Você está na página 1de 14

7/27/2019 Fatigue Life Prediction - an overview | ScienceDirect Topics

Journals & Books Create account Sign in

Fatigue Life Prediction

Related terms:
Fatigue Life, fatigue crack growth, S-N Curve, stress-intensity factor, Fatigue Behaviour, Laminate

View all Topics

Fatigue life prediction of wind turbine blade composite materials


A.P. Vassilopoulos, in Advances in Wind Turbine Blade Design and Materials, 2013

Abstract:
Fatigue life prediction of wind turbine rotor blades is a very challenging task, as blade failure is led by different failure types that act
synergistically. Inherent defects like wrinkles, fiber misalignments and voids, that can be introduced during fabrication, can
constitute potential damage initiation points and rapidly develop to failure mechanisms like matrix cracking, transverse-ply cracking,
interface cracking, debonding, fiber breakage, etc. Different methods have been established to address this problem, some based on
phenomenological and others on actual damage mechanics modeling. This chapter aims to provide an overview of fatigue life
modeling and prediction methodologies for the composite materials and structural composite elements that compose a wind turbine
rotor blade under complex loading conditions.

Polymer Matrix Composites


Ramesh Talreja, in Comprehensive Composite Materials, 2000

2.14.4.2.3 Other laminates


Fatigue life prediction of laminates other than cross-ply laminates is still not possible primarily because analytical models for stress
analysis of these laminates with damage do not yet exist. In fact the most advanced stress analysis model available is that of
McCartney (2000), which is seminumerical and applies to symmetric balanced or unbalanced laminates of all ply orientations but
admits cracks only in one symmetry direction, e.g., cracks in 90-plies of a (0, ±45, 90)s laminate. It cannot yet treat cracks in off-axis
plies of a (0, ±θ)s laminate. Although this model is justifiably viewed as having significantly advanced the stress analysis capabilities
with respect to previous models, which only applied to cross-ply laminates with transverse cracks, its use for fatigue of laminates is
severely restricted. In fact, fatigue damage of general laminates involves not only cracks in multiple orientations, it also has local
delaminations in different interfacial planes. Such a complexity of damage currently necessitates that either a numerical stress
analysis be conducted or a different approach be taken that does not require determining stress states in damaged plies. The first
option is not very desirable considering that each laminate configuration would require a separate treatment and that the analysis
would quickly become intractable. The second option is possible by an approach called continuum damage mechanics (CDM). This
approach and its potential to provide a methodology for fatigue life prediction are discussed in Talreja (1999). The following will
describe that briefly.
https://www.sciencedirect.com/topics/engineering/fatigue-life-prediction 1/14
7/27/2019 Fatigue Life Prediction - an overview | ScienceDirect Topics
describe that briefly.
(i) CDM approach to life predictionThe three essential steps in a CDM approach are (i) damage characterization, (ii) average property
changes with damage, and (iii) damage evolution. For fatigue life prediction we would then need the critical state of damage or
the critical degradation of a given property. A description of the three steps follows.
(a) Damage characterization. A procedure for damage characterization incorporating the essential physical aspects of damage
and suitable for use in an internal variables based formulation of materials response was used by Talreja (1990, 1994). It
defines a damage entity tensor as
(4)

where ai, placed at a point on the damage entity surface, is a vector representing a selected “influence” of the damage entity
and nj is a unit outward normal to the surface at that point (Figure 34).

Sign in to download full-size image

Figure 34. Characterization of a transverse crack with delamination by two vectors.

This tensor is the vehicle by which specific physical aspects of damage are brought into the CDM theory. A particular way to
accomplish this is by introducing an influence coefficient κ as illustrated in several recent works (Talreja, 1996; Varna et al.,
1999a, 1999b). The essential idea is to assume that all the effects of factors influencing the way an individual damage entity
affects material response at point P can be incorporated in the magnitude of the vector ai. Thus, writing ai = ani and a = κtc,
where tc is a characteristic size of damage entity, provision for the stated effects is made through the influence coefficient κ.
For a point P in the composite with damage, an average effect of all damage entities is represented by a damage mode tensor
Dij defined as

(5)

The volume V over which averaging is done is a representative volume element (RVE).
If multiple damage modes exist, e.g., ply cracking in different orientations and delamination, then Dij for each mode is
obtained by appropriately constructing the corresponding dij for each mode.
(b)

Material response with damage. The tensor components Dij act as internal variables in a thermodynamics based theory of
constitutive behavior.
F th i l f l h i l th f ti i th H l h l f d it itt
https://www.sciencedirect.com/topics/engineering/fatigue-life-prediction 2/14
7/27/2019 Fatigue Life Prediction - an overview | ScienceDirect Topics
For the simple case of purely mechanical response, the response function is the Helmholz free energy density written as a
function of strain and damage

