway entry instructions to the pilot of the Duchess. About 20 seconds later the pilot of the Cherokee reported at 2RN at 1,500 ft and was subsequently instructed by the ADC to join via crosswind for runway 29. At 0214:50 the ADC issued a take off clearance to the pilot of the Duchess. The instructor in the Cherokee recalled hearing the take off clearance for the Duchess and saw that aircraft when it was on the runway but lost sight of it behind the engine cowl of the Cherokee. At 0217:10 the instructor in the Cherokee reported to the ADC that the aircraft was `joining downwind and that they had just seen the Duchess'. The ADC acknowledged that report and at 0217:34 the pilot of the Duchess reported passing 2,000 ft.

The altitude of the Duchess could not be accurately ascertained, as the Mode C function of that aircraft's transponder was not activated. The Mode C function for the Cherokee was activated and the recorded radar information showed that the aircraft had maintained 1,500 ft until established on downwind. The instructor in the Duchess later estimated that the aircraft would have been at about 1,500 ft at the point where the tracks intersected.

The radar track of the Cherokee showed that the pilot had tracked directly from 2RN to an early downwind position (Figure 2 - Track 1). That track was about 0.5 NM west of the recommended crosswind track (Figure 2 - Track 2). The aircraft tracks recorded during a 4-hour period on the day of the occurrence showed that there were a number of other pilots who tracked west of the recommended crosswind track, depicted in the Bankstown Visual Pilots Guide, when they entered the circuit area.

The intention of the entry procedure (Figure 1) was to have pilots enter the circuit, via crosswind, in a position to sight aircraft on the runway, aircraft departing, and other aircraft in the circuit. Crossing the extended runway centreline at 90 degrees also minimised the potential for arrival/departure conflicts as departing aircraft would generally have been airborne for only a short period and consequently would not have climbed to the 1,500 ft inbound altitude. Aircraft climb performance is subject to various factors, including aircraft type and load but generally the closer an inbound aircraft tracks to the threshold of a departure runway the more likely that there would be some vertical spacing between it and departing aircraft.

During taxi for take-off, the crew of the BAe146-100 aircraft noticed a 'yellow' hydraulic system 'low quantity' warning light on the aircraft's master warning system panel.

At approximately the same time, a cabin crewmember opened the flight deck door to alert the flight crew to the presence of fumes in the cabin. Passengers and two off-duty cabin crew reported a slowly moving white haze, low on the right side of the passenger cabin, in the vicinity of row 6. The haze was acrid and transparent and caused coughing and breathing difficulties.

An off-duty cabin crewmember also went to the flight deck and told the captain that the situation in the cabin had worsened, that there was smoke on the right side of the cabin and that passengers were having difficulty breathing. Because of the urgency of the report the pilot stopped the aircraft on a taxiway and instructed the cabin crew to prepare to evacuate passengers through the left doors. After shutdown procedures were completed, he ordered the evacuation.

The two operating cabin crewmembers opened the forward and rear left doors and deployed the escape slides. The two off-duty cabin crewmembers evacuated first, one through each door, to assist passengers at the base of the slides.

A passenger reported that cabin crew who stood at the aircraft doors to control the evacuation and block access to the right doors were out of view of the cabin. Therefore, the cabin crew could not see other passengers attempt to retrieve cabin baggage; an action that clogged the aisle and slowed progress to the exits. However, cabin crewmembers reported that cabin baggage did not delay the evacuation.

A cabin crewmember at the base of a slide reported that early in the evacuation, some passengers struck others that had not yet cleared the slide. Some fell as they reached the slide base and she lifted people to avoid a bank-up and the possibility of injury. Later, the evacuation proceeded in a more orderly manner. Cabin crew reported that they did not request assistance from able-bodied passengers during the evacuation.

Medical assistance

The airport Rescue Fire Fighting Services attended shortly after the evacuation was completed. They offered medical assistance and administered oxygen to two passengers. Medical assistance was also offered to passengers and crew on arrival at the airport terminal. None of the passengers requested medical attention.

Aircraft crew actions

Company emergency procedures required flight crew to don oxygen masks at any time that smoke or fumes were detected in the cabin. The procedures also required the flight deck door to remain closed to avoid flight crew incapacitation from fumes.

Both the pilot and the cabin crewmember that opened the door to speak to the flight crew reported that they were aware of the emergency procedure requirements. However, the pilot reported that the flight crew did not don oxygen masks as there were no fumes in the area and because the urgency of the cabin crew messages conveyed the need for an immediate evacuation. The cabin crewmember reported that it was quicker to open the flight deck door and safe to do so as there were no fumes in the area.

Cabin crew who had inhaled vapours, or who had assisted passengers off the escape slide, reported that during the continued tour of duty they suffered effects that included extreme tiredness, sore muscles and minor throat and chest problems.

Hydraulic system

Two independent systems provided hydraulic power to the aircraft flight controls and landing gear. These hydraulic systems were designated 'green' (left) and 'yellow' (right).

The power generation components were housed in the hydraulic bay, situated immediately forward of the main landing gear bay, below the forward rows in the passenger cabin. A light on the flight deck instrument panel provided a low hydraulic quantity warning when the fluid level fell below the operating level.

An inspection by the operator found that a leak in a hydraulic coupling allowed fluid under pressure to escape as vapour into the hydraulic bay and enter the passenger cabin via gaps in the sidewall lining. The 'o' ring seal for the coupling was replaced and the leak stopped. After a number of subsequent flights the coupling leak re-occurred. On closer inspection it was found that the coupling had a crack along its threads. The coupling was replaced.

The company reported that a subsequent Non Destructive Test (NDT) examination of the cracked coupling revealed that the coupling had failed through the bottom of a thread due to overload, which was consistent with having been done up too tightly.

Hydraulic equipment bay sealing

The aircraft manufacturer had generated three service bulletins that either required or recommended remedial action to improve sealing between the hydraulic bay and the passenger cabin. A zonal inspection was also conducted in the area at regular intervals.

At the time of the occurrence the operator had incorporated the first two service bulletins and had scheduled, but had not commenced, the third.