The Repository Root Cause

The DC-10 Crack That No Rule Required Anyone to Report

September 11, 2026 · Runtime 13:02 · Watch on YouTube

The story everyone tells about Flight 191 is an airline that took a shortcut with a forklift. That is in the report. What is also in the report is that the shortcut was adopted on safety grounds, that it worked sixty-seven times out of seventy-six, and that the exact same flange had already been cracked twice before — loudly, in a hangar, by mechanics who did everything the rules asked of them.

Chapters

Transcript

Chicago, 15:04 0:00

Chicago O'Hare, May 25th, 1979. Three in the afternoon, clear sky, fifteen miles of visibility. Flight 191 is rolling down runway 32 right. At rotation, the left engine tears off the wing and lands on the runway behind it. The airplane keeps flying. It climbs to three hundred feet, wings level, and then it rolls left and does not stop. Thirty-one seconds after liftoff it is down. Two hundred seventy-three people are dead. The crack that did this was eight weeks old. It was made in a hangar, by a forklift, on an airplane standing still.

The story everyone tells, and a correction 0:44

The story everyone tells about Flight 191 is a maintenance shortcut. American was supposed to take the engine off the pylon first. Instead it lifted the whole assembly on a forklift, and broke the airplane. Every word of that is in the report. It is also the part of this accident that explains the least. And one correction to this channel before we go on. Last episode I called that shortcut a procedure the manufacturer had recommended. Reading the report, that is wrong. Douglas assumed the engine came off first, and told American it would not encourage doing otherwise. It simply had no authority to forbid it.

Seventy-nine disconnects, down to twenty-seven 1:28

The job itself was routine: two service bulletins, replacing the spherical bearings that carry the pylon on the wing. Neither was mandatory. American chose to do them during a C check at Tulsa, and wrote its own procedure for the work. And here is the first thing the common version leaves out. The shortcut was not laziness. The report gives two reasons for it. Lowering the engine and pylon as one unit saved about two hundred man-hours per airplane. The second reason matters more. Taking them down together cut the things that had to be disconnected from seventy-nine to twenty-seven. Hydraulic lines, fuel lines, electrical cables. American argued the change on safety grounds, and the argument was a real one.

Nine hundred pounds, or twenty thousand 2:19

What it traded away was not obvious, and it is a number. The pylon hangs from the wing on two bulkheads about six feet apart, with the weight of everything hanging ahead of both of them. With the engine removed, the most that arrangement can put on the aft bulkhead is about nine hundred pounds. With the engine still attached, it can reach over twenty thousand. The flange on that aft bulkhead fractures under eight thousand. That is the trade. Fifty-two fewer disconnects, bought with a ceiling that moved past the breaking point of the part.

The aft bolt, pulled first 2:55

The sequence is where it turns fatal. The written order was forward upper fitting, forward lower fitting, thrust link, and last the aft bolt. Pull the aft bolt first, and the forward bulkhead becomes a pivot. The back of the pylon now hangs on nothing but the forklift. The mechanics did exactly that, because the written order was hard. Getting the forward fittings out first demanded unusual precision. And the procedure never said the steps had to be in sequence. It read as a checklist with signoff boxes, not an order of work.

Half an inch of clearance 3:33

And the clearance there is almost nothing. The bottom of the wing clevis sits half an inch above the surface of the flange. Take the bushing out and let the bulkhead rise, and it rises six tenths of an inch. The clevis does not clear the flange. It lands on it. Four tenths of an inch of movement at the center of gravity makes a seven inch fracture. That was measured afterwards.

Alignment was a visual estimate 4:00

The machine holding all of this was a forklift carrying eighteen thousand pounds, with its lifting forks five feet apart. The load had to sit balanced exactly under the center of gravity of the engine and pylon. There was no mark on the forklift showing the midpoint between the forks. Alignment was a visual estimate. The operator took directions by voice or hand signal, and the bulkhead he was controlling was fifteen feet away from him.

Tulsa, the shift change 4:29

Here is what happened to this airplane, in Tulsa, on the twenty-ninth and thirtieth of March, 1979. The midnight shift removed the aft bolt and bushing, and went off duty. The engine and pylon spent the shift change hanging on the forklift, pivoting on the forward bulkhead. When the day shift arrived, two mechanics saw the upper lug of the aft bulkhead resting against the bolts that hold the clevis. To get there it had to move six tenths of an inch, and it could only move that far if the flange was already bent.