ɛ (6)

The incremental stress can be derived from this function and expressed as

ɛ (7)
where

(8)
ɛ ɛ

and

(9)
ɛ

Thus, for unloading response, where dDij = 0, the stiffness properties are given by Cijkl. For a cross-ply laminate with
damage, illustrated in Figure 34, the stiffness properties, expressed as engineering elastic constants are given by:
(10)

The constants c3, c7, c11, and c13 are the material coefficients obtained by an experimental procedure (cf. Talreja, 1990).
(c) Damage evolution. The earliest way to describe damage evolution was introduced by Kachanov (1958), who also laid the
foundation of a CDM approach. In treating creep of metals he proposed that the rate (or increment) of damage can be
written as a power law function of an “effective” stress, i.e.,

(11)
where σ is a remotely applied stress and D (= ω in Kachanov's work) is the void volume fraction. C and n are empirical
constants to be evaluated from creep tests.

More recent works on damage evolution have generalized Kachanov's notion of the local (“effective”) stress to the energy release
rate as the quantity driving the damage evolution. Clearly, this has been motivated by the works of Griffith and Irwin on fracture
from cracks in materials. Thus, the above equation for damage evolution in fatigue is replaced by
(12)

where G is the “energy release rate,” and A and m are material constants. Note that G here is not the energy release rate as
defined in fracture mechanics for an incremental growth of a crack tip. In a mode of damage consisting of multiple cracks,
evolution is not growth of crack size, but instead, increase in number of cracks. Thus, to relate G to crack-tip stress intensity factors
may be questionable. A thermodynamics based interpretation of energy release rate (or, equivalently, crack extension force) is the
so-called “force conjugate to damage.” Rice (1971) has treated this rigorously for general dissipative phenomena such as
plasticity, creep, and brittle cracking. Ladeveze (1994), among others, has used this notion to express damage evolution in
composites. It is worth noting that the force conjugate to damage is not unique but depends on the chosen damage variable.
In keeping with the CDM formulation a way to treat damage evolution is by considering the damage rate as a response function
much like any other response function, such as the Helmholz free energy density, and express it as a general polynomial
function. Thus, for isothermal response

ɛ (13)
where P is a polynomial function of the variables stated. In Talreja (1990), this formulation was treated for the case of a cross ply
laminate with 90-ply cracks. Experimental verification for that case was also provided. As is to be expected, additional data must
be generated for this formulation to be applied. The alternative would be to carry out a local stress analysis, based on which and a
suitable stress-based failure criterion the damage evolution rate can be determined. As explained above, the local stress analysis is
only possible for cross-ply laminates. Seminumerical and numerical stress analysis methods do not offer attractive alternatives to
the CDM approach.
https://www.sciencedirect.com/topics/engineering/fatigue-life-prediction 3/14
7/27/2019 Fatigue Life Prediction - an overview | ScienceDirect Topics
the CDM approach.
(ii) Fatigue life predictionOnce a damage evolution rate equation has been developed, one can integrate it to obtain the fatigue life if
the critical damage state is known. This procedure is analogous to what is done in the case of metal fatigue where the empirical
crack growth rate equation is used. In composite damage it is not always clear what the critical damage state is, particularly when
damage mode changes occur. The alternative way to proceed is to use a critical amount of degradation of a stiffness property as a
failure event and use the degradation rate for that property. As an example, if critical reduction in the elastic modulus in a
symmetry direction of a given laminate is used as a failure event, then the degradation rate equation for that modulus can be
derived from the stiffness–damage relationships, e.g., Equations (10). It may be noted that for laminates where a significant
proportion of the plies have been placed in the longitudinal direction, e.g., in (04, ±45, 90)s laminate, the longitudinal elastic
modulus may not change significantly even when all off-axis plies have been saturated with cracks. In such a case the longitudinal
Poisson's ratio, which shows greater percentage change, may be used instead.

In applications where the load-bearing capacity of a structure is more critical than the deformational response, a strength-based life
prediction may be more desirable. Methods for this purpose are well developed for metals and are commonly known by the field of
fatigue reliability where the statistical nature of material strength is also incorporated. In extending these methods to composites,
several difficulties are faced. First, while crack length is related to strength in metals, at least in the case of brittle failure, a
relationship between crack density and strength for composites cannot be easily established. What is instead done is to assume that
the stress state in the critical ply that governs failure is altered by crack density in the other plies and, if progressive fiber failure takes
place in the critical ply, then the strength of that ply also degrades. This approach has been taken by Reifsnider (1991). The problems
faced here consist of determining the local stress states in plies of laminates that undergo failure as well as conducting an analysis of
strength degradation due to fiber failure. It may be recalled that the stress analysis problem remains unsolved for general laminates
with damage, as discussed above.

