The aircrew's decision not to activate the fire extinguisher was not in accordance with the requirements of the operator's approved phase-one emergency checklist procedures. Those procedures are immediate actions, and the operations manual does not suggest that the memory items from an emergency checklist are optional, or that they should not be completed in certain circumstances once they have been commenced.

As the PIC had recently experienced two similar engine fire indications in this aircraft, his response on this occasion may have been influenced by those recent events. However, the PIC's decision not to activate the fire extinguisher placed heavy reliance on the extinguished fire-warning lights as an indicator that there was no longer a threat of fire. That decision, whilst supported by the company's chief pilot, did not appear to take into account the possibility that a malfunction of the fire-warning system was masking a real fire.

When the SA227 Metro aircraft was turning onto the downwind leg of the circuit, the crew noticed that the left-engine fire warning annunciator light was illuminated. The phase-one emergency procedures were immediately initiated. Those procedures included immediate recall and checklist prompted actions. Accordingly, the engine was shut down. Shortly after, the fire warning light extinguished. The pilot in command (PIC) then decided not to activate the fire bottle. A single-engine approach and landing was conducted.

This was the third left-engine fire indication the aircraft had experienced since September 1999, the previous two having occurred over a two-week period. The PIC was involved in all three events. During the first and second events, the fire warning indications had remained illuminated following engine shutdown, and the PIC had discharged the corresponding fire extinguisher into the engine fire zone on both those occasions. However, the PIC chose not to discharge the fire extinguisher during the most recent event as the fire-warning indications had extinguished following the engine shutdown. The company's chief pilot supported that decision.

Following the first event, maintenance investigation carried out by the operator revealed that the insulation on a wire in the fire-warning system wiring harness had chafed through while in contact with a nearby bracket. That damage was rectified. After the second event, the fire-detection system was checked for operation. No fault that could have contributed to the activation of the fire-warning system was found. Notwithstanding, it was considered that the lower-turbine fire detector was possibly too close to the engine, and may have resulted in a spurious fire indication. The detector was re-positioned and the aircraft returned to service.

Investigation of the most recent event showed that one of the fire detectors was activating at an incorrect temperature. The aircraft was returned to service following replacement of that detector with one capable of activating within the correct temperature range. At the time of writing this report, similar problems had not re-occurred.

The phase-one emergency checks are described in detail in the operator's aircraft flight operations manual and in the aircraft flight manual. The immediate recall actions require that the engine be shut down with the "stop and feather" control, and that the fuel and hydraulic oil be isolated, before the fire extinguishing system is discharged. The procedure is to then be continued from a written checklist. Items identified as recall actions are intended as a non-discretionary response to a fire warning.

History of flight

The float fitted Robinson R44 helicopter was being used to conduct a charter flight from Hayman Island to Reef World. The pilot reported that about 20 minutes after take-off, he noticed a burning smell, felt a slight shudder closely followed by the helicopter's clutch light dimly flickering. The pilot conducted a powered descent, transmitted Mayday calls and landed the helicopter in the water with minimal impact forces. The helicopter sustained damage to its aft cowling. The pilot and passengers were unhurt.

Examination of the helicopter revealed that the fan shaft bearing located on the fan shaft between the engine and the cooling fan had overheated, melted and seized.

The fan shaft had fractured just forward of the bearing. The fan shaft bearing is the lower actuator bearing of the clutch actuator assembly. The clutch actuator had fractured with the bottom half departing the aircraft. The sheave and tail rotor drive shaft were damaged by the V-belts. The engine had extensive overspeed damage and had moved off the aft mount. The upper frame was bent near the engine and the aft cowling was damaged as a result of excessive engine vibration and contact with separating components. The rocker assemblies on several cylinder heads were pushed through the rocker covers indicating the severity of the engine overspeed.

Maintenance history

It was reported that an earlier clutch actuator in the helicopter had experienced flickering clutch lights for a period of time. The actuator was replaced with a modified clutch actuator about 19 months previously and the clutch light problems disappeared until about a month prior to this incident when the clutch light began staying on for 8-10 seconds. The problem was attributed to a faulty tensioner and a new clutch actuator was fitted 21.6 flight hours previously. The helicopter had 926.4 hours of service. The fan shaft bearing was the original fitment to the helicopter. The manufacturer advised that they had overhauled the clutch actuator fitted at the time of the occurrence prior to fitment to VH-STO.

Examination by helicopter manufacturer

The helicopter was shipped to the U.S. to permit examination and repair by the manufacturer. The manufacturer provided an investigation report and photographs of damaged components to the ATSB. The manufacturer advised that the exact sequence of events was open to discussion since several events had occurred at virtually the same time.

