COVER STORY JAL 123: AUGUST 12, 1985 520 LOST IT’S 20 YEARS SINCE THE WORLD’S WORST SINGLE AIRLINER ACCIDENT. MACARTHUR JOB AND STEVE SWIFT REPORT. J apan Air Lines’ 747SR, registered JA8119, completed four uneventful inter-city trips on Monday, August 12, 1985, arriving back at Tokyo’s Haneda Airport at 5.17 pm. Its next service was Flight JAL123 to Osaka, 215nm (400 km) south-west of Tokyo. A senior training Captain was in command, supervising the upgrading of a former 747 first offi cer. With 509 passengers and 15 crew aboard, JAL 123 took off at 6.12 pm. Th e planned route was via the island of Oshima, 50 nm southwest of Tokyo, cruising at FL240 (24,000 ft). At 6.25, the controller saw the emergency code 7700 suddenly appear beside the 747’s radar target. Seconds later the aircraft called, requesting an immediate return to Haneda. Controller: “Roger – approved as requested.” JAL123: “Radar vector to Oshima, please.” Controller: “Turn right, heading 090.” But instead of making the expected turn back towards Oshima, the aircraft gradually turned to a north-west heading. Controller: “Negative, negative...confirm you are declare [sic] emergency?” JAL123: “Th at’s affi rmative!” Controller: “Request your nature of emergency?” Th ere was no immediate reply. Controller: “JAL123 – fly heading 090 radar vector to Oshima.” JAL123 (tensely): “But now uncontrol! [sic]” FLIGHT SAFETY AUSTRALIA JULY-AUGUST 2005 28 News that the flight was in trouble leaked to the media. Japanese television conducted a live-to-air telephone interview with an eyewitness watching the 747. He described it as “wavering and having trouble keeping to its flight path”. Meanwhile the aircraft had turned north towards the mountain ranges forming a spine along the main Japanese island of Honshu. Yokota, 77 nm distant. It was maintaining around 22,000 ft above scattered thunderstorms and rain showers. When 45 nm west of Haneda Airport, the aircraft entered a descending turn to the right, completing a full circle before straightening out on an easterly heading towards the airport. Its descent then continued, but at 13,500 ft one of the crew called in an agitated voice: “JAL123, JAL123 – uncontrollable!” Then, just on sunset at 6.56 Tokyo control: “Roger – understood. Do you pm, as its altitude fell to wish to contact Haneda (approach)?” JAL123 (frantically): “Ah ... stay with us!” 8400 ft, the controller was JAL123 (now down to 9000 ft): “JAL123 – horrified to see the target request radar vector to Haneda!” vanish from his screen. Controller: “Roger – I understand. It is runway 22, maintain heading 090.” At 6.34 pm the company called the 747. Still descending, the 747 now gradually Company operator: “JAL123 – this is Japan turned to the left on to a heading of about 340 Air. Tokyo control received an emergency call degrees. It was below the level of mountains 30 miles west of Oshima Island.” that now lay in its path. Th e flight engineer responded, obviously Controller: “Can you control now?” under pressure: “Ah ... the R5 door is broken. JAL123 (desperately): “JAL123 – uncontrolAh ... we are descending now...” lable. JAL123 – ah uncontrol. JAL123, unconCompany operator: “Roger – does the Cap- trol [sic].” tain intend to return to Tokyo?” Approach: “Your position ... ah ... 45 miles JAL123: “Ah ... just a moment ... we are making north-west of Haneda.” an emergency descent ... we’ll contact you JAL123 (anxiously, with altitude read-out now again. Ah ... keep monitoring.” 13,000 ft): “North-west of Haneda – ah – how Out of control: Th e 747 continued north about many miles?” 25 nm, then began a gradual turn north-east Approach: “According to our radar, 55 miles in the direction of the US Air Force base at north-west. I will talk in Japanese – we are ready for your approach anytime. Also Yokota landing Sprawling (above): The wreakage of Japan Airlines Flight 123 on the slopes of Mount is available – let us know your intentions.” Th ere was no reply. Th e 747’s height was Osutaka. Clearly visible is part of a wing. Aftermath (inset right): Rescue workers now decreasing again and, by 6.54 pm, its with debris from the accident. altitude read-out was 11,000 ft. Approach called the aircraft again, advising its position was “50 miles – correction 60 miles” to recover its balance to the right. It was north-west of Haneda Airport. But again flying just like a staggering drunk.” Because of the inaccessibility of the area, it there was no response. A minute later, its target suddenly devi- was not until 9 am, more than 14 hours after ated 90° to the right and, as its altitude read- the crash, that civil defence workers reached out rapidly decreased, it entered a tight turn the site. Fog had forced a temporary suspenof less than 2 nm radius. Th en, just on sunset sion of mountain flying, but when conditions at 6.56 pm, the controller was horrified to improved, army paratroopers arrived aboard Chinook helicopters, rappelling down to see the target vanish from his screen. Further calls to the 747 went unanswered. where the wreckage lay. Th