Fatigue and Fracture Analysis


Kuang-Hua Chang, in e-Design, 2015

9.6.3 Rain-Flow Counting


Most traditional fatigue life prediction models require a constant-amplitude loading as input. Constant-amplitude loading is usually
represented by repeating sinusoidal or triangular waveforms. In the case of fatigue crack initiation prediction, the conventional
strain-life relationship is used to relate a constant-amplitude strain loading Δε/2, shown in Figure 9.24, to the fatigue crack initiation
life 2Nf, using, for example, Eq. 9.18.

Sign in to download full-size image


https://www.sciencedirect.com/topics/engineering/fatigue-life-prediction 4/14
7/27/2019 Fatigue Life Prediction - an overview | ScienceDirect Topics

Figure 9.24. Constant-amplitude strain history.

However, components of mechanical systems usually experience a variable-amplitude loading that yields a stress or strain history
with variable amplitude, such as the one shown in Figure 9.23. To predict fatigue life of a component subject to a variable-amplitude
loading history, the variable-amplitude stress or strain history must be converted into several constant-amplitude cycles. The
procedure for this conversion is referred to as the cycle counting procedure. A number of such procedures have been proposed over
the years, but all of them must use some rules to decide when or how a cycle is defined from a variable-amplitude history. A well-
accepted procedure, called rain-flow counting (Downing and Socie, 1982), is discussed next. This procedure attempts to define cycles
that correspond to a closed stress-strain hysteresis loop.
The rain-flow counting algorithm is used in the analysis of fatigue data to reduce a spectrum of varying stress into a set of simple
stress reversals. Its importance is that it allows the application of Miner’s rule to assess the fatigue life of a structure subject to
complex loading. The algorithm was developed by Endo and Matsuiski in 1968; they describe the process in terms of rain falling off a
pagoda roof.
Figure 9.25(a) shows a typical stress history, composed of repetitive blocks. A stress block (stress points A to G) is identified and
rotated 90 deg. clockwise, as shown at the top of Figure 9.25(b). The rotated stress block (or loading history) (points A–G) resembles a
Japanese pagoda. The corresponding stress and strain history is plotted directly below the loading history. In the lower stress-strain
plot, three cycles are easily identified: one large overall cycle (A–D–G), one intermediate cycle in the center of the plot (C–B–C), and
one smaller cycle (E–F–E). Each cycle has its own strain range and mean stress.

Sign in to download full-size image

Figure 9.25. Rain-flow cycle counting: (a) stress history, (b) hysteresis loops, and (c) rain-flow counting.

From a deformation viewpoint, the process proceeds as follows. Start at A, the maximum strain, and load the material to B in
compression. Then reload to point C and compress to D. When the material reaches the strain at point B during the loading from C
to D, the material remembers its prior deformation and deforms along a path from A to D as if event C–B never happened. This is
better illustrated in the next part of the loading. Load from D to E and unload to F. Now load from F to G. When the material reaches
the strain at point E during the loading from F to G, the material remembers its prior deformation and deforms along a path from D
to G as if the event E–F never happened. As a result, rain-flow counting identifies three cycles: A–D–G, B–C–B, and E–F–E.
Several rules are imposed on raindrops falling down these sloping roofs so that the rain flow may be used to define cycles and half-
https://www.sciencedirect.com/topics/engineering/fatigue-life-prediction 5/14
7/27/2019 Fatigue Life Prediction - an overview | ScienceDirect Topics
Several rules are imposed on raindrops falling down these sloping roofs so that the rain flow may be used to define cycles and half
cycles of fluctuating stress in the spectrum. Rain flow is initiated by placing raindrops successively at the inside of each peak
(maximum) and valley (minimum). The rules are as follows:
1. The rain is allowed to fall on the roof and drip down to the next slope. For example, the rain flow begins at point A, drips to point
B, then to point D. After point D, there are no more roofs to drip to, so stress points A–D are counted as a half-cycle, cycle 1.
2. Step 1 is true except that, if the rain drop initiates at a valley (for example, point B shown in Figure 9.25(c)), it must be terminated
when it comes opposite a valley more negative (in this case point D) than the valley from which it initiated. Therefore, stress
points B–C are counted as a half-cycle, cycle 2.
3. Similarly, if the rain flow initiates at a peak, it must be terminated when it comes opposite a peak more positive than the peak
from which it initiated.
4. The rain flow must stop if it meets the rain from a roof above. For example, the rain begins at point C and ends beneath cycle 1,
right under point B. Therefore, stress points C–B are counted as a half-cycle, cycle 3.
5. Half-cycles of identical magnitude (but opposite sense) must be paired up to count the number of complete cycles.

