Aeroengine Safety

Figure "Historic case of disk fracture by vibration" (Ref. 12.6.3.3-4): The early Whittle engine had a double-flow radial compressor (middle diagram). At 14,000 RPM it began to make a screaming sound. If the engine was operated for even a short time while making this sound, radial cracks would occur in the transition from the blades to the disk (bottom right diagram). Emitted sounds can still be used for verification of dangerous vibrations today, albeit in an altered form. During dangerously intense vibrations, the parts emit characteristic sonic frequencies. Frequency analysis is capable of using these sounds to detect fatigued parts.
The bottom right diagram shows the damage symptoms after the blade failure. The blades of the axial turbine also fractured due to high-frequency vibrations (also see Fig. "Vibration dampers at roots of turbine rotor blades"). In order to avoid this damage, important methods for analyzing excitement possibilities were developed at the time. These include the Campbell diagram (Fig. "Development preventing vibrations"). In addition to the typical expected fractures at the blade transition to the root platform (fundamental flexural modes), dynamic fatigue fractures also occurred in unusual/unexpected areas of the turbine rotor blades, such as the center or tip of the blade. Whittle identified the cause as a thermal element that was located in the exhaust gas flow roughly one meter behind the turbine. The realization that flow disturbances can be located far behind the excited part is always startling. After the disturbance was removed, the dynamic fatigue fractures no longer occurred.

Figure "Vibrations by remote engine copmponents" (Ref. 12.6.3.3-5): In January 1941, the engine Jumo 004A was successfully accelerated to full thrust during development. At this time, dynamic fatigue fractures of the rotor blades were a problem (also see Fig. "Vibrations by shaft caused flow disturbance"). The affected blades were compressor stator vanes made of sheet metal and without shrouds.
In the summer of 1943, several dynamic fatigue fractures occurred in turbine rotor blades (also see Fig. "Vibration dampers at roots of turbine rotor blades"). They were incited through the six can-type combustion chambers and the second harmonic of the three housing braces behind the rotor. The 36 stator vanes between the combustion chamber and turbine exacerbated the problem. The eigenfrequency of the blades was estimated by a musician by the sound they made when rubbed with a violin bow. This was an early form of modal analysis. The problem was solved by tapering and shortening (1 mm) the rotor blades and reducing the maximum RPM.

Figure "Vibrations by flow disturbance at compressor inlet": Vibration-inciting flow disturbances in the compressor inlet can be caused by external influences in many different ways. Typical causes are:

• FOD (e.g. bird strikes)
• Uneven distribution of temperature, pressure, and velocity in the inlet flow (Chapter 11.2.1.2).
• Shifting of the inlet cross-section through foreign objects (Example "Flow disturbance"),
• Icing (Volume 1, Ill. 5.1.4-2).The poor visibility into inlet ducts favors the accidental disposition of foreign objects (e.g. tools).

Figure "Vibrations by shaft caused flow disturbance" (Example "Dynamic fatigue fracture in a fan blade"): Evidently vibration problems in blades are not always eliminated in the development phase. This is especially likely if the application of the engine changes. For example, if a helicopter engine is used as a turboprop, then forces from the propellor can act on the filigreed engine. Unexpected problems with clearances and rubbing occur due to housing deformations that are dependent on the peculiarities of operation and the aircraft type. If the clearances are too tight, rubbing can excite unallowably strong vibrations in the blades. If the blade tip clearances are too large and/or uneven, they can excite blade and shaft vibrations (Fig. "Self exciting rotor vibration"). Another excitement mechanism that is directly related to the tip clearance is the rotating stall (Fig. "Damages by rotating stall"). If there are flow disturbances in the compressor inlet, such as a power takeoff shaft (top diagram), they can have a damage-promoting effect in combination with the described influences.

Figure "Imbalance by fan blade dynamic fracture" (Example "Increased dynamic loads under climbing flight conditions"): In this case, a commercial aircraft crashed due to a dynamic fatigue fracture in a fan blade. The blade vibrations were explained by coupled, self-increasing vibrations of blades and disk in connection with shaft deflection (whirl, also see Ills. 12.6.3.1-9 and 12.6.3.1-15). In this case, it was evidently not possible to recognize this damage mechanism in testing rig tests on the ground during engine certification before serial use.

