Aviation Investigation Report A98H0003
- 126.96.36.199 - General
- 188.8.131.52 - Positioning of the IFEN Power Supply Cable Segments Exhibiting Arcing Events
- 184.108.40.206 - Three Identified Arced Wires from Aircraft Systems
- 220.127.116.11 - Nine Arced Wires Locations Not Determined
- 18.104.22.168 - Hella Map Lights
- 22.214.171.124 - Inside the Overhead Switch Panel Housing
- 126.96.36.199 - Forward Galleys
- 188.8.131.52 - Overhead Aisle and Emergency Light Fixtures
- 184.108.40.206 - Emergency Lights Battery Pack
The wreckage reconstruction mock-up helped delineate the boundaries of the fire damage, which were primarily located in the cockpit attic and forward cabin drop-ceiling areas. Within the primary fire-damaged area, the most prevalent potential ignition source was electrical energy. Numerous power cables and wires were present in this area, running to or from either the avionics compartment, the cockpit CB panels, or the overhead switch panels. There were also numerous other electrical components, such as module blocks, ground studs, light fixtures, battery packs, two electrically powered galleys, two lavatories, and electrically powered door mechanisms. Consideration was given to other ignition mechanisms, such as chemical reactions, thermal processes (e.g., conduction, convection, radiation) and mechanical operations (e.g., friction) but none were found. Only those related to electrical energy were assessed as being pertinent.
Each of the 20 wire or cable segments exhibiting arc damage was assessed for potential involvement as an ignition source. Several criteria were taken into account when making this assessment, such as information from the CVR about the cockpit and passenger cabin environment, the time frame between when the odour was first noted and when the fire first affected aircraft systems, the presence and proximity and quantity of flammable materials, the fire damage pattern, and the airflow patterns in the aircraft as determined through flight testing and research.
The initial odour and smoke were noticed only in the cockpit, and the pilots assessed it to be from an air conditioning source. The actions by the pilots, and the airflow patterns in the cockpit area, suggest that the smoke was most evident at or near the cockpit rear wall. (For additional airflow information, see Section 1.16.3.)
When the four IFEN PSU cable segments that exhibited melted copper were positioned parallel to one another, similarities were noticed that suggested that there were two matching pairs of cable segments (see Figure 22 and Section 220.127.116.11). Exhibits 1-3791 and 1-3793 were almost identical in length and colouration, as were 1-3790 and 1-3792.
Once the two pairs were matched together, there were notable differences that distinguished one pair from the other. The 1-3790/1-3792 cable pair was about 18 cm (7 inches) shorter than the 1-3791/1-3793 pair, and starting at about 61 cm (24 inches) from the straightened-out end the 1-3790/1-3792 pair had an almost identical span of approximately 30 cm (12 inches) where the tin coating was completely missing. Over the last 5 cm (2 inches), the tin was again present. The 30-cm (12-inch) region of missing tin was the result of exposure to a localized temperature that was higher than the temperature experienced in the area where the tin remained intact.
Exhibit 1-3790 had copper melt on a single wire 11 cm (4.3 inches) from its end, and additional copper melt on all three wires between 62 and 77 cm (24.3 and 30.5 inches) from their respective ends. Exhibit 1-3792 had copper melt on one wire at 22 cm (8.8 inches), and on a second wire at two locations: 22.6 cm (8.9 inches) and 23.4 cm (9.2 inches) from its end position. The second wire was severed between the two copper melts.
Cable 1-3791 had copper melt on one wire 9 cm (3.7 inches) from its end and at the 62-cm (24.5-inch) location. A second wire also had a copper melt at the 65-cm (25.8-inch) location. Cable 1-3793 exhibited melted copper on one single wire between 64 to 67 cm (25 to 26.5 inches) from its end.
The two pairs of cable segments could not be co-located by aligning all of them to either one end or the other, because the region of missing tin on the 1-3790/1-3792 pair was not duplicated on the 1-3791/1-3793 pair. As the cables had been installed as a contiguous bundle, it is not possible for only two of the four cables to be affected by a localized higher temperature.
