Download as pdf or txt
Download as pdf or txt
You are on page 1of 14

J Fail. Anal. and Preven.

DOI 10.1007/s11668-016-0162-6

TECHNICAL ARTICLE—PEER-REVIEWED

Fatigue in Rotating Equipment: Is it HCF or LCF?


Nicholas E. Cherolis . Daniel J. Benac . Wesley D. Pridemore

Submitted: 1 August 2016


Ó ASM International 2016

Abstract This article discusses practical differences distinction between HCF and LCF as a means to determine
between high-cycle fatigue and low-cycle fatigue at rela- the root cause of the failure and aid in seeking means to
tively lower temperatures in rotating equipment and the prevent future occurrences. This paper discusses the issues
equipment attached to it. Methods to identify the failure that can arise when trying to determine the mode of a
mode are discussed so proper prevention measures can be fatigue failure: was it slow growing due to start and stop
put in place to prevent future occurrences. Examples of cycles, or did it rapidly progress due to a high frequency
typical components are provided to show the methods in stimulus? The root causes of these two types of fatigue are
action. so different that the search for solutions takes one down a
different path, depending on the active fatigue mode.
Keywords High-cycle fatigue  HCF  Although examples from rotating equipment and associ-
Low-cycle fatigue  LCF  Fracture  Crack initiation  ated equipment will be used to demonstrate these
Fatigue origin  Striations  Facets principles, the concept can be applied to other equipment
as well.

Introduction
Low-cycle Stresses in Rotating Equipment
Rotating equipment is found in many industries, such as
aerospace and industrial gas turbines, industrial compres- The stresses that a rotating component sees can vary in a
sors, turbine expanders, generators, gearboxes, and linear fashion with the rotational speed, with some areas
machinery such as rolling mills. This rotating equipment is experiencing hoop stresses while others experience radial
subject to fatigue loading. Non-rotating equipment such as stresses or a combination of both. Stresses are pre-
tubing, studs, bolts, supports, and piping, however, is also dictable and can be accounted for in the design, usually
subjected to fatigue due to its association with the attached through finite element modelling.
rotating equipment. Numerical boundaries are often set Fatigue testing of materials results in plots of stress (S)
between high-cycle fatigue (HCF) and low-cycle fatigue as a function of fatigue life cycles (N), resulting in what is
(LCF) in such equipment as a means of categorization. In called an S–N curve. These curves are available for a
this article, we hope to take that further and use the variety of combinations of material, condition, tempera-
ture, and stress ratio. Highly stressed components typically
have multiple stressed areas competing to be the ‘life-
N. E. Cherolis (&)  D. J. Benac limiting area’, and each is considered when doing an
Baker Engineering and Risk Consultants, Inc., 3330 Oakwell analysis. For some equipment, the fatigue life is limited, so
Court, Suite 100, San Antonio, TX 78218, USA the components must be replaced at some point below the
e-mail: ncherolis@bakerrisk.com
calculated crack initiation life, with a safety factor added.
W. D. Pridemore Any infidelity in choosing the boundary conditions (friction
GE-Aviation, 1 Neumann Way, Cincinnati, OH 45215, USA