Following the rules, from Figure 9.25(c), stress points D–E–G are counted as a half-cycle (rule 1), cycle 4. Stress points E–F are
counted as half-cycle 5 (rule 1), and points F–E are counted as half-cycle 6, as shown in Figure 9.25(c). Half-cycles 1 and 4 form a
complete cycle (rule 5); hence, the hysteresis loop A–D–G shown in Figure 9.25(b). Similarly, half-cycles 2 and 3 form the second
cycle, the smaller loop B–C in Figure 9.25(b). Finally, half-cycles 5 and 6 form the third cycle, the smallest loop F–G in Figure 9.25(b).

Cyclic Loading and Fatigue


S.W. Case, K.L. Reifsnider, in Comprehensive Structural Integrity, 2003

4.16.5.2 Virtual Design


The application of fatigue life-prediction models to structures has another set of challenges. The approach outlined in Section
4.16.4.5 has been incorporated into FEA codes, and is being applied on an element-by-element basis to predict the remaining
strength and life of high temperature structures, such as combustor liners. There are many challenges to such procedures. When the
degradation and life-controlling processes are nonlinear combinations of time- and cycle-dependent micro-events, the sequence of
internal stepping and adjustments of stiffness and strength values in each element of a grid is significant—and largely unexplored.
Nevertheless, the prospect using a virtual design environment to design the composite material along with the structure, so that
both the material and the structure can be optimized at the same time, before the cost of characterization and testing is incurred, is a
strong motivation. Such a virtual design has the promise of greatly reducing design cost and of accelerating the implementation of
new materials and design concepts. These developments have brought us to the threshold of this opportunity.

Fatigue behavior of spot welded joints in steel sheets


S.K. Khanna, X. Long, in Failure Mechanisms of Advanced Welding Processes, 2010

4.7 Models for fatigue life prediction of spot welded joints


Various analytical and numerical fatigue life prediction models have been proposed in the literature. The analytical methods are the
stress–life method, strain–life method or the fracture mechanics method. The stress-based approaches are based on the assumption
that the elastic stress state controls the fatigue behavior and it is generally an accurate representation of high cycle fatigue, while it is
not accurate for low cycle fatigue applications. These typically involve phenomenological relations between stress amplitude, mean
stress, mechanical properties such as ultimate strength and yield strength, and fully reversed fatigue strength. Several existing
models are the modified Goodman, Gerber, Soderberg, Morrow and ASME-elliptic.54‒56 If the structure is subjected to loading
conditions resulting in multiaxial stress states, equivalent uniaxial alternating and mean stresses can be used in fatigue calculations
for proportional loading cases For non proportional loading cases more sophisticated numerical methods should be used such as
https://www.sciencedirect.com/topics/engineering/fatigue-life-prediction 6/14
7/27/2019 Fatigue Life Prediction - an overview | ScienceDirect Topics
for proportional loading cases. For non-proportional loading cases, more sophisticated numerical methods should be used, such as
the critical plane method.
The strain–life approach considers the plastic deformation that may occur during fatigue loading. This method accounts for localized
yielding, which is often the case in metal components displaying low cycle fatigue (that is a relatively short fatigue life). The stress–life
approach utilizes average or nominal stresses, while the strain–life approach uses local stresses and strains. Some of the common
strain-based approaches are the Coffin–Mason relationship, Morrow’s mean stress approach and the Smith, Watson and Topper
Parameter.55‒57 Neither of these two approaches includes analysis of crack growth, while the fracture mechanics approach does
through formulations such as the Paris law, Walker equation and Forman equation.
The above-mentioned stress–life and strain–life approaches have been applied to spot weld fatigue.4 Ertas et al.4 found better
correlation with experimental data using the strain–life approach compared to the stress–life approach, which was overly
conservative. It is believed that the stress–life approach is not a good representation since it does not account for localized plastic
deformation in the circumferential notch region. Among the strain–life approaches, the Coffin–Mason model captured the fatigue
life trend better than the Morrow mean stress model. This difference could contribute to the fact that the Coffin–Mason approach
does not take into account the mean stress, while the Morrow approach does account for the mean stress. Hence Ertas et al.4
concluded that the mean stress effects in spot weld fatigue, for example caused by residual stress or external loading, are not
important.
Bonnen et al.3 evaluated the performance of five fatigue damage parameters in the context of large fatigue data sets consisting of
hundreds of tests. One parameter by Swellam57 is based on a fracture mechanics approach, while the other four by Rupp et al.,58
Sheppard,59 Dong60 and Kang61 are based on a structural stress approach. One of the evaluation methods used by Bonnen et al.3 is
to estimate the specimen life using the parameter and compare the estimate against the actual fatigue life observed in laboratory
tests. Figure 4.16 shows a plot for each parameter of the observed specimen life versus that predicted by the parameter. All
parameters resulted in a reasonable fit to the experimental data for these specimen geometries, with the Swellam’s parameter
showing the most scatter among them.

https://www.sciencedirect.com/topics/engineering/fatigue-life-prediction 7/14
7/27/2019 Fatigue Life Prediction - an overview | ScienceDirect Topics

Sign in to download full-size image

4.16. Predicted vs. experimental fatigue lives for (a) Rupp et al.,58 (b) Sheppard,54 (c) Dong,60 (d) Kang,61 and (e) Swellam57 spot-weld fatigue parameters.3

Fatigue Design of Components


Anne-Sophie Béranger, ... Jean-François Vittori, in European Structural Integrity Society, 1997

Fatigue life prediction results.


