Are you sure?
This action might not be possible to undo. Are you sure you want to continue?
Br idge Alliance
A Fat igue Pr imer
for
St r uct ur al Engineer s
by
J ohn W. Fisher
Geoffr ey L. Kulak
Ian F. C. Smit h
A Note of Cauti on
All dat a, specificat ions, suggest ed pract ices, and
drawings presented herein, are based on the best
available information and delineated in accordance
with recognized professional engineering principles
and practices, and are published for general informa
tion only. Procedures and products, suggested or dis
cussed, should not be used without first securing com
petent advice respecting their suitability for any given
application.
Publication of the material herein is not to be con
strued as a warranty on the part of the National S teel
Bridge Alliance – or that of any person named herein
– that these data and suggested practices are suitable
for any general or particular use, or of freedom from
infringement on any patent or patents. Further, any
use of these data or suggested practices can only be
made with the understanding that the National S teel
Bridge Alliance makes no warranty of any kind
respecting such use and the user assumes all liability
arising therefrom.
Copyri ght © 1998, Nati onal Steel Bri dge Alli ance May 1998
A FATIGUE PRIMER
for
STRUCTURAL ENGINEERS
by
John W. Fisher
Lehigh University
Bethlehem, Pennsylvania, USA
Geoffrey L. Kulak
University of Alberta
Edmonton, Alberta, Canada
Ian F.C. Smith
Swiss Federal Institute of Technology
Lausanne, Switzerland
ii
ACKNOWLEDGMENTS AND DISCLAIMER
Several US state highway departments and highway design firms or consultants
supplied information on various bridge structures reported in this document. These include
transportation departments in the States of Connecticut, Maryland, Massachusetts,
Minnesota, Ohio, Pennsylvania, South Dakota, Virginia, and West Virginia and the firms
DeLeuw Cather and Company, Fay, Spofford and Thorndike Inc., Greiner Engineering
Sciences, Kozel Engineering Company, Modjeski and Masters Inc., and Wiss Janney Elstner
Associates Inc. Their contributions are noted with thanks.
The document was proofread with great care by Jeffrey DiBattista, Graduate Student
at the University of Alberta and helpful comments were provided on the technical content by
Prof. Gilbert Grondin, also of the University of Alberta. Jeffrey DiBattista did many of the
calculations associated with the example presented in Chapter 5. The contributions of
Manfred Hirt, Professor of Steel Construction and Director of ICOM, Swiss Federal Institute
of Technology, Lausanne, are gratefully recognized. Professor Hirt is the coauthor of a
principal source document (Reference 1). Figure 4 is taken from Hirt, M.A. "Anwendung der
Bruchmechanik für die Ermittlung des Ermüedungsverhaltens geschweisster
Konstruktionen," Bauingenieur, 57 (1982), and is used with his permission. The authors also
thank Dr. Peter Kunz, formerly at ICOM, for his help with the preparation of Example 7.
The authors have taken care to ensure that the material presented is accurate.
However, it must be understood that persons using the material assume all liability arising
from such use. Notification of errors and omissions and suggestions for improvements are
welcome. No part of this publication may be reproduced or distributed in any form or by any
means without the prior written permission of the authors.
iii
TABLE OF CONTENTS
1 Introduction ................................................................................................................1
2 Basic Fracture Mechanics Concepts
2.1 How to Account for a Crack ............................................................................ 2
2.2 Fracture Limit State ......................................................................................... 6
2.3 Fatigue Limit State .......................................................................................... 8
2.4 Fracture Mechanics Used as a Qualitative Design Tool .................................11
3 Fatigue Strength Analysis
3.1 Introduction and Historical Background.........................................................16
3.2 Sources of Flaws in Fabricated Steel Structures .............................................18
3.3 Basis for Design Rules ...................................................................................22
3.4 Design Rules Given by the AASHTO Specification.......................................25
3.5 Fracture Mechanics Analysis of Fatigue.........................................................29
4 Fatigue Assessment Procedures for Variable Stress Ranges
4.1 Cumulative Fatigue Damage...........................................................................34
4.2 Analysis of Stress Histories.............................................................................40
4.3 Fatigue Limits..................................................................................................47
5 Fatigue Design According to the American Association of State Highway and
Transportation Officials Specification (AASHTO)
5.1 Introduction .....................................................................................................51
5.2 Redundancy and Toughness............................................................................51
5.3 Fatigue Design in the AASHTO Specification ...............................................54
5.3.1 Fatigue Load and Frequency...............................................................54
5.3.2 Fatigue Resistance...............................................................................56
5.4 Summary of AASHTO Requirements.............................................................58
5.5 Design Example ..............................................................................................59
iv
6 DistortionInducted Fatigue Cracking
6.1 Introduction .....................................................................................................73
6.2 Examples of DistortionInduced Cracking......................................................73
6.3 Further Examples of DistortionInduced Cracking.........................................76
6.3.1 Web Gaps in Multiple Girder and Girder Floor Beam Bridges ..........76
6.3.2 Web Gaps in Box Girder Bridges........................................................79
6.3.3 Long Span Structures ..........................................................................81
6.3.4 Coped Beam Connections ...................................................................82
6.3.5 Connections for Lateral Bracing .........................................................83
6.3.6 Other Examples ...................................................................................83
6.4 AASHTO Specification Requirements Relating to
DistortionInduced Fatigue .............................................................................84
6.5 Design Examples.............................................................................................85
6.6 Summary..........................................................................................................87
7 Inspection and Repair of Fatigue Cracks
7.1 Introduction .....................................................................................................88
7.2 Protocol for Fatigue Crack Investigation ........................................................88
7.3 Identifying the Causes of Cracking.................................................................89
7.4 Cracking at Low Fatigue Strength Details ......................................................90
7.5 Methods for Inspection of Fatigue Cracking...................................................95
7.6 Repair of FatigueCracked Members ..............................................................97
7.7 Avoiding Future Cracking Problems............................................................. 100
8 Special Topics
8.1 Bolted or Riveted Members .......................................................................... 102
8.1.1 Bolted Members ................................................................................ 102
8.1.2 Threaded Rods................................................................................... 106
8.1.3 Riveted Connections.......................................................................... 108
v
8.2 Environmental Effects; Use of Weathering Steel.......................................... 110
8.3 Combined Stresses ........................................................................................ 114
8.4 Effect of Size on Fatigue Life ....................................................................... 115
8.5 Role of Residual Stress.................................................................................. 115
8.6 Quantitative Design Using Fracture Mechanics............................................ 121
REFERENCES ........................................................................................................ 124
INDEX...................................................................................................................... 128
1
Chapter 1. Introduction
Fatigue in metals is the process of initiation and growth of cracks under the action of
repetitive load. If crack growth is allowed to go on long enough, failure of the member can
result when the uncracked crosssection is sufficiently reduced such that the part can no
longer carry the internal forces. This process can take place at stress levels (calculated on the
initial crosssection) that are substantially less than those associated with failure under static
loading conditions. The usual condition that produces fatigue cracking is the application of a
large number of load cycles. Consequently, the types of civil engineering applications that
are susceptible to fatigue cracking include structures such as bridges, crane support
structures, stacks and masts, and offshore structures.
The first approach in the design and execution of structures is to avoid details that
might be prone to cracking, and then to inspect the structure for cracks, both during
fabrication and later in its life. However, it is inevitable that cracks or cracklike
discontinuities will be present in fabricated steel elements, and it is the responsibility of the
engineer to consider the consequences in terms of brittle fracture and in terms of fatigue. The
fatigue behavior of a fabricated steel engineering structure is significantly affected by the
presence of preexisting cracks or cracklike discontinuities. Among other things, it means
that there is little or no time during the life of the structure that is taken up with "initiating"
cracks.
Probably the most common civil engineering structures that must be examined for
fatigue are bridges. In North America and elsewhere, early steel bridge structures were
fabricated using mechanical fasteners, first rivets and later highstrength bolts. In these cases,
initial imperfections are relatively small. In addition, loading and load frequency were also
low by today's standards. Consequently, fatigue cracking in these early structures was
infrequent. In the 1950's welding began to be used as the most common method for
fabrication. This had two principal effects related to fatigue. First, welding introduces a more
severe initial crack situation than does bolting or riveting. Second, the continuity inherent in
welded construction means that it is possible for a crack in one element to propagate
unimpeded into an adjoining element.
Design rules at this period of time had been developed from a limited experimental
base and the mechanism of fatigue crack growth was not well understood. Furthermore, most
of the experimental results came from smallscale specimens. This is now known to be a
2
limitation in evaluating fatigue strength: reliance on smallscale specimens can result in
overestimates of fatigue strength.
During the 1970's and 1980's there were many examples of fatigue crack growth from
welded details now known to be susceptible to this phenomenon. Research revealed that the
type of cracking observed in practice was in agreement with laboratory test results and
supportable by theoretical predictions. Experience in the 1970's also exposed an unexpected
source of fatigue cracking, that from distortions. This is also a phenomenon related largely to
welded structures.
The purpose of this publication is to provide the student and the practicing engineer
with the background required to understand and use the design rules for fatigue strength that
are currently a standard part of design codes for fabricated steel structures. The approach
adopted establishes the basis for the problem in terms of fracture mechanics, that is, an
analytical tool that accounts for the presence of a crack in a structure [1]. The focus is then
directed specifically upon the issue of fatigue.
It is intended that fundamentals are presented in a general way, but applications will
refer to the specification for the design of steel bridges prepared by the American
Association of State Highway and Transportation Officials (AASHTO) [2]. This
specification is widely used in the United States, and the governing Canadian specification is
nearly identical to it.
3
Chapter 2. Basic Fracture Mechanics Concepts
Use of the fracture mechanics method of analysis is relatively recent. Originally
advanced to explain the rupture of glass specimens [3], its introduction into the field of
structural engineering practice started when it was used in the 1940's to help explain the
catastrophic failure of welded ship hulls. Currently, it is employed to assess the behavior of
elements used in machinery, pipelines, automotive parts, spacecraft, turbine blades, and
many other components.
In this Chapter, basic concepts of the fracture mechanics approach are described in
order to assist the reader in understanding the fatigue design rules. In addition, for those who
might need to design at higher levels of sophistication, it will provide the basis for further
reading and selfinstruction.
Only a summary of fracture mechanics concepts is given in this section. For
simplicity, the discussion is limited to cases where the loads are applied at locations remote
from the crack locations and normal to the crack surfaces, the socalled Mode 1 situation.
(The different ways in which a crack can open, or modes, will be explained in Example 1).
Excellent review articles provide more detailed information (for example, see Ref. [4–6]) and
several reference books are available [7–9].
2.1 How to Account for a Crack
Five cases of a loaded plate containing a crack are shown in Fig. 1. It requires no
knowledge of fracture mechanics to appreciate that cases 1 to 5 are placed in order of
increasing severity. Taking Case 1 as the basis for comparison, the following important
Decreasing Strength
Figure 1 Five Different Cases of a Plate Containing a Crack
1 2 3 4 5
4
fracture mechanics parameters can be identified: i) crack length (Case 2); ii) crack location
(at edge of plate in Case 3); iii) effect of bending (Case 4); and iv) presence of a stress
concentration (Case 5). Of course, the result of any of these parameters in weakening the
plate will depend on the actual circumstances. The effect can be significant, however. For
instance, the consequence of a sharp stress concentration in combination with a crack
(Case 5) could weaken a plate to less than onehalf of its uncracked strength.
A magnified view of the area around a crack tip in an infinitely wide plate is shown in
Fig. 2. This resembles closely the conditions of Case 1 when the crack length is small
compared with the plate width. When a remote stress, σ, is applied, the crack opens a certain
distance, d, and the stress that this crosssectional area would have carried is diverted to the
uncracked area of the plate. This diversion creates a high concentration of stress in the
vicinity of the crack tip. For an elastic material, theoretically this stress is infinite at the crack
tip: in real materials, plastic zones are formed since the strain exceeds the ability of the
material to behave elastically. This process—whereby i) an applied load causes a crack to
open, ii) the crack opening relieves crack surfaces of stress, and iii) crack tip plastic straining
is created—is the fundamental mechanism that weakens structures containing cracks or
cracklike discontinuities.
If plasticity is ignored, a description of the stress field near the crack tip can be
obtained. Using special stress functions, a solution containing the coordinates r and θ is
σ
σ
yy
τ
xy
σ
xx
r
θ
x
y
a
Detail
Centerline
of crack
Figure 2 A Crack in an Infinitely Wide Plate
A
A
Section A  A
.
crack surface
crack tip
cracktip
plastic zone
crack front
d
5
developed. For the particular case of θ = 0, that is, for the stress in the ydirection, the stress
is
σ
σ π
π
yy
a
r
=
2
(1)
provided that the crack length, a, is much larger than the distance from the crack tip, r. The
numerator in Eq. (1) determines the gradient of the (theoretical) stresses as they rise to
infinity when r approaches zero. This numerator is called the stress intensity factor, K. Thus:
K a = σ π (2)
The advantage of this model is that any combination of stress and crack length can be
characterized by the single parameter K. Analytical solutions are available for other
particular geometrical configurations and loading conditions; these are summarized in
handbooks [10]. However, many practical cases cannot be solved analytically. In such
instances, the following expression is used to approximate K:
K WY a = σ π (3)
where Y is an expression that corrects for plate and crack geometry and W corrects for non
uniform local stress fields caused by the presence of factors such as residual stresses, stress
concentrations, and stress gradients. Usually, such correction factors are determined using
numerical methods. Here, also, solutions can be found in handbooks.
Equation 1 is based on linearelastic material behavior and cannot account for
yielding at the crack tip. Furthermore, stress redistribution due to plasticity alters the stress
field outside the crack tip plastic zone. Nevertheless, if this zone is small, say less than 2% of
the plate thickness, of the crack length, and of the uncracked ligament, then the stress
intensity factor (K) approach is satisfactory.
These limitations are violated in many practical situations. For example, an elastic–
plastic analysis may be required when stress concentrations cause localized plasticity. The
most common analyses either use a parameter named J, which is an expression of the change
in potential energy with respect to crack length, or use a parameter called the crack tip
opening displacement (CTOD). Further description of elasticplastic analyses is available
elsewhere [9] and the American Society for Testing and Materials has produced standards for
6
determining these parameters experimentally. (See, for example, Ref. [11].) Occasionally, an
equivalent K is calculated [3]:
K
E
J
y
2
= = CTOD σ (4)
where E is Young's Modulus and σ
y
is the effective yield strength. When plate dimensions
are large enough to restrict behavior to essentially twodimensional straining (plane strain
conditions), the material constant, E, is replaced by
( )
E / 1
2
− υ , where υ is Poisson's ratio.
A further restriction on K exists for small crack lengths, where all of the approaches
explained above lose their validity. When the size of the crack or initial discontinuity is of
the order of grain size, microstructural properties such as grain orientation influence crack
growth [12]. Microstructural fracture mechanics models may then become necessary. These
models are not yet well defined, and no generally accepted design rules are available. For the
usual situation, when the crack length is greater than about five grain diameters, the models
that assume an isotropic continuum, i.e., those employing K or J, are sufficiently accurate.
2.2. Fracture Limit State
The analysis tools available will indicate that fracture occurs when the crack
parameter (i.e., K or J) exceeds a critical value, commonly referred to as the fracture
toughness. The designer should choose a steel with a fracture toughness level that is
sufficiently high for the intended application. The fracture toughness depends upon such
factors as microstructure and composition of the material, service temperature, loading rate,
plate thickness, and fabrication processes.
An accurate determination of the fracture toughness is complicated, especially in
most structural engineering design situations. This is primarily due to the fact that, for most
designs, plane strain conditions do not dominate; conditions of essentially twodimensional
stress (plane stress) have an influence, thereby disqualifying K as a model that characterizes
combinations of stress and crack length. Elasticplastic models are needed because the stress
at fracture produces a plastic zone size that exceeds the limitations cited for the validity of K
specified in Section 2.1.
Less sophisticated approaches are used for practical problems in structural
engineering. The most widely used method for approximating the toughness quality of a steel
is a procedure that was developed over 80 years ago, the Charpy impact test. In brief, this
7
method measures the energy absorbed by the rapid fracture of a small bar containing a
machined notch. The bar is broken by a swinging pendulum and the absorbed energy is
measured by the difference in swing height before and after fracture. The effect of
temperature is examined by repeating the test using physically identical specimens that have
been cooled to various temperatures. Several tests provide a relationship between absorbed
energy and temperature for the steel under investigation.
The Charpy test, and many other similar procedures [9] provide only qualitative
information because stress and crack length values cannot be assessed directly. However,
correlation with fracture mechanics models are available for certain situations [8]. For
example, Charpy impact data in the lower region of the energy vs. temperature response
curve can be converted into dynamic planestrain fracture toughness values, K
Id
, using an
empirical formula.
1
If required, the K
Id
values then can be converted to static planestrain
fracture toughness values, K
Ic
. In general, such correlations are valid only for steels of low to
medium strength. Moreover, the toughness value where conditions change from plane strain
to plane stress depends upon the yield strength and the thickness of the element. The location
of this change cannot be correlated to Charpy results.
At a given temperature, a given material will always exhibit higher fracture toughness
for plane stress conditions than for plane strain conditions in the elasticplastic region.
Therefore, K
Ic
and K
Id
are conservative first estimates of the fracture toughness. Elastic–
plastic analyses, which result in less conservatism, can be used when the design problem
justifies increased complexity of work.
A critical crack length, a
cr
, may be approximated by introducing the fracture
toughness at the minimum service temperature and the maximum possible stress, σ, in Eq. 3.
Fracture will not occur if the maximum size of the crack or cracklike discontinuity, a, is less
than a
cr
. Often, inspection technology influences this verification: a
cr
is compared with the
minimum verifiable crack length, a
o
. The engineer should ensure that the design
configuration and the inspection equipment and procedures permit reliable detection of crack
lengths less than a
cr
.
In most modern steel structures, the likelihood of fracture shortly after erection has
been completed is not high. Some problems may arise when severe weld discontinuities are
present or when very thick plates are used or when the structure is in a very cold
environment, but such situations are not common. Rational designs, which employ high
1
The subscripts used here for K are explained in Example 1.
8
strength finegrained steels and which use modern fabrication techniques, should provide
high toughness and, consequently, large critical crack lengths, a
cr
. At the same time, careful
assembly and improvements in the technology of nondestructive inspection are reducing the
minimum verifiable crack length, a
o
. As a result, the greatest risks of fracture in modern steel
structures arise when subcritical crack growth due to fatigue, corrosion, stress corrosion, etc.
is possible. Predicting the occurrence of the fracture limit state is becoming more dependent
upon the correct choice of the subcritical crackgrowth model than upon an accurate
estimate of the fracture toughness of the detail. Crack growth due to fatigue is discussed in
the next section.
2.3 Fatigue Limit State
Fatigue is the initiation and propagation of microscopic cracks into macro cracks by
the repeated application of stresses. (As has already been noted, the initiation portion of
fatigue life is essentially nonexistent for fabricated steel structures.) In civil engineering
practice, examination of the possibility of fatigue cracking must be a consideration for
bridges, cranes, towers, offshore platforms, and any such structure that is subjected to
repeated loading. Although steel structures today use highertoughness materials than was
common in the past, and are thus more resistant to fracture than ever before, many structural
elements remain susceptible to fatigue crack growth. Consequently, if the fracture limit state
is reached, it is often the result of fatigue crack growth after many years of troublefree
service. Such an occurrence is covered by the definition of the fatigue limit state.
In addition to highertoughness properties, development of lowalloy and finegrain
microstructures have increased the yield strengths of the steels used for construction.
Consequently, higher service stresses have been allowed in recently built structures.
Furthermore, welding is now used more often than formerly and this method of fastening
leads to a lower fatigue life than would apply if the connection was made using rivets or
bolts. Generally speaking, then, modern structures are more susceptible to fatigue cracking
than were older structures. Furthermore, the number of old structures, for example railway
bridges, that have exceeded their design fatigue life is growing exponentially. The combined
effect of these trends is increasing the importance of fatigue strength evaluation.
All elements of a fabricated steel structure contain metallurgical or fabricationrelated
discontinuities, and most also include severe stress concentrators such as weld toes.
Consequently, fatigue failure is often the result of slow crack growth from an existing
discontinuity at a stress concentration. This growth can even begin before the structure is put
9
into service (the result of transportation of a girder, for example). A description of the fatigue
crack growth phenomenon can be made on the basis of the fracture mechanics model
described in Section 2.2.
The stress intensity factor K can be modified to represent fatigue crack growth by
adapting Eq. 3 to account for repeated loading. For a constant stress range,
∆σ σ σ ) (
max min
= − , Eq. 3 becomes
∆ ∆ K WY a = σ π (5)
Equation 5 is related empirically to the crackgrowth rate, da dN, which is obtained
from the slope of the curve of crackgrowth measurements—see Fig. 3(a). This slope is used
as the ordinate in a plot of this parameter against ∆K, using a double logarithmic
representation—see Fig. 3(b). The values of ∆K are calculated using Eq. 5 for particular
magnitudes of crack length, a. At very low growth rates, the curve for crack growth (in steel)
becomes vertical, indicating a crackgrowth threshold at ∆K
th
, the threshold stress intensity
factor. At higher values of ∆K, the curve straightens to a nearconstant slope, and it becomes
vertical again when the fracture toughness is approached at the maximum stress in the cycle.
da
log
dN
m
1
log∆Κ
fracture toughness
K
min
DETAIL
log∆Κ
da
log
dN
extrapolated
design line
high R
low R
R =
K
min
K
max
∆Κ
th
intrinsic
da
dN
2a
a
N
time
∆σ
σ
σ
max
min
σ
σ
σ
∆Κ
th
(a)
(b)
(c)
Figure 3 StressIntensity Factor and Fatigue Crack Growth. (a) Crack length vs.
number of cycles; (b) Crack growth rate vs. stress intensity factor
range; (c) Magnification of the lower portion of the curve in (b)
10
Material properties, stress level, and environment have greater influence in the end
(vertical) portions of the curve than in the middle. In this center portion, which is of
considerable engineering interest, the Paris equation is useful [13]:
da
dN
A K
m
= ∆ (6)
where A and m are constants that are determined by means of regression analysis of test data.
These constants are reliable when similar materials, loadings, and environments are
compared. (The constants show up directly in the design specifications, as will be seen in
Section 3.5.) Also, the regression analysis is dependent upon the domain of crack growth
rates considered since the center portion of the curve in Fig. 3(b) is not perfectly straight.
Many structural applications involve repeated loading of over one million cycles: this
requires a precise knowledge of slow crack growth rates near ∆K
th
—see Fig. 3(c).
Conservative assessments result when Eq. 6 is extrapolated into this region. The error
resulting from the extrapolation is dependent upon the magnitude of the stress ratio,
R = σ σ
min max
. This ratio is often used to examine the effects of mean stress on crack
growth.
When ∆K is much larger than ∆K
th
, Eq. 6 can be integrated to calculate the crack
propagation fatigue life, N:
N
A
K
m
i
f
=
∫
1 1
da
a
a
∆
(7)
where a
i
and a
f
are the initial and final crack lengths. These integration limits may take the
values a
o
and a
cr
, respectively, as defined in Section 2.2.
It was pointed out in Section 2.2 that small crack sizes and excessive plasticity may
invalidate models that employ the stress intensity factor. This is equally true for fatigue
applications: nonconservative calculations may result if socalled short crack behavior
occurs [14]. Fortunately, such situations are less common when assessing the fatigue limit
state. Usually, the stress intensity factor remains a useful characterizing parameter for
conditions of fatigue crack growth since discontinuities are large and a high percentage of
crack growth occurs under conditions where K is valid. Moreover, structural engineering
applications have particular characteristics that reduce the occurrence of this anomalous
behavior.
11
2.4 Fracture Mechanics used as a Qualitative Design Tool
Engineering designers rarely use fracture mechanics as a design tool. Older concepts,
such as Charpy energy values for fracture toughness requirements and stress range models
for fatigue assessments (see Section 3.4), are the most practical tools for evaluating many
structures. Nevertheless, the concepts of fracture mechanics enable the designer to increase
his qualitative understanding of structures containing cracklike discontinuities. Because
more parameters are explicit in fracture mechanics analyses, designers can identify more
easily those parameters that influence the strength of the structure. Some examples, covering
the importance of discontinuities, parametric studies, and crack propagation behavior, are
presented below. Guidelines for simple linearelastic fracture mechanics models sufficient
for most structural engineering designs are available in some design codes [15] and one
illustration of the application of fracture mechanics analysis is given in Section 3.5.
The size of the discontinuity clearly influences the resistance of an element to
fracture. For example, two trends are illustrated by the curves in Fig. 4(a). First, increasing
the applied stress causes a decrease of the critical defect size. Second, an embedded
discontinuity (such as an inclusion) is less serious than a surface discontinuity (such as a
weld undercut). Fracture mechanics analysis clarifies the importance of discontinuities in
fracture assessments. This has resulted in an increased emphasis on quality assurance
guidelines and on detail designs that have small discontinuities.
20
40
60
80
0
200 400 600 800
σ
(MPa
)
2a
a
(a)
a
cr
(mm)
a
N (cycles)
1.0
0.1
10
4 5
10
7
10
a
f
= 40 mm
a
f
= 20 mm
= 10 mm
a
f
(b)
6
10
a
i
(mm)
10
Figure 4 Examples of Studies that Examined the Significance of Discontinuities (a) Critical crack size vs.
applied stress; (b) Effect of initial and final crack size on fatigue life
The effect of discontinuities is equally important when assessing the fatigue limit
state. The integral in Eq. 7 has the limits a
i
and a
f
. Small variations in the magnitude of the
12
final crack size, a
f
, may not significantly alter the resulting fatigue life. The initial crack size
is far more important than the final crack size—see Fig. 4(b).
Fracture mechanics analysis facilitates recognition
that any areas not welded become builtin cracks. Figure 5
shows an end plate welded to a beam. (The assembly will
subsequently be bolted to a column flange, not shown.)
Groove welding may not always be able to penetrate
completely into the zone through which welding is
intended. In the example, the flange groove weld will be
difficult to complete in the vicinity of the webtoflange
junction. (However, provision of cope holes will give
better access to the webtoflange junction.) Design
guidelines already recognize that flange fillet welds, an
alternative to the groove welds shown, produce a detail of
low fatigue strength because of the large unwelded areas (cracks).
A fatigue crack that starts at the surface of the material initially propagates very
slowly into the plate thickness. The stress concentration, modeled by the parameter W in
Eq. 5, affects the crack growth rate. Connections classified by specifications as falling into
different fatigue strength categories may have different crack propagation characteristics for
the same fatigue life—see Fig. 6.
In this example (Fig. 6), the detail with lowerseverity stress concentration (groove
welded plate) consumes a greater percentage of its total fatigue life during the propagation of
groove
weld
fillet
welds
Figure 5 Unwelded Areas can
become BuiltIn Cracks
cr
a (mm)
N (cycles)
2x10
6
0
a
a
a
a
Figure 6 Comparison of Fatigue Crack Growth Behavior for
Two Different Details Having the Same Fatigue Life
13
a crack to a size a
o
than does the detail with a higherseverity stress concentration
(attachment detail). This means that a crack of a given size may be identified earlier at the
attachment than at the groove weld. Other factors such as differences in inspection feasibility
influence the exact timing of crack identification. Nevertheless, such crack growth behavior
should be considered when establishing inspection intervals.
EXAMPLE 1
Crack loading modes are shown in Fig. 7. In structural engineering applications,
Mode I, the crack opening mode, will almost always be the applicable condition. In order to
distinguish among the various possible cases, the stress intensity factor K introduced in
Section 2.1 is subscripted, e.g., K
I
would be used for a Mode I case. The value of the stress
intensity factor at which brittle fracture will occur is designated as K
Ic
or as K
Id
, depending
upon whether the loading is essentially static (K
Ic
) or is dynamic (K
Id
).
Published solutions for the expressions for the stress intensity factor are available for
a wide variety of conditions [10]. Several common cases are shown in Fig. 8. (The term Q in
part (b) is a correction for the presence of an elliptical surface flaw.)
x
y
z
x
z
y
z
x
y
MODE I
MODE II MODE III
Figure 7 Basic Modes of Loading Involving Different Crack Surface Displacements
σ
a
(b) Surface crack
a
(c) Edge crack
σ
σ
2a
(a) Through
thickness crack
K a
I
= σ π K a Q
I
=112 . / σ π
K a
I
=112 . σ π
2c
Figure 8 K
I
Values for Various Crack Geometries (Infinitely Wide Plates)
14
Consider the element shown in Fig. 9, where a centrally located throughthickness crack is
present in a plate that is loaded by a uniform tensile stress. (It is to be understood that the
crack width is small relative to the plate width in the illustrations of Figs. 8 and 9. These
illustrations are intended to depict the case of an "infinitely wide plate.")
Given: σ
y
= 550 MPa
K
Ic
= 66 MPa m
1/2
(This information is provided by
the supplier of the material and is related to the
intended service temperature, loading rate, and
the particular plate thickness.)
Design stress = 140 MPa
Question: (a) What is the flaw size at which brittle fracture
might be expected?
Solution: It is given that the value of the critical stress intensity
factor is K
Ic
= 66 MPa m
1/2
. From Fig. 8, the general
expression for the stress intensity factor is
K a
Ic
= σ π . It is simply a matter of examining the situation K K
I Ic
→ , or
σ π
σ
MPa
Using = 140 MPa and solving, = 0.27
a m mm
a m
a m
=
= =
66
0 071 71 .
Thus, if the flaw size reaches 2a = 142 mm, failure by brittle fracture can occur.
Question: (b) If the design stress is increased to 310 MPa, what is the tolerable flaw size
now?
Solution: σ π a MPa m = 66
310 66 MPa a MPa m π =
Solving, 2a = flaw size = 29 mm.
Question: (c) Residual stress due to welding is present and it is estimated that the total stress
(design stress + residual stress) is 500 MPa. What is the tolerable flaw size
now?
Solution: σ π a MPa m = 66
σ
Throughthickness
crack: K a
I
= σ π
2a
Figure 9
15
500 66 MPa a MPa m π =
Solving, 2a = flaw size = 11 mm.
Comment: Often, it is not possible to superimpose stresses due to residual stresses because—
1. The distribution of residual stress is different from the applied stress. (Recall that the
expression in Fig. 8 is valid only for a uniform, remote stress);
2. There is a stress concentration present. This requires that a correction factor for K be used
on the applied loads, but not for the residual stresses;
3. There is a possibility of residual stress relaxation with crack growth. In such cases, the
value of K for residual stresses should be calculated separately and then added to the K for
the applied loads. Thus, superposition is applied at the level of K, the parameter that
characterizes the stress field at the crack tip.
16
Chapter 3. Fatigue Strength Analysis
3.1 Introduction and Historical Background
Fatigue cracking was observed in railroad equipment over 120 years ago. Studies
carried out at that time by Wöhler on railway rolling stock showed that stress concentrations
and sharp angles in the axle configuration resulted in failures even though the stress in the
material was well below its yield strength. The industrialization of society and the
subsequent increased use of machinery and equipment led to other examples of failures
resulting from fatigue cracking. As a result, studies into the phenomenon started in both
Europe and in North America. For example, in North America the observation of cracks in
railroad bridge truss hangers and in stringer end connection angles led to a number of
laboratory investigations between 1930 and 1960.
Welded details were first examined in the 1930's when tests were carried out on
welded steel details. These and later studies following World War II formed the basis for the
early fatigue design specifications in North America. Fatigue cracks forming in steel bridges
at a road test program conducted in the USA in the 1960's [16] became the genesis of the
fatigue test program sponsored by the National Cooperative Highway Research Program
(NCHRP) that began at Lehigh University in 1968. Prior to the NCHRP program, the fatigue
design rules that existed for welded steel bridge components were based on small specimens
and on a limited quantity of test data. This made it difficult to establish the significance of
stress variables, detail type, types of steel, and quality of fabrication. The early provisions for
fatigue life evaluation proved to be inadequate for a number of bridge details. This explains,
in part, the relatively large number of cases of fatigue cracking in bridges that were designed
prior to about 1975.
The approach taken by modernday specifications for the fatigue design of fabricated
steel structures is based primarily on work done in Great Britain [17] and in the USA [18–20]
in the late 1960's and early 1970's. Although many other investigators have contributed to
our understanding of the problem, both before and after the work cited, it was this research
that identified the influence of residual stress on fatigue life. These studies also revealed the
necessity to acknowledge that fabricated steel structures always contains cracks or cracklike
discontinuities.
Fatigue can be defined as the initiation and propagation of microscopic cracks into
macro cracks by the repeated application of stress. In terms of the fracture mechanics model
17
described in Chapter 2, an initial crack grows a small amount in size each time a load is
applied. A good explanation of the crack growth mechanism has been provided by Broek
[21]. Growth occurs at the crack front, which is initially sharp. Even at relatively low loads,
there will be a high concentration of stress at the sharp front, and plastic deformation (slip on
atomic planes) therefore occurs at the crack front. Continued slip results in a blunted crack
tip, and the crack grows a minute amount during this process. Upon unloading, not
necessarily to zero, the crack tip again becomes sharp. The process is repeated during each
load cycle.
Figure 10 shows the fracture surfaces of a
member that has an Ishaped crosssection. The
web of the member, which was 10 mm thick, was
filletwelded to 13 mm thick flange plates. (The
full thickness of the flange is not shown in
Fig. 10). The profiles of the fillet welds are
generally satisfactory and the flow lines of the
weld show good penetration of the base metal. In
this illustration, an internal flaw in the lefthand
fillet weld grew under the repeated application of
stress until the crack penetrated the outside surface
of the weld. Since this was a laboratory specimen,
at this point the beam was deliberately overloaded so that the remaining crosssection
fractured and could be exposed.
In the case illustrated in Fig. 10, the crack front eventually reached the exterior
surface of the weld. Experience in the laboratory shows that as much as 80% of the fatigue
life has been consumed by the time a fatigue crack emanating from an internal flaw reaches
the surface and can be observed.
If the test that produced the specimen shown in Fig. 10 had not been terminated by
the investigators, failure could have occurred in one of two ways. One possibility is that the
fatigue crack grows to such an extent that the loss of crosssection means the load simply can
no longer be carried by the uncracked portion of the beam. In this case, failure occurs by
yielding of the remaining material, or, exceptionally, by instability if the crack growth
produces a grossly unsymmetrical crosssection. The other way that the beam can fail is by
brittle fracture. As discussed in Section 2.3, growth of a crack by fatigue can lead to brittle
Figure 10 Fracture Surface of Ishaped Member
18
fracture if the crack reaches a critical size according to the particular conditions of material
toughness, temperature, and loading rate.
3.2 Sources of Flaws in Fabricated Steel Structures
The kinds of flaws that can occur in a filletwelded detail are shown pictorially in
Fig. 11. These include partial penetration and lack of fusion, porosity and inclusions (the
fatigue crack shown in Fig. 10 started at a nonmetallic inclusion), undercut or micro flaws at
the weld toe, and cracking or inclusions around a weld repair or at startstop locations or at
arc strikes. Although the fabricator of the structure and those responsible for the fabrication
inspection will attempt to minimize these defects, it is neither practical nor economically
possible to eliminate them.
Test data on welded details have demonstrated that all fatigue cracks commence at
some initial discontinuity in the weldment or at the weld periphery, and grow perpendicular
to the applied tensile stresses. In a welded beam without attachments (simply two flange
plates welded to a web), most laboratory fatigue cracks are observed to originate in the web
toflange fillet welds at internal discontinuities such as porosity (gas pockets), incomplete
fusion, or trapped slag. Figures 12 and 13 show fatigue cracks that have formed from
porosity and entrapped slag in longitudinal submerged arc fillet welds. These discontinuities
are always present to some degree, irrespective of the welding process and techniques used
during fabrication.
porosity
undercut
weld repair or startstop
of weld or arc strike
lack of fusion
partial penetration
Figure 11 Flaws in a FilletWelded Detail
19
Figure 13 Fatigue Crack Enlarged to ThreeEnded Crack From Internal Porosity
Attachments such as cover plates, gussets, stiffeners, and other components welded to
a web or flange introduce a transverse weld periphery (toe), thus forming a line of elevated
tension where fatigue cracking can start from small, sharp discontinuities. Figure 14 shows a
fatigue crack that has formed at a cover plate fillet weld toe.
Figure 12 Fatigue Cracks Forming from Internal Porosity in Web–Flange
Connection
20
(a) Fatigue Crack at End of Cover Plate Fillet Weld Toe
(b) Crack Surface Showing Fatigue Crack Growth
Figure 14 Fatigue Cracking at Fillet Weld Toe
In some cases, a "defect" is an expected result of the type of fabrication process and
has no effect on the life of the member. For instance, the partial penetration shown in Fig. 11
(which is also seen in the welded beam of Fig. 10) is a natural consequence of the fillet
welded connection: it is not expected that the two fillet welds will merge in the central region
of the connection. Furthermore, since the crack represented by the lack of penetration is
parallel to the direction of the (bending) stress field, a socalled Mode II crack (see Fig. 7),
the crack will not open up under the application of stress and failure by fatigue is unlikely.
21
Consider a detail involving mechanical fasteners—an Ishaped beam with a cover plate
fastened to the beam flange with bolts. The region between bolt lines could be described as a
"flaw" or "crack," but, since the discontinuity is parallel to the stress field, the "crack" does
not grow and therefore its presence does not affect the fatigue strength of the member.
The flaws that exist in all fabricated steel structures are a consequence of the
manufacturing process of the steel itself and the normal fabrication processes. Flaws in rolled
shapes arise from surface and edge imperfections, irregularities in mill scale, laminations,
seams, inclusions, etc., and from mechanical notches due to handling, straightening, cutting,
and shearing. In a rolled shape, fatigue crack growth can start from one of these sources.
Comparatively, the "unaltered" rolled shape presents the most favorable fatigue life situation.
However, there are not many practical cases in which a rolled shape does not have some kind
of attachment, connection, or some other kind of alteration.
Mechanical details, in which holes are drilled or punched and forces are transferred
by means of rivets or bolts, present a somewhat more severe fatigue life situation than the
bare rolled shape. Drilled or subpunched and reamed holes give some reduction in fatigue
life as compared with an unaltered member, but the difference usually is not very great. If
preloaded highstrength bolts are used, the disturbing effect of the hole is largely mitigated
by the presence of the high local compressive stresses introduced by the bolt. Punched holes
give a greater reduction in fatigue life than do drilled or subpunched and reamed holes
because of imperfections at the hole edge arising from the punching process. In this case, the
crack usually starts at the edge of the hole.
Broadly speaking, any mechanical detail has a better fatigue life than does its
equivalent welded detail. The types of flaws introduced when welding is used have already
been discussed. In addition to the fact that more flaws will be present when welding is used,
inspection for defects is more difficult than is the case when mechanically fastened details
are used. Repairing defects in welded details is also difficult. Prohibiting the use of welded
details in fatigue situations is not usually a practical option, however.
The task of the structural engineer is to be able to proportion those structural
members that have a potential for failure by fatigue crack growth so that they have a
sufficiently long life as compared with the design life of the structure. As will be seen, this
will be done in the environment that some probability of failure must be accepted: in real
terms, there is no structure that can be designed for zero probability of failure. The design
will be carried out in the expectation that flaws will be present initially in all fabricated steel
22
structures and that all such members will contain residual stresses of relatively high
magnitude. A concomitant feature is that in the design process it is possible to identify the
size of flaws that are permissible and then to use this information as the basis for both initial
inspection of the structure as well as ongoing inspection. This latter feature is not yet well
developed in design specifications, and the usual procedure is to accept as permissible flaw
sizes consistent with the specifications that accompany the fabrication processes, e.g., the
welding specifications.
3.3 Basis for Design Rules
To those working in the research area, it was evident that the features that needed to
be examined included the yield strength of the steel (reflected by the grade of steel), the
number of cycles of load to which the member was subjected, some aspect of the stress itself,
and the stress concentration present.
Figure 15 Effect of Grade of Steel on Fatigue Life of Beams with Transversely EndWelded Cover Plates
Figure 15 shows a plot of the stress feature versus the number of cycles to failure for
a given detail (beams with transversely endwelded cover plates) for three different grades of
steel. These particular steels represent the spectrum of steel strengths in use at the present
time—from 250 to 690 MPa. Given that a certain amount of scatter is always present in
results of this kind, it is evident that the yield strength of the steel does not have an important
23
influence on the results. Other information contained in this figure is significant and is
typical of all fatigue strength tests results. First, the data plot as a straight line in this log–log
representation. Second, the data are contained within a reasonably welldefined band about
the mean. In Fig. 15, the dashed parallel lines plotted two standard deviations (measured
relative to the horizontal axis) away from the mean line contain most of the test results. The
degree of scatter in the results will have to be considered when a choice is made for a design
life line. Customarily, this choice is made for the designer by the specification itself.
Figure 16 Effect of Stress Range and Minimum Stress on Fatigue Life for Welded End of Coverplated Beams
The way the stress feature should be represented is examined in the data contained in
Fig. 16. Considering again the endwelded coverplated beam (both asrolled and threeplate
welded beams are represented in Fig. 16, however), the effect of stress is introduced as the
stress range, that is, the algebraic difference between the maximum stress and the minimum
stress at the critical location (∆σ σ σ
r
= −
max min
). The tests represented in Fig. 16 were done
at values of minimum stress (σ
min
) equal to – 41.4 MPa, 13.8 MPa, and 68.9 MPa, and, of
course, a spectrum of stress range. Examine, for example, the data plotted for the stress range
of 80 MPa. For this stress range, this means that the maximum stresses had to have been
38.6 MPa, 93.8 MPa, and 149 MPa, respectively, for the three values of minimum stress. It is
obvious that the data are closely grouped and that the minimum stress per se does not
24
influence the results. As will be seen in Chapter 8, it is the presence of high levels of residual
stress that dictates that stress range is the controlling stress parameter for a description of
fatigue life, rather than maximum stress, minimum stress, or stress ratio (i.e., σ σ
max min
).
In accordance with the ideas developed in Chapter 2, Basic Fracture Mechanics
Concepts, it would be expected that stress intensity (K) should be represented in the fatigue
life evaluation. For the usual level of design, this is not a practical solution, however, and a
more expedient approach is taken. This is simply to arrange standard structural details into
categories relative to their expected fatigue life. For example, illustrated in Fig. 17 are the
fatigue life representations for two different categories—beams which have cover plates that
include a weld across their ends and beams made up of three plates welded together, such as
the beam illustrated by Fig. 10. Clearly, if the designer had one of the two types of members
shown in Fig. 17, it would be possible to determine the fatigue life of that member.
Figure 17 Fatigue Strength of Welded and Coverplated Beams
In summary, the fatigue life of a fabricated steel structure is determined by three
factors. These are:
25
1. The number of cycles of loading to which the member is subjected;
2. the type of detail under examination; and
3. the stress range at the location of the detail.
It has been implicit in the discussion so far that the stresses, which are the driving
force behind crack growth, are those corresponding to the loads on the structure. This is
indeed an important case, and for this situation the stress range to be calculated is simply that
corresponding to the nominal stress at the location of the detail. This is valid because
selection of the detail itself implies inclusion of the stress concentration for that detail. There
is another source of stresses in the structure that can produce crack growth, however. This is
the stresses (really, strains) that are produced as a result of displacements. Displacement
induced fatigue cracking is at least as important as loadinduced fatigue cracking, and it will
be discussed in a separate section (Chapter 6).
3.4 Design Rules Given by the AASHTO Specification
The basis of the fatigue life rules given in all codes, standards, or specifications for
the design of fabricated steel structures has been given in the preceding sections of this
Chapter. Further elaboration will be required on a number of points, but it is appropriate to
introduce next the rules given by one of the mostwidely used standards, that of the
American Association of State Highway and Transportation Officials (AASHTO)[2]. In this
section, the fundamentals of the AASHTO rules will be introduced. The subject is more
completely discussed in Chapter 5.
As in most contemporary standards, the AASHTO fatigue life rules reflect the two
issues—fatigue cracking induced by stress and fatigue cracking induced by displacements
within the structural system. Only the first has been explained so far, and the discussion in
this section will continue to be limited to loadinduced fatigue cracking. Displacement
induced fatigue cracking is presented in Chapter 6.
Figure 18 shows the fatigue life curves given in the AASHTO Specification. The plot
shows stress range on the vertical axis and number of cycles on the horizontal axis for seven
different Detail Categories. Both axes are logarithmic representations. Over some portion of
the range, each Detail Category is a sloping straight line with a slope constant m equal to 3.
Beyond a certain point, which depends on the Detail Category, the fatigue life line is
horizontal. This feature will be discussed subsequently.
26
The information in Fig. 18 must be used in conjunction with information like that
shown in Table 1 and Fig. 19, which give only a small portion of the relevant material in the
AASHTO Specification.
Figure 18 Fatigue Life According to the AASHTO Specification
A
B
B'
C
D
E
E'
10
100
Fatigue Life (cycles)
Stress
Range
(MPa)
500
10
5
10
6
10
7
10
8
27
Table 1 Example of Fatigue Categories used in the AASHTO Specification
Table 1 must be used in conjunction with the type of information shown in Fig. 19,
where one of the relatively large number of typical construction details classified by the
specification is shown. Application of the information is straightforward. For example,
suppose a designer proposes to use a beam made by joining three plates using continuous
fillet welds parallel to the direction of stress, such as was shown in Fig. 10. According to
Table 1 and Fig. 19, this is Detail Category B. If the number of cycles to which the beam will
be subjected is, say, 2 10
6
x , then the permissible range of stress for this detail is 120 MPa.
This number was obtained using Fig. 18 to estimate the stress range corresponding to 2 10
6
x
cycles. Working out the equation of the line for this detail category, the "exact" value of the
permissible stress range at 2 10
6
x cycles is 125 MPa. As will be seen later, the AASHTO
Specification provides information that allows the calculation of the permissible stress range
corresponding to a given number of cycles.
The fatigue strength curves
presented in the AASHTO Specification
(Fig. 18) are those corresponding to the
mean life of a detail, usually as obtained by
physical testing, shifted horizontally to the
left by two standard deviations. For
reasonably large numbers of test data, the
General
Condition
Situation
Detail
Category
Illustrative
Example
(see Fig. 16
AASHTO)
Builtup
Members
Base metal and weld metal in component, without
attachments, connected by:
• continuous full penetration groove welds with backing
bars removed, or
• continuous fillet welds parallel to the direction of
applied stress
• continuous fullpenetration groove welds with backing
bars in place, or
• continuous partialpenetration groove welds parallel to
the direction of applied stress.
B
B
B'
B'
3, 4, 5, 7
etc.
etc.
Fillet weld or
groove weld
Figure 19 Illustrative Example
for Table 1
Illustrative example 4
(refer to Table 1)
28
corresponding confidence limit is estimated to be approximately 95%.
EXAMPLE 2
The overhead crane in a small manufacturing operation uses a simplysupported crane girder
of 8 m span. The section used for the girder is to be made by filletwelding three plates into
an Ishape. The fillet welds will be continuous. The flange plates are 350 mm wide by 22 mm
thick and the web plate is 306 mm by 14 mm. The moment of inertia of this section is
448 10
6
x mm
4
. The main use of the crane will be to transport a 300 kN ladle from one end
of the shop to the other. The crane travels in such a position that the crane girder receives a
maximum 80% of the total load as a reactive force. It can be assumed that this force comes
onto the girder as a single concentrated load. Information from the owner is that the crane
will make no more than two trips per hour at this load level, this will be the only significant
load, the work schedule will not exceed 10 hours per day five days per week, and the design
life of the building is 40 years.
Is the fatigue life of this crane girder satisfactory? Use the AASHTO Specification.
Solution:
1. Number of stress cycles (equals number of load cycles, in this case) –
N = (2 cycles/hr.) (10 hr./day) (5 days/wk.) (52 wk./yr.) (40 yr.) = 208 000 cycles.
2. Detail classification – According to the AASHTO Specification, this is Detail Category B.
From Fig. 18, read the Detail Category B line at N = 208 000 cycles to find that the
permissible stress range is approximately 300 MPa.
3. Calculate actual stress range –
σ
min
= 0
σ
max
M = PL/4 = (300 x 10
3
N x 0.8) (8 000 mm)/4 = 480 x 10
6
N mm
∴ σ
max
= M y/ I = (480 x 10
6
N mm) (175 mm)/(448 x 10
6
mm
4
) = 188 MPa
Thus, ∆σ
r
= 188 − 0 = 188 MPa.
Since the actual range of stress (188 MPa) is less than the permissible range of stress
for this detail (300 MPa), the situation is satisfactory.
29
Comments:
1. The number of stress cycles is not always equal to the number of load cycles. Designers
should be alert for cases where a single passage of load produces more than one stress
cycle, as could occur, for example, when a multiple axle vehicle traverses a member or
when continuous beams are used.
2. Since stress due to dead load is always present in the member, the change in stress (∆σ
r
)
is always simply equal to the change in stress produced by the moving (i.e., live) loads.
3. Another way of looking at the problem is to compare the number of cycles that would be
permitted at the actual stress range of 188 MPa with the number of cycles that actually
occur. In this example, the number of cycles permitted by the AASHTO Specification for
a stress range ∆σ
r
= 188 MPa is N = 600 000 cycles, obtained from Fig. 18 or calculated
as will be shown in Section 3.5.
3.5 Fracture Mechanics Analysis of Fatigue
In Chapter 2, Basic Fracture Mechanics Concepts, it was put forward that brittle
fracture and fatigue are crack growth phenomenon that are characterized by the same
parameter, K. As such, it should be possible to use the fracture mechanics method of analysis
to deal with the fatigue strength problem.
The crack growth law was identified in Section 2.3 as
da
dN
A K
m
= ∆ (6)
where a = crack length
N = number of cycles
A, m = numerical constants determined from regression analysis of test
data
∆K = change in stress intensity factor corresponding to a given change in
stress range. In Section 2.3, this was also written (Eq. 5) as
∆ ∆ K WY a = σ π (5)
Integration of Eq. 6 gave the following expression for crack growth propagation (see
also Section 2.3):
N
K
m
i
f
=
1
A
da
a
a 1
∆
∫
(7)
30
where a
i
and a
f
are the initial and final crack length, respectively. Making the substitution
for ∆K, this becomes:
 
N
m
=
1
A
1
W Y a
da
a
a
i
f
∆σ π
∫
( )
( )
=
1
A
1
a
da
m
a
a
i
f
∆σ
−
−
∫
π
m
m
W Y
The terms 1/A and
( )
π
−m
are constants. Since the final crack size is always very
large as compared to the initial crack size, any term appearing with the limit a
f
can be
neglected since the limit will appear with a negative power. The terms W and Y vary with
the crack length, and of course the term a also contains the crack length. However, for a
given geometry and starting crack size, the term within the integral is also a constant.
Designating the product of all of the constant terms as M and using the more common
notation ∆σ
r
instead of ∆σ , the result can be written finally as
N M
r
m
=
−
∆σ (8)
or, alternatively, as
log N = log M − m log ∆σ
r
(9)
Equation 9 defines a sloping straight line on a plot of log stress range versus log
number of cycles. This is precisely what is observed in the physical tests. See, for example,
Figures 15, 16, and 17.
The AASHTO Specification provides values of the constant M in Eq. 8 for each
fatigue category. These values, termed "A" in the AASHTO Specification, are listed in Table
2 for each fatigue Detail Category. Also shown in the table are the values of the stress range
that identifies the horizontal portion of each curve, the socalled threshold stress. The
AASHTO Specification uses m = 3 for all fatigue categories.
Application of Eq. 8 and 9 can be made in Example 2. In step 2 of the Solution, the
permissible stress range for N = 208 000 cycles for this Detail Category B now is calculated
according to Eq. 8 as
∆σ
r
M
N
x
=

\

.
 =

\

.

1
3
11
1
3
393 10
208 000
.
= 266 MPa
31
The value of M ( 39 3 10
11
. x ) was obtained from Table 2, where it is listed according to the
Specification notation as "A." It should be obvious that it is much more expeditious to
calculate values of fatigue life or permissible stress range than to try to read them from a log
log plot (Fig. 18).
In Comment 3 of Example 2, the number of cycles permitted for ∆σ
r
= 188 MPa had
been estimated using Fig. 18 as 600 000. It now can be calculated (Eq. 8) as
( )
( ) N M = =
− −
∆σ
3 3
188 39.3x10 = 591 000 cycles
11
EXAMPLE 3
The beam whose failure surface is illustrated in Fig. 10 was one of a series of nine tested
[22]. Using the test data
1
, the regression line expressing the relationship between fatigue life
and stress range was determined to be log N = 12.32 − 2.73 log ∆σ
r
. The flaws from which
the cracks initiated were generally circular and the measured average size was 1.346 mm.
Use fracture mechanics analysis to verify the experimentally obtained regression line.
1
All the source data in this problem were expressed using U.S. Customary units. The necessary conversions
have been made.
Table 2 Constants for use with Figure 18
Detail Category
Constant, A
( ) MPa
3
Threshold Stress
(MPa)
A
82.0x10
11
165
B
39.3x10
11
110
B'
20.0x10
11
82.7
C
14.4x10
11
69.0
C'
14.4x10
11
82.7
D
7.21x10
11
48.3
E
3.61x10
11
31.0
E'
1.28x10
11
17.9
32
Solution:
It has been noted that the final size of the crack will not be influential in the result of Eq. 7.
This means that the stress intensity modifiers W and Y in Eq. 5 can be taken as a constant.
Calling the product WY C = , and substituting for ∆K in Eq. 7 gives
N
A
a da
m
r
m m
i
f
=
− − −
∫
1
2
C
a
a
∆σ
/
=
− −
−
1
1 2
1 2
A
a
m
m
r
m
m
a
a
i
f
C

∆σ
Calling m/2 – 1 = q , this can be written as
( )
N
A
a a
r
m
i
q
f
q
= −
− − −
1
C
1
q
m
∆σ
Since the final crack size is always very large as compared with the initial crack size, the
term a
f
can be neglected because it appears with a negative power in this equation. Thus, the
crack propagation equation can be further simplified to
N
A
a
m
r
m
i
q
=
− − −
1
C
1
q
∆σ
Rolfe and Barsom [8] suggest that the constant of proportionality, A, in the crack growth
equation (Eq. 6) can be taken as 218x10
13
.