123
J Fail. Anal. and Preven.

coefficients, temperatures, etc.) in the analysis can result in bending) and can develop fatigue after a high number of
inaccurate calculations and even selecting the wrong area cycles. This high frequency of loading is what brings these
as life-limiting. Although rare, a failure identified as LCF components into the possibility of HCF failure. These cracks
can cause re-evaluation of the analysis leading to identifi- are classified under the category HCF not only because the
cation of the source of inaccuracy, giving rise to a rational number of cycles involved is high, but because the cracking is
basis for the reduction in predicted life or, even worse, lead not directly related to start-stop and cycles, which are the
to a redesign of the affected component, in an attempt to main stress sources in LCF type cracks. Although the alter-
recover the stated life. nating stresses are lower, cycles can accumulate very quickly
Test bars are used to determine S–N fatigue curves for a depending on the frequency of loading. For instance, at
particular material and temperature, and stress ratio. Fig- 15,000 rpm, just one hour of operation exposes a shaft to
ure 1 shows the fracture surface from a test bar tested in 900,000 alternating stress cycles. Therefore, after just over a
higher alternating stress LCF conditions. In this case the thousand hours of operation, the equipment can be exposed to
test bar was coated with a thermal spray deposit. Obvi- almost a billion cycles. Normal fatigue testing for steels
ously, cracks initiated at multiple locations around the bar involves testing to 107 cycles, but in these instances, vibra-
circumference. If the cracks had been present in a real part tions can regularly expose material to operation above 1010
and discovered in service, the fracture features would cycles. In recent years, unexpected failures have been repor-
confirm it was LCF and analysis of the fatigue striations ted, even for structural components made from ferrous metals,
would indicate how fast it was growing. Knowing that this which were assumed to have a distinct fatigue endurance
is an LCF crack would mean that the component could be limit. In response to this, much effort has gone into research
removed from service in a rational manner based on into the very HCF regime longer than 107 cycles [1–3].
accumulated, higher alternating stress cycles typically Normally, stresses in components subject to vibration
related to the stop/start cycles or other alternating cycles. are intentionally kept below the endurance limit in order
to avoid failure in fatigue. Manufacturing issues can result
in residual stresses, undesirable surface finish, or even
High-cycle Fatigue Stresses in Rotating Equipment incipient cracks that can eventually result in the propa-
gation of fatigue cracks. The tubing shown in Fig. 2 was
Most life-limited rotating components are designed to coun- carrying vent air in a piece of rotating equipment.
teract the stresses at max rpm, such that they are not subject to Vibrations led to fatigue cracking from a single origin
inherent vibration modes within their normal operating range. region associated with a remnant tack weld on the OD
Imbalances are also carefully controlled and critical speed surface.
ranges are avoided to avoid rotational instabilities and Other components within rotating assemblies are subject
vibration modes (resonance). Shafts and gears, however, can to inherent vibration modes. Compressor and turbine
be loaded by imbalanced forces or periodic loads (as in tooth blades are usually removable components fitted into slots

Fig. 1 LCF test bar fracture. This fracture surface (a) shows multiple origins around the circumference of an LCF test bar. The test was
performed to determine the effect of the thermal sprayed coating on overall fatigue life. Some of the crack origins are shown in the SEM
backscatter image (b)

123
J Fail. Anal. and Preven.

Fig. 4 Natural separation between HCF and LCF [7]. It is important


to determine which fatigue type is operating so the cause can be
Fig. 2 HCF Fracture in vent tube. ‘‘Beachmark’’ features indicate the properly addressed
positions of the crack as it propagated around the tube circumference.
Crack propagation is perpendicular to the ‘‘beachmark’’ features.
the diagram to check if it is comfortably below the
Crack initiation occurred from a remnant tack weld. Arrows show the
directions of fatigue crack propagation Goodman line, which would indicate cracking in the
selected number of cycles. Another option is the Walker
approach, with an adjustable exponent that may be fitted to
test data, allowing superior accuracy [5].

Separating High from Low

Why discuss HCF and LCF, as the S–N curve includes all
ranges of life from the upper high stressed region to the low
stressed region with extremely long lives? Yes, there are
components that exist between classic high and LCF areas
of the curve, and 105 cycles is generally understood as the
divider between high and low cycles [6], but that is not
useful for our purposes. There is a natural division between
the two types of fatigue we have been discussing that
Fig. 3 Mean stress effects. This modified Goodman diagram shows separates the causes, and hence the method of prevention,
the effect of mean stress on fatigue life to various numbers of cycles
into two distinct classes as shown in Fig. 4.
The idea is to identify whether the driving force causing
in disks, and these components are designed not to operate the fatigue is a larger applied alternating stress occurring at
in any resonance vibration modes that could lead to a slower, manageable rate (LCF) or a lower stress resulting
increased stress amplitudes during operation. The vanes in from a vibration or other loading occurring at higher fre-
rotating equipment have similar requirements. Unfortu- quencies (HCF). So what is a high frequency? For
nately, operating conditions can change and are not always example, a shaft rotating at 15,000 rpm, would see 15,000
what was assumed in the original design. When the com- cycles per minute, giving 250 cycles per second (Hz),
bination of mean stress and vibrational amplitude exceed which is approximately middle C for those who are
the capability of the material, failures can occur. musically inclined. Consider the passing frequency of a
A key limitation to the S–N curve used in LCF analysis compressor blade in that same rotor going past a vane in
is the inability to predict life at mean stresses different the adjacent stator. If there are 39 vanes it is passing every
from those under which the curve was developed. In pre- revolution, the passing frequency is 15,000 9 39, or
dicting the life of a component in HCF, a more useful 585,000 pulses a minute, or 9750 Hz, or about the fre-
presentation of fatigue life test data is the modified quency of ringing in your ears. At that rate, cycles can add
Goodman diagram as shown in Fig. 3 [4]. Such a diagram up fast. Contrast this with a piece of machinery that is
allows the user to predict life at different mean stresses. brought up to speed and the disks stressed at maximum
The maximum mean stress and the alternating stress from a RPM with no significant change in stress until it is stopped
calculated or measured vibration mode can be plotted on and restarted, repeating the application of LCF stresses. It