Based on the presence of FEP remnants trapped in the wire strands, the PSU cable pairs were considered to have been either in, or in close proximity to, the conduit located above Galley 2. The recovered segments of wire runs FBC and FDC, which had been routed parallel to the conduit in the area over Galley 2, were positively relocated in the aircraft based on the installed position of the marriage clamp at approximately STA 427. It was noted that from approximately 30 cm (12 inches) aft and forward of the clamp along the FDC wire run, and for approximately 30 cm (12 inches) forward of the clamp along the FBC wire run, there was a region that appeared to have been subjected to a higher temperature than the remaining wires in the harness, similar to the region of missing tin on the 1-3790/1-3792 pair of cables. On both sides of the dual clamp, the polyimide film insulation remaining on the FBC and FDC wires was blackened, similar to test samples that had been exposed to a temperature of about 500°C (932°F) for 10 minutes.
A 38-cm (15-inch) segment of IFEN PSU cable with a 30-cm (12-inch) section of tin missing from all three wires was recovered and designated as Exhibit 1-4687. This cable segment was considered to have been from the area near the FBC/FDC clamp, and to be a continuation of either Exhibit 1-3791 or 1-3793. The location on Exhibit 1-4687, where there was a transition between tin and missing tin, was aligned at the aft end of the conduit such that it matched the start of the missing tin on the 1-3790/1-3792 pair. When the three cables were placed with their regions of missing tin aligned, it was noted that the individual wires within the region of missing tin were similarly embrittled. Furthermore, two of the cable segments, exhibits 1-3790 and 1-4687, had nearly identical twists in individual wires that were adjacent to each other in this alignment. The frayed (aft) end of Exhibit 1-3791 closely matched in colour the end of Exhibit 1-4687 where the tin was not missing. The lengths of the fractured ends were of similar lengths, suggesting that they could have been joined at one time. Based on this similarity, Exhibit 1-4687 was considered to be a continuation of Exhibit 1-3791. When the area of missing tin on the 1-3790/1-3792 pair and 1-4687 was aligned on either side of the clamp, and taking into account the known routing of the cables as they exited the conduit, the beginning of the missing tin was placed at approximately the aft end of either the middle or outboard conduit, or possibly just inside the conduit by less than 2.5 cm (1 inch).
Aligning the cables in the outboard conduit using the same criteria as above for placing the region of missing tin placed the single arc on 1-3790 near STA 410 and the remaining (aft) melted copper locations on all three wires from Exhibit 1-3790 approximately 25 cm (10 inches) along the wires aft of STA 420. This placed the copper melts on two of the wires from Exhibit 1-3791 adjacent to the copper melts on the single wire of Exhibit 1-3793 near STA 401. This also located the single copper melt on the forward end of Exhibit 1-3791 approximately 8 cm (3 inches) along the wire forward of the bracket at STA 383. This configuration is shown in Figure 30.
Aligning the cables in the middle conduit, the single (forward) melted copper on Exhibit 1-3790 was located near STA 407 inside the conduit, and the remaining (aft) melted copper locations on all three wires from Exhibit 1-3790 were about 7.5 cm (3 inches) outside the aft end of the conduit. In this configuration, the remaining areas of melted copper on Exhibit 1-3792 were in the conduit between STA 407 and STA 408. In this alignment, it placed the forward end of the 1-3791/1-3793 cables approximately 25 cm (10 inches) along the wire path past the bracket at STA 383. This placed the copper melts on two of the wires from Exhibit 1-3791 adjacent to the copper melts on the single wire of Exhibit 1-3793 near STA 398. This also located the single copper melt on the forward end of Exhibit 1-3791 approximately 15 cm (6 inches) along the wire path outside of the forward end of the conduit. This configuration is shown in Figure 31.