The lower actuator bearing lost lubrication after 926.4 hours of service. The dry bearing overheated the fan shaft and resulted in its fracture.

The excessive heat from the bearing partially melted the aluminium bearing spacers and the brass roller separator, and the bearing seized. Spinning of the outer bearing housing tore the bearing free and fractured the actuator.

Bearing and actuator failures resulted in a loss of drive belt tension and caused an engine overspeed and rotor RPM decay necessitating an autorotation onto water. The fractured actuator and loose belts caused secondary damage as they flapped around with the spinning clutch shaft.

Failure of the actuator and fan shaft allowed the tail rotor drive shaft and clutch shaft to move downward and the tail rotor driveshaft rubbed the steel tube frame. The spinning fan dropped down damaging the exhaust and heater ducts.

The engine overspeed caused damage to No. 1 and 2 cylinder intake valves and resulted in the No. 2 intake valve dropping into its cylinder. Subsequent severe engine vibration fractured the aft engine mount and also damaged the aft cowling.

The manufacturer advised that early bearings had been assembled with some seal rings non-concentric with the bearing. Non-concentric seals were being pinched during assembly resulting in distortion of the seal. Distorted seals may have allowed grease to leak out and water to leak in. The corrosion and/or loss of grease resulted in roughness and eventual failure of some bearings. New tooling was introduced in February 1999 to keep the seals centred during assembly.

Due to the severe heat damage to the bearing, the manufacturer was unable to determine why the lower actuator bearing lost lubrication. The distorted seal, loss of grease and water ingress was considered the most likely sequence of events.

A search of the Civil Aviation Safety Authority's Major Defect Report database revealed no other reported loss of lubrication to a Robinson R44 fan shaft bearing.

Passengers had boarded the BAe 146 prior to departure. When the pilot in command selected the start master switch to No. 1 engine in preparation for engine start, the aircraft's AC power supply immediately failed. The indications included the "APU GEN OFF LINE" annunciator and cockpit/cabin emergency lighting illuminating. The auxiliary power unit (APU) generator was reselected on to restore AC power but immediately after the switch selection was made, the AC power failed again. The crew also noticed a small amount of smoke drifting past the cockpit overhead emergency lighting. They immediately turned off the start power and began disembarking the passengers.

While the passengers were disembarking, the co-pilot checked the electrical equipment bay located on the outside of the aircraft. He found a small fire in an electrical rack, which he extinguished with the cockpit's portable fire extinguisher. The co-pilot also disconnected the aircraft batteries. The off-airport rescue and fire fighting service (RFFS) was called and remained in attendance until the arrival of engineering staff.

Inspection by maintenance personnel revealed that the remote control circuit breaker (RCCB) which controls the AC-powered hydraulic pump had failed.

The RCCB was forwarded to the ATSB and dismantled. It was found to have been subjected to extreme heat, which destroyed two of the three main AC contacts within the RCCB. The level of internal damage precluded determination of why the RCCB had failed. However, it was found that as a consequence of the RCCB design, the three main contactor chambers were open to air, dirt and moisture during normal operations. The investigation could not determine if this design feature was a factor in the electrical malfunction.

The AC-powered hydraulic pump internal thermal switch wire was found to be pinched between the impeller housing and the stator, effectively creating a short circuit to ground. The effect of this short circuit would only be noticed when the pump had exceeded an operating temperature of 204 degrees Celsius. Although the pump did not display any outward signs of excessive heat, it did exhibit a general state of deterioration commensurate with the extended time in service for this unit. Clearly the RCCB was subjected to excessive current load. This caused a catastrophic internal failure and the subsequent heat generated by the failure led to molten metal escaping from the RCCB main contactor compartment. The molten metal then flowed across two energised power cables, which resulted in the short-circuiting of two AC power phases.

The RCCB was located in an equipment bay that was not monitored by fire or smoke detection devices. The technical crew was alerted to the fire by smoke in the cockpit, system failures and a fire that the co-pilot noticed when he gained access to the RCCB through an external bay door. It was possible to access the equipment bay, which held the RCCB, from the cockpit. If a similar problem were to occur, opening the access door during flight would introduce more oxygen to the fire and vent smoke and noxious fumes into cabin, threatening crew and passengers.

The airframe manufacturer's failure-trend data for the RCCB was examined and it was determined that the equipment exhibited very high reliability in service. Consequently, the probability of recurrence of this type of failure was considered to be low.