123
J Fail. Anal. and Preven.

can take years at this rate to accumulate a significant


number of higher alternating stress cycles.
Determining you are dealing with an LCF crack tends to
lead to a life-based field campaign to remove cracked
components either by inspection or early retirement. Such a
tactic is hard to justify, however, for random fast growing
cracks that are caused by vibration affects. One problem is
the inspection intervals would have to be very frequent to
catch a HCF crack between the time it is detectable and
failure of the component. HCF cracks can grow so fast that
Fig. 5 Size of the fatigue region compared to overload region. This
some components can initiate and propagate to failure
schematic shows that the higher stresses associated with LCF usually
within one service cycle. Continuous monitoring of results in smaller fatigue fracture with a relatively large overload
vibrations can be useful to detect changes that could affect fracture (left, shaded area). For an HCF fracture, lower stresses result
airfoil separations or bearing spalling issues, giving a in a larger fatigue fracture with a relatively smaller final tensile region
(right)
reason to remove such equipment from service, reducing
the chance for additional damage.
The steadily accelerating growth of LCF cracks also detected, one cannot make such an easy judgment about
lends itself to the application of fracture mechanics to operating stress.
predict the future crack growth rate and critical crack size
at which the component would fail catastrophically. For the
fractographer, fatigue striations are key to confirming Number of Origins
actual crack growth rates. If good crack growth rate data
can be obtained from the fracture surface, the stresses and Multiple origins separated by steps are called ratchet marks
other assumptions can be confirmed and compared to a and tend to indicate LCF, as shown in Fig. 1. There are two
known stress model, if such a model exists. Changes can reasons for this. One is that the variability in crack initiation
then be made to lower the stress or lower the life of the life from one area to the next is narrower in the high stress
component to safe levels. region than in the low stress region of the SN curve, so as
On the other hand, HCF failures tend to be resolved by one area starts cracking, other nearby areas are close to
adjusting the component shape, adding clamps and brack- cracking. Multiple, smaller cracks initiate at nearly the same
ets to unsupported structures, or adding dampening features time and grow until they meet and join up, creating a ratchet
to ‘‘tune out’’ vibrations. For shafts, the stresses are less- mark which is a slight mismatch between the cracks planes.
ened by balancing the rotating assembly and avoiding For HCF cracks, the crack typically initiates from a single
critical speed ranges (resonance) in order to avoid dynamic origin region and can lead to a complete fracture before
instability. Another HCF strategy is to use surface modi- another nearby location can initiate a crack. One exception
fications such as shot peening that puts the surface into to this is reversed bending in thin section sheet metal
compression, thus shifting the near surface mean stress components. The bending stresses can cause multiple fatigue
experienced by the component. Most other surface treat- initiation sites on each surface. Cracking progresses from
ments such as coatings and dimensional restoration by both sides, typically meeting somewhere in the middle of the
flame spray, however, will likely lead to degraded LCF sheet thickness, depending on the symmetry of the bending.
properties. Rotating bending in notched shafts also can result in origins
all around the shaft, with propagation toward the center [8].
A large enough anomaly, such as an internal inclusion,
Fatigue Crack Size will tend to cause a fracture to start at one origin and may
even cause cracking to initiate remote from the predicted
If a component separates due to fatigue and the overload high stress location. In this case, the anomaly is usually the
area is small, then the maximum stress is probably low, real cause of the failure and the stresses may be found to be
indicating HCF. Conversely, components stressed high adequate for a normally unflawed structure.
enough to fracture in LCF typically exhibit smaller cracks
with much larger final fracture areas as indicated in Fig. 5.
There is a much more extensive schematic in the Metals HCF Facets
Handbook Volume 11 that shows the effect of rotation,
torsion, bending, notches, and stress level [8]. Of course, if Stage I fatigue facets clustered in the origin area supports
the crack has not gone to complete failure when it is HCF, as shown on the nickel-based turbine blade fracture

123
J Fail. Anal. and Preven.

Fig. 6 Nickel-based turbine blade fracture with faceted fatigue origin

in Fig. 6. Origin regions can sometimes be traced down to


small inherent defects such as the small residual casting
pore shown in Fig. 7. Faceted initiation regions can be both
surface and sub-surface in nature. Shot-peening tends to
promote sub-surface, faceted origins since the surface
possesses a more compressive condition. Faceted Stage I
origins are also more commonly observed as the grain size
increases. Some alloys like Waspaloy and IN901 are
known to exhibit extensive areas of faceted fracture when
tested in LCF conditions, but these are the exceptions.

