Aviation Investigation Report A98H0003
2.17.4 Fire Propagation from an Arc Fault Near STA 383
- 18.104.22.168 - General
- 22.214.171.124 - Initial Fire Propagation
- 126.96.36.199 - Initial Odour and Smoke in the Cockpit
- 188.8.131.52 - Fire Propagation Aft Out of the Cockpit
- 184.108.40.206 - Fire Affecting IFEN PSU Cables Above Galley 2
- 220.127.116.11 - Fire Affecting IFEN PSU Cable Outside Aft End of Conduit
- 18.104.22.168 - Fire Progression Riser Duct Area to the Left Side of the Fuselage and Aft
- 22.214.171.124 - Progression of the Fire into the Cockpit
A detailed assessment was conducted to determine whether the known conditions and subsequent events could be accounted for if the fire started just forward of the cut-out in the upper portion of the right cockpit rear wall near STA 383.
An arcing event just forward of STA 383 near the front of the 102-cm (40-inch) long conduits, would have the potential to ignite MPET-covered insulation blankets. Ignition of MPET-covered insulation blankets at the STA 383 arc location would initially generate a small creeping flame that would produce a small amount of smoke, with a relatively strong accompanying odour. Most of the smoke and odour would initially be carried by the airflow being continuously drawn down the air space adjacent to the ladder assembly into the avionics compartment, from where it would be filtered and exhausted overboard.
The small flame would slowly propagate across the underside of the MPET-covered over-frame insulation blankets. It could not travel very far forward before being blocked or redirected by a series of wire bundles that contact the insulation blankets in this area. The fire would not likely travel up and across the underside of the over-frame insulation blankets against a prevailing airflow down the ladder. Also, there are some wires and wire bundles that are routed on the cockpit ceiling that would act as fire barriers. It is possible that the fire could propagate down the ladder with the prevailing airflow; however, the flame front would encounter a series of horizontal wire support brackets that would act as fire barriers. Although it is possible that the fire could migrate around obstructions such as these, no physical evidence was found to indicate that the fire had propagated into the avionics compartment, because other than soot, there was no fire damage in that area.
If the flame front was able to propagate a relatively small distance inboard before predominantly propagating aft, it is likely that odour or smoke generated near STA 383 would momentarily be drawn into the cockpit through the holes and air gaps near the top of the avionics CB panel. Even if the flame front did not propagate sufficiently inboard for this to occur, it is likely that odour and smoke would enter the cockpit through these locations during the early stages of the fire. This would be expected to take place soon after the fire ignited the MPET-covered muff assembly adjacent to the smoke barrier or the insulation blanket on the riser duct assembly. In testing, smoke released over the front surface of the upper avionics CB panel near the rear right cockpit wall was initially blown downward before swirling back upward, eventually following a corkscrew path forward toward the flight crew seats.
The smoke would have to travel a relatively turbulent path before reaching the pilots. It would become more diffused, and it is likely that the smoke would initially present a weak visual indication. Therefore, the pilots would most likely detect an odour before seeing smoke. The most likely area where smoke would eventually be visible, would be in the vicinity of the avionics CB panel near the rear right cockpit wall. The density of the smoke would be greatest where it first enters the cockpit. The right overhead diffuser would accelerate the motion of the smoke, making it easier to detect. The smoke would also be in proximity to the overhead dome light, which would enhance the detection of airborne smoke particulates.
A small flame front could travel aft from the initiating location and pass over the cockpit rear wall through the cut-out. If present, the foam material used around the conduits and wire runs at the cockpit cut-out would either melt, or ignite and create additional smoke and odour.
When the small flame front passed over the cockpit rear wall, it is likely that for a short time the smoke was no longer migrating into the cockpit in sufficient quantity to be visible. Most of the smoke and odour would be expected to be exhausted down the ladder area during this interval. This condition would change once the fire ignited the MPET-covered muff assembly adjacent to the smoke barrier, or the insulation blanket on the riser duct assembly. Ignition of this insulation blanket would create additional smoke and odour that would ultimately be directed to areas near the smoke barrier. Airflow tests show that smoke and odour present in these areas could also migrate into the cockpit through openings in the smoke barrier.