−
for ferritepearlite steels. They also suggest the
value of C in the stress intensity factor expression can be taken as 2 0 . / π for a circular
crack in a plate of infinite width, and that the term m can be taken as 3, i.e., q = (m/2) − 1 =
0.5. The combined multipliers of ∆σ
r
are the term M in Eq. 8, and we can now solve for this
value. Using the alternative form, Eq. 9, we solve for log M as –
log log M
A q
m
i
q
= C a
1 1
− −
¸
(
¸
(
( ) =

\

.

¸
(
¸
(
(
−
−
−
2
1
0.5
1.346 log
.
.
1
213 10
13
3
0 5
x π
= 12.74
Thus, the equation of the fatigue strength line as obtained using the experimental data and a
fracture mechanics analysis is
33
log N = 12.74 − 3 log ∆σ
r
This is in good agreement with the fatigue strength line obtained from the experimental data
exclusively, log N = 12.32 − 2.73 log ∆σ
r
.
Comments:
Obviously, this type of exercise is not directly useful to a designer since it requires
knowledge of the type and size of flaw that is likely to result in fatigue crack growth and
failure. However, the results of this analysis and others like it do identify that the behaviour
observed in the laboratory can be substantiated by an analytical prediction—the fracture
mechanics analysis. It gives confidence in the prediction of cases not tested experimentally
when those predictions are based on presumption of reasonable starting flaw sizes and
shapes.
34
Chapter 4. Fatigue Assessment Procedures for Variable Stress Ranges
In all of the discussion so far, it has been implicitly assumed that the stress range at
the detail under investigation is unique and that counting or predicting the number of cycles
is straightforward. As might be anticipated, things are not simple in either of these
categories: fatigue loading is usually quite complex. The designer has to deal with the reality
that stress ranges of different magnitude take place at the detail and that these stress ranges
are applied for varying numbers of cycles. Methods for dealing with these problems are
outlined in this Chapter.
4.1 Cumulative Fatigue Damage
In this section, a method is presented that accounts for the damage that results when
fatigue loading is not applied at a constant harmonic amplitude. Although both linear and
nonlinear damage theories are available, the one that is customarily used in civil engineering
practice is a linear theory that is easy to understand and apply and which gives satisfactory
results. This is the linear damage rule first proposed by Palmgren in 1924 and further
developed by Miner in 1945 [23]. It is known as the PalmgrenMiner rule, and it assumes
simply that the damage fraction that results from any particular stress range level is a linear
function of the number of cycles that takes place at the stress range. The total damage from
all stress range levels that are applied to the detail is, of course, the sum of all such
occurrences. This can be written in equation form as:
n
N
i
i
∑
= 1 (10)
where n
i
= number of cycles that take place at stress range level i
N
i
= number of cycles that would cause failure at stress range level i.
The rule is obviously very simple. It has two major shortcomings [23]: it does not
consider sequence effects and it is independent of the average stress in the cycle. To at least
some degree, both of these factors are not consistent with observed behavior. However, when
residual stresses are high and when plasticity is restricted (usually the case in structural
engineering applications), it is known that these factors have only a small influence.
Moreover, the approach gives reasonable correlation with test data and it has the
considerable advantage that it is easy to use. The AASHTO Specification [2] advises that the
PalmgrenMiner rule can be used to account for cumulative damage. It should also be noted
that the term "failure" in these definitions is not intended to be taken literally. It is to be
35
interpreted as the permissible fatigue life, that is, the value represented by the mean life less
two standard deviations on the log stress range vs. log number of cycles plot.
EXAMPLE 4
The beam of Example 2 was designed, fabricated, and erected when the owner decided that,
in addition to the loads that had already been stipulated, it will be necessary for the crane to
be able to accommodate one trip per hour at a load level of 350 kN. (See Example 2 for all
other details.) Is the fatigue life of this crane girder still satisfactory? Use the AASHTO
Specification.
Solution:
1. According to Example 2, the number of cycles at the old load level was 208 000 ( = n
1
)
2. According to Example 2, the number of cycles to failure at the old load level was 588 000
( = N
1
)
3. Number of cycles at the new load level is –
N = (1 cycle/hr.) (10 hr./day) (5 days/week) (52 weeks/yr.) (40 yr.)
= 104 000 cycles ( = n
2
).
4. Number of cycles to failure at the new load level –
In Example 2, the stress range under the 300 kN loading was found to be 188 MPa. The
stress range for the 350 kN load can be calculated by proportion as (350/300) (188 MPa) =
219 MPa.
The detail that is under examination, the builtup beam composed of three plates joined by
continuous fillet welds, is a Category B detail according to the AASHTO specification.
Recalling that the AASHTO designation A of Table 2 is equivalent to the constant M in
Eq. 8, this means that the value of this constant is 393 10
11
. x . Thus, in Eq. 8, using m = 3
this becomes:
N =
−
M ∆σ
3
( )
( ) N x = =
−
cycles 393 10 219 374 162
11 3
. ( ) = N
2
Finally, checking Eq. 10:
36
n
N
n
N
n
N
i
i
∑
= 1 or,
+ =
000
588 000
+
104 000
374 162
= + 0.28 = 0.63
1
1
2
2
208
035 .
Since the total effect ("damage") of the two different stress ranges is less than 1.0, the crane
girder is still satisfactory under the new loading condition.
It is often convenient to express the MinerPalmgren cumulative fatigue damage rule
(Eq. 10) in terms of an equivalent stress range. We wish to calculate an equivalent constant
amplitude stress range, ∆σ
e
, that will display the same amount of damage as actually
produced by the variable amplitude stress ranges. Thus, using the PalmgrenMiner statement
of damage (Eq. 10), it is required that
n
N N
i
i
i
e
∑
=
n Σ
(11)
In Eq. 11, the left hand side represents the damage under the variable amplitude stress
cycles, for which the terms were defined under Eq. 10. The right hand side expresses the
damage under the constant amplitude equivalent stress range, i.e., ∆σ
e
. Each of N
i
and N
e
correspond to the number of cycles to failure—one for the variable amplitude stress ranges
and the other for the equivalent, constant amplitude, stress range.
Equation 8, which expressed the failure condition in a general way, can now be
applied to Eq. 11. Thus, N M
i i
m
= ∆σ
−
applies to the left hand side of Eq. 11 and
N M
e e
m
=
−
∆σ applies to the right hand side. Making the substitutions—
n
M M
i
i
m
e
m
∆σ
Σ
∆σ
− −
∑
=
n
i
The term M is a constant and can be eliminated from the equation. Then, solving for the
equivalent stress range
∆σ
∆σ
Σ
e
m i i
m
i
n
=
∑
n
(12)
Calling n
i i
Σ n
i
= γ , that is, γ
i
is the fraction that any particular portion of the stress range
is of the total number of cycles, then Eq. 12 is written as
37
∆σ ∆σ
e
m
i
m
=
∑
i
γ
Finally, solving for the equivalent stress range, we have
[ ]
∆σ ∆σ
e i i
m
m
=
∑
γ
1
(13)
In some specifications, for example the rules provided by the American Railway
Engineering Association (AREA) [24], the equivalent stress range ( S
Re
) is written as
[ ]
S
i Ri Re
/
= S γ
3
1 3
∑
(14)
where the symbol S
Ri
is used to indicate the stress range, ∆σ
i
. It should be clear by
inspection that Equation 14 is identical to Eq. 13, given that m = 3.
EXAMPLE 5
Use the equivalent stress method to determine the percentage of life that has been expended
by the loading applied to the beam of Example 4.
Solution:
All of the necessary data are available in the solutions to Examples 2 and 4. In summary,
these are –
n
1
= 208 000 cycles n
2
= 104 000 cycles
∆σ
1
= 188 MPa ∆σ
2
= 219 MPa
N = 208 000 + 104 000 = 312 000 cycles, and
γ γ
1
208
0 67
104
033 = = = =
000
312 000
000
312 000
2
. .
Now, using the expression given by Eq. 14 to calculate the equivalent stress range –
[ ] ( ) ( )
[ ]
S x x
re rMiner i Ri
= S = S MPa γ
3
1 3
3 3
1 3
0 67 188 033 219 1995
∑
= + =
/
. . .
For this Detail Category B and the equivalent stress range of 199.5 MPa, the number of
cycles to failure can be calculated (Eq. 8) as –
N =
−
M ∆σ
3
=
( )
393 10 494 953
11
. x x 199.5 cycles
3
=
Since the actual number of cycles is 312 000, the percentage of life expended is
(312 000/494 953) 100% = 63.0%.
38
Comments:
The same result was also seen in the solution to Example 4, where the Miner's summation
was 0.63. Whether the solution proceeds by the method shown in Example 4 (Miner's
summation) or by that shown in Example 5 (equivalent stress range method) is simply a
matter of choice.
EXAMPLE 6
A stress range histogram has been obtained from field measurement of stresses on a certain
simplespan girder bridge. This was obtained from a sample of 2064 truck crossing events.
The total truck traffic that is estimated to have taken place since the opening of the bridge is
35 x 10
6
cycles. It can be assumed that each passage of a truck causes one stress range at any
given detail. (This information is taken from Reference [25]).
The critical detail on this bridge,
and the one for which the stress
history is given, is that at the end
of a partial length coverplate. The
thickness of the flange of the
girder is 25 mm and the cover
plate is attached to the girder
flange by fillet welds.
a) Is fatigue cracking possible at
this detail?
b) Calculate how much of the
fatigue life of this detail has been used up. Use each of the two possible approaches (i.e.,
direct application of the Miner summation and use of a calculated equivalent stress range).
Use the AASHTO Specification in the solution of this problem.
Solution:
According to the AASHTO specification, this is a Detail Category E' (Consult the
specification: this information is not contained within this document.)
(a) According to Fig. 20, the maximum stress range recorded is 37.3 MPa. The threshold
stress for an E' detail is 17.9 MPa, which can be read from Fig. 18 or obtained directly
from Table 2. Since the actual stress is greater than the minimum stress above which
crack growth is possible, this indicates that fatigue cracking is possible in this girder.
events
Stress
%
55
25
Stress range, MPa
12.4 8.28 24.8 29.0 33.1 37.3 16.6 20.7 4.14
10
5
2
1 1 1
Figure 20 Example 6
39
(b) In order to calculate the fatigue life expended, it is suggested that the average stress in
each range be used.
The results of the calculations necessary to determine the remaining fatigue life are given in
Table 3. Sample calculations for the first row are shown following the table.
Table 3 Example 6
n
N
i
i
=
∑
0 60 . γ
i i
∑
= ∆σ
3
2194
Sample calculations for the first four items of Row 1 of the Table:
( ) ∆σ
1
414 828 2 6 21 = + = . . / . MPa (Refer to the stresses given in Fig. 20)
( )
( )
n x x
1
6 6
55% 35 10 19 25 10 = = cycles cycles .
( )
( )
N M
x
r 1
3
11 3 6
128 10 534x10
=
= = ≡
−
−
∆σ (Eq. 8)
6.21 cycles (M A from Table 2) .
n
N
1
1
19 25
535
0 04 = =
.
.
Since the Miner summation for all the events in this loading history is 0.60, this means that
60% of the fatigue life of this detail has been expended.
∆σ
i
average
MPa
n
i
cycles x 10
6
N
i
cycles x 10
6
n
N
i
i
γ
i
n
x
=
1
6
35 10
γ
i i
∆σ
3
6.21 19.25 534 0.04 0.55 131.7
10.3 8.75 117 0.07 0.25 273.2
14.5 3.5 42.0 0.08 0.10 304.9
18.6 1.75 19.9 0.09 0.05 321.7
22.7 0.70 10.9 0.06 0.02 233.9
26.9 0.35 6.57 0.05 0.01 194.6
31.0 0.35 4.30 0.08 0.01 297.9
35.2 0.35 2.93 0.12 0.01 436.1
40
Now, determine the fatigue life expended by calculating an equivalent stress range. The
information necessary for substitution into Eq. 13 has been generated in the last two columns
of Table 3 and recall also that Σ n
i
x = 35 10
6
cycles in this example.
[ ]
∆σ ∆σ
e i
m
m
= ( Eq. 13)
i
γ
∑
1/
( ) = = 2194 130
1 3
. MPa
Next, calculate the number of cycles that can be sustained by a Detail Category E' for an
equivalent stress range of 13.0 MPa. Using Eq. 8, and where we will substitute
∆σ
e
= 130 . MPa for the term ∆σ
r
:
( )
( )
N M
x
r
=
=
−
−
∆σ
3
3 6
583 10 = 1.28x10 13.0 cycles
11
.
Since the number of cycles that
can be sustained by the Category
E' detail is 58 3 10
6
. x cycles and
the actual number of cycles is
35 10
6
x cycles, the fatigue life
expended is 35.0/58.3 = 0.60. This
is the same as the value calculated
using the Miner summation.
Comments:
Figure 21 displays the equivalent
stress range calculation graph
ically. Note that in this example the equivalent stress range falls below the constant
amplitude fatigue limit. The calculation proceeds on the basis that all cycles are damaging,
that is, the extension of the sloping fatigue life line is used. There is a discussion of this issue
in Section 4.3.
4.2 Analysis of Stress Histories
Situations often arise where the applied loads create stress levels and stress counts
(number of cycles at a given stress level) that are much more complicated than those given in
Example 4. For example, if the crane in Example 4 is carried by a continuous beam over
Stress
Range
Number of Cycles
Detail
Category E'
Constant
amplitude fatigue
limit (17.9 MPa)
583 10
6
. x
∆σ
e
= 13.0
MPa
35 10
6
x
Figure 21 Example 6
41
several intermediate supports, more than one stress cycle is applied per trip at a given
location. This results because loading adjacent spans causes stress cycles in addition to the
cycle created when the crane passes directly over the location under examination. For this
more general case, one trip is termed a loading event and the stress variation at a given point
in the structure during such an event is called a stress history.
Very complicated stress histories can be caused by loading events such as the passage
of a truck over a bridge or a wave hitting an offshore oil platform. Furthermore, the
occurrence of a loading event brings about different stresses within different elements in a
structure, and, in addition, the number of cycles per loading event may be related to the type
and location of the element under consideration. Many different stress histories may need to
be evaluated for a complete fatigue assessment of a complex structure.
In fields other than civil engineering, elaborate stress measurements are often taken
on actual structures subjected to service loading in order to overcome uncertainties
associated with calculated values of stresses in elements. However, it is rare that such
measurements are carried out on civil engineering structures, usually because of the cost.
Consequently, civil engineers generally resort to stress analysis using load models and
dynamic magnification factors that are provided in codes and standards.
Figure 22 shows the stress
history at a certain location in a
member as a moving load passes
across a structure. Shown are the
maximum live load stress, the
minimum live load stress, and the dead
load stress. In any trace of stress versus
time there is one absolute maximum
live load stress and one absolute
minimum live load stress, of course.
The stress range between these two
extremes is shown on the figure. (It is noteworthy that the stress range is independent of the
dead load stress, and it has already been indicated (Section 3.3) that stress range is the
dominant stressrelated feature of fatigue life determination.) However, there can be other
"local" maximum and minimum live load stresses. The question is, how is the element under
examination for fatigue life influenced by these other stress excursions?
Figure 22 Typical Stress vs. Time Variation
Under a Moving Load
42
One method of interpreting the type of stress history shown in Fig. 22 is to simply use
the maximum live load stress range shown in the figure. However, as the "minor" stress
ranges increase in value, their influence becomes important at some point. Methods that are
used for counting stress ranges include the reservoir method, the rainflow method, the peak
count, and the meancrossing peak count method. The reservoir and rainflow methods are the
ones most commonly used in civil engineering applications [26, 27]. Generally, rainflow
counting is more suited to computer analyses of long stress histories, whereas the reservoir
method is most convenient for graphical analyses of short histories. The results obtained by
these two methods are the same.
The plot shown in Fig. 23 (a) is an example of stress variation in an element
subjected to a loading event. Typically, simple analyses result in stress histories having fewer
peaks than shown in this figure, whereas actual stress measurements usually reveal more
stresses due to one loading event
Stress
(a) An example of stress variation in an element
due to one loading event
22
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
1
(b) peaks and troughs numbered
for one loading event
1
2
3
4
12
20
21
22
1
(c) rainflow analysis
(d) reservoir analysis
Figure 23 Analysis of Stress Histories
43
complex stress histories.
The information shown in Fig. 23 (a) cannot be used directly in the PalmgrenMiner
rule, Eq. 10, without application of a stress counting method in order to tabulate values for
the number of cycles, n
i
, for different stress range levels. As already indicated, the most
widely used are the socalled rainflow counting method and the reservoir methods. Peaks and
troughs for one loading event are numbered in Fig. 23 (b) and this history is rotated 90
degrees in Fig. 23 (c) in order to perform the rainflow analysis.
The rainflow method is named for its analogy of rain drops flowing down a pagoda
roof. The following rules apply to rainflow counting:
1. A drop flows left from the upper side of a peak or right from the upper side of a trough
and onto subsequent "roofs" unless the surface receiving the drop is formed by a peak
that is more positive for left flow or a trough that is more negative for right flow. For
example, a drop flows left from point 1 off points 2, 4 and 12 until it stops at the end of
the loading event at point 22 since no peak is encountered that is more positive than
point 1. On the other hand, a drop flows right from point 2 off point 3 and stops since it
encounters a surface formed by a trough (point 4) that is more negative than point 2.
2. The path of a drop cannot cross the path of a drop that has fallen from above. For
example, a drop flowing left from point 3 stops at the horizontal position of point 2
because a path coming from point 2 is encountered.
3. The horizontal movement of a raindrop, measured in units of stress from its originating
peak to its stop position is counted as one half of a cycle in the stress spectrum.
The reservoir method, sonamed because of its analogy of water contained in
reservoirs formed by peaks and drained successively out of troughs, is shown in Fig. 23 (d).
To start, imagine that the area bounded by the highest peak in the loading event forms a
reservoir of water contained by it and the same peak in a following, identical, loading event.
Using the numbering shown in part (b) of Fig. 23, this is point 1. Next, drain water from the
reservoir out of the lowest trough in the spectrum, point 22 in Fig. 23 (b). Water caught
between other peaks forms smaller reservoirs. Drain water successively out of the lowest
troughs in the loading event until all reservoirs have been drained. The vertical distance,
measured in stress units, between a high water level and the drain (trough) that lowers it is
counted as one full cycle in the stress spectrum.
44
The rainflow and reservoir methods give identical results provided that rainflow
counting begins with the highest peak in the loading event, as is shown in Fig. 23.
EXAMPLE 7
The values of the peaks for the stress history shown in Fig. 23 are given below. Apply (a)
rainflow counting and (b) reservoir counting in order to identify the stress ranges in the stress
spectrum. Finally, (c) evaluate the effects of one million loading events of this stress history
acting on a beam that uses a continuous partial penetration groove weld to connect the
flanges to the web.
The given data are –
Peak / Trough No. Stress (MPa)
1 93
2 18
3 55
4 10
5 85
6 10
7 37
8 18
9 37
10 10
11 46
12 6
13 55
14 46
15 74
16 8
17 55
18 18
19 65
20 39
21 83
22 0
45
Solution: (a) Rainflow counting:
Solution: (b) Reservoir counting method:
Note that these stress ranges are the same as were determined in part a) using the rainflow
counting method (where halfcycles are reported).
From Peak or Trough No. To Horizontal Distance of
Point No.
Half Cycle
(MPa)
1 22 93
2 3 37
3 2 37
4 5 75
5 6 75
6 11 36
7 10 27
8 9 19
9 8 19
10 9 27
11 10 36
12 21 77
13 14 9
14 13 9
15 16 66
16 15 66
17 18 37
18 17 37
19 20 26
20 19 26
21 12 77
22 1 93
Drain from Trough No. Water Level at Peak Stress Range (MPa)
22 1 93
12 21 77
4 5 75
16 15 66
2 3 37
18 17 37
10 11 36
6 7 27
20 19 26
8 9 19
14 13 9
46
Solution: (c) Cumulative damage using the PalmgrenMiner rule:
According to Table 1 this is an AASHTO Detail Category B'. Table 2 indicates that the
fatigue life constant for use in Eq. 8 is 20 0 10
11
. x . (The reader should review Example 8
relative to this part of the problem after Section 4.3 has been studied.)
Stress Range
∆σ (MPa)
Fatigue Resistance
N =
( )
20 0 10
11 3
. x ∆σ
−
Damage Due to 1x10
6
Loading Events, n
i
/N
93 2 486 000 0.402
77 4 381 000 0.228
75 4 740 000 0.211
66 6 957 000 0.144
37 (twice) 39 484 000 0.051
36 42 867 000 0.025
27 101 610 000 0.010
25 128 000 000 0.008
19 291 588 000 0.003
9 2.7 x 10
9
0.000
Damage summation: n N
i
= ≥
∑
1 08 1 0 . .
Comments:
This analysis indicates that the fatigue evaluation has failed, meaning that the element may
not withstand one million loading events. At this point, the designer has several remedial
possibilities. In order of importance these are:
1. A better design detail should be investigated. Often, fatigue problems can be traced to poor
initial design detail choices. In this example, the use of continuous fillet welding rather
than continuous partial penetration groove welding would increase the fatigue strength of
this element by about 25% according to the AASHTO rules since the fatigue category
changes from Detail B to Detail B'. If this change could be made, in this example the
damage summation would be 0.55, a substantial reduction, and the detail would then pass
the fatigue evaluation.
47
2. Measures that reduce stresses due to dynamic loading should be studied. Such measures
could involve employing continuous construction, avoiding short members near
application of loads, reducing the number of expansion joints, improving road surfaces on
road bridges, and improving threedimensional loadcarrying capacity.
3. A more detailed fatigue analysis could be carried out. This is covered next, in Section 4.3.
4. Only as a nearlylast resort, the thickness of the member could be increased in order to
lower the stresses. Note that, if this is done, the fatigue strength may also decrease; see
Section 8.4.
4.3 Fatigue Limits
In Fig. 24 a given stress range
spectrum (distribution of stress ranges at a
particular structural element) is shown for
three possible cases. Case 1 refers to the
situation where all cycles are above the
constant amplitude fatigue limit (CAFL). For
Case 2, the same distribution of stress ranges
is located such that there are stress ranges both
above and below the CAFL. In Case 3, all
stress ranges in the spectrum are below the
CAFL.
The constant amplitude fatigue limit is simply that stress range level below which
there will be no fatigue crack growth under the condition of constant amplitude loading. In
laboratory testing, constant amplitude loading is the procedure most often used, primarily
because it is the easiest situation to create. Some real fatigue life cases will correspond to
this. For example, a machine part that is loaded in the same way for all of its operating cycles
corresponds to the constant amplitude loading case. On the other hand, civil engineering
applications such as bridges and offshore structures would hardly ever be loaded in constant
amplitude fatigue. Rather, as we have already seen (Section 4.2), the stress ranges that
actually act on a structural detail correspond to a case of variable amplitude fatigue. We have
seen that the use of an effective stress range (Eq. 13 or 14) is a way of expressing the
variable amplitude stress ranges as an equivalent constant amplitude stress range. The
σ
max
Stress Range
Number of Cycles
CASE 1
CASE 2
CASE 3
Constant amplitude
fatigue limit (CAFL)
σ
max
σ
max
σ
min
Figure 24 Three Cases of Variable Amplitude
Stress Spectra
48
question remains: does constant amplitude loading fairly represent variable amplitude
loading?
Studies done on the fatigue life of coverplated beams (AASHTO Detail Category E')
[28] showed that constant amplitude tests and variable amplitude tests (done at the equivalent
constant amplitude stress range) gave comparable results up to at least one million test
cycles. There are fewer data (from any sources) for stress ranges in the order of ten million
cycles upward. However, it appears that the maximum stress per se must also be considered
at these long lives.
For Case 1 shown in Fig. 24, all of the stress cycles in the spectrum contribute to
fatigue crack growth and must be considered in any fatigue life evaluation. This is how the
discussion and examples have been treated so far. Case 3 is also clearcut: because none of
the stress cycles in the spectrum are above the constant amplitude fatigue limit, no fatigue
assessment is required. (However, if a stress range in a spectrum that resulted from another
loading event does exceed the CAFL, then Case 3 spectra should be considered in the same
way as Case 2 spectra.)
Case 2 spectra are most often associated with situations where long fatigue lives are
required. For example, structures such as bridges, offshore platforms, chimneys, and
overhead highway signs are often subjected to millions of loading events. It is difficult to
guarantee that values of all stress ranges will remain less than the value for the CAFL
throughout the service life of the structure.
Two options are possible for the analysis of Case 2 spectra. First, it can be assumed,
conservatively, that all cycles contribute to crack propagation in the element from the first
cycle of fatigue loading. This is equivalent to assuming that the value for the fatigue
threshold stress intensity factor is zero. For this option, analysis proceeds in a fashion
identical to the analysis required for Case 1. In other words, all stress ranges in the spectrum
are assumed to contribute to fatigue crack growth. The analyses in Examples 5, 6, and 7
adopted this option. In effect, this option assumes that the fatigue strength curve continues
below the CAFL with a slope constant of 3.
The second option for analysis of Case 2 spectra is based on the more logical
assumption that not all cycles in the spectrum will contribute immediately to crack
propagation. This assumption is equivalent to admitting the presence of a nonzero value for
the fatigue threshold stress intensity factor. As crack propagation proceeds, more and more
49
cycles in the spectrum will exceed the threshold stress intensity factor range until all cycles
in the spectrum contribute to crack propagation at a certain crack length.
International opinion as to how the analysis should proceed in the Case 2 second
option is not unanimous. This is due to a lack of test data at sufficiently long fatigue lives so
as to be able to verify proposed analysis methods. Furthermore, fracture mechanics analyses
are hindered by difficulties in estimating values for the threshold stress intensity factor under
conditions of high tensile residual stress that vary with crack length. European practice, e.g.,
the ECCS rules [15], allow the use of a slope constant, m, of 5 for stress ranges below the
CAFL.
1
However, recent work indicates that this may be unsafe for stress spectra applied to
large beams with welded longitudinal attachments [29]. Nevertheless, test data for all other
cases investigated indicate that use of the slope constant of 5 is acceptable. It remains to be
seen whether the best rule is the first option or the second option qualified by a provision for
exceptional cases. More research and testing is needed.
The second limit discussed in this Section, the cutoff limit, is used to discard small
stress cycles from the stress spectrum. Small cycles in the spectrum do not contribute
significantly to fatigue damage: it is considered that by the time cracks grow to a length
where these stresses would induce stress intensity factor ranges greater than the threshold
stress intensity factor range, the vast majority of fatigue life has been expended.
The importance of the cutoff limit is greatest when automatic stress measuring
equipment is used to measure stress ranges. This equipment is capable of measuring several
thousand small cycles per loading event. When the number of small stress cycles (having
values less than half of the CAFL) is greater than two orders of magnitude more than the
number of large cycles, the PalmgrenMiner damage summation can produce an excessively
conservative evaluation. The Eurocode rules suggest a cutoff limit at the stress level that
corresponds to the intersection of the fatigue strength curve (m = 5) with one hundred million
cycles.
EXAMPLE 8
Part (c) of Example 7 required calculation of the cumulative damage for a given stress
spectrum. The PalmgrenMiner rule was used, and implicit in the solution was that all cycles
1
The fatigue life rules in ECCS and in the more recent Eurocode [46] are nearly identical. They are similar in
many major respects to those in the AASHTO Specification. However, a significant difference is that the
number of cycles at which the CAFL is taken to occur is a constant (5 million cycles) for all detail categories
in the European standards, whereas in the AASHTO rules the number of cycles at which the CAFL starts
varies with the detail category. The AASHTO approach is more consistent with the experimental evidence.
50
of the spectrum produce fatigue crack growth. Establish whether the situation is Case 1, 2, or
3 according to the terminology introduced in this Section.
Solution:
The detail given in Example 7 is an AASHTO Detail Category B'. Table 2 indicates that the
threshold stress for Category B' is 82.7 MPa. (The term threshold stress is synonymous with
CAFL.) According to the information given in Example 7, the maximum stress in the
spectrum is 93 MPa, and the minimum stress in the spectrum is 0 ksi.
Since the maximum stress (93 MPa) is greater than the CAFL (82.7 MPa), this means that
the stress spectrum does not correspond to Case 3 (see Fig. 24). Since the minimum stress
(zero) is less than the CAFL, the stress spectrum is not Case 1. Thus, the stress spectrum is
identified as Case 2, that is, the range of stresses within the spectrum straddles the CAFL.
Comments:
The calculations carried out in Part (c) of Example 7 counted all stress cycles as damaging.
As pointed out earlier in Section 4.3, this is equivalent to assuming that the sloping straight
line of the Detail B' category extends downward indefinitely at a slope constant m = 3 on the
log–log plot.
51
Chapter 5 Fatigue Design According to the American Association of State
Highway and Transportation Officials Specification (AASHTO)
5.1 Introduction
The discussion in the first four chapters of this document has centered around
concepts of fatigue crack growth and how to treat design tasks that are linked to loadinduced
crack growth. Distortioninduced fatigue crack growth is also an important consideration;
this is the subject of Chapter 6. The examples thus far have deliberately been kept simple, so
that the instructive points made could be developed in the mind of the reader.
In this chapter, the rules of the American Association of State Highway and
Transportation Officials (AASHTO) Specification [2] that reflect loadinduced fatigue are
examined. Chapter 6 will contain observations arising out of the AASHTO Specification that
relate to distortioninduced fatigue.
5.2 Redundancy and Toughness
Users of previous editions of the AASHTO Specification (e.g., 15
th
Edition, 1992)
will be aware that the concept of structural redundancy has been dropped in the fatigue
design rules of the 1994 Edition. In those earlier specifications, permissible stress ranges
were set forth for both redundant and nonredundant members. The stress range values for
nonredundant members were fixed, as a matter of engineering judgment, at 80% of the
values for redundant members. The rationale for this was that the consequences of failure
were greater in the former, and therefore they should be designed more conservatively. This
was a reasonable position, but, at the same time, greater fracture toughness was also
specified for nonredundant members. This constituted a double penalty, and therefore the
distinction of fatigue design with respect to nonredundant members and redundant members
has been dropped from the current (1994) edition of the Specification. (The distinction for
fracture toughness has been maintained, however. The relevant term used in that part of the
specification is "fracturecritical.")
Even though the differentiation between nonredundant and redundant members is no
longer part of loadinduced fatigue design, it is still instructive to examine the issue.
Whenever possible, and if economical, the designer should use details and concepts that
introduce "redundancy."
52
The 1992 AASHTO definition of a nonredundant member was that it was a member
whose tension failure would result in collapse of part, or all, of a structure.
1
The statement
was interpreted in a variety of ways, however. For example, continuous bridges that consist
of just two parallel girders would generally be regarded by designers as nonredundant. The
assumption is simply that such a bridge would fail catastrophically if one girder fractured.
Although this seems a reasonable position, it is contrary to experience with actual structures.
For example, the twogirder Lafayette Street Bridge in St. Paul, Minnesota carried traffic
without difficulty for several months even though a girder had fractured near the center span
inflection point [20]. Similarly, the I79 bridge across the Ohio River near Pittsburgh carried
traffic even though a crack was present that extended all the way up to the top flange [20]. (It
is not known how long the situation existed. However, traffic was carried without incident in
this condition for at least a few hours, between the time the crack was observed by the crew
on a passing tugboat and when traffic on the bridge was stopped.) In both of these structures,
their lateral bracing systems participated in carrying the load because the cracked plate girder
behaved like a torsionally stiff box.
There are at least three types of redundancy: only the first one listed is accepted by
some owners, however.
1. Statical redundancy (i.e., a statically indeterminate structure). A continuous girder
bridge is statically redundant.
2. Multiple load paths, usually as identified in the principal crosssection. This
implies that there are at least three parallel loadcarrying components such as
girder elements.
3. Internal member (crosssection) redundancy. Beams or truss members in which the
crosssectional components are fastened with rivets or bolts are considered to be
redundant internally. The severing of one component by a fatigue crack may not
lead to catastrophic failure since the other components are still intact. Of course,
instability due to the nonsymmetry of the cracked crosssection might be a
consideration in some cases. However, generally it is not until a fatigue crack
starts in second component that collapse becomes a concern.
The concept that multiple unwelded components of a beam, truss, or arch member
1
Obviously, tension failure could be the failure of a tension member, but it was intended that the definition
include the concept of the tension failure of a portion of the crosssection, e.g., the tension region of a beam
in bending.
53
builds in crack arresters and therefore improves performance has been
documented. For example, recent research has shown that the multicomponent
riveted members, which have a component fatigue category of D, can sometimes
be evaluated as Category C provided that the fasteners are tight and in good
condition [30]. However, the design specification does not yet make this
distinction. Good practice sometimes goes beyond the minimum requirements,
even though a perceived detailing improvement may not produce an advantage in
terms of the specification.
The redundancy in the case of statical indeterminacy is identified as a routine part of
the structural analysis. It is obvious that a statically indeterminate structure will be less likely
to suffer catastrophic collapse in the event of significant cracking due to fatigue. However,
the degree to which this vulnerability might exist can only be established with the aid of an
extensive, threedimensional computer analysis. Sometimes, owners are receptive to
demonstration of redundancy by this means. The objective is usually to show that a more
economical structure is also redundant, e.g., that a threegirder redesign alternative is
comparable to a fourgirder design. This requires design skill and good details.
Although a redundant system receives no benefit in the AASHTO Specification as
compared with a nonredundant system, good design practice incorporates as much
redundancy as can be justified economically.
Because of the possibility of fatigue crack formation, fracture toughness is also a
matter of concern in bridge structures. It is possible that fatigue crack growth can lead to
conditions that will trigger brittle fracture, a highly undesirable situation. Use of a steel that
has a good level of fracture toughness and avoidance of details that create triaxial stress
conditions are important to achieving the desired fatigue life and preventing premature
fracture. In addition, tough steels promote stable crack extension, which is a desirable feature
for fatigue crack inspection.
The goal is to avoid brittle fracture, which is a type of catastrophic failure. Low
fracture toughness bridge steels can result in brittle fracture under normal service loads when
cracks or cracklike geometric conditions develop. Bridge steels are strain rate sensitive,
which means that the loading rate from traffic is not as severe as the dynamic rate associated
with the standard Charpy Vnotch impact test commonly used to provide qualitative
information about the steel material. Fracture toughness is also dependent upon temperature.
As the temperature decreases, fracture toughness also decreases. Although fracture
54
mechanics provides a useful analysis tool to relate crack size and stress level to the material
fracture toughness characteristics, an accurate determination of fracture susceptibility for
structural engineering design is complicated. This is primarily because the model
characterizing combinations of stress and crack size results in plasticity levels such that the
elastic stress model becomes invalid.
Nevertheless, the Charpy Vnotch tests can provide a useful means of screening out
materials that would be susceptible to brittle fracture at small crack sizes, thereby preventing
achievement of adequate fatigue resistance.
5.3 Fatigue Design in the AASHTO Specification
In the material presented so far, it has been established that the fatigue life of a
structural steel component can be described in terms of just three elements. These are the
stress range that is acting on the detail under examination, the number of cycles to which the
detail is subjected, and a description of the fatigue life inherent in the detail.
The AASHTO rules were first introduced in Section 3.4, where the elements of the
design requirements were outlined so that examples could be carried out. In this section,
more information concerning the fatigue provisions in the AASHTO rules will be provided.
5.3.1 Fatigue Load and Frequency
The basis of the load to be used for the fatigue design of bridges according to the
AASHTO Specification is the socalled Design Truck. The Design Truck consists of three
axles. In order, these are one axle of 35 kN load followed by two of 145 kN load each. The
axle spacing between the 35 kN load and the first 145 kN load is 4.3 m and the spacing
between the two 145 kN axles is a variable quantity that can range between 4.3 and 9.0 m.
The Design Truck is a loading that applies to many different design cases. For fatigue, there
are two unique features that are applied to the Design Truck. One is that the axle spacing
between the two 145 kN axles is fixed at 9.0 m. The other feature is that the axle loads to be
used for calculation of the fatigue stress ranges are to be taken as 0.75 times the axle loads of
the Design Truck. This reflects the observation that the large number of truck passages
associated with the fatigue strength examination are commensurate with lighter trucks than
the Design Truck. These, and other related details, are in Article 3.6.1.2.2 of the
Specification.
The stress range at the particular detail under examination for fatigue strength is to be
55
evaluated according to the loading described, including also the dynamic load allowance
stipulated in Article 3.6.2 of the Specification. For consideration of fatigue, the dynamic load
allowance is designated simply as a 15% increase in the stress range calculated using the
Design Truck, modified as described above.
In general, each detail must satisfy the requirement that
( ) ( ) γ ∆ ∆ f F
n
≤ (15)
where ∆f = the live load stress range due to passage of the fatigue load. (In this document so
far, this has been designated ∆σ
r
)
( ) ∆F
n
= the nominal fatigue resistance
γ = the load factor for the fatigue load combination. The value γ = 0 75 . is to be used
for the fatigue case, as discussed above.
As we have already seen, the nominal fatigue resistance is linked to the number of
stress ranges that will be imposed. According to the AASHTO Specification, the loading is to
be taken as the singlelane average daily truck traffic, ADTT
SL
. Unless better information is
available, this quantity is to be calculated as:
ADTT p x ADTT
SL
= (16)
where ADTT = the number of trucks per day in one direction, averaged over the design life
ADTT
SL
= the number of trucks per day in a single lane, averaged over the design
life
p = a factor reflecting the number of lanes available to trucks. If only one lane is
available to trucks, p = 1; if two lanes are available then p = 0.85; if three or
more lanes are available, then p = 0.80.
The Specification provides other information regarding the loading. For instance, the
designer is advised that the average daily traffic (ADT), which includes all cars and trucks, is
physically limited to about 20 000 vehicles per lane per day. The fraction of truck traffic in
this total is estimated to be about 20% for rural interstate highways, 15% for urban interstate
or "other" rural highways, and 10% for "other" urban conditions. If sitespecific information
is not available, the usual situation, these values can be used to generate the frequency of
loading of the Design Truck.
56
5.3.2 Fatigue Resistance
Each detail must satisfy the requirement that the calculated stress range, termed ∆f in
the Specification, be equal to or less than the nominal resistance identified in the
Specification, ( ) ∆F
n
. As just discussed, the calculated stress range is that arising from
application of the fatigue Design Truck, including impact. The subscript n in the expression
for nominal resistance simply identifies that there must be a linkage with number of
cyclesthe fatigue resistance is a function of the number of cycles of the stress range.
The nominal fatigue resistance is to be taken as:
( ) ∆F
A
N
n
=
1 3 /
(17)
where A is the constant listed in Table 2 (≡ M in Eq. 8) and
( ) ( ) ( ) N ADTT
SL
= 365 75 n (18)
The constant A in Eq. 17, which is equivalent to the term M in Eq. 8, reflects the
severity of the detail, according to the Detail Categories identified in Fig. 18 and Table 2.
Given that the term ( ) ∆F
n
is simply the value of the term ∆σ
r
at its permissible value for
the given number of cycles, it can be seen that Eq. 17 is identical to Eq. 8. Equation 8 is the
fatigue life expression that was developed from a fracture mechanics approach. Thus, the
AASHTO requirement for fatigue life (Eq. 17) is consistent with the fatigue life of a given
detail as described on a theoretical basis and incorporating the results of physical testing by
the inclusion of the term A in Eq. 17.
Equation 18 is the number of cycles of stress range expected to be imposed on the
structure in a design life of 75 years. (At the option of the designer, another number can be
used in place of 75.) The term n is included in Eq. 18 as a way of acknowledging that more
than one stress cycle can result from a single passage of the fatigue Design Truck. Primarily,
this is as a result of the vibrations set up in the structure as the truck moves across. Values of
n are tabulated for 10 different cases in Table 6.6.1.2.5–2 of the Specification. For a simple
span girder, for example, n = 10 . is to be used for spans less than 12 m and n = 2 0 . is to be
used for spans greater than 12 m. For cantilever girders, n = 50 . must be used. (This high
number reflects the susceptibility of cantilevers to vibration under moving loads.) The
AASHTO Table can be consulted for other cases.
The AASHTO Specification rules provide that, in addition to Eq. 17, the following
57
requirement must also be met—
( ) ( ) ∆ ∆ F F
n TH
≥
1
2
(19)
where ( ) ∆F
TH
is the constant amplitude fatigue threshold stress. (This has been referred to
elsewhere in this document as the constant amplitude fatigue limit, CAFL.)
Viewing Equations 17 and 19 together, the latter says that the sloping straight line
represented by Eq. 17 need not extend below a horizontal straight line at the value of one
half the constant amplitude fatigue limit. According to Figure 18, it would appear that the
constant amplitude fatigue limit itself should be used. However, the requirement reflects the
fact that the stress ranges calculated according to the AASHTO rules are a simplified
depiction of the actual stress ranges that can take place. In particular, there can be occasional
stress ranges that are considerably greater than the calculated, nominal stress ranges. If even
a small number of these stress ranges exceed the CAFL, fatigue crack growth will take place.
The check using Eq. 19 determines whether or not the maximum stress range in the random
variable cycle exceeds the CAFL. It is assumed that the maximum stress range, ∆σ
max
, is
twice as great as the nominal stress range calculated from the fatigue truck, which represents
the effective stress range, ∆σ
e
. This is reflected in Eq. 19.
Figure 25 illustrates these points concerning the AASHTO rules. The AASHTO
permissible fatigue life line is the sloping solid straight line shown in Fig. 25, followed by
the dashed sloping straight line, and then by the dashed horizontal straight line. Also shown
∆σ ∆σ
max
= 2
e
∆σ
e
log number of cycles
Equation 17
Constant amplitude
fatigue limit (CAFL)
CAFL / 2
Stress range spectrum
log
stress
range
Figure 25 AASHTO Fatigue Limits
58
in the figure is a hypothetical random variable stress spectrum, plotted on the ordinate.
Shown is the equivalent stress range for the spectrum, ∆σ
e
, which in this case intercepts
Eq. 17 below the CAFL. We note, however, that the maximum stress range in the spectrum is
above the CAFL, i.e., fatigue crack growth will occur. The spectrum shown has been given
the unique relationship wherein ∆σ ∆σ
max
= 2
e
. If it is visualized that the stress spectrum is
moved down to a level such that ∆σ
e
is just below the level CAFL/2, then it is apparent that
the maximum stress in the spectrum is just below the CAFL. Thus, if the check of Eq. 19
indicates that ∆σ
e
is below the value CAFL/2, then all cycles are below the fatigue limit and
no fatigue cracking will occur.
The AASHTO rules point out (Article 6.6.1.2.1) that the force effect for the fatigue
design of a steel bridge shall be the live load stress range. This is consistent with the
discussion so far in this document. However, it is recognized that fatigue cracks will not
grow if the stress range is entirely compressive, an observation supported by both field and
laboratory experience
2
. Accordingly, a further provision in Article 6.6.1.2.1 allows the
designer to identify those regions where only a compressive stress range is expected and to
therefore eliminate them from further consideration. In accordance with the discussion just
concluded, it is again necessary to assume that the live load stress range can be as high as
twice that calculated using the AASHTO fatigue Design Truck. Thus, the statement in
Article 6.6.1.2.1 that allows the designer to reduce the areas of a bridge requiring
examination for fatigue is:
"In regions where the unfactored permanent loads produce compression, fatigue shall
be considered only if this compressive stress is less than twice the maximum tensile
live load stress resulting from the fatigue load…"
5.4 Summary of AASHTO Requirements
• Determine the number of design cycles. Use the AASHTO recommendations, or
load survey information, or information provided by the owner.
• Identify the Detail Category for each component of the structure that could be
sensitive to fatigue.
• Calculate the stress range at the details of interest.
• Determine the permissible stress range for the detail under examination and its
2
In practice, it is often difficult to ensure that stresses will always remain in compression, especially when fit
up stresses, stresses due to temperature changes, and stresses due to foundation settlement may be present.
59
number of stress range cycles. (Engineering judgment should be exercised: some
details can be eliminated from examination on the basis of comparison with
others of similar stress range and number of cycles.)
• If the actual stress range at the detail exceeds the permissible value, modify the
design and repeat the assessment.
5.5 Design Example
The following design example will be used to elaborate the fatigue requirements of
the AASHTO Specification. The problem is derived from Design Example 2 of Reference
[31]. In that example, the 1977 AASHTO Specification rules were used and the details were
laid out in US Customary units. Because the illustration that follows includes calculation of
the stresses and other design aspects that are not a central part of this document, reference
will be made from time to time to the specific articles of the Specification. This will enable
the reader to use that source for more information.
EXAMPLE 9
Design Information:
Twospan continuous girder bridge, 42.7 m spans, girder spacing at 2440 mm c/c,
four parallel girders
Rural interstate highway, two lanes of traffic in same direction
Unshored composite construction (positive moment region only)
Concrete slab: f
c
'
. = 20 7 MPa , modular ratio E E
s c
= 10
Reinforced concrete slab, 203 mm thick, plus a 12 mm thick integral wearing surface
Welded plate girder, ASTM A588 steel, F
y
= 345 MPa.
Design Assumptions:
For purposes of this example, assume that the given section (Fig. 26 and Table 4) is
satisfactory for noncyclic loading and that all steel meets the fracture toughness
requirements. Of course, both of these issues would be a normal part of the design
procedure.
Concrete over Field Section 1 (see Fig. 26) always acts compositely with steel girder.
Concrete over Field Section 2 never acts compositely with steel girder (i.e., no shear
studs).
60
Problem: Investigate the suitability of an interior girder with respect to fatigue.
Solution:
CrossSection Properties.
Table 4 lists the dimensions of the steel plate girder and the section properties calculated
from those dimensions. Sufficient information is given such that interested readers can
confirm the calculations if desired. These section properties are used subsequently to
calculate stress ranges. In a similar fashion, Table 5 lists the section properties of the
composite crosssection. For these calculations, it is necessary to identify the effective flange
width of the composite crosssection. This is obtained using Article 4.6.2.6.1 of the
AASHTO Specification, where, for an interior girder, the effective slab width is the least
of —
0.25 x effective span length (to point of permanent load inflection)
= 0.25 x 30 000 = 7 470 mm
or, 12 x average thickness of slab + greater of web thickness or 0.5 x top flange width
= (12x203.2)+(0.5x304.8) = 2 590 mm
or, the average spacing of adjacent beams = 2 440 mm Governs
61
Top Flange Stress Range
0
5
10
15
20
25
30
35
40
45
0 10 20 30 40
Distance along girder, m
Stress
range
MPa
Bottom Flange Stress Range
0
5
10
15
20
25
30
35
40
45
0 10 20 30 40
Distance along girder, m
Stress
range
MPa
Field Section 1
Field Section 2
Center Line
Station
→
0
30.1 37.9 42.7
42.7
30.1
7.8 4.8
No fatigue check req'd top flange
No fatigue check
req'd bottom flange
NOTE: All plate sizes given in Table 4
Figure 26 Plate Girder Details and Stress Range Diagrams
62
TABLE 4 –Properties of Steel Section
Plate Girder Only
END MIDDLE CENTER Notes:
Web
h (mm) 1524 1524 1524
w (mm) 11.11 11.11 11.11 END refers to Field
Section 1
( )
A mm
w
2
16 932 16 932 16 932
Top Flange
MIDDLE refers to the
b (mm) 304.8 381.0 609.6 portion of Field Section 2
t (mm) 15.88 25.40 31.75 that is away from the pier
A mm
f
( )
2
4 839 9 677 19 355
Bottom Flange
CENTER refers to the
b (mm) 457.2 381.0 609.6 portion of Field Section 2
t (mm) 23.81 25.40 31.75 that is over the pier
A
f
(mm
2
)
10 887 9 677 19 355
Neutral Axis (xx) (from web centerline, +ve up)
A mm
total
(
2
)
32 657 36 286 55 641
Top Flange: Ay (mm
3
)
3 725 497 747 082 15 055 615
Bottom Flange: Ay (mm
3
)
8 425 575 7 497 082 15 055 615
ΣAy (mm
3
)
4 700 079 0 0
y (mm)
143.9 0.0 0.0
Moment of Inertia (xx)
Web: wh
3
/12 (x10
6
mm
4
) 3 277 3 277 3 277 (web, about own axis)
Web: A y
w
( )
2
(x10
6
mm
4
)
350.7 0.0 0.0 (web, about girder
centroid)
TF: A h y
f
( / ) 2
2
− (x10
6
mm
4
)
3 971 5 619 11 238 (top flange, about girder
centroid)
BF:
A h y
f
( / ) 2
2
+ (x10
6
mm
4
)
4 159 5 619 11 238 (bottom flange, about
girder centroid)
Σ I
x
(x10
6
mm
4
) 11 758 14 515 25 754
63
TABLE 5 – Properties of Composite Section
Composite Section
Composite Section Note: Composite action develops in End Section
(Field Section 1) only
Modular ratio n (= E
s
E
c
/ ) 10
Concrete Thickness (mm) 203.2
Effective Concrete Width
(mm)
2 438
Effective Concrete Area
( ) mm
2
495 402
Equivalent Area of Steel
( ) mm
2
49 540
Composite Neutral Axis
(xx)
(measured from neutral axis of steel girder, +ve up)
A
total
(mm
2
) 82 198
Concrete: Ay (mm
3
) 50 699 188
Steel Girder: Ay (mm
3
) 0
ΣAy (mm
3
) 50 699 188
y (mm) 616.8 (from neutral axis of steel girder alone, +ve up)
or y (mm) 472.9 (from web centerline, +ve up)
or y (mm) 406.6 (from centroid of concrete slab, +ve up)
Composite Moment of
Inertia (xx)
Girder: (x10
6
mm
4
) 11 758
Girder: A y
girder
( )
2
(x10
6
mm
4
)
12 424
Concrete: (x10
6
mm
4
) 170.5
Concrete, A y
equiv
( )
2
(x10
6
mm
4
)
8 190
ΣI
x
(x10
6
mm
4
) 32 543
64
Number of Load Cycles
According to Eq. 16 (which is Eq. 3.6.1.4.2–1 of the AASHTO Specification), the number of
trucks per day in a single lane is
ADTT p x ADTT
SL
=
where p = 0.85 (AASHTO Table 3.5.1.4.2–1) since in this bridge two lanes are available to
trucks. The average daily truck traffic (ADTT) will be taken as 20% (rural interstate
highway) of the physical limit of 20 000 vehicles per lane per day. Thus,
( ) ADTT p x ADTT
SL
= = 085 0 20 20 . . x x 000 = 3 400 trucks / lane / day
According to Eq. 18 (AASHTO 6.6.1.2.5–2), the number of stress cycles is then—
( ) ( ) ( ) N ADTT
SL
= 365 75 n , or
N x = = ( ) ( ) ( ) . 365 75 3 400 93075 10
6
n n cycles / lane over an assumed 75year life.
In this expression, the term n reflects the number of stress cycles for a given truck passage.
According to AASHTO Table 6.6.1.2.5–2, for a continuous span bridge of 42.7 m the value
is n = 15 . for details located within L/ . 10 43 = m, or n = 10 . otherwise. (In fact, the Table
simply differentiates as "near interior support" or "elsewhere." However, Commentary to the
Table suggests that L/10 can be used to define locations near the interior support. In this
example, and since the two spans are of equal length, details within 4.3 m on each side of the
interior support can be categorized as "interior." )
Finally, it can be concluded that for locations within 4.3 m of the interior support,
N x x lane = 93075 10 93075 10 10 10
6 6 6 6
. . , , / n cycles / lane = x 1.5 =139.6x say 140x cycles ,
and for all other locations N x lane = 93075 10 10
6 6
. / n cycles / lane, say, 93x cycles .
Load Assumptions and Calculations
The Fatigue Design Truck is the Design Truck (as
shown in the sketch), multiplied by the load factor of
0.75 (AASHTO Table 3.4.1–1), and applying an impact
allowance of 0.15 (AASHTO Article 3.6.2.1).
The load distribution factor (i.e., the fraction of a truck in a given lane that accrues to an
interior girder) will be taken as 0.40. This is obtained from Article 3.6.1.4.3b and from
145 kN 145 kN 35 kN
4.3 m 9.0 m
65
Article 4.6.2.2.2b. In the calculation of the load distribution factor, the longitudinal stiffness
of the slab–girder system is involved. According to Article 4.6.2.2.2b, this term, K
g
, can be
taken as unity for the purpose of a preliminary analysis. Using this value, the load
distribution factor is 0.398, say 0.40. The "exact" analysis, obtained using the section
properties of Field Section 1, gives a load distribution value of 0.408. Obviously, in this
problem these results are sufficiently close to one another that either could be used.
Furthermore, in a socalled exact analysis, the section properties at the location where the
fatigue truck is placed on the influence line for moment at a given detail should be used. This
will vary, depending upon which detail is being examined. Considering the level of
confidence likely in the load distribution factor calculation, these refinements seem
unwarranted.
Calculation of Stress Ranges
It was pointed out earlier in this Chapter (Section 5.3.2) that fatigue need be considered only
when the dead load stresses (unfactored load) are compressive and are less than twice the
live load tensile stresses. It is advantageous for the designer to take advantage of this
provision in order that the bridge length over which fatigue needs to be examined is reduced.
In this example, the dead load of the steel section is estimated to be 2.5 kN/m for Field
Section 1 and 3.0 kN/m for Field Section 2. The initial dead load (unfactored) due to the
concrete slab and girder selfweight, applied to the girder alone, is 12.40 kN/m. The
additional dead load, including wearing surface, applied to the composite section, is 5.84
kN/m (unfactored). These values are used in conjunction with the fatigue live load to
calculate the stresses in the top and the bottom flanges of the steel section. In accordance
with AASHTO Article 6.6.1.2.1, this examination establishes that no fatigue check is
required between bridge stations 0 and 29.1 m for the top flange and between stations 33.3 m
and 42.7 m for the bottom flange. (The centerline of the twospan structure is bridge station
42.7 m).
