Professional Documents
Culture Documents
CHAPTER 14
Fatigue
FATIGUE FAILURES OCCUR due to the of the data in the literature is for fully reversed
application of fluctuating stresses that are much bending with no mean stress applied on top of it.
lower than the stress required to cause failure Another common stress cycle is the repeated
during a single application of stress. It has been stress cycle, in which there is a mean stress (sm)
estimated that fatigue contributes to approxi- applied on top of the maximum and minimum
mately 90% of all mechanical service failures.
Fatigue is a problem that can affect any part or
component that moves. Automobiles on roads,
aircraft wings and fuselages, ships at sea, nu-
clear reactors, jet engines, and land-based tur-
bines are all subject to fatigue failures. Fatigue
was initially recognized as a problem in the early
1800s when investigators in Europe observed
that bridge and railroad components were
cracking when subjected to repeated loading. As
the century progressed and the use of metals
expanded with the increasing use of machines,
more and more failures of components subjected
to repeated loads were recorded. Today, struc-
tural fatigue has assumed an even greater
importance as a result of the ever-increasing use
of high-strength materials and the desire for
higher performance from these materials.
stresses. The middle graph in Fig. 14.1 shows the Two ratios frequently used in presenting
condition where both stresses (cyclic and ap- fatigue data are:
plied) are tensile (greater than zero), but it is also
smin
possible to test with both stresses in compres- Stress ratio R= (Eq 14.4)
sion. In addition, the maximum and minimum smax
stresses in the cycle do not necessarily have to be
equal in value. The last type of loading cycle is sa 17R
Amplitude ratio A= = (Eq 14.5)
the random or irregular stress cycle, in which the sm 1+R
part is subjected to random loads during service,
as shown in the bottom graph of Fig. 14.1.
Although a majority of the fatigue data in the
literature is for fully reversed bending, there are
also axial test machines (Fig. 14.3) that are 14.2 High-Cycle Fatigue
capable of tension and compression loading in
both the high- and low-cycle fatigue ranges. High-cycle fatigue involves a large number of
These modern test frames are closed-loop cycles (N4105 cycles) and an elastically
servohydraulically controlled and can be pro- applied stress. High-cycle fatigue tests are
grammed with almost any desired fatigue spec- usually carried out for 107 cycles and sometimes
trum. 5 · 108 cycles for nonferrous metals. Although
A fluctuating stress is made up of two com- the applied stress is low enough to be elastic,
ponents: a mean or steady stress, sm, and an plastic deformation can take place at the crack
alternating or variable stress, sa. The stress tip. High-cycle fatigue data are usually pre-
range, sr, is the difference between the max- sented as a plot of stress, S, versus the number of
imum and minimum stress in a cycle: cycles to failure, N. A log scale is used for the
number of cycles. The value of stress, S, can be
sr =smax 7smin (Eq 14.1) the maximum stress, smax, the minimum stress,
smin, or the stress amplitude, sa. The S-N rela-
The alternating stress is one-half the stress tionship is usually determined for a specified
range: value of the mean stress, sm, or one of the two
ratios, R or A.
sr smax 7smin The fatigue life is the number of cycles to
sa = = (Eq 14.2) failure at a specified stress level, while the fati-
2 2
gue strength (also referred to as the endurance
The mean stress is the algebraic average of the limit) is the stress below which failure does not
maximum and minimum stress in the cycle: occur. As the applied stress level is decreased,
the number of cycles to failure increases.
smax +smin Normally, the fatigue strength increases as the
sm = (Eq 14.3)
2 static tensile strength increases. For example,
Fig. 14.2 Schematic of R.R. Moore reversed-bending fatigue machine. Source: Ref 1
Name ///sr-nova/Dclabs_wip/Metallurgy/5224_243-264.pdf/Chap_14/ 2/6/2008 9:57AM Plate # 0 pg 245
high-strength steels heat treated to over then the scatter becomes too great to be reliable
1400 MPa (200 ksi) yield strengths have much (Fig. 14.5). This does not necessarily mean it is
higher fatigue strengths than aluminum alloys wise to use as high a strength steel as possible
with 480 MPa (70 ksi) yield strengths. A com- to maximize fatigue life because, as the tensile
parison of the S-N curves for steel and aluminum strength increases, the fracture toughness de-
is shown in Fig. 14.4. Note that steel not only has creases and the environmental sensitivity
a higher fatigue strength than aluminum, but it increases. The endurance limit of high-strength
also has an endurance limit. Below a certain steels is extremely sensitive to surface
stress level, the steel alloy will never fail due condition, residual-stress state, and the presence
to cyclic loading alone. On the other hand, of inclusions that act as stress concentrations.