HCF Fractographic Progression Features

Directional features tend to dominate at high magnification


on HCF fractures in Stage II, usually observed as fan-like
or feathery features, as shown in Figs. 8 and 9. In titanium
alloys, the lower stress fractures and fractures nearer origin
areas for LCF fractures can be greatly affected by
microstructure. Actual fatigue striation features will be
hard resolve and may only be observed using high reso-
lution field emission gun (FEG) SEM equipment. Lower Fig. 7 Stage I fatigue facets at fatigue origin. These SEM frac-
alternating stresses responsible for HCF may not be sig- tographs show fatigue origin of the cast nickel-based superalloy
nificantly different close to the origin and well away during turbine blade in Fig. 6. The first image (a) shows the faceted Stage I
crack propagation. In this case, similar feathery fracture fatigue origin region consisting of multiple facets (F) which
transitioned to a Stage II fatigue fracture. Note Stage I fatigue is
morphology may be observed over most of the fracture, typically angled at *45° to the stress direction whereas Stage II
highlighting the vibration driven nature of HCF cracking. fatigue is perpendicular to stress direction. Arrows show the
HCF type cracks also tend to meander a bit, changing directions of fatigue propagation within in the Stage II region (a)
directions with the vibratory response, especially as the and within the facets (b) initiating at a microscopic subsurface pore
crack becomes larger.
can be correlated with striation spacing measurements [9].
If the striations are resolved in the SEM, the striation
LCF Fractography density data can be plotted as a function of depth. If the
data indicates rapid crack acceleration as the crack gets
Striations features tend to dominate on LCF fractures. bigger (basic fracture mechanics), and the total number of
Cracks tend to be planar, following the high stress direction calculated cycles makes sense for the slow accumulation of
(perpendicular to the stress) through the part. Often, the cycles on the component, this indicates that it is an LCF or
crack propagation rate and number of propagation cycles mission-driven crack. Striation counting has been

123
J Fail. Anal. and Preven.

Fig. 8 Direction of fracture features for HCF in Titanium. Fractograph (a) from the middle of a titanium alloy airfoil fracture shows typical HCF
propagation where the fracture direction is indicated by multiple fan-like features aligned with the propagation direction (arrows) and opening up
in the direction of propagation, but little in the way of striations or beach marks. Fractograph (b) closer to the origin shows that the fracture was
dominated by structure related features

Fig. 9 Fracture features for HCF in titanium weld. These SEM fractographs are from the origin area of a fracture in a titanium alloy tube weld at
the inner diameter surface. The fracture morphology was dominated by fracture features in the direction of propagation. The stress concentration
of the weld and the bending mode promoted multiple origin areas

successful in nickel, aluminum, and titanium alloy rotating equipment, like TEM and FEG SEM with in-lens detectors,
machinery. Figure 10 shows typical striation features from striations can often be seen and counted on HCF fractures.
the fracture on a nickel disk. Similar features exist in The estimated cycles might not be correlated to any
titanium alloy rotating components as shown in Fig. 11. ‘specific’ cycles, but comparison to the number of start and
With tungsten filament scanning electron microscopes, stop cycles for the component is enough to indicate whe-
the ability to resolve fatigue striations is usually sufficient ther HCF is acting. In one instance, 40,000 cycles were
to discriminate between HCF and LCF, with the pre- seen on a crack in a rotating seal component that had been
sumption that either the alternating stresses of HCF were in the field for just over 3000 start stop cycles. Eventually,
small enough to make the striations unresolvable for the analysis was able to identify the vibration mode that was
equipment at hand, or that striations are not being produced responsible for driving the crack.
at the lower delta K levels. With the use of high resolution

123
J Fail. Anal. and Preven.

Fig. 10 Nickel superalloy LCF striation fatigue features. The SEM fractograph from area (a) near the deepest part of the fracture on a nickel-
based superalloy disk shows the coarse striation fracture features consisting mainly of parallel cracking. Propagation direction is indicated by the
arrows. Closer to the origin at area (b), the striations were finer and required higher magnification. At elevated temperatures, oxidation can
interfere with fine striation resolution