The arced ends of the four 16 AWG control wires (1-3788, 1-3794, 1-3795, and 1-10503) could have been the result of two arcing events severing the wires. The arced ends of exhibits 1-3795 and 1-10503 were a possible match, based on the similarity of the missing tin over a short distance from each of their arced ends. Assuming the arced ends of 1-3788 and 1-10503 were also a match, and laying out the four segments as one continuous wire, their overall length was within a few centimetres of that of the 12 AWG 1-3790/1-3792 pair. Laying the combined 16 AWG control wire alongside the 1-3790/1-3792 pair also showed that their arc locations could be co-aligned, indicating they may have been caused at the same time. However, Exhibit 1-3795 was not missing tin over the same length as the 1-3790/1-3792 pair; this made it less likely that they could be co-aligned.
The regions of melted copper on the 1-3791/1-3793 pair were co-located, suggesting that the arcing events that produced the copper melts occurred at approximately the same time, or within seconds of each other. All of these arcs occurred where the cables would have been running through the conduit, suggesting phase-to-phase arcing occurred when the fire destroyed the conduit and wire insulation.
The one anomaly is the copper melt on a single wire near the forward end of Exhibit 1-3791. This same wire also exhibited melted copper approximately 51 cm (20 inches) further aft, within the conduit. The melted copper in two locations indicates that one of the arcing events did not trip the associated CB. The arcing events in the conduit had the appearance of being more severe, and it is believed that they would have tripped the CB. Therefore, the forward arc event on Exhibit 1-3791 must have occurred first, as electrical power would not be available to produce the forward arcing if the arcing in the conduit had occurred first and tripped the CB.
The melted copper on the 1-3790/1-3792 pair within the conduit also appears to be the result of the fire destroying the conduit and wire insulation. This allowed arcing to take place between phases, and also with the 16 AWG control wire. In the placement of the 1-3790/1-3792 pair, as described above, the single (forward) arc on Exhibit 1-3790 did not align exactly with the melted copper on Exhibit 1-3792 inside the conduit. The initial alignment resulted in an approximate 5-cm (2-inch) separation between these arcs.
As the single (forward) arc on Exhibit 1-3790 occurred inside the conduit, it had to have arced to either another IFEN PSU cable, or the 16 AWG control wire. Therefore, the original positioning of the 1-3790/1-3792 pair of cables had to be adjusted slightly (within the 5-cm (2-inch) range) to align all of the copper melts. As with Exhibit 1-3791, the fact that a single arc event took place on one wire that subsequently arced approximately 51 cm (20 inches) farther aft indicated that the initial arcing event did not trip the CB. The arcing on Exhibit 1-3790 that occurred outside the aft end of the conduit involved all three phases. This would have tripped the CB, indicating that the forward arc occurred first and did not trip the CB.
Various combinations were tried to determine the best placing for the arc-damaged IFEN PSU cable and 16 AWG wires segments, particularly with respect to their positioning in relation to the localized heat zone noted on either side of the two clamps. However, all of the combinations had to take into account the regions of missing tin on the 1-3790/1-3792 pair of PSU cables. For the various layouts, this region of missing tin was always matched to the area of the two clamps on FBC and FDC, as this was a known location from which to start. No other combination or layout could be supported by the physical evidence on the cables and wires.
As indicated above, based on the layout of the cables and wires, it is possible to assess the direction of fire propagation based on the sequence of the arcing events. The aft arcs on exhibits 1-3790 and 1-3791 would most likely have tripped their associated CBs; therefore, the forward arcs must have occurred first, and not tripped their associated CBs. This strongly suggests that the fire was moving forward to aft.
The region of missing tin represented an area of higher heat. If this had been an earlier event, it is strongly suspected that all three cables would exhibit similar arcing events in the vicinity of 1-3790 outside the aft end of the conduit. With the cables and control wire laid out as described above, all of the regions of melted copper could be attributed to arcing events between phases on the same cable, between the cables and the control wire except for the arc event that took place at the forward end of 1-3791, or both. There was no matching arc damage on cable 1-3793 in this area. However, the continuation of exhibits 1-3790/1-3792 from approximately STA 401, along with the 16 AWG control wire forward, were not identified. Therefore, no determination could be made with respect to exactly what the single wire from 1-3791 had come in contact with at the forward end.
The three exhibit numbers discussed in this section are 1-3029, 1-1733, and 1-12755.