Soon after the MPET-covered muff assembly was ignited, it is likely that a small propagating flame front breached, to some extent, the exposed Galley 2 silicone elastomeric vent cap. This vent cap is located immediately adjacent to the forward end of the muff assembly, next to the cockpit rear wall, adjacent to the smoke barrier. Air would immediately be drawn through the breached opening into the galley vent duct assembly, as soon as the vent cap was penetrated. This airflow would likely extinguish the small propagating flame front on the muff assembly in the area immediately adjacent to this vent cap. In testing of MPET-covered insulation blankets, air currents typically extinguished small propagating flame fronts.
It is also likely that the burning silicone elastomeric end cap would be extinguished at the same time owing to the sudden and continuous draw of air through the spot where initial fire penetration took place. Flame propagation along MPET-covered insulation blankets could still be taking place elsewhere, since by the time the Galley 2 vent cap was ignited, flame propagation would have likely spread over a much larger area including onto the riser duct assembly. The draw of air through the vent cap would also be expected to be small, causing only a localized air current effect. Once the fire intensified in the riser duct area and propagated onto the right fuselage side wall, the vent cap likely would have reignited or melted, resulting in the complete failure of the vent cap. This would create a much larger opening and a much larger draw of air into the vent duct system. The early draw of air and combustion by-products into the vent duct system could delay the return of smoke and odour into the cockpit, and delay the early detection of the fire in the passenger cabin.
The vertical air spaces adjacent to the centre riser duct, and between the aft side of the aft riser duct and the forward side of the R1 door frame, would channel hot combustion by-products, and create a chimney plume effect that would produce concentrated heat in areas above these air spaces. This plume effect would be further promoted by the vertical walled-in confinement of the MPET-insulated lower section of the riser duct assembly. Evidence of a high-temperature chimney plume effect was apparent in the wreckage that corresponded to these locations.
The eventual complete fire-related failure of the Galley 2 vent cap would cause a large volume of air to be drawn into the galley vent system at that location and would significantly change airflow patterns. A high-temperature fire damage pattern was found on ducts adjacent to the vent cap location. Overall damage patterns in the area are consistent with hot combustion by-products being drawn past the waterfall area, then forward underneath the riser ducts toward and into the galley vent duct system. The flow of hot combustion by-products, between the underside of the aft riser duct and the CD 207 ceiling panel below it, would create a significant localized convective heat effect on the panel. High-temperature fire damage of this nature was found on a piece of a CD 207 panel. This piece of panel most likely came from the sliding ceiling panel used at the R1 door location. Heat on this piece was consistent with a temperature exposure of 593°C (1 100°F) for a duration of 10 minutes.
The lower portions of the fuselage frames at STA 401 and STA 410, between plane 15 right and plane 15 left, exhibited high-temperature damage. In addition, localized high-temperature damage was also found on some polyimide-insulated wires in the FDC wire run, concentrated between STA 401 and STA 410. This wire run is routed adjacent to the middle conduit on the ceiling above Galley 2. These heat damage patterns are consistent with localized high-temperature chimney plume effects created by the presence of a vertical air space on each side of the centre riser duct along its outboard face. The tops of the two chimney plumes intersect the ceiling at STA 401 and STA 410.
When the IFEN control wire and PSU cables are positioned to simulate the outboard conduit, the cable layout matches the overall damage pattern in the assembly, with five wire arc locations aligning at approximately STA 401. This is consistent with the fluorinated ethylene-propylene conduit and ethylene-tetrafluoroethylene (ETFE) wire insulation being preferentially melted through at this location, causing multiple wire-to-wire arcing events. These events would likely trip the associated CBs and sever the wires at some arc locations, opening the electrical circuit and de-energizing these power cables.