aluminum does not have a true endurance limit. Fatigue cracking can occur quite early in the
It will always fail if tested to a sufficient number service life of the component by the formation of
of cycles. Therefore, the fatigue strength of a small crack, generally at some point on the
aluminum is usually reported as the stress level it external surface. The crack then propagates
can survive at a large total number of cycles, slowly through the material in a direction
usually 5 · 108 cycles. It should be noted that roughly perpendicular to the main tensile axis
there is a considerable amount of scatter in (Fig. 14.6). Ultimately, the cross-sectional area
fatigue test results. It is therefore important to of the member is reduced to the point that it can
test a sufficient number of specimens to obtain no longer carry the load, and the member fails in
statistically meaningful results. tension. The fracture surface of a fatigued high-
For a large number of steels, there is a direct strength part is shown in Fig. 14.7. The portion
correlation between tensile strength and fatigue of the fracture surface due to fatigue crack
strength; higher-tensile-strength steels have growth and the portion finally cracked due to
higher endurance limits. The endurance limit is overload are clearly evident.
normally in the range of 0.35 to 0.60 of the As previously mentioned, much of the fatigue
tensile strength. This relationship holds up data in the literature has been determined for
to a hardness of approximately 40 HRC completely reversed bending with sm = 0. How-
(~1200 MPa, or 180 ksi tensile strength), and ever, the effects of mean stress are important,
Alignment fixture
Load cell
Top grip
Heating coil
Assembly
Heat shield
Specimen Extensometer
Bottom grip
and an increase in mean stress will always cause usually fall closer to the Gerber parabolic curve;
a reduction in the fatigue life (Fig. 14.8). A however, because of the scatter in fatigue data
number of mathematical models have been and the fact that notched data fall closer to the
developed that allow the effects of mean stress Goodman line, the more conservative Goodman
on stress amplitude to be predicted from fully relationship is often used in practice. If the
reversed-bending data. Goodman developed a component design is based on yield rather than
linear model, while Gerber used a parabolic ultimate strength, as most are, then the even
model (Fig. 14.9). Test data for ductile metals more conservative Soderberg relationship can
be used. Mathematically, the three relationships
can be expressed by:
80 x
sm
sa =se 17 (Eq 14.6)
500 su
1045 Steel Endurance Stress Amplitude (MPa)
Stress Amplitude (ksi)
60 Limit
400
where x = 1 for the Goodman line, x = 2 for the
300
Gerber curve, su = sy for the Soderberg curve,
40
No Endurance
and se is the fatigue limit for completely
Limit 200 reversed bending.
2024-T6 Al
20
100
0 0
103 104 105 106 107 108 109 1010 14.3 Low-Cycle Fatigue
Number of Cycles
During cyclic loading within the elastic
Fig. 14.4 Comparison of steel and aluminum fatigue beha- regime, stress and strain are directly related
vior. Source: Ref 2 through the elastic modulus. However, for cyclic
140
130 900
All test results from
one laboratory
All specimens 0 to 2
micro-in. finish
120
800
110
Fatigue limit, (MPa)
Fatigue limit, (ksi)
700
100
90
600
80
Quenched and tempered steels
1054 4063 5140
2340 4068 5150 500
70 4032 4130 5160
4042 4140 8640
4053 4340 9262
60
400
20 25 30 35 40 45 50 55 60 65 70
Rockwell C hardness
Fig. 14.5 Endurance limit versus hardness for steels. Source: Ref 3
Name ///sr-nova/Dclabs_wip/Metallurgy/5224_243-264.pdf/Chap_14/ 2/6/2008 9:57AM Plate # 0 pg 247
loading that produces plastic strains, the re- De, consists of both elastic and plastic compo-
sponses are more complex and form a hysteresis nents:
loop (Fig. 14.10). From point O up to point A,
the component is in tension. On unloading, De=Dee +Dep (Eq 14.7)
the strain response of the specimen follows
the curve from A to D. At D, the component where Dee is the elastic strain and Dee = Ds/E,
is under no stress. As the component is subjected and Dep is the plastic strain and Dep is the width
to compressive stress, the strain response of the loop at its center, that is, the distance CD
follows the curve from D to B. Releasing the in Fig. 14.10. The area of the hysteresis loop is
compressive stress from B and reapplying equal to the work done or the energy loss per
tensile stress, the component stress-strain con- cycle.