Of the three identified arced wires from aircraft systems, two were accurately placed such that the arcs were located about 15 cm (6 inches) aft of the right oval hole in the overhead switch panel housing. (See Figure 23.) The arc on the third wire was located in the same general area. From a potential arcing perspective, this area is considered benign in that the wires, as installed, have little likelihood of being chafed or otherwise damaged. However, there are MPET-covered insulation blankets in close proximity to the wires. The three wires were assessed regarding their potential to be related to the lead ignition event.
Exhibit 1-3029 is a section of 10 AWG tin-coated, XL-ETFE (BXS7008) insulated wire from the left emergency AC bus feed cable. The area of melted copper was approximately 15 cm (6 inches) outside of the right oval hole in the overhead switch panel housing. The functions powered by the left emergency AC bus were lost at 1:25:06. The arc on this wire would have tripped the left emergency AC bus remote controlled CB. If this arcing was connected to the lead event, the loss of this bus would have been obvious to the pilots, and the loss of associated systems would have been recorded on the FDR much earlier than when they were actually recorded. The arc on this wire was the result of fire damage, and was not connected to the lead ignition event.
Exhibit 1-1733 is a section of 24 AWG nickel-coated, polyimide-insulated wire from the Engine 2 fire detection loop "A" circuit. The wire was severed by an arc event at one end. The area of melted copper occurred about 15 cm (6 inches) outside the right oval hole in the overhead switch panel housing. The arc on this wire would have caused the Engine 2 fire detection loop "A" circuit to open, or the associated CB to trip, causing loop "A" to be de-powered. The fire detection control unit would then send a fault to the DEU, which would be displayed to the crew as a Level 1 (amber) "FIRE DET 2 FAULT" alert on the EAD. A fire alarm would not be generated, nor would an overhead warning light illuminate. There was no mention by the crew on the CVR of any alerts being displayed before the smoke appeared in the cockpit. This arcing event most likely occurred as a result of fire-related damage to the wire, and was not likely connected to the lead ignition event.
Exhibit 1-12755 is a section of 22 AWG nickel-coated, polyimide-insulated wire from the high-intensity lights (supplemental recognition) wingtip strobe lights circuit. The high-intensity wingtip strobe lights are turned on in the cockpit by the HI-INT push button located in the overhead light switch panel. When the lights are off, the button illuminates a blue OFF legend. This wire is powered when the lights are on. The loss of power to the wire will result in the high-intensity lights shutting off, but the OFF legend in the switch will not illuminate as long as it is in the ON position and the left ground sensing relay R2-5009 is powered. Therefore, a shorting of this wire would be a hidden event for the crew unless a CB tripped and was noted.
Exhibit 1-12755 was severed by an arcing event at one end. In the reconstruction mock-up, the location of the wire could not be precisely determined; however, it had been installed in the wire run between the forward switch panel receptacle R5-204 in the overhead switch panel housing and the electrical connector plug P1-420 in the overhead disconnect panel. There is no indication that any wire arcing occurred within the overhead switch panel housing; therefore, the arc on this wire likely occurred between the overhead switch panel housing and the disconnect panel, most likely at about 15 cm (6 inches) outside the right oval hole in the same location as the other known fire-related arcing in that area. This suggests that the arcing event occurred as a result of fire-related damage, and was not connected to the lead ignition event.
The nine exhibit numbers discussed in this section are 1-3700, 1-3713, 1-3718, 1-3796, 1-4689, 1-11252, 1-11897, 1-12756, and 1-12809.
The exact installed location and system function of each of the nine arced wires could not be determined. It is highly unlikely that all nine wires arced at the same location and time.
Exhibits 1-4689 and 1-11897 were identified as sections of 10 AWG tin-coated wire, with the insulation missing. Based on the tin coating, it is likely these two sections were insulated with MIL-W-22759 type insulation. The left emergency AC bus displayed arcing damage; this wire was 10 AWG tin-coated, with MIL-W-22759/34 type insulation. It is also known that power to this bus feed was lost at 0125:06, when numerous functions powered by this bus were simultaneously lost.