Figure 26 shows the live load tensile stress range for the top and bottom flanges of the steel
section. The regions that do not have to be examined for fatigue are shown as broken lines in
this figure. This information can now be used to evaluate the fatigue life of the various
details. (Space does not permit tabulation of the stress ranges at discrete locations, but this
would be available to the designer, of course.)
66
Fatigue Life Assessment of Flange to Web Weld
A continuous fillet weld is used to join the flange to the web. According to AASHTO Fig.
6.6.1.2.3–1 and Table 6.6.1.2.3–1, this is Detail Category B. For Category B, the fatigue life
constant is A x = 393 10
11
. MPa
3
(Table 6.6.1.2.5–1) and the constant amplitude fatigue
threshold is 110 MPa (Table 6.6.1.2.5–1). Therefore, the fatigue strength according to Eq. 17
(AASHTO Eq. 6.6.1.2.5–5) is first calculated as
( ) ∆F
A
N
x
N
n
=
=
1 3 11
1 3
393 10
/
.
It has already been established that for locations within 4.3 m of the interior support (Sta.
38.4 m), N = 140x10
6
cycles. Everywhere else, N x = 93 10
6
cycles should be used.
Substitution of these values gives ( ) ∆F
n
= 30 4 . MPa when N = 140x10
6
cycles and
( ) ∆F
n
= 34 8 MPa . when N x = 93 10
6
cycles.
Application of Eq. 19 also says that the permissible stress range need not be less than one
half of the constant amplitude fatigue threshold, however. In this case, the value is
110 MPa 2 55 MPa = . This governs in both cases since it is larger than both 30.4 MPa and
38.4 MPa. (An alternative way to think of Eq. 19 would be to multiply the calculated stress
by a factor of 2, which then provides the estimated maximum stress range. This should be
less than the CAFL.)
The actual stress range for the top flange in the region Sta. 29.1 to 38.4 m is 44.9 MPa (see
Fig. 26). The actual stress range for the bottom flange in the region Sta. 33.3 to 42.7 m is also
44.9 MPa. Since the actual range of stress for this detail is less than the permissible value of
55 MPa, the flangetoweb fillet weld detail is satisfactory with respect to fatigue.
Comment: Details of Category A or B are seldom critical with respect to fatigue. With
experience, it should be possible to eliminate most cases by inspection. In fact, AASHTO
Article 6.6.1.2.3, which is the introductory statement setting out the fatigue requirements,
refers only to components and details with fatigue resistance less than or equal to Detail
Category C.
Fatigue Life Assessment of Flange Plate Splices
Figure 26 shows that the girder flange plates must be spliced at Stations 30.1 and 37.9 m.
Both these locations are located in the regions wherein N x = 93 10
6
cycles. The plates
change in both width and thickness at each location. The detail category is dependent upon
67
just how these transitions are made. If it is assumed that the transition slopes are not greater
than 1:2.5 and that the groove weld reinforcement is not removed, then Detail Category C is
applicable. According to the AASHTO rules, the fatigue life constant is 14 4 10
11
. x MPa
3
and
the constant amplitude fatigue threshold stress range is 69 MPa for this category. Thus, the
fatigue strength is the greater of 69 2 345 MPa = . or
( ) ∆F
A
N
x
x
n
=
=
=
1 3 11
6
1 3
14 4 10
93 10
249
/
.
. MPa
At Sta. 30.1 m, the maximum tensile stress range is 31.8 MPa (bottom flange) and at Sta.
37.9 the stress range is 25.6 MPa (either top or bottom flange). Both of these values are less
than the permissible value of 34.5 MPa. The groove weld at the transition in flange plate
sizes is therefore satisfactory in fatigue.
Fatigue Life Assessment of Transverse Stiffeners
Transverse stiffeners are present in a girder in order to stiffen the web and assist in carrying
the shear forces. If this is their only function, they will be welded or bolted to the web and
fitted tight against the underside of the compression flange. For economy, especially in the
fitting process, these transverse stiffeners will be cut short of the tension flange (Fig. 27).
When transverse stiffeners are present, it is
common practice to use them as the
connecting plate for diaphragms or cross
frames. If this is the case, then the stiffener
must be welded or bolted to each flange of the
girder. Often, the stiffeners are also used for
the connection of lateral bracing. Once again,
the transverse stiffener must be connected to
both flanges if it is used to receive lateral
bracing. Information on these requirements is
contained in AASHTO Article 6.6.1.3 and Article 6.10.8.1.1.
It will be assumed in this example that some of the transverse stiffeners will have diaphragms
or crossframes connected to them and some will not. This is a realistic circumstance for
most bridges. Later in the example there will also be a discussion about the connection of
lateral bracing members to the transverse stiffeners.
When the stiffener will not to be fitted to both flanges, then the end in the tension flange
t
w
= 11.1 mm
4 t
w
min
6 t
w
max
Assume 10
mm fillet
welds
Figure 27 Stiffener Detail
68
region must be cut short of the flangetoweb weld a distance not less than four times the web
thickness and not more than six times the web thickness. (AASHTO Article 6.10.8.1.1.) This
requirement relates to distortionbased fatigue and is discussed in Chapter 6.
According to Fig. 27, and assuming that the flange to web weld will be 10 mm, the maximum
gap (6 t
w
) will require that the end of the stiffener be ( ) 6 111 10 77 x mm . + = from the nearer
side of the flange. This information is required in order to be able to calculate the stress at the
toe of the stiffenertoweb fillet weld.
In accordance with AASHTO Fig. 6.6.1.2.3–1 and Table 6.6.1.2.3–1, the fillet weld at the
end of a transverse stiffener is a Category C' detail. The fatigue life constant is
14 4 10
11 3
. x MPa and the constant amplitude threshold stress limit is 82.7 MPa. Thus, the
fatigue strength of this detail is the greater of 872 2 413 . / . = MPa or
( ) ∆F
A
N
x
N
n
=
=
1 3 11
1 3
14 4 10
/
.
When N x = 93 10
6
cycles , the result is ( ) ∆F
n
= 249 . MPa and when N = 140x10
6
cycles
( ) ∆F
n
= 218 MPa . . For both these regions, then, the permissible stress range is 413 . MPa . In
Field Section 1, the maximum tensile stress range in the bottom flange is 37.7 MPa (Fig. 26).
Figure 28 shows the stress gradient at the detail in this region. The stress range at the critical
location is calculated as ( ) 1159 1260 37 7 34 7 x MPa . . MPa = . Since the permissible stress in
this region is 41.3 MPa, the transverse stiffeners in Field Section 1 are satisfactory with
respect to fatigue. It should also be obvious, with experience, that in this example the stress
range at the critical location in the transverse stiffener need not have been calculated. The
stress range at the level of the bottom flange (37.7 MPa) is already less than the permissible
stress range (41.3 MPa).
In Field Section 2, the permissible stress range is again calculated as 41.3 MPa (see
calculation above). The actual stress range in this zone is a maximum of 44.9 MPa (Fig. 26).
A calculation of the stress range at the extremity of the stiffener (similar to the calculation
illustrated in Fig. 28) will identify that the stress range at this location is 41.3 MPa. As it
happens, this is identical to the permissible stress range. Thus, any transverse stiffener in
Field Section 2 that does not extend to the tension flange is satisfactory with respect to
fatigue.
69
Undoubtedly, some transverse stiffeners will have to extend from flange to flange and be
fastened to those flanges. This is because diaphragms will be present, and the most
convenient way of attaching them to the girders is to use the transverse stiffeners. Thus, these
stiffeners will extend all the way to the flange and be welded to it. (Bolting is also a
possibility, if desired.) This is a Category C' detail, just as was the stiffener welded only to
the web. Accordingly, the permissible stress range is 41.3 MPa, as calculated for the
stiffeners that stopped short of the flange. In Field Section 1 the maximum stress range is
37.7 MPa for the tension flange and in Field Section 2 the maximum stress range is
44.9 MPa. By inspection, the condition in Field Section 1 is satisfactory. In Field Section 2,
the stress range at the end of the stiffener (i.e., at the level of the inner surface of the flange)
will be ( )
[ ]
1260 24 1260 449 44 0 − = x MPa MPa . . . See Fig. 27. Since the permissible stress
range is only 41.3 MPa, this means that the detail is not satisfactory. The designer can either
accept the small amount of underdesign (about 6%) or ensure that this type of transverse
stiffener is not located anywhere in the region of Field Section 2 where the flange stress
exceeds ( )
[ ]
413 1260 24 1260 421 . / . MPa MPa − = . According to Fig. 26, this will be only a
very short portion of the girder to the right of bridge station 30.1 m. Review of the calculated
stress (not given in the example) reveals that the stress range has decreased to 42.1 MPa at
bridge station 31.4 m.
Fatigue Life Assessment of Lateral Bracing Attachment
Lateral bracing can be attached to the girder in different ways. A horizontal plate can be
fastened to the web or to the flange by welding or by bolting. The lateral bracing members
77
24
1260
N.A.
1159
Critical location
Figure 28 Stress Gradient at Transverse Stiffener
37.7 MPa
70
are then bolted (usual case) or welded to this connection plate. (In the AASHTO
Specification, this is referred to as a "lateral connection plate.") If transverse stiffeners are
present and the lateral connection plate is to be located within the web depth, then it must be
attached to the stiffener when on the same side of the girder as the stiffener or centered on
the stiffener otherwise. (AASHTO Article 6.6.1.3.2). Although the use of lateral connection
plates within the web depth and centered on transverse stiffeners is a detail that has been
widely used in the past, it will be seen that it presents a number of difficulties related to
fatigue.
The fatigue category for this type of detail will depend upon the length of the plate (in the
direction of stress), the treatment of the weld ends (no treatment, transition radius on the
weld, or transition radius and welds ground smooth), and the thickness of the plate.
For the example girder, try a square plate that is less than 25 mm thick, and no special
finishing of the weld. According to AASHTO Table 6.6.1.2.3–1, this will be Category E
when the plate length is greater than either 12 times the plate thickness or 100 mm. It is
likely that the connection plate will have to be longer than either of these values for proper
attachment of the bracing members.
Category E is a severe fatigue detail. The fatigue life constant is 361 10
11 3
. x MPa and the
constant amplitude threshold stress limit is 31.0 MPa. Thus, the permissible fatigue stress is
the greater of 31.0/2 = 15.5 MPa or
( ) ∆F
A
N
x
N
n
=
=
1 3 11
1 3
361 10
/
.
When N x = 93 10
6
cycles , the result is ( ) ∆F
n
=15 7 MPa . and when N = 140x10
6
cycles
( ) ∆F
n
=137 . MPa . For convenience, simply take 15.5 MPa as the permissible fatigue stress
in both regions.
In Field Section 1, the stress gradient diagram is that shown in Fig. 28. In order to not exceed
the permissible stress of 15.5 MPa, a calculation will show that the horizontal plate must be
located 735 mm from the bottom. This is not a practical solution; the lateral bracing should
be much closer to the plane of the lower flange. Try a similar detail except that there will be
a transition radius >50 mm at the end of the plate and the welds ends will be ground smooth.
According to Table 6.6.1.2.3–1 the lateral connection plate is now a Category D detail. The
fatigue life constant is 7 21 10
11 3
. x MPa and the constant amplitude threshold stress limit is
71
48.3 MPa. Calculation will show that the criterion 48.3 MPa/2 = 24.1 MPa will govern
throughout. The necessary location of the horizontal plate is now 455 mm above the bottom
of the girder. Although this is a considerable
improvement over the previous condition, it still
places the connection plate a considerable distance
from the lower flange. Furthermore, in this girder it
is inevitable that some connection plates will have
to be placed around the transverse stiffeners. The
situation is shown in Fig. 29. Although the extreme
ends of the horizontal plate could have a radius and
the welds ground, the interior "ends" of the plate
near the transverse stiffener could not receive this
treatment. Thus, the condition reverts to that of a
square plate.
The type of lateral bracing attachment plate that attaches to a transverse stiffener must also
meet a distortionbased fatigue requirement. This topic is discussed in Chapter 6 and the
fatigue design of this lateral bracing attachment is completed in Example 11.
All things taken into account, the designer should consider simply attaching the horizontal
connection plate to the lower flange. According to AASHTO Fig. 6.6.2.3–1 and Table
6.6.2.3–1, if the horizontal connection plate is made the same thickness as the flange plate, a
transition radius of between 150 and 600 mm is used at the ends, and the welds ground
smooth, then this is a Category B detail. The fatigue life constant is 393 10
11 3
. x MPa and the
constant amplitude threshold stress limit is 110 MPa. Thus, the permissible stress range for
this detail is the greater of 110 2 550 / . = MPa or
( ) ∆F
A
N
x
N
n
=
=
1 3 11
1 3
393 10
/
.
When N x = 93 10
6
cycles , the result is ( ) ∆F
n
= 34 8 . MPa and when N = 140x10
6
cycles
( ) ∆F
n
= 30 4 . MPa . For both these regions, then, the permissible stress range is 550 . MPa . In
the bottom flange, the maximum tensile stress range is 37.7 MPa in Field Section 1 and
45.0 MPa in Field Section 2 (Fig. 26). Thus, this connection detail is satisfactory for use
anywhere along the length of the girder. A highstrength bolted connection is also a
possibility. It is likewise a Category B and the connection would be much simpler than the
bracing
forces
bracing forces
girder web
transverse
stiffener
horizontal
connection
plate
Figure 29 Transverse Stiffener and Horizontal
Connection Plate
72
welded one. However, the effect of the bolt holes on the strength of the flanges would have
to be taken into account.
Fatigue Life Assessment of the Shear Studs
In this girder, Field Section 1 is composite and will therefore have shear studs. Field Section
2 is noncomposite and so no studs will be present. Positive bending moment predominates
in Field Section 1. However, as Figure 26 identifies, there is also a region of positive stress
range (which results from loading in the opposite span). The examination of dead load stress
plus two times the live load stress identifies that the region from Sta. 0 to 29.1 is always in
compression, however, and so no fatigue check is required. This leaves the region between
Sta. 29.1 and 30.1 m (end of Field Section 1) in which a fatigue check is required for the
studs. Studs are a Detail Category C (AASHTO Table 6.6.1.2.3–1), which has a relatively
high fatigue life. The largest stress range in the short portion of the girder that must be
checked is only 8.1 MPa (Fig. 26). It will be taken by engineering judgment that the studs are
satisfactory in fatigue.
Fatigue Life Assessment of Bolted Splice
Figure 26 implies that the girder is spliced at Sta. 30.1 m, i.e., at the junction of Field
Section 1 and Field Section 2. The actual splice will likely be made just to the left of that
location, however, in order that filler plates will not be required for the flange splice plates.
The actual location of the center of the splice will depend upon the length of the splice plates
required for the flanges. Assume that the splice will be centered at Sta. 29.0. From inspection
of Fig. 26, the bottom flange stress range controls the design. The detailed calculations show
that the stress range in the bottom flange is 33.2 MPa at location 29.0 m.
The AASHTO Specification designates Category B for bolted connections that are slip
critical. The stress is to be calculated on the gross crosssection, which is the value 33.2 MPa
cited above. The splice is located in the region wherein N x = 93 10
6
cycles.
The flangetoweb weld was also a Category B, and the calculations done earlier for that
detail identified that the permissible stress range is 55 MPa. Since this is greater than the
actual stress range, the bolted splice will be satisfactory with respect to fatigue.
This example has dealt with the major elements of fatigue life assessment for the
welded plate girder. As the details are completed, the designer should be alert for other
situations that require a fatigue life assessment.
73
Chapter 6 DistortionInduced Fatigue Cracking
6.1 Introduction
Most of the topics so far have discussed the effect of stresses acting on preexisting
flaws, cracks, and geometrical discontinuities with respect to fatigue lives of fabricated steel
elements. The assumption has been that these stresses can be calculated, usually at an
elementary level. The loads used are the same as those associated with the strength design of
the members. In many instances, however, fatigue crack growth results from the imposition
of deformations, not loads. Although it is possible in some of these cases to calculate a stress
range, this is usually performed after the fact and requires that field measurements be made.
Designers are not likely to be able to identify the need for such calculations in the course of
their work. As will be seen, this type of fatigue crack growth results from the imposition of
relatively small deformations, usually outofplane, in local regions of a member. These
deformations are not anticipated in the design process. The main defense against this source
of cracking is proper detailing, and this, in turn, is dependent on experience.
6.2 Examples of DistortionInduced Cracking
An example of the phenomenon is illustrated in Fig. 30. Standard practice for many
years was to cut transverse stiffeners short of the bottom (tension) flange so as to avoid a
severe detail for fatigue if the stiffener is welded to the flange at that location
1
. (There are
also practical reasons for cutting the
stiffener short: the stiffener will have to
be made to a precise length if it is to
extend from flange to flange.) The
height of the gap between the end of the
stiffener and the girder flange is usually
quite small. If lateral movement of the
top flange relative to the bottom flange
takes place, large strains are imposed in
the gap region because of the
significant change in stiffness between
the stiffened and unstiffened (gap)
1
Experience gained over the past 20 years has shown that, in fact, the fatigue life of the detail is independent of
whether the stiffener terminates in the web or is extended down to the flange. This is reflected in current
specifications. The only difference, then, is the effect of the stress gradient.
Detail at Bottom of Transverse Stiffener
Girder and Transverse Stiffener
crack
Fig. 30 Fatigue Cracking from OutofPlane Movement
74
regions of the web. Typically, the strains that are produced are so large [32] that it may take
relatively few cycles for a crack to propagate. The flange movement could be the
consequence of transverse forces in a skew bridge, but it could even be due to shipping and
handling.
The detail in Fig. 30 shows the crack emanating from the weld toe at the bottom of
the stiffener. Often, the crack will also extend across the toe of the fillet weld at the
underside of the stiffener and for some distance into the web. Up to this stage, the crack is
more or less parallel to the direction of the main stress field that the girder will experience in
service. Thus, if the source of the displacementinduced fatigue can be identified and
eliminated, then further growth of the crack is unlikely. However, if crack growth has gone
on for some time, the crack may have turned upwards or downwards in the web and thus be
aligned in the most unfavorable orientation with respect to service load stresses.
The detail just described (Fig. 30) has been the source of many fatigue cracks in the
past. New designs accommodate the situation in different ways, reducing the possibility of
fatigue crack growth that is induced in this manner.
Another illustration of a case in
which outofplane movement can
produce fatigue cracks is shown in
Fig. 31, where a floor beam is attached to
a vertical connection plate that is welded
to the web of a girder. Under the passage
of traffic, the floor beam will rotate as
shown. As this rotation occurs, the
bottom flange of the floor beam lengthens
and the top flange shortens. Lengthening
of the bottom flange will not be
restrained because it is pushing into the
web of the girder, which is flexible in this
outofplane direction. However, because the top flange of the girder is restrained by the
deck slab, shortening of the top flange of the floor beam can only be accommodated by
deformation within the gap at the top of the connection plate. This type of deformation is
shown (exaggerated) in the detail in Fig. 31.
θ
∆
Detail at
gap
Girder
Floor
beam
Fig. 31 Floor BeamtoGirder Connection
75
The behavior illustrated in Fig. 31 has been confirmed by field measurements taken
by Fisher [32]. Moreover, the field study showed that each passage of an axle caused a
significant stress range at the top of the connection plate. In this situation, fatigue cracks
could develop either at the weld at the top of the connection plate or at the webtoflange
fillet weld of the girder, or both. The residual tensile stress in this small gap will tend to be
very high because of the proximity of the two welds. It can be expected that fatigue cracks
could occur under relatively few cycles of load, although of course the fatigue life will
depend largely upon the deformation, ∆, that actually takes place as a result of the rotation of
the floorbeam.
EXAMPLE 10
Given:
Figure 31 illustrated a floor beam to girder connection where out
ofplane cracking could occur. The detail is repeated here, as
Fig. 32. The web plate thickness is t = 12 mm, the length of the
gap is L = 15 mm, and a measurement of the displacement of the
gap shows that ∆ = 0.004 mm. What is the fatigue life of this
detail?
Solution:
According to the AASHTO Specification, the webtoflange fillet
weld is a Detail Category B for the inplane bending stresses ("continuous fillet weld parallel
to the direction of applied stress"). However, the weld is Category C when the outofplane
displacement depicted in Fig. 32 takes place ("filletwelded connection with weld normal to
the direction of stress"). The fillet weld that is used to attach the connection plate to the web
is a Detail Category ′ C ("transverse stiffener to web weld"). For the outofplane
displacement under examination, the stress range at the webtoflange weld and at the
stiffenertoweb weld will be identical. Therefore, Category ′ C will govern the design.
Application of the momentarea principle, or reference to a design handbook, will show that
the moment at the end of a fixed ended beam when it is displaced transversely an amount ∆
at one end is
M
L
=
EI 6
2
∆
∆
t
L
Fig. 32 Example 10
76
Thus, the stress is σ
I
=
6 EI
L
t
2
1
I
=
3 E t
L
2 2
=
M y ∆ ∆
Since the displacement ∆ varies each time the floorbeam rotates, the stress so calculated is
really the stress range, ∆σ. Thus, solving for ∆σ ( ≡ σ ) and using the specific values given:
( ) ( ) ( ) ( )
( )
∆σ =
3 200 000 MPa 0.004 mm 12 mm
15 mm
= 128 MPa
2
The number of cycles that a Detail Category ′ C can sustain a for stress range of 128 MPa
can now be calculated. From Table 2, the constant amplitude fatigue limit (threshold stress)
is 82.7 MPa for a Detail Category ′ C and the value of the fatigue life constant is
14 4 10
11 3
. x MPa . Because the actual stress range is greater than the CAFL/2 = 41.4 MPa,
the fatigue life is calculated according to Eq. 8:
( )
( ) N cycles = M = 14.4x10 128 =
3 11 3 3
∆σ 687 x 10
Comment:
This example shows that even a very small amount of outofplane displacement can have a
significant effect on fatigue life.
6.3 Further Examples of DistortionInduced Cracking
The distortioninduced fatigue conditions that can be present in fabricated steel
members was introduced in Section 6.2 and used in Example 10. Section 6.3 provides more
general comments and gives further examples for bridge structures.
6.3.1 Web Gaps in Multiple Girder and Girder Floor Beam Bridges
Diaphragms and crossframes in multibeam bridges are used to provide torsional
stiffness to the structure. The diaphragms or crossframes are connected to the longitudinal
members by means of transverse connection plates, and this often provides fatigue conditions
similar to the floor beamtogirder connection plate discussed in Section 6.2. The connection
is usually made to transverse stiffeners that are welded to the girder web. In the past, it was
customary that no connection be provided between the stiffener and the tension flange
because this would adversely affect the fatigue detail category of the girder. Sometimes these
stiffeners were not attached to either flange. Since adjacent beams deflect differing amounts
under traffic, the differential vertical movement produces an outofplane deformation in the
77
web gap at the stiffener ends if they are not attached to the beam flange. The magnitude of
this outofplane movement depends on the girder spacing, amount of bridge skew, and type
of diaphragm or crossframe.
Various types of diaphragms are used, ranging from rolled sections to simple X
bracing made of angles. Figure 33 shows the underside of a multiple girder bridge that has an
Xtype crossframe bracing system. In the web gap, cracking developed in the negative
moment region (i.e., top flange in tension) of this continuous span structure. This cracking is
illustrated in Fig. 34, which is a view along the length of the bracing toward the girder and its
transverse stiffener. The transverse stiffener to which the Xbracing is attached was not
welded to the top (tension) flange for the reasons described earlier. This permitted outof
plane displacement to occur in the web gap and, consequently, cracks formed along the web
toflange fillet weld and at the top of the connection plate.
Fig. 33 XBracing and Girders
78
Fig. 34 Fatigue Cracks in Web Gap of Diaphragm Connection Plate
One of the earliest and most common sources of fatigue cracks in welded bridges is
the cracking in the web gaps at the ends of floor beam connection plates. (See also Section
6.2.) These cracks have occurred in the web gap near the end reactions when the floor beam
connection plate was not welded to the bottom (tension) flange. However, the most extensive
cracking is observed in the negative moment regions of continuous girder bridges where the
connection plate is not welded to the top (tension) flange of the girder. One such case is
illustrated in Fig. 35. The view is toward the web of the girder. Cracking is seen along the
fillet weld toe at the webtoflange weld (horizontal crack) and at the top of the stiffener
(vertical crack). The illustration shows that the floor beam has been bolted to the transverse
stiffener and that the top of the stiffener has a small corner clip where it reaches the top of
the girder web. The clip is provided so that the stiffener clears the girder webtoflange fillet
weld. Thus, even though the stiffener extends to the underside of the girder flange, it is not
fastened to the flange and the presence of the stiffener clip creates the gap in which
deformation was concentrated.
79
Fig. 35 Fatigue Cracks at Floor Beam Connection Plate Web Gap
6.3.2 Web Gaps in Box Girder Bridges
Internal diaphragms in various types of box girder structures are a source of web gap
cracking as a result of crosssection distortion. This type of cracking has been seen both in
continuous and simple span spread box girder structures.
An example of this type of cracking is shown in Fig. 36. The structure is an elevated,
curved continuous span structure. Two curved steel box girders with internal diaphragms
support the reinforced concrete deck. Fatigue cracking occurred in the top web gaps
(negative moment region) near the piers and in the bottom web gaps (positive moment
region) at the diaphragm locations. In both cases, the cracking was the result of outofplane
movement in the web gap. This resulted from the box girder distortion and the resultant
diaphragm forces.
The cracking that occurred at the top web gap is shown in Fig. 36. The photograph,
which was taken inside the girder, shows the (sloping) girder web in the lefthand portion of
the figure, a transverse connection plate welded at right angles to it, and a gusset plate
(horizontal) that formed part of the lateral bracing used for shipping and construction. The
gusset plate is about 80 mm below the top of the connection plate and there is a small gap
between the top of the transverse connection plate and the underside of the girder flange.
(The photograph shows that a loose plate has been placed in this gap above the connection
plate. It was later fastened in place in order to prevent further movement of the girder web in
80
this region.)
Fig. 36 Web Gap Cracking at Box Girder Diaphragm Connection Plate
Fatigue cracking has also been observed in the bottom web gaps at the locations of
internal diaphragms in simple span steel box girders. An example of this type of cracking is
shown in Fig. 37. These cracks are also the result of outofplane distortion in the web gap.
The web gap displacement results primarily from torsional distortion of the box girder when
eccentric loads are applied to the structure.
81
Fig. 37 Web Gap Cracking at End of Transverse Connection Plate
6.3.3 Long Span Structures
A number of tied arch structures have been built in which the floor beams are framed
into the arch tie girder with web shear connections alone. No direct connection is provided
between the floor beam flange and the tie girder. In older structures, the floor beam end
connections were riveted. In such instances, the double angles are riveted to the floor beam
web and the outstanding angle legs are riveted to the tie girder. In more recent arch design, a
transverse welded connection plate on the tie girder webs is utilized. If this is the case, then
the floor beams are bolted to the welded transverse connection plate.
Cracks have been observed in the floor beams of structures with either riveted end
connections or welded transverse connection plates. The cracks form in the floor beam web
along the webtoflange connection at the floor beam web gap. This gap is present between
the end of the connection angles or at the end of the welded connection plates. The cracks
extend parallel to the floor beam flange along the length of the web gap and then begin to
turn and propagate toward the bottom flange. Figure 38 shows a crack in the web gap of the
floor beam just above the end connection to the tie girder. The photograph also shows a
drilled hole at the lefthand end of the crack. As will be discussed in Chapter 8, this is often
an effective way of stopping the further growth of a crack.
82
Fig. 38 Floor Beam Web Gap Crack at End Connection to a Tie Girder
83
Similar cracks have been observed in long span trusses where floor beams are
connected to the bottom chord panel points.
Cracks have also developed in box tie girders at internal diaphragms. The fatigue
cracking occurred in tie girder web gaps (created by internal diaphragms) at the floor beam
locations as a result of the web gap outofplane distortion. The web gap distortion is the
result of the bending of the floor beams and the relative movement between the two tie
girders. This creates diaphragm forces that torsionally deform the tie girders.
6.3.4 Coped Beam Connections
In order to facilitate the easy connection of one flexural
member to another, the flange of one of the members is often cut
back, as is illustrated in Fig. 39 (The detail shown was used
extensively in the past in throughgirder railway spans.) In other
cases, the flanges may simply be narrowed: this is called a
"blocked" beam. Fatigue cracking at coped beam locations is not
so much related to distortioninduced fatigue as it is illustrative of
cracking at a location where the calculated stress is zero.
In the case of coped beams, either the top or bottom
flange, or both, may be coped. The cope is generally made by
flamecutting the material, and experience shows that workmanship is often unsatisfactory.
The radius of the cope may be small and the cutting done unevenly. In addition to the
potential for fatigue cracking created by such workmanship, the flamecutting process can
leave a region of hardened and brittle material adjacent to the cut.
The coped end of the beam is at a location of theoretical zero stress since the
connection must necessarily be one that does not transmit moment and the shear force is
carried by the web. Nevertheless, the region of the cope will have stresses due to bending
because of the restraint at the connection. There are many examples of fatigue cracking at
cope locations [20], and the best solution in the case of new designs is to avoid copes
entirely. If copes must be included, execution of the work and inspection must be of a high
standard. Specific information on the fatigue life of a coped beam can be found in
Reference 34.
fatigue
crack
Figure 39 Bottom Flange of
Floorbeam Coped at
Connection to Girder
84
6.3.5 Connections for Lateral Bracing
The connection of lateral bracing to girders should be done as close to the plane of
the bottom flange as possible. (In this discussion, it is assumed that the top flange will be
braced by the deck.) Sometimes, the connection between the lateral bracing system and the
girder will be to a horizontal plate welded to the girder web. (In the AASHTO Specification,
this is termed a "lateral connection plate.") If transverse stiffeners are present, usually the
case, the arrangement will be as
shown in Fig. 40.
In the illustration, the
horizontal connection plate is fitted
around the vertical transverse
stiffener and welded to it where
they are in contact. In other
arrangements, the vertical stiffener
passes clear through a slot in the
horizontal plate without attachment
at that location. In either event, it is
highly likely that a gap will be left
in the region shown in order to
avoid intersection of the horizontal
and vertical welds. Consequently,
as the bracing forces push into and
pull on the web at this location, the web will rotate about a vertical axis formed by the back
of the vertical stiffener and the plane of the web. Because the web is very flexible in the out
ofplane direction, this causes large strains in the web in the gap region when the lateral
connection plate is not attached to the transverse stiffener. The region is also a zone of high
residual stresses because of the proximity of the vertical and horizontal welds. Lack of fusion
in the weld or other microdefects can also be anticipated at the weld terminations. Taken all
together, these conditions mean that the type of detail shown is very susceptible to fatigue
crack growth [25, 33].
6.3.6 Other Examples
Many end connections are designed as "simple," that is, it is assumed that no moment
is transmitted. However, even simple connections do carry some moment, and this means
bracing force
bracing force
girder web
transverse
stiffener
horizontal
connection plate
(lateral connection
plate)
gap
Fig. 40 Connection of Lateral Connection Plate
85
that the connection elements will deform under the moment. End rotation at web framing
angles causes the angles to deform and load the rivets, bolts, or welds in ways not
contemplated by the design. If the loading is cyclic, fatigue cracks may develop in the angles
themselves or in the bolts, rivets, or welds. In the cases of the mechanical fasteners, prying
forces develop that may cause fatigue cracking under the heads or nuts.
Another example of distortion concentrated in a small space is when a beam or girder
flange is loaded directly. The most obvious example of this is the case of railway beams or
girders. Particularly in older bridges, it was common to place the ties directly on the top
flange of the flexural members in the floor system. Thus, a stringer flange, for example, can
be rotated as traffic passes over the structure. If the flange is too flexible in this direction,
cracking can occur at the webtoflange junction.
6.4 AASHTO Specification Requirements Relating to DistortionInduced Fatigue
Based on the discussion and examples thus far in this Chapter, it can easily be
appreciated that it is difficult for a specification or design standard to provide very much in
the way of explicit rules relating to distortioninduced fatigue. The AASHTO design
specification provides a separate section on distortioninduced fatigue (Article 6.6.1.3). It
contains a general statement that stresses the importance of proper connection of transverse
components to longitudinal (i.e., main) components. Subsections then offer specific
information relating to transverse connection plates (Article 6.6.1.3.1), lateral connection
plates (Article 6.6.1.3.2), and orthotropic decks (Article 6.6.1.3.3).
Article 6.6.1.3.3 simply directs the reader to Article 9.8.3.7, which gives detailing
requirements for orthotropic decks. Those requirements are mainly a reflection of good
practice and experience derived from this type of deck construction. The articles relating to
connection plates provide rules for this important topic of girder detailing. As has already
been illustrated in this Chapter, cracking resulting from improper detailing of connection
plates is a significant source of fatigue cracking in bridges. Example 11 will illustrate some
of the features of this section. Article 6.6.1.3 also alerts the user to the possibility of fatigue
cracking as a result of excessive outofplane flexing of a girder web. This will be illustrated
in Example 12.
86
6.5 Design Examples
EXAMPLE 11
Given:
In Example 9 (Chapter 5), a lateral connection plate was examined for its loadinduced
fatigue life. (See "Fatigue Life Assessment of Lateral Bracing Attachment" in that example.)
One possible solution was to center a horizontal plate around a transverse stiffener, as shown
in Fig. 29. According to Example 9, the plate had to be located 455 mm above the bottom of
the girder in order to meet the requirements of the loadinduced stresses. Is this arrangement
satisfactory with respect to distortioninduced fatigue?
Solution:
The AASHTO requirements for distortioninduced fatigue of a lateral connection plate are
given in Article 6.6.1.3.2. If the connection plate is not attached to the flange, the preferable
solution, then in the case of a stiffened web it should be attached not less than onehalf the
flange width above or below the flange. This minimum distance requirement is to ensure that
distortioninduced fatigue cracking does not take place in the web between the lateral
connection plate and the flange. The maximum flange width in this girder is 609.6 mm (see
Example 9, Table 4). Thus, the AASHTO requirement is that the horizontal connection plate
be at least 305 mm above the flange. The distance identified as satisfactory for loadinduced
fatigue, 455 mm, therefore is satisfactory also for distortioninduced fatigue.
Comment: Recall that this stiffener arrangement was not the preferred solution, however. See
Example 9.
EXAMPLE 12
Given:
Does the web of the girder used in Example 9 (Chapter 5) meet the AASHTO distortion
related fatigue requirements for webs?
Solution:
The possibility of fatigue due to web flexing is first cited in AASHTO Article 6.6.1.3, but
this simply refers the reader to Article 6.10.4. The requirements for the case of flexure,
Article 6.10.4.3, are given only for girder webs that have no longitudinal stiffeners. (If
longitudinal stiffeners are present, it is considered that web flexing will not occur to any
significant extent.)
87
For flexural stresses, if
2
576
D
t
E
F
c
w yc
< . , then it is required that f
cd h yc
≤ R F
In these expressions, D
c
is the depth of the web in compression, t
w
is the web thickness, E
is the modulus of elasticity of steel, F
ys
is the yield strength in the compression flange, and
f
cd
is the maximum compressive stress, due to the fatigue load, in the compression flange.
The term R
h
is a flange stress reduction factor. For homogeneous girders, its value is unity.
(The AASHTO Article 6.10.4.3 actually gives a threepart rule, of which only the governing
one for this example has been presented here.)
The expression on the righthand side of the first inequality is equivalent to the allowable
web slenderness given in the allowable stress design provisions of the 1992 AASHTO
allowable stress design specification. In other words, when this requirement is met the web is
stocky enough that it will remain elastic under working loads. Note that the requirement
deals with stress, not stress range as is done generally for fatigue life checks. According to
Article 6.10.4.2, the flexural stress resulting from the fatigue load is to be taken as twice the
calculated value. This is consistent with the requirements explained in Section 5.3.2.
The depth of the web in compression will be greatest in Field Section 2, where the girder is
noncomposite. According to Table 4, D
c
= = 1524 2 762 mmand t
w
= 111 . mm. The yield
strength of the compression flange material is 345 MPa. Thus—
2 762
111
200 x
.
5.76
000
345
< or, 137 139 <
Since the inequality requirement is met for the quantities associated with this girder, then the
compressive stress (resulting from the fatigue load) in the flange can be as large as the yield
strength of the flange. Recall, however, that the compressive stress is to be taken as twice
that calculated for the fatigue load.
In this girder, calculations show that the maximum compressive stress in the compression
flange due to two times the fatigue load is 149 MPa. Since this is less than the permissible
(345 MPa), fatigue cracking due to web flexing is not a concern.
Comments:
1. An experienced designer would probably recognize that this girder has a relatively stocky
web and that flexural web buckling leading to fatigue is unlikely. Based on the
88
calculations already done, it will be concluded here, without calculation, that fatigue
resulting web buckling due to shear will likewise not be a concern. The AASHTO rule
for this case is presented in Article 10.6.4.4.
2. It should be recalled that the possibility of fatigue cracking due to web flexing was also
the subject of Section 6.3.5. In that case, the web flexing was local—the region in the
vicinity of a transverse stiffener and a horizontal connection plate was under
examination.
6.6 Summary
Outofplane distortion that is concentrated in small web gaps remains a large source
of fatigue cracking in bridge structures. It develops in nearly every type of bridge structure
including trusses, suspension bridges, plate girders, box girders and tied arches. It is
fortunate that most of the cracks that develop from local distortion lie in planes parallel to the
loadinduced stresses. In addition, since stress intensities around distortioninduced cracks
may decrease with increasing crack length, cracks can slow, and even stop, once a certain
flexibility has been provided. As a result, distortioninduced cracks have not caused
significant numbers of fractured flanges or hampered the loadcarrying capability of the
bridge member in which they form.
88
Chapter 7 Inspection and Repair of Fatigue Cracks
7.1 Introduction
It is the responsibility of the engineer to evaluate the fatigue life of an existing
structure, and then, if necessary, to advise on a protocol for inspection. If the inspection
reveals that cracking is present, then the responsibility will extend to recommendations for
remedial action. The previous chapters have given the scientific foundation for the fatigue
life evaluation. This chapter provides recommendations directed toward helping the engineer
devise a plan for inspection and then to decide what to do if a crack is discovered. Ideally,
decisions should be based upon a scientific understanding of crack propagation, combined
with knowledge gained through practical experience of examining steel structures and
dealing with cracks found in those structures. Often, when a crack is found in a steel
structure more careful inspection will reveal additional cracks in similar elements. If no
action is taken to eliminate the cause of cracking, more cracks usually appear at other
locations. For these reasons, the discovery of a crack should be taken seriously by the
inspection team and reported immediately to the engineer responsible for the structure. Quick
fixes, carried out by untrained personnel, often worsen the problem.
7.2 Protocol for Fatigue Crack Investigation
Based on the forgoing, a reasonable protocol for fatigue crack investigation is as
follows:
• Carry out a remaining fatigue life analysis based on loadinduced fatigue. If this
analysis identifies details that have little or no remaining fatigue life, these locations
are then candidates for an examination when a physical inspection of the structure is
carried out.
The principles of the remaining fatigue life assessment have been covered in earlier
chapters of this document. For important structures, it may be appropriate to carry out
field measurement of strains. This will provide a better estimate of loadinduced
stresses. Global strains calculated using the conventional methods of analysis are
usually conservative. At the same time, local strains are often not even identified by
the traditional methods of structural analysis.
• Use the shop drawings of the members that make up the structure to help identify
those details that are susceptible to distortioninduced fatigue.
89
• If a physical inspection of the structure is indicated, this should be carried out by
trained personnel. (Methods of inspection are discussed later in this chapter.)
• If cracks are located, consider appropriate repair methods and advise on inspection
procedures subsequent to the repairs.
7.3 Identifying the Causes of Cracking
Cracks in a structure may be identified as a result of an examination that starts with
fatigue life assessment, as described in Section 7.2, but they may simply be seen in a routine
inspection of the structure. The first decision that must be made following the discovery of
cracks is whether the structure is safe in its existing condition. The examination at this point
can lead to one of several options.
• The structure is deemed to be unsafe for the intended use and it must be shut down.
• The structure is deemed to be safe providing that load levels are reduced.
• Because of redundancy, the structure is deemed to be safe at existing load levels even
though cracks have been identified. It is possible that, although remedial action is not
required, continued monitoring of the structure is indicated.
In those instances where the structure is not shut down for use, an attempt should be
made to determine the causes of cracking before any corrective action is taken. Although this
course of action seems obvious, often it is not followed in practice. Fatigue crack growth
mechanisms are activated by factors such as large numbers of stress ranges, severe stress
concentrations, dynamic magnification (impact), outofplane displacements, corrosion
damage, inappropriate fabrication procedures, bad welding processes, and large defects.
Frequently, more than one factor contributes to a critical situation. It is important that the
repairs account for the causes of the cracking.
The parameters that effect the stress ranges at a given location can be difficult to
quantify. For example, the effect of truck passages over a worn expansion joint at the end of
a bridge may be impossible to evaluate without strain gauge measurements. Stress ranges
brought about by temperature changes and other types of deformationinduced stresses
(sometimes called secondary stresses) present similar problems. Occasionally, signs of high
stresses, such as wear marks on contact surfaces or visible outofplane movement, indicate
that the structure is subjected to high stress ranges. Often, the most accurate evaluation
possible can not do more than provide a list of candidate factors ordered according to their
importance. In some instances, examination of a crack segment removed from the structure
can assist a specialist in the determination of the most probable causes. See reference [25].
90
Knowledge of stress ranges is usually related to the accuracy of estimates of the
number of fatigue cycles applied to the structure. If possible, this information should be
checked against recommendations for fatigue strength in the design specification in order to
determine whether cracking would be expected under such circumstances. Note that if
cracking is visible, it may be assumed that there is no remaining fatigue life at that location,
and therefore a check can be performed using standard relationships between stress range
and total fatigue life. The reason for making such a check is that if fatigue cracking cannot be
explained in this way, all sources of fatigue loading may not have been identified. The fact
that the structure has survived without fatigue failure for a given period may be the only
piece of information available that is completely certain.
When fatigue loading information and code recommendations cannot explain fatigue
cracking, other factors not covered by fatigue strength curves—such as large defects due to
poor fabrication methods, outofplane movements, and the effects of corrosion—may have
influenced cracking. Special requirements exist for the fabrication of steel structures
subjected to fatigue loading. For example, eccentricities due to fitup that may be acceptable
for some structures might weaken a structure subjected to fatigue loading. The structure
should be examined for evidence of large eccentricities, corrosion damage, and fabrication
induced discontinuities. Note that lack of weld penetration which is oriented transversely to
the direction of fatigue loading is equivalent to an initial crack. (See Figure 5.) These
discontinuities are the result of poor fabrication and quality control.
7.4 Cracking at Low Fatigue Strength Details and Defects
Frequently, cracking in welded bridge structures is the result of i) flaws that escaped
detection, ii) the use of details more severe than assumed in the design, or iii) secondary and
displacementinduced stresses. (Except for secondary and displacementinduced stresses,
comparable conditions do not develop in riveted and bolted bridge structures.) Most critical
cracks exist in tension areas. When details are located in compression stress regions and no
possibility of stress reversal exists, there is a reduced fatigue risk. Under these conditions,
crack growth will be confined to the residual tensile stress zone of the detail unless outof
plane deformations occur.
It is appropriate to examine further some low fatigue strength details that have been
or are being used in steel bridges. Those design details that fall into AASHTO Category E or
E' are particularly susceptible to fatigue crack growth in steel bridges. The lowest fatigue
rating, Category E', is seen in coverplated beam bridges that have flange thickness greater
than 20 mm. Although the use of coverplated steel beams has almost disappeared from use
91
in new structures because of its low fatigue rating, many existing bridges have these beams.
However, details that are equivalent to the coverplated beam continue to be used. These
include certain longitudinally or transversely loaded groovewelded or filletwelded gusset
plates or other attachments.
Figure 41 shows the bottom flange of a rolled steel shape to which a cover plate has
been attached by welding. A fatigue crack is present in the rolled section. It started at the toe
of the fillet weld at the location of the cover plate termination. In addition to the usual
conditions of potential crack growth at the toe of a fillet weld, there will be a significant
stress concentration at this location of abrupt change in crosssection. In this case where a
transverse end weld is present, it is likely that multiple initiation sites are present. They
eventually link together as the small individual cracks grow and coalesce to form a single
large crack. Even if the transverse weld were not present, fatigue cracking from the
termination of the longitudinal welds (or the longitudinal weld end returns) is highly likely.
Figure 41 Fatigue Crack at End of Cover Plate
Another low fatigue strength detail is the welded web gusset plate, such as the one
illustrated in Fig. 42. These are particularly susceptible to crack growth when they are
adjacent, but not attached, to transverse stiffeners and connection plates. In the figure, the
horizontal gusset plate is the dark area at the top of the photo. A transverse (vertical) web
stiffener passes through a rectangular hole in the gusset plate. A fatigue crack is seen
emanating from the fillet weld region near the junction of the rectangular hole and the girder
web. As the crack moves downward, it is intercepted by a retrofitted hole intended to stop the
crack from moving into the lower flange of the girder.
92
Figure 42 Fatigue Crack at Web Gusset
The fatigue crack shown in Fig. 42 was the result of combined cyclic stresses—those
from the expected inplane stress and those from the unexpected outofplane stress in the
web gap. The latter stress develops from the lateral forces that cause the gusset plate to twist
and deform the web gap. This phenomenon was discussed earlier: see Section 6.3.5.
The use of low fatigue strength details such as coverplated beams and welded web
and flange gusset plates should be avoided on bridge structures that will experience large
numbers of stress cycles during their design life. If cover plates are used on high volume
bridges, they should extend full length of the girder.
There are several other Category E or E' details that are susceptible to fatigue
cracking at weld terminations. Typical of these structures and cracking is the small crack
shown in Fig. 43. It originated at the end of a longitudinal fillet weld that was used to attach
a 0.9 m long lateral connection plate to the underside of the floor beam flange. (The
longitudinal weld is located at the junction of the top side of the connection plate and the
edge of the flange. In Fig. 43, the edge of the floor beam flange is the element with the
descriptive marking "FB 28 S.") The cracks were small, and it was possible to retrofit these
details by grinding and peening the weld terminations.
Large initial defects and cracks that result from poor quality welds are a primary
source of fatigue crack growth. In many cases, however, it is not appreciated that the welding
of socalled secondary members or attachments must meet the same weld quality criteria and
nondestructive test requirements as for main members. Even though the secondary member
does not carry calculated stress, its welded connection to a stresscarrying member means
93
Figure 43 Fatigue Crack at Lateral Gusset Plate Welded to Floor Beam Web
that stress concentrations, initial defects, and so on directly influence the fatigue life of the
main member. An example of a detail this category is a splice in a longitudinal girder web
stiffener. The longitudinal stiffener fulfills its
purpose without carrying significant stress. If the
welding of the stiffener to the web is not expected
to meet high quality control and inspection
requirements because it is a secondary attachment,
fatigue cracking can be the result.
The girder web shown in Fig. 44 is the
fascia girder of a ninegirder fourspan bridge on an
Interstate highway [20]. The bridge was
approximately nine years old when the fatigue crack
was discovered. The fracture initiated at an unfused
groove weld that was used to join sections of the
longitudinal stiffener. It grew both upward (to about
the middepth of the girder) and downward. The
lower end of the crack had penetrated the bottom
flange of the girder. A second crack was discovered
later in the same girder. It had also started from an
unfused groove weld in the longitudinal stiffener,
Figure 44 Crack in Girder Web from Splice
in Longitudinal Stiffener
94
but had only severed the stiffener and not yet penetrated the web when it was discovered.
A condition related to that just described occurs when backing bars are used to make
a groove weld between transverse stiffeners and a lateral gusset plate. Lackoffusion often
exists adjacent to the girder web in the transverse groove welds. If the transverse welds
intersect the longitudinal welds, they provide a path for a crack to enter the girder web.
Figure 45 LackofFusion Defect Adjacent to Backing Bar at Gusset Stiffener Weld
Cracks that develop in the web at lateral connection plates generally start at the
location of intersecting welds. The horizontal gusset plate used to connect diaphragm and
lateral bracing members to a longitudinal girder is often slotted to fit around a transverse
stiffener. In the U.S., one of the first structures to exhibit this cracking was the Lafayette
Street Bridge over the Mississippi River at St. Paul, Minnesota [20]. Figure 45 shows the
junction of the horizontal gusset plate (foreground) with girder web (righthand side of
photo) and the vertical transverse stiffener. The connection between the horizontal gusset
plate and the vertical stiffener was made using a groove weld founded on a backup bar.
However, the groove weld did not penetrate to the backup bar, particularly near the girder
web. This meant that there was a significant lack of fusion in a plane perpendicular to the
main stress field in the girder. Since this weld also intersected both the vertical weld
attaching the stiffener to the web and the longitudinal weld attaching the gusset to the web, a
path was provided into the girder web. Examination of the fracture surface showed that a
brittle fracture occurred after the fatigue crack propagated into the web. The final condition
of the girder was that the crack had completely severed the lower flange and had stopped
95
about 150 mm above the horizontal stiffener, which itself was about 150 mm above the
flange.
7.5 Methods for Inspection for Fatigue Cracking
In Section 7.2, it was stressed that it is desirable that any inspection for fatigue
cracking be preceded (if possible) by an analysis of the structure and by identification of
details prone to outofplane cracking. Inspection is usually a difficult and expensive process,
and these preliminary steps will help to focus the inspection on the highrisk regions of a
structure. A knowledge of where cracking is likely to occur in a detail that has been
identified as prone to fatigue damage is likewise an important adjunct to the inspection
process itself. Clearly, the goal is that the inspector be able to concentrate on those details
that are candidates for cracking and thereby be able to direct resources to the locations where
fatigue cracking is likely to occur. It is unusual to be able to inspect every part of a given
structure for fatigue cracking: time and money do not permit this except in the case of
specialized structures.
The next step is to identify the inspection methods that are both appropriate for the
expected manifestation of cracking and are costeffective. As will be seen, some inspection
tools are better for picking up certain types of fatigue cracks than are others. The inspection
methods or tools that are in common use today include visual inspection (sometimes at low
power magnification), ultrasonic inspection, magnetic particle inspection, and use of acoustic
emission techniques.
Before cleaning the surfaces to be examined, a visual inspection is essential. Often
the most reliable sign of fatigue cracks is the oxide (i.e., rust) stains that develop after the
paint film has cracked. Similarly, oxides can be seen on unpainted weathering steel surfaces
when cracks develop. They are visually apparent because they differ in appearance from the
protective oxide coating that has formed normally on the weathering steel. Magnification
(10x) is helpful, as is supplementary lighting. Experience has shown that cracks have
generally propagated in depth to between onequarter to onehalf the thickness of the part
before the paint film is broken, which permits the oxide to form. Cracks smaller than this are
not likely to be detected visually unless the paint, mill scale, and dirt have been removed by
blastcleaning the suspect area. At weld terminations, very small cracks are difficult to detect
by visual inspection even on cleaned surfaces unless the crack depth is between 3 and 6 mm.
It should also be kept in mind that the inspection usually takes place when the live load
stresses are small. This means that an existing crack will not be held open by the stress field.
96
The following inspection methods are useful:
• Dye penetrant inspection. This is the most common and least expensive inspection
method. A low viscosity, high capillary fluid (usually a red dye) is used to penetrate
into surface cracks. After cleaning and drying, a second material, called a developer
(usually white), is sprayed on in order to draw, by capillary attraction, the penetrant
fluid from cracks. When the developer is applied to the opposite face, through
thickness cracks can be detected. When used correctly, this method is reliable for
detection of surface cracks emanating from smooth surfaces. However, for cracks
growing from weld toes, it is often difficult to distinguish between a crack and the
plate/weld toe interface.
• Magnetic particle inspection. When a magnetic field in a steel plate encounters a
discontinuity, the field becomes distorted. Fine magnetic particles are sprayed or
dusted on the object in order to detect such distortions. Clusters of particles form at
crack tip locations. As with dye penetrant inspection, this method works most
effectively on smooth surfaces. Accuracy is reduced when used on weld surfaces, but
reasonable results can be obtained by an experienced inspector. Even subsurface
cracks can be detected if the depth below the surface is not too great. Equipment is
simple and portable.
• Ultrasonic inspection. This method uses high frequency sound waves at a pitch
beyond that audible to the human ear. These waves are picked up by a receiver and a
crack will cause a distortion in the waves received. This method can be used for any
thickness of plate over 3 mm and, once calibrated, it can detect small cracks whether
they are embedded or not. Variations in surface roughness affect accuracy. The
orientation and size of the crack are not always easy to visualize and, therefore, use of
this method requires considerable experience.
• Acoustic emission. This method of assessment uses the high frequency sound wave
that is generated by crack extension. Sensitive piezoelectric sensors convert the very
high frequency vibrations (several hundred kilohertz) generated by a propagating
fatigue cracks into electrical signals that are detected and analyzed. Most available
systems are not directly engineered for bridge monitoring, however. In addition,
relatively large cracks must be present for the system to be effective.
• Xray Inspection. Under controlled conditions, radiography can be used to identify
the presence of defects and cracks. It is not particularly helpful when the actual sizes
97
of the crack are desired, however. Its primary use has been to evaluate groove welds
where lackoffusion may exist.
• Coring. An effective way to establish the type of defect or crack is to core and
remove a segment of the crack or discontinuity. This permits a destructive assessment
of the section within the cores. In this way, the actual size of the crack can be
established and this can often be used to calibrate and interpret ultrasonic and
radiographic results. After the core has been removed, the opening should be ground
smooth and then inspected with liquid penetrant to ensure that no crack extends
beyond the hole. In some cases, a highstrength bolt is inserted into the core opening
and preloaded.
