See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/242303636

FORENSIC ANALYSIS OF THE STEEL GIRDER FRACTURE IN THE I-95

BRANDYWINE RIVER BRIDGE

Article · January 2003

CITATION READS

1 386

1 author:

Spencer E. Quiel
Lehigh University
74 PUBLICATIONS   677 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

University Transportation Center for Underground Transportation Infrastructure View project

Blast Resistant Design of Cladding View project

All content following this page was uploaded by Spencer E. Quiel on 18 July 2014.

The user has requested enhancement of the downloaded file.

 FORENSIC ANALYSIS OF THE
STEEL GIRDER FRACTURE IN THE
I-95 BRANDYWINE RIVER BRIDGE

Spencer Quiel
University of Notre Dame

August 8, 2003

Project:
Research Experience for Undergraduates (REU)
Funded by the National Science Foundation

Advisor:
Dr. Michael Chajes
Center for Innovative Bridge Engineering
University of Delaware
301 DuPont Hall
Newark, DE 19716

1
Brandywine River Bridge Forensic Analysis 8/8/03

Table of Contents
ABSTRACT........................................................................................................................ 2
INTRODUCTION .............................................................................................................. 2
BACKGROUND ................................................................................................................ 2
Fatigue and Fracture ....................................................................................................... 3
Description of the Brandywine River Bridge ................................................................. 3
Description of the Fracture ............................................................................................. 5
Case Similarity: The I-91 Quinnipiac River Bridge ....................................................... 6
Bridge Retrofit ................................................................................................................ 6
Temporary Repair and Evaluation .................................................................................. 7
PERMANENT REPAIR..................................................................................................... 7
Jacking Procedure and Repairs ....................................................................................... 7
Data Collection ............................................................................................................... 9
RESULTS ........................................................................................................................... 9
Problems with Data File JS-26 ..................................................................................... 10
Jacking Data Results ..................................................................................................... 10
DISCUSSION ................................................................................................................... 12
Computer Modeling of the Brandywine River Bridge ................................................. 12
Ambient Traffic Load Analysis .................................................................................... 15
SUMMARY AND CONCLUSION ................................................................................. 15
ACKNOWLEDGEMENTS.............................................................................................. 16
REFERENCES ................................................................................................................. 16

1
Brandywine River Bridge Forensic Analysis 8/8/03

ABSTRACT
The following forensic analysis of the I-95 Brandywine River Bridge was conducted in
response to a recent fracture in one of the bridge’s steel girders. The break was attributed
to welding defects present in the girder’s tension zone. The bridge was retrofitted
accordingly and temporarily repaired with splice plates. Initial field tests determined that
the temporary repairs were adequate to resist traffic loads. While permanent repairs were
installed, the bridge was monitored for a redistribution of deflection, strain, and stress. It
was determined that the girder adjacent to the cracked girder had absorbed most of the
stress shed from the fractured member. The bridge appeared to sag slightly below its
original dead load position when released in its new repaired state. Computer models
were constructed to confirm the changes in behavior that occurred when the bridge
cracked. Modeling supported the general deflection and stress trends of the actual
structure but was unable to precisely mirror the bridge behavior.

INTRODUCTION
On Thursday, April 10, 2003, a Wilmington bird watcher walking along the
Brandywine River noticed sunlight peeking through one of the girders in the I-95 bridge
crossing overhead. Later that night, Delaware Department of Transportation inspection
crews confirmed the existence of a seven foot long vertical crack in a steel girder on the
bridge’s northbound extension. The crack occurred at the mid-span of the fascia
(outermost) girder in the bridge’s center span directly over the Brandywine River. The
right lane of northbound traffic was immediately closed and truck traffic was confined to
the left-most lane to avoid agitating the crack further.
With input from the consulting engineers of DMJM+Harris, temporary repairs
were erected. Splice plates were bolted to the cracked section to provide stability.
Bridge retrofit consisted of holes drilled into the girder in order to arrest the propagation
of the existing crack and to prevent the formation of new cracks. University of Delaware
technicians and faculty conducted field testing to evaluate the ability of the temporary
repairs to adequately resist the bridge’s traffic loads.
In June, the bridge was lifted to its original non-cracked position using hydraulic
jacks and fitted with permanent repairs. The condition of the bridge following these
repairs has become the focus of current research. Of primary concern is how much
loading the girders adjacent to the fractured girder sustained, how much load the
fractured girder reassumed after it was repaired, and how loads have been distributed
among the girders bridge both before and after the fracture and the repairs. The research
documented in this report makes use of strain data taken throughout the repair process as
well as computer modeling used to analyze the bridge’s final condition.

