History of Flight
On March 26, 2009, a McDonnell Douglas DC10-30F, registration N526MD, experienced an uncontained No. 2 engine failure during climb out after takeoff from Eduardo Gomes International Airport, Manaus, Brazil. According to the flight crew, about 30 minutes after takeoff, at approximately 8,000 feet above ground level, the No. 2 engine oil pressure steadily decreased. The captain shut down the engine and continued the flight. The flight was later diverted to José Maria Cordova International Airport, Medellin, Colombia, where an uneventful landing was accomplished.
Damage
A post-incident inspection revealed damage to the tail strut, elevators, rudder, horizontal stabilizer, and the No. 2 engine pylon and nacelle. The aft end of the No. 2 engine had separated roughly in plane with the low pressure turbine (LPT) case; all components aft of the separation were missing. Liberated engine parts impacted a densely populated area near Manaus, damaging 22 homes. There was no major structural damage to the airplane, but extensive tearing damage occurred to the No. 2 aft pylon fairing, aft engine support mounting bracket, and No. 2 engine core cowl doors. Perforations and dents were found in the rudder, upper and lower skins of the left and right wing elevators, and the right horizontal stabilizer.
Aircraft and Engine Information
The airplane, a McDonnell Douglas DC10-30F, was powered by three General Electric CF6-50E2 turbofan engines. It was registered to Miami Leasing, Inc. and operated by Arrow Air Holdings Corporation, doing business as Arrow Cargo. The incident engine, installed in the No. 2 (tail) position, was owned by AAR Corporation and leased to Arrow Cargo. Maintenance records showed the engine had accumulated 83,895 hours and 21,904 cycles since new, with an overhaul 7,883 hours and 2,703 cycles before the incident. Borescope inspections performed on May 5, August 8, and October 28, 2008, had found the high pressure turbine (HPT) in airworthy condition per airplane maintenance manual limits.
Flight Recorder Data
The flight data recorder, a Honeywell Model 980-4100, recorded 22 parameters. Data showed that the No. 2 engine fan speed began to decrease as the throttle was increased to climb power, continuing to decrease for about 65 seconds, reaching 14 percent RPM and remaining below 27 percent until the data ended. The flight crew reported no unusual engine vibration before the event.
Engine Examination
The incident engine was examined at Kelly Engine Center, San Antonio, Texas, after recovery of liberated parts by the Brazilian Air Force (CENIPA). External inspection found no obvious damage forward of the turbine midframe (TMF). Aft of the TMF, there was localized crush damage, and the LPT stator case was separated 360 degrees roughly in plane with the LPT S3 disk plane of rotation. The LPT S3 disk rim, web, bore, and aft spacer arm; 39 LPT S3 nozzle segments; all LPT stage 4 nozzle segments; the LPT S4 disk; and the turbine rear frame had liberated. The LPT S3 disk forward spacer arm was attached to the LPT forward shaft and fractured circumferentially approximately 1.54 inches aft of the flange forward face. Disassembly revealed that the total HPT rotor unbalance was 1,213 gram-inches, far exceeding the limit of 40 gram-inches. A single HPT stage 1 blade was severely eroded, with about 85% of its airfoil material missing. Combustor components showed wear consistent with aftward shifting during operation. The exhaust gas temperature (EGT) system was in poor condition. Metallurgical examination of the LPT S3 disk fracture surfaces showed features consistent with high-cycle, high-amplitude fatigue crack propagation over about 90% of the fracture, with the remaining 10% showing overstress features. The disk material met specifications except for a small deviation in grain size, which did not contribute to the fatigue fracture.
Research and Findings
According to General Electric, high HPT rotor unbalance can cause synchronous vibration forces that interact with the LPT rotor through a common bearing sump, exciting a bladed-disk mode in the LPT S3 disk. This mode results in forward spacer arm bending loads exceeding the fatigue limit, leading to crack initiation. This resonance response was first identified in the CF6-6 engine, which experienced four uncontained LPT S3 disk separations between 1975 and 1978, leading to a redesign. The CF6-50 engine had experienced ten instances of LPT S3 disk forward spacer arm cracking between 1973 and 2009, with two progressing to in-service uncontained failures. The DC-10 was certificated with an engine vibration monitoring (EVM) system, but it was considered marginal for detecting CF6-50 HPT rotor unbalance. The FAA had placed a special condition on the CF6-45/-50 type certificate requiring demonstration that the engine would operate without detrimental stresses from undetected increased vibration; this was satisfied by a certification report concluding that such unbalance would be easily detected by noise and vibration in the cabin and controls. EVM systems were later decertified in many CF6-50-powered airplanes. Arrow Cargo's EGT trend data for the incident engine lacked sufficient fidelity to determine a performance trend.
Regulatory Actions
As a result of this investigation and a similar event on July 4, 2008, the NTSB recommended on May 27, 2010, that the FAA require more frequent borescope inspections and fluorescent penetrant inspections of the LPT S3 disk, and that General Electric redesign the disk. The FAA issued several airworthiness directives: AD 2010-06-15 (March 17, 2010) requiring fluorescent penetrant inspection and borescope inspections; AD 2010-12-10 (June 9, 2010) adding EGT system checks; AD 2011-02-07 (February 7, 2011) requiring on-wing inspections, vibration surveys, and ultrasonic inspection; and AD 2012-02-07 (January 31, 2012) establishing a new lower life limit for the old-design disk. A new-design CF6-50 LPT S3 disk with improved tolerance to HPT unbalance forces was certificated in June 2011.