dition returns to point A along the curve de- In cyclic strain-controlled fatigue, the strain
fined by B, C, and A. Points A and B represent the amplitude is held constant during cycling.
cyclic stress and strain limits. The total strain, Since plastic deformation is not completely
σ
σ
σ σ
(σm)
σ
σ σ
σ
σ
Fig. 14.9 Comparison of Goodman, Gerber, and Soderberg models for relating mean stress to stress amplitude
Name ///sr-nova/Dclabs_wip/Metallurgy/5224_243-264.pdf/Chap_14/ 2/6/2008 9:57AM Plate # 0 pg 249
Typically, metals harden if su =sy 1:4 and e0f is approximately equal to the true fracture
soften if su =sy 1:2. strain, ef. 2N is the number of strain reversals to
The reason for hardening or softening is failure, where one cycle is two reversals, and c is
related to the dislocation microstructure of the the fatigue ductility exponent, which usually
metal. When the metal is highly work hardened varies between 0.5 and 0.7. A smaller value
and the dislocation density is high, cyclic strain of c results in longer fatigue lives.
allows the rearrangement of dislocations into The Basquin equation, which describes the
more stable networks, thereby reducing the high-cycle, low-strain regime where the nominal
stress at which plastic deformation occurs. strains are elastic, is:
Conversely, when the initial dislocation density
is low, the cyclic strain increases the dislocation Dee
sa = E=s0f (2N)b (Eq 14.9)
density, increasing the amount of elastic strain 2
and stress on the material.
Low-cycle fatigue test data are often pre- where sa is the alternating stress amplitude,
sented as a plot of the plastic strain range, Dep, Dee =2 is the elastic strain amplitude, E is the
versus cycles to failure, N. When plotted on log- modulus of elasticity, and s0f is the fatigue
log coordinates, a straight line is obtained that is strength coefficient, defined as the stress inter-
described by the Coffin-Manson relation: cept at 2N = 1. s0f is approximately equal to the
true fracture stress, sf . 2N is the number of load
Dep reversals to failure, and b is the fatigue strength
=e0f (2N)c (Eq 14.8) exponent, which varies for most metals between
2
0.05 and 0.12. A smaller b results in a longer
where Dep =2 is the plastic strain amplitude, and fatigue life.
e0f is the fatigue ductility coefficient defined by Combining Basquin’s equation and the
the strain intercept at 2N = 1. For many metals, Coffin-Manson equation gives an equation
Stress A
σa = ∆σ/2
∆σ/2E
C O D
Strain
∆σ/2E
∆σ
∆ε
toward the elastic curve at small strain am- of its life even if the stress is reduced to s2.
plitudes (Fig. 14.12). For high-cycle strain For example, at point E, 50 cycles of 200 are
conditions, ductile metals have the longest lives, consumed. If, when the component reaches the
while at low-cycle strain conditions, strong number of cycles defined by point E, the stress is
metals have the longest lives. reduced to s2, the component will be in the
condition defined by point F, with 100 of its 400
cycles consumed. The same kind of change can
be described for a low-to-high stress traverse.
14.4 Cumulative Damage Cumulative damage during fatigue is often
determined using the Palmgren-Miner rule,
Fatigue tests are often conducted under sim- which assumes that the total life of a part can be
ple conditions, such as constant amplitude and estimated by adding up the percentage of life
constant frequency. However, in real structures, consumed by each stress level:
the loading conditions are rarely simple. Many
structures are subjected to a range of fluctuating j=k
X
n1 n2 n nj
loads, mean stress levels, and variable fre- + + + k =1 or =1
quencies. Thus, it is important to be able to N1 N2 Nk N
j=1 j
predict the life of a component subjected to (Eq 14.12)
variable amplitude loading using data generated
in constant amplitude laboratory tests. Cumula- where n1, n2 nk represent the number of
tive damage theories consider the fatigue pro- cycles at a specific stress level, and N1,
cess to be one of damage accumulation until the N2 Nk represent the fatigue life in
life of the component is exhausted. Consider the cycles at the same stress level.