e disaster was now revealed. Flying a Moments later, a military jet reported “a huge westerly heading, the Boeing 747 had deburst of flame in the Nagano Mountains”. Impact in the mountains: It was dark by scended into a pine forest near the top of the the time two search helicopters reached the northern face of the 5400 ft Mt Osutaka, a area through showery weather. Attracted by narrow, steep-sided east-west ridge, exploda fire blazing near the top of the 5400 ft Mt ing into flames and breaking up as it bounced Osutaka in inaccessible ranges more than along the ridge line. Th ere was no sign of survivors. In Tokyo, 60 nm north-west of Tokyo, one helicopter pinpointed the site of the crash, reporting the fearful news was confirmed to waiting media – the highest death toll ever in a singleflames “over an area about 300 m square”. One witness, surprised at seeing an air- aircraft accident. Well down the mountain face, a fireman liner above his remote mountain village, described its erratic flight. “All of a sudden, stood on the steep slope surveying the wrecka big aeroplane appeared from between age. Suddenly he saw something that looked mountains,” he told police. “Four times like an arm waving! Sure enough, a young it leaned to the left, and each time it tried woman, conscious though suffering a broken Photo: AAP Photo:AAP Photo:AAP photo: newsweek COVER STORY pelvis and a fractured arm, was pinned between two sets of seats. Not long afterwards there was more good news – a 12-year-old schoolgirl was found wedged in a tree, suffering nothing more serious than cuts and bruises. Even more was to come – rescuers discovered another young woman and her daughter beneath wreckage. Both suffered fractures. All four survivors had been seated among the last seven rows of seats. Medical staff found some victims had clearly survived the impact but, wearing only light summer clothes, had died of exposure during the night. Investigation: Th e aircraft had been worked hard, flying 25,000 hours in the course of 18,800 cycles. Did this demanding utilisation show up some unknown flaw? Th e only clue to the loss of control was the tense radio transmission that the 5R cabin door – the rear most door on the starboard side – was “broken”. Could the door have broken away and struck the tail, disrupting the multiple hydraulic systems that actuate the aircraft’s control surfaces? FLIGHT SAFETY AUSTRALIA JULY-AUGUST 2005 29 COVER STORY There was no sign of survivors. In Tokyo, the fearful news was confirmed to waiting media – the highest death toll ever in a single-aircraft accident. 5 Th e finding of the door amongst the wreck age with its latches in the closed position only deepened the mystery. Why had the flight crew referred to it as “broken”? Could structural distortion of the fuselage have caused the door warning lamp to light up?. A photograph of the stricken aircraft, snapped from a mountain village shortly before the 747 crashed, provided new and dramatic evidence (see cover photo). A portion of its vertical fin, together with the section of the tailcone containing the auxiliary power unit (APU), was missing. Th e photographic evidence was confirmed when a 5 m piece of the aircraft’s fin was found floating in the bay where the aircraft had been passing when the emergency developed. Could the APU’s gas turbine have disintegrated, rupturing the hydraulic lines to the rudder and elevators? While Boeing and US investigators were on their way to join the Japanese team, other important evidence was emerging. One of the surviving passengers described what took place in the rear passenger cabin. She was an off-duty JAL flight attendant, sitting only four rows from the rear of the cabin. “Th ere was a sudden loud noise, somewhere to the rear and overhead,” she said. “It hurt my ears and the cabin filled with white mist. Th e vent hole at the cabin crew seat also opened.” Th e white mist was characteristic of sudden cabin decompressions. Th e “vent hole” was one of the modifications made to wide-bodied aircraft as a result of the Turkish Airlines DC-10 disaster near Paris 11 years before in 1974. (See Flight Safety Australia, March-April 2005). “Th ere was no sound of any explosion,” the witness continued, “But ceiling panels fell off, and oxygen masks dropped down.” Th en she felt the aircraft going into a “hira-hira” (Japanese for a falling leaf). Investigators soon discovered the flight data recorder (FDR) and cockpit voice recorder (CVR). A read-out of the FDR, and a transcription of the CVR tape, confirmed the flight attendant’s report. Th e explosive decompression occurred a few seconds past 6.24 pm, soon after the 747 reached its cruising level. After the aircraft was cleared to return to Haneda, the Captain exclaimed: “Hydraulic pressure has dropped!” Th e failure of the 747’s multiple hydrau lic control systems completely deprived the crew of primary control. Stabiliser and aileron trim were also rendered useless, and the yaw COVER STORY damper was no longer effective. With the aircraft’s stability also