Figure "Overheating and fusing at shafts by vibrations": The components (compressor and turbine disks) of this rotor from a small shaft-power engine are axially fixed with two centrical clamping bolts. The turbine-side clamping bolt had a stiff spring at the free end (top diagram). This was intended to ensure a sufficient axial pre-load. In addition, thermal strain differences between the turbine disks and the centrical clamping bolt had to be controlled, especially during fast instationary processes. During testing runs in the development phase, strong vibrations occurred during instationary operating states such as startup and shutdown. At almost constant rotor RPM, the frequency of the externally measurable vibrations changed. This appearance was explained by unmatched stiffnesses and thermal strain in the clamping bolt and compressor disks. Evidently, the assembly periodically loosened and changed the rotor eigenfrequency. This concerned very intense rotor vibrations, during which the contact surfaces between the compressor rotors experienced large relative movements. Overheating in the form of local welding beads occurred. Evidently large, high-frequency relative movements of the contact surfaces had caused this overheating (bottom diagrams). The overheated zones were evenly distributed around the circumference. Therefore, this at least temporarily involved a wave with several nodal diameters that travelled around with the rotor (Fig. "Vibration models of disks").
The solution was a spring on the free end of the clamping bolt on the compressor side (middle diagram).

Figure "Vibration dampers at roots of turbine rotor blades" (Example "Dampening vibrations to tackle fatigue cracks"): Dynamic fatigue fractures in turbine rotor blades have been a problem from the very beginnings (Figs. "Historic case of disk fracture by vibration" and "Vibrations by remote engine copmponents") of engine development. Unfortunately, in this case of a helicopter engine (middle diagram), the available literature gives no information regarding the source of the excitement, nor does it mention the turbine stage in which the fractures occurred. The excitement evidently occured via a flow disturbance and its scope was not recognized in the development phase. It is possible that the rotor blades of the first HPT stage (bottom right diagram) were excited by the turbine stator vanes in a certain RPM range. The detailed diagram at bottom left is a schematic depiction of the dampers made of metal sheeting that were introduced as a remedy for the problem (Fig. "Influences on damping of parts").

Figure "Identifyig dynamic fatigue in metal sheet parts": Thin, lamellar sheet metal constructions under high-frequency vibrations (sheet vibrations) typically exhibit branching cracks and material break-outs (top diagram, Volume 2, Ill. 7.2.2-24). In a turboprop engine type, these dynamic fatigue cracks occurred in the exhaust pipe flange after years of damage-free operation. This pipe was made from sheet metal that was about 1 mm thick.
The damage investigation revealed that the damage only occurred in newly installed parts. Their sheet metal thickness was within the specified tolerances, but was at the lower limit, unlike the older parts. Dynamic inspections revealed that the lower stiffness of the somewhat thinner metal sheets promoted dangerous vibrations.
Similar damage was observed in thin-walled labyrinth cones of another engine type (Volume 2, Ill. 7.2.2-24)
These examples demonstrate the influence of the sheet thickness on the stiffness and therefore also on the vibration-sensitivity of a part. This is influenced in two ways:
Changes in the eigenfrequencies: While the cube of the thickness goes into the stiffness, the frequency only increases slightly overproportionally when the vibrating mass does not become considerably larger, as well. This is the case if the change in wall thickness is limited to the fastening point (bottom right diagram). Contrary to expectations, this situation occurs relatively often in practice. It occurs in flange areas, for example, if the welded-on sheet metal wall has been deep-drawn for the necessary geometry.
Elastic deflection and stresses under excitement forces: This changes considerably with changes in stiffness. Excitement forces in thin-walled sheet metal structures are generally pressure vibrations in the gas flow and mechanical vibrations.
Therefore, it must be ensured that sheet metal thicknesses are correct during new part manufacture, repair, or when considering replacement parts. If in doubt, the guiding policy should be to use configurations that have been proven in operation.