BACKGROUND
This evaluation of the current state of the Brandywine River bridge will be based
in an understanding of bridge fatigue and fracture mechanics. Combined with knowledge
of the bridge’s structural design and the details of the crack, research can then attempt to
determine the origin and cause of the fracture. Also, this type of crack is not unique to

2
Brandywine River Bridge Forensic Analysis 8/8/03

this bridge; comparison to a similar case of damage on a bridge over the Quinnipiac
River in Connecticut can provide information about the nature of the Brandywine fracture.
Initial work with the cracked bridge, as mentioned above, has included retrofit of
the bridge, installation of temporary repairs, field testing, and permanent repair. Each of
these steps will be discussed in order to set the stage for an analysis of the new permanent
condition of the bridge.

Fatigue and Fracture

Bridges are prone to fatigue and fracture because of their constant exposure to
traffic loading. These repetitive loads, though producing stresses in the girders well
below the steel’s ultimate stress, can reduce the strength of the material. Eventually, at
the location of a flaw in the material, the ability of the steel to withstand the applied
traffic loads (fracture toughness) will decrease and allow for formation of the crack.
Welds are one of the most common material “flaws” that reduce fatigue strength.
Welding changes the metallurgical structure of the steel and introduces high residual
stress and the potential for defects (Gurney, 1968). These defects can accelerate the
effects of fatigue and provide an origin for the formation of fatigue cracking. Common
weld defects include lack of fusion or penetration, voids, inclusion of foreign material,
stress cracking, and imperfect shape (Gurney, 1968). Cracks such as the Brandywine
fracture are seldom the result of sudden inflicted damage and can usually be traced back
to the defect by examining the chevron patterns in the crack itself. This type of fatigue
cracking has three stages: (1) a small crack or pit will form at a defect after loading
begins, (2) the crack propagates slowly according to stress paths as loading continues,
and (3) sudden fracture will occur perpendicular to the plane of stress when the material
cross-section becomes too small to resist loading (Askeland and Phule, 2003).

Description of the Brandywine River Bridge

The I-95 bridge, constructed in the early 1960’s, crosses the Brandywine River in
New Castle County, DE. The bridge consists of two identical parallel spans that
independently carry the northbound and southbound traffic (Chajes et al, 2003). The
superstructure in each span has six parallel steel plate girders (twelve total) that are
spaced 10 feet apart on a 40° skew counterclockwise from north. An 8-inch concrete
deck with a thin bituminous wearing surface rests on the steel skeleton.

3
Brandywine River Bridge Forensic Analysis 8/8/03

Figure 1. Interstate 95 Brandywine River Bridge

The main span of the bridge is three-span continuous, with each of the back spans
measuring 200 feet and the center span reaching 245 feet across the river. The girders are
haunched over the concrete piers; each is 8 feet deep at the middle of the center span and
12 feet deep over the piers. Each girder has been fitted with longitudinal stiffeners,
welded to the web about 1 ½ feet below the compression (top) flange, for buckling
stability. Each of the fascia girders also have longitudinal stiffeners welded to their
exterior face 1 ½ feet above their tension (bottom) flanges. These stiffeners provide
buckling stability over the piers, where the bottom flange will actually be in compression.
A cross-sectional sketch, numbering the girders G1 through G12 going southbound to
northbound, is provided in Figure 2 below.