seriously impaired by the loss of a substantial part of its fin, it began combined “phugoid” and “Dutch roll” oscillations, settling into a pitching, yawing, and rolling motion. Th e pitching, in cycles of about 90 seconds, was taking the aircraft from about 15° noseup to 5° nose-down, with vertical accelerations varying between +1.4 g and -0.4 g. Variations in airspeed and altitude during the cycles were averaging around 70 kt and 3000 ft, with peaks of as much as 100 kt and 5000 ft. Th e yawing and rolling motion was much faster, the aircraft alternately rolling 50° either way in cycles of about 12 seconds. Delicate handling: Holding the aircraft’s attitude by increasing and decreasing power, the crew also achieved limited directional control by applying power asymmetrically. At 6.29 pm they achieved a bank to the right, turning the aircraft on to a northerly heading while maintaining an altitude between 23,000 and 25,000 ft. Desperate efforts: Th e altitude excursions reached a peak, with the nose pitching down and the aircraft diving from 25,000 to 20,000 ft in a little over half a minute as the airspeed rose from 200 kt to 300 kt. Just as quickly the motion then reversed, the speed falling off again to 200 kt as the nose rose and the aircraft began climbing again. Preoccupied with trying to maintain control, the flight crew had overlooked donning their oxygen masks. Nearly 10 minutes had passed since the decompression and they were undoubtedly suffering a degree of hypoxia and some deterioration in judgement. But at the flight engineer’s prompting, this was remedied. Th e oxygen took effect quickly, for the pilots now limited the pitching excursions to about 2000 ft in altitude and 60 kt in airspeed. But they could do nothing to dampen the continuous rolling from side to side. Th e CVR revealed the pilots’ increasingly desperate efforts to control the aircraft. Over and over again, the Captain instructed the co-pilot to “lower the nose”. Just before 6.39 pm, the flight engineer suggested lowering the undercarriage to help stabilise the motion, but both Captain and co-pilot countered: “We cannot decrease the speed!” A minute later, with the pitch oscillations reduced to about half, the pilots succeeded in turning the aircraft towards Haneda Airport, 42 nm distant. As they did so, the flight engi- neer, seizing the opportunity as the airspeed fell below 200 kt at the top of a pitch-up, selected the undercarriage down. Although the change of longitudinal trim required an immediate increase in engine power, the increased drag dampened the pitching, reducing the amplitude of the airspeed and altitude excursions as the aircraft entered a descent of about 3000 fpm. But the drag also dampened its response to directional control and, instead of continuing towards the airport, it entered a turn to the right, still descending. But after turning through 360 degrees, the pilots regained some measure of directional control at 15,000 ft. Th eir reprieve was shortlived – the aircraft began turning again, this time to the left. Now below 9000 ft and still descending, the 747 was heading north again towards mountainous country. “Hey – there’s a mountain – up more!” the Captain called anxiously. Th e co-pilot carefully applied more power, trying to juggle the aircraft’s attitude. But with the undercarriage down, this failed to check the descent. Captain: “Turn right! Up! We’ll crash into a mountain!” With the application of more power, the aircraft pitched nose-up, gaining 2000 ft, while the airspeed fell from 210 to 120 kt. Captain (urgently): “Maximum power!” But the coarse application of power trig- gered the phugoid oscillation again. Captain: “Nose down ... nose down!” Th e co-pilot reduced power again and the nose pitched down. Th e aircraft was plung ing to below 5000 ft with the airspeed rising quickly to around 280 kt, before recovering from the dive at a loading of 1.85 g. Th e air craft then climbed even more steeply to about 8000 ft and, as its airspeed fell sharply, the stall warning began sounding. Captain (dismayed): “Oh no!” (urgently): “Stall! Maximum power!” Calls from Tokyo, Approach and Yokota were ignored as the crew fought to prevent the 747 plunging out of control. Th e final 108 seconds of the CVR revealed a string of increasingly desperate instructions calling for “Nose up”, “Nose down, and “Flap” as the pilots tried to prevent the aircraft falling out of control. Finally, as it entered a tightening descending turn to the right, the ground proximity warning system began sounding. Fourteen seconds later there was the sound of the aircraft striking tree-tops, followed three seconds later by the sound of the crash. Wreckage examination: It was clear that the explosive decompression, the pre-impact damage to the fin and rudder, and the loss of all four hydraulic systems, were somehow linked. But what was the link – and what had precipitated it? As a precautionary measure, the Japanese Fatal flight: The explosive decompression occurred soon after the 747 reached it’s cruising