Fatigue life prediction has been performed on the test component. The result of the static linear elastic F.E. calculation is used to
generate the load cycle. This one is + 1 450 daN; 0; -720 daN; 0; + 1 450 daN.
The Dang Van fatigue model was used with a material damage line, whose coefficients are:

They correspond to the shot-peened material, for a life of 107 cycles, with a probability of failure of 50%. Several stress ratios are used
to identify these coefficients. Therefore any loading cycle, with various stress ratios, can be predicted with these coefficients using the
Dan Van model.
The result of the calculation is a value of a so called safety factor (Sf ). This factor can be defined as the normalised smallest distance
of the loading path in the (τ; p) Dang Van diagram to the damage line. When Sf is greater than 1 the component is supposed to be
safe. On the contruary when Sf of some elements become less than unity, failure may occur.
Isovalues of the safety factors are presented on Fig. 7 for the overmentionned fatigue test on the component.

https://www.sciencedirect.com/topics/engineering/fatigue-life-prediction 8/14
7/27/2019 Fatigue Life Prediction - an overview | ScienceDirect Topics

Sign in to download full-size image

Fig. 7. Isovalues of the safety factors for the suspension arm tested in fatigue.

The minimum value of the safety factor on the part is located in the area where the failure occured on the test rig. Its value is 1,02.
This result exhibits a very good correlation between the experiment and the fatigue life prediction since the theoretical value of 1 and
the actual safety factor are very close. Several damage lines representative of the various surface finish have been tested for the
predicition of the safety factors associated to the fatigue life. It was shown that the safety factors obtained are highly dependent on
the material data. The best results were derived from material data representative of the exact surface finish of the part, namely shot-
peened.

Cyclic Loading and Fatigue


D Davidson S Hudak in Comprehensive Structural Integrity 2003
https://www.sciencedirect.com/topics/engineering/fatigue-life-prediction 9/14
7/27/2019 Fatigue Life Prediction - an overview | ScienceDirect Topics
D. Davidson, ... S. Hudak, in Comprehensive Structural Integrity, 2003

4.05.7 Experimental Techniques


Fatigue life prediction that includes small-crack effects often uses baseline data collected from conventional large-crack and smooth
specimens, which are combined using the Kitagawa diagram. In some applications, however, it may be necessary to obtain direct
experimental evidence for small-crack behavior. Unfortunately, small-crack growth rates can be measured with the standard test
procedures developed for large cracks. Small-crack tests usually require different specimen geometries and different specimen
preparation techniques, different crack length measurement techniques and equipment, and different data analysis techniques.
The study of small cracks requires detection of crack initiation and growth while physical crack sizes are extremely small, and this
requirement influences specimen design. The preferred and most widely used specimens promote the initiation of naturally small
surface or corner cracks in rectangular or cylindrical specimens, rather than growing a large crack and then machining away material
in the crack wake to leave a small or short crack. However, it is inherently difficult to find a microscopically small crack in the test
specimen after it has initiated, but while it is still small, in order to monitor its growth. It is easier to find the crack if its probable
location can be somehow restricted. For example, a three-point bend specimen or a tensile specimen with a mild hourglass shape
restricts the initiation site to a narrow band along the region of highest stress. A sharper notch is more successful in localizing the
initiation site, but the notch itself can significantly influence the resulting crack growth behavior.
For materials with few natural defects, some investigators have put microdefects (such as extremely small electro-discharge
machined notches, environmental pits induced by a drop of a corrosive fluid, or ion beam milled holes) into the specimen to serve as
crack starters. However, it can be difficult to manufacture a microdefect that is small enough not to influence the growth behavior of
the resulting small fatigue crack in the size range of interest, while at the same being big enough to cause the fatigue crack to form
at the defect before another crack initiates at some other location in the specimen. Furthermore, any biasing of the crack origin (by
either means) to an “unnatural” location could have some influence on the resulting growth behavior for the smallest flaw sizes if the