7.6 Repair of FatigueCracked Members
Repairing a cracked element is an obvious option, but other possibilities should also
be considered. Appropriate measures could also include implementation of damage tolerance
strategies and element replacement. When damage tolerance strategies are used, no
immediate action is taken to repair the crack or replace the element. (However, often holes
are drilled at the crack tip in order to minimize crack extension.) Repair can be delayed,
thereby making it possible to avoid repairs at inconvenient times. Furthermore, this strategy
often provides the opportunity to determine the causes of cracking more precisely. However,
damage tolerance should not be employed when rupture of the element would cause general
collapse of the structure; when continued cracking would endanger human life through
increased deformations (for example, an increased risk of derailment due to cracking of a
railway bridge girder); when monitoring crack growth at regular intervals is not possible; or
when further cracking may substantially increase repair costs.
In order to ensure that a damage tolerance strategy is successful, additional measures
are recommended. Fracture mechanics analyses should be employed to predict crack growth
and to determine critical crack lengths, i.e., crack lengths at which fracture would be
expected. Laboratory testing of specimens (Charpy Vnotch, CT, etc.) made from steel taken
from the structure may be needed. Crack measurement equipment should be calibrated and
verified on a regular basis. Also, any modifications to the structure or changes to the applied
loading should be assessed in order to verify that the structure can safely tolerate further
cracking.
Repair of cracks, when carried out correctly, may result in an adequate solution to
fatigue cracking provided that the steel quality is appropriate for both present and future
service conditions. However, when repairing fatigue cracks, it is important to keep in mind
98
that repair welding is rarely successful, and therefore weld repair should be employed as a
last resort when no other solution is possible. The following is a list of measures that have
been employed successfully for steel structures containing fatigue cracks. These measures
are ordered (roughly) from best to worst.
• Place cover plates over the crack in order to provide a load path for forces and to
restrict movement of the crack surfaces during fatigue loading. Preferably, these
cover plates should be placed on both sides of the cracked plate, and they must be
attached with pretensioned highstrength bolts. If this solution is adopted, it is
essential to ensure that the fatigue crack growth has stopped. Crack growth may be
stopped by changes to the structure itself (reduce the stress range, stop outofplane
movement, etc.), or in some cases it may be done by drilling a hole at the end of the
existing crack. The latter procedure is discussed following.
• Drill a hole at the end of the crack and fill the hole with a pretensioned highstrength
bolt. The bolt should be pretensioned according to code provisions for bolted
assemblies. See reference [25] for guidance regarding appropriate hole diameters.
Care must be taken to ensure that the crack front is intercepted by the hole. (Drilling a
hole at the end of an existing crack changes the crack from sharp to blunt, thereby
greatly increasing the force required to drive the crack. Use of a pretensioned bolt
introduces a local compressive stress, which has the effect of masking the tensile
stresses that drive the crack.)
• Cut out and refabricate parts of elements in order to reproduce the same conditions
that existed at the crack site before cracking occurred. Employ fatigue strength
improvement methods to increase fatigue resistance. This measure is most effective
for cases in which fatigue cracks have grown from weld toes and when cracking in
another, unimproved location is unlikely.
• Airarc gouge the crack, fill the gouged area with weld metal, grind away weld
reinforcement, polish smooth and inspect for weld defects using ultrasonic and Xray
inspection technology. This measure should be accompanied by some modification
that reduces the fatigue stress range in the region surrounding the crack.
• Peening the toe of a weld termination that is perpendicular to the stress range is an
effective way to prevent small cracks (less than 3 mm deep) from propagating.
Peening introduces compressive residual stresses and also changes the size of the
weld toe crack. It effectively increases the fatigue resistance by a category.
99
• Gas tungsten arc remelting has been demonstrated to effectively remove the micro
discontinuities at the weld toe and decrease the stress concentration. It also improves
the fatigue resistance by about one category. Experimental field trials demonstrated
that it was difficult to execute while the bridge was in service because of vibration.
Also, an embedded crack sometimes resulted because the penetration did not follow
the crack path. The procedure seems best suited for shop use and new construction.
Finally, any measure that lowers the stress ranges in the area around the crack will
contribute to the success of a repair. Care should be taken to ensure that cracking at other
locations is not triggered as a result of these measures. Such consequences are most likely
when the local stiffness of the structure is modified, thereby altering the sensitivity of the
structure to dynamic loading.
When damage tolerance strategies are not appropriate and when no repair option can
be implemented, replacement of the cracked element may be warranted in those situations
where it is possible to remove and replace the element. In the presence of corrosive
environments, and when there is any doubt regarding the grade of steel used in the structure,
employ a steel grade that is more suited to the situation.
Replacement using a larger element than that originally present will result in lower
stresses at the location where the crack was detected, and, for this reason, use of a larger
element is often warranted. However, a larger element alters the stiffness of the structure.
Consequently, the effects of this change in stiffness should be studied in order to ensure that
problems are not created elsewhere.
After removal of the cracked element, all connections should be inspected for
cracking. Reuse old connection material only if it can be demonstrated that it has not been
damaged by the incidence of cracking in the element. Although connections may not be
cracked, they may be damaged due to excessive movement. Do not reuse bolts that were used
in the old element for connecting the new element.
Fabrication work on the new element should be performed where adequate quality of
workmanship can be assured. Work onsite should be avoided, especially if such activity
does not contribute to reducing builtinstresses and eccentricities. Therefore, site work
should be limited to actions such as drilling undersized holes to fit and other joining
techniques that are intended to ensure good fit and satisfactory alignment.
100
7.7 Avoiding Future Cracking Problems
If fatigue cracking has been discovered at one location in a structure, it is likely to be
present at other similar locations. Consequently, if nothing is done to improve conditions at
those places, development of fatigue cracks is probable in the future. Therefore, such
locations should be identified and measures for improvement of their fatigue resistance
should be considered.
Measures that lower the stress ranges in the structure are the most effective way of
preventing further crack growth. Care should be taken to ensure that susceptibility to fatigue
cracking is not increased elsewhere as a result of these measures. Such consequences are
most probable when the local stiffness of the structure is modified, thereby altering the
response of the structure to vibration and other loading.
Effective measures for avoiding future cracking problems usually fall into one of the
following categories:
• reduction of applied stress ranges
• reduction of dynamic magnification (impact)
• reduction of number of cycles of damaging stress ranges
• decrease in vibration through damping, addition of mass, or a change in restraint
• use of fatiguestrength improvement methods at critical locations (e.g., gas tungsten
arc remelting at weld toes, peening)
• holes drilled in webs will often be effective in prevention of future damage. Holes not
only arrest cracks, but they can be used to improve problem details such as flanges
that penetrate webs.
Corrosioninduced cracking can be avoided by protecting the structure from the
environment, not allowing water and debris to become trapped on the structure, by providing
adequate drainage (for example, for deicing salts) and, finally, through modifying areas
sensitive to electrolytic action.
The following general precautions are recommended for the structure as a whole:
• Identify areas in the structure where cracking would create a dangerous condition.
Use the same criteria as those used to judge whether further cracking can be tolerated
for implementation of a damage tolerance strategy. Mark these areas for modification
and inspection.
101
• If new information is received regarding the structure or if structural modifications
are significant, reconsider decisions made regarding the management of the structure.
• Inspect repair locations regularly in order to verify the effectiveness of repair
measures.
• Create conditions whereby the structure can be inspected as easily as possible. This
involves measures such as keeping the structure clean, providing safe access to
critical locations, and designing modifications considering inservice inspection.
• All modifications should be designed with the goal of making the structure as fail
safe as possible. Failsafe means that the structure can tolerate failure of one or more
elements before a catastrophic condition arises. This condition must be preceded by a
period that is long enough to identify cracking and to take appropriate action.
102
Chapter 8 Special Topics
This chapter contains a number of topics that will be of interest in specific cases. The
topics include fatigue strength of bolted or riveted members, threaded fasteners, the influence
of environmental effects upon fatigue life, the role of residual stresses in fatigue crack
growth, fatigue under combined stresses, and comments on the use of fracture mechanics in
fatigue strength evaluation.
8.1 Bolted or Riveted Members
Fatigue characteristics of bolted connections, riveted connections, anchor bolts, and
threaded rods are discussed in this section. The need for information on all of these topics is
selfevident, except perhaps for riveted members. Rivets have not been used in new
construction for a considerable period of time, but an understanding of the fatigue life
characteristics of riveted connections is needed because of the necessity for evaluating the
remaining fatigue life of existing structures. There are tens, perhaps hundreds, of thousands
of existing riveted structures still in use in North America. Most of these are bridges, and
their remaining fatigue life and safety is a topic of great importance to the owners and to
regulatory authorities.
8.1.1 Bolted Members
Highstrength bolted joints can be subdivided into two categories; those which are lap
or butt splices ("shear splices") and those which are tensiontype connections. In the former
case, the bolts can be either preloaded or not preloaded, although in new construction most
specifications require that the bolts be preloaded if fatigue loading is likely. It has always
been common practice in bridge construction to use preloaded bolts.
The fatigue strength of a bolted shear splice is directly influenced by the type of load
transfer in the connection. This load transfer can be completely by friction at the interface of
the connected parts (preloaded bolts), completely by bearing of the bolts against the
connected material (nonpreloaded bolts), or by some combination of these two mechanisms.
In the case where the load transfer is by friction, fretting of the connected parts occurs,
particularly on the faying surfaces near the extremities of the joint. Here, the differential
strain between the two components is highest and, consequently, minute slip takes place in
this location as load is applied repeatedly. Cracks are initiated and grow in this region, which
means that cracking takes place ahead of the first (or last) bolt hole in a line, and the crack
103
progresses through the gross crosssection of the component. The phenomenon is often
referred to as "fretting fatigue."
If the bolts are not preloaded, the load transfer is by shear in the fasteners and an
equilibrating bearing force in the connected parts. The local tensile stress in the region of the
connected part adjacent to the hole is high, and this is now the location where fatigue cracks
can start and grow. Some point at the edge of the hole or within the barrel of the hole is the
initiation site for the fatigue crack and growth is through the net crosssection of the
connected part.
Both types of fatigue crack behavior have been observed in laboratory tests and, in a
few cases, both types have been observed within the same test. If nonpreloaded bolts are
used, it is highly unlikely that fretting fatigue will occur, however. When preloaded bolts are
used, it is prudent that the designer check both possible types of failure.
It is worth noting that there is no history of fatigue failure of highstrength bolts
themselves in shear splices. Only the connected material is susceptible to fatigue cracking.
Figure 46 shows the test results that were used to develop the design rule given in the
AASHTO Specification. As compared with welded details, it can be observed that there is a
great deal of scatter in the results. The data come from a wide variety of sources and test
configurations, and this explains some of the dispersion. However, it is likely that most of the
scatter is due to two more fundamental reasons. First, the level of residual stress is negligible
Figure 46 Fatigue Resistance of High Strength Bolted Shear Splices
104
in mechanically fastened joints, whereas in welded joints it is very high. This means that any
portion of a stress range that is compressive will not be as damaging as are the tensile
portions. Second, the severity of initial flaws in a mechanically fastened connection is much
less than that in a typical welded joint.
In Fig. 46, the data were plotted using the stress range on the net crosssection or the
stress range on the gross crosssection, depending upon whether or not the connection had
slipped into bearing. This is consistent with the fatigue crack mechanisms discussed above.
Also shown in the figure is the AASHTO Detail Category B fatigue life line. It represents a
reasonable lower bound, albeit a conservative choice for many of the data. Since Category B
is already a superior fatigue life category, the design of bolted shear splices will not be at any
significant disadvantage because of this conservatism. A closer evaluation of the data would
be warranted only in the case of fatigue life evaluation of an existing structure. It should also
be noted that there are very few data in the longlife region.
The AASHTO Specification prescribes Detail Category B for bolted shear splices
(AASHTO Table 6.6.1.2.3−1), and requires that stresses be calculated in the gross cross
section of the base metal in slipcritical connections (i.e., preloaded bolts are used) and at the
net crosssection when the connections are "highstrength bolted nonslipcritical." Article
6.13.2.1.1 requires that joints subjected to fatigue loading must be designed as slipcritical,
i.e., Detail Category B will be used and stresses will be calculated in the gross crosssection.
It seems prudent to use slipcritical connections throughout a bridge structure, whether or not
a particular joint is subjected to fatigue. Since the structure will be dynamically loaded,
preloaded bolts are desirable so that there is no possibility that nuts will vibrate loose.
Moreover, since the majority of the joints within the bridge will have to use preloaded bolts
in any case, use of preloaded bolts throughout the bridge gives a consistency to the
installation process.
Connections that place the bolts in tension are infrequent in bridge construction, but
they may occur in bents and frames, for example. Although there are few, if any, reported
fatigue failures of highstrength bolted shear splices, fatigue failures of highstrength bolted
tensiontype connections have occurred from time to time. Some features of the behavior of
this type of connection will be reviewed here, but it will not be treated fully. Reference [35]
provides a more complete treatment of the subject than can be covered here.
105
Typical of the configurations that can place bolts into tension is the one illustrated in
Fig. 47. This situation arises when a tee is used to connect a tension hanger to a beam lower
flange, for example. A significant feature of the connection is that prying forces develop, and
this places an additional force in the bolt, thereby
increasing the nominal value (total external force divided
by the number of bolts). The amount of the prying force
is dependent upon the flexibility of the connection, as
illustrated in the sketch. The same flexibility introduces
bending into the bolt, and this also can have an effect on
the fatigue life of the bolt. The threaded portion of the
bolt provides the crack initiation location, which as a rule
is at the root of a thread.
The stress range experienced by the bolt as the
assembly undergoes repeated loading is significantly affected by the level of bolt preload
[35]. At one extreme, properly preloaded bolts in a very stiff connection will undergo little or
no stress range. On the other hand, if the connection is relatively flexible, bolt bending is
present, and the bolt preload is low, then the stress range in the bolt threads will be large. An
additional complication occurs if the applied load is high enough to produce yielding in the
fasteners. In this case, it has been shown that the stress range increases with each cycle [35].
Recommendations for the fatigue design of highstrength bolts in tensiontype
connections can be found in References 35 and 36. In addition to reflecting the features
already mentioned here, the recommendations distinguish between ASTM A325 bolts and
ASTM A490 bolts. The latter are somewhat less ductile than the former.
The AASHTO Specification requirements for bolts in tensiontype connections
follow the same general pattern as that for other details. However, the cases of ASTM A325
and A490 bolts in tension are not set out as separate Detail Categories. Instead, the necessary
information for calculating the fatigue life of a highstrength bolt in tension is simply listed
in AASHTO Tables 6.6.1.2.5–1 and 6.6.1.2.5–3. These tables provide the constant A and the
constant amplitude fatigue stress for use in Eq. 17 and 19 of this document. Other
information concerning fatigue of bolts in tension is given in AASHTO Article 6.13.2.10.3,
where, among other things, it is noted that the bolt prying force must not exceed 60% of the
nominal force in the bolt. It is also pointed out that the stress range is to be calculated used
the area of the bolt corresponding to the nominal diameter. This is simply a convenience that
can be employed because the ratio between the area through the threads and that
Bolt force
Prying force
Applied force
Figure 47 Bolts in Tension
106
corresponding to the nominal diameter of the bolt is relatively constant for the usual bolt
sizes.
The fatigue design of highstrength bolts that are in tensiontype connections should
reflect the following guidelines:
• Whenever possible, redesign the connection so that the bolts are in shear, not
tension.
• Design the connection so that prying forces are minimized. Reference [36]
suggests that the prying force should be less than 60% of the externally applied
force. Some specifications have even lower limits, although the AASHTO
Specification also uses the 60% value.
• Whenever possible, use ASTM A325 bolts rather than A490 bolts.
• Ensure that proper installation procedures are adhered to so that the prescribed
bolt preloads will be attained.
8.1.2 Threaded Rods
Threaded rods are used frequently in mast or sign base plates. Under the action of the
wind on the mast or sign, fatigue loading will be present, and this can produce a large
number of stress cycles, usually at a relatively low stress range. If highstrength bolts that are
not preloaded act in direct axial tension, they are equivalent to a threaded rod.
A large number of test results are available [37] and they are shown in Fig. 48. Most
of the tests represented in this figure were carried out at constant amplitude loading, but a
few tests were done under variable amplitude. (As is customary, the stress range was
calculated using the socalled "stress area" of the rod.) Many of the tests were done in a
doublenut configuration, that is, a nut was present on each side of a simulated base plate. In
this way, a short section of the rod is prestressed, but most of it is not. The effect of a small
amount of misalignment was also included in some tests. The results include axially loaded
rods (i.e., no preload), rods preloaded to the "snugtight" level, and rods preloaded by turning
the nut on 1/3 turn.
107
Figure 48 Fatigue Life of Anchor Bolts (Ref. 37)
There is considerable scatter in the results, but it clear that the threaded rod is an
element with poor fatigue resistance. The results do indicate that grade of steel, thread size,
method of thread forming, and bolt diameter did not in themselves have a major effect on the
fatigue life [37]. The misalignment used in the tests (1:40) likewise did not significantly
reduce the fatigue life, assuming that beveled washers are used. The fatigue life of rods
tested with no preload and rods tested in the snugtight condition was similar. A small
increase in fatigue life was observed for fully preloaded rods.
Figure 48 also shows AASHTO Detail Category E'. As can be seen, this is a
reasonable representation of the data in the region up to about 1 million cycles. The constant
amplitude fatigue limit for Category E' (17.9 MPa) clearly is too conservative for these
(mostly) constant amplitude tests: a better choice is the Category E constant amplitude
fatigue limit (31.0 MPa). Unfortunately, the region of low stress range and a large number of
cycles is reflective of where many of the structures that use threaded rods will have to
function. A sign baseplate, for example, could have a large number of cycles at low stress
range under wind loading. Until more variable stress range data are available, it would be
10
100
1000
10000 100000 1000000 10000000 100000000
Number of Cycles
S
t
r
e
s
s
R
a
n
g
e
,
M
P
a
Frank
Dusel
Dexter (concentric)
Dexter (eccentric)
Mean
Mean  2 Std. Dev.
Detail Category E'
108
prudent to use Detail Category E' throughout. The AASHTO Specification does not address
the fatigue life of anchor bolts (threaded rods).
8.1.3 Riveted Connections
It has already been explained that the need for fatigue life rules for riveted
connections rests in evaluation of existing structures. If a conventional fatigue life
examination indicates that a given riveted structure has long remaining fatigue life, then
obviously no further effort is necessary. However, when remaining fatigue life is shown to be
inadequate, it is necessary that the best possible information be used. The alternative might
be replacement of the structurea costly solution.
As was the case for highstrength bolted shear splices, the experimental evidence is
that fatigue cracking in riveted shear splices takes place in the connected material, not in the
rivet itself. Thus, the fatigue life can be expected to be a reflection of such features as the
size of the hole relative to the part, the method of hole forming (drilled, punched, or sub
punched and reamed), the bearing condition of the rivet with respect to the hole, and the
clamping force provided by the rivet.
At the present time, the influences of clamping force, bearing condition, and the
method of hole formation have not been examined in any systematic way. The influence of
the hole size, per se, is not likely to be strong, as long as the hole sizes and plate thicknesses
commonly used in structural practice pertain. Thus, the best data available are tests on
riveted connections of proportions that are consistent with usual structural practice and are of
full size, or at least large size. For the time being, the reflections of clamping force, bearing
condition, and hole formation must simply be part of the data pool. For this reason, and
because the "defect" presented by a riveted connection is not severe, it is to be expected that
the scatter of data will be relatively large.
Figure 49 shows the data: identification of the specific sources from which the test
data came can be obtained in Reference [38]. Most of the data come from tests of flexural
members, and most of these were members taken from service. For those cases where
members taken from service were tested, the previous stress history was examined and
deemed to have been nondamaging. A few of the test results are from tension members. In
the case of bending members, the moment of inertia of the crosssection included the effect
of holes. For the tension members, the stress range was calculated on the net crosssection.
(It is not yet clear whether this is justified since the fatigue cracks were observed to grow at
right angles to the crosssection.) As was the case with bolted connections, there is a
109
considerable amount of scatter in the data, for the reasons described earlier. Almost all the
data shown in Fig. 49 come from members made of mild steel, but results from some
wrought iron members are also present.
Figure 50 shows a fatigue crack passing through the net section of a riveted member
that was tested in tension.
The AASHTO Specification stipulates the use of Detail Category D for riveted
connections, and this is shown along with the data in Fig. 49. Category D is a reasonable
reflection of the data except for two test results at long fatigue lives. These were variable
amplitude tests. It has been suggested that a fatigue limit of 40 MPa should be used, rather
than the value of 48.3 MPa that is used for Category D. This change would capture the long
life data points better, but it is too conservative in the region of the "knee" of the bilinear
fatigue life representation. The issue is discussed further in Reference [38].
Figure 49 Fatigue of Riveted Connections
110
8.2 Environmental Effects; Use of Weathering Steel
The effects of the environment upon the fatigue life of steel members can show up in
one of several ways. Corrosion produces local effects that cause stress concentrations. For
instance, corrosive products (e.g., brine) can drip down from an upper level to a lower level
and produce both a local reduction in crosssection and a notch effect. Old railway bridge
floor system members sometimes show this type of corrosion. For example, the top surface
of the lower flange of a girder can often show deterioration of the crosssection as the result
of brine that has dripped down from the upper levels. (Ice and salt were formerly used as the
refrigerant in railway rolling stock). It is not uncommon to see depressions in the order of 25
to 50 mm diameter, with depths in the order of 3 to 8 mm. In extreme cases, particularly with
initially thin members (e.g., bracing members, lacing bars), the local corrosion may
completely penetrate the member. Although these are not desirable situations, they are
similar to the introduction of a hole: as such, they do not introduce a particularly severe flaw
into the member. Since this type of corrosion effect will generally be an issue in the case of
evaluation of old bridges, a value judgment can be made as to how much to degrade the
fatigue life as compared with a noncorroded member. Fisher et al. have reported on the
fatigue life of a number of corrosionnotched beams [39], although the number of test data in
which the corrosion effect was specifically examined is quite limited The ratio of corroded
part to original thickness ranged from 0.2 to 0.8. Fisher et al. reported that if the corrosion
reduced the thickness by more than onehalf, fatigue cracking was likely to start at a notch
within the corroded region rather than at the location of a net section (reduced by corrosion)
through rivet or bolt holes.
Figure 50 Fatigue Crack Through Net Section of a Tension Member
111
There are also instances in which the entire crosssection has been corroded, more or
less in a uniform way. The origin of the corrosion is these cases is a more general source,
such as high humidity, usually in a saltladen atmosphere, or corrosive fumes in proximity to
chemical process plants. In the case of a general reduction in crosssection, the fatigue life
evaluation can be based on an estimate of the reduced crosssection. In other words, the
fatigue strength examination of a corroded member can proceed on the basis of the usual
stress range vs. number of cycles relationships, given that the stress range will be calculated
using the properties of the reduced section. Of course, the designer should be alert to the
presence of notches in this case also.
Figure 51 shows the lower flange of a railroad stringer in which a fatigue crack has
occurred at a corrosion notch.
In principle, the useful life of a welded bridge structure can be affected by an
aggressive environment when it is under sustained load (stresscorrosion cracking), or by the
enhancement of an aggressive environment on the effects of cyclic stresses (corrosion
fatigue). However, general experience with bridge steels and weldments, including
successful use in longtime environmental service applications, have not demonstrated any
significant problems with either stresscorrosion cracking or corrosion fatigue.
The effect of a corrosive environment upon crack growth or as it increases stress
concentrations has not received much attention, particularly in civil engineering applications.
The available laboratory test results, often done at relatively high test rates, do not indicate
any significant effect, however [39]. Moreover, experience in the field does not support the
Figure 51 Fatigue Crack at a Corrosion Notch in a Railroad Stringer
112
need for inclusion of special rules for bridges or similar structures that are operating in the
types of corrosive environments normally associated with that type of structure. In two
specific cases, the consequences of operating in a corrosive environment are included
explicitly in the design rules. Both of these are clearcut—offshore structures and the
drillpipe used in oil exploration. In both instances, the environment is severe—salt water—
and its consequences cannot be avoided. There is not much test data available here either,
however, and it must be recognized that corrosion and its effect on fatigue life is time
dependent. Thus, testing in this area takes even longer than it does for the normal situation of
testing in air. Designers of civil engineering structures in particularly harsh environments
should take special care that detailing of the structure minimizes the possibility of locations
for fatigue and that a detailed maintenance strategy is established. In all cases, detailing
practice should be to promote drainage of water and to prevent of entrapment of moisture and
corrosion products.
Crack growth rate data has been acquired on most commonly used bridge steels under
constant amplitude and variable amplitude loading [40, 41]. The scatter bands of the test
specimens in air generally bracketed the test data acquired in an aqueous environment. The
latter was done with or without a 3 percent solution of sodium chloride. Most of the aqueous
data fell between the mean and upper limit of the data acquired in air. However, all of the test
data was acquired at relative high levels of ∆K and crack growth rate. Figure 52 shows the
Figure 52 Crack Growth in Aqueous Environment
113
test data for A588 steel in an aqueous environment with and without the sodium chloride
solution. This level of ∆K is only reached after very large cracks have formed. The crack
growth threshold appears to increase with decreasing load frequency when subjected to a
hostile environment. The general conclusion is that the crack propagation life of bridge steel
components under actual operating conditions (wet and dry) in aqueous environments is
equal to or greater than that obtained in a room temperature air environment. The laboratory
studies which led to the conclusion that environmental conditions have little effect upon the
fatigue behavior of bridge components are supported by field experience.
The influence of the environment upon rusting, pitting, and the start of corrosion
fatigue cracking (assuming no prior fatigue crack) may differ from its influence on
progressive crack extension when the crack crevice controls the crack tip behavior. As would
be expected, rusting and pitting introduces notches and this can affect the crack initiation
behavior of the base metal. Studies on weathering steel have demonstrated that only the
Detail Category A situation is significantly affected, however. Figure 53 shows the
experimental results of cruciform joints, which are a Category C detail, in their aswelded
and weathered conditions. It is apparent that the use of weathering steel does not in itself
produce a fatigue life that is inferior to that when ordinary grades of steel are used.
Many civil engineering structures must operate in low temperature environments.
Again, there are very few test data in this area. The work of Fisher et al. [39] did include
Figure 53 Comparison of Weathered and Unweathered Cruciform Joints
114
some tests in which a portion of the fatigue crack growth was done under very low
temperatures (–40°C and −73°C). The study showed that the low temperature did not result
in sudden fracture until the cracks had become very large.
In summary, it can be concluded that the civil engineering structures most likely to
have to endure corrosive environments, especially bridges, can be designed in accordance
with the rules that have been developed on the basis of tests in air. Evaluation of existing
structures in which corrosion has affected the crosssection can be done by applying
judgment and using the results of those tests that are available [39]. In the case of new
structures, great care should be taken to ensure that corrosion products from roadway decks
are directed away from the members and that road spray from traffic either above or below
the bridge is cleaned up periodically. Detailing should be done with a view to minimizing
those locations where corrosion products can form and sit, especially in crevices.
8.3 Combined Stresses
Throughout the discussion so far, it has been emphasized that the stresses to be used
in the fatigue life evaluation are those corresponding to the nominal stress as obtained from a
strength of materials level of analysis. The detail classification stipulated by the governing
specification includes the effect of local stress concentrations due to weld shape,
discontinuities, triaxial conditions, and so on. Only in exceptional cases, around large
openings, for example, will a more sophisticated analysis be required. Occasionally, also, it
might be necessary to investigate the combined effect of normal and shear stresses.
When normal and shear stresses are present at the same location but do not occur
simultaneously under a given loading event, the individual components of damage can be
added according to a Miner's summation, as follows:
∆σ
∆σ
∆τ
∆τ
e
r r
+
≤
e
1
3 3
(20)
where σ and τ refer to normal stresses and shear stresses, respectively. The subscript e
means a calculated equivalent stress range (see Eq. 13, for example) and the subscript r
refers to the permissible stress range for the detail.
For a case in which normal and shearing stresses are present in significant quantities
at a given location and are concurrently present and in phase at that location during a given
loading event, the principal stresses should be calculated. This stress range is then used in the
115
fatigue life evaluation. (Normal stress is defined as the stress perpendicular to the direction
of the potential fatigue crack.) When the principal stress changes direction during the stress
cycle, as it does at transverse stiffeners in plate girder webs, for instance, it has been shown
that the fatigue resistance need not consider the principal stress [19]. Generally, it should
only be necessary to consider the principal stress when its direction does not change during
the stress cycle. Hence, use of Eq. 20 should be focused on this case.
In unusual situations, shear stress alone can be significant. There are very few test
results, but they show that the general relationship between stress range and number of
cycles still exists. The AASHTO Specification recommends that Detail Category E be used
for the case of shear stress on a fillet weld throat.
8.4 Effect of Size on Fatigue Life
Most of the test results upon which the fatigue life rules are based were done on
specimens where the component parts were in the order of 12 to 25 mm thick. Fabricated
steel structures often use plate thickness considerably greater than this, however, particularly
in the case of flanges of welded builtup beams. It is generally agreed [42, 43, 44] that the
fatigue resistance of thicker elements is less than that of thinner elements. Although a thicker
element will have a greater statistical chance of containing flaws, this is not the main reason
for the decrease. Rather, it has to do with more severe residual stresses that result when using
thicker plates and from the observation that the stress intensity factors at the weld toe
increase as plate thickness increases (and the initial flaw size remains constant) [45].
There are several locations within the AASHTO Specification where it is required
that fatigue strength be decreased with an increase in plate thickness. For example, the
fatigue category for the base metal at the end of partial length cover plates is reduced from
Detail Category E to E' (case of plate wider than flange) when the plate thickness exceeds
20 mm. Likewise, the same decrease in fatigue detail category occurs when the plate
thickness exceeds 25 mm in a longitudinally loaded groove or fillet welded detail. This
would apply to the attachment of flange plates to girder web, for example.
8.5 Role of Residual Stress
Steel structures that are fabricated by welding contain "residual" or "lockedin"
stresses that have been introduced as a result of the welding process. These have
considerable influence on the propagation of fatigue cracks. The main effect is to
significantly reduce the effects of the applied stress levels, per se, for standard weldable
116
structural steels. For modern codes, this has brought about a return to the simple stress range
vs. cycle life model for fatigue strength suggested by Wöhler over one hundred years ago.
Furthermore, independence of steel grade justifies the use of the relatively large data base of
laboratory results taken from tests of different steel grades produced in different countries.
During recent modifications to fatigue codes, code developers have taken advantage of such
opportunities and, consequently, fatigue design guidelines have been greatly simplified and
harmonized internationally.
Consider a weld laid down as shown in
Fig. 54. As the weld cools, it tries to contract.
However, since the plate and the weld must
maintain compatibility of length, the plate restrains
the weld during the cooling and contraction
process. This puts the weld and a relatively small
volume of plate adjacent to the weld into tension.
Conversely, the main portion of the plate is being
pulled down by the contracting weld, thereby
placing it into compression. The stresses set up in
the plate and in the weld as a consequence of this
process are called residual stresses. Since there are
no external forces applied during this process, the equilibrium condition of the crosssection
must be reflected in the balance between residual tensile stress and the area over which it acts
and the residual compressive stress and its associated area. The actual distribution and
magnitude of the residual stress pattern depends upon such factors as the strength of the steel
and the weld metal, the geometry of the connected parts, and the size of the weld relative to
the connected parts. The important fact, however, is that the magnitude of the tensile residual
stress can reach the yield strength of the material.
It follows, of course, that the rolled shapes or builtup members used in structural
applications also contain regions of high residual tensile stress. For example, very large
residual tensile stresses are present at the junction of the flange and web of a beam that has
been built up by welding the component parts together [18]. This junction is also the location
of the flaws that are likely to be the source of fatigue crack growth, which means that the
flaw is under a condition of initial stress even before load is applied. For the usual condition
wherein this initial stress is at or near the yield stress level, this means that stress range,
rather than the maximum applied stress, the stress ratio (ratio of maximum stress to minimum
residual stress
pattern
tendency for
deformation
Figure 54 Residual Stresses in a Welded
Plate
117
stress), or some other parameter of applied stress, is the governing condition describing
fatigue crack growth. This will be illustrated by the following numerical example.
EXAMPLE 13
Figure 55 shows a rectangular plate and its assumed residual stress pattern. (This
could be an idealization of the flange of a welded beam.) According to the designer, the plate
will be subjected to a maximum tensile stress of 215 MPa and a minimum compressive stress
of 122 MPa. Calculate the actual range of stress that will exist for a flaw at the center
of the plate (cf., junction of the web and flange of a welded beam). Compare this range of
stress with the stress range that will be computed by the designer, that is,
( ) + − − = 215 122 337 MPa . The plate has a yield strength of 300 MPa.
Solution:
1. Initial conditions—The pattern of residual stress is shown in Fig. 55 (a). It shows that the
values of the initial stress are as high as the yield stress. Since the yield stress is not
exceeded, however, the corresponding strains can be calculated from the elastic
relationship, σ ε = E , and these strains are shown in Fig. 55 (b). There is no force on the
member, and we note (by inspection) that the condition P =
∫
σ dA = 0 is met. (This can
verified by calculation.)
σ
ε
E = 200 000 MPa
300
σ
y
=
ε
y
= 1.50x10
– 3
300 +
300 – MPa
MPa
(a)
Stress
Residual (b) Residual
Strain
125 mm
25 mm
–3
–1.50x10
–3
1.50x10
Figure 55 Initial Stress Condition in a Welded Plate
weld
118
2. Apply tensile force such that the design stress level is +215 MPa, i.e.,
P = σ A = 215 N/mm
2
x (25 x 125) mm
2
= 672 x 10
3
N = 672 kN
During application of this force, which is the equivalent of the imposition of a uniform
compressive strain over the residual strain of Fig. 55 (b), not all parts of the crosssection
respond in the same way. Specifically, the relationship ε σ = E will not be valid for those
regions of the resultant strain diagram where the strain is greater than the yield strain. In
Fig. 56, part (a) shows the residual strain (which is simply Fig. 55 (b) repeated), part (b)
shows the strain imposed by the tensile force, part (c) is the resultant strain, and the
corresponding stress diagram is shown in Fig. 56 (d).
The value of the uniform strain corresponding to imposition of the applied force, 14 10
3
. x
−
,
was determined by trial so to satisfy the requirement that σ dA = 000 N 672
∫
. That the
integral is satisfied can be confirmed by first calculating the dimensions of the stress
diagram, Fig. 56 (d), and then calculating the force volume obtained when the stress
diagram is superimposed on the crosssection.
3. Now, apply a force such that the nominal stress in the crosssection goes to 122 MPa
compression. This requires that the 672 kN force be removed and an additional force of
( )
( ) − = − 122 25 125 381 10
2 2 3
N mm x x mm x N / be applied. Thus, the change in force is
( ) ∆P = + = 672 381 1053 kN. As before, a uniform strain must be applied of sufficient
magnitude such that σ dA = P = 381 kN
∫
. By trial, this has been determined to be a
strain of 170 10
3
. x
−
.
+ =
(a) (b) (c)
1.50x10
– 3
1.50x10
– 3
–
1.40 x10
– 3
2.90x10
– 3
0.10x10
–3
–
(d)
300 MPa
–20 MPa
ε
y
0 0 0
0
Figure 56 Superimposed Strain (Step 2) and Corresponding Stress
119
Figure 57 (a) shows the strains that existed at the end of step 2. Figure 57 (b) is the
uniform compressive strain that must be applied to satisfy the requirement that a force of
381 kN be imposed. The resultant strains are shown in Fig. 57 (c). The reader can verify
the stress values shown in Fig. 57 (d) and that the imposition of these stresses on the
crosssection does, in fact, produce the value 381 kN.
4. Finally, examine the stresses in the
middepth fibre of the crosssection.
As shown in Fig. 58, the stress
initially was 300 MPa, that is, it was
the yield value. Upon initial loading
to 672 kN force, the stress remained
at this value, even though the strains
increased as shown in Fig. 58. The
unloading that occurred in Step 3
took the stresses down to 40 MPa
compression. Thus, the actual
change in stress at the middepth
fibre was from +300 MPa down to
–40 MPa, a total change of
340 MPa. The result is that the actual change of stress, 340 MPa is equal (nearly) to the
change of stress calculated by the designer, 337 MPa. The difference between the two
values is attributable to roundoff errors in the calculations.
(Still with reference to Fig. 58, it should be borne in mind that unrestricted plastic flow of
the crosssection cannot occur. The small zone of weld and plate that is at yield is limited
+
=
2.90x10
–3
– 0.10x10
–3
–1.70x10
–3
1.20x10
–3
1.80x10
–3
–
ε
y
300 MPa
(a) (b) (c) (d)
–
40 MPa –
Figure 57 Superimposed Strain (Step 3) and Corresponding Stress
σ
ε x10
–3
σ
y
= 300
– 40
MPa
Unloading
strain applied in step 3
i.e., 1.70
0
1.50
final stress condition
2.90
initial (no load) stress
after application
of 672 kN force
Figure 58 Strains and Stresses at MidDepth Fibre
120
to some finite value of strain because he material adjacent to it remains elastic. Hence,
the defect in a high residual tensile stress region is cycled to a limited strain.)
To summarize, in large welded structures there are very high tensile residual stresses
near fatigue crack sites and their presence significantly reduces the effects of applied stress
levels and steel grade upon crack propagation for standard weldable structural steels. As a
result, it is generally agreed that stress range is the only dominant stress parameter. However,
there has been a great deal of discussion regarding the best way to treat the following cases:
• the influence of compressive applied stresses
• the fatigue strength of nonwelded elements
• the fatigue strength of stressrelieved welded elements
• the difference between results taken from testing small specimens and those taken
from testing fullscale elements.
Some of the differences between various codes can be traced to different
interpretations of the effect of residual stresses (and other builtin stresses) upon the fatigue
strength of structural elements. A selection of approaches to these cases is described next.
Nearly all fatigue guidelines for steel structures recommend that nonwelded
elements, stressrelieved welded elements, and welded elements be treated in the same
manner when the applied stress ranges are tensile. When the applied stress ranges are
compressive, or partly compressive, recommendations vary. According to the AASHTO
recommendations [2], it is possible to ignore all fully compressive stress ranges. (This was
discussed in Section 5.3.2.) However, if a part of the stress range is tensile, then the whole
stress range must be considered. European recommendations suggest that the whole stress
range should be employed, regardless of whether it is completely or partially compressive,
when welded elements are assessed. Eurocode 3 (EC 3) [46] allows a reduction of the
compressive stress range to 60% of its magnitude for nonwelded and stress relieved
elements. The ECCS recommendations [15] are the most conservative; they recommend that
100% of the stress range be used for all elements and all cases. It can be observed that there
are no known cases of loadinduced fatigue cracking in bridge structures when the cyclic
stresses are totally in compression [20].
Authors of the AASHTO provisions argue that, even if fatigue cracks start within
tensile residual stress fields, under fully compressive loading cracks will eventually stop
when they propagate beyond the influence of these stresses. The developers of the EC 3
121
document note that a crack growing from an aswelded joint in an element may be dangerous
when applied stress ranges are full compressive. Indeed, some test elements have failed due
to cracks growing under fully compressive applied loading. Also, the EC 3 provisions reflect
an attempt to accommodate those cases where there are no high tensile residual stresses.
For the ECCS document, it was decided to provide conservative recommendations
because in most complex structures the magnitudes of builtin stresses are unknown. Builtin
stresses do not only originate from welding related residual stresses, but they may also be
caused by other effects such as lack of fit, settlement of supports, temperature gradients, and
inefficient expansion joints. Therefore, it was decided that no advantage should be given to
cases where applied stresses are in compression or where elements contain no tensile residual
stresses.
The final point in this section relates to the physical testing of fatigue specimens.
Compared with elements that have no tensile residual stresses, the presence of tensile
residual stress near potential crack sites in an element reduces its fatigue strength, especially
at long lives. Small test specimens may not contain the same high level of tensile residual
stress found in fullscale elements. For this reason, most recent research programs have
concentrated upon testing of fullscale elements in order to establish code recommendations.
Therefore, results taken from tests of small specimens should be employed with caution for
design assessments. However, small specimens are useful when exploring variables such as
improvement procedures, use of new materials, and so on.
Test programs are sometimes used to obtain a more accurate fatigue strength
relationship than is provided in fatigue codes. Whenever possible, in these circumstances it is
desirable that the test specimen have the same dimensions as the element under consideration
for the real structure.
8.6 Quantitative Design Using Fracture Mechanics
The fatigue strength curves described in Chapter 3 are not capable of providing an
engineer with information necessary for several special cases related to fatigue assessments.
For example, these curves cannot evaluate the influence of an unusually large defect; they
are not able to evaluate the influence of inspection precision upon fatigue reliability; they
provide little help in fixing inspection intervals; and, finally, they cannot be used to predict
the remaining life of a cracked structure. Fracture mechanics provides the central analytical
model employed in fatigue assessments for dealing with these aspects.
122
Some defects that may be rejected during standard fabrication controls would not, in
reality, influence the performance of the structure. The economic consequences of remedial
measures are often severe and resulting delays to a project can be long. Quality assurance
provisions in many codes are based more upon accepted standard practice, experience, and
detection capability than upon scientific accuracy.
Fracture mechanics offers a more rational means of defect assessment. In such
analyses, the value for the initial crack length in Eq. 7, Section 2.3 is normally taken to be
equivalent to the size of the defect. Equation 7 is then integrated, usually numerically using
equal increments of cycles, in order to determine cycle life. If the cycle life is greater than
the number of cycles expected over the life of the structure, the defect may be regarded as
being acceptable. Reference 47 provides guidelines for appropriate values of W and Y for
use in Eq. 3 and for additional criteria, such as those which determine whether two closely
spaced defects should be considered to be equivalent to one large initial crack.
A similar approach can be used for the design of structures when fabrication quality
cannot be guaranteed and when it is known that inspection technology is unable to detect
defects that are more severe than those normally known to occur in steel structures. In these
evaluations, it is assumed that defects are present and the minimum detectable defect size
becomes the initial crack size for use in Eq. 7. When the cycle life resulting from the
integration is less than the number of cycles applied over the life of the structure, either the
design is modified or inspection technology that enables greater precision is justified. Such
analysis has been used to determine inspection criteria for steel structures in the North Sea,
for example. This approach is sometimes called a "fitness for purpose assessment."
Often, determination of intervals
between inservice inspections does not
account for the behavior of a potentially
cracked structure. Crackgrowth curves can
be predicted for all critical details in a
structure, as shown schematically in
Fig. 59. Between the crack length when
first detection is possible, a
o
, and the
critical crack size, a
cr
, a set number of inspections, n, should be performed. A constant
cracklength increment, ∆a, should be fixed rather than a constant time interval. The number
of inspections should be determined using reliability analyses that take into account crack
detection probabilities and the consequences of failure. More frequent inspections are
∆
a
∆
a
∆
a
∆
a
a
0
a
cr
a
time (years)
100
∆
a
∆
a =
a
cr
a
0
n
n = number of
inspections
crack
size
Figure 59 Crack Growth Curve
123
required as the structure ages. Therefore, fracture mechanic analysis enables a more rational
approach for writing those parts of specifications related to inspection.
As the number of structures that approach or exceed their design life grows
exponentially each year, so does the number of structures which have a high probability of
containing fatigue cracks. Fracture mechanics will become an essential part of the
assessment of these structures and, consequently, the decisions concerning their
strengthening, repair and replacement. Cracking in several North American steel bridges was
reviewed in a collection of case studies [20]. Finally, fracture mechanics analyses may assist
in the assessment of complex structures that have connections which are not covered by test
results. Test data obtained from simple beams and small specimens may be inapplicable to
large structures and to complex loading. This is particularly true for offshore structures, and
some of these have been analyzed extensively using fracture mechanics.
124
REFERENCES
1. Smith, I.F.C. and Hirt, M.A., "Fracture Mechanics in Structural Engineering," Steel
Structures; Recent Research Advances and Their Application to Design, M.N. Pavolic,
Editor, Elsevier, 1986, Chapter VI, 381402.
2. American Association of State Highway and Transportation Officials, "AASHTO LRFD
Bridge Design Specifications," SI Units, First Edition, Washington, D.C., 1994.
3. Griffith, A.A., "The Phenomena of Rupture and Flow in Solids," Phil. Trans. Royal Soc.
London (Series A), 221 (1921), 163198.
4. Irwin, G.R. and de Wit, R., "A Summary of FractureMechanics Concepts," J. Testing
Evaluation, 11 (1983), 5665.
5. Rolfe, S.T., "Fracture and Fatigue Control in Steel Structures," Eng. J. Am. Inst. Steel
Constr., 14 (1983), 215.
6. Smith, R.A., "An Introduction to Fracture Mechanics for Engineers – Parts I, II, and III,"
Materials Eng. Applic., 1 (197879), 121128, 227235 and 316322.
7. Broek, D., Elementary Engineering Fracture Mechanics, Martinus Nijhoff, The Hague,
Netherlands, 1984.
8. Rolfe, S.T. and Barsom, J.M., Fracture and Fatigue Control in Structures, 2nd Ed.,
PrenticeHall, Englewood Cliffs, New Jersey, 1987.
9. Pellini, W.S., Guidelines for FractureSafe and FatigueReliable Design of Steel
Structures, Welding Inst., Cambridge, UK, 1983.
10. Tada, H., Paris, P., and Irwin, G., The Stress Analysis of Cracks Handbook, Paris
Productions, St. Louis, 1985.
11. "J, a Measure of Fracture Toughness," in ASTM E 81381 (ASTM Book of Standards
3.01) (1983), 762780.
12. Lankford, J., "The Influence of Microstructure on the Growth of Small Fatigue Cracks,"
Fatigue Fracture Eng. Materials Struct., 8 (1985), 161175.
13. Paris, P. and Erdogan, F., "A Critical Analysis of Crack Propagation Laws," Trans.
ASME (Series D), 85 (1963), 528534.
125
14. Suresh, S. and Richie, R.O. , "Propagation of Short Fatigue Cracks," Int. Metals Rev., 29
(1984), 445476.
15. European Convention for Constructional Steelwork, "Recommendations for the Fatigue
Design of Steel Structures," ECCS Technical Committee 6, Rotterdam, 1985.
16. Fisher, J.W. and Viest, I.M., "Fatigue Life of Bridge Beams Subjected to Controlled
Truck Traffic," Preliminary Publication, 7
th
Congress, IABSE, 1964.
17. Gurney, T.R., Fatigue of Welded Structures, 2nd Edition, Cambridge University Press,
1979.
18. Fisher, J.W., Frank, K.H., Hirt, M.A., and McNamee, B.M., "Effect of Weldments on the
Fatigue Strength of Steel Beams," National Cooperative Highway Research Program
Report 102, Highway Research Board, Washington, D.C., 1970.
19. Fisher, J.W., Albrecht, P.A., Yen, B.T., Klingerman, D.J., and McNamee, B.M., "Fatigue
Strength of Steel Beams With Transverse Stiffeners and Attachments," National
Cooperative Highway Research Program, Report 147, Highway Research Board,
Washington, D.C., 1974.
20. Fisher, J.W., Fatigue and Fracture in Steel Bridges: Case Studies, Wiley Interscience,
1984.
21. Broek, D., The Practical Use of Fracture Mechanics, Kluwer Academic Publishers,
Dordrecht, The Netherlands, 1989.
22. Bardell, G.R., and Kulak, G.L., "Fatigue Strength Behaviour of Steel Beams with Welded
Details," Structural Engineering Report No. 72, Dept. Civil Engrg., University of
Alberta, September, 1978.
23. Bannantine, J.A., Comer, J.J., and Handrock, J.L., Fundamentals of Metal Fatigue
Analysis, PrenticeHall, Englewood Cliffs, (New Jersey), 1990.
24. American Railway Engineering Association, "Manual for Railway Engineering," Chapter
15; Steel Structures, AREA, 1994.
25. Fisher, J.W., Yen, B.T. and Wang D., "Fatigue of Bridge Structures – a Commentary and
Guide for Design, Evaluation and Investigation Of Cracking," ATLSS Report No 8902,
Lehigh University, Bethlehem, PA, 1989.
26. Matsuishi, M. and Endo, T., "Fatigue of Metals Subject to Varying Stress," Jap. Soc.
Mech. Engrg, Fukoda, Japan, March 1968.
27. Downing, S.E., and Socie, D.F., "Simple Rainflow Counting Algorithms," International
Journal of Fatigue, Vol. 4. No. 1, 1982.
126
28. Schilling, C.G., Klippstein, Barsom, J.M., and Blake, G.T., "Fatigue of Welded Steel
Bridge Steel Members Under Variable Amplitude Loadings," Final Report, NCHRP
Project 12–12 Transportation Research Board, Aug. 1975.
29. Smith I.F.C., Castiglioni, C.A., and Keating, P.B. "An Analysis of Fatigue
Recommendations Considering New Data," IABSE Periodica, 13, International
Association of Bridge and Structural Engineering, Zurich, 1989.
30. Fisher, J.W., Yen, B.T., Wang, D., and Mann, J.E., "Fatigue and Fracture Evaluation for
Rating Riveted Bridges," National Cooperative Highway Research Program, NCHRP
Report 302, Highway Research Board, Washington, D.C., 1974
31. Fisher, J.W., "Bridge Fatigue Guide: Design and Details," American Institute of Steel
Construction, AISC Pub. No. T112–11/77, 1977.
32. Fisher, J.W., "Fatigue Cracking in Bridges from OutofPlane Displacements," Canadian
Journal of Civil Engineering, Vol. 5, No. 4, December 1978.
33. Fisher, J.W., Jian Jin, Wagner, D.C., and Yen, B.T., "DistortionInduced Fatigue
Cracking in Steel Bridges," National Cooperative Highway Research Program, NCHRP
Report 336, Highway Research Board, Washington, D.C., 1990.
34. Cheng, J.J., "Design of Steel Beams with End Copes," Journal of Constructional Steel
Research, Vol. 25, 1993.
35. Kulak, G.L., Fisher, J.W., and Struik, J.H.A., Guide to Design Criteria for Bolted and
Riveted Joints, Second Edition, John Wiley & Sons, Inc., New York, 1987.
36. LRFD Specification for Structural Joints Using ASTM A325 or A490 Bolts. Research
Council on Structural Connections, 1988. (Available through the American Institute of
Steel Construction, Chicago, IL.)
37. Van Dien, J.P., Kaczinski, M.R., and Dexter, R.J., "Fatigue Testing of Anchor Bolts,"
Proceedings, Structures Congress XIV, ASCE, Chicago, 1996.
38. Kulak, G.L., "Fatigue Strength of Riveted Connections," IABSE Workshop on
Evaluation of Existing Steel and Composite Bridges, IABSE Report No. 76, Lausanne,
Switzerland, March 1997.
39. Fisher, J.W., Yen, B.T., and Wang D., "Fatigue Strength of Riveted Bridge Members,"
Journal of Structural Engineering, ASCE, Vol. 116, No. 11, November, 1990.
40. Carter, C.S., Hyatt, M.V., and Cotton, J.E., "StressCorrosion Susceptibility of Highway
Bridge Construction Steels, Phase I," Report FHWA–RD–F73–003, FHWA, April 1972.
41. Barsom, J.M. and Novak, S.R., "Subcritical Crack Growth and Fracture of Bridge
Steels," NCHRP Report 181, Transportation Research Board, 1977.
127
42. Gurney, T.R., "Theoretical Analysis of the Influence of Attachment Size on the Fatigue
Strength of Transverse NonLoadCarrying Fillet Welds," Members Rept. 91/1979.
Welding Inst. Cambridge, 1979.
43. Berge, S., "Effect of Plate Thickness in Fatigue of Cruciform Welded Joints," Rept.
MK/R 67, Norwegian Inst. Tech., Trondheim, 1983.
44. Smith, I.F.C. and Gurney, T.R., "Numerical Study of Geometrical Effects in Longitudinal
NonLoadCarrying FilletWelded Joints," Welding Journal, Vol. 65, American Welding
Society, Miami, 1985.
45. Maddox, S.J., "Fatigue Strength of Welded Structures," Second Edition, Abington
Publishing, Cambridge, U.K., 1991.
46. European Committee for Standardisation (1992). "Eurocode 3: Design of Steel
Structures," ENV 1993–1–1, Brussels.
47. PD 6493 "Guidance on Some Methods for the Derivation of Acceptance Levels for
Defects in Fusion Welded Joints," Welding Standards Committee, BSI, UK, 1991.