I 9 5 S O U T H B O U N D I 9 5 N O R T H B O U N D

G1 G2 G3 G4 G5 G6 G7 G8 G9 G10 G11 G12

L O O K I N G N O R T H

Figure 2. Typical cross-section of the Brandywine River Bridge

4
Brandywine River Bridge Forensic Analysis 8/8/03

Description of the Fracture

The crack found in the fascia girder on the bridge’s northbound span, referred to
as G12, was in the middle of the center span. The vertical crack penetrated the lower
seven feet of the web and completely severed both the tension flange of the girder and the
longitudinal stiffener in the lower part of the web. The crack gap had reached a
maximum width of half an inch. The cracked section of G12 was therefore carrying no
significant amount of load.
Upon inspection, the origin of the crack was determined to be the full penetration
groove weld (or butt weld) connecting adjacent ends of the lower-web longitudinal
stiffeners. The most probable cause was thought to have been incomplete fusion in the
weld. Tension present in the stiffener, for which it was not designed, helped pull the
weld apart. Also, traffic redirected to the shoulder during previous bridge construction
could have put unexpectedly high loads directly on G12 and aided the more sudden
stages of crack propagation. Inspection of the fractured surface revealed a small, smooth
fracture surface in the suspected weld, indicating a possible fatigue fracture. The
photograph below shows a cored out section of the longitudinal stiffener at the crack
origin. The smooth fatigue area displayed the greatest amount of rust, suggesting that it
had been exposed the longest and was indeed where the fracture started.

Figure 3. Side view of the fracture

5
Brandywine River Bridge Forensic Analysis 8/8/03

Figure 4. Cored section of the longitudinal stiffener

Case Similarity: The I-91 Quinnipiac River Bridge

In November 1973, a similar crack was discovered in an I-91 bridge spanning the
Quinnipiac River near New Haven, Connecticut. Opened in 1964, the bridge is four-span
continuous with a 220-foot center span. The midspan includes a suspended girder section
attached to cantilevered sections that are haunched at the concrete piers. Longitudinal
stiffeners were welded to the web in a similar pattern as the Brandywine River bridge.
The vertical crack occurred in the south fascia girder of the suspended section of the
midspan. The crack had penetrated the lower five feet of the web as well as the lower-
web longitudinal stiffener but had not broken through the tension flange.
Inspections of the damaged bridge determined that, identical to the Brandywine
case, the crack had originated in the full penetration groove weld made across the width
of the lower-web longitudinal stiffener (Fisher, 1984). Investigators cited incomplete
fusion in the weld as the primary defect leading to fatigue fracture. The sudden stage of
fracture was believed to have occurred during the exceptionally cold winter of 1972-73,
when subzero temperatures would have reduced the steel’s ductility (Fisher, 1984). This
case helps confirm the prognosis for the source of failure in the Brandywine River Bridge.
Both bridges experienced conditions that would have accelerated the fatigue and fracture
process. Together, the two cases suggest that the installation of a non-functional
longitudinal stiffener in a tension zone for aesthetic reasons is a hazard to the bridge’s
structural integrity.

Bridge Retrofit
Prevention of further cracking in the bridge became an immediate concern after
the initial crack was discovered. Holes were drilled at the propagating tip of the primary
crack and fitted with pre-tensioned bolts in order to arrest the crack’s growth. Engineers
inspected the bridge for ways to prevent similar cracks from forming elsewhere on the

6
Brandywine River Bridge Forensic Analysis 8/8/03

bridge. As a result of this inspection, all of the butt welds in the aesthetic lower
longitudinal stiffeners in the fascia girders were drilled out perpendicularly to the web
from the backside of the girder. Because these stiffeners had no structural function,
removal of their connections to one another has no adverse effect on the bridge’s
performance.