Figure "Vibrations excited in ring shaped spaces 1" (Ref. 12.6.3.3-10): The fighter aircraft engine depicted in the middle diagram showed a rotor-dynamic instability when it went into serial operation in the mid-1980s. Up to 50% of the engines were rejected during the acceptance phase due to very strong non-integral vibration (NIV). Only the introduction of an oil-damped high-pressure turbine (HPT) bearing (bottom diagram) solved the problem.
The predecessor of the affected engine had already been in operation for years without similar problems. The rotor instability coincided with the introduction of the new HPT, even though the rotor dynamic was not noticeably affected by this.
Comprehensive investigations of over 30 engine configurations were conducted in order to understand the damage mechanism and the remedy and its verification. Differences between the cooling systems of the two engine variants were especially interesting. In the affected variant (top diagram), the cooling air (“D”) for the HPT rotor blades jumps over a tangential onboard injector (TOBI; “E”) on the front side of the first rotor disk (“F”). The labyrinth seals “A” and “B” form a ring-shaped space “C” ahead of the turbine blade, which distributes the cooling air onto the blades. The ring chamber “C” directs the air to the blades through holes in the plate that carries the rotor of the labyrinth “A”. Measurements in the ring chamber “C” showed a pressure vibration with the NIV frequency. This revealed that changes in the fan inlet pressure (engine inlet) influenced the amplitude of the vibrations. This led to the damage hypothesis of a cross-coupled self-increasing excitement/instability in the HPT system.
An analytical investigation dealt with the possible causes of instability (middle left diagram) in the high pressure rotor.
One characteristic was fretting wear in the flange threads (“G”) between the turbine disk and the shaft. This wear indicates that internal friction in this flange connection promoted rotor vibrations (Ills. 12.5.3.1-10 and 12.6.3.1-16). Internal friction of the flange can promote instability if the shaft that rotates in itself begins deflected rotation (“whirl”, Ills. 12.6.3.1-13 and 12.6.3.1-15).
It must be noted that the cross-coupled stiffness incites a force in the direction of rotation of the deflected shaft (orbiting movement of the whirl), creating an instability. In order to quantify the forces responsible for the instability, it takes a good “engineer`s sense” (!) in addition to the mathematical analysis. It was discovered that the frequencies of the system react sensitively to the stiffness of the suspension of the first turbine stator assembly.
A second possible instability-promoting influence was the different labyrinths around the HPT. Labyrinth seals affect the stiffness and damping of the system both directly on integral rotor components and via gas coupling. In contrast, the Alford force (Fig. "Tip clearance influencing 'orbiting'") only acts via gas coupling. The stiffness and damping of a labyrinth depend on clearances, pressure drops at the labyrinth, geometry, and rotating speed (see Volume 2, Chapter 7.2.1). They are not easy to accurately adjust due to the many influences.
A third suspected influence was a reciprocal aerodynamic influence between the instability and TOBI (top diagram). The pressure measurements in the ring duct “C” revealed a considerable influence of the turbine on the temporal pressure progression in the gas flow. This observation led to the assumption of a sinusoidal pressure vibration in the ring duct (Fig. "Vibrations excited in ring shaped spaces 2"). In this way, a moment was able to act on the turbine.
The individual instability-promoting influences were analyzed in tests in which specific design characteristics were changed for each test.

The final analysis of the analytical observations and the tests resulted in the following conclusion:

• It is extremely difficult to determine the cause of a non-integral vibration. Estimates demand iterative analytical and test-technical steps.
• The primary cause for the vibrations was an aeromechanical coupling of the cooperation of three instability-promoting propulsive forces in roughly equal amounts: Alford forces (Fig. "Tip clearance influencing 'orbiting'"), influence of TOBI, forces related to the labyrinth seals of the ring duct “C” and pressure vibrations in the cooling air inflow. Other realizations were:
• Variation in the influences can lead to variously strong vibrations (Fig. "Cyclic spin test disk fatigue crack ").
• Instability-promoting influences can be surpressed with sufficient safety through the use of a specifically designed damped bearing (middle right diagram, Fig. "Estimating allowable fatigue loads").

Figure "Vibrations excited in ring shaped spaces 2" (Refs. 12.6.3.3-10 and 12.6.3.3-12): This diagram shows the results of tests to determine the damage mechanism of the case described in Ill. 12.6.3.3-9.
In the ring duct between the TOBI and the turbine disk (top left diagram), a gas vibration forms that has six nodal diameters and rotates non-integrally relative to the shaft (top right diagram, Refs. 12.6.3.3-12 and 12.6.3.3-13). This is excited by a labyrinth vibration at low fan inlet pressure, and therefore low pressure levels around the TOBI, via an increased clearance gap flow (also see Volume 2, Chapter 7.2).