fatigue life diagram in Fig. 14.13. At a constant This rule should be used with caution. It is a
stress, s1, the life is 200 cycles, while at s2, the linear relationship that does not consider the
life is 400 cycles. According to cumulative effects of understressing or overstressing.
damage theory, fatigue life is gradually exhau- Understressing, in which relatively low stress
sted. That is, at points A and C, 100% of the life levels are initially applied, can result in longer
is available, while at points B and D, the re- fatigue lives at a higher stress level, due to
spective lives are completely exhausted. If fati- localized strain hardening. On the other hand,
gue damage accumulates in a linear manner, overstressing, in which high stresses are initially
each cycle contributes the same amount of applied, can result in shorter fatigue lives at a
damage at a given stress level. For example, at low stress level, due to damage induced by the
s1, on cycling the metal from A to E, one-fourth higher stress levels.
of the fatigue life is consumed (50 of 200
cycles). If the stress level is reduced to s2, the
percentage of life exhausted at s2 is equivalent
to the percentage of life exhausted at s1. That is,
if one-fourth of the fatigue life is consumed at
s1, the component still will have lost one-fourth
14.5 Fatigue Crack Nucleation direction normal to the applied stress. Crack
and Growth growth proceeds by a continual process of crack
sharpening followed by blunting, as shown in
Fatigue cracks almost always initiate at free the Fig. 14.15 illustration. Crack propagation
surfaces, usually external surfaces but also during crack growth often produces a pattern of
internal surfaces if the metal contains defects fatigue striations (Fig. 14.16), with each striation
such as voids and cracked second-phase parti- representing one cycle of fatigue. Although
cles. Common external surface defects include striations are indicative of fatigue, fatigue fail-
geometric notches and surface roughness. ures can occur without the formation of stria-
Fatigue crack nucleation and growth occurs tions. Striations are microstructural details that
in the following stages. are best examined with a scanning electron
Stage I. Crack initiation usually starts at microscope and are not visible to the naked eye.
a notch or other surface discontinuity. Even in Frequently, visible examination of a fatigued
the absence of a surface defect, crack initiation surface will reveal a series of concentric mark-
will eventually occur due to the formation of ings on the surface, referred to as beach marks
persistent slip bands (PSBs), so called because (Fig. 14.6). These are present as a result of stress
traces of the bands persist even when the surface changes during fatigue, for example, the starting
damage is polished away. Slip bands are a result and stopping of a rotating shaft. Each of the
of the systematic buildup of fine slip movements beach marks can contain thousands or even tens
on the order of only 1 nm. However, the plastic of thousands of fatigue cycles.
strain within the PSB can be as much as 100 Stage III. Ultimate failure occurs when the
times greater than that in the surrounding fatigue crack becomes long enough that the
material. The back-and-forth movement of the remaining cross section can no longer support
slip bands leads to the formation of intrusions the applied load.