128
INDEX
AASHTO:
average daily truck traffic (ADTT), 55
basis of design rules, 22, 25
bolted shear splices, 103
compressive stresses, 58, 120
constant amplitude fatigue limit
(CAFL), 57
design example, 59
design truck, 54
distortioninduced fatigue, 25, 73, 84
dynamic load allowance, 55
effect of component size, 115
fatigue category, 25
fatigue design truck, 54, 64
fatigue resistance, 51, 56
fracture toughness, 51
lateral bracing, 69
load cycles, 64
redundancy, 51
riveted connections, 108
shear studs, 72
specifications, 2, 25, 54
stress ranges, 65
summary of requirements, 58
tensiontype connections, 104
threaded rods, 107
webs, 85
Backing bar, 94
Bolted members, 102
preloaded, 103
nonpreloaded, 104
tension, 104
Bracing, 77
Bridges
box girder, 79
girder, 76
longspan, 81
multibeam, 76
Butt weld (see Groove weld)
Brittle fracture, 1, 17
Charpy Vnotch, 6, 11, 54
Combined stresses, 114
Connections:
bolted, 72, 102
diaphragm, 76, 80
groove weld, 93
fillet weld, 18, 66, 75
lateral bracing, 69, 83
lateral connection plate, 69, 83
riveted, 81, 108
shear studs, 72
splice plates, 66
Constant amplitude fatigue limit
(CAFL), 47, 57
Coped beam, 82
Corrosion, 100, 110
Cover plates, 19, 20, 22, 48, 91
Crack:
compression zone, 90
corrosion, 100, 110
critical length, 7
discontinuities, 1, 8
environmental effects, 110
future cracking, 100
growth (Paris equation), 9, 10
growth law, 29
initiation, 1
inspection, 7, 88, 95
length, 4
location, 4
low temperature, 113
mean stress effect, 10
modes, 3, 13
preexisting, 18
propagation, 10
repair, 97
residual stresses, 115
small, 6
stress concentration, 4
stress intensity factors, 5, 9
subcritical, 8
Cumulative damage, 34
Cross frames (see Diaphragms)
Cutoff limit, 49
129
Diaphragms, 76, 80
Distortioninduced fatigue cracking, 73
summary, 87
web gaps, 76, 79
Environmental effects, 110, 113
Equivalent stress range, 36
Eurocode, 120
European Convention for Constructional
Steelwork (ECCS), 49, 120
Examples, 13, 28, 31, 35, 37, 38, 44, 49, 59,
75, 85, 117
Fatigue:
constant amplitude, 47
definition, 1, 8, 16
displacementinduced, 25
distortion, 2,73
historical background, 16
limits, 47
limit state, 8
loadinduced, 25
Fatigue life (fatigue life assessment):
AASHTO, 56, 84
anchor bolts, 107
basis of design rules, 22
bolted, 72, 102
combined stresses, 114
controlling factors, 24
coped beam, 82
corrosion fatigue, 111
cumulative damage, 34
flange splice plates, 66
flangetoweb weld, 66
fracture mechanics analysis, 29
grade of steel, 22
lateral bracing attachment, 83
minimum stress, 23
PalmgrenMiner rule, 34
redundancy, 51
residual stress, 115
riveted, 108
shear studs, 72
size of component, 115
stress corrosion, 111
stress cycle counting, 41
stress histories, 40
stress range, 23
threaded rods, 106
transverse stiffeners, 67, 73
type of detail, 24
variable stress ranges, 34
web gap, 76
Fatigue limits, 47
constant amplitude, 47
Fillet welds, 17, 18, 90
Flaws:
in fabricated steel structures, 18,
21
mechanical details, 21
welded details, 18
Fracture:
toughness, 6, 51, 53
surfaces, example, 17, 19, 20,
Fracture mechanics, 3
analysis of fatigue life, 29
Charpy test, 6
crackgrowth model, 17
cracktip opening displacement
(CTOD), 5
defect assessment, 122
inspection interval, 122
Jintegral, 5
limit state, 6
microstructural, 6
modes, 3, 13
quantitative design, 11, 121
Groove weld, 93
Gusset plate, 91, 94
Inspection, 95
acoustic emission, 96
coring, 97
dye penetrant, 96
fracture mechanics, 122
magnetic particle, 96
methods, 95
ultrasonic, 96
visual, 95
130
Xray, 96
Kfactor (see Stress intensity factor)
Long span structures, 81
Longitudinal stiffener, 93
Limit state:
fracture, 6
fatigue, 8
National Cooperative Highway Research
Program (NCHRP), 16
PalmgrenMiner rule, 34
Rainflow cycle counting, 42
Redundancy, 51
Repair, 97
Cover plates, 98
Gas tungsten arc remelting, 99
Gouge, 98
Hole drilling, 91, 98
Methods, 98
Peening, 98
Refabricate, 98
Replacement, 99
Reservoir cycle counting, 42
Residual stresses, 115
compressive applied stresses, 120
effect on fatigue life, 24
stressrelieved elements, 120
nonwelded elements, 120
Riveted connections, 108
Rods (see Threaded rods)
Secondary members, 92
Shear splices, 102
Steel:
effect of grade upon fatigue life, 22
toughness, 8
Stiffeners:
longitudinal, 93
transverse, 67, 73
Stress:
combined, 114
compressive, 58, 120
constant amplitude, 47, 57, 106
cycle counting, 41
dead load, 41
equivalent, 36
gross section, 104
histories, 40
intensity factor, 5, 9
live load, 58
maximum, 23, 116
mean, 10
minimum, 23
net section, 104, 108
range, 23, 54, 67, 116
ratio, 24, 116
residual, 115
shear, 114
spectra, 47
stressrelieved, 120
tension, 104
threshold, 50
variable, 34
variable amplitude, 47, 57, 106
Stress Corrosion, 111
Temperature, effect of, 113
Threaded rods, 106
Threshold stress, 50
Toughness, 6, 53
Variable amplitude stress ranges, 34
Web gaps, 76, 92
Web stiffener, 67, 73, 93
Weathering steel, 110
Welding:
defects, 18, 20
effect on fatigue, 1, 8, 16, 18
H050
Pr int ed in U.S.A.
May 1998
Amer ican Institute
of Steel Construction
National Steel
Br idge Alliance
A Note of Caution
All data, specifications, suggested practices, and drawings presented herein, are based on the best available information and delineated in accordance with recognized professional engineering principles and practices, and are published for general information only. Procedures and products, suggested or discussed, should not be used without first securing competent advice respecting their suitability for any given application. Publication of the material herein is not to be construed as a warranty on the part of the National Steel Bridge Alliance – or that of any person named herein – that these data and suggested practices are suitable for any general or particular use, or of freedom from infringement on any patent or patents. Further, any use of these data or suggested practices can only be made with the understanding that the National Steel Bridge Alliance makes no warranty of any kind respecting such use and the user assumes all liability arising therefrom.
Copyright © 1998, National Steel Bridge Alliance
May 1998
A FATIGUE PRIMER for STRUCTURAL ENGINEERS
by
John W. Fisher Lehigh University Bethlehem, Pennsylvania, USA Geoffrey L. Kulak University of Alberta Edmonton, Alberta, Canada Ian F.C. Smith Swiss Federal Institute of Technology Lausanne, Switzerland
The authors also thank Dr. Their contributions are noted with thanks. Virginia. Minnesota. Massachusetts. Notification of errors and omissions and suggestions for improvements are welcome. Pennsylvania. "Anwendung der Bruchmechanik für die Ermittlung des Ermüedungsverhaltens geschweisster Konstruktionen. for his help with the preparation of Example 7. The document was proofread with great care by Jeffrey DiBattista. Graduate Student at the University of Alberta and helpful comments were provided on the technical content by Prof. Professor Hirt is the coauthor of a principal source document (Reference 1). Spofford and Thorndike Inc. Ohio. Maryland. The contributions of Manfred Hirt. Professor of Steel Construction and Director of ICOM. However. Greiner Engineering Sciences. Gilbert Grondin. ii . it must be understood that persons using the material assume all liability arising from such use.A. M. and West Virginia and the firms DeLeuw Cather and Company. South Dakota. Kozel Engineering Company. Swiss Federal Institute of Technology. and is used with his permission. Peter Kunz. are gratefully recognized. Figure 4 is taken from Hirt.. and Wiss Janney Elstner Associates Inc. The authors have taken care to ensure that the material presented is accurate. Jeffrey DiBattista did many of the calculations associated with the example presented in Chapter 5." Bauingenieur.. Fay.ACKNOWLEDGMENTS AND DISCLAIMER Several US state highway departments and highway design firms or consultants supplied information on various bridge structures reported in this document. No part of this publication may be reproduced or distributed in any form or by any means without the prior written permission of the authors. Modjeski and Masters Inc. 57 (1982). also of the University of Alberta. These include transportation departments in the States of Connecticut. Lausanne. formerly at ICOM.
..........1 2....40 Fatigue Limits...............4 3 How to Account for a Crack ..........59 iii ..........................................54 Fatigue Resistance..................5 Fatigue Load and Frequency .....................................3 Cumulative Fatigue Damage ...................................................................................................1 4....2 2................................... 2 Fracture Limit State ...............................22 Design Rules Given by the AASHTO Specification.....................................................................................................TABLE OF CONTENTS 1 2 Introduction ...3....11 Fatigue Strength Analysis 3........................................18 Basis for Design Rules .......................................................56 Summary of AASHTO Requirements.................................................................................................................4 3.................54 5......................... 6 Fatigue Limit State ..............................................1 5.......4 5............. 8 Fracture Mechanics Used as a Qualitative Design Tool .......3 3................51 Fatigue Design in the AASHTO Specification ....................................1 Basic Fracture Mechanics Concepts 2..........16 Sources of Flaws in Fabricated Steel Structures .....................................47 5 Fatigue Design According to the American Association of State Highway and Transportation Officials Specification (AASHTO) 5................................................34 Analysis of Stress Histories..........3.........51 Redundancy and Toughness.....................................................................25 Fracture Mechanics Analysis of Fatigue .................................................................................................................................................3 2..............3 Introduction ..............5 Introduction and Historical Background ............................................................................2 5............................58 Design Example .......................................................2 4......................................................................2 5...........2 3...............................29 4 Fatigue Assessment Procedures for Variable Stress Ranges 4..1 3.........................1 5................................
...............................................3 7..........................................................................2 6........1..................73 Examples of DistortionInduced Cracking..6 6..... 108 iv ........................76 6............................................1...................3............79 Long Span Structures ........................................................................95 Repair of FatigueCracked Members .......83 AASHTO Specification Requirements Relating to DistortionInduced Fatigue .................................................................................3............................84 Design Examples .................................................2 6....73 Further Examples of DistortionInduced Cracking ..................1................ 102 8..........................1 8...........4 6.....90 Methods for Inspection of Fatigue Cracking..................................1 Bolted or Riveted Members ..................................................5 7...7 Introduction .................2 7...89 Cracking at Low Fatigue Strength Details .........................................................4 7..3 Bolted Members ....................3..........................................6 Web Gaps in Multiple Girder and Girder Floor Beam Bridges .................. 100 8 Special Topics 8........85 Summary...........................................................................1 6.....................................6 DistortionInducted Fatigue Cracking 6.................4 6.......3 Introduction ..2 8.........................................................83 Other Examples ......3 6......................................................................88 Protocol for Fatigue Crack Investigation .................81 Coped Beam Connections ............76 Web Gaps in Box Girder Bridges..............3.....1 6................................ 102 Threaded Rods...........................3...........................3..............................................88 Identifying the Causes of Cracking ........................5 6....................................97 Avoiding Future Cracking Problems..........................5 6...........................87 7 Inspection and Repair of Fatigue Cracks 7.........................................................................1 7.................................................................................6 7..82 Connections for Lateral Bracing ............................................................. 106 Riveted Connections........
............................ 124 INDEX ......8..................................................................................2 8....... 115 Quantitative Design Using Fracture Mechanics.........................................................6 Environmental Effects.............3 8.............................5 8.. 128 v ........................................................ 114 Effect of Size on Fatigue Life ............................................................................................. 121 REFERENCES .............................................................4 8....... 110 Combined Stresses ....................... Use of Weathering Steel..................................... 115 Role of Residual Stress.......................................................
Chapter 1. Introduction
Fatigue in metals is the process of initiation and growth of cracks under the action of repetitive load. If crack growth is allowed to go on long enough, failure of the member can result when the uncracked crosssection is sufficiently reduced such that the part can no longer carry the internal forces. This process can take place at stress levels (calculated on the initial crosssection) that are substantially less than those associated with failure under static loading conditions. The usual condition that produces fatigue cracking is the application of a large number of load cycles. Consequently, the types of civil engineering applications that are susceptible to fatigue cracking include structures such as bridges, crane support structures, stacks and masts, and offshore structures. The first approach in the design and execution of structures is to avoid details that might be prone to cracking, and then to inspect the structure for cracks, both during fabrication and later in its life. However, it is inevitable that cracks or cracklike discontinuities will be present in fabricated steel elements, and it is the responsibility of the engineer to consider the consequences in terms of brittle fracture and in terms of fatigue. The fatigue behavior of a fabricated steel engineering structure is significantly affected by the presence of preexisting cracks or cracklike discontinuities. Among other things, it means that there is little or no time during the life of the structure that is taken up with "initiating" cracks. Probably the most common civil engineering structures that must be examined for fatigue are bridges. In North America and elsewhere, early steel bridge structures were fabricated using mechanical fasteners, first rivets and later highstrength bolts. In these cases, initial imperfections are relatively small. In addition, loading and load frequency were also low by today's standards. Consequently, fatigue cracking in these early structures was infrequent. In the 1950's welding began to be used as the most common method for fabrication. This had two principal effects related to fatigue. First, welding introduces a more severe initial crack situation than does bolting or riveting. Second, the continuity inherent in welded construction means that it is possible for a crack in one element to propagate unimpeded into an adjoining element. Design rules at this period of time had been developed from a limited experimental base and the mechanism of fatigue crack growth was not well understood. Furthermore, most of the experimental results came from smallscale specimens. This is now known to be a
1
limitation in evaluating fatigue strength: reliance on smallscale specimens can result in overestimates of fatigue strength. During the 1970's and 1980's there were many examples of fatigue crack growth from welded details now known to be susceptible to this phenomenon. Research revealed that the type of cracking observed in practice was in agreement with laboratory test results and supportable by theoretical predictions. Experience in the 1970's also exposed an unexpected source of fatigue cracking, that from distortions. This is also a phenomenon related largely to welded structures. The purpose of this publication is to provide the student and the practicing engineer with the background required to understand and use the design rules for fatigue strength that are currently a standard part of design codes for fabricated steel structures. The approach adopted establishes the basis for the problem in terms of fracture mechanics, that is, an analytical tool that accounts for the presence of a crack in a structure [1]. The focus is then directed specifically upon the issue of fatigue. It is intended that fundamentals are presented in a general way, but applications will refer to the specification for the design of steel bridges prepared by the American Association of State Highway and Transportation Officials (AASHTO) [2]. This specification is widely used in the United States, and the governing Canadian specification is nearly identical to it.
2
Chapter 2. Basic Fracture Mechanics Concepts
Use of the fracture mechanics method of analysis is relatively recent. Originally advanced to explain the rupture of glass specimens [3], its introduction into the field of structural engineering practice started when it was used in the 1940's to help explain the catastrophic failure of welded ship hulls. Currently, it is employed to assess the behavior of elements used in machinery, pipelines, automotive parts, spacecraft, turbine blades, and many other components. In this Chapter, basic concepts of the fracture mechanics approach are described in order to assist the reader in understanding the fatigue design rules. In addition, for those who might need to design at higher levels of sophistication, it will provide the basis for further reading and selfinstruction. Only a summary of fracture mechanics concepts is given in this section. For simplicity, the discussion is limited to cases where the loads are applied at locations remote from the crack locations and normal to the crack surfaces, the socalled Mode 1 situation. (The different ways in which a crack can open, or modes, will be explained in Example 1). Excellent review articles provide more detailed information (for example, see Ref. [4–6]) and several reference books are available [7–9]. 2.1 How to Account for a Crack Five cases of a loaded plate containing a crack are shown in Fig. 1. It requires no knowledge of fracture mechanics to appreciate that cases 1 to 5 are placed in order of increasing severity. Taking Case 1 as the basis for comparison, the following important
1 2 3 4 5
Decreasing Strength
Figure 1 Five Different Cases of a Plate Containing a Crack
3
ii) the crack opening relieves crack surfaces of stress. ii) crack location (at edge of plate in Case 3). however. the result of any of these parameters in weakening the plate will depend on the actual circumstances. A magnified view of the area around a crack tip in an infinitely wide plate is shown in Fig. d. 2. Centerline of crack σ yy σ Detail d y τxy σ xx r θ x cracktip plastic zone A crack tip A a . This resembles closely the conditions of Case 1 when the crack length is small compared with the plate width. the consequence of a sharp stress concentration in combination with a crack (Case 5) could weaken a plate to less than onehalf of its uncracked strength. Section A . For an elastic material. Using special stress functions. and iv) presence of a stress concentration (Case 5). plastic zones are formed since the strain exceeds the ability of the material to behave elastically.fracture mechanics parameters can be identified: i) crack length (Case 2).A crack front crack surface Figure 2 A Crack in an Infinitely Wide Plate If plasticity is ignored. theoretically this stress is infinite at the crack tip: in real materials. For instance. and the stress that this crosssectional area would have carried is diverted to the uncracked area of the plate. the crack opens a certain distance. σ. Of course. When a remote stress. is applied. and iii) crack tip plastic straining is created—is the fundamental mechanism that weakens structures containing cracks or cracklike discontinuities. a solution containing the coordinates r and θ is 4 . iii) effect of bending (Case 4). This diversion creates a high concentration of stress in the vicinity of the crack tip. This process—whereby i) an applied load causes a crack to open. a description of the stress field near the crack tip can be obtained. The effect can be significant.
the stress is σ yy = σ πa 2π r (1) provided that the crack length. an elastic– plastic analysis may be required when stress concentrations cause localized plasticity. Nevertheless. Usually. The most common analyses either use a parameter named J. of the crack length. and stress gradients.developed. K. also. Further description of elasticplastic analyses is available elsewhere [9] and the American Society for Testing and Materials has produced standards for 5 . These limitations are violated in many practical situations. the following expression is used to approximate K: K = WYσ πa (3) where Y is an expression that corrects for plate and crack geometry and W corrects for nonuniform local stress fields caused by the presence of factors such as residual stresses. For example. that is. Furthermore. solutions can be found in handbooks. r. and of the uncracked ligament. which is an expression of the change in potential energy with respect to crack length. then the stress intensity factor (K) approach is satisfactory. if this zone is small. Thus: K=σ πa (2) The advantage of this model is that any combination of stress and crack length can be characterized by the single parameter K. such correction factors are determined using numerical methods. is much larger than the distance from the crack tip. The numerator in Eq. Here. a. Analytical solutions are available for other particular geometrical configurations and loading conditions. However. these are summarized in handbooks [10]. In such instances. for the stress in the ydirection. say less than 2% of the plate thickness. This numerator is called the stress intensity factor. Equation 1 is based on linearelastic material behavior and cannot account for yielding at the crack tip. stress redistribution due to plasticity alters the stress field outside the crack tip plastic zone. stress concentrations. For the particular case of θ = 0. many practical cases cannot be solved analytically. or use a parameter called the crack tip opening displacement (CTOD). (1) determines the gradient of the (theoretical) stresses as they rise to infinity when r approaches zero.
In brief. Ref.. When plate dimensions are large enough to restrict behavior to essentially twodimensional straining (plane strain conditions).e. These models are not yet well defined.determining these parameters experimentally. service temperature. thereby disqualifying K as a model that characterizes combinations of stress and crack length. for most designs. when the crack length is greater than about five grain diameters.1. an equivalent K is calculated [3]: K2 = J = σ y CTOD E (4) where E is Young's Modulus and σ y is the effective yield strength. microstructural properties such as grain orientation influence crack growth [12]. A further restriction on K exists for small crack lengths. for example. (See. An accurate determination of the fracture toughness is complicated. this 6 . loading rate. the material constant. Microstructural fracture mechanics models may then become necessary. The designer should choose a steel with a fracture toughness level that is sufficiently high for the intended application. where υ is Poisson's ratio.) Occasionally. plate thickness. those employing K or J. E. This is primarily due to the fact that. and no generally accepted design rules are available. especially in most structural engineering design situations. For the usual situation. and fabrication processes. The fracture toughness depends upon such factors as microstructure and composition of the material. the models that assume an isotropic continuum. Less sophisticated approaches are used for practical problems in structural engineering. where all of the approaches explained above lose their validity. K or J) exceeds a critical value. Fracture Limit State ( ) The analysis tools available will indicate that fracture occurs when the crack parameter (i. When the size of the crack or initial discontinuity is of the order of grain size. the Charpy impact test. is replaced by E / 1 − υ 2 . Elasticplastic models are needed because the stress at fracture produces a plastic zone size that exceeds the limitations cited for the validity of K specified in Section 2. are sufficiently accurate.e. 2. [11]. The most widely used method for approximating the toughness quality of a steel is a procedure that was developed over 80 years ago.2.. commonly referred to as the fracture toughness. conditions of essentially twodimensional stress (plane stress) have an influence. i. plane strain conditions do not dominate.
The Charpy test. which result in less conservatism. In most modern steel structures. KIc. the likelihood of fracture shortly after erection has been completed is not high. inspection technology influences this verification: acr is compared with the minimum verifiable crack length.1 If required. In general. acr. At a given temperature. The engineer should ensure that the design configuration and the inspection equipment and procedures permit reliable detection of crack lengths less than acr. is less than acr. Therefore. correlation with fracture mechanics models are available for certain situations [8]. KIc and KId are conservative first estimates of the fracture toughness. Several tests provide a relationship between absorbed energy and temperature for the steel under investigation. temperature response curve can be converted into dynamic planestrain fracture toughness values. Elastic– plastic analyses. in Eq. Some problems may arise when severe weld discontinuities are present or when very thick plates are used or when the structure is in a very cold environment. a. Charpy impact data in the lower region of the energy vs. the toughness value where conditions change from plane strain to plane stress depends upon the yield strength and the thickness of the element. but such situations are not common. Moreover. Rational designs. the KId values then can be converted to static planestrain fracture toughness values. 3. A critical crack length. Fracture will not occur if the maximum size of the crack or cracklike discontinuity. The location of this change cannot be correlated to Charpy results. However. ao. For example.method measures the energy absorbed by the rapid fracture of a small bar containing a machined notch. a given material will always exhibit higher fracture toughness for plane stress conditions than for plane strain conditions in the elasticplastic region. The effect of temperature is examined by repeating the test using physically identical specimens that have been cooled to various temperatures. which employ high1The subscripts used here for K are explained in Example 1. and many other similar procedures [9] provide only qualitative information because stress and crack length values cannot be assessed directly. σ. The bar is broken by a swinging pendulum and the absorbed energy is measured by the difference in swing height before and after fracture. can be used when the design problem justifies increased complexity of work. using an empirical formula. 7 . KId . Often. such correlations are valid only for steels of low to medium strength. may be approximated by introducing the fracture toughness at the minimum service temperature and the maximum possible stress.
stress corrosion.) In civil engineering practice. if the fracture limit state is reached. Generally speaking. Consequently. Consequently. The combined effect of these trends is increasing the importance of fatigue strength evaluation. In addition to highertoughness properties. Furthermore. the initiation portion of fatigue life is essentially nonexistent for fabricated steel structures.strength finegrained steels and which use modern fabrication techniques. the greatest risks of fracture in modern steel structures arise when subcritical crack growth due to fatigue. and any such structure that is subjected to repeated loading. is possible. offshore platforms. This growth can even begin before the structure is put 8 . Such an occurrence is covered by the definition of the fatigue limit state. corrosion. higher service stresses have been allowed in recently built structures. acr. development of lowalloy and finegrain microstructures have increased the yield strengths of the steels used for construction. and are thus more resistant to fracture than ever before. welding is now used more often than formerly and this method of fastening leads to a lower fatigue life than would apply if the connection was made using rivets or bolts. Crack growth due to fatigue is discussed in the next section. 2. many structural elements remain susceptible to fatigue crack growth. the number of old structures. towers. examination of the possibility of fatigue cracking must be a consideration for bridges. As a result. careful assembly and improvements in the technology of nondestructive inspection are reducing the minimum verifiable crack length. All elements of a fabricated steel structure contain metallurgical or fabricationrelated discontinuities. Predicting the occurrence of the fracture limit state is becoming more dependent upon the correct choice of the subcritical crackgrowth model than upon an accurate estimate of the fracture toughness of the detail. At the same time. large critical crack lengths. Consequently. should provide high toughness and. Although steel structures today use highertoughness materials than was common in the past. (As has already been noted. that have exceeded their design fatigue life is growing exponentially. etc. cranes. modern structures are more susceptible to fatigue cracking than were older structures. it is often the result of fatigue crack growth after many years of troublefree service. fatigue failure is often the result of slow crack growth from an existing discontinuity at a stress concentration. Furthermore. consequently. and most also include severe stress concentrators such as weld toes.3 Fatigue Limit State Fatigue is the initiation and propagation of microscopic cracks into macro cracks by the repeated application of stresses. then. for example railway bridges. ao.
At very low growth rates. the threshold stress intensity factor. The values of ∆K are calculated using Eq. a σ 2a da σ σmax ∆σ σmin (a) σ dN time N log da dN m 1 fracture toughness K min DETAIL log da dN (c) (b) extrapolated design line R= high R low R K min K max log ∆ Κ ∆ Κ th intrinsic log ∆ Κ ∆ Κ th Figure 3 StressIntensity Factor and Fatigue Crack Growth. (a) Crack length vs. a. Eq. ∆σ ( = σ max − σ min ) . 3 becomes ∆K = W Y ∆σ π a (5) Equation 5 is related empirically to the crackgrowth rate. number of cycles. the curve for crack growth (in steel) becomes vertical. At higher values of ∆K. which is obtained from the slope of the curve of crackgrowth measurements—see Fig. (c) Magnification of the lower portion of the curve in (b) 9 . indicating a crackgrowth threshold at ∆Kth . For a constant stress range.into service (the result of transportation of a girder. for example). 3(b). and it becomes vertical again when the fracture toughness is approached at the maximum stress in the cycle. This slope is used as the ordinate in a plot of this parameter against ∆K.2. using a double logarithmic representation—see Fig. A description of the fatigue crack growth phenomenon can be made on the basis of the fracture mechanics model described in Section 2. 3 to account for repeated loading. da dN . (b) Crack growth rate vs. stress intensity factor range. the curve straightens to a nearconstant slope. The stress intensity factor K can be modified to represent fatigue crack growth by adapting Eq. 3(a). 5 for particular magnitudes of crack length.
) Also. R = σ min σ max . N: N= 1 af 1 ∫ ai ∆K m da A (7) where ai and af are the initial and final crack lengths. 3(b) is not perfectly straight. This is equally true for fatigue applications: nonconservative calculations may result if socalled short crack behavior occurs [14]. stress level. (The constants show up directly in the design specifications. These integration limits may take the values ao and acr . the Paris equation is useful [13]: da = A ∆Km dN (6) where A and m are constants that are determined by means of regression analysis of test data. Usually. The error resulting from the extrapolation is dependent upon the magnitude of the stress ratio. When ∆K is much larger than ∆Kth . as defined in Section 2. This ratio is often used to examine the effects of mean stress on crack growth.5.2. loadings. the stress intensity factor remains a useful characterizing parameter for conditions of fatigue crack growth since discontinuities are large and a high percentage of crack growth occurs under conditions where K is valid. 3(c). These constants are reliable when similar materials.Material properties. 6 can be integrated to calculate the crack propagation fatigue life. such situations are less common when assessing the fatigue limit state. as will be seen in Section 3. Fortunately. and environments are compared. 6 is extrapolated into this region. which is of considerable engineering interest.2 that small crack sizes and excessive plasticity may invalidate models that employ the stress intensity factor. the regression analysis is dependent upon the domain of crack growth rates considered since the center portion of the curve in Fig. In this center portion. Many structural applications involve repeated loading of over one million cycles: this requires a precise knowledge of slow crack growth rates near ∆Kth —see Fig. structural engineering applications have particular characteristics that reduce the occurrence of this anomalous behavior. Conservative assessments result when Eq. Eq. It was pointed out in Section 2. Moreover. and environment have greater influence in the end (vertical) portions of the curve than in the middle. respectively. 10 .
0 a = 10 mm f 0 σ 200 400 600 800 0. are presented below. The integral in Eq. This has resulted in an increased emphasis on quality assurance guidelines and on detail designs that have small discontinuities. such as Charpy energy values for fracture toughness requirements and stress range models for fatigue assessments (see Section 3. Older concepts.1 10 4 N (cycles) 10 5 (MPa ) 10 6 10 7 Figure 4 Examples of Studies that Examined the Significance of Discontinuities (a) Critical crack size vs. 4(a). an embedded discontinuity (such as an inclusion) is less serious than a surface discontinuity (such as a weld undercut). Because more parameters are explicit in fracture mechanics analyses.2. designers can identify more easily those parameters that influence the strength of the structure. Guidelines for simple linearelastic fracture mechanics models sufficient for most structural engineering designs are available in some design codes [15] and one illustration of the application of fracture mechanics analysis is given in Section 3.4 Fracture Mechanics used as a Qualitative Design Tool Engineering designers rarely use fracture mechanics as a design tool. Second.4). two trends are illustrated by the curves in Fig. applied stress. parametric studies. Small variations in the magnitude of the 11 . First. Some examples. the concepts of fracture mechanics enable the designer to increase his qualitative understanding of structures containing cracklike discontinuities. a 80 cr (mm) a a (a) 10 i (mm) a f = 40 mm a (b) a f = 20 mm 60 2a 40 20 1. covering the importance of discontinuities. Nevertheless. Fracture mechanics analysis clarifies the importance of discontinuities in fracture assessments. The size of the discontinuity clearly influences the resistance of an element to fracture. (b) Effect of initial and final crack size on fatigue life The effect of discontinuities is equally important when assessing the fatigue limit state. increasing the applied stress causes a decrease of the critical defect size.5. are the most practical tools for evaluating many structures. 7 has the limits ai and af . For example. and crack propagation behavior.
Figure 5 shows an end plate welded to a beam. provision of cope holes will give better access to the webtoflange junction. In the example. Connections classified by specifications as falling into different fatigue strength categories may have different crack propagation characteristics for the same fatigue life—see Fig. The initial crack size is far more important than the final crack size—see Fig. Fracture mechanics analysis facilitates recognition groove weld that any areas not welded become builtin cracks. 5. modeled by the parameter W in Eq. 6). (The assembly will subsequently be bolted to a column flange.) Design Figure 5 Unwelded Areas can guidelines already recognize that flange fillet welds. not shown. A fatigue crack that starts at the surface of the material initially propagates very slowly into the plate thickness. may not significantly alter the resulting fatigue life.) fillet Groove welding may not always be able to penetrate welds completely into the zone through which welding is intended. af. (However. the detail with lowerseverity stress concentration (groovewelded plate) consumes a greater percentage of its total fatigue life during the propagation of 12 . produce a detail of low fatigue strength because of the large unwelded areas (cracks). 4(b). the flange groove weld will be difficult to complete in the vicinity of the webtoflange junction. affects the crack growth rate. an become BuiltIn Cracks alternative to the groove welds shown. The stress concentration. 6. a (mm) a cr a a a0 N (cycles) 2x10 6 Figure 6 Comparison of Fatigue Crack Growth Behavior for Two Different Details Having the Same Fatigue Life In this example (Fig.final crack size.
) a σ (a) Through2a thickness crack 2c σ (b) Surface crack σ a (c) Edge crack K I = σ πa K I = 112 σ πa / Q .g.1 is subscripted. Figure 8 KI Values for Various Crack Geometries (Infinitely Wide Plates) 13 . KI would be used for a Mode I case. will almost always be the applicable condition. In structural engineering applications. 7. the crack opening mode. depending upon whether the loading is essentially static (KIc) or is dynamic (KId). EXAMPLE 1 Crack loading modes are shown in Fig. such crack growth behavior should be considered when establishing inspection intervals. Other factors such as differences in inspection feasibility influence the exact timing of crack identification.. e. The value of the stress intensity factor at which brittle fracture will occur is designated as KIc or as KId . Several common cases are shown in Fig. This means that a crack of a given size may be identified earlier at the attachment than at the groove weld.a crack to a size ao than does the detail with a higherseverity stress concentration (attachment detail). Nevertheless. y x y x z y x z z MODE I MODE II MODE III Figure 7 Basic Modes of Loading Involving Different Crack Surface Displacements Published solutions for the expressions for the stress intensity factor are available for a wide variety of conditions [10]. 8. In order to distinguish among the various possible cases. K I = 112 σ πa . (The term Q in part (b) is a correction for the presence of an elliptical surface flaw. the stress intensity factor K introduced in Section 2. Mode I.
a = 0. failure by brittle fracture can occur.) Design stress = 140 MPa Question: (a) What is the flaw size at which brittle fracture might be expected? Solution: It is given that the value of the critical stress intensity factor is KIc = 66 MPa m1/2. These illustrations are intended to depict the case of an "infinitely wide plate. What is the tolerable flaw size now? Solution: σ π a = 66 MPa m 14 . It is simply a matter of examining the situation KI → KIc .") Given: σ y = 550 MPa KIc = 66 MPa m1/2 (This information is provided by the supplier of the material and is related to the intended service temperature. 8. where a centrally located throughthickness crack is present in a plate that is loaded by a uniform tensile stress.071 m = 71 mm Thus. 2a = flaw size = 29 mm. or σ π a = 66 MPa m Using σ = 140 MPa and solving. the general expression for the stress intensity factor is σ 2a Throughthickness crack: K I = σ π a Figure 9 KIc = σ π a . loading rate. Question: (b) If the design stress is increased to 310 MPa. what is the tolerable flaw size now? Solution: σ π a = 66 MPa m π a = 66 MPa m 310 MPa Solving. 8 and 9.27 m a = 0. (It is to be understood that the crack width is small relative to the plate width in the illustrations of Figs. 9. if the flaw size reaches 2a = 142 mm. From Fig. Question: (c) Residual stress due to welding is present and it is estimated that the total stress (design stress + residual stress) is 500 MPa. and the particular plate thickness.Consider the element shown in Fig.
it is not possible to superimpose stresses due to residual stresses because— 1. The distribution of residual stress is different from the applied stress. superposition is applied at the level of K. 3. Comment: Often.500 MPa π a = 66 MPa m Solving. the parameter that characterizes the stress field at the crack tip. There is a stress concentration present. remote stress). the value of K for residual stresses should be calculated separately and then added to the K for the applied loads. Thus. 2a = flaw size = 11 mm. but not for the residual stresses. 15 . (Recall that the expression in Fig. 2. In such cases. This requires that a correction factor for K be used on the applied loads. 8 is valid only for a uniform. There is a possibility of residual stress relaxation with crack growth.
1 Introduction and Historical Background Fatigue cracking was observed in railroad equipment over 120 years ago. detail type. The approach taken by modernday specifications for the fatigue design of fabricated steel structures is based primarily on work done in Great Britain [17] and in the USA [18–20] in the late 1960's and early 1970's. in part. The industrialization of society and the subsequent increased use of machinery and equipment led to other examples of failures resulting from fatigue cracking. For example.Chapter 3. it was this research that identified the influence of residual stress on fatigue life. The early provisions for fatigue life evaluation proved to be inadequate for a number of bridge details. the relatively large number of cases of fatigue cracking in bridges that were designed prior to about 1975. As a result. in North America the observation of cracks in railroad bridge truss hangers and in stringer end connection angles led to a number of laboratory investigations between 1930 and 1960. Fatigue cracks forming in steel bridges at a road test program conducted in the USA in the 1960's [16] became the genesis of the fatigue test program sponsored by the National Cooperative Highway Research Program (NCHRP) that began at Lehigh University in 1968. types of steel. In terms of the fracture mechanics model 16 . Fatigue can be defined as the initiation and propagation of microscopic cracks into macro cracks by the repeated application of stress. studies into the phenomenon started in both Europe and in North America. the fatigue design rules that existed for welded steel bridge components were based on small specimens and on a limited quantity of test data. Studies carried out at that time by Wöhler on railway rolling stock showed that stress concentrations and sharp angles in the axle configuration resulted in failures even though the stress in the material was well below its yield strength. Although many other investigators have contributed to our understanding of the problem. These and later studies following World War II formed the basis for the early fatigue design specifications in North America. Prior to the NCHRP program. These studies also revealed the necessity to acknowledge that fabricated steel structures always contains cracks or cracklike discontinuities. and quality of fabrication. both before and after the work cited. This explains. Fatigue Strength Analysis 3. This made it difficult to establish the significance of stress variables. Welded details were first examined in the 1930's when tests were carried out on welded steel details.
One possibility is that the fatigue crack grows to such an extent that the loss of crosssection means the load simply can no longer be carried by the uncracked portion of the beam. which is initially sharp. Growth occurs at the crack front. The process is repeated during each load cycle. Figure 10 shows the fracture surfaces of a member that has an Ishaped crosssection. (The full thickness of the flange is not shown in Fig. and the crack grows a minute amount during this process. the crack front eventually reached the exterior surface of the weld. As discussed in Section 2. A good explanation of the crack growth mechanism has been provided by Broek [21]. 10. failure occurs by yielding of the remaining material.described in Chapter 2. at this point the beam was deliberately overloaded so that the remaining crosssection fractured and could be exposed. Experience in the laboratory shows that as much as 80% of the fatigue life has been consumed by the time a fatigue crack emanating from an internal flaw reaches the surface and can be observed. 10 had not been terminated by the investigators. In the case illustrated in Fig. or. by instability if the crack growth produces a grossly unsymmetrical crosssection. was filletwelded to 13 mm thick flange plates. Continued slip results in a blunted crack tip. exceptionally. The web of the member. growth of a crack by fatigue can lead to brittle 17 . Upon unloading. If the test that produced the specimen shown in Fig. 10). The profiles of the fillet welds are generally satisfactory and the flow lines of the weld show good penetration of the base metal. In this illustration. The other way that the beam can fail is by brittle fracture. an internal flaw in the lefthand fillet weld grew under the repeated application of stress until the crack penetrated the outside surface Figure 10 Fracture Surface of Ishaped Member of the weld. and plastic deformation (slip on atomic planes) therefore occurs at the crack front. the crack tip again becomes sharp. not necessarily to zero. failure could have occurred in one of two ways. which was 10 mm thick. Even at relatively low loads. In this case.3. there will be a high concentration of stress at the sharp front. an initial crack grows a small amount in size each time a load is applied. Since this was a laboratory specimen.
In a welded beam without attachments (simply two flange plates welded to a web). 10 started at a nonmetallic inclusion). or trapped slag. temperature. weld repair or startstop of weld or arc strike undercut lack of fusion partial penetration porosity Figure 11 Flaws in a FilletWelded Detail Test data on welded details have demonstrated that all fatigue cracks commence at some initial discontinuity in the weldment or at the weld periphery.2 Sources of Flaws in Fabricated Steel Structures The kinds of flaws that can occur in a filletwelded detail are shown pictorially in Fig. and grow perpendicular to the applied tensile stresses.fracture if the crack reaches a critical size according to the particular conditions of material toughness. Although the fabricator of the structure and those responsible for the fabrication inspection will attempt to minimize these defects. and cracking or inclusions around a weld repair or at startstop locations or at arc strikes. Figures 12 and 13 show fatigue cracks that have formed from porosity and entrapped slag in longitudinal submerged arc fillet welds. incomplete fusion. most laboratory fatigue cracks are observed to originate in the webtoflange fillet welds at internal discontinuities such as porosity (gas pockets). 3. 11. it is neither practical nor economically possible to eliminate them. undercut or micro flaws at the weld toe. porosity and inclusions (the fatigue crack shown in Fig. These discontinuities are always present to some degree. 18 . These include partial penetration and lack of fusion. and loading rate. irrespective of the welding process and techniques used during fabrication.
19 .Figure 12 Fatigue Cracks Forming from Internal Porosity in Web–Flange Connection Figure 13 Fatigue Crack Enlarged to ThreeEnded Crack From Internal Porosity Attachments such as cover plates. thus forming a line of elevated tension where fatigue cracking can start from small. Figure 14 shows a fatigue crack that has formed at a cover plate fillet weld toe. stiffeners. and other components welded to a web or flange introduce a transverse weld periphery (toe). gussets. sharp discontinuities.
since the crack represented by the lack of penetration is parallel to the direction of the (bending) stress field. Furthermore. the crack will not open up under the application of stress and failure by fatigue is unlikely. 10) is a natural consequence of the filletwelded connection: it is not expected that the two fillet welds will merge in the central region of the connection. 7). the partial penetration shown in Fig.(a) Fatigue Crack at End of Cover Plate Fillet Weld Toe (b) Crack Surface Showing Fatigue Crack Growth Figure 14 Fatigue Cracking at Fillet Weld Toe In some cases. a "defect" is an expected result of the type of fabrication process and has no effect on the life of the member. 20 . For instance. 11 (which is also seen in the welded beam of Fig. a socalled Mode II crack (see Fig.
any mechanical detail has a better fatigue life than does its equivalent welded detail. The types of flaws introduced when welding is used have already been discussed. In a rolled shape. The task of the structural engineer is to be able to proportion those structural members that have a potential for failure by fatigue crack growth so that they have a sufficiently long life as compared with the design life of the structure. and shearing. In this case. or some other kind of alteration. However. Broadly speaking. etc. however. the disturbing effect of the hole is largely mitigated by the presence of the high local compressive stresses introduced by the bolt. fatigue crack growth can start from one of these sources. Comparatively. present a somewhat more severe fatigue life situation than the bare rolled shape. straightening." but. The design will be carried out in the expectation that flaws will be present initially in all fabricated steel 21 . As will be seen. In addition to the fact that more flaws will be present when welding is used. irregularities in mill scale. but the difference usually is not very great. laminations. in which holes are drilled or punched and forces are transferred by means of rivets or bolts. the "crack" does not grow and therefore its presence does not affect the fatigue strength of the member. this will be done in the environment that some probability of failure must be accepted: in real terms. cutting. there is no structure that can be designed for zero probability of failure. there are not many practical cases in which a rolled shape does not have some kind of attachment. Repairing defects in welded details is also difficult. inspection for defects is more difficult than is the case when mechanically fastened details are used. Punched holes give a greater reduction in fatigue life than do drilled or subpunched and reamed holes because of imperfections at the hole edge arising from the punching process. connection. If preloaded highstrength bolts are used. and from mechanical notches due to handling. Prohibiting the use of welded details in fatigue situations is not usually a practical option. since the discontinuity is parallel to the stress field. inclusions. the crack usually starts at the edge of the hole. seams.. The flaws that exist in all fabricated steel structures are a consequence of the manufacturing process of the steel itself and the normal fabrication processes. Flaws in rolled shapes arise from surface and edge imperfections. Mechanical details. the "unaltered" rolled shape presents the most favorable fatigue life situation. The region between bolt lines could be described as a "flaw" or "crack.Consider a detail involving mechanical fasteners—an Ishaped beam with a cover plate fastened to the beam flange with bolts. Drilled or subpunched and reamed holes give some reduction in fatigue life as compared with an unaltered member.
the welding specifications. the number of cycles of load to which the member was subjected. some aspect of the stress itself. Figure 15 Effect of Grade of Steel on Fatigue Life of Beams with Transversely EndWelded Cover Plates Figure 15 shows a plot of the stress feature versus the number of cycles to failure for a given detail (beams with transversely endwelded cover plates) for three different grades of steel. e. This latter feature is not yet welldeveloped in design specifications. Given that a certain amount of scatter is always present in results of this kind. A concomitant feature is that in the design process it is possible to identify the size of flaws that are permissible and then to use this information as the basis for both initial inspection of the structure as well as ongoing inspection. and the stress concentration present.g. it was evident that the features that needed to be examined included the yield strength of the steel (reflected by the grade of steel). it is evident that the yield strength of the steel does not have an important 22 .structures and that all such members will contain residual stresses of relatively high magnitude.. and the usual procedure is to accept as permissible flaw sizes consistent with the specifications that accompany the fabrication processes. These particular steels represent the spectrum of steel strengths in use at the present time—from 250 to 690 MPa.3 Basis for Design Rules To those working in the research area. 3.
this means that the maximum stresses had to have been 38. 15. Customarily. however). that is. 16.influence on the results. In Fig. First. this choice is made for the designer by the specification itself. the data plot as a straight line in this log–log representation. Second. Examine. 16. the algebraic difference between the maximum stress and the minimum stress at the critical location ( ∆σ r = σ max − σ min ). 16 were done at values of minimum stress ( σ min ) equal to – 41.8 MPa. for example. For this stress range. the data plotted for the stress range of 80 MPa.9 MPa. The tests represented in Fig.4 MPa. respectively. a spectrum of stress range.6 MPa. and 149 MPa. the dashed parallel lines plotted two standard deviations (measured relative to the horizontal axis) away from the mean line contain most of the test results. for the three values of minimum stress. and 68. of course. Other information contained in this figure is significant and is typical of all fatigue strength tests results. Figure 16 Effect of Stress Range and Minimum Stress on Fatigue Life for Welded End of Coverplated Beams The way the stress feature should be represented is examined in the data contained in Fig.8 MPa. the data are contained within a reasonably welldefined band about the mean. 93. and. the effect of stress is introduced as the stress range. The degree of scatter in the results will have to be considered when a choice is made for a design life line. It is obvious that the data are closely grouped and that the minimum stress per se does not 23 . 13. Considering again the endwelded coverplated beam (both asrolled and threeplate welded beams are represented in Fig.
This is simply to arrange standard structural details into categories relative to their expected fatigue life. it would be possible to determine the fatigue life of that member. These are: 24 . Figure 17 Fatigue Strength of Welded and Coverplated Beams In summary. Basic Fracture Mechanics Concepts. illustrated in Fig. In accordance with the ideas developed in Chapter 2. if the designer had one of the two types of members shown in Fig. As will be seen in Chapter 8. however. this is not a practical solution.. it is the presence of high levels of residual stress that dictates that stress range is the controlling stress parameter for a description of fatigue life. Clearly.e. and a more expedient approach is taken.influence the results. σ max σ min ). or stress ratio (i. For example. such as the beam illustrated by Fig. the fatigue life of a fabricated steel structure is determined by three factors. it would be expected that stress intensity (K) should be represented in the fatigue life evaluation. rather than maximum stress. 10. 17. 17 are the fatigue life representations for two different categories—beams which have cover plates that include a weld across their ends and beams made up of three plates welded together. minimum stress. For the usual level of design.
Both axes are logarithmic representations. and for this situation the stress range to be calculated is simply that corresponding to the nominal stress at the location of the detail. and 3.1. are those corresponding to the loads on the structure. and it will be discussed in a separate section (Chapter 6). the fundamentals of the AASHTO rules will be introduced. and the discussion in this section will continue to be limited to loadinduced fatigue cracking. This feature will be discussed subsequently. which are the driving force behind crack growth. the fatigue life line is horizontal. Over some portion of the range. strains) that are produced as a result of displacements. In this section. The subject is more completely discussed in Chapter 5. which depends on the Detail Category. There is another source of stresses in the structure that can produce crack growth. It has been implicit in the discussion so far that the stresses. Further elaboration will be required on a number of points. 3. however. Beyond a certain point.4 Design Rules Given by the AASHTO Specification The basis of the fatigue life rules given in all codes. or specifications for the design of fabricated steel structures has been given in the preceding sections of this Chapter. the AASHTO fatigue life rules reflect the two issues—fatigue cracking induced by stress and fatigue cracking induced by displacements within the structural system. Displacementinduced fatigue cracking is presented in Chapter 6. Displacementinduced fatigue cracking is at least as important as loadinduced fatigue cracking. but it is appropriate to introduce next the rules given by one of the mostwidely used standards. Only the first has been explained so far. The plot shows stress range on the vertical axis and number of cycles on the horizontal axis for seven different Detail Categories. This is indeed an important case. 2. As in most contemporary standards. the type of detail under examination. the stress range at the location of the detail. This is the stresses (really. This is valid because selection of the detail itself implies inclusion of the stress concentration for that detail. each Detail Category is a sloping straight line with a slope constant m equal to 3. The number of cycles of loading to which the member is subjected. standards. that of the American Association of State Highway and Transportation Officials (AASHTO)[2]. 25 . Figure 18 shows the fatigue life curves given in the AASHTO Specification.
The information in Fig. 18 must be used in conjunction with information like that shown in Table 1 and Fig. which give only a small portion of the relevant material in the AASHTO Specification. 500 Stress Range (MPa) 100 A B B' C D E E' 10 10 5 10 6 10 Fatigue Life (cycles) 7 10 8 Figure 18 Fatigue Life According to the AASHTO Specification 26 . 19.
19. the "exact" value of the permissible stress range at 2x10 6 cycles is 125 MPa. 7 Builtup Members Base metal and weld metal in component. This number was obtained using Fig. this is Detail Category B.General Condition Situation Detail Category Illustrative Example (see Fig. According to Table 1 and Fig. As will be seen later. the AASHTO Specification provides information that allows the calculation of the permissible stress range corresponding to a given number of cycles. shifted horizontally to the left by two standard deviations. For reasonably large numbers of test data. If the number of cycles to which the beam will be subjected is. For example. or • continuous fillet welds parallel to the direction of applied stress • continuous fullpenetration groove welds with backing bars in place. The fatigue strength curves presented in the AASHTO Specification (Fig. etc. Working out the equation of the line for this detail category. connected by: • continuous full penetration groove welds with backing bars removed. B B B' B' Table 1 Example of Fatigue Categories used in the AASHTO Specification Table 1 must be used in conjunction with the type of information shown in Fig. etc. 16 AASHTO) 3. 10. Application of the information is straightforward. where one of the relatively large number of typical construction details classified by the specification is shown. suppose a designer proposes to use a beam made by joining three plates using continuous fillet welds parallel to the direction of stress. usually as obtained by physical testing. 18 to estimate the stress range corresponding to 2x10 6 cycles. 2x10 6 . such as was shown in Fig. 18) are those corresponding to the mean life of a detail. 5. or • continuous partialpenetration groove welds parallel to the direction of applied stress. without attachments. say. then the permissible range of stress for this detail is 120 MPa. 4. the Fillet weld or groove weld Illustrative example 4 (refer to Table 1) Figure 19 Illustrative Example for Table 1 27 . 19.
The crane travels in such a position that the crane girder receives a maximum 80% of the total load as a reactive force. Detail classification – According to the AASHTO Specification. Since the actual range of stress (188 MPa) is less than the permissible range of stress for this detail (300 MPa). From Fig. 28 . Information from the owner is that the crane will make no more than two trips per hour at this load level. and the design life of the building is 40 years. The moment of inertia of this section is 448x10 6 mm 4 . The section used for the girder is to be made by filletwelding three plates into an Ishape. The flange plates are 350 mm wide by 22 mm thick and the web plate is 306 mm by 14 mm. read the Detail Category B line at N = 208 000 cycles to find that the permissible stress range is approximately 300 MPa. the work schedule will not exceed 10 hours per day five days per week./yr. 2. Calculate actual stress range – σmin = 0 σmax M = PL/4 = (300 x 103 N x 0.) (52 wk. in this case) – N = (2 cycles/hr. Number of stress cycles (equals number of load cycles. 18. Is the fatigue life of this crane girder satisfactory? Use the AASHTO Specification. It can be assumed that this force comes onto the girder as a single concentrated load. the situation is satisfactory. Solution: 1.) (10 hr./day) (5 days/wk. The fillet welds will be continuous. 3.) (40 yr.8) (8 000 mm)/4 = 480 x 106 N mm ∴ σmax = M y/ I = (480 x 106 N mm) (175 mm)/(448 x 106 mm4) = 188 MPa Thus.) = 208 000 cycles. this is Detail Category B. The main use of the crane will be to transport a 300 kN ladle from one end of the shop to the other.corresponding confidence limit is estimated to be approximately 95%. ∆σ r = 188 − 0 = 188 MPa. EXAMPLE 2 The overhead crane in a small manufacturing operation uses a simplysupported crane girder of 8 m span. this will be the only significant load.
the number of cycles permitted by the AASHTO Specification for a stress range ∆σ r = 188 MPa is N = 600 000 cycles.5 Fracture Mechanics Analysis of Fatigue In Chapter 2. In Section 2. obtained from Fig. Another way of looking at the problem is to compare the number of cycles that would be permitted at the actual stress range of 188 MPa with the number of cycles that actually occur. 6 gave the following expression for crack growth propagation (see also Section 2. 18 or calculated as will be shown in Section 3. the change in stress ( ∆σ r ) is always simply equal to the change in stress produced by the moving (i. The number of stress cycles is not always equal to the number of load cycles. this was also written (Eq.. K.3.Comments: 1. Since stress due to dead load is always present in the member. The crack growth law was identified in Section 2. when a multiple axle vehicle traverses a member or when continuous beams are used.3): N = 1 af 1 da A ∫ a i ∆K m 29 (7) .5. 3. Basic Fracture Mechanics Concepts.e. m = numerical constants determined from regression analysis of test data ∆K = change in stress intensity factor corresponding to a given change in stress range. as could occur. In this example. it should be possible to use the fracture mechanics method of analysis to deal with the fatigue strength problem. 2. it was put forward that brittle fracture and fatigue are crack growth phenomenon that are characterized by the same parameter. for example. Designers should be alert for cases where a single passage of load produces more than one stress cycle. live) loads. As such. 3. 5) as ∆K = W Y ∆σ πa (5) Integration of Eq.3 as da = A ∆K m dN (6) where a = crack length N = number of cycles A.
Also shown in the table are the values of the stress range that identifies the horizontal portion of each curve. any term appearing with the limit af can be neglected since the limit will appear with a negative power. Since the final crack size is always very large as compared to the initial crack size. Application of Eq. The AASHTO Specification uses m = 3 for all fatigue categories. the term within the integral is also a constant. 8 and 9 can be made in Example 2. Designating the product of all of the constant terms as M and using the more common notation ∆σ r instead of ∆σ . This is precisely what is observed in the physical tests. Making the substitution for ∆K. for a given geometry and starting crack size.3x1011 = 208 000 1 3 = 266 MPa 30 . These values. The AASHTO Specification provides values of the constant M in Eq. termed "A" in the AASHTO Specification. In step 2 of the Solution. the result can be written finally as N = M ∆σ − m r ( ) −m (8) or. However. The terms W and Y vary with the crack length. and of course the term a also contains the crack length. 8 for each fatigue category. alternatively. See. 8 as M ∆σ r = N 1 3 39. this becomes: N = 1 A ∫ a [W Y ∆σ i af 1 π f a ]m da = 1 ∆σ − m A ( π ) −m ∫ aa W Y 1 a m ( ) i da The terms 1/A and π are constants. respectively. and 17. as log N = log M − m log ∆σr (9) Equation 9 defines a sloping straight line on a plot of log stress range versus log number of cycles. Figures 15. the permissible stress range for N = 208 000 cycles for this Detail Category B now is calculated according to Eq. are listed in Table 2 for each fatigue Detail Category. for example. the socalled threshold stress.where ai and af are the initial and final crack length. 16.
