1. David Dye Department of Materials, Imperial College Royal School of Mines, Prince Consort Road, London SW7 2BP, UK +44 (207) 594-6811, david.dye@imperial.ac.uk © Imperial College London Engineering Alloys (307) Lecture 8 Case Study 2: Sioux City

2. © Imperial College London Page 2 Why Case Studies? MSE307 is about how alloys are designed and used, as well as (less important) a survey of alloy systems –How engineering systems fail has a large influence on the alloy design and certification process i.e. We always fight the last war! Aim to –Make you think about how a failure investigation is run –Make you realise that a production error, undetected at the time, may result in court appearances by the supervising managers (you) later! –Illustrate that systems fail as systems, and that usually no one part is to blame (so everybody is to blame)

3. © Imperial College London Page 3 How will these Aims be Assessed? May be a question on its own, asking you to discuss ONE of the cases that you know of (incl. further study), or May be part of a connected technical question, e.g. –defect acceptance limits –defect types –lifing philosophy –predicted defect growth rates Will ask you to give the direct materials failure AND outline the fault sequence MAY ask you to discuss the indirect causes MAY ask you to discuss WHY the failure occurred (who do you think was to blame?) MAY ask you to discuss what might be done differently MAY ask you what is meant by a ‘Safety Culture’ and compare that to a risk-taking or innovation culture

4. © Imperial College London Page 4 UA232 a.k.a. The Sioux City Disaster

5. © Imperial College London Page 5 What Happened? A Ti fan rotor disc failure Debris disabled all 3 hydraulic flight control acutation systems Crew could not effect a controlled landing 111/296 persons aboard died when the airplane broke up on landing

6. © Imperial College London Page 6 Fault Sequence UA232 was a DC-10 delivered in 1971 that was 26 years old with an airframe life of 43,000 hrs and 17,000 flight cycles The flight was from Denver, CO to Philadelphia, PA ~1hr into flight at 15:16, the tail-mounted engine failed with a loud bang. Hydraulic power was lost, and could not be restored using backup systems. Airplane roll could be corrected via wing engine power settings Crew elected to land at Sioux Gateway Airport, closer than Des Moines Int’l. Sioux City Category B, rather than Category D at Des Moines – not rated for DC10, only smaller 727 / 737 etc. Fire cover provided by Air Nat’l Guard. Crew concerned about (i) structural stability of airframe, (ii) ability to land plane. Difficulty in controlling speed, heading and rate of descent simultaneously.

7. © Imperial College London Page 7 Aircraft ground track

8. © Imperial College London Page 8 Fault Sequence (2) At ~15:42, dumped remaining fuel down to reserve levels Landed on (disused) runway 22, 6,600 ft long, rather than runway 31, 9,000 ft long. Crew wanted to avoid a go-around, and happened to align on runway 22. Landed at 16:00, left of the centerline on runway 22. Contact by right wing tip, right engine nacelle, right main landing gear. Airplane skidded, rolled and broke up. Debris scattered over an approx 1mi. long track. Fire crews attended, ran out of retardant, but provided blanket cover for many passenger to escape. Failure of the reserve tender’s resupply hose led to a ~10min gap in the application of foam, due to rotation of a 50¢ plastic stiffener. This meant it took a further few hours to control the fire, but probably resulted in only a few more deaths.

9. © Imperial College London Page 9 Debris Track

10. © Imperial College London Page 10 Seat Map

11. © Imperial College London Page 11 Might the Crew have been expected to be able to land safely? Simulator tests with experienced DC-10 pilots, trainers and test pilots concluded that effecting a landing with 2 engines and no control surfaces was not trainable Crew experience was significant Although landed at ~1500 ft/min rate-of-descent and 200+ kts IAS, very little they could have done last DC10 check yrs - airline hrs - lifeyrs – DC10 hrs – DC10 Capt.3mo3330k137k 1 st 11mo3020k5+0.6k 2 nd 1mo315k2k Check3mo314.5k+3k

12. © Imperial College London Page 12 Should a bigger airport have been tried? With hindsight, the airframe was probably OK to fly for some time –could have vectored to a bigger airport, –assessed the condition of the plane better, and –prepared emergency crews more fully Stresses the importance of snap decisions by aircraft crew, Air Traffic Control and United Ops center

13. © Imperial College London Page 13 Should the fire crews have extinguished the fire? Used more retardant than were required to have on hand Bigger incident than would have trained for Had never tested tender resupply hose in operation Once the plane is on the ground, in pieces and burning, these are trivial, besides-the-point issues

14. © Imperial College London Page 14 Much comment at the time that the airplane design was deficient In this case, triple redundancy in the hydraulics did not resolve the problem – all 3 failed due to turbine debris striking the rear left horizontal stabiliser Should the disc failure have been ‘safe’?

