The US National Transportation Safety Board has reported that a military vehicle carried in a National Airlines Boeing 747-400F “most likely” broke free from restraints during the take-off rotation at Camp Bagram airfield, Afghanistan, rolled aft and punctured the aft pressure bulkhead, damaging control systems moments before the aircraft pitched up unsustainably and crashed.

The agency believes that factors in addition to a shift in the aircraft’s centre of gravity contributed to the aircraft’s loss of control. These NTSB and Boeing findings were made available in advance of the final report.

The 747-400 converted freighter (N949CA) crashed on 29 April 2013 shortly after take-off on a flight to Dubai, killing all seven crewmembers.

Video from a passing ground vehicle’s dash camera shows the aircraft pitching steeply nose-up shortly after take-off, then rolling right, falling and exploding on impact with the ground.

Documents from the NTSB and a report from Boeing, which examined the flight data recorder at Afghanistan’s request, show that the cargo consisted of five MRAP (mine-resistant ambush protected) armoured military vehicles. The aircraft was loaded at Camp Bastian and made a fuel stop at Bagram en route to Dubai.

The reports note that shortly before take-off the pilots discussed that one of the straps had broken and one of the vehicles had shifted “a couple inches” during the previous flight. The vehicles could not be driven on to the aircraft, so technicians loaded them on pallets. The 12t vehicles were loaded on single pallets, while the 18t vehicles were loaded on a “double pallet”, which was constructed from two pallets with plywood between them to reduce friction, documents say. The pallets were secured to the 747’s deck using “dozens” of 5,000lb straps.

Boeing’s report says: “Data and physical evidence… indicate that the most likely scenario involved at least one MRAP (aft-most) breaking loose of its restraints shortly after take-off rotation, shifting aft and damaging the FDR/CVR [flight data recorder/cockpit voice recorder], before penetrating the aft pressure bulkhead.”

The company notes that the FDR, which is installed just forward of the aft pressure bulkhead, stopped working when the aircraft was just 33ft off the ground.

That finding is consistent with another NTSB document noting that crash investigators found streaks of the FDR’s orange paint on the rearmost vehicle, and an imprint of that vehicle’s rear spare tyre on the aft-pressure bulkhead.

The investigation also found aircraft parts, including those related to hydraulic systems, on the runway and ground along the aircraft’s flight path, suggesting the vehicle damaged the aircraft’s control systems.

“The MRAP’s aft movement compromised at least hydraulic systems number one and number two and may have contacted the stabiliser jackscrew actuator assembly, shearing the jackscrew actuator from its fuselage attach points,” says Boeing’s report. “If the stabiliser jackscrew actuator had been liberated from its attach points on the fuselage by the MRAP, the horizontal stabiliser control system would most likely have been compromised to the point that continued safe flight and landing would not have been possible.”

The report adds, however, that loss of FDR data makes it impossible to determine the cause with absolute certainty.

Following publication of findings on the August 2013 UPS Airbus A300-600 freighter crash in Alabama, the US Federal Aviation Administration has advised all operators to ensure they have the latest software updates for enhanced ground-proximity warning systems.

Although the freighter, which struck terrain while attempting to land at Birmingham, Alabama, had been fitted with an approved Honeywell warning system, it did not feature the most recent available software. The bulletin states that if the software had been updated, the aircraft would have entered the terrain alert envelope about 200ft above the ground, some 1.3nm from the runway threshold.

As it was, the crew of the A300 received a “sink rate” warning about 8s before an initial collision with trees, with a “too low, terrain” alert sounding just after that impact.

In the special bulletin dated 13 March 2015 the FAA states that the latest software would have provided a “too low, terrain” alert some 6.5s earlier, when the aircraft was 150ft higher”, explaining: “Although it is not clear if the later version of the software would have prevented the accident, it would have provided a significantly improved margin of safety.”

The bulletin, which has also been highlighted by the European Aviation Safety Agency, has not been elevated to a formal airworthiness directive. NTSB investigators pointed out that the A300’s high descent rate would nevertheless have “compromised” the effectiveness of the warning system, even with the updated software.

But their analysis of the crash determined that an immediate activation of the go-around switch, or an aggressive manual response to the terrain alert, would have enabled the aircraft to avoid the impact.

Investigators have been unable to determine why a Kazakh Bombardier CRJ200 operated by SCAT crashed while attempting a go-around at Almaty two years ago, with the loss of all 21 on board.

Analysis of the aircraft’s dynamics shows that, instead of climbing away from the abandoned approach to runway 23R, it dived from a height of about 150m (500ft), striking the ground and disintegrating.

“It has not been possible to clearly identify the reason for the aircraft’s transition to a dive – instead of a climb – from the available data,” says Russia’s Interstate Aviation Committee. The inquiry points out that the aircraft dived after its elevators moved into the nose-down demand position.

But it says there was no evidence of technical failure on the SCAT aircraft, nor any indication that the jet was affected by icing, windshear or wake turbulence.

The CRJ200 had been attempting to land in poor visibility following a service from Kokshetau on 29 January 2013. While flight data shows the aircraft was stable on its approach, the cockpit voice recorder indicates that the captain was extremely frustrated by the weather conditions at Almaty, which he had hoped would improve before the arrival.

The inquiry says that, although the aircraft continued on its approach, the cockpit recordings show that there was no intention to pursue a landing below minima, and the captain called for a go-around.

Fifteen seconds before impact, at a height of 180m, the autopilot was disengaged, when the aircraft was flying at 140kt (259km/h) and descending at around 600-800ft/min. Its flaps started retracting while the engines’ thrust increased. But just 4s after the autopilot disconnection the elevator began deflecting into the pitch-down position. An automated call-out told the crew that the CRJ200 was at 500ft and the airspeed increased to 150kt. The horizontal stabiliser’s position remained unchanged but the aircraft pitched increasingly nose-down. “Despite the increase in airspeed and [descent], the crew did not take action to put [the aircraft] into a climb,” says the inquiry.

Less than 5s after the aircraft entered the dive, its ground-proximity warning system cautioned on the sink rate and then began issuing urgent terrain alerts, instructing the pilots to pull up. But the aircraft’s pitch demand remained nose-down, the attitude eventually exceeding 20° nose-down, from which the CRJ failed to recover. It struck the ground about 1.6km before the threshold of 23R.

Bombardier modelled the last 30s of the ill-fated flight as part of the effort to explain the accident. The simulations revealed nothing unusual in the aircraft’s aerodynamic behaviour, and that it had responded as designed to the elevator deflection. There were “no other external factors” contributing to the downward pitch, says the inquiry.

It states that, in the absence of any system failure, the elevator deflection could have resulted from control inputs by the crew. The inquiry has struggled to resolve the mystery, saying that there is “no logical reason” for the initiation of the fatal dive when the aircraft should have been climbing for the missed approach. Investigators have considered somatogravic illusion – a strong but false perception of the aircraft’s pitch caused by linear acceleration – but rejected it because the flight-data recorder revealed slight changes in the CRJ’s elevator deflection during the dive that are not characteristic of the effects of the illusion.

Download the half-year safety report here