Table of Contents
18.104.22.168 Damages due to LCF
Damages due to LCF loads are already treated in the part-specific chapters (compressor, combustion chamber, and turbine). In Volume 2, Chapter 8.2 also deals with LCF in the context of fragment containment. Volume 2, Chapter 6.2 contains examples of the influence of fretting on LCF (Ills. 6.2-6 and 6.2-7). The special situation of thermal fatigue is dealt with in Chapter 12.6.3. Therefore, the following are merely several selected examples to illustrate the causes and consequences of LCF damage in rotors.
A focal point for causes of LCF crack initiation is material flaws and peculiarities. In the examples in this chapter, these flaws were evidently not detected or were not understood when the material came into serial use. Effects such as ingot-metallurgically conditional inclusions, pore formation, or structural characteristics (Fig. "When is a weak point a flaw limiting LCF"), combined with forging and heat-treating processes, can catastrophically reduce the life (required number of cycles) of engine parts. This reduction in life span can be related to the flaw size and therefore have a very short incubation time. In extreme cases, crack growth begins in the first few load cycles (Fig. "Long time effect of material flaws"). Crack growth can also be unexpectedly accelerated ( ).
Figure "Long time effect of material flaws" (Example "Turbine blade seperating from disk"): In this case of an older fan engine with a small bypass ratio, blades were thrown free in the low-pressure turbine. A radial crack spread from the hub to the blade annulus. The disk split open without bursting and released the blades. The crack was initiated at hard, ingot-metallurgical inclusions. This type of hard particle is characteristic of the disk material in question and decrease LCF strength considerably (Fig. "When is a weak point a flaw limiting LCF").
Excerpt 1, June 1983 (Ref. 12.6.1-5): “.NTSB is investigating a failure …in a… engine that caused an aborted takeoff…A preliminary visual check of the engine has focused the investigation on `a discrepancy in the hot section, localized in the second stage turbine disk'…Blades from the damaged engine…punched holes in the fuselage and tail of the …(aircraft). …(the pilot)was able to turn the aircraft into the high-speed accessed taxiway and execute a chute evacuation…“
Excerpt 2, June 1996 (Ref. 12.6.1-4): “The U.S. NTSB is calling for immediate and repetitive inspections of…turbine hubs produced before 1989 to detect fatigue cracks that could cause uncontained failures…As a result of its probe, the safety board is recommending that the FAA require immediate, nondestructive inspection of turbine hubs…if the hubs were manufactured from Incoloy 901alloy and cerium or lanthanum used as deoxidation agents. Periodic inspections would be accomplished at intervals not to exceed 5,900 cycles.
Investigation has revealed that a fatigue crack in the fourth-stage low-pressure turbine disk extended from the hub's core to a blade slot an NTSB official said. At a takeoff power setting, the crack expanded and allowed turbine blades to leave the disk, producing a large hole in the upper cowling section and inflicted shrapnel damage to the left side of the vertical stabilizer. The hub had accumulated 19,381 cycles…97 % of its approved service life of 20,000 cycles. It was produced in 1980 to 1981 from a one-piece machined forging of nickel-based Incoloy 901 alloy.
A detailed metallurgical examination of the hub indicated that the crack emanated from the bore of the turbine disk that contained `inclusions rich in cerium and lanthanum', the officials said. These latter elements were used during the foundry process to help deoxidize the alloy. When the ingot was initially cast, they rise to the top as dross and are later discarded…the failed hub was fabricated from a mult (forging blank) located near the top of the ingot adjacent to the dross. The …(OEM) began using only cerium as a deoxidant and in 1989 stopped using cerium in the forging process of Incoloy 901.
Similar fatigue cracks were found in two other …(same engine type) turbine disks prior to the…incident…As a result, investigators are concerned that the fatigue crack could occur `at any time during the service life' of a hub whose billet was exposed to a cerium/lanthanum mixture.”
Comments: Both of these excerpts concern the same engine type in the same aircraft type. It is not clear if the affected turbine stages were the same. Evidently the disk did not burst in either case. In the case discussed in Excerpt 2, blades were released by the widening crack. This would also explain the release of blades in Excerpt 1. The fact that the disks did not burst despite the large radial cracks is evidence of the extreme toughness of the disk material. It is interesting that 13 years elapsed between the two cases. Such a long interval before effective measures were implemented casts a shadow of doubt on the equivalence of these two safety-relevant cases.