The instability-promoting influences in the various test variations of the engine, ring duct, and the surrounding labyrinths are assigned to the involved components in the bottom diagram:

“I” highest fan inlet pressure
“II” reduced fan inlet pressure
“III” inner static labyrinth carrier removed
“IV” opposite facing labyrinth tips (detail)
“V” as “III”
“VI” complete inner labyrinth seal removed

Figure "Rotor deflection during flexural mode" (Ref. 12.6.3.3-20): This diagram shows an example of the movement of a deflected rotor (orbiting). The available literature does not state which forces caused orbiting in this case.
Orbiting is a process in which a rotor is deflected by flexure, similar to situations involving a bent flexible shaft (Fig. "Orbiting mechanism and cause"). In contrast to orbiting, the deflection of an unbalanced shaft occurs at RPM rate of the shaft. The frequency of the flexure (orbiting frequency) is considerably below the rotor RPM, so that the rotor rotates around its bent axis during orbiting (Ill. 12.6.3.1-15).
In an 18-stage axial compressor (top diagram is only suggestive), rotor deflections of up to half a millimeter occurred in the eighth stage (bottom diagram). The rotor was running at 7700 RPM (128 s^-1), which is 100% RPM. The vibration frequency was attributable to the fundamental flexural mode of the system: rotor/bearing support. The tip clearance was measured at the 12 - and 3-o`clock positions (top right diagram). The orbiting frequency was in the range between 50 and 65 Hz in several tests, which is equal to 30% to 50% of the rotor RPM. The greatest deflections were recorded at the higher frequencies.
The primary cause was determined to be an overly soft rotor. Stiffening the rotor and the housing solved the problem.

Figure "Cyclic spin test disk fatigue crack ": The vibration excitement of a rotor disk can also occur with flexural modes of the rotor. In the depicted example, a specimen disk was cyclically stressed in a vacuum vertical centrifugal testing rig by an annulus that had been “drained” by keyhole bores (to simulate blade loads). After a short operating time, penetrative crack detection discovered several cracks (right diagram) in the annulus area that were primarily radially oriented and periodically distributed (bottom left diagram). The findings from the crack surface inspection indicated dynamic loads in the HCF range. This was plausibly explained by a disk vibration with three nodal diameters. The vacuum in the testing rig made vibration excitement due to aerodynamic forces impossible. However, there was also no effective air damping. Therefore, the only possible excitement was from the testing rig spike that powered the disk. The spike is elastically suspended in the testing rig on the free end, together with the disk. In the case of imbalance, it runs into a safety bearing. Evidently the rotor vibrated within itself in a way that no externally noticeable imbalances occurred.

Illustrations 12.6.3.3-13 and 12.6.3.3-14: This case occurred during the development phase of a small, low-performance turbine for a helicopter (top diagram). Without any indication of imbalances or other anomalies in operation, the integral cast turbine disk of the last stage burst at full RPM in a steady operating state. Fragments of the annulus had broken off the disk. The entire fracture surface could be seen as a combination of two fractures that ran together at an angle to one
another that was tangential to the balancing ring (bottom diagram). The fracture surface analysis revealed a bow-shaped limited fracture zone in the middle of each of the annulus fragments near the balancing ring. It also showed stage 1 characteristics (cleavage cracks) that are typical for this Ni-based cast material (Fig. "Stage 1 at dynamic fatigue cracks"). A testing-technical vibrational analysis (Fig. "Experimental determinating disk vibration modes" ) with heaps of powder resulted in a distribution pattern that corresponds with the bottom left diagram. This Chladni figure showed that a three nodal diameter vibration (Fig. "Vibration models of disks") is excited exactly at full RPM by the three bearing braces of the exit housing located behind the disk.
A successful provisional solution was the installation of a disturbing brace with no load-bearing function. It was only intended to dampen the vibration in the right rhythm with a disturbing pulse. The final solution was four equally-spaced supporting braces, from which no dangerous excitement was expected (middle diagram). The effectiveness of this solution was confirmed by later experience.

Figure "Fixing of centered rotor damping bolts": “Mysterious” fractures of the first stage turbine disk occurred several times during the development phase of a small shaft-power engine (middle diagram). Every failure occurred after very few operating hours. The fracture split the integral turbine disk, which was made of a Ni-based cast alloy with no central bore (!), completely in half (bottom diagram). At the base of the thread for the compressor-side clamping bolt, a zone with facet-like fracture structures was located almost in the center. These fracture structures are a material-specific sign of a dynamic fatigue crack. In the short operating times before damage occurred, only a few startup/shutdown cycles had taken place. Therefore, a high-frequency dynamic load in the shape of a one-nodal-diameter vibration of the turbine disk was suspected to be the cause of damage (bottom left diagram).
The causal vibration excitement of the very solid disk most likely occurred through flexural modes of the relatively thick and, therefore, heavy compressor-side clamping bolt. These vibrations can induce very strong axial forces on the ends where the bolt is fixed (top diagram and bottom example). This effect is even used in dynamic testing machines. The flexural mode of the bolt probably resonated with the vibration of the turbine disk. However, with such an intense excitement mechanism, it is also thinkable that there were other dangerously large loads apart from the resonance.