15. © Imperial College London Page 15 No: Failure should never have occurred Rotating parts should not fail: review reveals that airplane loss occurs ~25% of the time after a disc failure GEAE expected life for a defect-free disc was 54k cycles, FAA rated to 1/3 rd of that: 18k cycles Energy contained in a disc (170kg at ~40,000 rpm?) is so large that the fragments are uncontainable Alloy Ti-6Al-4V: very well established and widely used So what happened?

16. © Imperial College London Page 16 History of #2 Engine and Stage 1 Disc Engine #2, a GEAE CF6 SN 451-243 had 42.4k hours total time The stage 1 disk, SN MPO 00385 manufactured at GEAE 3-Sep to Dec-11 1971. Accumulated 41.0k hrs / 15.5k cycles (originally installed in another engine). Inspected 6 times during routine overhauls, most recently at 38.8k hrs / 14.7k cycles (not required at all for disc until 14k cycles).

17. © Imperial College London Page 17 Disc GA

18. © Imperial College London Page 18 Disc Reconstruction Disc reconstructed from pieces found in fields near the initial in-air failure, and from crash site

19. © Imperial College London Page 19 Fractography

20. © Imperial College London Page 20 Fracture Mechanics of the Failure NTSB could count the striations on the crack and match to the number of cycles: defect had existed since new Crack grew in-line with GEAE fracture mechanics calculations At last inspection, crack was 0.5in long and 0.05in across in the forward bore area Dye penetrant fluid residue was found on the crack surface Why was it not found by the inspector? –working alone –piecework – appropriate for a critical job? –difficult inspection location / procedure –low awareness of potential cracking issue at bore, focus was on blade root cracking (but blade release is containable)

21. © Imperial College London Page 21 So where did the crack come from? Microstructure of region ‘C’ a nitrogen-stabilised hard-alpha inclusion, approximately 1mm in diameter Should have been found in the manufacturing process, one would think? –disc ultrasonically inspected and macroetched in Rough Machined Form (RMF), before final finishing. Not subsequently etched. -> defect missed. –NTSB concluded that hard alpha inclusion was knocked out during final shot peening.

22. © Imperial College London Page 22 Heat Tracking GE had records for 2 discs with SN MPO 00385 Inspection of 6 of the 7 sister discs from the same forging by Alcoa, Timet heat K8283, resulted in 3 with rejectable defects, 2 alpha case and one overheating during forging Confusion as to whether Timet or RMI heat Possibility that the set of 8 discs from heat K8283 actually came from two heats RMI heat Ar-melted, Timet Vacuum-melted (Ar melting withdrawn in late 1971). (not in report) Some inspectors convinced that discs were swapped between the two heats on Alcoa’s factory floor

23. © Imperial College London Page 23 Implications for Manufacturing Switchover to Skull melting / hearth melting process (80s) Mandate vacuum triple melting (1972) specify no remelted material in critical rotating components COST major stumbling block to use of Ti

24. © Imperial College London Page 24 Implications for Lifing Philosophy NTSB suggested moving to a damage tolerance philosophy, e.g. assume a defect present, then retire for cause based on inspection Advantage: –avoids arbitrary 1/3 rd life retirement of >90% of components: $$$ Disadvantage: –requires the inspectors to catch every flaw, which they didn’t in this case At present: continue to assume no flaws, ramp cost by continuous improvement in quality of manufacture -> space shuttle approach

25. © Imperial College London Page 25 Future Directions – Damage Aware Materials Another approach is to use a material that will tell you about its flaws, on the flight line e.g. incorporate conducting wires into composites, detect breaks then can retire only for cause, and catch every defect!

26. © Imperial College London Page 26 Conclusions - 1 UA232 was lost due to the failure of a Ti-6Al-4V stage one fan disc in flight that lead to the loss of the hydraulic control systems and made lading the aircraft impossible –UA failed to detect the flaw –GEAE failed to track their components properly –Alcoa may have swapped remelt / non-remelt and vacuum / Ar remelt material on their factory floor –Douglas failed to design a hydaulics system that was damage tolerant –The flight crew could have selected a better crash location –The fire crews suffered from equipment failure -> Everybody was at fault in some way

27. © Imperial College London Page 27 Conclusions - 2 Records matter! Materials tracking matters! Processing matters! Lifing philosophy is a diffcult issue! What you do on a factory floor as a 20-something can bite you 27 years later when you are divisional director! Failure investigations involve a huge CYA operation on all sides, making the inspectors’ lives hard Metallurgists, fatigue prediction, process design and QA procedures all become critical in an investigation

28. © Imperial College London Page 28 Review: Nickel Superalloys II (L5-6) Theory of Cast Ni Superalloys Manufacture Coarsening & Strength Ordering 5nm  ’ r dr