Forks creeping down under load 5:03

Testimony says the forklift ran out of fuel during the operation, and sat unpowered for a while. Its maintenance log, three weeks later: trouble shooting, forks creeping down under load. No record of any corrective action. They finally tested it in June, after the crash. The forks drifted an inch in thirty minutes, and a defective check valve was replaced.

Nobody looked at it afterwards 5:30

After the pylon went back on the wing, nobody looked at it. The inspector's signoff blocks carried no requirement to inspect the attach fittings after reinstallation. Once he cleared the pylon, his remaining checks were connection integrity and a look for fuel and hydraulic leaks. The airplane went back into service with a ten inch crack through that flange, and flew three hundred forty-one hours. Fatigue grew from both ends of it, until the damaged length was about thirteen inches. Then, at rotation, the rest of it let go.

What left with the pylon 6:05

Losing an engine on takeoff is a trained maneuver. Losing the pylon with it is not the same event. Three feet of the wing's leading edge went too, and inside those three feet ran the lines to the left outboard slats. Four hydraulic lines and two cables. Two lines fed by the number one system, two by the number three, for redundancy. All six ran through the same three feet of structure. The redundancy was in the plumbing and not in the routing. And the only thing holding those slats extended is fluid trapped in a cylinder. There is no mechanical lock in the design. With the lines cut, air loads pushed the left outboard slats closed, and the left wing quietly became a different wing.

One stickshaker motor 6:54

The same separation took the number one generator bus and the number one direct current bus, and they were never restored. On those buses were the captain's flight director, the left stall warning computer, the stickshaker motor, and the slat disagree light. The Safety Board found that the airplane had exactly one stickshaker motor, and that the two stall warning computers did not talk to each other. So the one wing that had changed was the one wing whose warning had just lost power.

Flying the book into the stall 7:25

Now put the crew inside it. They are climbing on two engines, doing what the manual says: hold V two, one hundred fifty-three knots. With its outboard slats retracted, the left wing now stalls at one hundred fifty-nine. The first officer flew down toward the speed the procedure demanded, and at one hundred fifty-nine the wing stalled. He could not see the wing from the cockpit. The slat disagree light was dark. The stickshaker never fired. The Board's conclusion is blunt. It was not reasonable to expect this crew to recognize it, or to recover from it.

It had already happened twice 8:05

And here is the part the common version leaves out entirely. This had already happened. Twice. December nineteenth, 1978. Continental Airlines, a DC-10 in the hangar, same bulkhead, same forklift, same idea. The pin came out, the aft end of the pylon moved up, and the mechanics heard a loud pop, described as a pistol shot. They found the fracture, repaired the flange, and returned the airplane to service. February twenty-second, 1979. Continental again. The forklift operator moved the wrong lever. Same movement, same sound, same crack. Continental investigated both, classified both as maintenance error, and reported neither to the FAA.

"Not a service related problem" 9:00

That sounds like concealment. It was not. Federal rules do require carriers to report cracks and structural damage. The regulation lists sixteen reportable conditions. But the system it feeds is a mechanical reliability system, and mechanical reliability means service. An FAA maintenance specialist was asked about those two Continental cracks, and his answer is printed in the report. Under that concept, quote, we would not consider it because it was not a service related problem. The system collects what fails after the engines are started with intent for flight. Damage done in a hangar is not that. There was no box to put it in.

"Pylons had failed" 9:45

The same shape shows up at the other end of this story, years earlier, in the drawing office. McDonnell Douglas never analyzed what the loss of an engine and pylon would do to the rest of the airplane's systems. Not from carelessness. It had classified that event with the loss of a wing: unacceptable, and therefore not worth analyzing. For a wing, the Safety Board says, that logic holds. Without a wing, further flight is impossible and the rest is academic. For a pylon it does not hold, because the airplane keeps flying. This one flew for thirty-one seconds. And the report answers the objection in three words. Pylons had failed.

Twelve on a hoist, seventy-six on a forklift 10:32

After the crash, investigators counted what the industry had actually been doing. American carriers had removed and reinstalled one hundred seventy-five pylon assemblies. Eighty-eight came down as a single unit. Twelve of those eighty-eight were lowered on an overhead hoist. None of them were damaged. Seventy-six were lowered on a forklift. Nine of those seventy-six cracked the flange. So the procedure was not reckless. It worked sixty-seven times out of seventy-six. It was a loaded coin, and nobody had measured the bias, because nobody was collecting the throws.