and extrusions at the surface, eventually leading
to the formation of a crack (Fig. 14.14). The
initial crack propagates parallel to the slip bands. 14.6 Fatigue Crack Propagation
The crack propagation rate during stage I is very
low, on the order of 1 nm per cycle, and produces Linear elastic fracture mechanics assumes
a practically featureless fracture surface. The that all structures contain flaws. Cracks grow
crack initially follows the slip bands at from an initial size, ao, to a critical size, ac, cor-
approximately 45 to the principal stress direc- responding to failure as a function of the number
tion. When the crack length becomes sufficient of load cycles (Fig. 14.17). The crack growth
for the stress field at the tip to become dominant, rate, da/dn, can be determined from the slope of
the overall crack plane changes and becomes the curve. Initially, the crack growth rate is slow
perpendicular to the principal stress, and the but increases with increasing crack length. Of
crack enters stage II. course, the crack growth rate is also higher for
Stage II crack growth occurs when the stage I higher applied stresses. If one can characterize
crack changes direction and propagates in a the crack growth, it is then possible to estimate
Fig. 14.14 Development of extrusions and intrusions during fatigue. Source: Ref 4
Name ///sr-nova/Dclabs_wip/Metallurgy/5224_243-264.pdf/Chap_14/ 2/6/2008 9:57AM Plate # 0 pg 253
Striation
1
4
2 5
3 6
where Y depends on the specific specimen geo- 7075-T651 aluminum plate. Since the Paris
metry. Thus, the Paris law becomes: equation does not account for the load ratio,
there are other expressions that do account for
da pffiffiffiffiffiffi the sensitivity of crack growth rate to the load
=C(YDs pa)m (Eq 14.15)
dn ratio. One of the most utilized is the one devel-
oped by Foreman and his associates:
One of the goals of fatigue analysis is to be
able to predict the fatigue life of structures. The da C(DK)m
fatigue life, n, can be solved for by rearranging = (Eq 14.19)
dn (17R)Kc 7DK
Eq 14.13:
In these expressions, it is necessary to deter-
da mine the initial crack length, ao, and the final
dn= (Eq 14.16)
C(DK)m or critical crack length, ac. Crack lengths can
be detected using a number of nondestructive
which may then be integrated to give: testing techniques. If no cracks are detected, it
must be assumed that a crack exists below the
nðf aðc
da resolution of the detection system being used. A
nf = dn= (Eq 14.17) subcritical crack will eventually grow to a length
C(DK)m
0 ao at which the metal immediately fails, that is:
Fatigue Life
ac
ac
Crack Length a
ao da
dn
ao
nf
Inspection Retirement
Number of Cycles
Note that even if a component contains a a longer fatigue life than would normally be
detectable crack, it can remain in service, expected. Crack closure mechanisms effectively
provided that it is periodically inspected. This reduce the DK during fatigue cycling, thus
philosophy forms the basis for what is known as retarding the crack growth. Since DK is equal to
the damage-tolerance design approach. Kmax Kmin and the actual DK is reduced be-
Finally, in region III, the crack growth rate cause crack closure is hindered, DK is effectively
accelerates, since the fracture toughness of the smaller, and crack growth rate is then reduced.
material is approached, and there is a local ten- Four mechanisms that can cause crack closure
sile overload failure. are shown in Fig. 14.20. Plasticity-induced
closure results from compressive residual stres-
ses developing in the plastic wake. Roughness-
14.7 Crack Closure induced closure is due to crack meandering, in
which the crack deviates from mode I dis-
During fatigue cycling, the crack growth rate placements and is subject to mode II shear dis-
can slow due to crack closure. For example, if the placements. These displacements cause a
structure is subjected to a series of high-tensile mismatch between the upper and lower crack
overloads that are followed by loads at lower faces, which results in contact between the crack
tensile stress levels, the crack may temporarily faces during cycling. Oxide debris-induced
stop propagating, leading to a decrease in the closure results from corrosion products becom-
crack growth rate, da/dn, and consequently have ing wedged in the crack interface. Finally, fluid
10–2 3×10–2
R = σmin/σmax
10–3 3×10–3
R = 0.1
da/dN (cm/cycle)
da/dN (in./cycle)
R = +0.5
10–4 3×10–4
10–5 3×10–5
R = –0.5
10–6 3×10–6
1 10 (12) 100 (120)
pressure can act as a wedge, preventing total fatigue life of normalized 4340 steel sheet con-
crack closure. taining different types of stress concentrations is
shown in Fig. 14.21.
The effect of notches on fatigue strength is
14.8 Geometrical Stress Concentrations determined by comparing the S-N curves of
notched and unnotched specimens. The data for
In most structures, fatigue cracking usually the notched specimens are usually plotted in
initiates at a stress concentration. The stress terms of nominal stress based on the net cross
concentration may by inherent in the design, section of the specimen. The effect of the notch
such as a fillet, hole, thread, or other geometrical in decreasing the fatigue strength is reported as
feature, or the stress concentration can result the fatigue strength reduction factor, or the
from a manufacturing process, such as a rough fatigue notch factor, Kf:
surface finish or residual tensile stresses intro-
duced by heat treatment. Fatigue limit unnotched
Kf = (Eq 14.23)
The effect of geometrical stress concentra- Fatigue limit notched
tions on fatigue is often studied by testing not-
ched specimens. When a notch is present in a For metals that do not have a definite fatigue
specimen under uniaxial loading, three effects limit, the fatigue notch factor is based on the
are present: (1) there is an increase or con- fatigue strength at a specified number of cycles.
centration of stress at the root of the notch, (2) Values of the fatigue notch factor vary with the
there is a stress gradient from the notch toward severity of the notch, the type of notch, the
the center of the specimen, and (3) a triaxial state material, the type of loading, and the applied
of stress exists. The dramatic reduction in stress level.
Mode II Displacement
Fluid
Corrosion Product
Notched fatigue data are also reported using a on the fatigue strength. Since the early days
notch sensitivity factor, q: of fatigue investigations, it was recognized
that the fatigue life of a component is very
Kf 71 dependent on the surface finish produced by
q= (Eq 14.24) machining or grinding operations. As shown
Kt 71
in Fig. 14.23, highly polished steel specimens
This relationship compares the theoretical perform much better in fatigue than even care-
stress-concentration factor, Kt, to the fatigue fully machined surfaces. Further degradation
notch factor, Kf. In this relationship, a material in fatigue strength occurs for hot rolled and
that experiences no reduction in fatigue due to a forged surfaces. Finally, when corrosion is
notch will have a notch sensitivity factor of introduced, the fatigue life can be seriously
q = 0, while one that experiences a reduction in degraded.
fatigue up to the full theoretical value will have a Parts that are formed at room temperature will
notch sensitivity factor of q = 1. The value of q contain residual stresses. For example, the sur-
is dependent on the material and the radius face of a part that was formed in tension will
of the notch root, as illustrated by the plots contain residual compressive stresses, and a
shown in Fig. 14.22 for steels. The notch sensi- surface that was in compression during forming
tivity factor, q, significantly decreases with will contain residual tensile stresses. Since fati-
smaller notch radii. Although this seems coun- gue always occurs under tensile loading, the
terintuitive, it occurs because Kf increases more surface with the residual tensile stresses will be
slowly than Kt with decreasing notch radius. the most prone to fatigue cracking. Similar to
For notches with large radii, Kf is almost equal forming, some quenching operations during heat
to Kt. For materials with small notches, Kf is treatment can result in a tensile residual-stress
less than Kt. In addition, lower-strength metals pattern on the surface that will adversely affect
are less affected than high-strength metals by fatigue strength.
geometric discontinuities that reduce the fatigue The fatigue data shown in Table 14.1 for 4340
resistance, because high-strength metals have a steel forgings tested in fully reversed bending
limited capacity for deformation and crack tip illustrate two important points. First, vacuum
blunting. melting, which removes inclusions, improves
both the longitudinal and transverse endurance
limits. Second, since forging tends to elongate
inclusions in the longitudinal or working direc-
14.9 Manufacturing Stress tion, the fatigue endurance limit is higher in the
Concentrations longitudinal direction. The lower transverse
endurance limit in steels containing inclusions is
Because almost all fatigue failures start on a attributed to stress concentrations at the inclu-
surface, the surface finish and residual-stress sions, which can be quite high when an elon-
state near the surface can have a profound effect gated stringer is oriented transverse to the tensile
stress. For the case of the vacuum-melted steel,
4340 Steel Normalized the transverse limit is still somewhat lower than
80
500
Stress Amplitude (MPa)
Stress Amplitude (ksi)
60 400
Unnotched
300
40
200
Hole, Kt = 2
20 Fillet, Kt = 4
Edge Notch, Kt = 4
100
0
104 105 106 107 108 109
Cycles to Failure
Fig. 14.21 Effect of geometrical stress concentrations on Fig. 14.22 Notch sensitivity versus notch radius for steels.