Concentrated high heat damage was found directly above the R1 door, flapper door ramp deflector and above the adjacent wire bundles in the waterfall area. This damage was manifested in the form of broomstraw-like features on a localized region of the R1 forward door track that is attached to the bottom of the fuselage frames. On such a robust part, high heat is required over a relatively long time to create damage of this type. The high heat damage at this location is consistent with a chimney plume effect that channelled hot combustion by-products upward and impinged on the ceiling. The location of the broomstraw-like heat damage corresponds to the area above the inside radius of the elbow connection for the riser duct assembly. This geometry, together with other factors, such as the alignment of a vertical air space between the aft side of the aft riser duct and the cabin interior wall panel, favoured the formation of a plume as the MPET-covered insulation blanket combusted on the riser duct assembly. Further corroboration of such an event having taken place was the presence of other broomstraw-like features immediately adjacent to the same location, on the lower portions of an intercostal between STA 427 and STA 435, and nearby along the lower portions of a frame at STA 442 between plane 15 right and plane 15 left. Additional localized heating in the waterfall area would occur as hot combustion by-products were drawn under the riser ducts to the Galley 2 vent duct.
This localized heating would account for the missing tin coating on the three recovered IFEN PSU cables between STA 420 and STA 427, just aft of where they exited the conduit. At this location, the Exhibit 1-3790 PSU cable had arcs on each of its three phases. This same cable had also arced further forward within the conduit. The two separate arc locations on this cable are consistent with the fire propagating in the fore-to-aft direction, first causing an arc to take place at the forward position, not tripping the CB, then subsequently causing a second arc near the waterfall area that tripped the CB. The absence of arcs on the other two recovered PSU cables from the waterfall location, specifically in the area where the tin coating was missing from these cables, is consistent with these two cables being de-energized when the arc occurred on Exhibit 1-3790 at the waterfall location. This latter observation is also consistent with the two other PSU cables being previously de-energized by the tripping of their respective CBs when the multiple arcing events took place in the conduit. The latter further supports a fore-to-aft direction of fire propagation.
The burning of relatively large quantities of MPET insulation blanket cover material in the vicinity of the riser duct assembly would create a significant heat release. Although some of the combustion by-products would continue to be drawn down the ladder area and into the breached Galley 2 vent duct system, most of the by-products would flow upward in hot buoyant plumes. These by-products would form a hot buoyant layer along the upper attic air space above the forward cabin drop-ceiling.
The smoke barrier assembly would initially prevent most of the hot combustion by-products from flowing forward into the cockpit attic air space. Some leakage would be expected to take place, which would allow some by-products to penetrate into the flight crew compartment. Combustion by-products would also be drawn down the engine fire shut-off cable drop, where smoke could also leak into the cockpit interior. An indication of that flow having taken place was the presence of soot on some of the interior surfaces of the recovered cable drop shroud pieces.
The hot combustion by-products would heat the ceiling insulation and other items, including those below the hot buoyant layer, by processes such as radiant heating. Preheating and subsequent ignition of other materials would take place, including the ignition of the metallized polyvinyl fluoride (MPVF) insulation blanket cover material and splicing tape on the ducts. This in turn would be expected to cause ignition of the silicone elastomeric end cap situated on the end of the conditioned air branch duct, located approximately 30 cm (12 inches) aft of the cockpit door, above the ceiling panels. Failure of the end cap would cause a continuous release of conditioned air, which would further exacerbate the fire. An indication that this end cap had been breached by the fire was the presence of high heat damage on recovered portions of the branch duct where the cap was situated.