Figure "Unpleasant surprises by 'standard' materials" (Example "Dwell time fatigue"): Large forged parts, especially made from titanium alloys, are required for the fan disks and rotors of high-pressure compressors (drum rotors; middle diagram) in large fan engines (top diagram). During the manufacturing process (casting and forging), many different material problems occur that are related to the size of the unfinished parts. In the depicted case, an unfavorable structure formed, primarily in the especially highly stressed hub area. As far as the damage mechanism and its causes are understood so far, the following conclusion can be reached: large grains with a special lamellar structure and an orientation perpendicular to
the main load direction are bad for the LCF behavior. The part-specific load progression (bottom diagram) causes unexpectedly rapid crack growth in the case of the described structural properties (Fig. "Damage accumulation and dwell times"). This can be explained by a very low fracture toughness. This leads to crack growth even with very small flaws and weak points (threshold; Fig. "Characteristic crack growth").
Dwell times of the cyclical loads (startup/shutdown cycle) seem to have an influence on crack growth (Fig. "Damage accumulation and dwell times"). According to Ref. 12.6.1-19, Ti-6Al-4V is sensitive to compressive stress (see page 22.214.171.124). This means that the dwell times under LCF-induced compressive residual stresses evidently have an especially strong decreasing effect on part life. This is a possible reason why the problem was evidently only recognized relatively late (in serial operation). Therefore, verification and design-relevant material data require sufficient dwell times at standstill during cyclical tests. Surprisingly, the effect is more pronounced at lower temperatures in the described case. Therefore, accumulated creep damage is not a plausible explanation. The accepted theory is that there is a relationship with hydrogen that has been dissolved in the material (Ref. 12.6.1-8).
The remedy is exchanging the rotors for others which are welded together from smaller seperate sections.
Example "Dwell time fatigue" (Figure "Unpleasant surprises by 'standard' materials"): In this example, several excerpts from various sources are provided to illustrate the causes and effects of the problem. The order of the excerpts corresponds to the dates of publication.
Excerpt 1 October 93 (Ref. 12.6.1-13): “Following takeoff, the airplane was climbing through 6,200 feet MSL over the ocean when the crew heard a loud bang and the number one engine fire annunciator light illuminated. The engine power decreased to zero. The engine was secured and the flight landed without further problem at the departure airport… Examination revealed that the engine had sustained an uncontained failure of the high pressure compressor. The 6th stage disk of the 3-to-9 compressor spool was missing completely from the engine. The airplane received minor damage. Similar disk failures have been attributed to a phenomenon known as dwell time fatigue. This may occur when titanium crystals form into microstructure colonies during the forging and manufacturing process. The disk has not been recovered and the exact failure modus has not been determined.
…As a result of previous spool separations, all one piece spools are subject to various service bulletins that call for one-time immersion ultrasonic inspection. As a result…12…16-inch billet spools have been found with cracks or crack-like indications…One 13-inch…billet spool was found cracked prior to the incident. The cracks have been attributed to a phenomenon known as “dwell time fatigue” (also referred to as “Quasi-Cleavage Cracking”).
Dwell time fatigue is characterized by flat, faceted internal crack initiation that may occur in areas subjected to high stress at low temperature over a period of time. Susceptibility to dwell time fatigue is associated with the presence of regions of microscopically aligned colonies of alpha phase titanium crystals. The colonies are aligned so that the basal plane is perpendicular to the axis of stress. The colonies form naturally during the billet manufacturing process. Subsequent billet reduction and part forging normally breaks up and randomizes the colonies; however, they may persist into the final part, resulting in a structure that is susceptible to fatigue over time (dwell time) under certain conditions.
…Including this incident, there have been four uncontained failures involving the 3-to-9 spool in… (two models of the same engine type).”
Excerpt 2, September 95 (Ref. 12.6.1-18): ”…the failure was caused by a fatigue fracture on the aft web of the stage 6 disc (of the 3-9 spool)…The crack propagated in a radial direction forward through the web, as well as inward toward the disk bore and outboard toward the disk rim, until it reached a length of 1.54 in. Further analysis indicated the crack `began propagating very early' in the spool's service life…“
Excerpt 3, April 96 (Ref. 12.6.1-7): ”…The safety board and the FAA disagree about how often the stage 3-9 high pressure compressor rotor assembly (spool) should be inspected to detect fatigue cracks…NTSB Chairman James Hall said the agency's current inspection interval of 3,500 cycles…'is insufficient and inappropriate to ensure detection of dwell-time fatigue cracks before failure….Tests conducted by …(the OEM) do indicate that the fatigue life of titanium alloy material subjected to dwell-time fatigue conditions was significantly reduced compared with conventional fatigue cycling, according to the safety board.”