Fig. 11 Titanium Alloy LCF Striation Fatigue Features. The SEM fractography from area (a) near the origin of the fracture on a titanium alloy
disk shows fine striation fracture features. Propagation direction is indicated by the arrows. Closer to the final fracture at area (b), the striations
were extremely coarse and required less magnification

Time Factors Additionally the slower propagation results in near-simul-


taneous formation of cracks in a component with multiple
HCF cracks tend to be found as single isolated large cracks stress concentration sites or cracking will occur in different
and can become quite large in a short period of time. The installations at a similar number of cycles. This is the
higher frequencies contribute to fast crack growth and the ‘‘signature’’ of an LCF situation.
large variability of the initiation life leads to a large crack
being found or leading to separation before any other
cracks are formed. Other Considerations
Large numbers of small cracks suggest slow growing
LCF cracks. As mentioned previously, with LCF there is One caution: not all fatigue failure modes have been cov-
less variability in fatigue crack initiation lives. ered in this discussion. Some fatigue modes do not land in

123
J Fail. Anal. and Preven.

the same areas of the plot in Fig. 4, such as high amplitude


fatigue (HAF) which can occur during severe out of bal-
ance events where high alternating stress cycles can be
accumulated at a high rate. Also, there exist other types of
fatigue such as corrosion fatigue, wear-related fatigue,
elevated temperature fatigue, and others that have not been
covered here.

Component Examples of LCF and HCF

Knowing how components are stressed and how they tend


to crack can go a long way to help the investigator identify
the type of fatigue. Several examples from typical com-
ponents from rotating equipment are listed below.

LCF Disk

As discussed earlier, rotating parts like fan, compressor,


and turbine disks are highly stressed, so it is normal to
expect LCF for such large rotating components. One
incident in 1989 was the in-flight separation of a fan disk
that was analyzed by the National Transportation Safety
Board [10] and has been the subject of documentaries and,
more recently, a book [11]. This event is commonly
referred to as the Sioux City fan disk burst. The titanium
alloy fan disk in the tail engine of a DC-10 aircraft sepa- Fig. 12 Titanium alloy fan disk fracture. The 2.1 m (7 ft) diameter
rated at cruise altitude. Fragments of the disk and fan fan rotor recovered from a cornfield in 1989 (a) exhibited three disk
blades exited the engine, incapacitating all of the hydraulic origin areas. Primary origin O1 shown in the higher magnification
view (b) exhibited a thumbnail shaped fatigue area as outlined by the
systems. The aircraft flew on for 44 minutes with the pilots
dashed line where striations were measured. For scale, the bore was
using differential control of the left and right wing engines. approximately 76 mm (3 inches) wide. Images from publicly
Unfortunately, the aircraft broke up as the pilots attempted available NTSB report [10]
to land at Sioux Gateway Airport, leading to multiple
fatalities.
Major sections of the fan disk were recovered from Iowa nitrogen stabilized hard alpha melt-related defect that was
cornfields months later during harvest; the largest of which present in disk from initial manufacturing.
is shown in Fig. 12. The final fatigue crack size was Additionally, the striation data and fracture mechanics
approximately 33.0 mm (1.3 inch) wide and 14.2 mm (0.56 was useful in determining the size of the crack at the final
inch) deep as the crack went critical. The fracture surface shop inspection, approximately 760 cycles before this
exhibited a single origin area, mainly because there was a flight. Proper identification of this fracture as LCF was key
material defect and associated cavity at the bore surface, as in developing a field plan to inspect and remove similar
shown in Fig. 13. Measurements were made from scanning disks from the fleet. This event led to a consortium of
electron microscope fractographs showing striations, manufacturers and users of titanium disks called the AIA
including those shown in Fig. 14.. Typical of LCF frac- Rotor Integrity Subcommittee, which has greatly decreased
tures, the striations were coarser near the end of the the likelihood of such a defect recurring in engines [12].
fracture and finer near the origin. The data was used to
determine the fatigue cycles on the fracture, and the total HCF of Shafts
was estimated at 15,000 [11]. It turns out that the disk had
been in service for approximately 18 years and had accu- Shafts see one stress cycle of bending per rotation, which
mulated 15,503 flight cycles. The reason this disk had been usually classifies them in the HCF regime. Some shafts can
propagating since early in its life was the presence of a see large bending moments, however, in addition to the
defect at the bore of the disk at the cavity feature. Sec- primary stress in torsion. The torsion portion of the stress
tioning through the origin confirmed the presence of a causes the fracture plane to spiral as shown in Fig. 15,