Release of conditioned air out of the branch duct would be directed slightly upward and laterally across the aircraft, toward the Galley 1 vent duct plenum, situated approximately 46 cm (18 inches) away from the end cap. This forced air ventilation would be expected to not only deliver conditioned air, but also entrain hot combustion by-products with it. A hot airflow (convective oven effect) would likely be created in certain areas along flow lines. Owing to the geometry of the ducts in the area, the flow would be channelled along a tapered path toward, and over, the top of the exposed portion of Galley 1. Indications that a hot airflow existed in this area were the broomstraw-like features that were present on the lower surfaces of intercostals and frames in the vicinity, and heat damage to the top of Galley 1. There was also high heat damage concentrated on the inboard side of the intercostals, which face the branch duct.
It is likely that the fire-induced failure of the branch duct silicone elastomeric end cap preceded and contributed to the failure of at least one, if not both, of the Galley 1 vent duct hose connections. The hoses were constructed from a fibreglass cloth, which was impregnated with a red-coloured silicone-like rubber material. Failure of the hose connection or connections would draw air and combustion by-products into the vent duct assembly. This exhaust ventilation would further exacerbate the fire. Indications that hose failure had taken place at some point during the fire was indicated by the abrupt cessation of heat damage along the top outboard face of Galley 1 at an elevation that corresponded to one of the hose connections, and by the presence of high-temperature heat damage on pieces of vent duct assembly just outboard of Galley 1.
The presence of high heat in the attic air spaces would likely cause the nylon fasteners holding the MPET-covered over-frame and between-frame insulation blankets to melt and fail. This would allow portions of the insulation blankets, or whole insulation blanket assemblies, to fall free, exposing more flammable MPET cover material to the fire. This in turn would significantly add to the growth and intensity of the fire.
The hot buoyant layer above the forward cabin drop-ceiling would be free to flow aft toward the empennage of the aircraft above the passenger cabin ceiling. Some of these by-products would continue to be drawn into the recirculation fan intakes, while these systems were operating. After passing through the intakes, the by-products would be delivered to areas within the passenger cabin. Soot patterns found on items such as wire support brackets, and on and within an overhead stowage bin located at STA 1780, indicate that the attic space above the passenger cabin probably became filled with combustion by-products. No smoke was reported in the passenger cabin prior to the flight recorders stopping.
Before the pilots selected the CABIN BUS switch to the OFF position at 0123:45, the airflow above the forward cabin drop-ceiling would have predominantly been in an aft direction, toward the input of the recirculation fans. Some smoke and combustion by-products would have been migrating into the cockpit. Several soot deposits were found in various places to indicate such seepage. At about this time, it is likely that the fire breached the silicone elastomeric end cap on a short branch stub on an air conditioning duct located immediately aft and overhead of the cockpit door. This would have allowed a large volume of conditioned air to enter the area and augment the fire. This additional airflow would have rapidly accelerated the propagation of the fire, as indicated by the high heat damage observed on the surrounding ducts and aircraft structure.
Selecting the CABIN BUS switch to the OFF position would shut down the recirculation fans, and result in a reversal of the airflow above the forward cabin drop-ceiling. With the airflow then moving predominantly forward, hot combustion by-products would have been drawn toward the cockpit attic air space.
It is likely that the weakened smoke barrier would have completely failed shortly after the CABIN BUS switch was selected to the OFF position. Although it is possible that the smoke barrier could have failed earlier, this is considered unlikely, because such a failure would have likely led to an earlier failure of the thermoformable plastic cockpit ceiling liner, which melts within a relatively low temperature range. Following the breach of the smoke barrier, hot combustion by-products could then freely fill the cockpit attic air space. The fill rate would likely exceed the exhaust rate, and a significant rapid build-up of heat and combustion by-products would occur. Hot combustion by-products would penetrate, fill, and rapidly heat the air spaces behind the avionics CB panel, overhead CB panel, and overhead panel housing. The MPET- and MPVF-covered insulation blankets above the cockpit ceiling would provide additional combustible material to the fire. This would further contribute to the amount of smoke entering the cockpit through passageways, such as the engine fire shut-off handle slots, and the various cut-outs in the cockpit ceiling liner.