Excerpt 4, March 98 (Ref. 12.6.1-6): “The NTSB urged the FAA to mandate inspections of all …(Engines and derivates of a civil high bypass type) series engines at intervals of less than 4000 cycles, to head off fatigue problems that have triggered about a half-dozen uncontained failures… Although they have categorized the failure mode, industry officials said,..(the OEM), the FAA and operators still do not fully understand what triggers dwell-time fatigue and at what rate it propagates to failure of critical rotating parts such as Stage 3-9 spools…“
Excerpt 5, August 2000 (Ref. 12.6.1-11): ”…Approximately 1,400 engines are affected by the airworthiness directive issued last week that calls for accelerated inspection of an interior portion of the engine's 3-9 compressor spool…titanium compressor spools have been subjected to inspections for years, but those inspection techniques were aimed at finding cracks resulting from hard alpha defects, cracks that can develop because of flaws in the compressor's titanium. The accelerated inspections are unrelated to the hard alpha defects. Instead the new inspections are aimed at finding cracks related to dwell time, fatigue cracks that develop and propagate over time. Up to the time of the June 767 failure, no …engine had suffered from dwell-time fatigue problems, but …laboratory tests had indicated that this problem could develop.”
Comments: This serious problem which profoundly affects security has evidently been present for about ten years, according to Excerpts 1 and 5. Apparently there have been many similar instances of damage. Far more than 1000 engines are affected. This indicates the complexity and demands of the problem. It is interesting that, according to Excerpt 4, the crack initiation mechanism and crack growth process are not completely understood by experts even after five years. In Excerpt 3, the considerably smaller number of load cycles to crack initiation in dwell time-overspeed tests relative to tests without dwell times is mentioned. Similar writings existed as early as the early 1970s (see Example "Importance of proper testing"). Evidently the knowledge from 20 years earlier did not lead to sufficiently safe verification procedures.
Figure "LCF fracture of a fan disk" (Example "Importance of proper testing"): In the two described cases, the bladed fan disk fractured (bottom left diagram). In one case, fragments struck the parallel engine (top diagram). An investigation revealed that a combination of undetectable flaws (pores) from the production process of the blank part, life span verifications without sufficient relevance, and residual stresses caused the damage (Fig. "Residual stresses and thermal strength"). These influences were exacerbated by the special constructive shape of the disk and the disk material, which is evidently susceptible to this type of damage (also see Fig. "Damage accumulation and dwell times").
A similar phenomenon, which is evidently closely related to dwell time fatigue, is cracking around pores. In the technical literature, this occurrence is referred to as sustained-load cracking (SLC) in titanium materials (Ref. 12.6.1-21). The pores were discovered in forged materials and welds. The cracks were evidently initiated by residual stresses or operating loads.
Example "Importance of proper testing" (Ill. 12.6.1-17):
Excerpt 1 (Ref. 12.6.1-10.1): “…The two fans which failed were both wing engines. They had completed 250 and 300 flights; and because one had desintegrated over the sea and the other over desert there has been little to go on…
The big longterm question for designers and airworthiness authorities is the adequacy of big-fan certification-testing. Were the fatigue tests (eight …hubs were cycled to an average 8,500 simulated flights each) fully representative of flight loads? Are the loads on engines mounted on flexing wings greater than, or different from, the stresses imposed by overspeed static tests? Are flying test beds of wing pods necessary?”
Excerpt 2 (Ref. 12.6.1-10.2): “…It is believed that about 40 fan hubs had been… (non-destructive tested) in the following way:
1 Visual inspection for cracks after surface-etching in dilute acid;
2 Binocular microscope inspection for surface cracks;
3 Zyglo surface crack detection …(penetrant inspection);
4 Eddy-current testing to detect deeper flaws:
5 Ultra-sonic testing, also to find still deeper cracks.
As a result several cracks have been discovered in discs removed from the 150-cycle plus engines…The most probable cause is considered to be fatigue failure of the titanium-alloy forged fan disc…The …pre-certification ground tests did not simulate all flight loads, but design took into account all loads with 10-20 per cent overspeeds (=40% higher maximum stress).
In..(one) incident the No. 1 engine fan left the pod, travelled forward (Volume 1, Ill. 4.5-6) under the fuselage and then aft, striking the No 3 nacelle (Fig. "LCF fracture of a fan disk").”
Excerpt 3 (Ref. 12.6.1-9): “…it is worth summarising briefly what happened and the lessons that have been learned.
The source of failure was quickly traced to the fan hub or disc - a machined titanium forging in IMI 685 material. The flight cycles in each case were low (335 and 279 cycles) and the failures were quite unexpected in view of results of cyclic tests on discs which has shown no failures below 9500 cycles.
A vast programme of spin tests with simulated and actual defects were undertaken but for a long time it was not found possible to reproduce the failure. Attention was concentrated on the mechanism of crack propagation and by use of all past data on faults leading to crack initiation…several significant facts came to light.
(a) Cyclic fatigue testing led to very significantly reduced number of cycles to failure if the test included a dwell time at maximum load representative of cruising time per flight. This was a vital fact as far as speed on this engine is high in cruise conditions. It represented an entirely new failure characteristic for this type of component.