fatigue life. Source: Ref 8 Source: Ref 9
Name ///sr-nova/Dclabs_wip/Metallurgy/5224_243-264.pdf/Chap_14/ 2/6/2008 9:57AM Plate # 0 pg 258
Mirror Polished
0.9
Fine Ground or
Commercially Polished
0.8
0.7 Machined
Surface Retention Factor
0.6
0.4
0.3 As-Forged
Corroded in
0.2 Tap Water
0.1
Corroded in Salt Water
0
60 100 140 180 220 260
Ultimate Tensile Strength (ksi)
Fig. 14.25 Effect of corrosion on fatigue performance Fig. 14.26 Residual-stress state around cold-worked hole
Name ///sr-nova/Dclabs_wip/Metallurgy/5224_243-264.pdf/Chap_14/ 2/6/2008 9:57AM Plate # 0 pg 261
imparts greater fatigue life during high-cycle fastener holes in metallic structures, methods
fatigue than it does for higher-stress low-cycle such as cold working fastener holes and
fatigue. In low-cycle fatigue, large stresses in interference fit fasteners have been developed
the plastic range cause “fading” of the residual- to improve fatigue life. Both cold working and
stress pattern. interference fit fasteners set up residual com-
Steel parts subjected to wear conditions are pressive stress fields in the metal immediately
often carburized or nitrided to harden the surface adjacent to the hole (Fig. 14.26). The applied
for greater wear resistance. Since both of these tension stress during fatigue loading must then
processes also introduce residual compressive overcome the residual compressive stress field
stresses on the surface, there is an improvement before the hole becomes loaded in tension. The
in fatigue life. The greatest improvement occurs fatigue improvement due to cold working in
where a high stress gradient occurs, as in bend- 2024-T851 aluminum is shown in Fig. 14.27.
ing or torsion, with less improvement in axial Cold working of holes is usually conducted
loading. Nitriding is extremely effective in using either the split-sleeve or split-mandrel
improving the fatigue strength of notched spec- method. Both methods involve pulling a man-
imens. Other surface-hardening methods, such drel through the hole that expands the hole dia-
as flame or induction hardening, produce similar meter, creating plastic deformation of material
effects. Case hardening is more effective for around the hole and a resulting residual com-
components that have a stress gradient, as in pressive stress field. The residual-stress field,
bending, or where there is a notch. For axially depending on the material and the amount of
loaded components, fatigue cracks can form expansion, will extend approximately one radius
under the case in the softer, weaker core at the from the edge of the hole. In the split-sleeve
boundary between the surface and bulk. process (Fig. 14.28), a stainless steel split sleeve
Many large structures are assembled with is placed over a tapered mandrel and inserted
mechanical fasteners. In the aerospace industry, into the hole. The hole is cold worked when the
fatigue life improvement of aluminum structures largest part of the mandrel is drawn back through
is often accomplished with cold working of the sleeve. After cold working, the sleeve is
fastener holes and/or installing interference fit removed and discarded. In the split-mandrel
fasteners. Since fatigue cracks often initiate at process, a collapsible mandrel is placed in the
2024-T851 Al
45
Metal With Cold Worked 300
No Hole Open Hole
40
30
200
Reamed
25 Open Hole
150
Tight Fit Fastener in
20 Reamed Hole
Fig. 14.27 Fatigue life improvement with cold working. Source: Ref 13
Name ///sr-nova/Dclabs_wip/Metallurgy/5224_243-264.pdf/Chap_14/ 2/6/2008 9:57AM Plate # 0 pg 262
hole, and as the mandrel is withdrawn, it argon pressure at 105 MPa (15 ksi) and tem-
expands to cold work the hole. peratures in the range of 480 to 525 C (900 to
Interference fit fasteners can also be used in 980 F). The use of HIP usually results in
metallic structures to improve fatigue life. When improved mechanical properties, especially
an interference fit fastener is installed in metal, it fatigue strength, but, of course, it adds to the cost
also plastically deforms a small zone around the and cycle time. The improvement in fatigue life
hole, setting up a compressive stress field, which for the aluminum casting alloy A201.0-T7 as a
again is beneficial when fatigue loading is pri- result of HIP is shown in Fig. 14.29.
marily tensile. The amount of interference can
vary, depending on structural requirements, but
it is usually in the range of 0.07 to 0.10 mm 14.12 Fatigue Design Methodologies
(0.003 to 0.004 in.). In some highly loaded holes,
both cold working and interference fit fasteners Since the 1800s, a number of design philo-
are used. While cold working and the use of sophies or methodologies have evolved to deal
interference fit fasteners are proven methods of with design against fatigue failures.
improving fatigue resistance, both increase Infinite-Life Design. This is the oldest of the
assembly costs and should only be specified design philosophies and is based on maintaining
when they are really needed. the stresses at some fraction below the fatigue
Castings can be hot isostatic pressed (HIP) to strength of the metal. The initial material is
help reduce internal porosity. The HIP of alu- assumed to be free of flaws. This methodology is
minum castings is usually conducted using based on the classic S-N curve. It is most
Fig. 14.28 Split-sleeve cold working process. Reprinted with permission from SAE Paper # 982145 # 1998 SAE International.