123
J Fail. Anal. and Preven.

Fig. 13 Fatigue origin area of titanium alloy fan disk. The upper photograph (a) shows the fracture surface with an origin area at a cavity marked
C on the bore surface with arrows showing fatigue propagation direction from this area and a ridge R (ratchet mark) near the origin indicating
cracking started on at least two different planes. The lower photograph (b) shows the corresponding bore surface with the same cavity C
indicated. The rumpled surface texture was from shot peening. Images from publicly available NTSB report [10]

where one arrow represents rotation direction and the other machine. Certain speeds that excite these modes can be
represents the resistance of the load. passed through quickly (transient). Additionally, mean
stresses must be kept low to remain on the proper side of
HCF Gear the Goodman diagram, which makes it highly unlikely that
LCF would be the cause for these components to fail in the
Gears fail most often due to either tooth bending fatigue or absence of higher temperatures, where creep effects may
surface contact fatigue (spalling). Gears tend to be hard- become dominant.
ened steel, so their fatigue fracture features near the origin Conditions outside of what is normally considered in the
are usually smeared, and do not generally display mea- design can lead to HCF cracks, for example, miss-rigged
surable striations. Since the stresses that propagate these variable vanes in compressors or inlet blockages that can
cracks occur once every revolution, the cracking rate would produce a 1/rev excitation. Cast turbine blades experienc-
put gear tooth failures in the high cycle fatigue regime. ing HCF conditions can result in faceted stage I fatigue
Most gear tooth failures are due to the bending stresses origins as shown in Figs. 6 and 17.
alone, and a gear tooth generally fractures due to a crack Foreign object damage (FOD), domestic object damage
near the tooth root or from poorly machined areas on the (DOD), and erosion and corrosion pitting result in notches
gear that can lead to stress concentrations such as burrs or that act as stress concentrations, providing sites for HCF
sharp edges. Occasionally gears will fail away from the cracking. If they are large enough, such pre-cracks and
predicted location, as shown in Fig. 16. Here, HCF of a notches can result in cracks that initiate quickly and skip
steel gear initiated sub-surface from an inclusion stringer, the normal initiation process and likely will not contain
resulting in premature failure. facets at the origin. Erosion and pitting can be eliminated
by proper material selection or coatings.
HCF Compressor and Turbine Blades High temperatures in turbine blades and surrounding
static structures, particularly those under constraint, lead to
Airfoils (both blades and vanes), like those in rotating thermal strains that can initially yield the material in
equipment such as compressors and turbines, tend to be compression. Upon cooling, the material attempts to return
long, thin parts that can vibrate at multiple frequencies in to its original position resulting in a tensile stress. This
various modes (resonance). There is a great difference in repetitive heating and cooling can lead to LCF-type
the ease with which the various modes can be excited [13]. cracking known as thermal fatigue or in the presence of
Great care is taken to tune out these vibrations and make additional mechanical loading, thermal-mechanical fatigue
HCF less likely, but sometimes unavoidable vibration (TMF). It must be noted that these thermal fatigue cracks
modes exist within the normal operating range of the can act as pre-cracks for HCF cracking and one can often

123
J Fail. Anal. and Preven.

Fig. 14 SEM fractographs of fatigue striations from titanium fan disk. Typical course fatigue striations from deep in the crack near overload are
shown in image (a). Intermediate striations from the mid depth of fatigue cracking are shown in (b). Finely spaced striations from close to the
primary origin site are shown in (c). Arrows show the direction of fatigue crack propagation. Images from publicly available NTSB report [10]

find highly-oxidized, slow-growing thermal fatigue type Refining, chemical, and nuclear plants all experience
cracks at the fatigue origins of turbine blades that eventu- small bore pipe failures. For example, according to the
ally failed in HCF. Electric Power Research Institute [14], approximately 80%
of the fatigue failures in the nuclear power industry have
been associated with high-cycle vibration fatigue of socket-
HCF of Tubing, Small-Bore Pipe and Brackets welded connections in small-bore piping. These fatigue-
induced failures occur at the rate of two or three per plant-
Small-bore socket welded piping can have natural fre- year. With over 100,000 socket-welded joints in a typical
quency vibration modes as well as forced vibration from nuclear generating plant, it is difficult to know which one
the attached equipment. Small-bore piping is usually sup- could fail.
ported, in order to tune out the vibrations. When the Often, HCF is invisible to proven commercial nonde-
vibration source is high and there is not enough support, structive evaluation (NDE) methods during crack initiation
however, cracks usually result immediately adjacent to a and the initial phases of crack growth. Under a constant
weld. Typically, the weld is not the root of the problem but applied stress, cracks grow at accelerating rates, which
merely the location of the highest local stress concentration means cracks extend from a detectable size to a through-
at the end of a long lever. wall crack in a relatively short time. When fatigue cracks