Thirty-seven days grounded 11:14

On June sixth, 1979, the FAA suspended the DC-10's type certificate, and the fleet was grounded for thirty-seven days. Emergency directives went out to inspect every pylon attach point by eddy current, with the sealant stripped so it could not hide a crack. And the Board asked for something that had never been in type certification at all. Consideration of the failures that can follow structural damage in areas through which essential systems are routed. Six months before this crash, the damage tolerance rule had already been amended to require that analysis. The DC-10 was certified in 1971.

The warning that was never an event 11:56

This channel has now spent three episodes on warnings that existed and did not arrive. Two were documents that moved too slowly. One was a memo a contract made too expensive to pass along. This one is worse, because the warning here was not a document at all. It was the damage itself, twice, loud enough that the men standing under the wing heard it break. Nothing had to be hidden or overruled. It simply was not a reportable kind of event, so it never became an event at all. Ten months later, in the North Sea, a different structure fails for a reason that rhymes with this one. March twenty-seventh, 1980. The Alexander L. Kielland, a platform housing offshore workers, turns over in about twenty minutes. A hundred and twenty-three people die. The crack starts in a small weld, holding a bracket, for a sonar device that carried no load at all. It was welded onto a brace that carried everything.

Description and sources

On May 25th, 1979, American Airlines Flight 191 lost its left engine and pylon during takeoff rotation at Chicago-O'Hare. The airplane climbed to about three hundred feet, rolled left, and was down thirty-one seconds after liftoff. Two hundred seventy-one people aboard and two on the ground were killed.

The crack that ended it was made eight weeks earlier, in a hangar in Tulsa, while the airplane was standing still. American had devised its own way of replacing the pylon's spherical bearings: lower the engine and pylon together on a forklift, rather than taking the engine off first as the manufacturer's service bulletin assumed. The reason the popular version leaves out is that taking them down as one unit cut the hydraulic, fuel and electrical disconnects from seventy-nine to twenty-seven. It was argued on safety grounds, and the argument was real.

What it traded away is a number. With the engine removed, the most the geometry can put on the pylon's aft bulkhead is about nine hundred pounds. With the engine still attached, it can reach over twenty thousand — and that bulkhead's upper flange fractures under eight thousand. The mechanics then pulled the aft bolt first, out of the written order, because the written order was harder — turning the forward bulkhead into a pivot, with everything hanging on a forklift whose forks, a log entry three weeks later records, were creeping down under load.

The airplane flew three hundred forty-one hours with a ten-inch crack through that flange. When it let go, three feet of wing leading edge went with the pylon, severing the four hydraulic lines and two cables that held the left outboard slats out — lines fed by two independent systems, all routed through the same three feet of structure. The same separation unpowered the only stickshaker motor on the airplane and the slat disagreement light. The crew flew the procedure exactly: hold V2, one hundred fifty-three knots. The left wing now stalled at one hundred fifty-nine.

And it had happened twice before, in a Continental hangar. Both cracks were correctly classified as maintenance error, and neither was reported — the federal reporting system collects what fails after the engines are started with intent for flight. An FAA specialist, asked about those two cracks: "we would not consider it because it was not a service related problem." There was no box to put it in.

PRINT-READY, FROM THIS CHANNEL

The Failure Atlas, Vol. 01 — Tacoma Narrows · Citicorp Center · Millennium Bridge · Apollo 13 · the 2003 blackout · Hyatt Regency

https://therepository.gumroad.com/l/failure-atlas

PRIMARY SOURCES

NTSB-AAR-79-17, American Airlines DC-10-10 N110AA, Chicago-O'Hare International Airport, May 25, 1979, adopted December 21, 1979 — read directly, page by page, from the scanned original, which carries no text layer: https://www.ntsb.gov/investigations/AccidentReports/Reports/AAR7917.pdf
The 37-day suspension of the DC-10 type certificate (June 6 to July 13, 1979) is cross-checked against the FAA's own Lessons Learned record for N110AA: https://www.faa.gov/lessons_learned/transport_airplane/accidents/N110AA

The 37 technical plates in this video are illustrations generated for the channel by a diffusion image model, styled to match its cyanotype identity. They are diagrams of the system, not photographs of the hardware, and no person is depicted in any of them.

Root Cause investigates why engineered systems fail, using official investigation reports and primary technical literature. Sources for this episode are linked above.

The technical drawings in this video are cyanotype-style illustrations produced for the channel — diagrams, not photographs of the real hardware, documents, or wreckage. Charts and dimensioned comparisons are drawn from the sources listed above.

#engineering #aviation #rootcause