The rapid heating of the air spaces and electrical components behind the CB panels, and within other assemblies, would have caused aircraft systems to malfunction. The heat would have thermally tripped some CBs that would likely have resulted in many of the anomalies that were subsequently recorded. For example, the Autopilot 2 disconnect event took place approximately 20 seconds after the CABIN BUS switch was selected to the OFF position; this was followed shortly thereafter by a series of recorded system anomalies.
The right side of the cockpit ceiling primarily consists of several aluminum panels that would act, in combination with other metallic assemblies such as the conditioned air diffusers, as a physical barrier to the fire and its combustion by-products. In contrast, the cockpit ceiling on the left side consists mainly of liner material that would soften, sag, and melt when exposed to high temperatures.
As the fire entered the cockpit attic area, the heat would have first affected the most exposed surfaces of the ceiling liner just forward of the cockpit door, in the area aft of the diffuser, and in the left overhead region. Very little liner material from these areas was identified in the wreckage. The few pieces that were identified, such as a portion of the cockpit spare-lamps cover and hinge (located in the ceiling adjacent to the cockpit coat closet), showed signs of melting. Some of the liner material may have been consumed by the fire. Pieces of the liner from other areas appear blackened and burned along some edges. Other pieces were also melted, and some had flowed until their cross-sectional area was reduced to the thinness of paper.
Based on the high temperatures involved, it is likely that the breach of the ceiling liner occurred approximately one minute after the smoke barrier failed. The breach of the ceiling liner may have corresponded to the time the pilots declared the emergency at 0124:42. Most of the fire damage on the cockpit carpet was likely the result of portions of the melted ceiling liner dropping on it. Larger amounts of dense noxious smoke and hot combustion by-products would be expected to have immediately penetrated through openings in the cockpit ceiling liner.
Once the cockpit liner had been breached, the openings in the liner would be expected to progressively expand, allowing a further increase in the volume of dense noxious smoke and combustion by-products into the cockpit. The smoke would be drawn down through the openings next to the rudder pedals into the avionics compartment. Visibility within the cockpit would be expected to become progressively worse.
It is likely that the fire would have breached the silicone elastomeric end cap situated on the end of a conditioned air branch duct, located above the lower assembly of the left overhead cockpit ceiling liner, just forward of the cockpit coat closet. The insulation cover splicing tape installed over the MPVF-covered muff assembly, which fixes the muff assembly in place over the end cap, would provide a source of combustible material. Once the tape was ignited, the integrity of the muff assembly would be lost and ignition of the silicone end cap would likely soon ensue. Heat damage and melting were found on the edges of the recovered liner at a location immediately adjacent to the end cap's position. Failure of the end cap would cause conditioned air to be continuously blown out the branch duct above the ceiling liner, in close proximity to the MPET-covered over-frame and between-frame insulation blankets, exacerbating the fire situation. It is likely that the hose to the individual air outlet near the centre of the left overhead cockpit ceiling liner was also breached, causing a similar effect.
Portions of a fuselage frame and conditioned air duct assembly near the hose and hose connection exhibited high-temperature damage. Cone calorimeter tests indicate that material similar to the hose in question ignites at a heat flux of 25 kW/m2 (which is approximately equivalent to a surface equilibrium temperature of 591°C (1 095°F)), and it is probable that the hose would not withstand exposure to such high temperatures. The same would also apply to the hose for the other individual air outlet located further forward in the liner, to the left of the overhead CB panel. Similar high-temperature fuselage frame damage was found nearby. Failure of the silicone elastomeric end cap and the individual air outlet hoses would introduce conditioned air to the fire, which would likely contribute to the deteriorating environment within the cockpit.
As the MPET cover material was consumed by the fire, the underlying fibreglass batting in the insulation blankets would become exposed and then badly scorched by the high temperatures and flames. The forcible release of conditioned air in close proximity to the insulation blankets would likely disturb and release fibreglass particulates from the ashen surfaces and from the less-damaged areas underneath these surfaces where the adhesive binder would be degraded.