4x1011 14. the number of cycles permitted for ∆σr = 188 MPa had been estimated using Fig." It should be obvious that it is much more expeditious to calculate values of fatigue life or permissible stress range than to try to read them from a loglog plot (Fig.7 69.3x1011 ) was obtained from Table 2.3 31.S.346 mm.7 48.3x1011 20. 1 All the source data in this problem were expressed using U.0 17.28x1011 Threshold Stress (MPa) 165 110 82.21x1011 3. Customary units.The value of M ( 39.73 log ∆σr.3x1011 (188) −3 = 591 000 cycles ( ) Table 2 Constants for use with Figure 18 Detail Category A B B' C C' D E E' Constant.9 EXAMPLE 3 The beam whose failure surface is illustrated in Fig. The necessary conversions have been made. The flaws from which the cracks initiated were generally circular and the measured average size was 1.4x1011 7. Use fracture mechanics analysis to verify the experimentally obtained regression line.32 − 2.61x1011 1. A ( MPa) 3 82. 8) as N = M ∆σ −3 = 39. where it is listed according to the Specification notation as "A. 31 . Using the test data1. the regression line expressing the relationship between fatigue life and stress range was determined to be log N = 12. 18 as 600 000. 10 was one of a series of nine tested [22].0x1011 39. It now can be calculated (Eq. 18).0 82. In Comment 3 of Example 2.0x1011 14.
the term af can be neglected because it appears with a negative power in this equation. in the crack growth equation (Eq. the equation of the fatigue strength line as obtained using the experimental data and a fracture mechanics analysis is 32 . 5 can be taken as a constant.346) − 0. The combined multipliers of ∆σ r are the term M in Eq.m 2 A Calling m/2 – 1 = q .74 Thus.5 = log −13 0. 6) can be taken as 218x10 −13 for ferritepearlite steels. value of C in the stress intensity factor expression can be taken as 2.5 π 213x10 . = 12. This means that the stress intensity modifiers W and Y in Eq. 8.Solution: It has been noted that the final size of the crack will not be influential in the result of Eq. 7. They also suggest the . Eq. 7 gives N= 1 −m C ∆σ − m r A ∫a af i a − m / 2 da af ai 1− m 2 1 −m −m a = C ∆σ r 1. A. q = (m/2) − 1 = 0.. 0 / π for a circular crack in a plate of infinite width. 9. the crack propagation equation can be further simplified to N= 1 − m − m 1 −q a C ∆σ r A q i Rolfe and Barsom [8] suggest that the constant of proportionality.5. i. and we can now solve for this value. this can be written as N= 1 m 1 −q C ∆σ − m a i − a −q r f q A ( ) Since the final crack size is always very large as compared with the initial crack size. Thus.e. we solve for log M as – 1 1 −q log M = log C−m a q i A −3 1 1 2 (1. and substituting for ∆K in Eq. Calling the product WY = C . Using the alternative form. and that the term m can be taken as 3.
It gives confidence in the prediction of cases not tested experimentally when those predictions are based on presumption of reasonable starting flaw sizes and shapes. However.73 log ∆σr.32 − 2.74 − 3 log ∆σr This is in good agreement with the fatigue strength line obtained from the experimental data exclusively. the results of this analysis and others like it do identify that the behaviour observed in the laboratory can be substantiated by an analytical prediction—the fracture mechanics analysis. 33 .log N = 12. log N = 12. this type of exercise is not directly useful to a designer since it requires knowledge of the type and size of flaw that is likely to result in fatigue crack growth and failure. Comments: Obviously.
when residual stresses are high and when plasticity is restricted (usually the case in structural engineering applications). the approach gives reasonable correlation with test data and it has the considerable advantage that it is easy to use. Methods for dealing with these problems are outlined in this Chapter. This is the linear damage rule first proposed by Palmgren in 1924 and further developed by Miner in 1945 [23]. things are not simple in either of these categories: fatigue loading is usually quite complex. However. It is known as the PalmgrenMiner rule. The total damage from all stress range levels that are applied to the detail is. It is to be 34 .Chapter 4. of course. the one that is customarily used in civil engineering practice is a linear theory that is easy to understand and apply and which gives satisfactory results. the sum of all such occurrences.1 Cumulative Fatigue Damage In this section. and it assumes simply that the damage fraction that results from any particular stress range level is a linear function of the number of cycles that takes place at the stress range. It should also be noted that the term "failure" in these definitions is not intended to be taken literally. Fatigue Assessment Procedures for Variable Stress Ranges In all of the discussion so far. 4. The designer has to deal with the reality that stress ranges of different magnitude take place at the detail and that these stress ranges are applied for varying numbers of cycles. This can be written in equation form as: ∑ Ni n = 1 (10) i where ni = number of cycles that take place at stress range level i Ni = number of cycles that would cause failure at stress range level i. As might be anticipated. both of these factors are not consistent with observed behavior. It has two major shortcomings [23]: it does not consider sequence effects and it is independent of the average stress in the cycle. it is known that these factors have only a small influence. Moreover. The AASHTO Specification [2] advises that the PalmgrenMiner rule can be used to account for cumulative damage. a method is presented that accounts for the damage that results when fatigue loading is not applied at a constant harmonic amplitude. Although both linear and nonlinear damage theories are available. it has been implicitly assumed that the stress range at the detail under investigation is unique and that counting or predicting the number of cycles is straightforward. To at least some degree. The rule is obviously very simple.
Number of cycles to failure at the new load level – In Example 2. 4. the stress range under the 300 kN loading was found to be 188 MPa.) (40 yr.3x1011 . this means that the value of this constant is 39. the builtup beam composed of three plates joined by continuous fillet welds. According to Example 2. it will be necessary for the crane to be able to accommodate one trip per hour at a load level of 350 kN. fabricated. 10: ( ) 35 . EXAMPLE 4 The beam of Example 2 was designed. the value represented by the mean life less two standard deviations on the log stress range vs. Recalling that the AASHTO designation A of Table 2 is equivalent to the constant M in Eq.) (10 hr. 8. log number of cycles plot.) Is the fatigue life of this crane girder still satisfactory? Use the AASHTO Specification.3x1011 ( 219) −3 = 374 162 cycles ( = N 2 ) Finally. the number of cycles at the old load level was 208 000 ( = n1 ) 2. Solution: 1. checking Eq. Number of cycles at the new load level is – N = (1 cycle/hr.) = 104 000 cycles ( = n 2 ). using m = 3 this becomes: N = M ∆σ −3 N = 39. The stress range for the 350 kN load can be calculated by proportion as (350/300) (188 MPa) = 219 MPa. 8.interpreted as the permissible fatigue life. According to Example 2. that is. in Eq. is a Category B detail according to the AASHTO specification. and erected when the owner decided that. the number of cycles to failure at the old load level was 588 000 ( = N1 ) 3. The detail that is under examination. (See Example 2 for all other details. Thus. in addition to the loads that had already been stipulated./day) (5 days/week) (52 weeks/yr.
We wish to calculate an equivalent constant amplitude stress range.e. It is often convenient to express the MinerPalmgren cumulative fatigue damage rule (Eq. N i = M ∆σi− m applies to the left hand side of Eq. can now be applied to Eq.28 = 0. The right hand side expresses the damage under the constant amplitude equivalent stress range. ∆σ e .. the crane girder is still satisfactory under the new loading condition. 10). it is required that ∑ Ni n = i Σ ni Ne (11) In Eq. Thus. the left hand side represents the damage under the variable amplitude stress cycles. that is. Making the substitutions— ∑ M ∆σi − m i n = M ∆σ e − m Σ ni The term M is a constant and can be eliminated from the equation. Then. 11. constant amplitude. γ i is the fraction that any particular portion of the stress range is of the total number of cycles. 11. 12 is written as 36 .63 588 000 374 162 N1 N2 Since the total effect ("damage") of the two different stress ranges is less than 1.35 + 0. stress range. that will display the same amount of damage as actually produced by the variable amplitude stress ranges. 10) in terms of an equivalent stress range. Each of N i and N e correspond to the number of cycles to failure—one for the variable amplitude stress ranges and the other for the equivalent. using the PalmgrenMiner statement of damage (Eq. i n1 n2 208 000 104 000 + = + = 0. 10. solving for the equivalent stress range ∆σ e m =∑ n i ∆σ im Σ ni (12) Calling n i Σ n i = γ i . 11 and N e = M∆σ e − m applies to the right hand side. then Eq.0. ∆σ e . Thus.∑ Ni n = 1 or. i. for which the terms were defined under Eq. Equation 8. which expressed the failure condition in a general way.
for example the rules provided by the American Railway Engineering Association (AREA) [24].33x2193 )] ) 13 = 199.5 MPa. the percentage of life expended is (312 000/494 953) 100% = 63. ∆σ i . solving for the equivalent stress range. using the expression given by Eq. It should be clear by inspection that Equation 14 is identical to Eq.∆σ e m = ∑ γ i ∆σ im Finally.5 MPa For this Detail Category B and the equivalent stress range of 199. we have ∆σ e = [∑ γ i ∆σ im ]1 m [∑ γ i S3Ri ]1/3 (13) In some specifications. and n 2 = 104 000 cycles ∆σ 2 = 219 MPa γ1 = 208 000 = 0. these are – n1 = 208 000 cycles ∆σ1 = 188 MPa N = 208 000 + 104 000 = 312 000 cycles. EXAMPLE 5 Use the equivalent stress method to determine the percentage of life that has been expended by the loading applied to the beam of Example 4. 8) as – N = M ∆σ −3 = 39.53 = 494 953 cycles Since the actual number of cycles is 312 000. given that m = 3.67x1883 ) + (0. the number of cycles to failure can be calculated (Eq.33 312 000 Now. 13. In summary.0%. Solution: All of the necessary data are available in the solutions to Examples 2 and 4. the equivalent stress range ( S Re ) is written as S Re = (14) where the symbol S Ri is used to indicate the stress range. 37 ( .67 312 000 γ2 = 104 000 = 0.3x1011 x 199. 14 to calculate the equivalent stress range – S re = S rMiner = [∑ γ i S3Ri ]1/3 = [(0.
) (a) According to Fig. The thickness of the flange of the girder is 25 mm and the coverplate is attached to the girder flange by fillet welds.4 16.28 12. 38 .3 4..3 MPa. and the one for which the stress history is given. The threshold stress for an E' detail is 17.7 24. Since the actual stress is greater than the minimum stress above which crack growth is possible.Comments: The same result was also seen in the solution to Example 4. The total truck traffic that is estimated to have taken place since the opening of the bridge is 35 x 106 cycles. the maximum stress range recorded is 37.1 1 37.e.0 1 33.6 Stress range. 20.14 8. Use the AASHTO Specification in the solution of this problem. 18 or obtained directly from Table 2. is that at the end of a partial length coverplate. Whether the solution proceeds by the method shown in Example 4 (Miner's summation) or by that shown in Example 5 (equivalent stress range method) is simply a matter of choice. MPa Figure 20 Example 6 b) Calculate how much of the fatigue life of this detail has been used up.8 1 29.9 MPa.63. where the Miner's summation was 0. It can be assumed that each passage of a truck causes one stress range at any given detail. This was obtained from a sample of 2064 truck crossing events. direct application of the Miner summation and use of a calculated equivalent stress range). The critical detail on this bridge. EXAMPLE 6 A stress range histogram has been obtained from field measurement of stresses on a certain simplespan girder bridge. Use each of the two possible approaches (i. which can be read from Fig. (This information is taken from Reference [25]). this indicates that fatigue cracking is possible in this girder. Solution: According to the AASHTO specification. this is a Detail Category E' (Consult the specification: this information is not contained within this document. a) Is fatigue cracking possible at this detail? 55 Stress events % 25 10 5 2 20.
n 1 = (55% ) 35x10 6 cycles = 19.9 321.75 0.1 0.55 0. Table 3 Example 6 ∆σ i average MPa 6.70 0.04 0.7 26.21) −3 = 534x10 6 cycles (M ≡ A from Table 2) .5 1.5 18.06 0.07 0.2 ni cycles x 10 19.25 = = 0.2 304.93 6 ni Ni 0.3 14.09 0.25x10 6 cycles ( ) N 1 = M ∆σ r −3 (Eq.(b) In order to calculate the fatigue life expended.35 0.21 MPa (Refer to the stresses given in Fig. this means that 60% of the fatigue life of this detail has been expended.0 35.9 6.35 0.08 0.6 297. Sample calculations for the first row are shown following the table.21 10.9 194.10 0.75 3.01 ∑ Ni n = 0.08 0.28) / 2 = 6.7 233.05 0.02 0. The results of the calculations necessary to determine the remaining fatigue life are given in Table 3. 8) = 128 x1011 ( 6.01 0.6 22.01 0.35 6 Ni cycles x 10 534 117 42. it is suggested that the average stress in each range be used.7 273.25 0. 39 .05 0.60.57 4.12 γi = n1 35x10 6 γ i ∆σ 3 i 131.30 2.9 436.04 N1 535 ( ) Since the Miner summation for all the events in this loading history is 0.25 8. n 1 19.60 i ∑ γ i ∆σ 3 = 2194 i Sample calculations for the first four items of Row 1 of the Table: ∆σ1 = ( 414 + 8.0 19.9 31. 20) .9 10.
3. determine the fatigue life expended by calculating an equivalent stress range. For example. This is the same as the value calculated using the Miner summation. The calculation proceeds on the basis that all cycles are damaging.2 Analysis of Stress Histories Figure 21 Example 6 Situations often arise where the applied loads create stress levels and stress counts (number of cycles at a given stress level) that are much more complicated than those given in Example 4. Note that in this example the equivalent stress range falls below the constant amplitude fatigue limit.0 MPa Next.0 MPa. Comments: Stress Range Detail Category E' Constant amplitude fatigue limit (17. ∆σ e = [∑ γ i ∆σ im ]1/ m ( Eq.Now.3x10 6 cycles ( ) Since the number of cycles that can be sustained by the Category E' detail is 58. and where we will substitute ∆σ e = 13. 13 has been generated in the last two columns of Table 3 and recall also that Σ n i = 35x10 6 cycles in this example. if the crane in Example 4 is carried by a continuous beam over 40 . There is a discussion of this issue in Section 4. The information necessary for substitution into Eq. that is. 13) = ( 2194)1 3 = 13.0/58. calculate the number of cycles that can be sustained by a Detail Category E' for an equivalent stress range of 13.0 MPa for the term ∆σ r : N = M ∆σ −3 r = 1. the fatigue life expended is 35. the extension of the sloping fatigue life line is used.60.3x10 6 cycles and the actual number of cycles is 35x10 6 cycles .3 = 0.3x10 6 Figure 21 displays the equivalent stress range calculation graphically.0 MPa 35x10 6 Number of Cycles 58. Using Eq. 8.0) −3 = 58. 4.28x1011 (13.9 MPa) ∆σ e = 13.
This results because loading adjacent spans causes stress cycles in addition to the cycle created when the crane passes directly over the location under examination. Furthermore. one trip is termed a loading event and the stress variation at a given point in the structure during such an event is called a stress history. the occurrence of a loading event brings about different stresses within different elements in a structure.several intermediate supports. in addition. the number of cycles per loading event may be related to the type and location of the element under consideration. Very complicated stress histories can be caused by loading events such as the passage of a truck over a bridge or a wave hitting an offshore oil platform. usually because of the cost. In any trace of stress versus time there is one absolute maximum live load stress and one absolute Figure 22 Typical Stress vs.) However. and the dead load stress. In fields other than civil engineering. Under a Moving Load The stress range between these two extremes is shown on the figure. the minimum live load stress. Time Variation minimum live load stress. and.3) that stress range is the dominant stressrelated feature of fatigue life determination. For this more general case. Shown are the maximum live load stress. Figure 22 shows the stress history at a certain location in a member as a moving load passes across a structure. there can be other "local" maximum and minimum live load stresses. civil engineers generally resort to stress analysis using load models and dynamic magnification factors that are provided in codes and standards. However. Consequently. it is rare that such measurements are carried out on civil engineering structures. Many different stress histories may need to be evaluated for a complete fatigue assessment of a complex structure. how is the element under examination for fatigue life influenced by these other stress excursions? 41 . elaborate stress measurements are often taken on actual structures subjected to service loading in order to overcome uncertainties associated with calculated values of stresses in elements. (It is noteworthy that the stress range is independent of the dead load stress. of course. and it has already been indicated (Section 3. The question is. more than one stress cycle is applied per trip at a given location.
and the meancrossing peak count method. Generally. However. The plot shown in Fig. the peak count. The results obtained by these two methods are the same. as the "minor" stress ranges increase in value. the rainflow method. The reservoir and rainflow methods are the ones most commonly used in civil engineering applications [26. Methods that are used for counting stress ranges include the reservoir method.One method of interpreting the type of stress history shown in Fig. 22 is to simply use the maximum live load stress range shown in the figure. whereas the reservoir method is most convenient for graphical analyses of short histories. 23 (a) is an example of stress variation in an element subjected to a loading event. whereas actual stress measurements usually reveal more Stress 1 3 5 15 19 13 11 7 9 8 14 18 10 12 16 17 21 1 20 2 stresses due to one loading event 22 4 6 22 (b) peaks and troughs numbered (a) An example of stress variation in an element due to one loading event for one loading event 2 4 1 3 12 20 22 21 1 (d) reservoir analysis (c) rainflow analysis Figure 23 Analysis of Stress Histories 42 . rainflow counting is more suited to computer analyses of long stress histories. Typically. simple analyses result in stress histories having fewer peaks than shown in this figure. 27]. their influence becomes important at some point.
The rainflow method is named for its analogy of rain drops flowing down a pagoda roof. between a high water level and the drain (trough) that lowers it is counted as one full cycle in the stress spectrum. The horizontal movement of a raindrop. Eq. The reservoir method. a drop flows left from point 1 off points 2. 10. measured in units of stress from its originating peak to its stop position is counted as one half of a cycle in the stress spectrum. 4 and 12 until it stops at the end of the loading event at point 22 since no peak is encountered that is more positive than point 1. is shown in Fig. For example. The following rules apply to rainflow counting: 1. without application of a stress counting method in order to tabulate values for the number of cycles. point 22 in Fig. for different stress range levels. sonamed because of its analogy of water contained in reservoirs formed by peaks and drained successively out of troughs. 23. On the other hand. 3. The path of a drop cannot cross the path of a drop that has fallen from above. A drop flows left from the upper side of a peak or right from the upper side of a trough and onto subsequent "roofs" unless the surface receiving the drop is formed by a peak that is more positive for left flow or a trough that is more negative for right flow. loading event. The information shown in Fig. a drop flowing left from point 3 stops at the horizontal position of point 2 because a path coming from point 2 is encountered. this is point 1. the most widely used are the socalled rainflow counting method and the reservoir methods. 23 (b) and this history is rotated 90 degrees in Fig. For example. 23 (c) in order to perform the rainflow analysis. Drain water successively out of the lowest troughs in the loading event until all reservoirs have been drained. drain water from the reservoir out of the lowest trough in the spectrum.complex stress histories. 43 . imagine that the area bounded by the highest peak in the loading event forms a reservoir of water contained by it and the same peak in a following. 23 (d). ni. 23 (a) cannot be used directly in the PalmgrenMiner rule. Next. 23 (b). As already indicated. Using the numbering shown in part (b) of Fig. Water caught between other peaks forms smaller reservoirs. The vertical distance. measured in stress units. identical. To start. 2. a drop flows right from point 2 off point 3 and stops since it encounters a surface formed by a trough (point 4) that is more negative than point 2. Peaks and troughs for one loading event are numbered in Fig.
23. The given data are – Peak / Trough No. EXAMPLE 7 The values of the peaks for the stress history shown in Fig. Finally. Apply (a) rainflow counting and (b) reservoir counting in order to identify the stress ranges in the stress spectrum. (c) evaluate the effects of one million loading events of this stress history acting on a beam that uses a continuous partial penetration groove weld to connect the flanges to the web. as is shown in Fig. 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 Stress (MPa) 93 18 55 10 85 10 37 18 37 10 46 6 55 46 74 8 55 18 65 39 83 0 44 .The rainflow and reservoir methods give identical results provided that rainflow counting begins with the highest peak in the loading event. 23 are given below.
Water Level at Peak Stress Range (MPa) 22 12 4 16 2 18 10 6 20 8 14 1 21 5 15 3 17 11 7 19 9 13 93 77 75 66 37 37 36 27 26 19 9 Note that these stress ranges are the same as were determined in part a) using the rainflow counting method (where halfcycles are reported). 45 . 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 To Horizontal Distance of Point No.Solution: (a) Rainflow counting: From Peak or Trough No. 22 3 2 5 6 11 10 9 8 9 10 21 14 13 16 15 18 17 20 19 12 1 Half Cycle (MPa) 93 37 37 75 75 36 27 19 19 27 36 77 9 9 66 66 37 37 26 26 77 93 Solution: (b) Reservoir counting method: Drain from Trough No.
008 0.55.144 0.402 0.0x1011 . Table 2 indicates that the fatigue life constant for use in Eq. 0 This analysis indicates that the fatigue evaluation has failed.) Stress Range ∆σ (MPa) 93 77 75 66 37 (twice) 36 27 25 19 9 Fatigue Resistance N = 20. At this point. (The reader should review Example 8 relative to this part of the problem after Section 4. If this change could be made.211 0.003 0. In order of importance these are: 1.Solution: (c) Cumulative damage using the PalmgrenMiner rule: According to Table 1 this is an AASHTO Detail Category B'. and the detail would then pass the fatigue evaluation. A better design detail should be investigated.7 x 109 Damage summation: Comments: ∑ ni N = 1. Often. the use of continuous fillet welding rather than continuous partial penetration groove welding would increase the fatigue strength of this element by about 25% according to the AASHTO rules since the fatigue category changes from Detail B to Detail B'.010 0. meaning that the element may not withstand one million loading events. 8 is 20. the designer has several remedial possibilities.3 has been studied.051 0. 46 . a substantial reduction. ni/N 0.228 0. In this example. in this example the damage summation would be 0.000 2 486 000 4 381 000 4 740 000 6 957 000 39 484 000 42 867 000 101 610 000 128 000 000 291 588 000 2.025 0. 08 ≥ 1.0x10 11 (∆σ ) −3 Damage Due to 1x106 Loading Events. fatigue problems can be traced to poor initial design detail choices.
In laboratory testing. Note that. 4. and improving threedimensional loadcarrying capacity. On the other hand. 4. A more detailed fatigue analysis could be carried out. avoiding short members near application of loads. Case 1 refers to the situation where all cycles are above the constant amplitude fatigue limit (CAFL). a machine part that is loaded in the same way for all of its operating cycles corresponds to the constant amplitude loading case. constant amplitude loading is the procedure most often used. For example. Only as a nearlylast resort. civil engineering applications such as bridges and offshore structures would hardly ever be loaded in constant amplitude fatigue. 3. the same distribution of stress ranges is located such that there are stress ranges both above and below the CAFL. For Case 2. 24 a given stress range spectrum (distribution of stress ranges at a particular structural element) is shown for three possible cases. see Section 8. in Section 4. The 47 . In Case 3. This is covered next. 13 or 14) is a way of expressing the variable amplitude stress ranges as an equivalent constant amplitude stress range. Such measures could involve employing continuous construction. reducing the number of expansion joints.2). all stress ranges in the spectrum are below the CAFL. as we have already seen (Section 4. Stress Range σ max σ min CASE 1 Constant amplitude fatigue limit (CAFL) σ max σ max CASE 2 CASE 3 Number of Cycles Figure 24 Three Cases of Variable Amplitude Stress Spectra The constant amplitude fatigue limit is simply that stress range level below which there will be no fatigue crack growth under the condition of constant amplitude loading.3 Fatigue Limits In Fig. Measures that reduce stresses due to dynamic loading should be studied.3. if this is done.4. the thickness of the member could be increased in order to lower the stresses. improving road surfaces on road bridges. Some real fatigue life cases will correspond to this.2. primarily because it is the easiest situation to create. Rather. the fatigue strength may also decrease. We have seen that the use of an effective stress range (Eq. the stress ranges that actually act on a structural detail correspond to a case of variable amplitude fatigue.
This is equivalent to assuming that the value for the fatigue threshold stress intensity factor is zero. In effect. that all cycles contribute to crack propagation in the element from the first cycle of fatigue loading. offshore platforms. For example. In other words. if a stress range in a spectrum that resulted from another loading event does exceed the CAFL.question remains: does constant amplitude loading fairly represent variable amplitude loading? Studies done on the fatigue life of coverplated beams (AASHTO Detail Category E') [28] showed that constant amplitude tests and variable amplitude tests (done at the equivalent constant amplitude stress range) gave comparable results up to at least one million test cycles. First. it appears that the maximum stress per se must also be considered at these long lives. analysis proceeds in a fashion identical to the analysis required for Case 1. all of the stress cycles in the spectrum contribute to fatigue crack growth and must be considered in any fatigue life evaluation. chimneys. As crack propagation proceeds. 6. This is how the discussion and examples have been treated so far. structures such as bridges. However. This assumption is equivalent to admitting the presence of a nonzero value for the fatigue threshold stress intensity factor. It is difficult to guarantee that values of all stress ranges will remain less than the value for the CAFL throughout the service life of the structure. The analyses in Examples 5. this option assumes that the fatigue strength curve continues below the CAFL with a slope constant of 3. Case 3 is also clearcut: because none of the stress cycles in the spectrum are above the constant amplitude fatigue limit. 24. For Case 1 shown in Fig.) Case 2 spectra are most often associated with situations where long fatigue lives are required. There are fewer data (from any sources) for stress ranges in the order of ten million cycles upward. (However. no fatigue assessment is required. and 7 adopted this option. conservatively. The second option for analysis of Case 2 spectra is based on the more logical assumption that not all cycles in the spectrum will contribute immediately to crack propagation. then Case 3 spectra should be considered in the same way as Case 2 spectra. and overhead highway signs are often subjected to millions of loading events. Two options are possible for the analysis of Case 2 spectra. all stress ranges in the spectrum are assumed to contribute to fatigue crack growth. For this option. it can be assumed. more and more 48 .
and implicit in the solution was that all cycles 1 The fatigue life rules in ECCS and in the more recent Eurocode [46] are nearly identical. e. Small cycles in the spectrum do not contribute significantly to fatigue damage: it is considered that by the time cracks grow to a length where these stresses would induce stress intensity factor ranges greater than the threshold stress intensity factor range. This equipment is capable of measuring several thousand small cycles per loading event.1 However. They are similar in many major respects to those in the AASHTO Specification. test data for all other cases investigated indicate that use of the slope constant of 5 is acceptable. fracture mechanics analyses are hindered by difficulties in estimating values for the threshold stress intensity factor under conditions of high tensile residual stress that vary with crack length. The importance of the cutoff limit is greatest when automatic stress measuring equipment is used to measure stress ranges. This is due to a lack of test data at sufficiently long fatigue lives so as to be able to verify proposed analysis methods. The PalmgrenMiner rule was used. It remains to be seen whether the best rule is the first option or the second option qualified by a provision for exceptional cases. Nevertheless. the PalmgrenMiner damage summation can produce an excessively conservative evaluation. EXAMPLE 8 Part (c) of Example 7 required calculation of the cumulative damage for a given stress spectrum. a significant difference is that the number of cycles at which the CAFL is taken to occur is a constant (5 million cycles) for all detail categories in the European standards. the vast majority of fatigue life has been expended. the ECCS rules [15]. More research and testing is needed. 49 . European practice. the cutoff limit. Furthermore. The AASHTO approach is more consistent with the experimental evidence.. When the number of small stress cycles (having values less than half of the CAFL) is greater than two orders of magnitude more than the number of large cycles. of 5 for stress ranges below the CAFL. International opinion as to how the analysis should proceed in the Case 2 second option is not unanimous. allow the use of a slope constant. The Eurocode rules suggest a cutoff limit at the stress level that corresponds to the intersection of the fatigue strength curve (m = 5) with one hundred million cycles. The second limit discussed in this Section. m. whereas in the AASHTO rules the number of cycles at which the CAFL starts varies with the detail category. recent work indicates that this may be unsafe for stress spectra applied to large beams with welded longitudinal attachments [29]. However. is used to discard small stress cycles from the stress spectrum.cycles in the spectrum will exceed the threshold stress intensity factor range until all cycles in the spectrum contribute to crack propagation at a certain crack length.g.
this is equivalent to assuming that the sloping straight line of the Detail B' category extends downward indefinitely at a slope constant m = 3 on the log–log plot. 2. the stress spectrum is not Case 1.3. and the minimum stress in the spectrum is 0 ksi. or 3 according to the terminology introduced in this Section.7 MPa. Comments: The calculations carried out in Part (c) of Example 7 counted all stress cycles as damaging. 50 . (The term threshold stress is synonymous with CAFL. As pointed out earlier in Section 4. Establish whether the situation is Case 1. that is.of the spectrum produce fatigue crack growth. the range of stresses within the spectrum straddles the CAFL. Since the maximum stress (93 MPa) is greater than the CAFL (82. Table 2 indicates that the threshold stress for Category B' is 82. 24).) According to the information given in Example 7. this means that the stress spectrum does not correspond to Case 3 (see Fig. the maximum stress in the spectrum is 93 MPa. Since the minimum stress (zero) is less than the CAFL. the stress spectrum is identified as Case 2. Thus. Solution: The detail given in Example 7 is an AASHTO Detail Category B'.7 MPa).
it is still instructive to examine the issue. this is the subject of Chapter 6. and therefore they should be designed more conservatively.") Even though the differentiation between nonredundant and redundant members is no longer part of loadinduced fatigue design. but..Chapter 5 Fatigue Design According to the American Association of State Highway and Transportation Officials Specification (AASHTO) 5. The rationale for this was that the consequences of failure were greater in the former. Whenever possible. greater fracture toughness was also specified for nonredundant members. however. and if economical. The relevant term used in that part of the specification is "fracturecritical. 5. In those earlier specifications. the rules of the American Association of State Highway and Transportation Officials (AASHTO) Specification [2] that reflect loadinduced fatigue are examined.2 Redundancy and Toughness Users of previous editions of the AASHTO Specification (e. This was a reasonable position. 15 Edition." th 51 .1 Introduction The discussion in the first four chapters of this document has centered around concepts of fatigue crack growth and how to treat design tasks that are linked to loadinduced crack growth.g. as a matter of engineering judgment. at the same time. so that the instructive points made could be developed in the mind of the reader. 1992) will be aware that the concept of structural redundancy has been dropped in the fatigue design rules of the 1994 Edition. The examples thus far have deliberately been kept simple. Chapter 6 will contain observations arising out of the AASHTO Specification that relate to distortioninduced fatigue. and therefore the distinction of fatigue design with respect to nonredundant members and redundant members has been dropped from the current (1994) edition of the Specification. The stress range values for nonredundant members were fixed. permissible stress ranges were set forth for both redundant and nonredundant members. This constituted a double penalty. (The distinction for fracture toughness has been maintained. at 80% of the values for redundant members. Distortioninduced fatigue crack growth is also an important consideration. the designer should use details and concepts that introduce "redundancy. In this chapter.
however. This implies that there are at least three parallel loadcarrying components such as girder elements. For example. 52 . There are at least three types of redundancy: only the first one listed is accepted by some owners. Although this seems a reasonable position. the tension region of a beam in bending. The concept that multiple unwelded components of a beam. Paul. the I79 bridge across the Ohio River near Pittsburgh carried traffic even though a crack was present that extended all the way up to the top flange [20]. it is contrary to experience with actual structures.. Similarly. traffic was carried without incident in this condition for at least a few hours.The 1992 AASHTO definition of a nonredundant member was that it was a member whose tension failure would result in collapse of part. generally it is not until a fatigue crack starts in second component that collapse becomes a concern.e. usually as identified in the principal crosssection. the twogirder Lafayette Street Bridge in St. The assumption is simply that such a bridge would fail catastrophically if one girder fractured. Internal member (crosssection) redundancy. For example.) In both of these structures. tension failure could be the failure of a tension member. However. The severing of one component by a fatigue crack may not lead to catastrophic failure since the other components are still intact. but it was intended that the definition include the concept of the tension failure of a portion of the crosssection. truss. 2. (It is not known how long the situation existed. or arch member 1 Obviously. e. Multiple load paths. between the time the crack was observed by the crew on a passing tugboat and when traffic on the bridge was stopped. 1. Beams or truss members in which the crosssectional components are fastened with rivets or bolts are considered to be redundant internally. 3. a statically indeterminate structure).g. their lateral bracing systems participated in carrying the load because the cracked plate girder behaved like a torsionally stiff box. A continuous girder bridge is statically redundant. Minnesota carried traffic without difficulty for several months even though a girder had fractured near the center span inflection point [20]. instability due to the nonsymmetry of the cracked crosssection might be a consideration in some cases. however. Of course. or all. of a structure. Statical redundancy (i..1 The statement was interpreted in a variety of ways. However. continuous bridges that consist of just two parallel girders would generally be regarded by designers as nonredundant.
The redundancy in the case of statical indeterminacy is identified as a routine part of the structural analysis. Use of a steel that has a good level of fracture toughness and avoidance of details that create triaxial stress conditions are important to achieving the desired fatigue life and preventing premature fracture. Fracture toughness is also dependent upon temperature. This requires design skill and good details.. fracture toughness is also a matter of concern in bridge structures. Good practice sometimes goes beyond the minimum requirements. even though a perceived detailing improvement may not produce an advantage in terms of the specification. can sometimes be evaluated as Category C provided that the fasteners are tight and in good condition [30]. It is obvious that a statically indeterminate structure will be less likely to suffer catastrophic collapse in the event of significant cracking due to fatigue. the design specification does not yet make this distinction. which means that the loading rate from traffic is not as severe as the dynamic rate associated with the standard Charpy Vnotch impact test commonly used to provide qualitative information about the steel material. However. the degree to which this vulnerability might exist can only be established with the aid of an extensive. Sometimes.builds in crack arresters and therefore improves performance has been documented. Bridge steels are strain rate sensitive. For example. In addition. tough steels promote stable crack extension. owners are receptive to demonstration of redundancy by this means. The objective is usually to show that a more economical structure is also redundant. Low fracture toughness bridge steels can result in brittle fracture under normal service loads when cracks or cracklike geometric conditions develop. which is a type of catastrophic failure. a highly undesirable situation. Because of the possibility of fatigue crack formation. The goal is to avoid brittle fracture. threedimensional computer analysis. good design practice incorporates as much redundancy as can be justified economically. It is possible that fatigue crack growth can lead to conditions that will trigger brittle fracture. fracture toughness also decreases. Although a redundant system receives no benefit in the AASHTO Specification as compared with a nonredundant system. However.g. As the temperature decreases. that a threegirder redesign alternative is comparable to a fourgirder design. recent research has shown that the multicomponent riveted members. Although fracture 53 . which have a component fatigue category of D. e. which is a desirable feature for fatigue crack inspection.
the Charpy Vnotch tests can provide a useful means of screening out materials that would be susceptible to brittle fracture at small crack sizes. This is primarily because the model characterizing combinations of stress and crack size results in plasticity levels such that the elastic stress model becomes invalid. These are the stress range that is acting on the detail under examination.6.2 of the Specification. One is that the axle spacing between the two 145 kN axles is fixed at 9. it has been established that the fatigue life of a structural steel component can be described in terms of just three elements.2. the number of cycles to which the detail is subjected. In this section. there are two unique features that are applied to the Design Truck.0 m. These. In order. where the elements of the design requirements were outlined so that examples could be carried out. 5. thereby preventing achievement of adequate fatigue resistance.1 Fatigue Load and Frequency The basis of the load to be used for the fatigue design of bridges according to the AASHTO Specification is the socalled Design Truck. The stress range at the particular detail under examination for fatigue strength is to be 54 . more information concerning the fatigue provisions in the AASHTO rules will be provided. 5.3. an accurate determination of fracture susceptibility for structural engineering design is complicated. For fatigue.4. are in Article 3.3 Fatigue Design in the AASHTO Specification In the material presented so far. and a description of the fatigue life inherent in the detail.1.0 m.3 and 9. and other related details. The Design Truck consists of three axles. The Design Truck is a loading that applies to many different design cases. This reflects the observation that the large number of truck passages associated with the fatigue strength examination are commensurate with lighter trucks than the Design Truck.3 m and the spacing between the two 145 kN axles is a variable quantity that can range between 4. The axle spacing between the 35 kN load and the first 145 kN load is 4.75 times the axle loads of the Design Truck. Nevertheless. these are one axle of 35 kN load followed by two of 145 kN load each. The AASHTO rules were first introduced in Section 3.mechanics provides a useful analysis tool to relate crack size and stress level to the material fracture toughness characteristics. The other feature is that the axle loads to be used for calculation of the fatigue stress ranges are to be taken as 0.
(In this document so far. is physically limited to about 20 000 vehicles per lane per day.75 is to be used for the fatigue case. the dynamic load allowance is designated simply as a 15% increase in the stress range calculated using the Design Truck. which includes all cars and trucks. averaged over the design life p = a factor reflecting the number of lanes available to trucks. these values can be used to generate the frequency of loading of the Design Truck. averaged over the design life ADTTSL = the number of trucks per day in a single lane. this has been designated ∆σ r ) ( ∆F) n = the nominal fatigue resistance γ = the load factor for the fatigue load combination. this quantity is to be calculated as: ADTTSL = p x ADTT (16) where ADTT = the number of trucks per day in one direction.6. the usual situation. the loading is to be taken as the singlelane average daily truck traffic. the designer is advised that the average daily traffic (ADT). 15% for urban interstate or "other" rural highways.85. If only one lane is available to trucks. if three or more lanes are available. If sitespecific information is not available. the nominal fatigue resistance is linked to the number of stress ranges that will be imposed. In general.80. modified as described above. As we have already seen. as discussed above. For consideration of fatigue. p = 1. The value γ = 0. and 10% for "other" urban conditions. The fraction of truck traffic in this total is estimated to be about 20% for rural interstate highways. if two lanes are available then p = 0. For instance. each detail must satisfy the requirement that γ ( ∆f ) ≤ ( ∆F) n (15) where ∆f = the live load stress range due to passage of the fatigue load. ADTTSL .evaluated according to the loading described. The Specification provides other information regarding the loading. then p = 0.2 of the Specification. According to the AASHTO Specification. 55 . including also the dynamic load allowance stipulated in Article 3. Unless better information is available.
17. including impact.2 Fatigue Resistance Each detail must satisfy the requirement that the calculated stress range.3. this is as a result of the vibrations set up in the structure as the truck moves across. Primarily. Equation 18 is the number of cycles of stress range expected to be imposed on the structure in a design life of 75 years. As just discussed. 18 as a way of acknowledging that more than one stress cycle can result from a single passage of the fatigue Design Truck. for example. 17 is identical to Eq. which is equivalent to the term M in Eq. The subscript n in the expression for nominal resistance simply identifies that there must be a linkage with number of cyclesthe fatigue resistance is a function of the number of cycles of the stress range. The AASHTO Specification rules provide that. 17.2. ( ∆F) n .0 is to be . 17. (This high number reflects the susceptibility of cantilevers to vibration under moving loads. 8) and N = ( 365) ( 75) n ( ADTT) SL (18) The constant A in Eq. Equation 8 is the fatigue life expression that was developed from a fracture mechanics approach. 8. the following 56 . For cantilever girders. For a simple span girder. used for spans greater than 12 m. the AASHTO requirement for fatigue life (Eq.) The term n is included in Eq. (At the option of the designer. in addition to Eq.5. it can be seen that Eq. n = 5.5–2 of the Specification. another number can be used in place of 75. reflects the severity of the detail.6. Given that the term ( ∆F) n is simply the value of the term ∆σ r at its permissible value for the given number of cycles.1. termed ∆f in the Specification. Thus.0 must be used. The nominal fatigue resistance is to be taken as: A ( ∆F) n = N 1/ 3 (17) where A is the constant listed in Table 2 (≡ M in Eq.) The AASHTO Table can be consulted for other cases. the calculated stress range is that arising from application of the fatigue Design Truck. according to the Detail Categories identified in Fig. be equal to or less than the nominal resistance identified in the Specification. Values of n are tabulated for 10 different cases in Table 6. 8. n = 10 is to be used for spans less than 12 m and n = 2. 17) is consistent with the fatigue life of a given detail as described on a theoretical basis and incorporating the results of physical testing by the inclusion of the term A in Eq. 18 and Table 2.
nominal stress ranges. fatigue crack growth will take place. Also shown Equation 17 log stress range ∆σ max = 2 ∆σ e Constant amplitude fatigue limit (CAFL) CAFL / 2 ∆σ e Stress range spectrum log number of cycles Figure 25 AASHTO Fatigue Limits 57 . ∆σ e . 17 need not extend below a horizontal straight line at the value of onehalf the constant amplitude fatigue limit. the latter says that the sloping straight line represented by Eq. and then by the dashed horizontal straight line.requirement must also be met— ( ∆F) n ≥ ( ∆F) TH 1 2 (19) where ( ∆F) TH is the constant amplitude fatigue threshold stress. the requirement reflects the fact that the stress ranges calculated according to the AASHTO rules are a simplified depiction of the actual stress ranges that can take place. 19 determines whether or not the maximum stress range in the random variable cycle exceeds the CAFL. This is reflected in Eq. followed by the dashed sloping straight line. 25. The check using Eq. (This has been referred to elsewhere in this document as the constant amplitude fatigue limit. According to Figure 18. However. The AASHTO permissible fatigue life line is the sloping solid straight line shown in Fig. is twice as great as the nominal stress range calculated from the fatigue truck.) Viewing Equations 17 and 19 together. it would appear that the constant amplitude fatigue limit itself should be used. ∆σ max . there can be occasional stress ranges that are considerably greater than the calculated. which represents the effective stress range. It is assumed that the maximum stress range. Figure 25 illustrates these points concerning the AASHTO rules. 19. If even a small number of these stress ranges exceed the CAFL. In particular. CAFL.
6. Determine the permissible stress range for the detail under examination and its In practice.e. 58 . if the check of Eq. fatigue shall be considered only if this compressive stress is less than twice the maximum tensile live load stress resulting from the fatigue load…" 5.1) that the force effect for the fatigue design of a steel bridge shall be the live load stress range. Calculate the stress range at the details of interest. which in this case intercepts Eq. i. ∆σ e . it is recognized that fatigue cracks will not grow if the stress range is entirely compressive. especially when fitup stresses. or load survey information. Thus. 17 below the CAFL.2. Shown is the equivalent stress range for the spectrum. it is often difficult to ensure that stresses will always remain in compression. The AASHTO rules point out (Article 6. a further provision in Article 6. and stresses due to foundation settlement may be present.1 that allows the designer to reduce the areas of a bridge requiring examination for fatigue is: "In regions where the unfactored permanent loads produce compression.in the figure is a hypothetical random variable stress spectrum. Identify the Detail Category for each component of the structure that could be sensitive to fatigue. stresses due to temperature changes. If it is visualized that the stress spectrum is moved down to a level such that ∆σ e is just below the level CAFL/2. plotted on the ordinate.6. the statement in Article 6. fatigue crack growth will occur. Thus.2.6.1 allows the designer to identify those regions where only a compressive stress range is expected and to therefore eliminate them from further consideration. Use the AASHTO recommendations.2. The spectrum shown has been given the unique relationship wherein ∆σ max = 2 ∆σ e . an observation supported by both field and laboratory experience2. Accordingly. We note.1. or information provided by the owner. that the maximum stress range in the spectrum is above the CAFL. then all cycles are below the fatigue limit and no fatigue cracking will occur. In accordance with the discussion just concluded. however. then it is apparent that the maximum stress in the spectrum is just below the CAFL. However. This is consistent with the discussion so far in this document. 19 indicates that ∆σ e is below the value CAFL/2. it is again necessary to assume that the live load stress range can be as high as twice that calculated using the AASHTO fatigue Design Truck.4 Summary of AASHTO Requirements • • • • 2 Determine the number of design cycles.1..1.
Concrete over Field Section 2 never acts compositely with steel girder (i. 203 mm thick. four parallel girders Rural interstate highway. Design Assumptions: For purposes of this example. two lanes of traffic in same direction Unshored composite construction (positive moment region only) ' Concrete slab: f c = 20. This will enable the reader to use that source for more information..number of stress range cycles. reference will be made from time to time to the specific articles of the Specification. The problem is derived from Design Example 2 of Reference [31]. 26) always acts compositely with steel girder. 59 . modify the design and repeat the assessment. no shear studs). Concrete over Field Section 1 (see Fig. Of course.e. ASTM A588 steel.) • If the actual stress range at the detail exceeds the permissible value. plus a 12 mm thick integral wearing surface Welded plate girder.5 Design Example The following design example will be used to elaborate the fatigue requirements of the AASHTO Specification. Because the illustration that follows includes calculation of the stresses and other design aspects that are not a central part of this document. both of these issues would be a normal part of the design procedure. EXAMPLE 9 Design Information: Twospan continuous girder bridge. modular ratio E s E c = 10 Reinforced concrete slab. girder spacing at 2440 mm c/c. 42. assume that the given section (Fig. the 1977 AASHTO Specification rules were used and the details were laid out in US Customary units.7 MPa . 5. 26 and Table 4) is satisfactory for noncyclic loading and that all steel meets the fracture toughness requirements. In that example.7 m spans. (Engineering judgment should be exercised: some details can be eliminated from examination on the basis of comparison with others of similar stress range and number of cycles. Fy = 345 MPa.
For these calculations. the average spacing of adjacent beams = 2 440 mm Governs 60 .25 x 30 000 = 7 470 mm or. Sufficient information is given such that interested readers can confirm the calculations if desired. Solution: CrossSection Properties.2.5x304.Problem: Investigate the suitability of an interior girder with respect to fatigue.6. for an interior girder.5 x top flange width = (12x203.6.8) = 2 590 mm or. the effective slab width is the least of — 0.1 of the AASHTO Specification. Table 4 lists the dimensions of the steel plate girder and the section properties calculated from those dimensions. where.2)+(0. These section properties are used subsequently to calculate stress ranges. Table 5 lists the section properties of the composite crosssection.25 x effective span length (to point of permanent load inflection) = 0. In a similar fashion. 12 x average thickness of slab + greater of web thickness or 0. This is obtained using Article 4. it is necessary to identify the effective flange width of the composite crosssection.
45 40 35 S tre s s 3 0 ra n g e 2 5 M P a 20 15 10 5 0 0 T o p F la n g e S tre s s R a n g e 10 20 30 40 D is ta n c e a lo n g g ird e r.1 3 7 .8 Figure 26 Plate Girder Details and Stress Range Diagrams 61 . m N o fa tig u e c h e c k re q 'd b o tto m fla n g e N o fa tig u e c h e c k re q 'd to p fla n g e C e n te r L in e N O T E : A ll p la te siz es g iv e n in T a b le 4 S ta tio n → 0 F ie ld S e c tio n 1 3 0 .9 4 2 . m 45 40 35 30 S tre s s 25 ra n g e 20 M Pa 15 10 5 0 B o tto m F la n g e S tre s s R a n g e 0 10 20 30 40 D is ta n c e a lo n g g ird e r.7 F ie ld S e c tio n 2 7 .7 3 0 .8 4 .1 4 2 .
2 23.75 19 355 CENTER refers to the 457.0 25.0 5 619 5 619 14 515 3 277 0.11 16 932 Notes: 1524 11.0 Top Flange: Ay (mm 3 ) Bottom Flange: Ay (mm 3 ) ΣAy (mm 3 ) y (mm) Moment of Inertia (xx) Web: wh3/12 (x106mm4) Web: A w ( y ) 2 (x10 mm ) 6 4 3 277 350.40 9 677 609.6 31.88 4 839 381.0 55 641 15 055 615 15 055 615 0 0.75 19 355 portion of Field Section 2 that is over the pier portion of Field Section 2 that is away from the pier Bottom Flange b (mm) t (mm) A f (mm 2 ) Neutral Axis (xx) A total ( mm 2 ) (from web centerline.40 9 677 609.0 11 238 11 238 25 754 (web. about girder centroid) (bottom flange.81 10 887 381.11 16 932 CENTER 1524 11.TABLE 4 –Properties of Steel Section Plate Girder Only END Web h (mm) w (mm) A w mm 2 MIDDLE 1524 11.11 END refers to Field Section 1 ( ) 16 932 Top Flange b (mm) t (mm) A f ( mm ) 2 MIDDLE refers to the 304. about own axis) (web.8 15.6 31.7 3 971 4 159 11 758 3 277 0.0 25. about girder centroid) (top flange. about girder centroid) TF: A f ( h / 2 − y ) 2 (x106mm4) BF: 6 4 A f ( h / 2 + y ) 2 (x10 mm ) Σ I x (x106mm4) 62 .9 36 286 747 082 7 497 082 0 0. +ve up) 32 657 3 725 497 8 425 575 4 700 079 143.
+ve up) 2 A total (mm2) Concrete: Ay (mm3) Steel Girder: Ay (mm3) ΣAy (mm3) y (mm) or y (mm) or y (mm) Composite Moment of Inertia (xx) Girder: (x106mm4) Girder: A girder ( y)2 11 758 12 424 170.5 8 190 32 543 (x106mm4) Concrete: (x106mm4) Concrete.2 2 438 495 402 49 540 Modular ratio n (= E c / E s ) Concrete Thickness (mm) Effective Concrete Width (mm) Effective Concrete Area ( mm ) Equivalent Area of Steel 2 ( mm ) Composite Neutral Axis (xx) (measured from neutral axis of steel girder.6 (from neutral axis of steel girder alone. A equiv ( y )2 (x106mm4) ΣI x (x106mm4) 63 . +ve up) 82 198 50 699 188 0 50 699 188 616. +ve up) (from web centerline. +ve up) (from centroid of concrete slab.8 472.9 406.TABLE 5 – Properties of Composite Section Composite Section Composite Section Note: Composite action develops in End Section (Field Section 1) only 10 203.
According to AASHTO Table 6.3 m 9.6x10 6 .075x10 6 n cycles / lane = 93.5–2.075x10 6 x 1.0 m The load distribution factor (i. Load Assumptions and Calculations The Fatigue Design Truck is the Design Truck (as shown in the sketch).3 m on each side of the interior support can be categorized as "interior.75 (AASHTO Table 3.1–1). the Table . The average daily truck traffic (ADTT) will be taken as 20% (rural interstate highway) of the physical limit of 20 000 vehicles per lane per day.1. the number of stress cycles is then— N = ( 365) ( 75) n ( ADTT) SL . 35 kN 145 kN 145 kN 4." ) Finally. or n = 10 otherwise.6.2–1 of the AASHTO Specification).3 m of the interior support.1. for a continuous span bridge of 42.075x10 6 n cycles / lane over an assumed 75year life.1." However. 16 (which is Eq.6. In this example.e. This is obtained from Article 3.. Thus. details within 4.5–2).4. In this expression. say.4. ADTTSL = p x ADTT = 0.85 (AASHTO Table 3. and applying an impact allowance of 0.075x10 6 n cycles / lane.40. the number of trucks per day in a single lane is ADTTSL = p x ADTT where p = 0.85 x ( 0. it can be concluded that for locations within 4.1.2.1. say. multiplied by the load factor of 0.7 m the value is n = 15 for details located within L / 10 = 4. 18 (AASHTO 6. 93x10 6 cycles / lane .2–1) since in this bridge two lanes are available to trucks. 3.20 x 20 000) = 3 400 trucks / lane / day According to Eq.3 m .5.4.4.3b and from 64 . N = 93. simply differentiates as "near interior support" or "elsewhere. and since the two spans are of equal length. Commentary to the Table suggests that L/10 can be used to define locations near the interior support.5 = 139. and for all other locations N = 93. the fraction of a truck in a given lane that accrues to an interior girder) will be taken as 0. .Number of Load Cycles According to Eq.2. or N = (365) ( 75) n (3 400 ) = 93. the term n reflects the number of stress cycles for a given truck passage.15 (AASHTO Article 3.6. 140x10 6 cycles / lane .2.6.6. (In fact.1).
6.1 m for the top flange and between stations 33.2b. say 0.7 m).398.2b.Article 4.0 kN/m for Field Section 2. In this example.) 65 . in a socalled exact analysis. applied to the composite section. K g . (Space does not permit tabulation of the stress ranges at discrete locations.6. the dead load of the steel section is estimated to be 2. Furthermore. but this would be available to the designer.408.40. depending upon which detail is being examined. In the calculation of the load distribution factor.2. is 12. (The centerline of the twospan structure is bridge station 42. in this problem these results are sufficiently close to one another that either could be used. applied to the girder alone.2. The initial dead load (unfactored) due to the concrete slab and girder selfweight. Obviously. of course. Figure 26 shows the live load tensile stress range for the top and bottom flanges of the steel section. Considering the level of confidence likely in the load distribution factor calculation. the load distribution factor is 0.3. obtained using the section properties of Field Section 1.2.84 kN/m (unfactored). It is advantageous for the designer to take advantage of this provision in order that the bridge length over which fatigue needs to be examined is reduced. The regions that do not have to be examined for fatigue are shown as broken lines in this figure. these refinements seem unwarranted.3 m and 42.5 kN/m for Field Section 1 and 3. In accordance with AASHTO Article 6.7 m for the bottom flange.6. the longitudinal stiffness of the slab–girder system is involved.2) that fatigue need be considered only when the dead load stresses (unfactored load) are compressive and are less than twice the live load tensile stresses. These values are used in conjunction with the fatigue live load to calculate the stresses in the top and the bottom flanges of the steel section. Using this value. including wearing surface. This information can now be used to evaluate the fatigue life of the various details.1. this term. this examination establishes that no fatigue check is required between bridge stations 0 and 29.40 kN/m. The "exact" analysis. the section properties at the location where the fatigue truck is placed on the influence line for moment at a given detail should be used. can be taken as unity for the purpose of a preliminary analysis. According to Article 4.2. is 5. This will vary. gives a load distribution value of 0.2. The additional dead load. Calculation of Stress Ranges It was pointed out earlier in this Chapter (Section 5.1.