The pilot, who was based at Mornington Island, was tasked to convey passengers from Mornington Island to Normanton in Cessna 206 registered VH-XGR. The pilot's intended track would have resulted in the aircraft overflying Bentinck Island, south-east of Mornington Island.

The flight departed at about 0915. Other pilots operating aircraft in the area reported that the weather conditions were not favourable for visual flight to the south of Mornington Island, but were acceptable to the north. Rain started to fall when XGR taxied for departure. After a take-off to the east the aircraft was observed to turn left and set course from overhead the airstrip.

Soon after departure the pilot was advised by the pilot of an aircraft maintaining 5,500 ft, and tracking for Karumba that the weather conditions enroute were a line of light showers.

At 0929 a pilot tracking from Doomadgee to Bentinck Island in a Cessna 206 operated by the same company as the accident aircraft was advised by the company through Brisbane Flight Service that visibility at Bentinck Island had reduced to 1,000 m in rain. That pilot subsequently reported in the circuit area at Bentinck Island at 0949.

At 0935, after obtaining the current position of the pilot tracking to Bentinck Island, the pilot of XGR reported that he was now diverting to Burketown, passing 3,500 ft on climb to 5,500 ft. The two pilots then changed to a company frequency. No further report from the pilot of XGR was recorded.

The aircraft did not arrive at either Burketown or Normanton. A subsequent search found articles, identified as being from XGR, in the water to the south of Bentinck Island. The damage sustained by these articles was consistent with the aircraft having struck the water at high speed.

Examination of recorded weather radar data available from a radar site at Mornington Island indicated a band of weather extending from just south of Mornington Island to north of Bentinck Island, and beyond. To conduct the flight as intended the pilot would have been required to negotiate this weather. Pilots who had arrived at Mornington Island before the departure of XGR reported that the cloud did not contain any thunderstorm formations. Conditions under the cloud were assessed as not suitable for VFR flight due to reduced visibility in rain.

The pilot was appropriately qualified and met the recent experience requirements to conduct the flight. His medical history did not indicate any condition that may have led to incapacitation. He had obtained a command multi-engine instrument rating nine months before the accident flight, but had not since conducted a flight in instrument meteorological conditions and had only flown a limited number of hours in simulated instrument flight conditions.

The aircraft was not certified to operate in instrument meteorological conditions and was not fitted with an autopilot, nor was it required to be. Although it was equipped with appropriate instrumentation, there was no requirement that they be maintained to instrument flight standards. No evidence was found to indicate that the aircraft was other than serviceable for flight in visual meteorological conditions.

Following an unsuccessful search for the aircraft, the investigation by the Australian Transport Safety Bureau was terminated. At that time, in the absence of sufficient aircraft material to enable a comprehensive examination, it was not possible to determine the factors that led to the accident.

In late 2001, Queensland Police sent the ATSB photographs and items of wreckage recovered from an area on the seabed near Sweers Island, east of Bentinck Island. Each photograph was examined and image enhancement techniques were employed to assist with closer inspection. On 8 March 2002, Queensland Police sent the ATSB a video of aircraft wreckage taken underwater at the accident site.

The items of wreckage sent to the ATSB's engineering laboratory for examination, and the wreckage viewed on photographs and video, were consistent with having originated from a Cessna 206 aircraft. While positive identification of the aircraft VH-XGR was not possible, on the balance of probability it is likely that the wreckage was that of VH-XGR.

The examination of the attitude indicator revealed that the aircraft impacted the water at an angle of bank of 135 degrees right-wing down (45 degrees inverted) and 35 degree nose-down attitude. That was consistent with the structural damage viewed on the photographs taken by the police.

The damage to the airframe components and propeller blades, and the evidence from the attitude indicator were consistent with a high velocity impact with the water following a loss of control of the aircraft. This damage signature has been evident in numerous accidents investigated by the Bureau where an aircraft has crashed following flight from visual into non-visual meteorological conditions. Scenarios which have led to similar accidents have included where the pilot was either not trained for such flight operations, or had limited or no recent experience in instrument flight conditions, or the aircraft was not appropriately instrumented.

The pilot's limited experience in instrument flight conditions may have been insufficient to prevent a loss of aircraft control had he inadvertently entered an area of low visibility in the Sweers Island area.

The ATSB's engineering report is available above in the Technical Analysis Report tab.

1. The flight crew and air traffic control staff did not brief effectively before the calibration flight.
2. The Astra crew was unaware that Brisbane Centre was responsible for management of a portion of R577.
3. The Astra crew was unable to maintain continuous two-way communications with flight service international.
4. The provision of radar coverage and use of secondary surveillance radar transponders in oceanic airspace assisted in the resolution of the conflict.