After the completion of burn tests, the release of clouds of small particulates was observed to take place whenever the burnt insulation blankets were disturbed or removed from the test fixtures.
The captain's location would have been more directly in line with the area of the cockpit ceiling liner that was first breached. A higher percentage of the combustion by-products would be expected to flow directly toward the captain's seat location, and be drawn down into the avionics compartment through the captain's rudder pedal openings. The situation in the cockpit would have continued to deteriorate as systems malfunctioned and failed, owing to the effects of the fire.
Eventually, molten aluminum began to drip in the area of the right observer's seat, as indicated by the presence of resolidified aluminum deposits that were found on the recovered pieces of the seat. A 2024 aluminum alloy deposit was found on a screw on the right side of the seat pedestal, as well as on the right lap belt. There were also remnants of other aluminum deposits immediately adjacent to the 2024 aluminum alloy deposit on the belt. The type of alloy or alloys on these remnants could not be determined, as there was insufficient material available for analysis. Also, a 6061 aluminum alloy deposit was found on the CB for the seat, which is located near the rear right corner of the seat pedestal.
The sources of the aluminum deposits could not be conclusively determined; however, the possible areas from which 2024 aluminum alloy deposits could fall onto the right observer's seat are limited. Assuming that the integrity of the overhead dome light and the 6061 aluminum alloy diffuser assemblies were not compromised during the fire, the only major opening that could be created in the ceiling directly above the right observer's seat would be along a narrow rectangular-shaped area that predominantly comprises ceiling liner material. This area is approximately 7.5 cm (3 inches) in width by 76 cm (30 inches) in length. High heat damage is evident on portions of the recovered pieces of the diffuser assemblies and fuselage frames from above this location. Most of the upper edge of the recovered pieces of the avionics CB panel also exhibited high heat damage.
One known potential source of 2024 aluminum alloy is the AN 929-6 cap assembly, which is located on the crew oxygen supply line at STA 374. The end cap is situated above, and immediately adjacent to (within about 2.5 cm (1 inch)) the narrow rectangular-shaped area described above. This oxygen line cap assembly is in close vertical alignment with the right edge of the right observer's seat, near the right lap belt location and near the right side of the seat pedestal, when the seat is in the forward-facing position with the armrests stowed upright. Indications of broomstraw-like features were found along the bottom edges of the fuselage frame at STA 374 just above, and adjacent to, the cap. This further indicated that high temperatures had existed at this location.
Testing on the oxygen cap assembly indicated that before leakage or failure of the cap occurred, it would have to be heated at elevated temperatures for several minutes (see Section 1.14.13). These elevated temperatures were below the temperature at which external melting was visible on the cap. Therefore, if the time at which the CABIN BUS switch was selected to the OFF position is used as a reference for when significant elevated heating took place in the cockpit attic air space (approximately 8 minutes, 30 seconds, prior to the time of impact), the most likely time frame that melting of the cap could take place would correspond to the final stages of the flight.
If pure oxygen leaked from the cap during the fire, there would almost certainly be a quick and dramatic increase in the fire intensity. This would be expected to rapidly lead to a complete failure of the cap. A complete failure of the cap would result in a loss of pressure in the line and would abruptly stop the flow of oxygen to both pilots' oxygen masks. In addition, full venting of the line would be expected to quickly lead to a flashover within the cockpit, or an intense conflagration, or both. There was little physical evidence of an overall high-temperature damage pattern in the cockpit interior; therefore, it is likely that if this occurred, it was of a very short duration, and it occurred immediately prior to the time of impact.
 Cone calorimeter is a bench-scale test apparatus consisting of a cone heater, spark ignitor, sample holder, and a load cell located under a hood. It is widely used to determine the heat release rate of combustible solids.
- Date modified :
- Date modified: