Metal Fatigue Caused Airbus A380 Engine Breakup, BEA Says
The BEA’s final report on an Air France-operated Airbus A380 accident is a striking example of how rewarding it can be to conduct an investigation to the end, despite the high costs.
The findings by the French BEA accident investigation office may have consequences for engine design and manufacturing. The aircraft suffered an inflight breakup of one of its four Engine Alliance GP7200 turbofans when climbing over Greenland on Sept. 30, 2017. Investigators determined the root cause of the accident—characterized as such because it was deemed considerably more serious than most engine failures—was dwell fatigue of a metal that had been viewed as immune to such deterioration.
- Titanium alloy was thought to be immune to known metallurgic failure mode
- New systems developed to find metal part under thick layer of ice
Flight Air France 66 (AF066) took off from Paris to Los Angeles with 521 people on board, including 497 passengers. The fan rotor on engine No. 4 (outboard on the right wing) separated during climb from Flight Level 370. No occupants were harmed by the debris, and it caused little damage on the surrounding airframe structure. The crew diverted to Goose Bay, Newfoundland, without further damage.
The probe offers another example of the BEA’s persistence—an echo of the four search phases that eventually led to the discovery of the wreckage of an Air France A330 that crashed in the South Atlantic in 2009 (often referred to as the Rio de Janeiro-Paris flight, or Air France 447). This time, safety experts at the BEA had to conduct three research phases in the Arctic ice sheet to find the component holding the telltale sign: the fan hub.
Again, the outcome of the investigation proved worth the effort.
The results of the A380 inquiry are generating key lessons in metallurgy for large turbofans.
Before the analysis determined the root cause was in design and manufacturing, most stakeholders in the investigation—including specialists from the BEA and various OEMs—believed the origin of the failure was in maintenance. They theorized that a mechanic had inadvertently created a dent or pit in the fan hub’s front face, and that the defect had developed into a crack.
In fact, the problem stemmed from a surprising weakness that can occur during the manufacturing process.
To understand how close the industry came to missing an important engineering lesson if the inquiry had been dropped, consider the context. When the investigation began, the A380 was still in production. The fleet of aircraft powered by the Engine Alliance GP7200 (a GE Aviation-Pratt & Whitney joint venture) stood at 130 out of 239 A380s in service. The event was the first major incident involving a GP7200, and both Airbus and the engine-maker still hoped to sell many more of their products.
In early 2019, Airbus announced the termination of the A380, before discovering the missing piece of the puzzle. Twice the BEA sought additional funding for the next phase of the search in Greenland. Discouraging factors included the new status of the program and the aforementioned belief that a poorly conducted servicing operation caused the accident.
But the BEA persevered. With his Danish counterpart (who had delegated the investigation to the BEA), Executive Director Remi Jouty expected significant lessons if the fan hub was found. The investigation’s total cost was “close to” €5 million ($6 million), Jouty says.
Massive numbers surround the event. Virtually the entire fan module, weighing more than 1.2 metric tons, detached. The powerplant system, including the engine and the nacelle, lost 5.5 m (18 ft.)—more than half of its length. Of the 240-kg (530-lb.) fan hub, 220 kg were missing.
Large pieces of debris jettisoned vertically, and only by sheer luck, they were not ejected horizontally, which would probably have brought the aircraft down, Jouty says.
In the first remains the investigators could study, every piece of evidence was suggesting all damage was the consequence of something that happened at the hub level, he recalls. Analysis showed a crack developed from inside.
Metallurgy aims at producing isotropic materials, which have the same properties in every direction. This was a key design driver when a titanium alloy called Ti-6-4, including 6% aluminum and 4% vanadium, was invented.
Sometimes, however, anisotropic zones develop, creating local fragility. But these weakened zones had been demonstrated to remain small enough in Ti-6-4 to solve a premature failure problem seen in the 1970s with other alloys.
The A380 accident proved the demonstration had been flawed. The accident hub logged 3,500 cycles, well below its design life, Stephane Otin, an investigator at the BEA’s materials and failure analysis laboratory, explains. Yet it had performed 1,650 cycles after the crack formed, including 700 cycles during which the crack was invisible. The microscopic fissure initiated 1.5 mm (0.06 in.) beneath the surface.
An analysis determined that a so-called macro zone of anisotropy—where the crack originated—was created during the part’s forging process. It was 10 times greater and stronger than what statistics from analyzed production samples suggest.
The larger the part, the greater the probability to find a hazardous macro zone, Jouty adds.
Compounding the macro zone’s existence was that it was orienting the weakness in the same direction as the load the part undergoes. Moreover, loads were highest at the macro zone location.
Then, the fatigue process took place in constant-load phases (essentially when the engine runs at constant rotation speed), as opposed to transient loads, leading to dwell fatigue and the crack’s expansion. But due to the supposed properties of Ti-6-4, fatigue tests on production samples included virtually no constant-load phase. Such evaluations focused on transients.
The consequence for engine manufacturers, Jouty says, is that they will have to review the way they design rotating parts on large engines. Manufacturers may need to find an acceptable, perhaps lower load in the future, altering a part’s geometry and opting for a different material. These changes could, however, result in heavier parts. Manufacturing processes, especially forging, will have to be reassessed and an inspection regime created, Jouty points out.
The BEA thus recommends that the European Union Aviation Safety Agency (EASA) and the FAA ensure design and manufacturing processes keep the risk of failure under control. In a second recommendation, the BEA asked the EASA and FAA to “carry out a review of engine rotor-grade critical parts” made of titanium alloys of the same family as Ti-6-4 and ensure engines are inspected accordingly.
For the GP7200’s fan hub, frequent inspections were required after the failure process was understood. Since Nov. 21, 2019, eddy-current and ultrasonic inspections must be conducted every 330 cycles. Because fan blades must be removed during inspections, the blade lock ring was redesigned to reduce the risk of damaging the hub’s front face.
To understand the failure mode, the investigators needed to first find the accident fan hub, which eluded them for almost 20 months. Debris that had fallen from the aircraft had scattered, since their speed and aerodynamic characteristics were different. The hub’s location was thus distinct from the area where most other pieces hit the ground, and the region’s percolation phenomenon caused the part to sink.
For these reasons, search helicopter flights in early October 2017 found other pieces, such as fan blade fragments. But snowfall quickly covered the surface, leading investigators to find other means of recovering the missing debris.
The second phase began, but several methods proved ineffective.
Eventually, an airborne synthetic aperture radar, developed by French aerospace research center ONERA, was successful. The experimental system uses hyperfrequency-imaging remote detection (SETHI) and can “see” a large swath under the ice and snow surface with a 20-cm (8 in.) resolution. The radar system was carried by a modified Dassault Falcon 20 business jet.
The heterogeneous layers of ice and snow, however, disturbed the radar signal, Jouty explains. The probe’s organization and the weather slot gave ONERA limited time during the second phase. An algorithm developed on site proved helpful as well as the use of three frequencies and the combination of images, Jouty says. ONERA could suggest six target locations, but engineers expressed limited confidence in their results. The targets turned out to be false positives.
ONERA engineers, convinced that further SETHI data processing would provide solid results, persisted.
Meanwhile, Pratt & Whitney refined the failure analysis. Because the shape, weight and speed of the missing part were calculated more accurately, the ballistic computation could also be refined, leading to a more precise idea of where to explore.
The final link in the search chain was a towed transient electromagnetic system developed by Denmark’s Aarhus University. It was modified for BEA’s purposes and tested on a Swiss glacier. On May 23, 2019, an “unambiguous signal” was acquired, and the hub was subsequently discovered 3.3 m under the surface.