This is a look back at a serious incident from 2018, not a new event. On 14 May 2018 the right windshield of a Sichuan Airlines Airbus A319 burst and left the aircraft at 9,800 metres over south-west China. All 128 people on board survived. The final report of the Civil Aviation Administration of China (CAAC), adopted on 8 May 2020, is a detailed public account of how a double-ply windshield failed in service and what a flight deck at cruise altitude becomes without one1.

What happened on flight 3U8633?

A319-133 B-6419 left Chongqing for Lhasa at 06:27:18 Beijing time with 119 passengers and nine crew. The commander flew from the left seat, the first officer was pilot monitoring and a second captain had moved to seat 1F for the cruise1. At 07:07:05 the cockpit voice recorder caught a dull bang and the crew saw radial cracking across the right windshield. The first officer began looking up the procedure on his electronic flight bag, the commander took control at 07:07:11 and the crew asked to descend and divert to Chengdu1.

Forty seconds after the first crack, at 07:07:45, the windshield departed. Cabin altitude rose from 6,256 ft to 24,320 ft by 07:07:50 and peaked at 26,368 ft at 07:08:14. The autopilot disconnected at 07:07:46 and autothrust at 07:07:481. The first officer, wearing his lap belt but not his shoulder straps, was pulled out of his seat by the outflow with his upper body through the opening. He told investigators he pushed himself back in with both hands. The CAAC and EASA both record that he returned to his seat by his own effort12. A large forward and right input on the right sidestick at the moment of separation was most likely caused by his body striking it1.

Which systems did the crew lose?

Seventeen circuit breakers on panel 120VU tripped. The CAAC traced the resulting failures to those breakers: DC BUS 1 and 2, all windshield and window heating, autobrake and anti-skid, flight director 2, SEC 2 and 3, spoilers 1, 2 and 5 and both thrust reversers1. ELAC 2 failed in pitch and roll, taking pitch control to alternate law. The FCU was bent to the right and panel 130VU, including the right glareshield, was lost1. The cockpit door unlocks by design when it senses a rapid pressure drop. The report asks Airbus to assess protection against contact between the door and panel 120VU1.

How was the descent flown?

The commander hand-flew the entire descent. Bank reached 51.7 degrees left at 07:08:09. Descent stayed below 3,000 ft/min for the first 65 seconds, then increased to a maximum of 10,279 ft/min with a maximum of 349 knots indicated. Investigators linked that profile to the flight deck environment: the crew wanted to get down fast, but more speed meant more ram air through the opening. The commander later reported that at 280 to 290 knots his body could barely stand it. It became bearable at 250 to 260 knots. Speed brakes were not used, which the report attributes to the difficulty of judging which systems remained available1.

The aircraft held above 23,600 ft from 07:09:57 to 07:16:39 and stayed above MORA and radar minimum vectoring altitude throughout1. Noise made speech impossible, so the three pilots worked with gestures and shoulder taps. Radar showed squawk 7700 at 07:10:39. ATC called without reply from 07:08:17 to 07:17:08 and the crew transmitted MAYDAY blind on the frequency. The aircraft landed on runway 02R at Chengdu at 07:41:05, at flap 3 and 63 tonnes against a 62.5-tonne limit, with a ground speed of 156 knots. With reduced deceleration devices and no anti-skid, the fuse plugs on main wheels 3 and 4 melted and the tyres deflated1.

What were the physiological margins?

The commander never donned oxygen. He tried with his right hand, but the mask stowage was behind him on the left while his left hand held the sidestick. He spent 19 minutes 54 seconds above a cabin altitude of 10,000 ft, including 1 minute 21 seconds between 25,000 and 26,368 ft. The first officer's first breaths on oxygen appear on the recording at 07:10:571. The report compares those exposures with published time of useful consciousness and finds no evidence of incapacitation, a conclusion supported by the continuity of the commander's control inputs1.

Why was it survivable at all? At Mach 0.76 the total pressure at the opening was about 399 hPa, close to static pressure at 7,200 metres rather than the 272.6 hPa at 9,800 metres. The air conditioning packs also kept delivering warm air. Simulation put flight deck temperature between minus 24 and plus 8 degrees Celsius during the descent, with no more than 1.5 minutes at minus 24. None of the pilots suffered frostbite, though all three later showed symptoms the report considers possibly linked to decompression sickness and received more than 20 hyperbaric sessions1.

Why did the windshield fail?

The CAAC's most probable cause is that the weather seal or silicone seal of the right windshield was possibly damaged. With cavities inside the unit, moisture entered and stayed at the lower edge. Long immersion degraded the insulation of the power braid and a sustained arc in a humid environment developed at the lower left corner. Local overheating broke both structural glass plies and the windshield could no longer hold the pressure differential12.

The report found Sichuan Airlines' maintenance programme compliant with Airbus requirements. The windshield was the original unit with 19,942 flight hours and 12,937 cycles, nothing was deferred on the day. A C-check in February 2017 had found and repaired weather seal erosion on both sides. The last detailed inside and outside window inspections, in April 2018 and October 2017, were normal1.

The design and certification findings carry the wider lesson. The windshield heating computer could not detect an arc within normal current range. The A319's compliance showing under JAR 25 Amendment 11 did not need to consider the effect of heating system failure on windshield structure. With both structural plies broken, the unit could not hold cruise differential1. Airbus data listed six earlier double-ply failures in its fleet in which the windshield stayed in place, with arcing among the main causes. After the event the manufacturer SGS inspected 298 returned A320-family windshields and found moisture in the terminal block of 311.

How does it compare with British Airways 5390?

The obvious precedent is the BAC One-Eleven G-BJRT over Didcot on 10 June 1990. There the left windscreen, replaced before the flight, blew out while climbing through 17,300 ft because 84 of its 90 bolts were of smaller than specified diameter. The commander was sucked halfway out and was restrained by cabin crew while the co-pilot landed at Southampton. The AAIB identified a safety critical task done by one individual and not tested until a passenger flight3.

The outcomes look alike but the failure paths do not. BA 5390 was a recent installation error at lower altitude. 3U8633 was an original windshield degrading in service, passing its inspections and failing at cruise. One pilot remained half outside until landing. The other returned to his seat, set 7700 and left most of his duties to the second captain.

What changed after 2018?

In December 2018 Airbus updated the AFM, FCOM, QRH and FCTM for cracked windshields and windows: fasten shoulder harnesses in every case, use oxygen masks if a crack in the inner structural ply is confirmed and remove them when differential pressure reaches 5 psi. The CAAC also asked operators to require a pilot who has donned a mask to help a colleague who cannot. It further asked them to record cabin altitude at one sample per second1.

Regulation followed. CAAC recommendation SWCAAC-ASR-2018-1-6, logged as CHIN-2020-001, asked EASA to consider revising AMC 25.775(d). EASA cited the B-6419 event when it adopted CS-25 Amendment 27 through ED Decision 2021/015/R, which amended that AMC so that applicants properly address windshield system failure conditions that may have structural effects2. In the consultation the UK CAA pointed to water ingress in the Chinese report and asked EASA to consider its interaction with lightning and static build-up6. The commander, Liu Chuanjian, was named at the time and described the first officer as having "half his body suspended out of the window"4. His story became the 2019 film The Captain5.

Crew resource management under these conditions is part of what multi-crew cooperation training prepares for. Type rating courses are where abnormal procedures become drills. For a recent windshield case with a very different outcome, see United UA219's return to Chicago O'Hare.