With experience.1.6.4 MPa when N = 140x10 6 cycles and ( ∆F) n = 34.4 MPa and 38.2. In this case.2.Fatigue Life Assessment of Flange to Web Weld A continuous fillet weld is used to join the flange to the web. 6.2. Fatigue Life Assessment of Flange Plate Splices Figure 26 shows that the girder flange plates must be spliced at Stations 30.4 m). which is the introductory statement setting out the fatigue requirements.4 MPa.1. In fact. the value is 110 MPa 2 = 55 MPa . 6. Therefore. it should be possible to eliminate most cases by inspection.3 m of the interior support (Sta. the fatigue strength according to Eq. N = 140x10 6 cycles.8 MPa when N = 93x10 6 cycles.5–1) and the constant amplitude fatigue threshold is 110 MPa (Table 6. Everywhere else. Application of Eq. The detail category is dependent upon 66 .3–1. 38. (An alternative way to think of Eq.1. 33. Substitution of these values gives ( ∆F) n = 30. however. the flangetoweb fillet weld detail is satisfactory with respect to fatigue.9 MPa (see Fig. 26).3x1011 = N 13 It has already been established that for locations within 4.2. The actual stress range for the bottom flange in the region Sta.4 m is 44. 19 also says that the permissible stress range need not be less than onehalf of the constant amplitude fatigue threshold.5–1).2.3–1 and Table 6.9 m.9 MPa.6.1.1 to 38.5–5) is first calculated as A ( ∆F) n = N 1/ 3 39.3x1011 MPa 3 (Table 6.7 m is also 44. AASHTO Article 6. This should be less than the CAFL.6.) The actual stress range for the top flange in the region Sta. 29.1. 17 (AASHTO Eq. This governs in both cases since it is larger than both 30. According to AASHTO Fig. refers only to components and details with fatigue resistance less than or equal to Detail Category C. which then provides the estimated maximum stress range. The plates change in both width and thickness at each location. this is Detail Category B.1. the fatigue life constant is A = 39. 19 would be to multiply the calculated stress by a factor of 2. Both these locations are located in the regions wherein N = 93x10 6 cycles.1 and 37.3. Since the actual range of stress for this detail is less than the permissible value of 55 MPa. For Category B.6.6. N = 93x10 6 cycles should be used.6.3 to 42.2. Comment: Details of Category A or B are seldom critical with respect to fatigue.
8.4 tw min frames.1 m. Both of these values are less than the permissible value of 34.5 and that the groove weld reinforcement is not removed. then Detail Category C is applicable. Fatigue Life Assessment of Transverse Stiffeners Transverse stiffeners are present in a girder in order to stiffen the web and assist in carrying the shear forces. they will be welded or bolted to the web and fitted tight against the underside of the compression flange. it is tw = 11. According to the AASHTO rules.1.5 MPa or A ( ∆F) n = N 1/ 3 14. Often. 37. the transverse stiffener must be connected to Figure 27 Stiffener Detail both flanges if it is used to receive lateral bracing.just how these transitions are made.1.4 x1011 = 93x10 6 13 = 24. If it is assumed that the transition slopes are not greater than 1:2.8 MPa (bottom flange) and at Sta. If this is the case. For economy. the fatigue life constant is 14. Information on these requirements is contained in AASHTO Article 6. these transverse stiffeners will be cut short of the tension flange (Fig.9 the stress range is 25.6. Thus. When transverse stiffeners are present. Once again.4x1011 MPa 3 and the constant amplitude fatigue threshold stress range is 69 MPa for this category. then the end in the tension flange 67 . then the stiffener 6 tw max Assume 10 must be welded or bolted to each flange of the mm fillet welds girder.10.5 MPa.1 mm common practice to use them as the connecting plate for diaphragms or cross. The groove weld at the transition in flange plate sizes is therefore satisfactory in fatigue.1. the fatigue strength is the greater of 69 2 = 34. When the stiffener will not to be fitted to both flanges. 30.9 MPa At Sta. the maximum tensile stress range is 31. If this is their only function. This is a realistic circumstance for most bridges.6 MPa (either top or bottom flange). It will be assumed in this example that some of the transverse stiffeners will have diaphragms or crossframes connected to them and some will not. the stiffeners are also used for the connection of lateral bracing. especially in the fitting process. Later in the example there will also be a discussion about the connection of lateral bracing members to the transverse stiffeners. 27).3 and Article 6.
and assuming that the flange to web weld will be 10 mm. 26). (AASHTO Article 6.2 / 2 = 413 MPa or .9 MPa and when N = 140x10 6 cycles ( ∆F) n = 218 MPa .4x1011 MPa 3 and the constant amplitude threshold stress limit is 82. A ( ∆F) n = N 1/ 3 14. For both these regions.3–1. The actual stress range in this zone is a maximum of 44.) This requirement relates to distortionbased fatigue and is discussed in Chapter 6.7 MPa) is already less than the permissible stress range (41.2. 28) will identify that the stress range at this location is 41. 27. In Field Section 2. that in this example the stress range at the critical location in the transverse stiffener need not have been calculated.1.6.1.3 MPa). the permissible stress range is again calculated as 41. As it happens. the fillet weld at the end of a transverse stiffener is a Category C' detail.6.9 MPa (Fig. Thus. This information is required in order to be able to calculate the stress at the toe of the stiffenertoweb fillet weld. Field Section 1.1. the maximum tensile stress range in the bottom flange is 37. the transverse stiffeners in Field Section 1 are satisfactory with respect to fatigue. It should also be obvious. The stress range at the level of the bottom flange (37. 26). 68 .region must be cut short of the flangetoweb weld a distance not less than four times the web thickness and not more than six times the web thickness. any transverse stiffener in Field Section 2 that does not extend to the tension flange is satisfactory with respect to fatigue.7 MPa (Fig. the maximum gap (6 tw) will require that the end of the stiffener be ( 6 x 111) + 10 = 77 mm from the nearer . In . the fatigue strength of this detail is the greater of 87.3 MPa.7 MPa. .3–1 and Table 6.4 x1011 = N 13 When N = 93x10 6 cycles .7MPa = 34. The stress range at the critical location is calculated as (1159 1260) x 37. In accordance with AASHTO Fig. the result is ( ∆F) n = 24. with experience. The fatigue life constant is 14. the permissible stress range is 413 MPa .7 MPa .10. Thus. then. this is identical to the permissible stress range. Since the permissible stress in this region is 41.3 MPa.8. According to Fig.2. 6. Figure 28 shows the stress gradient at the detail in this region. side of the flange.3 MPa (see calculation above). A calculation of the stress range at the extremity of the stiffener (similar to the calculation illustrated in Fig.1.
In Field Section 2.0 MPa . 1159 1260 77 24 Critical location 37. A horizontal plate can be fastened to the web or to the flange by welding or by bolting.) This is a Category C' detail.N. Fatigue Life Assessment of Lateral Bracing Attachment Lateral bracing can be attached to the girder in different ways.9 MPa = 44. The designer can either very short portion of the girder to the right of bridge station 30. 26. Thus. and the most convenient way of attaching them to the girders is to use the transverse stiffeners. just as was the stiffener welded only to the web.1 m.9 MPa. 27. . as calculated for the stiffeners that stopped short of the flange.7 MPa for the tension flange and in Field Section 2 the maximum stress range is 44. Review of the calculated stress (not given in the example) reveals that the stress range has decreased to 42. some transverse stiffeners will have to extend from flange to flange and be fastened to those flanges. Accordingly. See Fig.4 m. (Bolting is also a possibility.A. According to Fig. this will be only a range is only 41. the condition in Field Section 1 is satisfactory. exceeds 413 MPa / [(1260 − 24) 1260] = 421 MPa . the stress range at the end of the stiffener (i.e.7 MPa Figure 28 Stress Gradient at Transverse Stiffener Undoubtedly. the permissible stress range is 41. this means that the detail is not satisfactory.3 MPa. This is because diaphragms will be present. Since the permissible stress accept the small amount of underdesign (about 6%) or ensure that this type of transverse stiffener is not located anywhere in the region of Field Section 2 where the flange stress . if desired.1 MPa at bridge station 31..3 MPa. By inspection. In Field Section 1 the maximum stress range is 37. at the level of the inner surface of the flange) will be [(1260 − 24) 1260] x 44. The lateral bracing members 69 . these stiffeners will extend all the way to the flange and be welded to it.
In Field Section 1. (AASHTO Article 6. According to Table 6.2). According to AASHTO Table 6. the stress gradient diagram is that shown in Fig. Category E is a severe fatigue detail.3–1.1. It is likely that the connection plate will have to be longer than either of these values for proper attachment of the bracing members.6.0/2 = 15. then it must be attached to the stiffener when on the same side of the girder as the stiffener or centered on the stiffener otherwise. the result is ( ∆F) n = 15.1.3–1 the lateral connection plate is now a Category D detail. For the example girder. the lateral bracing should be much closer to the plane of the lower flange. Thus. Try a similar detail except that there will be a transition radius >50 mm at the end of the plate and the welds ends will be ground smooth. In order to not exceed the permissible stress of 15. this is referred to as a "lateral connection plate. transition radius on the weld.7 MPa . simply take 15.61x1011 MPa 3 and the constant amplitude threshold stress limit is 31. a calculation will show that the horizontal plate must be located 735 mm from the bottom. it will be seen that it presents a number of difficulties related to fatigue.2.1.3.are then bolted (usual case) or welded to this connection plate. The fatigue life constant is 3.5 MPa or A ( ∆F) n = N 1/ 3 3.2. and the thickness of the plate. the permissible fatigue stress is the greater of 31. the treatment of the weld ends (no treatment.") If transverse stiffeners are present and the lateral connection plate is to be located within the web depth. 28. this will be Category E when the plate length is greater than either 12 times the plate thickness or 100 mm. The fatigue category for this type of detail will depend upon the length of the plate (in the direction of stress).5 MPa as the permissible fatigue stress in both regions.6.6. and no special finishing of the weld.61x1011 = N 13 When N = 93x10 6 cycles . (In the AASHTO Specification. or transition radius and welds ground smooth).21x1011 MPa 3 and the constant amplitude threshold stress limit is 70 .7 MPa and when N = 140x10 6 cycles ( ∆F) n = 13. try a square plate that is less than 25 mm thick. Although the use of lateral connection plates within the web depth and centered on transverse stiffeners is a detail that has been widely used in the past.5 MPa. The fatigue life constant is 7. For convenience.0 MPa. This is not a practical solution.
48. According to AASHTO Fig.3 MPa/2 = 24. Furthermore.3 MPa. The type of lateral bracing attachment plate that attaches to a transverse stiffener must also meet a distortionbased fatigue requirement.8 MPa and when N = 140x10 6 cycles ( ∆F) n = 30. A highstrength bolted connection is also a possibility.0 MPa or A ( ∆F) n = N 1/ 3 39. Calculation will show that the criterion 48.3 x1011 MPa 3 and the constant amplitude threshold stress limit is 110 MPa. Although this is a considerable improvement over the previous condition. it still places the connection plate a considerable distance transverse from the lower flange. the result is ( ∆F) n = 34. if the horizontal connection plate is made the same thickness as the flange plate.7 MPa in Field Section 1 and 45.3x1011 = N 13 When N = 93x10 6 cycles . 26). Thus.0 MPa in Field Section 2 (Fig. a transition radius of between 150 and 600 mm is used at the ends. the maximum tensile stress range is 37.4 MPa . then.2.2. the interior "ends" of the plate connection bracing forces plate near the transverse stiffener could not receive this treatment. For both these regions. 6.1 MPa will govern throughout.6. Although the extreme bracing forces ends of the horizontal plate could have a radius and horizontal the welds ground.3–1. the permissible stress range is 55. Thus. In the bottom flange. The situation is shown in Fig. this connection detail is satisfactory for use anywhere along the length of the girder. then this is a Category B detail. 29. the condition reverts to that of a Figure 29 Transverse Stiffener and Horizontal Connection Plate square plate. Thus.6.0 MPa . It is likewise a Category B and the connection would be much simpler than the 71 . This topic is discussed in Chapter 6 and the fatigue design of this lateral bracing attachment is completed in Example 11. All things taken into account. and the welds ground smooth. the designer should consider simply attaching the horizontal connection plate to the lower flange. the permissible stress range for this detail is the greater of 110 / 2 = 55.3–1 and Table 6. in this girder it stiffener is inevitable that some connection plates will have girder web to be placed around the transverse stiffeners. The fatigue life constant is 39. The necessary location of the horizontal plate is now 455 mm above the bottom of the girder.
26). which is the value 33. there is also a region of positive stress range (which results from loading in the opposite span).1 is always in compression. in order that filler plates will not be required for the flange splice plates. i. as Figure 26 identifies.e. 26.1 MPa (Fig. However. The stress is to be calculated on the gross crosssection. 30. 29. Positive bending moment predominates in Field Section 1. and the calculations done earlier for that detail identified that the permissible stress range is 55 MPa. however. the designer should be alert for other situations that require a fatigue life assessment. which has a relatively high fatigue life. The AASHTO Specification designates Category B for bolted connections that are slipcritical. The largest stress range in the short portion of the girder that must be checked is only 8. and so no fatigue check is required..2 MPa at location 29. however.1 m. The examination of dead load stress plus two times the live load stress identifies that the region from Sta.3–1). As the details are completed. the effect of the bolt holes on the strength of the flanges would have to be taken into account. Fatigue Life Assessment of Bolted Splice Figure 26 implies that the girder is spliced at Sta. However. the bottom flange stress range controls the design.0.2 MPa cited above. Fatigue Life Assessment of the Shear Studs In this girder. The detailed calculations show that the stress range in the bottom flange is 33.1 m (end of Field Section 1) in which a fatigue check is required for the studs.welded one. 29. The splice is located in the region wherein N = 93x10 6 cycles. Field Section 1 is composite and will therefore have shear studs. From inspection of Fig. The actual splice will likely be made just to the left of that location. This example has dealt with the major elements of fatigue life assessment for the welded plate girder. Assume that the splice will be centered at Sta.2.1.0 m. This leaves the region between Sta.1 and 30. 0 to 29. Field Section 2 is noncomposite and so no studs will be present. The flangetoweb weld was also a Category B. Studs are a Detail Category C (AASHTO Table 6. at the junction of Field Section 1 and Field Section 2. The actual location of the center of the splice will depend upon the length of the splice plates required for the flanges. Since this is greater than the actual stress range. It will be taken by engineering judgment that the studs are satisfactory in fatigue. 72 . the bolted splice will be satisfactory with respect to fatigue.6.
This is reflected in current specifications. The only difference. The assumption has been that these stresses can be calculated. The main defense against this source of cracking is proper detailing. this is usually performed after the fact and requires that field measurements be made. The loads used are the same as those associated with the strength design of the members. 73 . (There are also practical reasons for cutting the stiffener short: the stiffener will have to be made to a precise length if it is to extend from flange to flange. is the effect of the stress gradient. then. and geometrical discontinuities with respect to fatigue lives of fabricated steel elements.2 Examples of DistortionInduced Cracking An example of the phenomenon is illustrated in Fig. fatigue crack growth results from the imposition of deformations. large strains are imposed in the gap region because of the significant change in stiffness between the stiffened and unstiffened (gap) 1 crack Detail at Bottom of Transverse Stiffener Girder and Transverse Stiffener Fig. 30 Fatigue Cracking from OutofPlane Movement Experience gained over the past 20 years has shown that.1 Introduction Most of the topics so far have discussed the effect of stresses acting on preexisting flaws. the fatigue life of the detail is independent of whether the stiffener terminates in the web or is extended down to the flange. These deformations are not anticipated in the design process. however. Designers are not likely to be able to identify the need for such calculations in the course of their work. and this. in turn. Standard practice for many years was to cut transverse stiffeners short of the bottom (tension) flange so as to avoid a severe detail for fatigue if the stiffener is welded to the flange at that location1. 6. As will be seen. cracks. If lateral movement of the top flange relative to the bottom flange takes place. 30. Although it is possible in some of these cases to calculate a stress range.) The height of the gap between the end of the stiffener and the girder flange is usually quite small. not loads. In many instances. is dependent on experience.Chapter 6 DistortionInduced Fatigue Cracking 6. in fact. this type of fatigue crack growth results from the imposition of relatively small deformations. in local regions of a member. usually at an elementary level. usually outofplane.
but it could even be due to shipping and handling. 74 . This type of deformation is shown (exaggerated) in the detail in Fig. Thus. the crack may have turned upwards or downwards in the web and thus be aligned in the most unfavorable orientation with respect to service load stresses. 31 Floor BeamtoGirder Connection web of the girder. the floor beam will rotate as shown. New designs accommodate the situation in different ways. As this rotation occurs. the crack will also extend across the toe of the fillet weld at the underside of the stiffener and for some distance into the web. Under the passage of traffic. 30 shows the crack emanating from the weld toe at the bottom of the stiffener. where a floor beam is attached to beam a vertical connection plate that is welded ∆ to the web of a girder. However. the strains that are produced are so large [32] that it may take relatively few cycles for a crack to propagate. The flange movement could be the consequence of transverse forces in a skew bridge. if crack growth has gone on for some time. The detail just described (Fig. 31. Up to this stage. Another illustration of a case in θ which outofplane movement can produce fatigue cracks is shown in Floor Fig. the crack is more or less parallel to the direction of the main stress field that the girder will experience in service. which is flexible in this outofplane direction. shortening of the top flange of the floor beam can only be accommodated by deformation within the gap at the top of the connection plate. then further growth of the crack is unlikely. 30) has been the source of many fatigue cracks in the past. The detail in Fig.regions of the web. Often. reducing the possibility of fatigue crack growth that is induced in this manner. the Detail at bottom flange of the floor beam lengthens gap Girder and the top flange shortens. Lengthening of the bottom flange will not be restrained because it is pushing into the Fig. because the top flange of the girder is restrained by the deck slab. if the source of the displacementinduced fatigue can be identified and eliminated. Typically. However. 31.
∆. It can be expected that fatigue cracks could occur under relatively few cycles of load. 31 has been confirmed by field measurements taken by Fisher [32]. the webtoflange fillet weld is a Detail Category B for the inplane bending stresses ("continuous fillet weld parallel to the direction of applied stress"). What is the fatigue life of this detail? Solution: ∆ L t Fig.The behavior illustrated in Fig. The web plate thickness is t = 12 mm. the weld is Category C when the outofplane displacement depicted in Fig. or reference to a design handbook. Application of the momentarea principle. and a measurement of the displacement of the gap shows that ∆ = 0. as Fig. Therefore. 32 Example 10 According to the AASHTO Specification. Moreover. However. The detail is repeated here. the stress range at the webtoflange weld and at the stiffenertoweb weld will be identical. The fillet weld that is used to attach the connection plate to the web is a Detail Category C ′ ("transverse stiffener to web weld"). 32 takes place ("filletwelded connection with weld normal to the direction of stress"). Category C ′ will govern the design. the field study showed that each passage of an axle caused a significant stress range at the top of the connection plate. For the outofplane displacement under examination. fatigue cracks could develop either at the weld at the top of the connection plate or at the webtoflange fillet weld of the girder. or both. the length of the gap is L = 15 mm. although of course the fatigue life will depend largely upon the deformation. The residual tensile stress in this small gap will tend to be very high because of the proximity of the two welds. In this situation. 32. EXAMPLE 10 Given: Figure 31 illustrated a floor beam to girder connection where outofplane cracking could occur. will show that the moment at the end of a fixed ended beam when it is displaced transversely an amount ∆ at one end is M = 6 EI ∆ L2 75 .004 mm. that actually takes place as a result of the rotation of the floorbeam.
Thus.2 and used in Example 10. From Table 2.4x1011 (128) 3 = 687 x 10 3 cycles Comment: ( ) This example shows that even a very small amount of outofplane displacement can have a significant effect on fatigue life. Because the actual stress range is greater than the CAFL/2 = 41. the constant amplitude fatigue limit (threshold stress) is 82. 6.4 MPa. it was customary that no connection be provided between the stiffener and the tension flange because this would adversely affect the fatigue detail category of the girder. Section 6. the stress so calculated is really the stress range.1 Web Gaps in Multiple Girder and Girder Floor Beam Bridges Diaphragms and crossframes in multibeam bridges are used to provide torsional stiffness to the structure. the differential vertical movement produces an outofplane deformation in the 76 .7 MPa for a Detail Category C ′ and the value of the fatigue life constant is 14.004 mm ) (12 mm ) = 128 MPa (15 mm ) 2 The number of cycles that a Detail Category C ′ can sustain a for stress range of 128 MPa can now be calculated. the stress is σ = My 3E∆t 6 EI ∆ t 1 = 2 = L 2 I I L2 Since the displacement ∆ varies each time the floorbeam rotates. solving for ∆σ ( ≡ σ ) and using the specific values given: ∆σ = ( 3) ( 200 000 MPa) ( 0.2.4 x 1011 MPa 3 .3 provides more general comments and gives further examples for bridge structures. The connection is usually made to transverse stiffeners that are welded to the girder web.3 Further Examples of DistortionInduced Cracking The distortioninduced fatigue conditions that can be present in fabricated steel members was introduced in Section 6. the fatigue life is calculated according to Eq. The diaphragms or crossframes are connected to the longitudinal members by means of transverse connection plates. ∆σ . 8: N = M ∆σ 3 = 14. Since adjacent beams deflect differing amounts under traffic. 6. and this often provides fatigue conditions similar to the floor beamtogirder connection plate discussed in Section 6.3. Sometimes these stiffeners were not attached to either flange. Thus. In the past.
web gap at the stiffener ends if they are not attached to the beam flange. The magnitude of this outofplane movement depends on the girder spacing, amount of bridge skew, and type of diaphragm or crossframe. Various types of diaphragms are used, ranging from rolled sections to simple Xbracing made of angles. Figure 33 shows the underside of a multiple girder bridge that has an Xtype crossframe bracing system. In the web gap, cracking developed in the negative moment region (i.e., top flange in tension) of this continuous span structure. This cracking is illustrated in Fig. 34, which is a view along the length of the bracing toward the girder and its transverse stiffener. The transverse stiffener to which the Xbracing is attached was not welded to the top (tension) flange for the reasons described earlier. This permitted outofplane displacement to occur in the web gap and, consequently, cracks formed along the webtoflange fillet weld and at the top of the connection plate.
Fig. 33 XBracing and Girders
77
Fig. 34 Fatigue Cracks in Web Gap of Diaphragm Connection Plate One of the earliest and most common sources of fatigue cracks in welded bridges is the cracking in the web gaps at the ends of floor beam connection plates. (See also Section 6.2.) These cracks have occurred in the web gap near the end reactions when the floor beam connection plate was not welded to the bottom (tension) flange. However, the most extensive cracking is observed in the negative moment regions of continuous girder bridges where the connection plate is not welded to the top (tension) flange of the girder. One such case is illustrated in Fig. 35. The view is toward the web of the girder. Cracking is seen along the fillet weld toe at the webtoflange weld (horizontal crack) and at the top of the stiffener (vertical crack). The illustration shows that the floor beam has been bolted to the transverse stiffener and that the top of the stiffener has a small corner clip where it reaches the top of the girder web. The clip is provided so that the stiffener clears the girder webtoflange fillet weld. Thus, even though the stiffener extends to the underside of the girder flange, it is not fastened to the flange and the presence of the stiffener clip creates the gap in which deformation was concentrated.
78
Fig. 35 Fatigue Cracks at Floor Beam Connection Plate Web Gap
6.3.2 Web Gaps in Box Girder Bridges
Internal diaphragms in various types of box girder structures are a source of web gap cracking as a result of crosssection distortion. This type of cracking has been seen both in continuous and simple span spread box girder structures. An example of this type of cracking is shown in Fig. 36. The structure is an elevated, curved continuous span structure. Two curved steel box girders with internal diaphragms support the reinforced concrete deck. Fatigue cracking occurred in the top web gaps (negative moment region) near the piers and in the bottom web gaps (positive moment region) at the diaphragm locations. In both cases, the cracking was the result of outofplane movement in the web gap. This resulted from the box girder distortion and the resultant diaphragm forces. The cracking that occurred at the top web gap is shown in Fig. 36. The photograph, which was taken inside the girder, shows the (sloping) girder web in the lefthand portion of the figure, a transverse connection plate welded at right angles to it, and a gusset plate (horizontal) that formed part of the lateral bracing used for shipping and construction. The gusset plate is about 80 mm below the top of the connection plate and there is a small gap between the top of the transverse connection plate and the underside of the girder flange. (The photograph shows that a loose plate has been placed in this gap above the connection plate. It was later fastened in place in order to prevent further movement of the girder web in
79
These cracks are also the result of outofplane distortion in the web gap. The web gap displacement results primarily from torsional distortion of the box girder when eccentric loads are applied to the structure. An example of this type of cracking is shown in Fig.this region. 36 Web Gap Cracking at Box Girder Diaphragm Connection Plate Fatigue cracking has also been observed in the bottom web gaps at the locations of internal diaphragms in simple span steel box girders. 80 . 37.) Fig.
This gap is present between the end of the connection angles or at the end of the welded connection plates. then the floor beams are bolted to the welded transverse connection plate. In older structures.3 Long Span Structures A number of tied arch structures have been built in which the floor beams are framed into the arch tie girder with web shear connections alone. Figure 38 shows a crack in the web gap of the floor beam just above the end connection to the tie girder. The cracks form in the floor beam web along the webtoflange connection at the floor beam web gap. the double angles are riveted to the floor beam web and the outstanding angle legs are riveted to the tie girder. 37 Web Gap Cracking at End of Transverse Connection Plate 6. In such instances. If this is the case. The cracks extend parallel to the floor beam flange along the length of the web gap and then begin to turn and propagate toward the bottom flange. As will be discussed in Chapter 8. a transverse welded connection plate on the tie girder webs is utilized. No direct connection is provided between the floor beam flange and the tie girder.3. Cracks have been observed in the floor beams of structures with either riveted end connections or welded transverse connection plates. The photograph also shows a drilled hole at the lefthand end of the crack.Fig. In more recent arch design. this is often an effective way of stopping the further growth of a crack. 81 . the floor beam end connections were riveted.
38 Floor Beam Web Gap Crack at End Connection to a Tie Girder 82 .Fig.
Cracks have also developed in box tie girders at internal diaphragms.Similar cracks have been observed in long span trusses where floor beams are connected to the bottom chord panel points. The fatigue cracking occurred in tie girder web gaps (created by internal diaphragms) at the floor beam locations as a result of the web gap outofplane distortion. If copes must be included. The cope is generally made by flamecutting the material. may be coped. There are many examples of fatigue cracking at cope locations [20]. the region of the cope will have stresses due to bending because of the restraint at the connection. Fatigue cracking at coped beam locations is not so much related to distortioninduced fatigue as it is illustrative of cracking at a location where the calculated stress is zero. as is illustrated in Fig. and experience shows that workmanship is often unsatisfactory. the flange of one of the members is often cut back. Specific information on the fatigue life of a coped beam can be found in Reference 34. Nevertheless. the flanges may simply be narrowed: this is called a "blocked" beam. This creates diaphragm forces that torsionally deform the tie girders. and the best solution in the case of new designs is to avoid copes entirely.) In other cases. The radius of the cope may be small and the cutting done unevenly. fatigue crack In the case of coped beams.4 Coped Beam Connections In order to facilitate the easy connection of one flexural member to another. The web gap distortion is the result of the bending of the floor beams and the relative movement between the two tie girders. the flamecutting process can leave a region of hardened and brittle material adjacent to the cut. execution of the work and inspection must be of a high standard.3. In addition to the potential for fatigue cracking created by such workmanship. either the top or bottom flange. The coped end of the beam is at a location of theoretical zero stress since the connection must necessarily be one that does not transmit moment and the shear force is carried by the web. or both. 39 (The detail shown was used extensively in the past in throughgirder railway spans. Figure 39 Bottom Flange of Floorbeam Coped at Connection to Girder 83 . 6.
(In this discussion. Taken all together. In the illustration. However. these conditions mean that the type of detail shown is very susceptible to fatigue crack growth [25. 40. Because the web is very flexible in the outofplane direction. 40 Connection of Lateral Connection Plate and vertical welds.3. the connection between the lateral bracing system and the girder will be to a horizontal plate welded to the girder web. it is assumed that the top flange will be braced by the deck. even simple connections do carry some moment. as the bracing forces push into and pull on the web at this location. this is termed a "lateral connection plate. The region is also a zone of high residual stresses because of the proximity of the vertical and horizontal welds. the web will rotate about a vertical axis formed by the back of the vertical stiffener and the plane of the web. In either event.) Sometimes.6. the arrangement will be as shown in Fig. In other gap arrangements. Consequently. 33].") If transverse stiffeners are present.6 Other Examples Many end connections are designed as "simple. Lack of fusion in the weld or other microdefects can also be anticipated at the weld terminations. usually the case. the transverse horizontal connection plate is fitted stiffener around the vertical transverse girder web stiffener and welded to it where bracing force they are in contact. and this means 84 . this causes large strains in the web in the gap region when the lateral connection plate is not attached to the transverse stiffener. 6. the vertical stiffener passes clear through a slot in the horizontal plate without attachment horizontal at that location." that is. (In the AASHTO Specification. it is connection plate (lateral connection highly likely that a gap will be left bracing force plate) in the region shown in order to avoid intersection of the horizontal Fig.3.5 Connections for Lateral Bracing The connection of lateral bracing to girders should be done as close to the plane of the bottom flange as possible. it is assumed that no moment is transmitted.
1. it was common to place the ties directly on the top flange of the flexural members in the floor system.6. or welds in ways not contemplated by the design. a stringer flange.3 simply directs the reader to Article 9. The AASHTO design specification provides a separate section on distortioninduced fatigue (Article 6.3). As has already been illustrated in this Chapter. for example. and orthotropic decks (Article 6. can be rotated as traffic passes over the structure. main) components.3.3.1.6. Example 11 will illustrate some of the features of this section.1.6. it can easily be appreciated that it is difficult for a specification or design standard to provide very much in the way of explicit rules relating to distortioninduced fatigue. If the loading is cyclic. 6. This will be illustrated in Example 12. 85 .that the connection elements will deform under the moment.2). rivets. lateral connection plates (Article 6. cracking resulting from improper detailing of connection plates is a significant source of fatigue cracking in bridges. fatigue cracks may develop in the angles themselves or in the bolts. which gives detailing requirements for orthotropic decks. In the cases of the mechanical fasteners.1. cracking can occur at the webtoflange junction.1. Another example of distortion concentrated in a small space is when a beam or girder flange is loaded directly.1). The articles relating to connection plates provide rules for this important topic of girder detailing.3 also alerts the user to the possibility of fatigue cracking as a result of excessive outofplane flexing of a girder web. prying forces develop that may cause fatigue cracking under the heads or nuts.3.7. Particularly in older bridges. It contains a general statement that stresses the importance of proper connection of transverse components to longitudinal (i.3. Thus. Article 6.1. End rotation at web framing angles causes the angles to deform and load the rivets. The most obvious example of this is the case of railway beams or girders. Article 6.6.6.4 AASHTO Specification Requirements Relating to DistortionInduced Fatigue Based on the discussion and examples thus far in this Chapter.6. Subsections then offer specific information relating to transverse connection plates (Article 6. bolts. or welds. Those requirements are mainly a reflection of good practice and experience derived from this type of deck construction.3.3).. If the flange is too flexible in this direction.8.e.
a lateral connection plate was examined for its loadinduced fatigue life.4. (See "Fatigue Life Assessment of Lateral Bracing Attachment" in that example. but this simply refers the reader to Article 6.4. The requirements for the case of flexure. See Example 9. The distance identified as satisfactory for loadinduced fatigue. Article 6. Comment: Recall that this stiffener arrangement was not the preferred solution.6. then in the case of a stiffened web it should be attached not less than onehalf the flange width above or below the flange. Is this arrangement satisfactory with respect to distortioninduced fatigue? Solution: The AASHTO requirements for distortioninduced fatigue of a lateral connection plate are given in Article 6.3. This minimum distance requirement is to ensure that distortioninduced fatigue cracking does not take place in the web between the lateral connection plate and the flange.10.10.3. therefore is satisfactory also for distortioninduced fatigue.) One possible solution was to center a horizontal plate around a transverse stiffener. the AASHTO requirement is that the horizontal connection plate be at least 305 mm above the flange. EXAMPLE 12 Given: Does the web of the girder used in Example 9 (Chapter 5) meet the AASHTO distortionrelated fatigue requirements for webs? Solution: The possibility of fatigue due to web flexing is first cited in AASHTO Article 6. If the connection plate is not attached to the flange. Thus.1. it is considered that web flexing will not occur to any significant extent. are given only for girder webs that have no longitudinal stiffeners. however. The maximum flange width in this girder is 609. the plate had to be located 455 mm above the bottom of the girder in order to meet the requirements of the loadinduced stresses.3. (If longitudinal stiffeners are present.5 Design Examples EXAMPLE 11 Given: In Example 9 (Chapter 5).6.1. Table 4). 29.6 mm (see Example 9. 455 mm. the preferable solution.) 86 .6.2. According to Example 9. as shown in Fig.
The term R h is a flange stress reduction factor.2. of which only the governing one for this example has been presented here. Comments: 1. when this requirement is met the web is stocky enough that it will remain elastic under working loads.4. strength of the compression flange material is 345 MPa.2. E is the modulus of elasticity of steel. if 2D c E < 5. For homogeneous girders.76 111 .) The expression on the righthand side of the first inequality is equivalent to the allowable web slenderness given in the allowable stress design provisions of the 1992 AASHTO allowable stress design specification. where the girder is noncomposite. D c = 1524 2 = 762 mm and t w = 111 mm . According to Table 4. t w is the web thickness. the flexural stress resulting from the fatigue load is to be taken as twice the calculated value. its value is unity. however. fatigue cracking due to web flexing is not a concern.10. due to the fatigue load. and f cd is the maximum compressive stress.4. Thus— 2x762 < 5.3 actually gives a threepart rule. in the compression flange. An experienced designer would probably recognize that this girder has a relatively stocky web and that flexural web buckling leading to fatigue is unlikely. Since this is less than the permissible (345 MPa). not stress range as is done generally for fatigue life checks. Note that the requirement deals with stress. then the compressive stress (resulting from the fatigue load) in the flange can be as large as the yield strength of the flange. In other words. D c is the depth of the web in compression. According to Article 6. 137 < 139 Since the inequality requirement is met for the quantities associated with this girder.3. In this girder. The yield . The depth of the web in compression will be greatest in Field Section 2. 200 000 345 or. calculations show that the maximum compressive stress in the compression flange due to two times the fatigue load is 149 MPa. that the compressive stress is to be taken as twice that calculated for the fatigue load. This is consistent with the requirements explained in Section 5. tw Fyc then it is required that f cd ≤ R h Fyc In these expressions. Recall.For flexural stresses.10. (The AASHTO Article 6. Fys is the yield strength in the compression flange. Based on the 87 .76 .
the web flexing was local—the region in the vicinity of a transverse stiffener and a horizontal connection plate was under examination. It is fortunate that most of the cracks that develop from local distortion lie in planes parallel to the loadinduced stresses. 88 . it will be concluded here. 6. cracks can slow.5. As a result.3. since stress intensities around distortioninduced cracks may decrease with increasing crack length.6 Summary Outofplane distortion that is concentrated in small web gaps remains a large source of fatigue cracking in bridge structures.4. plate girders. box girders and tied arches. that fatigue resulting web buckling due to shear will likewise not be a concern. and even stop. once a certain flexibility has been provided.calculations already done. without calculation. distortioninduced cracks have not caused significant numbers of fractured flanges or hampered the loadcarrying capability of the bridge member in which they form. The AASHTO rule for this case is presented in Article 10. 2.6. It should be recalled that the possibility of fatigue cracking due to web flexing was also the subject of Section 6. It develops in nearly every type of bridge structure including trusses. In that case. suspension bridges. In addition.4.
7. when a crack is found in a steel structure more careful inspection will reveal additional cracks in similar elements. If no action is taken to eliminate the cause of cracking. a reasonable protocol for fatigue crack investigation is as follows: • Carry out a remaining fatigue life analysis based on loadinduced fatigue.2 Protocol for Fatigue Crack Investigation Based on the forgoing. Quick fixes. If the inspection reveals that cracking is present. If this analysis identifies details that have little or no remaining fatigue life. Global strains calculated using the conventional methods of analysis are usually conservative. Ideally. The previous chapters have given the scientific foundation for the fatigue life evaluation.1 Introduction It is the responsibility of the engineer to evaluate the fatigue life of an existing structure. At the same time.Chapter 7 Inspection and Repair of Fatigue Cracks 7. these locations are then candidates for an examination when a physical inspection of the structure is carried out. local strains are often not even identified by the traditional methods of structural analysis. often worsen the problem. 88 . combined with knowledge gained through practical experience of examining steel structures and dealing with cracks found in those structures. For important structures. the discovery of a crack should be taken seriously by the inspection team and reported immediately to the engineer responsible for the structure. if necessary. more cracks usually appear at other locations. • Use the shop drawings of the members that make up the structure to help identify those details that are susceptible to distortioninduced fatigue. decisions should be based upon a scientific understanding of crack propagation. Often. carried out by untrained personnel. This will provide a better estimate of loadinduced stresses. then the responsibility will extend to recommendations for remedial action. This chapter provides recommendations directed toward helping the engineer devise a plan for inspection and then to decide what to do if a crack is discovered. The principles of the remaining fatigue life assessment have been covered in earlier chapters of this document. and then. For these reasons. it may be appropriate to carry out field measurement of strains. to advise on a protocol for inspection.
often it is not followed in practice. It is possible that. The first decision that must be made following the discovery of cracks is whether the structure is safe in its existing condition. more than one factor contributes to a critical situation. The examination at this point can lead to one of several options.• If a physical inspection of the structure is indicated. Occasionally. consider appropriate repair methods and advise on inspection procedures subsequent to the repairs. severe stress concentrations.) If cracks are located. In some instances. 89 . Although this course of action seems obvious. examination of a crack segment removed from the structure can assist a specialist in the determination of the most probable causes. corrosion damage. Because of redundancy. this should be carried out by trained personnel. continued monitoring of the structure is indicated. Often. In those instances where the structure is not shut down for use. Stress ranges brought about by temperature changes and other types of deformationinduced stresses (sometimes called secondary stresses) present similar problems.3 Identifying the Causes of Cracking Cracks in a structure may be identified as a result of an examination that starts with fatigue life assessment. the structure is deemed to be safe at existing load levels even though cracks have been identified. an attempt should be made to determine the causes of cracking before any corrective action is taken. inappropriate fabrication procedures. The parameters that effect the stress ranges at a given location can be difficult to quantify. • • • The structure is deemed to be unsafe for the intended use and it must be shut down.2. (Methods of inspection are discussed later in this chapter. The structure is deemed to be safe providing that load levels are reduced. indicate that the structure is subjected to high stress ranges. as described in Section 7. although remedial action is not required. • 7. but they may simply be seen in a routine inspection of the structure. outofplane displacements. Frequently. dynamic magnification (impact). bad welding processes. Fatigue crack growth mechanisms are activated by factors such as large numbers of stress ranges. signs of high stresses. the most accurate evaluation possible can not do more than provide a list of candidate factors ordered according to their importance. the effect of truck passages over a worn expansion joint at the end of a bridge may be impossible to evaluate without strain gauge measurements. For example. such as wear marks on contact surfaces or visible outofplane movement. See reference [25]. It is important that the repairs account for the causes of the cracking. and large defects.
and therefore a check can be performed using standard relationships between stress range and total fatigue life. corrosion damage. and the effects of corrosion—may have influenced cracking. The structure should be examined for evidence of large eccentricities. comparable conditions do not develop in riveted and bolted bridge structures. Note that if cracking is visible. cracking in welded bridge structures is the result of i) flaws that escaped detection. For example. (Except for secondary and displacementinduced stresses. Category E'.) Most critical cracks exist in tension areas. eccentricities due to fitup that may be acceptable for some structures might weaken a structure subjected to fatigue loading. this information should be checked against recommendations for fatigue strength in the design specification in order to determine whether cracking would be expected under such circumstances. If possible. 7. is seen in coverplated beam bridges that have flange thickness greater than 20 mm. Note that lack of weld penetration which is oriented transversely to the direction of fatigue loading is equivalent to an initial crack. outofplane movements. When details are located in compression stress regions and no possibility of stress reversal exists. and fabricationinduced discontinuities. Those design details that fall into AASHTO Category E or E' are particularly susceptible to fatigue crack growth in steel bridges. It is appropriate to examine further some low fatigue strength details that have been or are being used in steel bridges. The reason for making such a check is that if fatigue cracking cannot be explained in this way. The fact that the structure has survived without fatigue failure for a given period may be the only piece of information available that is completely certain. Under these conditions. ii) the use of details more severe than assumed in the design. it may be assumed that there is no remaining fatigue life at that location. Although the use of coverplated steel beams has almost disappeared from use 90 .) These discontinuities are the result of poor fabrication and quality control. all sources of fatigue loading may not have been identified. other factors not covered by fatigue strength curves—such as large defects due to poor fabrication methods. The lowest fatigue rating. there is a reduced fatigue risk. or iii) secondary and displacementinduced stresses.4 Cracking at Low Fatigue Strength Details and Defects Frequently. When fatigue loading information and code recommendations cannot explain fatigue cracking. crack growth will be confined to the residual tensile stress zone of the detail unless outofplane deformations occur. Special requirements exist for the fabrication of steel structures subjected to fatigue loading.Knowledge of stress ranges is usually related to the accuracy of estimates of the number of fatigue cycles applied to the structure. (See Figure 5.
it is intercepted by a retrofitted hole intended to stop the crack from moving into the lower flange of the girder. to transverse stiffeners and connection plates. Figure 41 shows the bottom flange of a rolled steel shape to which a cover plate has been attached by welding. In the figure. These include certain longitudinally or transversely loaded groovewelded or filletwelded gusset plates or other attachments.in new structures because of its low fatigue rating. 42. such as the one illustrated in Fig. Figure 41 Fatigue Crack at End of Cover Plate Another low fatigue strength detail is the welded web gusset plate. details that are equivalent to the coverplated beam continue to be used. but not attached. However. They eventually link together as the small individual cracks grow and coalesce to form a single large crack. the horizontal gusset plate is the dark area at the top of the photo. In addition to the usual conditions of potential crack growth at the toe of a fillet weld. 91 . As the crack moves downward. it is likely that multiple initiation sites are present. It started at the toe of the fillet weld at the location of the cover plate termination. A fatigue crack is seen emanating from the fillet weld region near the junction of the rectangular hole and the girder web. A fatigue crack is present in the rolled section. These are particularly susceptible to crack growth when they are adjacent. Even if the transverse weld were not present. there will be a significant stress concentration at this location of abrupt change in crosssection. fatigue cracking from the termination of the longitudinal welds (or the longitudinal weld end returns) is highly likely. many existing bridges have these beams. In this case where a transverse end weld is present. A transverse (vertical) web stiffener passes through a rectangular hole in the gusset plate.
Figure 42 Fatigue Crack at Web Gusset The fatigue crack shown in Fig. 42 was the result of combined cyclic stresses—those from the expected inplane stress and those from the unexpected outofplane stress in the web gap. Large initial defects and cracks that result from poor quality welds are a primary source of fatigue crack growth. its welded connection to a stresscarrying member means 92 . The latter stress develops from the lateral forces that cause the gusset plate to twist and deform the web gap. In Fig. (The longitudinal weld is located at the junction of the top side of the connection plate and the edge of the flange. however.9 m long lateral connection plate to the underside of the floor beam flange. Even though the secondary member does not carry calculated stress.5.") The cracks were small. the edge of the floor beam flange is the element with the descriptive marking "FB 28 S. 43. It originated at the end of a longitudinal fillet weld that was used to attach a 0. they should extend full length of the girder. Typical of these structures and cracking is the small crack shown in Fig. 43. In many cases. it is not appreciated that the welding of socalled secondary members or attachments must meet the same weld quality criteria and nondestructive test requirements as for main members. This phenomenon was discussed earlier: see Section 6. The use of low fatigue strength details such as coverplated beams and welded web and flange gusset plates should be avoided on bridge structures that will experience large numbers of stress cycles during their design life. and it was possible to retrofit these details by grinding and peening the weld terminations. There are several other Category E or E' details that are susceptible to fatigue cracking at weld terminations. If cover plates are used on high volume bridges.3.
If the welding of the stiffener to the web is not expected to meet high quality control and inspection requirements because it is a secondary attachment. A second crack was discovered later in the same girder. initial defects. 93 Figure 44 Crack in Girder Web from Splice in Longitudinal Stiffener . It grew both upward (to about the middepth of the girder) and downward. 44 is the fascia girder of a ninegirder fourspan bridge on an Interstate highway [20]. fatigue cracking can be the result. The bridge was approximately nine years old when the fatigue crack was discovered. The fracture initiated at an unfused groove weld that was used to join sections of the longitudinal stiffener. It had also started from an unfused groove weld in the longitudinal stiffener. The longitudinal stiffener fulfills its purpose without carrying significant stress. The girder web shown in Fig. An example of a detail this category is a splice in a longitudinal girder web stiffener. The lower end of the crack had penetrated the bottom flange of the girder. and so on directly influence the fatigue life of the main member.Figure 43 Fatigue Crack at Lateral Gusset Plate Welded to Floor Beam Web that stress concentrations.
If the transverse welds intersect the longitudinal welds. However. Lackoffusion often exists adjacent to the girder web in the transverse groove welds. The horizontal gusset plate used to connect diaphragm and lateral bracing members to a longitudinal girder is often slotted to fit around a transverse stiffener. Figure 45 shows the junction of the horizontal gusset plate (foreground) with girder web (righthand side of photo) and the vertical transverse stiffener.but had only severed the stiffener and not yet penetrated the web when it was discovered. This meant that there was a significant lack of fusion in a plane perpendicular to the main stress field in the girder. In the U. they provide a path for a crack to enter the girder web. Examination of the fracture surface showed that a brittle fracture occurred after the fatigue crack propagated into the web. A condition related to that just described occurs when backing bars are used to make a groove weld between transverse stiffeners and a lateral gusset plate. the groove weld did not penetrate to the backup bar. Minnesota [20]. The final condition of the girder was that the crack had completely severed the lower flange and had stopped 94 ..S. Paul. Figure 45 LackofFusion Defect Adjacent to Backing Bar at Gusset Stiffener Weld Cracks that develop in the web at lateral connection plates generally start at the location of intersecting welds. a path was provided into the girder web. particularly near the girder web. The connection between the horizontal gusset plate and the vertical stiffener was made using a groove weld founded on a backup bar. one of the first structures to exhibit this cracking was the Lafayette Street Bridge over the Mississippi River at St. Since this weld also intersected both the vertical weld attaching the stiffener to the web and the longitudinal weld attaching the gusset to the web.
Magnification (10x) is helpful. They are visually apparent because they differ in appearance from the protective oxide coating that has formed normally on the weathering steel. mill scale. Similarly. 7. some inspection tools are better for picking up certain types of fatigue cracks than are others. At weld terminations. magnetic particle inspection. As will be seen. It should also be kept in mind that the inspection usually takes place when the live load stresses are small. very small cracks are difficult to detect by visual inspection even on cleaned surfaces unless the crack depth is between 3 and 6 mm. as is supplementary lighting.2. rust) stains that develop after the paint film has cracked. Cracks smaller than this are not likely to be detected visually unless the paint..5 Methods for Inspection for Fatigue Cracking In Section 7. and use of acoustic emission techniques. Experience has shown that cracks have generally propagated in depth to between onequarter to onehalf the thickness of the part before the paint film is broken. 95 . oxides can be seen on unpainted weathering steel surfaces when cracks develop. A knowledge of where cracking is likely to occur in a detail that has been identified as prone to fatigue damage is likewise an important adjunct to the inspection process itself. and these preliminary steps will help to focus the inspection on the highrisk regions of a structure. it was stressed that it is desirable that any inspection for fatigue cracking be preceded (if possible) by an analysis of the structure and by identification of details prone to outofplane cracking. It is unusual to be able to inspect every part of a given structure for fatigue cracking: time and money do not permit this except in the case of specialized structures. Inspection is usually a difficult and expensive process. which permits the oxide to form. Often the most reliable sign of fatigue cracks is the oxide (i. This means that an existing crack will not be held open by the stress field. the goal is that the inspector be able to concentrate on those details that are candidates for cracking and thereby be able to direct resources to the locations where fatigue cracking is likely to occur. The next step is to identify the inspection methods that are both appropriate for the expected manifestation of cracking and are costeffective. and dirt have been removed by blastcleaning the suspect area. a visual inspection is essential. which itself was about 150 mm above the flange. ultrasonic inspection. The inspection methods or tools that are in common use today include visual inspection (sometimes at low power magnification).about 150 mm above the horizontal stiffener.e. Before cleaning the surfaces to be examined. Clearly.
once calibrated. by capillary attraction. called a developer (usually white). Acoustic emission. In addition. Clusters of particles form at crack tip locations. is sprayed on in order to draw. a second material. this method is reliable for detection of surface cracks emanating from smooth surfaces. relatively large cracks must be present for the system to be effective.The following inspection methods are useful: • Dye penetrant inspection. Equipment is simple and portable. Accuracy is reduced when used on weld surfaces. After cleaning and drying. When the developer is applied to the opposite face. The orientation and size of the crack are not always easy to visualize and. throughthickness cracks can be detected. When a magnetic field in a steel plate encounters a discontinuity. this method works most effectively on smooth surfaces. When used correctly. radiography can be used to identify the presence of defects and cracks. It is not particularly helpful when the actual sizes • • • • 96 . This method of assessment uses the high frequency sound wave that is generated by crack extension. Magnetic particle inspection. for cracks growing from weld toes. the penetrant fluid from cracks. Even subsurface cracks can be detected if the depth below the surface is not too great. Most available systems are not directly engineered for bridge monitoring. However. This is the most common and least expensive inspection method. Under controlled conditions. As with dye penetrant inspection. Xray Inspection. These waves are picked up by a receiver and a crack will cause a distortion in the waves received. high capillary fluid (usually a red dye) is used to penetrate into surface cracks. Variations in surface roughness affect accuracy. Fine magnetic particles are sprayed or dusted on the object in order to detect such distortions. Ultrasonic inspection. This method uses high frequency sound waves at a pitch beyond that audible to the human ear. use of this method requires considerable experience. Sensitive piezoelectric sensors convert the very high frequency vibrations (several hundred kilohertz) generated by a propagating fatigue cracks into electrical signals that are detected and analyzed. This method can be used for any thickness of plate over 3 mm and. however. but reasonable results can be obtained by an experienced inspector. it can detect small cracks whether they are embedded or not. A low viscosity. therefore. it is often difficult to distinguish between a crack and the plate/weld toe interface. the field becomes distorted.
However. Also.) Repair can be delayed. when repairing fatigue cracks. when monitoring crack growth at regular intervals is not possible. no immediate action is taken to repair the crack or replace the element. may result in an adequate solution to fatigue cracking provided that the steel quality is appropriate for both present and future service conditions. the actual size of the crack can be established and this can often be used to calibrate and interpret ultrasonic and radiographic results. • Coring. however. This permits a destructive assessment of the section within the cores. a highstrength bolt is inserted into the core opening and preloaded. or when further cracking may substantially increase repair costs.6 Repair of FatigueCracked Members Repairing a cracked element is an obvious option. Fracture mechanics analyses should be employed to predict crack growth and to determine critical crack lengths. thereby making it possible to avoid repairs at inconvenient times. i. Appropriate measures could also include implementation of damage tolerance strategies and element replacement. Its primary use has been to evaluate groove welds where lackoffusion may exist. often holes are drilled at the crack tip in order to minimize crack extension. Furthermore.) made from steel taken from the structure may be needed. (However. 7. In this way. this strategy often provides the opportunity to determine the causes of cracking more precisely. additional measures are recommended. Crack measurement equipment should be calibrated and verified on a regular basis. In some cases. Laboratory testing of specimens (Charpy Vnotch. When damage tolerance strategies are used. the opening should be ground smooth and then inspected with liquid penetrant to ensure that no crack extends beyond the hole. CT. However. an increased risk of derailment due to cracking of a railway bridge girder). After the core has been removed. crack lengths at which fracture would be expected.of the crack are desired. any modifications to the structure or changes to the applied loading should be assessed in order to verify that the structure can safely tolerate further cracking. In order to ensure that a damage tolerance strategy is successful. etc. when carried out correctly. Repair of cracks.e. when continued cracking would endanger human life through increased deformations (for example.. An effective way to establish the type of defect or crack is to core and remove a segment of the crack or discontinuity. it is important to keep in mind 97 . but other possibilities should also be considered. damage tolerance should not be employed when rupture of the element would cause general collapse of the structure.
and therefore weld repair should be employed as a last resort when no other solution is possible. stop outofplane movement. these cover plates should be placed on both sides of the cracked plate. Employ fatigue strength improvement methods to increase fatigue resistance. grind away weld reinforcement. The bolt should be pretensioned according to code provisions for bolted assemblies. fill the gouged area with weld metal. and they must be attached with pretensioned highstrength bolts. See reference [25] for guidance regarding appropriate hole diameters. etc. Use of a pretensioned bolt introduces a local compressive stress. unimproved location is unlikely. (Drilling a hole at the end of an existing crack changes the crack from sharp to blunt. The following is a list of measures that have been employed successfully for steel structures containing fatigue cracks. • • • • 98 . Care must be taken to ensure that the crack front is intercepted by the hole. or in some cases it may be done by drilling a hole at the end of the existing crack. Crack growth may be stopped by changes to the structure itself (reduce the stress range. • Place cover plates over the crack in order to provide a load path for forces and to restrict movement of the crack surfaces during fatigue loading. polish smooth and inspect for weld defects using ultrasonic and Xray inspection technology.). it is essential to ensure that the fatigue crack growth has stopped. which has the effect of masking the tensile stresses that drive the crack. Drill a hole at the end of the crack and fill the hole with a pretensioned highstrength bolt. Airarc gouge the crack. This measure should be accompanied by some modification that reduces the fatigue stress range in the region surrounding the crack.that repair welding is rarely successful. The latter procedure is discussed following. These measures are ordered (roughly) from best to worst. If this solution is adopted.) Cut out and refabricate parts of elements in order to reproduce the same conditions that existed at the crack site before cracking occurred. thereby greatly increasing the force required to drive the crack. Peening introduces compressive residual stresses and also changes the size of the weld toe crack. This measure is most effective for cases in which fatigue cracks have grown from weld toes and when cracking in another. Peening the toe of a weld termination that is perpendicular to the stress range is an effective way to prevent small cracks (less than 3 mm deep) from propagating. It effectively increases the fatigue resistance by a category. Preferably.
use of a larger element is often warranted. Although connections may not be cracked. the effects of this change in stiffness should be studied in order to ensure that problems are not created elsewhere. they may be damaged due to excessive movement. Do not reuse bolts that were used in the old element for connecting the new element. Work onsite should be avoided. an embedded crack sometimes resulted because the penetration did not follow the crack path. Experimental field trials demonstrated that it was difficult to execute while the bridge was in service because of vibration. Reuse old connection material only if it can be demonstrated that it has not been damaged by the incidence of cracking in the element. Such consequences are most likely when the local stiffness of the structure is modified. replacement of the cracked element may be warranted in those situations where it is possible to remove and replace the element. It also improves the fatigue resistance by about one category. for this reason. any measure that lowers the stress ranges in the area around the crack will contribute to the success of a repair. Care should be taken to ensure that cracking at other locations is not triggered as a result of these measures. especially if such activity does not contribute to reducing builtinstresses and eccentricities. Fabrication work on the new element should be performed where adequate quality of workmanship can be assured. After removal of the cracked element. 99 . thereby altering the sensitivity of the structure to dynamic loading. and when there is any doubt regarding the grade of steel used in the structure. When damage tolerance strategies are not appropriate and when no repair option can be implemented. Therefore. Replacement using a larger element than that originally present will result in lower stresses at the location where the crack was detected. site work should be limited to actions such as drilling undersized holes to fit and other joining techniques that are intended to ensure good fit and satisfactory alignment. The procedure seems best suited for shop use and new construction. Consequently. all connections should be inspected for cracking. a larger element alters the stiffness of the structure. employ a steel grade that is more suited to the situation. In the presence of corrosive environments. Also. and.• Gas tungsten arc remelting has been demonstrated to effectively remove the microdiscontinuities at the weld toe and decrease the stress concentration. Finally. However.