Loss of the left emergency AC bus would result in the left emergency DC bus being powered by the aircraft battery, and the left emergency AC bus being powered through a static inverter also powered by the battery. Although the left emergency AC remote control CB (RCCB) most likely tripped as a result of the arcing event at 0125:06, there is no CB protection for the left emergency AC bus when powered from the static inverter. The current is limited only by the inverter itself; therefore, electrical power would continue to be fed to any short-circuit until the inverter itself failed. Therefore, it is possible that exhibits 1-4689 and 1-11897 could be from the left emergency AC bus feed.
Other 10 AWG wires routed within the fire-damaged area were also assessed. Three systems or components were powered by 10 AWG wires. One 10 AWG wire was installed completely within the overhead switch panel housing, where it is unlikely that any arcing took place. The remaining two 10 AWG circuits were associated with the tail tank alternate fuel pump and the Tank 2 left aft fuel pump. Both of these three-phase power circuit wires were routed through the right side of the forward cabin drop-ceiling.
Of these two pumps, only the Tank 2 left aft pump would have been powered at the time of the initial odour. According to the wiring diagram for this circuit, the three 10 AWG wires, C104-147(148) (149)-10, were routed between STA 475 aft to STA 1059. Also, according to the Boeing conduit list, the 10 AWG Tank 2 left aft fuel pump wires were polyimide insulated and nickel coated, not tin coated. As the most forward location for these wires was STA 475, they were considered to have been located too far aft to have been involved as a potential source of ignition.
Exhibit 1-11252 was a section of 24 AWG nickel-coated, polyimide-insulated wire. The physical appearance of Exhibit 1-11252 was almost identical to Exhibit 1-1733 (a segment of the Engine 2 fire detection loop "A" circuit) and could be a matching end, and as such a continuation of that wire.
Exhibits 1-3700 and 1-3718 were sections of 20 AWG nickel-coated, polyimide-insulated wires, and exhibits 1-3713 and 1-12809 (see Figure 20 for a photograph of 1-12809) were sections of 24 AWG nickel-coated, polyimide-insulated wires. These wires could not be associated with any particular circuit or system. During the final 92 seconds of the FDR operation, numerous systems anomalies were recorded. Two of the recorded anomalies were related to systems that were powered by 20 AWG wires, and six were related to 24 AWG wires. Any of these anomalies could have resulted from either an arc or from thermal tripping of the system CB.
Exhibit 1-12756 was identified as a section of 18 AWG tin-coated, polyimide-film-insulated wire. This wire was unusual in that nickel-coated, polyimide-film-insulated wire was the standard used during the manufacture of the aircraft. Exhibit 1-12756 could be related to a modification made to the aircraft following manufacture, but it could not be positively identified. As this wire could not be placed in the reconstruction mock-up, its potential involvement in the lead event could not be assessed from a systems perspective.
Exhibit 1-3796 was recovered as part of the bundle of entangled wires (Exhibit 1-4372) that contained the arced IFEN PSU cable segments, suggesting that it may have been installed in the same area of the aircraft. It was determined that these PSU cable segments had been installed above Galley 2 (see Section 18.104.22.168), starting in the vicinity of the cut-out in the top of the cockpit rear wall and running aft to about STA 420. Exhibit 1-3796 was assessed for its potential to be involved with the lead arcing event.
The area between the cut-out in the top of the cockpit rear wall aft to STA 420 (above Galley 2) was inspected on all of Swissair's fleet of MD-11s for potential anomalies that could lead to arcing. The wire bundles and conduits run relatively straight in that area, and the area is not considered to be susceptible to damage from routine maintenance operations or contamination. The threat from mechanical wire chafing in that area was considered to be low. During the inspections, no potential anomalies that could lead to arcing were found.
To create the 2-cm copper melt on Exhibit 1-3796 would require a significant arc-tracking event. Such an arc-tracking event would almost certainly have involved at least one other wire arcing, and would have resulted in significant damage to a number of adjacent wires. Significant collateral damage to nearby wiring is often seen when similar arc tracking occurs in laboratory testing, and may account for some of the other arced wires that were found but not positively identified as to location or circuit function.