local microstructure differs from locations where natural initiation would have occurred. This limitation can be circumvented by
studying a sufficiently large number of cracks that many microstructural variations are sampled.
Specimen fabrication (e.g., machining and polishing the surface) introduces near-surface residual stresses of some magnitude.
Residual stresses that are small enough to have only a minor effect on the behavior of large cracks or the total lifetime of a smooth
specimen may still be large enough to have a pronounced effect on the growth behavior of a small crack. Therefore, the specimen
preparation methods should be carefully chosen and carried out in a manner that will minimize residual stresses (Larsen et al., 1992).
Preparation methods that terminate with low-stress mechanical and then nonmechanical operations (e.g., electropolishing or
chemical polishing) may be preferable. However, since the growth rates of small surface cracks in engineering components can be
influenced by residual stress fields arising from fabrication of the component, the residual stress similitude—or lack thereof—
between test coupon and component residual stresses should be considered when the laboratory data are applied to component life
prediction.
Specimen preparation methods may also influence the visibility of the small crack on the specimen surface, which can be critical for
any visual method of crack length measurement. It is easier to find small cracks visually using polished surfaces, rather than etched,
and the etch may influence the initiation process. The relationship between microstructure and the crack is then determined by
etching after crack growth.
Several well-established experimental techniques are available for measuring the size of small fatigue cracks, and hence deducing
their growth rates. These techniques include replication, optical microscopy, scanning electron microscopy, and potential drop. Each
technique has unique strengths and limitations, and different techniques are optimum for different circumstances. All are useful for
measuring the growth of fatigue cracks with the size of the order of 50 μm and greater, and some are applicable to even smaller
cracks. Detailed descriptions of these and other techniques are available in Larsen and Allison (1992).
Replication (Swain, 1992) is one of the most commonly used techniques. While fatigue cycling is interrupted and a static force is
applied to the specimen, a small piece of thin cellulose acetate sheet is softened with acetone, gently applied to the specimen surface,
and allowed to dry for a few minutes. The acetate replica forms a permanent record of the surface topography, and is subsequently
viewed in an optical or scanning electron microscope to measure surface crack length. Replication is a useful means to record the
growth of small, naturally initiated flaws by working backwards from a larger crack in a chronological series of replicas. However,
replication has been shown to influence crack growth rates artificially in some materials due to chemical interactions, and replication
can also be very labor intensive and time consuming.
Direct microscopic methods, either optical (Larsen et al., 1992) or SEM (Davidson, 1992), permit direct inspection of the crack and
https://www.sciencedirect.com/topics/engineering/fatigue-life-prediction 10/14
7/27/2019 Fatigue Life Prediction - an overview | ScienceDirect Topics
p , p ( , ) ( , ), p p
may facilitate the permanent capture of a photographic record, but require advance knowledge of the crack origin and may be limited
in resolution by the available microscopic equipment. Nonvisual techniques such as potential drop (Gangloff et al., 1992) offer
automation advantages but again require prior knowledge of where the crack will form (since probe location is critical) and may also
be limited in resolution by the available instrumentation. It may be possible to use replication to locate a crack while it is still small,
and then use a high-resolution method such as the SEM to study ensuing growth in more detail, but the resolution of the replica
itself is limited, and usually less than the SEM.
Many small surface cracks develop shapes that are approximately semi-elliptical, and the standard K solutions for these geometries
can be applied during data analysis. However, variations in the crack shape can be a source of inaccurate ΔK calculation, especially for
microstructurally small cracks, and some confirmation of crack shape is desirable. Fractography can often be used to obtain crack
shape as a function of size. Interactions between closely spaced multiple cracks that affect growth rates are more likely to occur in the
small-crack regime (especially under higher cyclic stresses) and must be addressed. Special attention must be given to the minimum
interval between successive crack length measurements, Δa. Closely spaced measurements are often needed to capture key crack-
microstructure interactions, but measurement error can significantly influence apparent variations in da/dN for extremely small Δa
values (Larsen et al., 1992). Some of the alleged evidence for large scatter in small-crack growth rate data is actually just the result of
measurement intervals that are too small relative to the measurement error. Variation in crack growth rates also occurs when cracks
arrest during part of the measurement interval, which can be common for small fatigue cracks.
Further information is available in appendix X3 to ASTM Test Method E 647 (2002) (“Guidelines for measuring the growth rates of
small fatigue cracks”). This appendix does not prescribe complete, detailed test procedures. Instead, it provides general guidance on
the selection of appropriate experimental and analytical techniques and identifies aspects of the testing process that are of particular
importance when fatigue cracks are small.

Fatigue and Fracture Analysis


Kuang-Hua Chang, in Product Performance Evaluation with CAD/CAE, 2013

4.6.3 Rain-Flow Counting


Most traditional fatigue life prediction models require a constant-amplitude loading as input. Constant-amplitude loading is usually
represented by repeating sinusoidal or triangular waveforms. In the case of fatigue crack initiation prediction, the conventional
strain-life relationship is used to relate a constant-amplitude strain loading Δε/2, shown in Figure 4.24, to the fatigue crack initiation
life 2Nf, using, for example, Eq. 4.18.