123
J Fail. Anal. and Preven.

Fig. 15 Typical hollow shaft fatigue. The left sketch shows a spiral
crack in a hollow shaft. The fracture at right shows that the fatigue
cracking initiated at the outer surface and beach marks show the crack
size at various times. The curved arrows indicate the direction of
rotation and loading

grow large enough to be visible to NDE, the piping has


likely already leaked.
Figure 18a illustrates the geometry of a typical socket-
welded joint [15]. Fatigue cracks in this type of joint can
initiate at the weld root, then extend through the deposited
weld metal as in this 13 mm (‘-inch) stainless steel, seal
Fig. 16 Gear tooth fatigue fracture. Macroscopic image (a) shows a
water return line, socket weld fitting shown at (b). The next gear fatigue fracture exhibiting sub-surface initiation. Arrows show
most common event is a fatigue crack that initiates in the the propagation direction. The backscatter SEM image (b) shows the
weld metal near the axial weld toe and then extends inclusion stringer in the origin region
through the base metal as shown at (c). Fatigue cracks in
stainless steel typically propagate in a transgranular man- assemblies, bolts hold connecting rods together and in
ner, are unbranched, and are very tight and oxide free. bearing blocks, the bolts hold down shafts; both are
Runs of pipe are typically supported by brackets and the examples of HCF loading.
proper spacing of the brackets is one control measure for Two examples of bolt failures from rotating equipment
these types of cracks. Brackets themselves are also subject are shown in Fig. 19. The presence of multiple ratchet
to the same vibrations and cracking modes that excite the marks in the origin area could cause one to consider LCF,
pipe. Sheet metal parts like brackets can suffer from but the notch from the threads can promote more origins.
reversed bending and exhibit fatigue origins on both sur- Both of these examples exhibited little or no overload
faces of the sheet with an area of overload somewhere in regions, which indicate a low mean stress level. To keep
the middle. the stress below the fatigue endurance limit, bolts are
tightened with a pre-determined torque value (clamp load)
HCF in Bolts based upon the yield strength of the bolt and diameter. This
clamp load, also called pre-load, is a percentage of the bolt
Bolts and studs are used to secure rotating components to yield stress in a cyclic condition. If the cyclic stress does
one another. They are also used at pipe flange joints or to not exceed the pre-load stress, then the mean fatigue stress
join structural items. Although bolts and fasteners may be is reduced, and the probability for a fatigue failure is also
the smallest item in a design, that does not minimize their reduced. However, if the preload is not adequate, then the
importance. Failures of bolts have resulted in fatal acci- bolt could fail in fatigue, which was the case for these two
dents in vehicles and leaking of fluids in process equipment example bolts.
that can ignite and result in a fire or an explosion [16]. For Failure of nuts is far rarer than the bolts they are
the most part, like shafts, bolts see one cycle per revolu- attached to, but a nut can occasionally split due to liquid
tion, which categorizes them as HCF. In reciprocating metal embrittlement (LME) or stress corrosion cracking

123
J Fail. Anal. and Preven.

Fig. 17 Turbine blade attachment fatigue. Cast nickel- based turbine blade, (a) exhibited fatigue fracture in attachment area. The origin area
shown in image (b) and SEM view of slightly sub-surface, stage I faceted origin region (c)