The briefing planned between the Astra crew and Williamtown Air Traffic Control staff did not occur and as a result, the air traffic control staff were not fully conversant with the Astra crew's intentions. Also, the Astra crew were unaware of the unique delineation in airspace management responsibilities for the area where they intended to operate. These aspects made it difficult for the crew to adequately prepare for contingencies and for air traffic controllers to understand the complex nature of the calibration task.

The Astra crew complied with the clearance issued by Williamtown Air Traffic Control. However, the clearance was issued without air traffic control being fully conversant with the flight profile. The misunderstanding was compounded by the Astra crew believing that their flight would be managed by Williamtown Air Traffic Control. If the crew had been aware of the ramifications of the Jordy Release, it was likely that they would have appreciated that the initial stages of the flight would pass back and forth between R577 and non-controlled airspace. Had they been aware of this fact and that consequently, they were responsible for their own traffic avoidance, the Astra crew may have ensured that air traffic services were advised of the intention to change level well in advance of commencing the climb to FL200. If the crew had been able notify flight service international of their intention to climb from 10,000 ft to FL200 it was probable that they would have been issued with traffic information on the Dash 8 and the conflict may not have occurred.

The operation of the Astra's transponder on the assigned code assisted in the resolution of the situation. An alternative for the crew was to select the transponder to the radio failure code of 7600, which was part of the aviation safety net. Given the complex nature of the task and the communication difficulties being experienced by them at the time, this action may have provided a more timely alert to air traffic agencies. Had air traffic agencies been alerted by the change in code to 7600, it is likely that air traffic control or flight service international officers would have either provided traffic information or initiated action to separate other aircraft in the vicinity of the Astra.

The Brisbane Ocean sector controller saw on radar an unidentified aircraft climbing through flight level (FL)180. The controller issued traffic information to the crew of a Dash 8 operating an instrument flight rules (IFR) flight from Lord Howe Island to Sydney at FL200. The crew of the Dash 8 sighted the unidentified aircraft at an approximate distance of 5 NM and 1,500 ft below them. The two aircraft passed with 3 NM lateral separation at the same level. The unidentified aircraft was subsequently identified as an Astra conducting an IFR calibration flight. It appeared that the crew of the Astra had not complied with Aeronautical Information Publication (AIP) procedures for changing levels in non-controlled airspace. These procedures require pilots of IFR category aircraft operating in non-controlled airspace to notify the relevant air traffic agencies of any changes in altitude/level. Investigation of the occurrence revealed that the Astra crew were not notified of a change in airspace management for their area of operations.

The Astra crew was conducting a calibration of a new radar that had been installed at Williamtown aerodrome. This required the aircraft to track outbound from Williamtown at 1,500 ft, in a direction clear of local operating areas and air routes, until radar coverage was lost and then return on the reciprocal track while climbing. Once the next altitude or level was reached the crew would immediately turn the aircraft to track outbound again. The manoeuvre was to be repeated at 10,000 ft, FL200, FL300 and FL400 and would conclude with the aircraft completing the task at approximately 220 NM from Williamtown.

The Astra crew had prepared and dispatched written briefs on the task to Williamtown and Brisbane Air Traffic Control (ATC) agencies. However, the calibration task schedule was amended and subsequently conducted 24 hours earlier than originally notified. The brief depicted the calibration flight as being a series of steps with an increase in altitude/level as the distance from Williamtown increased but did not depict that the aircraft would fly the reciprocal track at any stage. After the Astra had departed Williamtown to commence the calibration, the Williamtown approach coordinator briefed the Brisbane Flight Service International officer on the task, based on the briefing and other details as subsequently advised by the crew.

The general practice of the Astra crew was to conduct personal briefings with the responsible air traffic control staff on the day of a calibration task and this was their intention on the day of the occurrence. However, when the Astra arrived at Williamtown, the senior air traffic control officer who had planned to attend the briefing was not available. Consequently, the crew only briefed a technical ground party.

The Astra crew was issued a clearance by Williamtown Air Traffic Control to track via the 100 tactical air navigation aid radial on climb to FL210. On this track the Astra would transit Williamtown restricted area R577 which was promulgated as continuously active, 24 hours a day. The intended track of the Astra entered and exited R577 respectively at 25 NM and 130 NM east of Williamtown. The vertical limits for the area are from sea level to FL600. When not required for military operations a notice to airman (NOTAM) deactivating R577 for specific periods was normally issued by Williamtown Air Traffic Control. A deactivation NOTAM was not issued for the period of the calibration flight.