level (point of rupture). The crew lost primary control as a result of loss of hydraulic control systems. FLIGHT SAFETY AUSTRALIA JULY-AUGUST 2005 31 COVER STORY Photo: Kjell Nilsson DESTRUCTION OF THE REAR PRESSURE BULKHEAD Tear stop straps LOWER Line of cracks Bulkhead 4.55 m diameter UPPER Tail cone blown off: A Japan Air Lines 747 SR. The red line shows the rear tail cone part of which was destroyed when the rear pressure bulkhead failed. UPPER BULKHEAD Fracture LOWER BULKHEAD Upper doubler plate splice Tear stop straps Tremendous force: The rear pressure bulkhead (shown left in blue) must contain tremendous force from the pressure difference at altitude between the cabin and outside air. The 4.55 m bulkhead is constructed of reinforced aluminium alloy sheets in a domed shape to resist pressure.The line of cracks (right) formed mid way between the upper and lower parts of the dome. Lower doubler plate splice Line of cracks: A side view of the cracks formed prior to rupture of the bulkhead. The bulkhead was destroyed after the cracks ran past the tear stop straps. INCORRECT REPAIR CORRECT REPAIR UPPER NORMAL BULKHEAD Fillet seal JA8119 REPAIR Upper doubler plate Stiffener Filler sealant Gap filled with filler sealant Doubler plate Lower doubler plate Up LOWER Rivets Forward Over-loaded rivets: A “section” through the bulkhead showing normal construction (left) a correct repair (centre) and an incorrect repair (right) that led to the JAL 123 tragedy. In the wrong repair, technicians tried to connect two doubler plates, which forced the middle row of rivets to carry too much load. FLIGHT SAFETY AUSTRALIA JULY-AUGUST 2005 32 Civil Aviation Board ordered inspections of all 69 of Japan’s 747s. Boeing, in a telex to 747 operators world-wide, suggested they inspect the aft portion of the pressure hull. Airworthiness authorities world-wide issued airworthiness directives. Although Boeing 747s had no history of bulkhead failure, a major bulkhead fracture seemed to fit the evidence. Pressurised air, escaping from such a fracture, could have burst the fin. Th e rear pressure bulkhead was certainly badly damaged. But had the damage all been sustained in the impact? Although Boeing investigators argued that the design had been tested to a simulated service life of 20 years, and that the wreckage exhibited no evidence of corrosion, a profound shock lay in store. Examining the wreckage, Boeing’s structures engineer picked up a broken off section of pressure bulkhead plating. It had been repaired, and the repair didn’t look right. Electron microscope photographs of the fracture surfaces of the doubler plate revealed striations indicative of metal fatigue. Th e discovery posed two vital questions: Why was the bulkhead repaired in the first place? And why had it been wrongly repaired? An examination of JAL maintenance records revealed the rear fuselage of the crashed 747SR had scraped the ground during a nose-high landing seven years before. Th e impact had been severe enough to remove skin panels and crack the rear pressure bulkhead. Th e aircraft had been grounded for a month while Boeing engineers supervised repairs at JAL’s maintenance facility. Th e repair included replace ment of the lower part of the rear fuselage and a portion of the lower half of the damaged bulkhead. Close examination of the bulkhead repair showed that two separate doubler plates, instead of one continuous one, were used as reinforcement. Th e result was excessive load on one row of rivets. JAL’s maintenance planning manager said the repairs were examined by the Japanese Civil Aviation Bureau, and the aircraft test flown after the work was done. No shortcomings were detected. Moreover, in the seven years the aircraft had flown since, six 3000 hourly “C-checks” had been carried out. Yet these had found nothing. Lessons: Boeing changed its 747 design to make it more forgiving to failures like this in the future – in other words to improve its “fail safety”. Th e manu facturer strengthened tear-stop straps in the bulkhead to stop cracks running. It improved venting of the tail compartment behind the bulkhead to reduce pressure if a bulkhead failed. And it provided a cover for an internal access hole to prevent pressurised air from entering the vertical fin. COVER STORY to the approved data, check with the designer. In JAL 123’s case, the Boeing repair team did not communicate well enough with their own company’s designers. Internal communication can be a problem for large companies. Operators should ask for assurance that the advice they are getting (including the NTO, or no technical objection) has engineering support, especially the support of company regulatory delegates. It is a good idea to check your old repairs. Be suspicious of all structural repairs. Be especially concerned about patches that are unusually old, large or thick. And be alert for small cracks emerging from under the edge of a patch repair. Look for signs of loose or working rivets, and be wary of stains streaking from under a patch. Th ey might be the