(SCC) or strip out due to cross threading or creep if By combining these fracture features with knowledge of
overheated. In our experience the authors have never seen a how certain components are loaded, an investigator can
fatigue fractured nut, but a loosened nut will contribute to properly identify the failure mode. For example, knowing
loss of pre-load and failure of the bolt or can lead to loss of if a particular component tends to vibrate can alert you to
clamp and fatigue at a shaft joint. the possibility of an HCF cause. Knowing where the
stresses come from and how rapidly cycles are accumu-
lated will determine if there is a chance to identify cracks
Summary with non-destructive testing (NDT).
Rotating components such as shafts, turbines, and gears
In summary, there is a natural division between HCF and are components that can be subjected to either HCF or LCF.
LCF that separates the causes, and hence the method of Even components that are associated with rotating equip-
prevention. High cycle fatigue tends to be caused by higher ment such as supports, bolts, and piping can experience HCF
frequency, lower stress cycles, such as vibrations or gear or LCF conditions. Understanding the design considerations
tooth bending. Low cycle fatigue tends to be caused by the and operating conditions and fracture features can assist in
slowly applied higher stresses that occur with start and stop making the proper fix to avoid repeat failures.
type cycles. Simply labeling a fracture as fatigue does not tell the
Certain fracture features can help you distinguish whole story. The root cause and potential solution to the
between these types of fatigue. Consider features such as cracking problem are dependent on knowing whether the
striations, origin quantity, size of the overload fracture, and crack is the result of HCF or LCF. Educating your failure
more. By paying attention to these macro and microscopic analysis team and customer as to which type of fatigue is
fracture features, as listed in Table 1, you can usually acting can help in identifying the actual root cause and help
determine which failure type is active, HCF or LCF. guide you or the customer to make effective changes.

123
J Fail. Anal. and Preven.

Fig. 18 Socket-welded small


bore pipe. The schematic (a)
shows the geometry of a typical
socket-welded small bore pipe.
Cracks usually initiate within
the weld root as in (b) with the
black arrows pointing out the
crack moving from left to right,
or at the axial weld toe, as in (c)
propagating from the surface
from right to left

123
J Fail. Anal. and Preven.

Fig. 19 Bolt fractures. These photos show the fracture surfaces of bolts from an electric motor (a) and a centrifugal compressor (b). Note the
beach marks and arrows indicating the directions of fatigue crack propagation from primary and secondary origin areas. There was a small
overload area on the electric motor bolt, but no distinct overload area for the compressor bolt

4. R. Stone, Fatigue life estimates using goodman diagrams. SMI


Table 1 Typical fracture features and what they tend to indicate encyclopedia of spring design.
5. N. Dowling, Mean stress effects in stress-life and strain-life
HCF LCF
fatigue, SAE Technical Paper 2004-01-2227.
Faceted origin regions No facets 6. P. Lukas, Low cycle fatigue and elasto-plastic behavior of
materials, Elsevier, 1998.
Single origin Multiple origins 7. N. E. Cherolis, Fatigue in the aerospace industry – is it HCF or
Smooth fracture texture Rougher fracture texture LCF? MS&T 2012, unpublished presentation.
Small overload area Large overload area 8. S.D. Antolovich, A. Saxena, ‘‘Fatigue Failures,’’ Metals Hand-
book, Ninth Edition, Volume 11, p. 111.
Feathery fracture features Striations that coarsen with depth
9. N.E. Cherolis, Fatigue in the aerospace industry: Striations. J.
Single large faster growing crack Multiple slower growing cracks Fail. Anal. Prev 8(3), 255–258 (2008)
Lower alternating stress or Higher alternating stress or 10. NTSB Materials Laboratory Report No. 90-2, NTSB/AAR-
vibration driven mission driven 90/06.
11. L. Gonzalez, Flight 232: A Story of Disaster and Survival (W. W.
Norton & Company, New York, 2014)
12. Guidelines to Minimize Manufacturing Induced Anomalies in
References
Critical Rotating Parts, DOT/FAA/AR-06/3, 2006.
13. S.S. Manson, A.J. Meyer Jr., H.F. Calvert, M.P. Hanson, Factors
1. T. Sakai, Review and prospects for current studies on very high affecting vibration of axial-flow compressor blades. Proceedings
cycle fatigue of metallic materials for machine structure use. of the society for experimental stress analysis, vol. 7, No. 2, 1948.
Fourth international conference on very high cycle fatigue, 2007. 14. Vibration Fatigue Testing of Socket Welds, EPRI Report TR-
2. G. Chai, Subsurface non-defect fatigue crack origin and local 107455.
plasticity exhaustion. 13th international conference on fracture, 15. D.N. Hopkins, D.J. Benac, Investigation of fatigue-induced,
June 16–21, Beijing, China. socket-welded joint failures for small-bore piping used in power
3. V. Kazymyrpvych, Very high cycle fatigue of engineering plants. Pract. Fail. Anal. 1(2), 71–82 (2001)
materials (literature review), Karlstad University Studies, 16. D.J. Benac, Technical brief: avoiding bolt failures. J Fail. Anal.
2009:22. Prev. 7(2), 79–80 (2007)

123

You might also like