Sign in to download full-size image

Figure 4.24. Constant-amplitude strain history.

However, components of mechanical systems usually experience a variable-amplitude loading that yields a stress or strain history
with variable amplitude, such as the one shown in Figure 4.23. To predict fatigue life of a component subject to a variable-amplitude
loading history, the variable-amplitude stress or strain history must be converted into several constant-amplitude cycles. The
https://www.sciencedirect.com/topics/engineering/fatigue-life-prediction 11/14
7/27/2019 Fatigue Life Prediction - an overview | ScienceDirect Topics
procedure for this conversion is referred to as cycle counting procedure. A number of such procedures have been proposed over the
years, but all of them must use some rules to decide when or how a cycle is defined from a variable-amplitude history. A well-
accepted procedure, called rain-flow counting (Downing and Socie 1982), is discussed next. This procedure attempts to define cycles
that correspond to a closed stress-strain hysteresis loop.
The rain-flow counting algorithm is used in the analysis of fatigue data to reduce a spectrum of varying stress into a set of simple
stress reversals. Its importance is that it allows the application of Miner’s rule to assess the fatigue life of a structure subject to
complex loading. The algorithm was developed by Endo and Matsuiski in 1968; they describe the process in terms of rain falling off a
pagoda roof.
Figure 4.25a shows a typical stress history, composed of repetitive blocks. A stress block (stress points A to G) is identified and rotated
90 deg. clockwise, as shown at the top of Figure 4.25b. The rotated stress block (or loading history) (points A–G) resembles a Japanese
pagoda. The corresponding stress and strain history is plotted directly below the loading history. In the lower stress-strain plot, three
cycles are easily identified: one large overall cycle (A–D–G), one intermediate cycle in the center of the plot (C–B–C), and one smaller
cycle (E–F–E). Each cycle has its own strain range and mean stress.

Sign in to download full-size image

Figure 4.25. Rain-flow cycle counting: (a) stress history, (b) hysteresis loops, and (c) rain-flow counting.

From a deformation viewpoint, the process proceeds as follows. Start at A, the maximum strain, and load the material to B in
compression. Then reload to point C and compress to D. When the material reaches the strain at point B during the loading from C
to D, the material remembers its prior deformation and deforms along a path from A to D as if event C–B never happened. This is
better illustrated in the next part of the loading. Load from D to E and unload to F. Now load from F to G. When the material reaches
the strain at point E during the loading from F to G, the material remembers its prior deformation and deforms along a path from D
to G as if the event E–F never happened. As a result, rain-flow counting identifies three cycles: A–D–G, B–C–B, and E–F–E.
Several rules are imposed on raindrops falling down these sloping roofs so that the rain flow may be used to define cycles and half-
cycles of fluctuating stress in the spectrum. Rain flow is initiated by placing raindrops successively at the inside of each peak
(maximum) and valley (minimum). The rules are as follows:
1. The rain is allowed to fall on the roof and drip down to the next slope. For example, the rain flow begins at point A, drips to point
B, then to point D. After point D, there are no more roofs to drip to, so stress points A–D are counted as a half-cycle, cycle 1.
https://www.sciencedirect.com/topics/engineering/fatigue-life-prediction 12/14
7/27/2019 Fatigue Life Prediction - an overview | ScienceDirect Topics

2. Step 1 is true except that, if the rain drop initiates at a valley (for example, point B shown in Figure 4.25c), it must be terminated
when it comes opposite a valley more negative (in this case point D) than the valley from which it initiated. Therefore, stress
points B–C are counted as a half-cycle, cycle 2.
3. Similarly, if the rain flow initiates at a peak, it must be terminated when it comes opposite a peak more positive than the peak
from which it initiated.
4. The rain flow must stop if it meets the rain from a roof above. For example, the rain begins at point C and ends beneath cycle 1,
right under point B. Therefore, stress points C–B are counted as a half-cycle, cycle 3.
5. Half-cycles of identical magnitude (but opposite sense) must be paired up to count the number of complete cycles.

Following the rules, from Figure 4.25c, stress points D–E–G are counted as a half-cycle (rule 1), cycle 4. Stress points E–F are counted
as half-cycle 5 (rule 1), and points F–E are counted as half-cycle 6, as shown in Figure 4.25c. Half-cycles 1 and 4 form a complete
cycle (rule 5); hence, the hysteresis loop A–D–G shown in Figure 4.25b. Similarly, half-cycles 2 and 3 form the second cycle, the
smaller loop B–C in Figure 4.25b. Finally, half-cycles 5 and 6 form the third cycle, the smallest loop F–G in Figure 4.25b.