(b) The design and method of manufacture led to the possibility of high locked in tensile stress on the recessed rear face of the disc.
© The forgings made from each billet were dissimilar in that the upper partner of the pair was prone to microscopic voids (approx .0001in dia) in the material which were not eliminated by the forging process.
Under high stress this could originate `fish eye' cracks from which fatigue cracks could propagate…
Meanwhile disc redesign and revised manufacturing techniques, to relieve the locked in stress was completed….it was decided as a result…on the redesigned disc in American 6/4 material (Ti-6Al-4V)…“
“In retrospect it can be seen that this trouble was encountered due to a combination of factors.
(a) The standard fatigue testing technique was not capable of detecting the weakness (also see Volume 1, Chapter 4.4).
(b) The production of the billet produced a proportion with undetectable internal voids which could originate fatigue cracks.
© Propagation of cracks from these origins was accelerated by the presence of high residual locked in stress resulting from the combination of manufacturing technique, material properties and the particular disc design.”
Comments: These damages first occurred in the early 1970s. They show the importance of selecting proper specimens for testing and also the importance of sufficiently relevant tests to operating behavior.
The small pores in the crack initiation zone are most likely related to hydrogen formation during the casting process of the primary material. It is not clear if the considerably larger fisheye cracks around the pores can be traced back to cyclical crack growth or spontaneous cracking. This could occur under the influence of hydrogen. This would mean that a relatively large initial flaw would suddenly appear, and the crack growth rate from this would be very high.
The disk material in this case, IMI 685, is characterized by very high thermal resistance. The change in material to Ti-6Al-4V indicates that this material was considered to be less sensitive to the damage-causing flaws. It is interesting to note that Ti-6Al-4V is the material that is causing major, possibly related, problems in the drum rotors of high-pressure compressors in other engine types (Example "Dwell time fatigue").
Figure "Characteristic damage by flutter vibrations": Several fan blades of the first stage of a fighter engine (left diagram) broke simultaneously during a test run during the engine development process. Evidently a range was reached that is critical for flutter vibrations ( ). The narrow, long blades do not have any supporting clappers. The damage symptoms can be seen as typical for flutter vibrations:
Several blades were broken in the radius of the transition from blade leaf to shaft.
All broken blades had a comparable fracture shape:
In the blade edge zone, arch-shaped dynamic fatigue fracture zones without lines of rest. The dynamic fatigue fracture at the leading edge is considerably larger and shows where the crack initiated. There are no flaws in the crack initiation zone that could explain the damage.
The fact that several blades failed simultaneously due to dynamic fatigue fractures indicates extremely high dynamic loads. Although this must have involved a high-frequency vibration, the loads were in the LCF range (see Fig. "LCF as lifespan determining"). The lack of lines of rest indicates a constant crack growth rate. If the crack growth rate had changed, it would have created lines of rest. The uniform fracture symptoms in all blades indicate that the vibrational mode was the same. The arrangement of the dynamic fatigue fractures indicates dynamic torsional loads. Therefore, a purely flexural mode can be ruled out as the cause of the overload (Fig. "Forms of typical blade vibrations").
All of these characteristics can be explained most easily with flutter vibrations. The introduction of clappers presented a logical remedy, and these then proved successful in operation.
Figure "Rotor blade fracture by LCF": This is a further example of an LCF fracture under high-frequency vibrations. As a consequence of a rotor blade leaf fracture, typical LCF fractures can occur in the remaining blades of the same stage (bottom right diagram). The blades run over the fragment that is lying against the inside of the housing (bottom left diagram) and are plastically bent.
The top diagram depicts the damage sequence in the remaining blading:
The rotor blade “A” is bent so far backward in phase “B” that the most heavily stressed flexural zone in the root area is plastically compressed. When it springs back and flexes open under centrifugal force (“C”), powerful tensile stresses with plastic strain are created in the compressed zone. This process repeats itself with every rotor rotation, which experience has shown leads to LCF crack initiation after a few hundred flexural cycles, and often also results in the fracture of cracked blades (“E”).
Figure "Ceramic coating outbrakes at a rotor spacer": Not only the disks, rings, and blades of a rotor are subject to LCF loads. Coatings themselves can also become fatigued and fail. Coatings can crack and/or spall. Cracks in firmly bonded coatings can initiate cracks in the base material and thereby worsen the LCF strength of the part.
The depicted case concerns the spalling of hard ceramic rub coatings on a rotor spacer. These damages occurred after several hundred startup/shutdown cycles in the engine of a fighter. They are most likely related to unfavorable parameters of the thermal spraying process. If the temperature control of the ring is not optimal during the spraying process, the result is damaging residual stresses between the coating and substrate. These stresses can act in combination with overlaying cyclical operating loads and lead to premature spalling.