signature of pressure or fluid leaks. And take any available opportunity to check internally for cracks hidden under external repairs. An improperly treated scratch on the aircraft pressure vessel skin, especially if covered under a repair doubler, could be hidden damage that might develop into fatigue cracking, eventually causing structural failure (see “Hidden hazard”, Flight Safety Australia, September-October 2003). DIAMOND STANDARD MAINTENANCE Diamond standard maintenance is based on a systematic analysis of how fatigue is likely to affect all the safety-critical parts of the airframe. Th e analysis has five elements: Site: Where could cracks start? Th is is a pre dictive element that requires analysis, testing and service experience, if available. Scenario: How will cracks grow? For example, will there be one or many? Will they interact? Will cracks in one part start cracks in another? Detectable: What is the smallest detectable size, considering the nature of the inspection method and other factors? Once you know this you can more effectively design your crack detection regime. Beware of optimism: For every lucky find of a small crack, there may be many more misses of large ones. Dangerous: As a crack continues to grow, sooner or later it starts to become “dangerous”, because the structure is about to lose the strength we want to assure. Duration: Th is is the time it will take a crack to grow from “detectable” to “dangerous”. It is the “safety window”. Th e inspection interval must be narrower, and must account for uncertainty and variability. Airworthiness authorities worldwide are By Steve Swift A t an international conference on aeronautical fatigue held in Hamburg, Germany, in June this year the concept of diamond standard maintenance received a lot of attention. Site Scenario Dangerous Detectable Duration The “diamond”: a new way of describing the “damage tolerance” rules for managing structural fatigue. Anyone operating a large airliner should check their compliance with Airworthiness Directive AD/GENERAL/82 Amdt 1. Th e scope of this airworthiness directive is likely to expand in the future. Finally, JAL123 warns owners and maintainers that fatigue cracks can stay hidden, even from the most thorough general maintenance. You need to adopt a systematic approach (see “Diamond standard maintenance” below). Th at’s why airworthiness authorities around the world are progressively requiring aircraft manufacturers to upgrade the maintenance programs they publish for the types they support. And that’s why Australia’s safety regulator insists that aircraft owners follow them, unless they have done a similarly systematic engineering analysis to justify the safety equivalence of their alternative. Fatigue is indiscriminate and inevitable. If not carefully managed, it can result in catastrophe for any aircraft, large or small, repaired or not. Macarthur Job is an aviation writer and aviation safety specialist. Steve Swift is a CASA structural engineer. Crack size Modifications were also developed to prevent a total loss of hydraulic fluid from the four independent hydraulic control systems if the lines were severed for any reason, and to provide additional protection for control cables. Boeing had thought about most of these things when designing the 747, but events proved they had not tested them suffi ciently. Aircraft manufacturers have since learned the value of testing to prove design assumptions. Th e JAL 123 tragedy reminds us how un forgiving structural fatigue continues to be in aviation, long after the infamous Comet crashes of the 1950s. Accidents resulting from structural fatigue have killed thousands, including many in Australia: In 1945, a Stinson A2W lost a wing, killing 10; and in 1968, a Vickers Viscount lost a wing, killing 26. In 1990, an Australian-built Nomad lost a tailplane, killing the pilot. Anyone repairing an aircraft needs to carefully follow approved data. To those repairing the bulkhead of the Japan Airlines’ 747, the improvised doubler plate repair probably looked strong enough – and it was, for a while. But, the fatigue aspects of a design are not always obvious. If you can’t install the repair exactly Dangerous Duration Detectable Time The time between inspections must be shorter than the “Duration”. So, for safety, we must know “Detectable”, “Dangerous” and how fast a crack will grow. working with aircraft manufacturers and operators to put in place diamond-standard maintenance programs for all aircraft, including their repairs and modifications. Diamond standard maintenance programs are usually called airworthiness limitations or supplementary inspection documents (SIDs). With two out of three Australian aircraft having seen a quarter century of hard service, the dangers of fatigue are ever present. Th ere is no room for complacency. For a copy of the full paper on diamond standard maintenance (called “Rough Diamond”), and other safety-related papers and reports on structural fatigue, visit CASA’s web site. FLIGHT SAFETY AUSTRALIA JULY-AUGUST 2005 33