On 30 September 2015, a Bombardier DHC-8 400 being operated by an airline based in Luxemburg and holding a Luxemburg AOC on a scheduled passenger flight from Hamburg to Luxemburg via Saarbrucken failed to get airborne in normal day visibility at the end of its take-off roll when the landing gear unexpectedly retracted. It then slid almost to the end of the runway before stopping. There was no fire but an immediate passenger emergency evacuation was initiated by the cabin crew because of "smoke and an acrid metallic smell" which had developed during the slide along the runway. None of the 20 occupants were injured but the aircraft sustained severe damage.
The aircraft in its final stopping position. The superimposed lateral lines indicate the distance of various runway skid marks from the first fuselage runway contact point. The superimposed longitudinal lines adjacent to the aircraft track show right and then left propeller impact marks. [Reproduced from the Official Report]
An Investigation was carried out by the German BFU. The 500 hour 260 parameter FDR and the 2 hour CVR were "seized and read out" by the Investigation. It was found that both recorders had stopped recording 6 seconds after the first fuselage ground contact following gear retraction. The reason for this was noted to have been the operation of a 5.5g inertia switch which was removed for function testing and found free of any defects.
It was found that the 45 year-old male Captain had about 11,927 total flying hours which included 3,649 hours on type. The 27 year-old female First Officer had about 3,295 total flying hours which included 1,483 hours on type. It was noted that the day of the accident was her first flying day following 16 consecutive days leave. Both pilots held Luxemburg licences.
It was established that the flight crew had operated from Luxemburg to Hamburg via Saarbrucken and were returning the same way. The first three flights had been uneventful. The fourth sector began with the aircraft taxiing well ahead of schedule and a reduced power take-off was planned from runway 09 in benign weather conditions with the Captain as PF. Normal callouts were made by the First Officer but 2 seconds after the "V1, rotate" call, a sound commensurate with the landing gear lever being moved was recorded on the CVR followed immediately by the First Officer saying "oops sorry".
FDR data showed that the gear had retracted as the aircraft was being rotated through approximately 5° nose up and a tail strike had followed almost immediately which led to the tail strike warning light illuminating. After this the fuselage bounced three times. The aircraft then skidded on the fuselage for just over 800 metres. The aircraft’s final stopping position was abeam the runway 27 PAPI installation, approximately 875 metres after the point where the initial tail strike had occurred and approximately 425 metres from the end of the 1,990 metre-long runway. The main landing gear was completely retracted with engine cowling bay doors closed too. The nose landing gear was also retracted but the front nose landing gear doors were open and damaged. The landing gear selector lever in the flight deck was in the "UP" position. The blades on both propellers had been shortened by about 40 cm (left side propeller) and about 1 cm (right side propeller). Commensurate propeller impact marks were found on the runway. The entire length of the left lateral fin below the fuselage was abraded to a depth of about 3 cm and the bottom surface of the fuselage was "scratched, dented and completely abraded" between about 2 metres aft of the nose landing gear bay and the area of the aft baggage compartment.
The damaged bottom of the fuselage looking forward [reproduced from the Official Report]
None of the approximately 1,630 kg of fuel on board the aircraft leaked. After recovery of the aircraft to a hangar, tests showed that there was no evidence of any malfunction.
It was confirmed that as per design, the landing gear would retract if selected up when the nose gear was airborne even if both main gears were still on the ground. Relevant QAR data from the airline was examined and no instances of premature gear lever actuation at take-off were found. However, none of a number of the Operator's DHC8-400 type-rated pilots who were interviewed were found to be "aware that the landing gear may retract even if only the nose landing gear is airborne". The simulator used by the Operator for pilot training was function tested and it was found that early selection of gear up with the nose gear airborne but the main gear still on the ground did not cause gear retraction.
The history of unintended premature gear up selection was reviewed and it was confirmed that the vast majority of such occurrences had been recorded during landing rather than take-off. In the investigated case, the First Officer was unable to explain her action and the Captain had no warning that she was going to grasp and immediately activate the gear selector. It was noted that human performance specialists usually describe an error of this type by a relaxed and normally competent pilot as a "slip" typically defined as an incorrect or ill-timed action which occurs unintentionally in place of either a correct action/timing which is a routine, frequently trained and regularly repeated action. It was noted that actions which are regularly repeated can "generate reduced concentration [and] are susceptible to this kind of error [and] additional training and checks take no effect because no one is immune to such errors". Minimising the prevalence of such errors sometimes depends on warnings of incorrect action being generated but as much as possible on error tolerant design. The gear selector in the DHC8-400 was found to be "unusual compared to others in transport aviation" in that it has a lock release button which has to be pushed before the lever can be moved. Bombardier stated that this feature "should prevent unintentional actuation on the ground" in cases where at least one weight on wheels sensor is signalling 'air'. However, on most other aircraft types, all landing gear assemblies have to not be in ground contact before the gear will retract in response to gear lever movement and release buttons are only used as override enablers in a situation where normal selector movement is not possible.
The fact that the (correctly functioning) inertia switch had stopped both crash recorders running when they would otherwise have been able to provide information which would have been useful to the Investigation, was noted as having been identified in other investigations carried out by the BFU, the UK AAIB, the NTSB and Canadian Transportation Safety Board. The 2003 EUROCAE MOPS specification for these recorders already states that such 'g' switches should not be used and the UK AAIB have made a 2008 Safety Recommendation inviting the FAA and the EASA to review corresponding certification requirements. As a result of this concern, it was noted that the switch manufacturer involved was reported to be "committed to updating their g-switch product line to incorporate time-delay features that would continue to supply power to the system for a predetermined interval following a switch activation".
The formally-recorded Conclusion of the Investigation was that the accident "was the result of an early retraction of the retractable landing gear during take-off, which was not prevented by the landing gear selector lever and the retracting control logic".
Four Contributory Factors were also identified:
- The First Officer's reduced concentration level
- A break in the callout process / task sequence on the part of the First Officer
- The actuation of the landing gear lever to the UP Position too early
- Landing Gear control logic design allows retraction of the landing gear with one wheel airborne.
Safety Action taken as a result of the investigated event was noted as including the following:
- The Aircraft Operator has added an OFDM event aimed at tracking the prevalence of premature landing gear retraction
- the Aircraft Manufacturer has devised a modification to the design logic on which gear selector function is predicated so that all gear assemblies must be airborne before gear selector movement to 'UP' will begin gear retraction. This change has been mandated by a Transport Canada AD which requires compliance within 18 months of publication.
The Final Report was completed on 31 October 2016 and published the following month. In view of the Safety Actions noted, no Safety Recommendations were made.