As an aid to airspace management, Brisbane and Williamtown Air Traffic Control agencies had agreed to automatically release portions of R577 and other restricted areas to Brisbane when the restricted areas were active. This agreement was published in the Northern New South Wales Manual of Air Traffic Services Supplement (MATS SUPP) which in turn was part of both Williamtown and Brisbane operating procedures. The agreement effectively transferred airspace management responsibility for the south-east portion of R577, known as the Jordy Release, from Williamtown Air Traffic Control to Brisbane Centre. The planned track of the Dash 8 passed through the Jordy Release.

The lateral and vertical limits for Australian airspace were published in the Airservices Australia Designated Airspace Handbook (DAH) and depicted on AIP charts. The DAH, charts or a NOTAM can be used for flight crew pre-flight briefing. None of the documents included the MATS SUPP provisions.

A letter of agreement between Brisbane Ocean sector, Brisbane Flight Service International and Sydney Flight Service 4 detailed the airspace management responsibilities for the Jordy Release when transferred to Brisbane Centre. The division of responsibilities was:

• Ocean sector
  That portion of the release above FL200 between 90 NM and 150 NM from Sydney and above FL245 outside of 150 NM Sydney,
• Flight Service International
  That portion of the release at FL200 and below that level outside of 90 NM from Sydney, and
• Sydney Flight Service 4
  That portion of the release below the control area steps inside 90 NM from Sydney.

The Ocean sector controller was required to separate IFR aircraft as if they were in controlled airspace. The officers of both flight service units were required to issue traffic information to pilots of IFR category flights, just as they would have for pilots of IFR flights in non-controlled airspace, in accordance with the Manual of Air Traffic Services (MATS) criteria. Part of the MATS criteria required traffic information to be issued when aircraft were climbing or descending through the level of another aircraft when these aircraft were within 15 NM or 10 minutes longitudinally or laterally. Ocean sector uses radar and procedural control methods to separate aircraft. Radar coverage extends to beyond the eastern limit of R577. Neither of the flight service units had access to radar services.

Because of the flexible airspace management arrangements for the Jordy Release, flight service international officers usually pre-empted requests for confirmation of the availability of planned routes for aircraft inbound to Sydney. Normal practice was to advise crews as early as possible when the area was active with military operations to enable them to commence tracking via alternative routes. The crew of the Dash 8 was not advised that their planned route was unavailable. The crew was subsequently advised to contact the Ocean controller at 140 NM from Sydney for a clearance.

The Astra crew reported that due to R577 being active, they believed that, their flight would be managed by Williamtown air traffic control. The Williamtown coordinator advised flight service international that the crew had been cleared to FL210 but that initially they would maintain 1,500 ft. The coordinator subsequently advised flight service international that the Astra had climbed to 10,000 ft. The Astra crew was instructed that they were leaving the Williamtown restricted area and to transfer to flight service international on high frequency (HF) radio. This was required as they were operating in that portion of R577 being managed by flight service international.

After several attempts, the Astra crew contacted flight service international on HF radio and was advised of another HF frequency to use if they were having communication difficulties. There was no further contact with the crew on HF radio. The flight service international officer attempted unsuccessfully to contact the crew. The problems with the HF radio communication were believed to be the result of propagation difficulties. Once it was apparent to the crew that HF communication was not viable they tried to establish communications with Brisbane Centre via VHF. The crew reported after the occurrence that because they were unable to advise any air traffic service agency of the next climb, from 10,000 ft to FL200, they broadcasted their intention on the VHF emergency frequency, 121.5 MHz. The investigation could not establish whether other crews heard the transmission.

The crew contacted the Nambucca Sector Controller and reported that their position was 120 NM east of Williamtown on the 100 radial and that the aircraft had left 10,000 ft on climb to FL200. The Nambucca controller confirmed that the Astra was tracking west and that at 110 NM from Williamtown the crew would commence a turn to track outbound. The Nambucca controller contacted the Ocean controller to report the Astra's position. At approximately the same time the Ocean controller was issuing a clearance to the crew of EA261 and shortly after noticed an unidentified aircraft approaching the Dash 8. The Nambucca controller issued traffic information to the crew of the Astra, who arrested the climb of their aircraft and then descended to FL185.

The Astra crew did not select the radio failure code of the aircraft's secondary surveillance radar transponder. The investigation did not establish why the crew did not change the transponder code.