Measures that lower the stress ranges in the structure are the most effective way of preventing further crack growth. The following general precautions are recommended for the structure as a whole: • Identify areas in the structure where cracking would create a dangerous condition. 100 . addition of mass. Care should be taken to ensure that susceptibility to fatigue cracking is not increased elsewhere as a result of these measures. development of fatigue cracks is probable in the future. Holes not only arrest cracks.7 Avoiding Future Cracking Problems If fatigue cracking has been discovered at one location in a structure. Such consequences are most probable when the local stiffness of the structure is modified.g. Consequently. Therefore. Mark these areas for modification and inspection. Use the same criteria as those used to judge whether further cracking can be tolerated for implementation of a damage tolerance strategy. through modifying areas sensitive to electrolytic action. finally. thereby altering the response of the structure to vibration and other loading. or a change in restraint use of fatiguestrength improvement methods at critical locations (e. Effective measures for avoiding future cracking problems usually fall into one of the following categories: • • • • • reduction of applied stress ranges reduction of dynamic magnification (impact) reduction of number of cycles of damaging stress ranges decrease in vibration through damping. if nothing is done to improve conditions at those places. it is likely to be present at other similar locations. gas tungsten arc remelting at weld toes. but they can be used to improve problem details such as flanges that penetrate webs. peening) holes drilled in webs will often be effective in prevention of future damage.. such locations should be identified and measures for improvement of their fatigue resistance should be considered.7. for deicing salts) and. by providing adequate drainage (for example. not allowing water and debris to become trapped on the structure. • Corrosioninduced cracking can be avoided by protecting the structure from the environment.
This involves measures such as keeping the structure clean. Inspect repair locations regularly in order to verify the effectiveness of repair measures. and designing modifications considering inservice inspection. Failsafe means that the structure can tolerate failure of one or more elements before a catastrophic condition arises. providing safe access to critical locations. This condition must be preceded by a period that is long enough to identify cracking and to take appropriate action. Create conditions whereby the structure can be inspected as easily as possible. • • • 101 . reconsider decisions made regarding the management of the structure.• If new information is received regarding the structure or if structural modifications are significant. All modifications should be designed with the goal of making the structure as failsafe as possible.
There are tens. those which are lap or butt splices ("shear splices") and those which are tensiontype connections. and the crack 102 .Chapter 8 Special Topics This chapter contains a number of topics that will be of interest in specific cases. which means that cracking takes place ahead of the first (or last) bolt hole in a line. of thousands of existing riveted structures still in use in North America. It has always been common practice in bridge construction to use preloaded bolts. minute slip takes place in this location as load is applied repeatedly. except perhaps for riveted members.1 Bolted or Riveted Members Fatigue characteristics of bolted connections.1 Bolted Members Highstrength bolted joints can be subdivided into two categories. and their remaining fatigue life and safety is a topic of great importance to the owners and to regulatory authorities. In the former case. In the case where the load transfer is by friction. consequently. 8. although in new construction most specifications require that the bolts be preloaded if fatigue loading is likely. fatigue under combined stresses. completely by bearing of the bolts against the connected material (nonpreloaded bolts). The need for information on all of these topics is selfevident. The fatigue strength of a bolted shear splice is directly influenced by the type of load transfer in the connection. Rivets have not been used in new construction for a considerable period of time. Cracks are initiated and grow in this region. threaded fasteners. or by some combination of these two mechanisms. and threaded rods are discussed in this section. the differential strain between the two components is highest and. Here. particularly on the faying surfaces near the extremities of the joint.1. 8. the influence of environmental effects upon fatigue life. perhaps hundreds. anchor bolts. and comments on the use of fracture mechanics in fatigue strength evaluation. The topics include fatigue strength of bolted or riveted members. but an understanding of the fatigue life characteristics of riveted connections is needed because of the necessity for evaluating the remaining fatigue life of existing structures. the bolts can be either preloaded or not preloaded. the role of residual stresses in fatigue crack growth. riveted connections. This load transfer can be completely by friction at the interface of the connected parts (preloaded bolts). fretting of the connected parts occurs. Most of these are bridges.
progresses through the gross crosssection of the component. Some point at the edge of the hole or within the barrel of the hole is the initiation site for the fatigue crack and growth is through the net crosssection of the connected part. Only the connected material is susceptible to fatigue cracking. both types have been observed within the same test. The phenomenon is often referred to as "fretting fatigue. the level of residual stress is negligible Figure 46 Fatigue Resistance of High Strength Bolted Shear Splices 103 . It is worth noting that there is no history of fatigue failure of highstrength bolts themselves in shear splices. However. it is prudent that the designer check both possible types of failure. The data come from a wide variety of sources and test configurations. As compared with welded details." If the bolts are not preloaded. Both types of fatigue crack behavior have been observed in laboratory tests and. the load transfer is by shear in the fasteners and an equilibrating bearing force in the connected parts. The local tensile stress in the region of the connected part adjacent to the hole is high. When preloaded bolts are used. and this explains some of the dispersion. If nonpreloaded bolts are used. it is highly unlikely that fretting fatigue will occur. however. in a few cases. it can be observed that there is a great deal of scatter in the results. and this is now the location where fatigue cracks can start and grow. it is likely that most of the scatter is due to two more fundamental reasons. Figure 46 shows the test results that were used to develop the design rule given in the AASHTO Specification. First.
whereas in welded joints it is very high. i. if any. the data were plotted using the stress range on the net crosssection or the stress range on the gross crosssection.e. but it will not be treated fully. Connections that place the bolts in tension are infrequent in bridge construction. It seems prudent to use slipcritical connections throughout a bridge structure.3−1). 104 . It should also be noted that there are very few data in the longlife region. albeit a conservative choice for many of the data.1.1.13.2. Some features of the behavior of this type of connection will be reviewed here. preloaded bolts are desirable so that there is no possibility that nuts will vibrate loose.. Moreover. In Fig. A closer evaluation of the data would be warranted only in the case of fatigue life evaluation of an existing structure.e. fatigue failures of highstrength bolted tensiontype connections have occurred from time to time. since the majority of the joints within the bridge will have to use preloaded bolts in any case. The AASHTO Specification prescribes Detail Category B for bolted shear splices (AASHTO Table 6. preloaded bolts are used) and at the net crosssection when the connections are "highstrength bolted nonslipcritical. for example. It represents a reasonable lower bound. 46. use of preloaded bolts throughout the bridge gives a consistency to the installation process.6. depending upon whether or not the connection had slipped into bearing. Also shown in the figure is the AASHTO Detail Category B fatigue life line. and requires that stresses be calculated in the gross crosssection of the base metal in slipcritical connections (i. Detail Category B will be used and stresses will be calculated in the gross crosssection. Although there are few." Article 6.in mechanically fastened joints. Reference [35] provides a more complete treatment of the subject than can be covered here. This means that any portion of a stress range that is compressive will not be as damaging as are the tensile portions. the design of bolted shear splices will not be at any significant disadvantage because of this conservatism. whether or not a particular joint is subjected to fatigue.2. the severity of initial flaws in a mechanically fastened connection is much less than that in a typical welded joint. but they may occur in bents and frames.1 requires that joints subjected to fatigue loading must be designed as slipcritical. Second. This is consistent with the fatigue crack mechanisms discussed above.. reported fatigue failures of highstrength bolted shear splices. Since Category B is already a superior fatigue life category. Since the structure will be dynamically loaded.
In addition to reflecting the features already mentioned here.2. In this case.5–1 and 6. The threaded portion of the bolt provides the crack initiation location. as illustrated in the sketch. However.1. At one extreme. and this places an additional force in the bolt. On the other hand. and this also can have an effect on the fatigue life of the bolt.Typical of the configurations that can place bolts into tension is the one illustrated in Fig.6. Figure 47 Bolts in Tension The stress range experienced by the bolt as the assembly undergoes repeated loading is significantly affected by the level of bolt preload [35].3. the cases of ASTM A325 and A490 bolts in tension are not set out as separate Detail Categories. The AASHTO Specification requirements for bolts in tensiontype connections follow the same general pattern as that for other details. Other information concerning fatigue of bolts in tension is given in AASHTO Article 6. The amount of the prying force is dependent upon the flexibility of the connection. A significant feature of the connection is that prying forces develop. properly preloaded bolts in a very stiff connection will undergo little or no stress range. Recommendations for the fatigue design of highstrength bolts in tensiontype connections can be found in References 35 and 36. Instead. An additional complication occurs if the applied load is high enough to produce yielding in the fasteners. where. which as a rule Bolt force Prying force is at the root of a thread.1. then the stress range in the bolt threads will be large. the recommendations distinguish between ASTM A325 bolts and ASTM A490 bolts.6. 47. bolt bending is present. it has been shown that the stress range increases with each cycle [35]. This is simply a convenience that can be employed because the ratio between the area through the threads and that 105 .5–3.13. the necessary information for calculating the fatigue life of a highstrength bolt in tension is simply listed in AASHTO Tables 6. This situation arises when a tee is used to connect a tension hanger to a beam lower flange.2. The latter are somewhat less ductile than the former. it is noted that the bolt prying force must not exceed 60% of the nominal force in the bolt. and the bolt preload is low.2. for example. It is also pointed out that the stress range is to be calculated used the area of the bolt corresponding to the nominal diameter. if the connection is relatively flexible. among other things.10. 17 and 19 of this document. These tables provide the constant A and the constant amplitude fatigue stress for use in Eq. thereby increasing the nominal value (total external force divided Applied force by the number of bolts). The same flexibility introduces bending into the bolt.
usually at a relatively low stress range. that is. A large number of test results are available [37] and they are shown in Fig.1.e. 48. Some specifications have even lower limits. a nut was present on each side of a simulated base plate. (As is customary. not tension. use ASTM A325 bolts rather than A490 bolts. • • • 8. fatigue loading will be present. Design the connection so that prying forces are minimized. Most of the tests represented in this figure were carried out at constant amplitude loading.2 Threaded Rods Threaded rods are used frequently in mast or sign base plates. although the AASHTO Specification also uses the 60% value. redesign the connection so that the bolts are in shear. Ensure that proper installation procedures are adhered to so that the prescribed bolt preloads will be attained. a short section of the rod is prestressed. and this can produce a large number of stress cycles. The effect of a small amount of misalignment was also included in some tests. If highstrength bolts that are not preloaded act in direct axial tension. In this way. but a few tests were done under variable amplitude. the stress range was calculated using the socalled "stress area" of the rod.corresponding to the nominal diameter of the bolt is relatively constant for the usual bolt sizes.. Under the action of the wind on the mast or sign. 106 . The results include axially loaded rods (i. they are equivalent to a threaded rod. The fatigue design of highstrength bolts that are in tensiontype connections should reflect the following guidelines: • Whenever possible. and rods preloaded by turning the nut on 1/3 turn. Reference [36] suggests that the prying force should be less than 60% of the externally applied force.) Many of the tests were done in a doublenut configuration. Whenever possible. rods preloaded to the "snugtight" level. no preload). but most of it is not.
A small increase in fatigue life was observed for fully preloaded rods. Dev. The constant amplitude fatigue limit for Category E' (17. this is a reasonable representation of the data in the region up to about 1 million cycles. and bolt diameter did not in themselves have a major effect on the fatigue life [37]. 37) There is considerable scatter in the results.0 MPa).2 Std.1000 Frank Dusel Dexter (concentric) Dexter (eccentric) Stress Range. Unfortunately. Until more variable stress range data are available. method of thread forming. The fatigue life of rods tested with no preload and rods tested in the snugtight condition was similar. the region of low stress range and a large number of cycles is reflective of where many of the structures that use threaded rods will have to function. As can be seen. thread size. MPa Mean Mean .9 MPa) clearly is too conservative for these (mostly) constant amplitude tests: a better choice is the Category E constant amplitude fatigue limit (31. but it clear that the threaded rod is an element with poor fatigue resistance. could have a large number of cycles at low stress range under wind loading. Detail Category E' 100 10 10000 100000 1000000 Number of Cycles 10000000 100000000 Figure 48 Fatigue Life of Anchor Bolts (Ref. it would be 107 . assuming that beveled washers are used. Figure 48 also shows AASHTO Detail Category E'. The misalignment used in the tests (1:40) likewise did not significantly reduce the fatigue life. The results do indicate that grade of steel. for example. A sign baseplate.
as long as the hole sizes and plate thicknesses commonly used in structural practice pertain.1. the fatigue life can be expected to be a reflection of such features as the size of the hole relative to the part. Figure 49 shows the data: identification of the specific sources from which the test data came can be obtained in Reference [38]. bearing condition. and most of these were members taken from service. (It is not yet clear whether this is justified since the fatigue cracks were observed to grow at right angles to the crosssection. is not likely to be strong. the experimental evidence is that fatigue cracking in riveted shear splices takes place in the connected material. or subpunched and reamed). the method of hole forming (drilled. If a conventional fatigue life examination indicates that a given riveted structure has long remaining fatigue life. Thus. For those cases where members taken from service were tested. For the tension members.prudent to use Detail Category E' throughout. In the case of bending members. At the present time. Thus. or at least large size. then obviously no further effort is necessary. However. per se. For the time being. it is necessary that the best possible information be used. The AASHTO Specification does not address the fatigue life of anchor bolts (threaded rods). the stress range was calculated on the net crosssection. the influences of clamping force. Most of the data come from tests of flexural members. there is a 108 . 8. punched. the best data available are tests on riveted connections of proportions that are consistent with usual structural practice and are of full size. The influence of the hole size. For this reason. when remaining fatigue life is shown to be inadequate. bearing condition. it is to be expected that the scatter of data will be relatively large. not in the rivet itself. The alternative might be replacement of the structurea costly solution. the reflections of clamping force. and because the "defect" presented by a riveted connection is not severe. and the method of hole formation have not been examined in any systematic way.) As was the case with bolted connections. A few of the test results are from tension members. the moment of inertia of the crosssection included the effect of holes. the bearing condition of the rivet with respect to the hole.3 Riveted Connections It has already been explained that the need for fatigue life rules for riveted connections rests in evaluation of existing structures. the previous stress history was examined and deemed to have been nondamaging. and the clamping force provided by the rivet. and hole formation must simply be part of the data pool. As was the case for highstrength bolted shear splices.
rather than the value of 48. Category D is a reasonable reflection of the data except for two test results at long fatigue lives. 49 come from members made of mild steel. 109 . 49. Almost all the data shown in Fig.3 MPa that is used for Category D. Figure 50 shows a fatigue crack passing through the net section of a riveted member that was tested in tension.Figure 49 Fatigue of Riveted Connections considerable amount of scatter in the data. The issue is discussed further in Reference [38]. but results from some wrought iron members are also present. but it is too conservative in the region of the "knee" of the bilinear fatigue life representation. It has been suggested that a fatigue limit of 40 MPa should be used. and this is shown along with the data in Fig. This change would capture the long life data points better. The AASHTO Specification stipulates the use of Detail Category D for riveted connections. These were variable amplitude tests. for the reasons described earlier.
. It is not uncommon to see depressions in the order of 25 to 50 mm diameter. For example. a value judgment can be made as to how much to degrade the fatigue life as compared with a noncorroded member. corrosive products (e. Since this type of corrosion effect will generally be an issue in the case of evaluation of old bridges.Figure 50 Fatigue Crack Through Net Section of a Tension Member 8. Fisher et al. reported that if the corrosion reduced the thickness by more than onehalf. Old railway bridge floor system members sometimes show this type of corrosion.g.8. 110 . (Ice and salt were formerly used as the refrigerant in railway rolling stock). they are similar to the introduction of a hole: as such. fatigue cracking was likely to start at a notch within the corroded region rather than at the location of a net section (reduced by corrosion) through rivet or bolt holes. For instance. they do not introduce a particularly severe flaw into the member.2 Environmental Effects. In extreme cases. bracing members. although the number of test data in which the corrosion effect was specifically examined is quite limited The ratio of corroded part to original thickness ranged from 0.2 to 0. brine) can drip down from an upper level to a lower level and produce both a local reduction in crosssection and a notch effect. Although these are not desirable situations. with depths in the order of 3 to 8 mm.g.. Use of Weathering Steel The effects of the environment upon the fatigue life of steel members can show up in one of several ways. Corrosion produces local effects that cause stress concentrations. lacing bars). the local corrosion may completely penetrate the member. Fisher et al. have reported on the fatigue life of a number of corrosionnotched beams [39]. the top surface of the lower flange of a girder can often show deterioration of the crosssection as the result of brine that has dripped down from the upper levels. particularly with initially thin members (e.
Figure 51 shows the lower flange of a railroad stringer in which a fatigue crack has occurred at a corrosion notch. have not demonstrated any significant problems with either stresscorrosion cracking or corrosion fatigue. the designer should be alert to the presence of notches in this case also. usually in a saltladen atmosphere. do not indicate any significant effect. more or less in a uniform way. general experience with bridge steels and weldments. the useful life of a welded bridge structure can be affected by an aggressive environment when it is under sustained load (stresscorrosion cracking). The effect of a corrosive environment upon crack growth or as it increases stress concentrations has not received much attention. In the case of a general reduction in crosssection. particularly in civil engineering applications. the fatigue strength examination of a corroded member can proceed on the basis of the usual stress range vs. However. the fatigue life evaluation can be based on an estimate of the reduced crosssection. such as high humidity. including successful use in longtime environmental service applications. The origin of the corrosion is these cases is a more general source. however [39]. experience in the field does not support the 111 . or by the enhancement of an aggressive environment on the effects of cyclic stresses (corrosion fatigue).There are also instances in which the entire crosssection has been corroded. Of course. In other words. often done at relatively high test rates. Moreover. number of cycles relationships. Figure 51 Fatigue Crack at a Corrosion Notch in a Railroad Stringer In principle. The available laboratory test results. or corrosive fumes in proximity to chemical process plants. given that the stress range will be calculated using the properties of the reduced section.
need for inclusion of special rules for bridges or similar structures that are operating in the types of corrosive environments normally associated with that type of structure. In two specific cases, the consequences of operating in a corrosive environment are included explicitly in the design rules. Both of these are clearcut—offshore structures and the drillpipe used in oil exploration. In both instances, the environment is severe—salt water— and its consequences cannot be avoided. There is not much test data available here either, however, and it must be recognized that corrosion and its effect on fatigue life is timedependent. Thus, testing in this area takes even longer than it does for the normal situation of testing in air. Designers of civil engineering structures in particularly harsh environments should take special care that detailing of the structure minimizes the possibility of locations for fatigue and that a detailed maintenance strategy is established. In all cases, detailing practice should be to promote drainage of water and to prevent of entrapment of moisture and corrosion products. Crack growth rate data has been acquired on most commonly used bridge steels under constant amplitude and variable amplitude loading [40, 41]. The scatter bands of the test specimens in air generally bracketed the test data acquired in an aqueous environment. The latter was done with or without a 3 percent solution of sodium chloride. Most of the aqueous data fell between the mean and upper limit of the data acquired in air. However, all of the test data was acquired at relative high levels of ∆K and crack growth rate. Figure 52 shows the
Figure 52 Crack Growth in Aqueous Environment
112
test data for A588 steel in an aqueous environment with and without the sodium chloride solution. This level of ∆K is only reached after very large cracks have formed. The crack growth threshold appears to increase with decreasing load frequency when subjected to a hostile environment. The general conclusion is that the crack propagation life of bridge steel components under actual operating conditions (wet and dry) in aqueous environments is equal to or greater than that obtained in a room temperature air environment. The laboratory studies which led to the conclusion that environmental conditions have little effect upon the fatigue behavior of bridge components are supported by field experience. The influence of the environment upon rusting, pitting, and the start of corrosionfatigue cracking (assuming no prior fatigue crack) may differ from its influence on progressive crack extension when the crack crevice controls the crack tip behavior. As would be expected, rusting and pitting introduces notches and this can affect the crack initiation behavior of the base metal. Studies on weathering steel have demonstrated that only the Detail Category A situation is significantly affected, however. Figure 53 shows the experimental results of cruciform joints, which are a Category C detail, in their aswelded and weathered conditions. It is apparent that the use of weathering steel does not in itself produce a fatigue life that is inferior to that when ordinary grades of steel are used.
Figure 53 Comparison of Weathered and Unweathered Cruciform Joints
Many civil engineering structures must operate in low temperature environments. Again, there are very few test data in this area. The work of Fisher et al. [39] did include
113
some tests in which a portion of the fatigue crack growth was done under very low temperatures (–40°C and −73°C). The study showed that the low temperature did not result in sudden fracture until the cracks had become very large. In summary, it can be concluded that the civil engineering structures most likely to have to endure corrosive environments, especially bridges, can be designed in accordance with the rules that have been developed on the basis of tests in air. Evaluation of existing structures in which corrosion has affected the crosssection can be done by applying judgment and using the results of those tests that are available [39]. In the case of new structures, great care should be taken to ensure that corrosion products from roadway decks are directed away from the members and that road spray from traffic either above or below the bridge is cleaned up periodically. Detailing should be done with a view to minimizing those locations where corrosion products can form and sit, especially in crevices.
8.3 Combined Stresses
Throughout the discussion so far, it has been emphasized that the stresses to be used in the fatigue life evaluation are those corresponding to the nominal stress as obtained from a strength of materials level of analysis. The detail classification stipulated by the governing specification includes the effect of local stress concentrations due to weld shape, discontinuities, triaxial conditions, and so on. Only in exceptional cases, around large openings, for example, will a more sophisticated analysis be required. Occasionally, also, it might be necessary to investigate the combined effect of normal and shear stresses. When normal and shear stresses are present at the same location but do not occur simultaneously under a given loading event, the individual components of damage can be added according to a Miner's summation, as follows:
∆σ e ∆σ r
3
∆τ + e ∆τ r
3
≤ 1
(20)
where σ and τ refer to normal stresses and shear stresses, respectively. The subscript e means a calculated equivalent stress range (see Eq. 13, for example) and the subscript r refers to the permissible stress range for the detail. For a case in which normal and shearing stresses are present in significant quantities at a given location and are concurrently present and in phase at that location during a given loading event, the principal stresses should be calculated. This stress range is then used in the
114
it has been shown that the fatigue resistance need not consider the principal stress [19]. 20 should be focused on this case. 43.4 Effect of Size on Fatigue Life Most of the test results upon which the fatigue life rules are based were done on specimens where the component parts were in the order of 12 to 25 mm thick. The main effect is to significantly reduce the effects of the applied stress levels. Rather.5 Role of Residual Stress Steel structures that are fabricated by welding contain "residual" or "lockedin" stresses that have been introduced as a result of the welding process. The AASHTO Specification recommends that Detail Category E be used for the case of shear stress on a fillet weld throat.fatigue life evaluation. but they show that the general relationship between stress range and number of cycles still exists. Likewise. Hence.) When the principal stress changes direction during the stress cycle. this is not the main reason for the decrease. In unusual situations. particularly in the case of flanges of welded builtup beams. it should only be necessary to consider the principal stress when its direction does not change during the stress cycle. (Normal stress is defined as the stress perpendicular to the direction of the potential fatigue crack. for example. it has to do with more severe residual stresses that result when using thicker plates and from the observation that the stress intensity factors at the weld toe increase as plate thickness increases (and the initial flaw size remains constant) [45]. For example. as it does at transverse stiffeners in plate girder webs. for instance. 8. It is generally agreed [42. There are very few test results. for standard weldable 115 . the same decrease in fatigue detail category occurs when the plate thickness exceeds 25 mm in a longitudinally loaded groove or fillet welded detail. however. the fatigue category for the base metal at the end of partial length cover plates is reduced from Detail Category E to E' (case of plate wider than flange) when the plate thickness exceeds 20 mm. Generally. Fabricated steel structures often use plate thickness considerably greater than this. 44] that the fatigue resistance of thicker elements is less than that of thinner elements. shear stress alone can be significant. This would apply to the attachment of flange plates to girder web. There are several locations within the AASHTO Specification where it is required that fatigue strength be decreased with an increase in plate thickness. per se. Although a thicker element will have a greater statistical chance of containing flaws. These have considerable influence on the propagation of fatigue cracks. use of Eq. 8.
the geometry of the connected parts. Furthermore. very large residual tensile stresses are present at the junction of the flange and web of a beam that has been built up by welding the component parts together [18]. code developers have taken advantage of such opportunities and. fatigue design guidelines have been greatly simplified and harmonized internationally. that the rolled shapes or builtup members used in structural applications also contain regions of high residual tensile stress. which means that the flaw is under a condition of initial stress even before load is applied. 54. of course. pattern However. thereby placing it into compression. As the weld cools. The actual distribution and magnitude of the residual stress pattern depends upon such factors as the strength of the steel and the weld metal. For the usual condition wherein this initial stress is at or near the yield stress level. the stress ratio (ratio of maximum stress to minimum 116 . independence of steel grade justifies the use of the relatively large data base of laboratory results taken from tests of different steel grades produced in different countries. rather than the maximum applied stress. is that the magnitude of the tensile residual stress can reach the yield strength of the material. the equilibrium condition of the crosssection must be reflected in the balance between residual tensile stress and the area over which it acts and the residual compressive stress and its associated area. this means that stress range. however. Consider a weld laid down as shown in residual stress Fig. The important fact. the plate restrains the weld during the cooling and contraction process. During recent modifications to fatigue codes. it tries to contract.structural steels. this has brought about a return to the simple stress range vs. tendency for Conversely. Since there are no external forces applied during this process. It follows. For example. since the plate and the weld must maintain compatibility of length. For modern codes. This junction is also the location of the flaws that are likely to be the source of fatigue crack growth. cycle life model for fatigue strength suggested by Wöhler over one hundred years ago. consequently. and the size of the weld relative to the connected parts. The stresses set up in the plate and in the weld as a consequence of this Figure 54 Residual Stresses in a Welded Plate process are called residual stresses. This puts the weld and a relatively small volume of plate adjacent to the weld into tension. the main portion of the plate is being deformation pulled down by the contracting weld.
Since the yield stress is not exceeded. however. Calculate the actual range of stress that will exist for a flaw at the center of the plate (cf. and we note (by inspection) that the condition P = ∫ σ dA = 0 is met. Solution: 1. that is. This will be illustrated by the following numerical example. is the governing condition describing fatigue crack growth. EXAMPLE 13 Figure 55 shows a rectangular plate and its assumed residual stress pattern. the corresponding strains can be calculated from the elastic relationship. σ = E ε . junction of the web and flange of a welded beam). It shows that the values of the initial stress are as high as the yield stress. Compare this range of stress with the stress range that will be computed by the designer. the plate will be subjected to a maximum tensile stress of 215 MPa and a minimum compressive stress of 122 MPa.. + 215 − ( − 122) = 337 MPa .50x10 –3 –1. There is no force on the member. 55 (b). Initial conditions—The pattern of residual stress is shown in Fig.) σ σy = 300 E = 200 000 MPa ε y = 1.stress).50x10 –3 125 mm weld (b) Residual Strain Figure 55 Initial Stress Condition in a Welded Plate 117 . and these strains are shown in Fig. or some other parameter of applied stress.50x10 25 mm –3 ε – 300 MPa + 300 MPa (a) Residual Stress 1. The plate has a yield strength of 300 MPa. (This can verified by calculation. 55 (a).) According to the designer. (This could be an idealization of the flange of a welded beam.
. P = σ A = 215 N/mm2 x (25 x 125) mm2 = 672 x 103 N = 672 kN During application of this force.e. not all parts of the crosssection respond in the same way. That the integral is satisfied can be confirmed by first calculating the dimensions of the stress diagram. 55 (b) repeated). i.90x10 –3 300 MPa 0 0 (a) (c) (d) Figure 56 Superimposed Strain (Step 2) and Corresponding Stress The value of the uniform strain corresponding to imposition of the applied force. the relationship ε = σ E will not be valid for those regions of the resultant strain diagram where the strain is greater than the yield strain. By trial. and then calculating the force volume obtained when the stress diagram is superimposed on the crosssection.50x10 –3 + 0 0 = (b) 2. part (b) shows the strain imposed by the tensile force. Specifically. strain of 170x10 −3 . which is the equivalent of the imposition of a uniform compressive strain over the residual strain of Fig. Thus. Apply tensile force such that the design stress level is +215 MPa. be applied. 14x10 −3 . This requires that the 672 kN force be removed and an additional force of (− 122N / mm 2 )x( 25x125) mm 2 = −381x103 N ∫ σ dA = P = 381 kN . and the corresponding stress diagram is shown in Fig.2. part (a) shows the residual strain (which is simply Fig. 55 (b). this has been determined to be a 118 .10x10 –3 –20 MPa εy 1.50x10 –3 –3 – 0. 56 (d). 56. apply a force such that the nominal stress in the crosssection goes to 122 MPa compression. As before. part (c) is the resultant strain. 3. the change in force is ∆P = ( 672 + 381) = 1053 kN . Now. . Fig. 56 (d). a uniform strain must be applied of sufficient magnitude such that .40 x10 – 1. In Fig. 1. was determined by trial so to satisfy the requirement that ∫ σ dA = 672 000 N .
Thus. Upon initial loading to 672 kN force. 58.90x10 –3 –3 –1. the stress σ = 300 y initially was 300 MPa. the actual i. even though the strains increased as shown in Fig. The small zone of weld and plate that is at yield is limited 119 . a total change of 340 MPa. 58. 337 MPa.– 0. 1.50 unloading that occurred in Step 3 – 40 final stress condition took the stresses down to 40 MPa strain applied in step 3 compression. examine the stresses in the initial (no load) stress σ MPa after application middepth fibre of the crosssection. of 672 kN force As shown in Fig. Finally. Figure 57 (b) is the uniform compressive strain that must be applied to satisfy the requirement that a force of 381 kN be imposed.10x10 2. it was the yield value. The 0 ε x10 –3 2.e.70 change in stress at the middepth fibre was from +300 MPa down to Figure 58 Strains and Stresses at MidDepth Fibre –40 MPa. 57 (c). that is.. in fact. The resultant strains are shown in Fig.80x10 –3 – 300 MPa + (a) (b) = 1.90 1. 340 MPa is equal (nearly) to the change of stress calculated by the designer. The reader can verify the stress values shown in Fig. the stress remained Unloading at this value. 57 (d) and that the imposition of these stresses on the crosssection does. produce the value 381 kN. it should be borne in mind that unrestricted plastic flow of the crosssection cannot occur. The difference between the two values is attributable to roundoff errors in the calculations. (Still with reference to Fig. The result is that the actual change of stress. 4.20x10 –3 – 40 MPa (c) εy (d) Figure 57 Superimposed Strain (Step 3) and Corresponding Stress Figure 57 (a) shows the strains that existed at the end of step 2. 58.70x10 –3 – 1.
(This was discussed in Section 5. even if fatigue cracks start within tensile residual stress fields.) However. stressrelieved welded elements.) To summarize.3. they recommend that 100% of the stress range be used for all elements and all cases. It can be observed that there are no known cases of loadinduced fatigue cracking in bridge structures when the cyclic stresses are totally in compression [20]. As a result. When the applied stress ranges are compressive. the defect in a high residual tensile stress region is cycled to a limited strain. Authors of the AASHTO provisions argue that. it is possible to ignore all fully compressive stress ranges. or partly compressive.to some finite value of strain because he material adjacent to it remains elastic. The ECCS recommendations [15] are the most conservative. when welded elements are assessed. The developers of the EC 3 120 . However. regardless of whether it is completely or partially compressive. Eurocode 3 (EC 3) [46] allows a reduction of the compressive stress range to 60% of its magnitude for nonwelded and stress relieved elements. there has been a great deal of discussion regarding the best way to treat the following cases: • • • • the influence of compressive applied stresses the fatigue strength of nonwelded elements the fatigue strength of stressrelieved welded elements the difference between results taken from testing small specimens and those taken from testing fullscale elements. Nearly all fatigue guidelines for steel structures recommend that nonwelded elements.2. under fully compressive loading cracks will eventually stop when they propagate beyond the influence of these stresses. if a part of the stress range is tensile. recommendations vary. Hence. European recommendations suggest that the whole stress range should be employed. and welded elements be treated in the same manner when the applied stress ranges are tensile. A selection of approaches to these cases is described next. According to the AASHTO recommendations [2]. in large welded structures there are very high tensile residual stresses near fatigue crack sites and their presence significantly reduces the effects of applied stress levels and steel grade upon crack propagation for standard weldable structural steels. it is generally agreed that stress range is the only dominant stress parameter. then the whole stress range must be considered. Some of the differences between various codes can be traced to different interpretations of the effect of residual stresses (and other builtin stresses) upon the fatigue strength of structural elements.
they cannot be used to predict the remaining life of a cracked structure. Therefore. Small test specimens may not contain the same high level of tensile residual stress found in fullscale elements. 8. some test elements have failed due to cracks growing under fully compressive applied loading. the presence of tensile residual stress near potential crack sites in an element reduces its fatigue strength. Builtin stresses do not only originate from welding related residual stresses. For the ECCS document. 121 . and so on. Whenever possible. finally. Compared with elements that have no tensile residual stresses. especially at long lives.document note that a crack growing from an aswelded joint in an element may be dangerous when applied stress ranges are full compressive. and inefficient expansion joints. most recent research programs have concentrated upon testing of fullscale elements in order to establish code recommendations. but they may also be caused by other effects such as lack of fit. Fracture mechanics provides the central analytical model employed in fatigue assessments for dealing with these aspects. the EC 3 provisions reflect an attempt to accommodate those cases where there are no high tensile residual stresses. they provide little help in fixing inspection intervals. Therefore. they are not able to evaluate the influence of inspection precision upon fatigue reliability. Indeed. these curves cannot evaluate the influence of an unusually large defect. and. it was decided to provide conservative recommendations because in most complex structures the magnitudes of builtin stresses are unknown. it was decided that no advantage should be given to cases where applied stresses are in compression or where elements contain no tensile residual stresses. results taken from tests of small specimens should be employed with caution for design assessments. Also. temperature gradients. in these circumstances it is desirable that the test specimen have the same dimensions as the element under consideration for the real structure. The final point in this section relates to the physical testing of fatigue specimens. However.6 Quantitative Design Using Fracture Mechanics The fatigue strength curves described in Chapter 3 are not capable of providing an engineer with information necessary for several special cases related to fatigue assessments. For example. settlement of supports. small specimens are useful when exploring variables such as improvement procedures. For this reason. use of new materials. Test programs are sometimes used to obtain a more accurate fatigue strength relationship than is provided in fatigue codes.
n. A similar approach can be used for the design of structures when fabrication quality cannot be guaranteed and when it is known that inspection technology is unable to detect defects that are more severe than those normally known to occur in steel structures. usually numerically using equal increments of cycles. Quality assurance provisions in many codes are based more upon accepted standard practice. Such analysis has been used to determine inspection criteria for steel structures in the North Sea. either the design is modified or inspection technology that enables greater precision is justified. in order to determine cycle life. influence the performance of the structure. 7. Equation 7 is then integrated. Reference 47 provides guidelines for appropriate values of W and Y for use in Eq. The number of inspections should be determined using reliability analyses that take into account crackdetection probabilities and the consequences of failure. Section 2. 3 and for additional criteria. This approach is sometimes called a "fitness for purpose assessment. and the Figure 59 Crack Growth Curve critical crack size. When the cycle life resulting from the integration is less than the number of cycles applied over the life of the structure.3 is normally taken to be equivalent to the size of the defect. the defect may be regarded as being acceptable. The economic consequences of remedial measures are often severe and resulting delays to a project can be long. acr . a set number of inspections. and detection capability than upon scientific accuracy. More frequent inspections are 122 . experience. as shown schematically in a 0 time (years) Fig. for example. determination of intervals ∆a = n between inservice inspections does not acr n = number of ∆a inspections account for the behavior of a potentially crack ∆a size ∆a cracked structure. 7. in reality. it is assumed that defects are present and the minimum detectable defect size becomes the initial crack size for use in Eq. If the cycle life is greater than the number of cycles expected over the life of the structure. ∆a." a cr a 0 a Often. In such analyses. should be performed. such as those which determine whether two closely spaced defects should be considered to be equivalent to one large initial crack. Fracture mechanics offers a more rational means of defect assessment. the value for the initial crack length in Eq. In these evaluations. should be fixed rather than a constant time interval. A constant cracklength increment. 59. Between the crack length when 100 first detection is possible. Crackgrowth curves can ∆a be predicted for all critical details in a ∆a structure. ao.Some defects that may be rejected during standard fabrication controls would not.
repair and replacement. Cracking in several North American steel bridges was reviewed in a collection of case studies [20]. fracture mechanics analyses may assist in the assessment of complex structures that have connections which are not covered by test results. As the number of structures that approach or exceed their design life grows exponentially each year. and some of these have been analyzed extensively using fracture mechanics. Finally. Fracture mechanics will become an essential part of the assessment of these structures and. 123 . so does the number of structures which have a high probability of containing fatigue cracks. the decisions concerning their strengthening. consequently.required as the structure ages. Test data obtained from simple beams and small specimens may be inapplicable to large structures and to complex loading. fracture mechanic analysis enables a more rational approach for writing those parts of specifications related to inspection. This is particularly true for offshore structures. Therefore.
"J. American Association of State Highway and Transportation Officials.. 381402. Elementary Engineering Fracture Mechanics." Steel Structures. G. D. 221 (1921). "Fracture and Fatigue Control in Steel Structures.. 121128..T. 124 .. UK. J. 4. 9. Martinus Nijhoff." SI Units. The Stress Analysis of Cracks Handbook. Editor.N. 161175.01) (1983).R.. 6. P. St. J.. 1 (197879). 5." Materials Eng. Smith.." J. "A Critical Analysis of Crack Propagation Laws.A.A. 163198.. R. 1984. 8 (1985)." Phil. 1985. and Irwin. The Hague. New Jersey.. "AASHTO LRFD Bridge Design Specifications. Steel Constr. 1986. Guidelines for FractureSafe and FatigueReliable Design of Steel Structures.. Pavolic.. Pellini. D. Applic.C. P. Recent Research Advances and Their Application to Design. I. Testing Evaluation. Paris. London (Series A). 1983. "Fracture Mechanics in Structural Engineering. 1987. Welding Inst.. and III." Trans. 12. and Erdogan. 3. H. 762780. Netherlands. Rolfe. G. Lankford. 528534. 13.REFERENCES 1.. "The Influence of Microstructure on the Growth of Small Fatigue Cracks. Am.. and de Wit. Chapter VI. W. M. "The Phenomena of Rupture and Flow in Solids. Washington. Louis. a Measure of Fracture Toughness..T. Griffith. Paris Productions. and Hirt. II. Irwin.A. Cambridge. 2. J. Elsevier. Inst.. 7.. and Barsom. 2nd Ed. Paris. 11 (1983). Rolfe. Tada. First Edition.S. "An Introduction to Fracture Mechanics for Engineers – Parts I. Broek. Smith. 1994." Eng. Englewood Cliffs. M. Fracture and Fatigue Control in Structures. ASME (Series D)." in ASTM E 81381 (ASTM Book of Standards 3. F.. 85 (1963). S.M. 227235 and 316322. 215. PrenticeHall. 10. Trans. Royal Soc. "A Summary of FractureMechanics Concepts." Fatigue Fracture Eng. 5665. S. 11. 14 (1983).F. R.. Materials Struct.C. A. 8.
" ATLSS Report No 8902. Fisher. 1984. S.J. B." Int. Fisher.J... Bardell. 20." ECCS Technical Committee 6. R. and Socie. 1985. 21. 1989. J. J.T. Fukoda. "Fatigue of Metals Subject to Varying Stress. No. Wiley Interscience. 26. 25.A.14. 16. 7th Congress." Structural Engineering Report No.C. 1. 23. 72.." Preliminary Publication. D. Rotterdam.L. 2nd Edition. Comer. "Effect of Weldments on the Fatigue Strength of Steel Beams. The Netherlands. .R. Soc.. September. Albrecht." Jap. Engrg.E. March 1968. Fisher.. 15.. Steel Structures.W. Downing. American Railway Engineering Association..C. Bannantine... Yen.. 19. PA." National Cooperative Highway Research Program. Kluwer Academic Publishers. "Fatigue Strength Behaviour of Steel Beams with Welded Details. 4. Metals Rev. G. Klingerman. J.. Fisher. and Kulak. I." Chapter 15. Gurney. M. 29 (1984). Fatigue of Welded Structures. Englewood Cliffs.M. K." National Cooperative Highway Research Program Report 102. and Wang D. Highway Research Board. P. AREA.H. J.. University of Alberta.. Broek.T. Highway Research Board. IABSE.W. Vol. Hirt..L. G. "Fatigue of Bridge Structures – a Commentary and Guide for Design. 24. and Handrock. Dept. 18. 1990. Matsuishi. Civil Engrg. M. S.W. and McNamee. B.... 125 . 1982. and Richie. The Practical Use of Fracture Mechanics.M.R. Cambridge University Press. 1964. 1978. Yen..F. J... Evaluation and Investigation Of Cracking. 27. "Propagation of Short Fatigue Cracks. Japan. Mech. 1989. 1974. J. Fatigue and Fracture in Steel Bridges: Case Studies. "Simple Rainflow Counting Algorithms. B. Lehigh University. D. Dordrecht. PrenticeHall. J. Report 147.W.. 1979.A. and Endo. T. 1994. Frank. Bethlehem. Fisher. "Manual for Railway Engineering. D... B. and Viest. European Convention for Constructional Steelwork. "Fatigue Life of Bridge Beams Subjected to Controlled Truck Traffic. Suresh. 445476. and McNamee. "Recommendations for the Fatigue Design of Steel Structures. T.... Washington. D.M.O. "Fatigue Strength of Steel Beams With Transverse Stiffeners and Attachments. 17.A." International Journal of Fatigue.. 22. 1970. J. (New Jersey). D. Washington. Fundamentals of Metal Fatigue Analysis.W.
No. and Novak.. J. Washington. Klippstein. Structures Congress XIV.C.. 38.. 1987. S..E. 76. J. B. C.C.P. Castiglioni. "Fatigue Cracking in Bridges from OutofPlane Displacements.G.A. NCHRP Report 302.W. Cheng. 1977.W. and Dexter. J.J. Jian Jin. 36.. "Fatigue and Fracture Evaluation for Rating Riveted Bridges. G.28. Chicago.S.. 41. Highway Research Board. J. Barsom. and Cotton.. and Mann. 40. G.. and Blake." Final Report. NCHRP Report 336. 1977.L." Proceedings. NCHRP Project 12–12 Transportation Research Board.. "StressCorrosion Susceptibility of Highway Bridge Construction Steels. John Wiley & Sons. 30. 1996.. Vol.. Yen. Fisher. Van Dien." American Institute of Steel Construction." Journal of Structural Engineering. No. FHWA. "DistortionInduced Fatigue Cracking in Steel Bridges. 35. Chicago. "Bridge Fatigue Guide: Design and Details. J.W. Second Edition. LRFD Specification for Structural Joints Using ASTM A325 or A490 Bolts. IABSE Report No. ASCE. Kulak." NCHRP Report 181. Highway Research Board." Journal of Constructional Steel Research. 34." IABSE Periodica. 11. 1990. and Yen. Research Council on Structural Connections. (Available through the American Institute of Steel Construction.J. 5.. 39.. "Subcritical Crack Growth and Fracture of Bridge Steels. March 1997.F..H.. D. D." National Cooperative Highway Research Program. "Fatigue Testing of Anchor Bolts.V. B. Phase I.C.M. Schilling..C. AISC Pub. 13. B. 33.R. Inc..B. New York.W. J. and Keating. 116. C." IABSE Workshop on Evaluation of Existing Steel and Composite Bridges. D. 32. Yen. Fisher.E.T... Lausanne. R.T. Fisher. Hyatt. 1993.W. 29.) 37. Guide to Design Criteria for Bolted and Riveted Joints. Fisher. "Fatigue Strength of Riveted Connections. J. International Association of Bridge and Structural Engineering. "An Analysis of Fatigue Recommendations Considering New Data.T.. T112–11/77. J. G.T. April 1972. M.R.. Barsom." Report FHWA–RD–F73–003..M.. C. 126 ... Smith I. 25. Fisher. 4. Kaczinski. and Wang D. Zurich. J. Switzerland. J. Fisher. Kulak. ASCE. 1990. J. 1988. Washington.. 1974 31. November.. Aug. Carter. D. Wagner. and Struik.. December 1978. Wang.W.. "Fatigue of Welded Steel Bridge Steel Members Under Variable Amplitude Loadings.. Vol. Transportation Research Board.. No. 1989. M. "Fatigue Strength of Riveted Bridge Members." National Cooperative Highway Research Program..A. J. Vol. P. 1975." Canadian Journal of Civil Engineering. J. IL. "Design of Steel Beams with End Copes.L.
Berge. "Theoretical Analysis of the Influence of Attachment Size on the Fatigue Strength of Transverse NonLoadCarrying Fillet Welds. Tech. "Numerical Study of Geometrical Effects in Longitudinal NonLoadCarrying FilletWelded Joints. European Committee for Standardisation (1992). Abington Publishing. 1979. 1991. Maddox. Trondheim. Norwegian Inst. 91/1979. Welding Inst. and Gurney. 47. Vol." ENV 1993–1–1.K.J. 46. 1983. 1991.. Miami. Cambridge. S." Welding Standards Committee.42. PD 6493 "Guidance on Some Methods for the Derivation of Acceptance Levels for Defects in Fusion Welded Joints. 65.R. I.C. Gurney. Smith. 43. "Fatigue Strength of Welded Structures." Second Edition." Rept. 127 ." Welding Journal. MK/R 67. 45. S. 1985. U.. UK. "Effect of Plate Thickness in Fatigue of Cruciform Welded Joints. American Welding Society.F.R. Cambridge.. Brussels." Members Rept.... 44. T. "Eurocode 3: Design of Steel Structures. T. BSI.
56 fracture toughness. 58. 104 Bracing. 115 fatigue category. 4 stress intensity factors. 18. 100. 79 girder. 66 Constant amplitude fatigue limit (CAFL). 29 initiation. 110 future cracking. 81 multibeam. 103 compressive stresses. 51 lateral bracing. 55 effect of component size. 69. 113 mean stress effect. 1 inspection. 88. 18 propagation. 69. 75 lateral bracing. 104 threaded rods. 102 diaphragm. 3. 90 corrosion. 114 Connections: bolted. 77 Bridges box girder. 22. 54 Combined stresses. 81. 59 design truck. 47. 107 webs. 93 fillet weld. 76. 110 Cover plates. 95 length. 94 Bolted members. 57 Coped beam. 83 lateral connection plate. 80 groove weld. 83 riveted. 10 modes. 9 subcritical. 25 bolted shear splices. 4 location. 72 splice plates. 82 Corrosion. 6. 49 128 . 6 stress concentration. 58 tensiontype connections. 54 distortioninduced fatigue. 25. 65 summary of requirements. 13 preexisting. 64 fatigue resistance. 25. 7 discontinuities. 102 preloaded. 72 specifications. 48. 1. 54. 20. 108 shear studs. 104 tension. 8 Cumulative damage. 8 environmental effects. 73. 64 redundancy. 10 growth law. 11. 54 stress ranges. 85 Backing bar. 10 repair. 7. 55 basis of design rules. 1. 97 residual stresses. 69 load cycles. 57 design example. 110 critical length. 76 Butt weld (see Groove weld) Brittle fracture. 4 low temperature. 66. 19. 76 longspan. 115 small. 120 constant amplitude fatigue limit (CAFL). 5.INDEX AASHTO: average daily truck traffic (ADTT). 100. 108 shear studs. 100 growth (Paris equation). 51. 72. 17 Charpy Vnotch. 22. 103 nonpreloaded. 34 Cross frames (see Diaphragms) Cutoff limit. 9. 25 fatigue design truck. 91 Crack: compression zone. 84 dynamic load allowance. 2. 51 riveted connections.
72. 72 size of component. 23 threaded rods. 8 loadinduced. 76 Fatigue limits. 87 web gaps. 79 Environmental effects. 51. 22 bolted. 120 Examples. 122 Jintegral. 16 displacementinduced. 82 corrosion fatigue. 29 Charpy test. 25 distortion. 56. 76. 24 variable stress ranges. 11. 40 stress range. 97 dye penetrant. 80 Distortioninduced fatigue cracking. 121 Groove weld. 3 analysis of fatigue life. 28. 36 Eurocode. 22 lateral bracing attachment. 47 Fillet welds. 21 mechanical details. 47 limit state. 24 coped beam. 38. 20. 66 flangetoweb weld. 23 PalmgrenMiner rule.Diaphragms. 17 cracktip opening displacement (CTOD). 84 anchor bolts. 96 coring. 59. 93 Gusset plate. 115 riveted. 85. 13. 8. 5 defect assessment. 34 web gap. 18. 1. 111 stress cycle counting. example. 107 basis of design rules. 5 limit state. 19. 115 stress corrosion. 90 Flaws: in fabricated steel structures. 6 crackgrowth model. 6. 122 inspection interval. 66 fracture mechanics analysis. 96 visual. 35. 120 European Convention for Constructional Steelwork (ECCS). 41 stress histories. 47 definition. 122 magnetic particle. 31. 110. 114 controlling factors. 91. 2. 111 cumulative damage. 25 Fatigue life (fatigue life assessment): AASHTO. 21 welded details. 13 quantitative design. 17. 96 fracture mechanics. 76. 95 acoustic emission. 95 129 . 83 minimum stress. 106 transverse stiffeners. 34 redundancy. 108 shear studs.73 historical background. 16 limits. 49. 53 surfaces. 96 methods. 34 flange splice plates. 3. 73 type of detail. 18 Fracture: toughness. 44. 6 modes. 102 combined stresses. 75. 94 Inspection. 37. 29 grade of steel. 47 constant amplitude. 6 microstructural. 18. 73 summary. 113 Equivalent stress range. 17. 117 Fatigue: constant amplitude. 51 residual stress. 49. 67. Fracture mechanics. 95 ultrasonic.
6 fatigue. 91.Xray. 8. 98 Refabricate. 54. 120 Riveted connections. 93 Limit state: fracture. 8 Stiffeners: longitudinal. 58 maximum. 9 live load. 6. 96 Kfactor (see Stress intensity factor) Long span structures. 1. 114 compressive. 98 Methods. 18. 97 Cover plates. 24. 23. 47 stressrelieved. 98 Peening. 5. 116 mean. 106 Stress Corrosion. 92 Web stiffener. 93 transverse. 114 spectra. 67. 102 Steel: effect of grade upon fatigue life. 116 ratio. 98 Hole drilling. 42 Redundancy. 51 Repair. 50 variable. 24 stressrelieved elements. 34 Web gaps. 106 cycle counting. 120 nonwelded elements. 67. 58. 34 variable amplitude. 104 threshold. 108 range. 108 Rods (see Threaded rods) Secondary members. 57. 16. 20 effect on fatigue. 115 shear. 99 Gouge. 76. 34 Rainflow cycle counting. 93 Weathering steel. 73 Stress: combined. 57. 16 PalmgrenMiner rule. 98 Replacement. 18 130 . 120 constant amplitude. 104. 40 intensity factor. 67. 41 dead load. 50 Toughness. 115 compressive applied stresses. 22 toughness. 116 residual. 47. effect of. 98 Gas tungsten arc remelting. 111 Temperature. 113 Threaded rods. 41 equivalent. 47. 8 National Cooperative Highway Research Program (NCHRP). 120 effect on fatigue life. 36 gross section. 42 Residual stresses. 120 tension. 81 Longitudinal stiffener. 99 Reservoir cycle counting. 10 minimum. 23 net section. 92 Shear splices. 106 Threshold stress. 104 histories. 110 Welding: defects. 53 Variable amplitude stress ranges. 23. 73.
A. May 1998 .S.National Steel Bridge Alliance American Institute of Steel Construction H050 Printed in U.
This action might not be possible to undo. Are you sure you want to continue?
Use one of your book credits to continue reading from where you left off, or restart the preview.