Exhibit 1-3796 was also assessed for its potential to be connected to a lead arcing event in the area forward of the cut-out in the top of the cockpit rear wall, behind the avionics CB panel. Such a scenario is considered unlikely for the following reasons. In laboratory testing, similar arcing events are known to produce a series of loud snapping-type sounds. These loud snapping sounds are typically accompanied by brilliant flashes of light similar to arc welding. In the MD-11, such flashes could potentially be seen in the (darkened) night cockpit lighting conditions through the small openings around the CB panel, or around the edges of unused CB holes. It is unlikely that the type of arcing event that produced Exhibit 1-3796 could be mistaken by the pilots as an air conditioning anomaly if it occurred behind the avionics CB panel. If this type of arcing event occurred behind the avionics CB panel, it would be expected that the arcing would produce significant damage to adjacent wiring. During such an event, it is most likely that one or more systems alert messages would appear, anomalies would be recorded on the FDR, or CBs would trip. Again, no such anomalies were mentioned by the pilots or recorded on the FDR for some 13 minutes.
The first officer's and the right observer's map lights were recovered, examined, and ruled out as potential sources of ignition for the fire. The captain's and the left observer's map lights were not found in the recovered wreckage. The map light fixture installed at the left observer's position did not have any history of electrical anomalies similar to the Hella map lights, and it was ruled out as a potential ignition source.
The Hella map light in close proximity to the MPET cover material at the captain's map light position presented a potential lead ignition-event scenario. Airflow flight testing showed that some of the test smoke generated above the cockpit ceiling at the captain's map light position would enter the cockpit around the left window air diffusers. It would almost certainly enter from one or more other locations also, including the map light housing, the left-side window diffuser slide control, and the engine fire handles.
If the fire started immediately overhead of the captain, it would be expected that he would detect the odour first; it appears that this was not the case. Also, if the initial smoke was coming from the area of the map light, it would not have been necessary for the first officer to stand to inspect the suspected area. Smoke entering through openings remote from the diffusers, especially through the map light housing itself, would be less likely to be mistaken for air conditioning smoke. Furthermore, a lead ignition event this far forward in the aircraft would not lead to the substantial fire-related damage that occurred in the attic area of the forward passenger cabin in the known time frame of the fire. The available information indicates that the fire did not start in any of the map lights.
The examination of the recovered material from inside the overhead switch panel housing showed little heat damage other than to a localized area at the aft end. There was no indication that the fire started inside the housing and propagated out. The heat damage pattern shows that the heat originated outside the housing, and entered through the aft cut-outs.
Galley 2 was not electrically powered, and none of the recovered wires from inside galleys 1 and 3 displayed any heat or fire damage. All of the heat damage and soot accumulation on the top portions of these galleys was from exposure to an external fire. There was no arcing damage to any of the identified galley wiring that was recovered.
The galleys would have been in use at the time of the detection of the initial odour in the cockpit. If a galley power feed wire were to arc, the galley load control unit would sense a differential between the input and output, and would shut off power to the galley. The appropriate galley OFF light in the cockpit overhead control panel would illuminate, and a galley OFF Level 1 (amber) alert would be generated. This would likely have been apparent to the flight crew. There were no CVR references to any galley problems. The available information indicates that the fire did not start within one of the galleys.
Discolouration was found on some of the cabin ceiling panel assemblies, both in the wreckage from SR 111 and during subsequent examinations of other MD-11 aircraft. The discolouration was caused by an overheating of the overhead aisle and emergency light assemblies by the lamp. Other than the potential for dust and lint deposits, there is no flammable material, such as MPET cover material, in close proximity to the light fixtures. It is assessed that the fire was not initiated by one of these light fixtures.
Examination of the battery pack showed that, although it was extensively heat damaged, the heat occurred from the outside in (see Figure 27). This indicates that the fire did not start from a heat condition, such as thermal heating, within the battery pack.
- Date modified :
- Date modified: