FATIGUE STRENGTH O F DETERIORATED S T E E L

H I G H W A Y BRIDGES
By Patrick D . Zuraski, 1 Member, A S C E , and John E. Johnson, 2
Downloaded from ascelibrary.org by Indian Inst of Tech - Guwahati on 03/29/15. Copyright ASCE. For personal use only; all rights reserved.

Fellow, A S C E

ABSTRACT: The effect of long-term environmental exposure and an extended his-

tory of highway loading was quantitatively determined for steel bridge beams that
had been in service more than 50 years. Fatigue tests were conducted on specimens
removed from the tension flange of truss stringers salvaged from four bridge re-
placement projects. Based on the fatigue strength reduction factor Kf, determined
on a basis of mean fatigue life for all specimens in one bridge, the fatigue sen-
sitivities encountered were less than that associated with an AASHTO Category C
welded connection detail. For individual specimens, the maximum Kf value de-
termined was 3.0. At 36-ksi (248-MPa) stress range, average specimen lives ex-
ceeded 300,000 cycles for all bridges, for both center-span (maximum prior stress
history) and near-support (insignificant prior stress history) specimens. The cor-
rosion deterioration experienced by the plain (nonwelded) simple-span steel bridge
beams in this study caused the average fatigue strength to be reduced by approx-
imately one AASHTO fatigue category.

INTRODUCTION

The existence of a large number of deficient highway bridges has been

widely publicized. Approximately 40% of the 575,000 bridges in the United
States are categorized as structurally deficient or functionally obsolete. The
United States Department of Transportation issues annual reports (Highway
Bridge 1986) concerning deficient and obsolete bridges, and a critical eval-
uation of estimated bridge expenditures has been prepared (Pisarski 1987)
for the National Council on Public Works Improvement. To reach the best
decisions when assigning priorities for replacement or rehabilitation among
the large number of bridges requiring attention, a knowledge of the behavior
of deteriorated materials and requirements for safety must be employed. Un-
fortunately, according to the Transportation Research Board (Research Prob-
lem 1981), uniform qualitative and quantitative criteria for the identification
of deteriorated components, and the amount of the deterioration, are not
available.
The research described herein was conducted to examine the effect of long-
term environmental exposure and an extended history of highway loading
on bridge steel with no enhanced atmospheric corrosion resistance. The steel
suffered corrosion deterioration after being in service more than 50 years.
The results effectively complement investigations by others to determine the
effect of long-term exposure on the performance of ASTM A588 weathering
steel. Work conducted by Albrecht and colleagues (Albrecht and Naeemi
1984; Albrecht and Cheng 1983; Friedland et al. 1982) are some examples.
The fatigue behavior of 80-year-old, deteriorated railroad bridge stringers of
riveted, built-up cross section has also been reported (Out et al. 1984).
'Asst. Prof, of Civ. Engrg., Univ. of Akron, Akron, OH 44325.
2
Prof. of Civ. and Envir. Engrg., Univ. of Wisconsin, Madison, WI 53706.
Note. Discussion open until March 1, 1991. To extend the closing date one month,
a written request must be filed with the ASCE Manager of Journals. The manuscript
for this paper was submitted for review and possible publication on February 22,
1989. This paper is part of the Journal of Structural Engineering, Vol. 116, No.
10, October, 1990. ©ASCE, ISSN 0733-9445/90/0010-2671/$1.00 + $.15 per page.
Paper No. 25164.
2671

J. Struct. Eng. 1990.116:2671-2690.

 OBJECTIVES AND SCOPE

The two primary objectives of this investigation were to quantify the effect
of corrosion deterioration on fatigue strength in several steel bridges, and to
determine whether decades of previous stress history had an effect on the
Downloaded from ascelibrary.org by Indian Inst of Tech - Guwahati on 03/29/15. Copyright ASCE. For personal use only; all rights reserved.

fatigue strength of the steel in these bridges. To achieve these objectives,

fatigue tests were conducted on specimens removed from beams salvaged
from bridge replacement projects. Details regarding the specimens appear in
the next section, and the testing program is described in a later section. The
scope was restricted to plain (nonwelded) rolled beams that are typically in
use in older, short-span, off-system highway bridges, and, furthermore, only
single-span girders were considered. Consequently, the findings apply most
directly to steel deck-girder structures and to portions of the superstructure
for more complex steel bridges, such as stringer spans in through trusses.

TEST SPECIMENS

Nineteen bridge beams, salvaged from four rural bridge replacement proj-
ects on secondary roads in southern Wisconsin, provided the specimens used
for conducting the research. Average daily truck traffic for these locations
is less than 40 vehicles per day. All beams were plain (nonwelded) rolled
sections that had functioned as single-span stringers within truss bridges. The
spacing between beams varied from 18-26 in. (46-66 cm). Beam size, span
length, and age data are provided in Table 1. The maximum computed stress
range, occurring at center span from an HS20 truck, was between 18-24
ksi (124-165 MPa), based on a one-lane distribution factor and impact factor
of 1.3.
Monotonic tensile properties were determined for the Steel of bridge C
during strain-controlled tests. Based on the mean of eight specimens, the
upper yield (yield point) was 47.2 ksi (325 MPa) and the lower yield (flat
plateau) was 39.1 ksi (270 MPa). Since fatigue performance of rolled beams
is not sensitive to yield strength (Fisher et al. 1970), no additional yield data
were obtained.
Test specimens were removed from the bottom (tension) flange of sal-
vaged beams from the location shown in Fig. 1. A portion of the deteriorated
surface of the bottom flange in a salvaged beam was incorporated as one
entire surface of a specimen. The configuration of a specimen is shown in
Fig. 2. During specimen fabrication, precautions were taken to avoid intro-
ducing significant residual stresses. These measures included using proper
speeds for cutting saws and milling machines, and the use of lubricants to

TABLE 1. Salvaged Beam Data

Bridge Number of beams Depth (in.) Approximate length (ft) Age (yr)
(1) (2) (3) (4) (5)
C 10 7 18 77
B 3 6 12 83
H 3 8 13 50
W 3 8 13 65
Note: 1 in. = 25.4 mm; 1 ft = 0.305 m.

2672

J. Struct. Eng. 1990.116:2671-2690.

 ^ / j / % = Sect. B-B in Fig. 2
Downloaded from ascelibrary.org by Indian Inst of Tech - Guwahati on 03/29/15. Copyright ASCE. For personal use only; all rights reserved.

FIG. 1. Specimen Location within Tension Flange

6 in. (152 mm) ^_

i>- A [>• B

a"'
< I! E2I 4 v

!>• A
LB A-A B-B

Fillet Radius = 4 in. (102 mm) 1/4" = 6 mm

FIG. 2. Fatigue Test Specimen Configuration

reduce temperature. Specimen fillets were designed to blend smoothly into

the test zone to minimize the geometrical stress concentration.
The maximum depth of surface pitting caused by corrosion (defined herein
as the maximum peak-to-valley distance) was measured for 18 of 173 total
specimens, using a surface-grinding procedure. Working with the halves of
the fractued test specimens, material was removed in 0.001-in. (0.025-mm)
increments, in layers parallel to the smooth surface of the specimen, until
all evidence of corrosion disappeared. It was not possible to determine the
depth of pitting that existed at the fracture site because localized necking
caused a decrease in thickness in that vicinity. Consequently, only those
portions of the surface located greater than 0.25 in. (6 mm) from the fracture
were included in the surface-grinding observations.
The maximum depth of corrosion pitting among the 18 specimens, deter-
mined in the manner described, ranged from 0.021-0.052 in. (0.53 mm-
1.32 mm). The irregularity encountered in a typical surface profile is shown
in the cross section in Fig. 3, wherein the maximum depth of pitting, peak
to valley, is 0.032 in. (0.81 mm).
With regard to position along the span, specimens were removed from
both near-support and center-span locations. As shown in Fig. 4, near-sup-
port locations are defined to lie within the first 18 in. (46 cm) of the clear
span at each end of the salvaged beams, while center-span locations are
defined to lie within the central 24 in. (61 cm) of the clear span. Center-
span specimens experienced the combined effects of both corrosion and pre-

2673

J. Struct. Eng. 1990.116:2671-2690.

Downloaded from ascelibrary.org by Indian Inst of Tech - Guwahati on 03/29/15. Copyright ASCE. For personal use only; all rights reserved.

It '1 II
r J I :- i / «. | | 3
* ; i Scale (ram) f s f » ';
FIG. 3. Typical Surface Profile for Fatigue Test Specimen

55^^ ^ ^
>#-
18 in. (46 cm) 24 in. (61 cm) 18 in. (46 cm)

Near-Support Specimens - Removed from 18 in. Zones

Center-Span Specimens -- Removed from 24 in. Zones
FIG. 4. Specimen Location along Span

viously applied stresses, while near-support specimens were almost solely

affected by corrosion since the flexural stresses near simple supports are
quite small. The role of specimen location is described in greater detail in
the following.

BASIS FOR EVALUATING DETERIORATION

The notion of deterioration may be intuitively expressed as an accumu-

lation of damage, accompanied by a reduction in remaining life. Objectively,
one manifestation of structural deterioration is a decrease in the number of
applications (cycle life) of a given stress range necessary to cause failure in
a component. The decrease may be observed by comparing the cycle life in

2674

J. Struct. Eng. 1990.116:2671-2690.

 a new component to that for one removed from service. Furthermore, one
may compare the deterioration experienced at different locations within the
same element. The difference in cycle life for material samples removed
from each location provides a measure of the differences in deterioration.
The implementation of these approaches within separate facets of this in-
Downloaded from ascelibrary.org by Indian Inst of Tech - Guwahati on 03/29/15. Copyright ASCE. For personal use only; all rights reserved.

vestigation is described in the following.

Fatigue Strength Reduction
Specimens removed from near-support zones (see Fig. 4) provided an op-
portunity to isolate and examine the effect of corrosion deterioration on fa-
tigue performance. The low flexural stresses that can be expected to occur
close to the support are too small to exert any confounding effect that could
be attributed to previous stress history.
The fatigue strength reduction factor Kf is an appropriate parameter for
quantifying the effect of corrosion deterioration on fatigue strength. It is
defined, at a given life, as (Sandor 1972)
fatigue strength without a stress concentration
Kf = — (1)
fatigue strength with a stress concentration
In Eq. 1, test results for corroded, near-support specimens are used in the
denominator, and a mean regression line for stress range versus cycles-to-
failure in plain rolled beams is used for the numerator. The regression line
is the same one used by the AASHTO specification (Standard Specifications
1989) to obtain the allowable stress range for the category A (plain rolled
beam) fatigue detail, except the AASHTO equation has been shifted two
standard deviations to account for design uncertainties.

Fatigue Category A
Base metal in plain, rolled beams may be used as a fatigue reference con-
dition (Albrecht and Simon 1981), i.e., Kf = 1. To compute the Kf asso-
ciated with a particular surface condition, one may obtain the numerator for
the right side of Eq. 1 from Eq. 2, developed by Fisher et al. (1970) for
stress range versus cycles-to-failure for base metal
log N = 11.121 - 3.178 log SR (ksi) (2a)
or
log AT = 13.785 - 3.178 log SR (MPa) (2b)
where N = the number of cycles to failure; and SR = stress range. The
denominator for the fraction is equal to the stress range required to produce
failure in the specimen at the given life. The relationship is shown in Fig.
5, wherein the equation is evaluated at the mean cycle life for a group of
specimens tested at a given stress range.
The reduction in fatigue strength attributable to corrosion pits and defects
may be effectively compared, via Kf, to the reductions associated with the
flaws and abrupt changes in geometry that accompany various welded de-
tails. Values are reported in a later section.
Effect of Previous Stress History
A comparison between mean cycle lives obtained for near-support and
center-span specimens, tested at the same stress range, provides the basis to
2675

J. Struct. Eng. 1990.116:2671-2690.

 STRESS RANGE vs. CYCLES TO FAILURE

Category "A" Mean Regression Line

/ Log 1 0 N-11.121 -3.178 L o g 1 0 S R
Downloaded from ascelibrary.org by Indian Inst of Tech - Guwahati on 03/29/15. Copyright ASCE. For personal use only; all rights reserved.

a = stress range in fatigue test

b = mean life of specimens
c = stress range for plain rolled beams at N^

Cycles to Failure, Log N

FIG. 5. Use of Category A Mean Regression Line to Determine K,

determine whether the stress history of the beams has affected their fatigue
strength. Assuming a relatively uniform distribution of corrosion deteriora-
tion over the clear-span length of a beam, any difference in mean cycle lives
between the two locations must be attributable to a difference in the mag-
nitude of bending stresses that each region experienced in the past.

TESTING PROGRAM

Specimens were subjected to cyclic loading in a closed-loop servohy-

draulic axial-load system, enhanced with a microcomputer and plug-in data
acquisition board. Hydraulically actuated fatigue grips were used to achieve
repeatability of results, proper specimen alignment, and constant clamping
force. Zero-to-tension loading (minimum stress equal to zero) was employed
to be nearly compatible with the service load stresses experienced in simply
supported beams with small dead-load stresses.
Stresses were computed by dividing the applied load by the nominal area
of the specimens. The width of the test zone was measured directly. Thick-
ness was determined by placing the corroded surface of each specimen against
a flat, smooth plate, measuring the total thickness of plate and specimen,
and subtracting the plate dimension from the total. No correction was applied
to account for a loss in area caused by corrosion.
The primary reason for computing stress in this simple manner is that loss
of strength due to a reduction in the actual dimensions of a component has
been observed (Gough 1932) to be of a negligible order when compared with
the actual damage caused by the formation of sharp pits under corrosion
fatigue conditions. Similar conclusions were reached in a more recent in-
vestigation by Out et al. (1984), which noted that the severity of corrosion
cannot be accounted for by considering loss of area alone. Surface roughness
and the associated stress concentrations are major factors in creating a fatigue
detail that is more severe than might otherwise be anticipated.
2676

J. Struct. Eng. 1990.116:2671-2690.

 AASHTO Category A
t/ Mean Line 2 = = number of specimens
at indicated cycle life range
Downloaded from ascelibrary.org by Indian Inst of Tech - Guwahati on 03/29/15. Copyright ASCE. For personal use only; all rights reserved.

5 runouts

<n
0)
AASHTO Category B
Mean Line
1 = 7 specimen lives between
43,000 and 72,000 cycles

10
10°
Cycles to Failure

FIG. 6. Specimen Lives Compared to Mean Lines for AASHTO Fatigue Cate-
gories A and B

Constant-amplitude fatigue tests were conducted at 33- and 36-ksi (228-

and 248-MPa) stress range. The objective was to obtain stress range versus
cycles-to-failure data at two different stress ranges of sufficient magnitude
to achieve failure in less than approximately 18 hours while running at 40
Hz. Testing at 30- and 33-ksi (207- and 228-MPa) range was attempted first,
but after a time it was apparent that the higher stress ranges indicated pre-
viously were necessary.
Test Results
The results from 173 fatigue tests are shown in Fig. 6, which provides a
comparison to the mean regression lines for AASHTO fatigue categories A
(plain rolled beam) and B (welded beam). Individual results from 42 tests
conducted at 33-ksi (228-MPa) stress range are presented in Table 2. Data
from all four bridges, from both near-support (NS) and center-span (CS)
locations, are included, arranged in ascending order. The data have been
ranked (ordered), assuming that the results for both NS and CS specimens
are from the same population. The ranking was done to facilitate rank-sum
statistical tests, described in the next section.
Results from 131 tests conducted at 36 ksi (248 MPa) are summarized in
Table 3. The greater number of specimens at this stress range facilitated a
comparison of NS and CS results according to individual bridges. The min-
imum number of specimens for any combination of bridge and specimen
location (NS or CS) was 14. The results obtained for the individual bridges
at the 36-ksi (248-MPa) stress range are presented in the form of cumulative
distribution diagrams in the following section.

ANALYSIS OF RESULTS
Quantitative Assessment of Fatigue Strength Reduction
The use of the fatigue strength reduction factor Kf to quantify the effect
of corrosion deterioration was described earlier, and it was defined in Eq.
2677

J. Struct. Eng. 1990.116:2671-2690.

 TABLE 2. Cycles to Failure at 33-ksi Stress Range*

Near Support Center Span

Cycles to failure Cycles to failure
(kilocycles) Bridge" Rank" (kilocycles) Bridge0 Rank"
Downloaded from ascelibrary.org by Indian Inst of Tech - Guwahati on 03/29/15. Copyright ASCE. For personal use only; all rights reserved.

(1) (2) (3) (4) (5) (6)

311 C 1 450 H 4
316 B 2 474 C 6
362 H 3 515 H 9
452 H 5 590 C 11
480 H 7 659 H 12
482 C 8 825 H 14
572 H 10 959 B 17
665 H 13 1,247 C 18
883 B 15 1,377 C 20
926 C 16 1,438 H 22
1,365 B 19 2,341 W 31
1,402 W 21 2,904 w 33
1,488 C 23 3,108 B 34
1,628 c 24 3,425 W 35
1,696 H 25 3,687 H 36
f
1,806 W 26 4,250* W
f
1,979 c 27 4,337 W
f
1,995 c 28 4,430" c f
2,200 c 29 5,025" c f
2,245 w 30 5,850° c f
2,372 w 32 6,400" w
a
Zero-to-tension loading, i.e., mean stress = 16.5 ksi.
b
Number of near-support specimens = 21; bridges C, B, H, and W provided 8, 3, 6,
and 4, respectively.
"Number of center-span specimens = 21; bridges C, B, H, and W provided 7, 2, 6,
and 6, respectively.
d
Rank based on placing all near-support and center-span specimens into one 33-ksi stress
range sample.
"Runout.
f
Rank is greater than 36, but otherwise unknown.
Note: 1 ksi = 6.89 MPa.

1. To evaluate the denominator in the Kf ratio, i.e., to compute the fatigue

strength for a component containing stress concentrations, one may begin
with

in which b and m = empirical constants for a given fatigue detail; and N

and SR were previously defined in Eq. 2. Eq. 3 is the commonly accepted
("Fatigue Reliability" 1982) power law expression for high cycle fatigue,
often referred to as Basquin's equation.
By taking the log of both sides and rearranging terms, one obtains the
following equation, which is linear in log of N and SK

2678

J. Struct. Eng. 1990.116:2671-2690.

 TABLE 3. Mean Values and Coefficients of Variation for Cycles to Failure at 36-
ksl Stress Range
Near Support Center Span
Coefficient of Coefficient of
Downloaded from ascelibrary.org by Indian Inst of Tech - Guwahati on 03/29/15. Copyright ASCE. For personal use only; all rights reserved.

Bridge Sample mean, N variation, SN/N Sample mean, N variation, SN/N

(1) (2) (3) (4) (5)
C 592,700 0.82 561,400 1.03
B 307,500 0.51 629,700 0.72
H 343,900 0.53 301,400 0.53
W 990,800 0.38 866,400 0.29
Note: N = mean cycle life; and SN = standard deviation of cycle life. 1 ksi = 6.89
MPa.

log N = b - m log SR (4)

where b = the intercept of the line when log SR = 0; and m = the slope of
the line. Eq. 4, then, is the general expression for cycles-to-failure versus
applied stress range. The stress range required to cause failure at a given
number of cycles may be computed if the slope and intercept are known for
the particular fatigue detail.
Fatigue details covered in the AASHTO specification {Standard Specifi-
cations 1989) have slope m ranging from approximately 3.1-3.4, with a
mean value of 3.2 (Albrecht and Naeemi 1984). Substituting this mean slope
into Eq. 4, and evaluating the equation at the average cycle lives obtained
at 36-ksi (248-MPa) stress range for near-support specimens (taken from
Table 3), one obtains the intercept b.
Thus, a stress range versus cycles-to-failure (SR-N) mean line may be es-
timated for the average level of corrosion deterioration associated with each
of the four bridges. Albrecht (1982) demonstrated that such a procedure was
used to compute allowable stress range for Category E' details in AASHTO.
Intercepts for each of the four bridges are shown in Table 4.
Values for Kf were computed from the fatigue strengths calculated from
the Category A mean regression line and estimated SR-N lines, according to
Eq. 1, and are reported in Table 5. A cycle life equal to 500,000 was se-
lected. The number of cycles is not of great importance because the slopes

TABLE 4. Value for Intercept b for Estimated SR-N Line8

Average life of near -support
Bridge specimens at 36-ksi stress range Intercept b
(1) (2) (3)
C 592,700 10.7530
B 307,500 10,4680
H 343,900 10.5166
W 990,800 10.9762
'Based on assuming a slope of 3.2 for the SR-N line.
Note: 1 ksi = 6.89 MPa.

2679

J. Struct. Eng. 1990.116:2671-2690.

 TABLE 5. Fatigue Strength Reduction Factor Kf for Bridges with Corrosion De-
terioration, Determined from Average Lives
Stress range at
Fatigue detail 500,000 cycles (ksi) Reduction factor, Kf
Downloaded from ascelibrary.org by Indian Inst of Tech - Guwahati on 03/29/15. Copyright ASCE. For personal use only; all rights reserved.

(1) (2) (3)

AASHTO Category A 50.83 1.00
Bridge C 37.97 1.34
Bridge B 30.93 1.64
Bridge H 32.03 1.59
Bridge W 44.58 1.14
Note: 1 ksi = 6.89 MPa.

of the estimated lines and Category A line, 3.2 and 3.178, respectively, are
nearly equal.
AASHTO Category B (beams fabricated from three-plate weldments, with
no other welded attachments) and Category C (condition at the toe of trans-
verse stiffener welds on girder webs or flanges) details have Kf values of
1.49 and 1.95, respectively (Albrecht and Simon 1981). For the bridges
examined in this study, the fatigue sensitivity (computed at mean life) of
defects caused by corrosion deterioration was less than that of a Category
C detail for all four bridges.
For individual specimens, some Kf values were much higher than those
determined from the mean fatigue life for the entire population of samples
from one bridge. The highest value obtained for Kf was 3.04, for a specimen
from bridge C that failed after 43,000 cycles at a stress range of 36 ksi (248
MPa). Thus, it is necessary that older bridges, and also those subjected to
an environment of aggressive corrosion, receive adequate inspection to de-
tect individual occurrences of high fatigue sensitivity caused by severe cor-
rosion pitting. Values of Kf greater than 3.0 correspond to fatigue perfor-
mance nearly as poor as a Category E welded cover plate detail, the worst
AASHTO fatigue category.
Albrecht and Naeemi (1984) reported Kf values of approximately 1.2 and
1.1 for two different weathering steels. The tests were conducted by other
investigators on base metal specimens that had been subjected to weathering
for periods of four and six years, respectively, prior to testing. Four years
of weathering prior to testing a grade of steel possessing no atmospheric-
corrosion resistance resulted in Kf equal to approximately 1.3.

Effect of Previous Stress History

Greater fatigue life for NS specimens, compared to CS specimens, might
be considered the most likely event. Intuitively, one might expect that CS
specimens would fail at shorter lives, for a given stress, because a greater
portion of the fatigue life of the steel would have been consumed by pre-
viously applied stresses during decades of service history. Without regard to
the absolute maximum stress to which any of the salvaged beams may have
been subjected, certainly the steel in the CS specimens experienced much
higher stresses than that in the NS specimens.
Inspection of Tables 2 and 3, however, reveals differences between the
behavior of NS and CS specimens, tested at the same stress range, that are
2680

J. Struct. Eng. 1990.116:2671-2690.

 not consistent. In some instances NS specimens exhibited greater fatigue
life, and in others the greater life was shown by CS specimens.
With regard to the results at 33-ksi (228-MPa) stress range (Table 2),
obtained for a composite population including specimens from all four bridges,
Downloaded from ascelibrary.org by Indian Inst of Tech - Guwahati on 03/29/15. Copyright ASCE. For personal use only; all rights reserved.

CS specimens exhibited much longer fatigue life than NS specimens. The

average life for the 21 NS specimens was 1,220,000 cycles. The average
for the 21 CS specimens cannot be determined because several runouts (stop-
ping at more than 4,200,000 cycles) occurred. If, however, one assumes
that the CS cycle lives at 33 ksi (228 MPa) are from a two-parameter Weibull
probability distribution, and based on a Weibull slope determined to be 1.2,
the calculated mean for the CS specimens, after Johnson (1964), is approx-
imately 2,700,000.
At 36-ksi (248-MPa) stress range, Table 3 reveals that a change in spec-
imen location from near support to center span resulted in a decrease in
fatigue life of 5.3%, 12.4%, and 12.6% for bridges C, H, and W, respec-
tively. Bridge B, on the other hand, exhibited a 104.7% increase in fatigue
life, comparing center span to near support. Lumping all data at 36-ksi (248-
MPa) range into NS and CS populations, the mean lives were 573,300 and
592,600, respectively.
With regard to coefficients of variation (COV), a typical value ("Fracture
Reliability" 1982) for constant-amplitude fatigue tests is considered to be
approximately 0.6, although that is probably for smooth specimens. Except
for bridge C, all COV values are definitely within reason. Values of 0.82
and 1.03 for bridge C are high, but considering the variability in surface
conditions for bridge steel after 80 years of exposure, such COV values are
not entirely unexpected.
Cumulative distribution (CD) diagrams for all tests are shown in Figs. 7 -
11. In each diagram the line plot shows the sample cumulative frequency
distribution for near-support specimens. Open squares have been used for
center-span specimens, for clarity, to eliminate any confusion that might
result from overlapping distributions. The CS cumulative distribution dia-
grams would otherwise have the same stepped appearance.
In Fig. 7, composed of data from all four bridges, a significant difference
in the CD diagram may be noted, with a divergence appearing at approxi-
mately 1,500,000 cycles. At 33-ksi (228-MPa) stress range, considering all
four bridges on a composite basis, specimens removed from center span
demonstrated superior fatigue life compared to near-support specimens.
In Figs. 8-11, individual distributions are shown for bridges C, B, H and
W, respectively. The CD diagrams for bridges C, H and W are sufficiently
close to each other that it is impossible to determine by inspection whether
NS and CS specimens from these bridges actually have different fatigue lives
at 36-ksi (248-MPa) stress range. In Fig. 9, for bridge B, the significant
divergence between the CD diagrams reveals that CS specimens have greater
fatigue life than NS specimens at 36-ksi (248-MPa) stress range.

Statistical Tests
Two statistical tests were conducted to examine the level of significance
associated with the difference in fatigue lives for NS and CS specimens. The
Kolmogorov-Smirnov two-sample test (Breiman 1973) utilizes the maximum
difference between sample cumulative distribution functions, over the entire
range of data, as the test statistic. The rank-sum test (Blank 1980), fre-

2681

J. Struct. Eng. 1990.116:2671-2690.

 # 6 more ' a ' specimens
100 - -I had lives exceeding
90 ;
80 -
1 70-
Downloaded from ascelibrary.org by Indian Inst of Tech - Guwahati on 03/29/15. Copyright ASCE. For personal use only; all rights reserved.

60
i£ :
/
- 50 - 4 million cycles
8 40 '- a #
w
D
I 30: a
20 ; a
a
10 \ r1 Q COMPOSITE SAMPLE
0 J
J 1000
a
a ° CUMULATIVE
2000 DISTRIBUTION
3000 DIAGRAM
4000
n 33 ksi STRESS RANGE
a Cycles to Near-Support
Failure (kilocycles)
Specimens -- Line Plot
1
Center-Span Specimens - Open Squares
FIG. 7. Cumulative Distribution Diagram for Composite Sample at 33-ksi Stress
Range

quently referred to as the Mann-Whitney test, utilizes a test statistic that is

the sum of the number of times a data value of one sample precedes a value
in the second sample. All tabulated data for cycles-to-fallure versus stress
range were ranked within the individual tables to facilitate this latter test.
The results from both statistical tests, which compared very favorably, are
summarized as follows. The hypothesis that there is no difference in fatigue
lives between NS and CS samples should be accepted for bridges C, H, and
W at 36-ksi (248-MPa) stress range. On the other hand, based on results
obtained at approximately the 0.0125 significance level, it is proper to con-
clude that center-span specimens had fatigue lives greater than those of near-
support specimens for bridge B at 36 ksi (248 MPa) and the composite sam-
ple at 33 ksi (228 MPa). Stated differently, to conclude for the two last cases
that CS lives are greater than NS lives, one would be incorrect approximately
only one time in 80.

Center-Span Behavior
Progressive strengthening of weak regions (at the root of the most critical
corrosion defects) in center-span zones of the deteriorated beams might be
the reason for longer fatigue lives of CS specimens in selected instances.
Such behavior may possibly be governed by time-dependent localized effects
from strain-aging. Examples of increases in fatigue strength that some in-
vestigators have obtained by planned programs of "understressing and coax-
ing," and the relationship to strain-aging, will be briefly described within
this section. A discussion of the relevance of these concepts to bridges is
included.
Understressing refers to repeated stressing at or below the fatigue limit.
Coaxing involves gradually increasing the amplitude of the alternating stress
in small increments. Sinclair (1952) conducted rotating bending tests in-
volving understressing and coaxing on SAE 1045 steel and was able to ob-
2682

J. Struct. Eng. 1990.116:2671-2690.

 0
Downloaded from ascelibrary.org by Indian Inst of Tech - Guwahati on 03/29/15. Copyright ASCE. For personal use only; all rights reserved.

c BRIDGE C
0 CUMULATIVE DISTRIBUTION DIAGRAM
o 36 ksi STRESS RANGE
0
a. Near-Support Specimens -- Line Plot
Center-Span Specimens - Open Squares

500 1000 1500 2000 2500

Cycles to Failure (kilocycles)

FIG. 8. Cumulative Distribution Diagram for Bridge C at 36-ksl Stress Range

T3
0

C
0 a BRIDGE B
O p CUMULATIVE DISTRIBUTION DIAGRAM
0
36 ksi STRESS RANGE
Near-Support Specimens - Line Plot
Center-Span Specimens - Open Squares

500 1000 1500 2000 2500

Cycles to Failure (kilocycles)
FIG. 9. Cumulative Distribution Diagram for Bridge B at 36-ksl Stress Range

tain significant increases in fatigue strength. One result is shown in Fig. 12,
where the ordinate in the plot is stress amplitude, which is equal to one-half
the range. Two coaxing curves are shown, both beginning at 40 ksi (276
MPa) [3 ksi (21 MPa) below the fatigue limit], and increasing by 1 ksi (7
MPa) at each change in stress amplitude. The conventional stress versus
cycles-to-failure (S-N) curve is indicated by the dashed line.
The magnitude of the increase in fatigue strength was dependent on the
cycle increment, which is the number of cycles of stress to occur after each
new increase in load. In the first coaxing curve (ending with an X), with a
cycle increment of 2,000,000 cycles, it was possible to attain 2,000,000
2683

J. Struct. Eng. 1990.116:2671-2690.

 100
90
80
Downloaded from ascelibrary.org by Indian Inst of Tech - Guwahati on 03/29/15. Copyright ASCE. For personal use only; all rights reserved.

CO 70
60
c 50 BRIDGE H
CO
o 40 CUMULATIVE DISTRIBUTION DIAGRAM
CD 30 36 ksi STRESS RANGE
a. 20 Near-Support Specimens -- Line Plot
Center-Span Specimens - Open Squares
10

0 500 1000 1500 2000 2500

Cycles to Failure (kilocycles)

FIG. 10. Cumulative Distribution Diagram for Bridge H at 36-ksi Stress Range

00 -
90 -
80 -
-a
cu 70
60 -
50 -
c BRIDGE W
CO 40 - CUMULATIVE DISTRIBUTION DIAGRAM
o 30
CO 36 ksi STRESS RANGE
a. 20 Near-Support Specimens - Line Plot
n1 Center-Span Specimens - Open Squares
10
0' | >
500 1000 1500 2000 2500
Cycles to Failure (kilocycles)

FIG. 11. Cumulative Distribution Diagram for Bridge W at 36-ksi Stress Range

cycles at a maximum amplitude of 48 ksi (331 MPa), plus millions of cycles

at the intervening stress amplitudes. The increase was dramatic because, ac-
cording to the S-N line, had stressing started at 48 ksi (331 MPa), it would
only have been possible to achieve approximately 370,000 total stress cycles.
A more dramatic increase in fatigue strength was achieved when the cycle
increment was increased to 10,000,000 cycles. The second curve (ending
with an open circle) indicates raising the fatigue strength to as high as 60-
ksi (414-MPa) stress amplitude through coaxing.
Sinclair ascertained that the benefits achieved by understressing and coax-
ing are associated with materials capable of being strain-aged. The strain-
2684

J. Struct. Eng. 1990.116:2671-2690.

 Coaxing Curves -- SAE 1045 Steel
60 dashed line = conventional S-N curve
co
stepped lines = coaxing curves
cycle increment: 1°
Downloaded from ascelibrary.org by Indian Inst of Tech - Guwahati on 03/29/15. Copyright ASCE. For personal use only; all rights reserved.

CD
X = 2 million cycles
Z)
+-•
"o.
E
50
***.,
O = 10 million cycles
i
<
CO
en
0)
v_
•+-'
X f J
£<[!
CO 40 ^
•
10 10° 10' 10° 10'
Number of Cycles
FIG. 12. Effect of Coaxing on SAE 1045 Steel (Sinclair 1952)

aging phenomenon is associated with regions of localized lattice distur-

bances, called edge dislocations, that are responsible for the yield point in
mild steel. One may consider the disturbance as being associated with an
extra half-plane of atoms in the lattice network.
Carbon and nitrogen atoms, which are attracted to dislocations because of
a low-energy state (Levy and Sinclair 1955), are presumed to strengthen the
metal by diffusing to dislocations in the crystal lattice. The occurrence of a
well-defined yield point in mild steel is associated with dislocations being
pulled away from the pinning atoms, and then continuing to move under a
lower force (Pascoe 1978). Such a well-defined yield point is shown at the
end of the initial slope in the stress-strain diagram in Fig. 13.
Loading paths X and Y are shown in Fig. 13, with path Y illustrating
increased strength due to strain aging. The hold time in Y is sufficient to
allow the carbon and nitrogen atoms to diffuse back through the lattice and
repin the dislocations. Further, greater numbers of dislocations are present
after having reached Y, due to dislocation multiplication as yielding pro-
gresses.

Relevance to Bridges
It is possible that understressing and strain-aging were responsible for pro-
gressive strengthening at the sites of the most critical corrosion defects in
center-span zones of the deteriorated beams. Applications of moderate-to-
heavy service loads might serve to strengthen, on a selective basis, those
areas operating under the greatest stress concentrations. A rest period of
several days between applications would allow sufficient time for solute at-
oms of carbon and nitrogen to diffuse through the crystal lattice and repin
the dislocations located in the vicinity of localized defects.
A natural limitation on the amount of strengthening is the available duc-
tility in the material. Each yield excursion diminishes remaining ductility.
Also, different amounts of strengthening are to be expected for different
2685

J. Struct. Eng. 1990.116:2671-2690.

 STRESS vs. STRAIN FOR MILD STEEL
Y/ I. _
x__

f
Downloaded from ascelibrary.org by Indian Inst of Tech - Guwahati on 03/29/15. Copyright ASCE. For personal use only; all rights reserved.

/ if i
/ -5/ i?/
/ ••!?/ _ /

/ / / / /
/ f 1/
/
/ v
V S/
-sy Note

/ <•?/ '""V X andY denote two

/ I I separate loading paths

Strain

FIG. 13. Effect of Strain-Aging in Mild Steel

bridges. The magnitude of the average truck load for a particular structure,
and frequency of application, will be factors in the amount of progressive
strengthening that occurs.
Based on mean lives obtained by lumping all specimens together at each
stress range, the 21 CS specimens had greater fatigue life at 33-ksi (228-
MPa) stress range than the 21 NS specimens, while at 36-ksi (248-MPa)
stress range, the mean lives for the 66 NS and 65 CS specimens, 573,300
and 592,700, respectively, were approximately equal. One possible expla-
nation is that the new level of strength achieved through strain-aging was
greater than the true stress occurring at the major defects when the nominal
33 ksi (228 MPa) was applied, but less than the true stress caused by the
nominal 36 ksi (248 MPa). In the context of Fig. 13, the actual stresses
caused by the nominal 33 ksi (228 MPa) might be stopping short of the yield
point at Y, while stresses caused by the 36-ksi (248-MPa) range might be
enough to go past the strain at the new yield point. Greater cyclic plasticity
in the 36-ksi (248-MPa) tests would, then, quickly eliminate the benefits
derived from the earlier strengthening.
With regard to bridge B, special circumstances related to truck service
loads might explain how CS specimens at 36 ksi (248 MPa) could achieve
105% greater fatigue life (Table 3) compared to NS specimens, while NS
specimens had slightly greater fatigue lives compared to CS specimens for
the three other bridges tested at the same stress range. A higher, and also
more sustained, level of truck traffic could cause such a difference. In ad-
dition, noting that bridge B was constructed in 1899, the magnitude of pres-
ent-day truck loads cause more frequent overloads on a bridge designed for
the loads of more than 90 years ago.
In an attempt to explain the unusual performance of bridge B, a cursory
2686

J. Struct. Eng. 1990.116:2671-2690.

 examination was conducted to determine whether beams from this early bridge
might be wrought iron. A comparison of photomicrographs from bridges B
and C revealed that both were constructed of steel.
Downloaded from ascelibrary.org by Indian Inst of Tech - Guwahati on 03/29/15. Copyright ASCE. For personal use only; all rights reserved.

CONCLUSIONS

This study was conducted to provide an understanding of the effects of

aging and long-term exposure on the fatigue performance of deteriorated,
steel deck-girder highway bridges. By using the results obtained from tests
conducted on specimens removed from salvaged bridge beams with service
histories exceeding 50 years, it was possible to quantify the effects of cor-
rosion deterioration and examine the effect of previous stress history on fa-
tigue strength. The conclusions listed in the following pertain to deteriorated
plain (nonwelded), single-span, rolled-section steel bridges that are typically
in use in older, short-span, off-system highway bridges. Furthermore, the
conclusions must also be restricted to components that experience small dead-
load (mean load) stresses, since the short-span bridges investigated are in
that category.

1. The fatigue strength reduction factor Kf provides a basis to quantify the

effect of decades of corrosion deterioration on the fatigue strength of beams
salvaged from existing bridges. Furthermore, with respect to fatigue strength, it
provides an effective means for comparing the effect of naturally occurring flaws,
such as corrosion-induced defects, to the effect of flaws induced during welding
operations in bridges.
The range in Kf values, determined for mean fatigue lives at 36-ksi (248-
MPa) stress range, was 1.14-1.64. Thus, for mean life only, the greatest re-
duction in fatigue strength encountered herein was between that associated with
AASHTO (Standard Specifications 1989) Category B and Category C fatigue
details in welded bridges. The highest value obtained for an individual specimen
was 3.04, which is between Category D and Category E.
2. An adequate fatigue strength was exhibited by tensile specimens that were
removed from deteriorated, simple-span steel bridge beams with service histories
ranging between 50-83 yr. Average lives for near-support and center-span spec-
imens tested at 33- and 36-ksi (228- and 248-MPa) stress ranges are shown in
Fig. 14. The averages incorporate results from four different bridges.
The salvaged steel exhibited a capacity to withstand hundreds of thousands
of applications of stress range greater than or equal to 30 ksi (207 MPa). Only
seven of 173 specimens failed to achieve a life of 100,000 cycles, and all seven
were tested at 36-ksi (248-MPa) stress range.
3. A history of bending stress in center-span regions did not reduce the fatigue
strength (compared to near-support specimens) of the steel to levels significantly
below which they had already been reduced by corrosion.
Based on 173 tests conducted at 33- and 36-ksi (228- and 248-MPa) stress
ranges, the potentially damaging effect of previously applied stresses never ex-
erted more than a 13% reduction in mean fatigue life for any bridge. In two
instances, accounting for 72 of the 173 specimens, the fatigue life exhibited by
specimens removed from center-span regions was superior, at the 95% confi-
dence level, to that demonstrated by specimens taken from near-support loca-
tions.
4. Observed incidences of superior fatigue strength for center-span steel, com-

2687

J. Struct. Eng. 1990.116:2671-2690.

 36
N fifi = 573,000 N65= 593,000
ksi

33 N = 1,220,000 N 2 1 = 2,700,000
21
ksi
Downloaded from ascelibrary.org by Indian Inst of Tech - Guwahati on 03/29/15. Copyright ASCE. For personal use only; all rights reserved.

SSSST ^ ^
Notes: The subscript with N indicates the number
of specimens utilized in the determination
of average life.

The averages incorporate results from four

different bridges.

FIG. 14. Average Lives for Near-Support and Center-Span Specimens at 33- and
36-ksi Stress Ranges

pared to near-support material, may possibly be governed by a time-dependent

mechanism of localized strengthening, i.e., strain-aging. This phenomenon in-
volves disturbances (dislocations) in the atomic lattice network that are respon-
sible for the yield point in mild steel.
It is important to note, however, that superior fatigue resistance for some
center-span locations is with respect to near-support material that has experienced
similar levels of deterioration, and not with respect to new steel with a relatively
smooth, as-rolled, surface condition.
5. The corrosion deterioration experienced by the plain (nonwelded) simple-
span steel bridge beams in this study caused the average fatigue strength to be
reduced by approximately one AASHTO fatigue category. Thus, it is unlikely
that fatigue will cause an abrupt end to the life of plain rolled beams experiencing
similar levels of corrosion deterioration, assuming the stress ranges experienced
by secondary road bridges are comparable to the stress ranges that have been
detected in larger bridges. Daniels et al. (1986) measured stress ranges in four
major bridges and found that the largest stress range to occur during more than
19,000 truck passages was 6.2 ksi (43 MPa).
Given the level of fatigue strength summarized in conclusion two, and con-
sidering that stress ranges greater than 15 ksi (103 MPa) have not been recorded
on real bridges, remaining life for the type of bridges investigated herein will
be controlled by severe overload, heavy flooding, or functional obsolescence.

ACKNOWLEDGMENTS
Support was provided by the National Science Foundation, grant CEE
2688

J. Struct. Eng. 1990.116:2671-2690.

 8407248, and the Wisconsin Department of Transportation. The work was
conducted at the facilities of the department of civil and environmental en-
gineering and the department of engineering mechanics at the University of
Wisconsin-Madison.
Downloaded from ascelibrary.org by Indian Inst of Tech - Guwahati on 03/29/15. Copyright ASCE. For personal use only; all rights reserved.

APPENDIX I. REFERENCES

Albrecht, P. (1982). "Fatigue reliability analysis of highway bridges." Transporta-

tion Research Record 871, Transportation Research Board, Washington, D.C.,
73-80.
Albrecht, P., and Cheng, J.-G. (1983). "Fatigue tests of 8-yr weathered A588 steel
weldment." J. Struct. Engrg., ASCE, 109(9), 2048-2065.
Albrecht, P., and Naeemi, A. H. (1984). "Performance of weathering steel in bridges."
NCHRP Report 272, Transportation Research Board, Washington, D.C.
Albrecht, P., and Simon, S. (1981). "Fatigue notch factors for structural details."
J. Struct. Div., ASCE, 107(7), 1279-1296.
Blank, L. (1980). Statistical procedures for engineering, management, and science.
McGraw-Hill Book Company, New York, N.Y.
Breiman, L. (1973). Statistics: With a view toward applications. Houghton Mifflin
Company, Boston, Mass.
Daniels, J. H., et al. (1986). "WIM + response study of four in-service bridges."
Proc, Third Annual Int. Bridge Conf., Engineers' Society of Western Pennsyl-
vania, Pittsburgh, Pa., 136-142.
ASCE Committee on Fatigue and Fracture Reliability "Fatigue reliability: Introduc-
tion." (1982). J. Struct. Div., ASCE, 108(1), 3-23.
Fisher, J. W., et al. (1970). "Effects of weldments on the fatigue strength of steel
beams." NCHRP Report 102, Transportation Research Board, Washington, D.C.
Friedland, I. M., Albrecht, P. and Irwin, G. R. (1982). "Fatigue of two-year weath-
ered A588 stiffeners and attachments." J. Struct. Div., ASCE, 108(1), 125-144.
Gough, H. J. (1932). "Corrosion-fatigue of metals." J. Inst. Metals, 42(2), 17-92.
"Highway bridge replacement and rehabilitation program." (1986). Seventh Annual
Report of the Secretary of Transportation to the Congress of the United States,
U.S. Government Printing Office, Washington, D.C.
Johnson, L. G. (1964). The statistical treatment offatigue experiments. Elsevier Pub-
lishing Company, Amsterdam, the Netherlands.
Levy, J. C , and Sinclair, G. M. (1955). "An investigation of strain aging in fa-
tigue." Proc, ASTM, 55, 866-890.
Out, J. M. M., Fisher, J. W., and Yen, B. T. (1984). "Fatigue strength of weathered
and deteriorated riveted members." Transportation Research Record 950, 2,
Transportation Research Board, Washington, D . C , 10-20.
Pascoe, K. J. (1978). An introduction to the properties of engineering materials.
Van Nostrand Reinhold Company, London, U.K.
Pisarski, A. E. (1987). The Nation's public works: Report on highways, streets,
roads and bridges. National Council on Public Works Improvement, Washington,
D . C , 38-44.
"Research problem statements." (1981). Transportation Research Circular No. 227,
Transportation Research Board, Washington, D.C.
Sandor, B. I. (1972). Fundamentals of cyclic stress and strain. University of Wis-
consin Press, Madison, Wis.
Sinclair, G. M. (1952). "An investigation of the coaxing effect in fatigue of metals."
Proc, ASTM, 52, 743-758.
Standard specifications for highway bridges. (1989). 14th Ed., American Association
of State Highway and Transportation Officials, Washington, D.C.

APPENDIX II. NOTATION

The following symbols are used in this paper:

2689

J. Struct. Eng. 1990.116:2671-2690.

 b = empirical constant; intercept of log 5R-log JV line;
CS = designation to indicate center-span test specimen;
Kf = fatigue strength reduction factor;
m = empirical constant; slope of log 5^-log N line;
N = number of cycles to failure; mean cycle life;
Downloaded from ascelibrary.org by Indian Inst of Tech - Guwahati on 03/29/15. Copyright ASCE. For personal use only; all rights reserved.

NS = designation to indicate near-support test specimen;

SN = standard deviation of cycle life; and
SR = stress range.

2690

J. Struct. Eng. 1990.116:2671-2690.