Mode Reversion
As the aircraft neared the initial approach fix (LAPAR) for the Darwin Airport runway 29 instrument landing system (ILS)‑Z approach, the approach, localiser and glideslope modes were armed with the correct ILS frequency set as the armed navigation source. While flying by LAPAR, the localiser mode engaged for one second before the autoflight system reverted to lateral roll and vertical flight path angle modes. The captain recalled that the reversion occurred because they received a second ATC clearance for the ILS; however, at that time they had only been cleared once. A second clearance came about 50 seconds later and a third 40 seconds after that. The pushbutton that would disarm approach mode was not recorded in the data, but pressing it when approach mode was armed would be consistent with the reversion.
Embraer's analysis of the recorded data found that the ILS navigation display source had not engaged when localiser mode became active. It was most likely that this non‑synchronisation led the flight control and guidance system to invalidate the capture and revert to basic modes. The avionics manufacturer, Honeywell, advised that because the recorded data did not update as frequently as the system status, it was not possible to determine whether a synchronisation issue existed. The mode reversion resulted in the aircraft maintaining the 20° right bank and approximately level flight path angle that were present at the time of reversion.
Automation and Approach Continuation
Alliance’s procedures stated that if the autoflight system was not doing what the pilot expected, the pilot should disconnect the autopilot and manually fly the aircraft. Following the unexpected mode change, the captain, who was the pilot flying (PF), did not re‑engage approach mode or disconnect the autopilot until about 200 ft above the aerodrome elevation, in accordance with normal landing procedures. This resulted in the aircraft deviating beyond the permitted tolerance from the localiser course and above the glideslope.
It took less than 3 minutes from the mode reversion at 3,000 ft until the aircraft was established on the ILS and in landing configuration about 400 ft above aerodrome elevation. During that time, there were several changes to lateral and vertical modes, selected heading and altitude, and significant variations in vertical speed as the PF manipulated the autoflight system to recapture the localiser and glideslope. The flight crew reported that en‑route weather deviations and thunderstorms in the vicinity of the airport may have led to perceived pressure to continue the approach, but there was adequate fuel to discontinue and conduct a second approach.
Once outside the lateral tolerance of the approach, the aircraft was operating contrary to the ATC clearance and descended below the minimum safe altitude, likely while in instrument meteorological conditions (IMC). The pilot monitoring (PM) did not call out the course deviations, as the PF was actively working to correct them. After descending about 500 ft, the PM advised ATC they were re‑intercepting the localiser and was assigned the minimum vector altitude to ensure terrain separation. The aircraft then maintained altitude while the glideslope diverged beneath the flight path. The PM did not read back the ATC clearance to descend once they had intercepted the glideslope, consistent with being focused on monitoring tasks and their self‑assessed 10/10 workload during the approach.
The PM selected flap 2 as the aircraft captured the localiser, about 800 ft below the normal 3,000 ft flap 2 selection height. Alliance reported that this late selection had a similar effect to other recent approach incidents, where the associated delayed speed reduction reduced the time available to stabilise the approach. As the aircraft passed the final approach fix (FAF) and descended through 1,500 ft, it was established on the localiser but 3 dots above the ILS glideslope. Alliance’s procedures required the ILS to be flown within 1 dot of the glideslope and stated that aircraft ‘should be’ on the 3° glideslope no later than 1,500 ft.
The PF reported stating to the PM that they were visual at about 2,500 ft, but the PM reported this occurred just as the aircraft approached 1,000 ft. Without a cockpit voice recording, it could not be determined when visual flight conditions were established. At 1,000 ft, the stabilisation height for an instrument approach, the PF incorrectly assessed that although several stabilised criteria were not met, as they were then in visual conditions, the approach could continue.