Figure "Fraction of turbine disk by nodal cycle vibration": This case concerns the fracture of an integral cast turbine disk from the second stage of a low-performance helicopter engine (top diagram). The fracture occurred following the initiation of several cracks at the circumference in the transition radius of a balancing ring (bottom diagram). The explanation for the damage was a nodal circle-vibration with a nodal diameter (Fig. "Vibration models of disks"). This vibration could have been excited by a neighboring disk that is connected to the damaged part by the centrical tie rod. The turbine disks of the first and third stages of the same engine type had fractured several times due to disk vibrations (Figs. "Cyclic spin test disk fatigue crack " and "Experimental determinating disk vibration modes").

Figure "Vibrations at seal membrane": In the area of the labyrinth bracket below the stator assembly of the first stage of the low-pressure turbine (top right diagram) of a large shaft-power engine, several damaging dynamic fatigue fractures occurred in the HCF range following a constructive change. These were dynamic cracks along the circumference (bottom right diagram) on the front wall of the ring duct that is formed by the seal bracket (top right detail).
The constructive change connected with the damage was a wire seal ring between the stator vanes and the labyrinth bracket. The seal effect against leakage air flow from the space with higher
pressure levels ahead of the wall to the ring duct merely relied on the resilient force of the disk-shaped front wall of the labyrinth bracket. After several hundred operating hours, the seal ring and the contact surfaces of the seal groove in the labyrinth bracket showed heavy fretting wear (Volume 2, Chapter 6.2).
The damage cause was thought to be vibration excitement of the resilient seal wall following sufficiently advanced fretting wear. This excitement principle (hydrodynamic paradox, Ref. 12.6.3.3-11 and Fig. "Vibrations of flat seals") is based on the resilient action of the cover and the pressure drop in the leakage air flow (Bernoulli) and is viewed as an exemplary physical laboratory experiment (bottom left diagram).
The damage hypothesis was confirmed with the aid of a simple demonstratory experiment. A stator assembly with a seal ring was set up and realistic pressure ratios similar to those in operation were simulated. As predicted, with sufficient wear and reduced resiliency, a strong vibration excitement of the covering plate with a corresponding pulsating leakage air flow could be observed.
After the damage mechanism was understood and reproduced using original parts, the requirements for a guaranteed solution were met.

Figure "Fatigue crack at compressor stator": In a small shaft-power engine (right diagram), compressor damage occurred during a trial run in the development phase. The inlet edge of the radial compressor disk of the second stage had broken out following an HCF dynamic crack (left detail). The dynamic crack did not have any pronounced lines of rest, which indicated constant crack growth. In addition, the stator (sheet metal) located behind the damaged disk had several dynamic fatigue fractures in its vanes and fastening bolts. This damage had already occurred frequently without cracks in the compressor disk. This indicated the following damage sequence:
First, the vanes of the compressor stator broke due to dynamic overstress (bottom diagram). Break-outs evidently created strong flow disturbances that acted against the airflow into the compressor disk. Here, it resulted in vibration excitement in the blade in a manner typical of the fundamental flexural modes of radial compressor blades, with crack initiation in the inlet edge area.

Figure "Gear excited fatigue fractures at cooler disk": After blades of an oil cooler disk made from a high-strength forged Al alloy (about 12 cm diameter, 20,000 RPM) were frequently affected by dynamic fatigue fractures, disks made of fiber-reinforced synthetic material were introduced. These disks, which are made from short-fiber molding material (fiberglass/epoxide), ran for a long period without problems. In one case, however, a disk suffered a dynamic fatigue fracture (top left diagram) despite the high inner damping of the synthetic material. The disk was replaced by an identical new part. This disk also fractured in a similar way within about 100 hours. This made a damage-causing flaw in the disk unlikely.
The damage anlysis revealed that a toothed gear in the drive chain (bottom right diagram) was probably damaged during installation. The damage was a hardly noticeable deformation of the edge of the roll-off surface. This “minor” tooth damage evidently led to the extreme dynamic loads. After the gear was replaced, no further dynamic fatigue fractures occurred in the oil cooler disk even after long run times. This example demonstrates the danger of vibration excitement due to tooth frequencies.