How cracks in jet engines elude exams, with deadly results

As regulators around the world order urgent inspections of certain Boeing Co. 777 engines after one failed dramatically over suburban Denver last Saturday, they must also wrestle with a decades-old quandary:

The complex technology necessary to find cracks buried within metal parts can produce notoriously inconsistent results.

Not only are the tests prone to vague readings, but the results are only as good as the training and attention span of the human examiner. Indeed, after a 2018 failure of an engine fan blade identical to the one involved in the Denver episode, investigators found inspections had twice missed a crack that began years earlier.

“It probably was there and they just didn’t find it,” James Wildey, a retired National Transportation Safety Board metallurgical expert, said of the internal imperfections that led to the most recent failure. “The indications are the crack should have or could have been detected.”

The Federal Aviation Administration on Tuesday night issued an emergency order calling on United Airlines, the only U.S. operator flying 777s with the Pratt & Whitney engine that failed, to inspect each fan blade before it can carry passengers on the widebody workhorse. Other nations where the plane is based, Japan and South Korea, have issued similar directives.

Preliminary evidence indicates that a fatigue crack at the base of one of the blades led to the violent failure, NTSB Chairman Robert Sumwalt said. The front section of the engine, made up mostly of lighter weight materials that smooth the flow of air, broke loose, hitting parts of the plane and showering debris over a populated area near Denver International Airport as the plane began a flight to Hawaii.

The blade had flown about 3,000 times since its last inspection or from when it was installed, according to two people familiar with the preliminary results of the probe. They asked not to be named while discussing the sensitive data.

That was well below the threshold required for an inspection, which was between 6,500-7,000 cycles, including flights or any time the engine is started. That could suggest the existing inspection period was inadequate, but it will take months for investigators to determine that.

A Japan Airlines 777 with a Pratt & Whitney engine also had a fan blade failure on Dec. 4 near Okinawa, Japan. Pratt is a division of Raytheon Technologies Corp.

The FAA ordered United to use high-technology tests employing sound waves to detect even small flaws within the hollow, 40.5-inch (103-centimeter) titanium blades. An image is created that shows potential cracks and they’re reviewed by an inspector.

“Any tests like this that require a human interpretation of the image are tricky,” said John Hansman, a professor of aeronautics and astronautics at the Massachusetts Institute of Technology.

The image might be fuzzy or vague, Hansman and Wildey said. The person reviewing it could be tired or under stress. And examining such images for days on end, particularly if flaws appear only rarely, can lead to complacency or inattention, they added.

A general rule of thumb in the aviation industry is that there should be at least two tests conducted on a part before a crack can be expected to grow large enough to fail, Wildey said. That way, at least one of them would be expected to catch the problem.

In its emergency order on Tuesday, the FAA said it was an interim step and it “anticipates” taking additional action. Among them will be a review of how the tests are conducted, the agency said in a statement Wednesday.

“The FAA always looks at human factors as part of our standard Continuous Operational Safety process,” it said in the statement.

Testing Pitfalls

The 2018 failure of a fan blade on the same type of engine on another United flight bound for Hawaii illustrates the pitfalls of these testing techniques. The NTSB found in that case that the crack leading to the failure had been detectable for years, but it was twice missed during inspections.

In March 2010, almost eight years before the failure, an inspection found indications of a crack in the spot where it originated, though it was small, the NTSB said in a report on the case. The inspector indicated it was likely paint and the blade was reinstalled.

By July 2015, the crack had grown to nearly 1/10th of an inch and was visible on Pratt & Whitney testing equipment, according to the NTSB. An inspector saw the hidden flaw, but attributed it to paint used in the inspection process.

The blade was returned to service, only to fail two and a half years later.

In a report released on June 30, the NTSB concluded the engine failure was caused by poorly developed inspection guidelines at Pratt & Whitney and a lack of training for inspectors. Engineers at the company also didn’t provide feedback to inspectors to help them differentiate between genuine flaws in the metal and false alarms, the NTSB said.

The inspector who had conducted the 2015 examination told investigators he sometimes “had felt pressure to get the blades out quicker” because “managers want to get their numbers up,” according to the report. But he said he’d never been forced to accept a blade with a flaw in it.

Pratt & Whitney said at the time it had taken multiple corrective actions to address the cause of the failure.

Metal Disk

Problematic inspections have long played a role in NTSB accident investigations.

A large metal disk within an engine on a McDonnell Douglas DC-10 broke on a United flight in 1989, forcing pilots to make a crash landing in Sioux City, Iowa. Of the 296 people aboard, 112 died.

The NTSB concluded that the failure to spot a crack during an earlier inspection had led to the failure and cited “inadequate consideration given to human factors limitations.”

Wildey, who became head of NTSB’s Materials Laboratory, was an NTSB investigator on that case. The inspection used for the disk involved painting it with a dye to make cracks more visible.

When Wildey examined the fractured remains, the dye was clearly visible on the crack that led to the failure, he said. It should have been easily identified.

“There are all kinds of examples,” he said.