Source: Ref 14
Name ///sr-nova/Dclabs_wip/Metallurgy/5224_243-264.pdf/Chap_14/ 2/6/2008 9:57AM Plate # 0 pg 263
Fig. 14.29 Effect of hot isostatic pressing (HIP) on fatigue life of A201.0-T7 aluminum casting. Source: Ref 15
applicable when the component is only stressed incorporate multiple load paths and crack
in the elastic range and displays a distinct stoppers in the structure. In other words, if a
endurance limit, such as for steels. Although it primary load path fails, the load will be picked
has largely been superseded by later methodol- up by an alternate load path to prevent the struc-
ogies, it is simple to use and can be useful ture from failing. A major part of this metho-
where periodic inspection is either impractical dology is rigid certification criteria along with
or not warranted. However, since it is a con- the capability to detect and inspect cracks.
servative methodology, it usually results in a Damage-tolerant design is the latest design
heavier design than some of the later meth- methodology and is an extension of fail-safe
odologies. design. The refinement of the damage-tolerant
Safe-life design is based on the assumption methodology is the assumption that structures
that the part is initially flaw-free and has a fine do contain cracks and that fracture mechanics
life in which to develop a critical crack size. Like approaches can be used to determine crack
the infinite-life methodology, safe-life assumes growth rates. If the crack growth rate can be cal-
an initial flaw-free component. This methodol- culated, it is then possible to inspect the structure
ogy was developed to account for parts that were at various intervals and either repair or retire it
subjected to higher loads that produced plastic prior to the crack reaching critical dimensions.
strains. Under these conditions, the description Although much progress has been made in the
of local events in terms of strain made more design against fatigue failures, they still occur
sense and resulted in the development of with disturbing frequency. For any design, it is
assessment techniques that used strain as a imperative that part or component testing be
determining quantity, that is, log strain (e) ver- conducted prior to placing the part in service. In
sus log number of cycles (N). The failure cri- addition, it is important that the test conditions
terion is usually the detection of a small crack or are representative of the actual environment that
some equivalent measure related to a substantial the component will experience in service.
change in load-deflection response, although
failure may also be defined as fracture.
Fail-Safe Design. The fail-safe design phi-
losophy assumes that fatigue cracks will REFERENCES
be detected and repaired before they lead to
failure. This methodology was developed in 1. Fatigue Failures, Failure Analysis and
the aircraft industry where large safety margins Prevention, Vol 11, ASM Handbook, ASM
were weight-prohibitive. Fail-safe designs International, 2002
Name ///sr-nova/Dclabs_wip/Metallurgy/5224_243-264.pdf/Chap_14/ 2/6/2008 9:57AM Plate # 0 pg 264
Terms of Use. This publication is being made available in PDF format as a benefit to members and
customers of ASM International. You may download and print a copy of this publication for your
personal use only. Other use and distribution is prohibited without the express written permission of
ASM International.
No warranties, express or implied, including, without limitation, warranties of merchantability or
fitness for a particular purpose, are given in connection with this publication. Although this
information is believed to be accurate by ASM, ASM cannot guarantee that favorable results will be
obtained from the use of this publication alone. This publication is intended for use by persons having
technical skill, at their sole discretion and risk. Since the conditions of product or material use are
outside of ASM's control, ASM assumes no liability or obligation in connection with any use of this
information. As with any material, evaluation of the material under end-use conditions prior to
specification is essential. Therefore, specific testing under actual conditions is recommended.
Nothing contained in this publication shall be construed as a grant of any right of manufacture, sale,
use, or reproduction, in connection with any method, process, apparatus, product, composition, or
system, whether or not covered by letters patent, copyright, or trademark, and nothing contained in this
publication shall be construed as a defense against any alleged infringement of letters patent,
copyright, or trademark, or as a defense against liability for such infringement.