Fatigue Life Prediction Methods of Seam-Welded Joints


Hong-Tae Kang, Yung-Li Lee, in Metal Fatigue Analysis Handbook, 2012

Publisher Summary
This chapter discusses fatigue life prediction methods of seam-welded joints. Welding strongly affects the materials by the process of
heating and subsequent cooling as well as by the fusion process with additional filler material, resulting in inhomogeneous and
different materials. Furthermore, a weld is usually far from being perfect, containing inclusions, pores, cavities, undercuts, and so on.
The shape of the weld profile and nonwelded root gaps creates high stress concentrations with varying geometry parameters.
Moreover, residual stresses and distortions due to the welding process affect the fatigue behavior. Nominal stress approaches are
relatively simple but have limitations to disclose stresses and strains at the critical regions at the welded joint. It also appears difficult
to determine weld class for complex weld shapes and loadings.

Advances in Crystals and Elastic Metamaterials, Part 1


Javier Segurado, ... Javier LLorca, in Advances in Applied Mechanics, 2018

7.2.1 Prediction of the Fatigue Life in IN718 Ni-Based Superalloy


To illustrate the fatigue life prediction process using CHP, an example of prediction of the fatigue life in IN718 Ni-based superalloy in
the LCF regime at 400°C is presented next (Cruzado et al., 2017, 2018a, 2018b). The CP model used to represent the cyclic plastic
behavior (Cruzado et al., 2017) included the mechanisms leading to Bauschinger effect, mean stress relaxation and cyclic softening
and a summary of the constitutive equations was already presented in Eqs. (25)–(27). The fatigue life estimation was based on the
plastic energy dissipated by cycle FIP, Eq. (186), averaged in slip bands (Castelluccio & McDowell, 2015). The resulting FIP of a
particular RVE, was given by
(191)

where ki (= 1, 2, 3) corresponds to the three different slips systems contained in the slip plane parallel to the band i, Vi is the volume
of that band and nb is the total number of bands in the microstructure, which is four times the number of elements in the RVE. The
fatigue life estimation for this particular RVE (the number of cycles Ni) is obtained as function of Wcyc using a power-law relation,
introduced by Cruzado et al. (2018b), to account for the change in deformation mechanism that controls the nucleation of fatigue
cracks in this material: from localized deformation in a few grains at small cyclic strain ranges to homogeneous plastic deformation
https://www.sciencedirect.com/topics/engineering/fatigue-life-prediction 13/14
7/27/2019 Fatigue Life Prediction - an overview | ScienceDirect Topics
cracks in this material: from localized deformation in a few grains at small cyclic strain ranges to homogeneous plastic deformation
at large cyclic strain ranges. Thus,
(192)

where the parameters Wcrit and m were obtained from two independent fatigue experiments corresponding to two different cyclic
strain ranges under the same strain ratio Rɛ = −1 (Fig. 13A). To this end, the FIPs (Wcyc) obtained by averaging the results of 20
different RVEs for each of the two loading conditions selected were used.

Sign in to download hi-res image

Fig. 13. Experimental results and model predictions for the fatigue life of IN718 Ni-based superalloy at 400°C. (A) Rɛ = −1 and (B) Rɛ = 0. The cyclic strain amplitude,
Δϵ, is normalized by Δϵmin, the minimum cyclic strain amplitude used in the tests.
Reprinted from Cruzado, A., Lucarini, S., LLorca, J., & Segurado, J. (2018b). Microstructure-based fatigue life model of metallic alloys with bilinear Coffin-Manson
behavior. International Journal of Fatigue, 107, 40–48.

Finally, the model was used to estimate the fatigue life under different loading conditions. Predictions for each loading condition
were obtained by averaging the FIPs computed using different RVEs. The ability of the proposed model to predict the fatigue life of
IN718 alloy at 400°C is depicted in Fig. 13A and B, in which the experimental and the model results for the fatigue life are plotted as
a function of the cyclic strain range in tests carried out with Rɛ = −1 and 0, respectively. The model was able to predict accurately the
fatigue life under small and large cyclic strain ranges for all the loading cases considered. Moreover, the numerical predictions with
different RVEs led to estimations of the scatter in fatigue life which followed the same trends that the experimental observations: the
scatter decreased as the applied cyclic strain range increased.

About ScienceDirect Remote access Shopping cart Advertise Contact and support Terms and conditions
Privacy policy

We use cookies to help provide and enhance our service and tailor content and ads. By continuing you agree to the
use of cookies.
Copyright © 2019 Elsevier B.V. or its licensors or contributors. ScienceDirect ® is a registered trademark of Elsevier
B.V.

https://www.sciencedirect.com/topics/engineering/fatigue-life-prediction 14/14

Você também pode gostar