Temporary Repair and Evaluation

Steel splice plates were bolted to the top and bottom of the severed tension flange
to temporarily repair the cracked section. The girder could now carry stresses caused by
live traffic loading, though it was still not carrying any significant dead load stresses.
Field tests, performed by University of Delaware technicians and faculty, were then
conducted to determine if the temporarily repaired bridge could adequately withstand two
lanes of traffic loading. The primary objective was to assess the temporary splice plate
repairs on the northbound span using strain measurements and compare the results to
similar measurements taken on the undamaged southbound span (Chajes et al, 2003).
The effects of traffic loading were simulated by driving three heavy trucks over
the bridge at crawl speed. For each trial, the trucks drove over the same location on the
northbound and southbound spans. Six different trials, each inducing different amounts
of strain in each of the girders, were conducted for comparison. Strain measurements
were taken in girders 9 through 12, instrumented prior to testing with strain gauges near
the center of their midspan, and in their sister girders in the undamaged southbound span.
Results of testing revealed that both the undamaged and the damaged bridge spans
were experiencing a similar amount of strain due to the truck passes. The strain readings
from the northbound and southbound never differed by more than 15 microstrain (Chajes
et al, 2003). It was clear that the temporary splice plate was absorbing traffic load and
was an adequate solution to the damage until permanent repairs could be made.

PERMANENT REPAIR
Permanent splices could not be immediately installed onto the cracked section.
The crack in G12 had allowed the superstructure to sag a couple of inches below its
undamaged height. The bridge would need to be lifted back to its neutral position in
order to close the crack and allow the girder to again carry dead load. Strain data taken
during this process would give some indication as to how stress had been redistributed
through the girders when G12 fractured. The inverse of the stress response caused by
lifting the bridge back into place should indicate how much stress has gone in and out of
the girders during fracture.

Jacking Procedure and Repairs

Two temporarily constructed steel truss tower were installed directly under the
cracked section. Each tower was anchored into four pile cylinders containing earth and
concrete. A temporary earthen causeway was put into the river to provide access for
heavy equipment to the towers. Each tower supported a 100-ton hydraulic jack which
was used to lift the bridge from its damaged position. The jacks pushed directly on G12,
making contact with the girder about 8 feet each side of the crack. Figure 5 below shows
the towers following their installation under G12.

7
Brandywine River Bridge Forensic Analysis 8/8/03

After the jacks were seated onto the girder, the lift procedure was conducted in
three steps. First, the bridge was jacked from its damaged position back to its undamaged
dead load position. The temporary splice plates were removed prior to this step. Timber
vertical stiffeners had been inserted between flanges of G12 over the jack contacts to
reduce buckling. At this point, the diaphragm connecting G12 and G11 at their midpoints
was disconnected. In the second step, the bridge was lifted to its estimated skeleton
position, i.e. the bridge height had there been no concrete deck to provide dead load.
Permanent splice plates were then bolted to both sides of the web, flange, and
longitudinal stiffeners. Figure 6 displays these splice plates during their installation as
well as the timber stiffeners. The splices were installed at the skeleton position so that
the parts would be in tension under the dead load of the deck, similar to the original
girder. During the third and final step, the jacks were slowly lowered and the bridge was
released to assume its new permanent position. Before the bridge was finally released,
the G12 to G11 diaphragm was reattached. Detailed documentation of the entire jacking
procedure, including the girder elevations at each step, is available in Appendix A.

Figure 5. Steel trussed jacking towers

8
Brandywine River Bridge Forensic Analysis 8/8/03

Figure 6. Permanent steel splice plates

Data Collection
Girders 9 through 12 were instrumented with an array of strain gauges in order to
monitor their strain response to the jacking process. Each gauge is identified with a three
digit number. The gauge setup for each step of the jacking procedure is available in
Appendix B. While the overall configuration remained unchanged from step to step, the
removal/installation of splices and the repositioning of equipment required that some
gauges be repositioned. Of primary concern would be the strain readings taken from the
bottom flanges of each girder, indicating the load-bearing distribution in the girders.

RESULTS
Strain data was taken during each portion of the jacking process. For each stage,
including Steps 1 through 3, several increments of jacking occurred. Thirty one total
strain data files were recorded for this investigation. In addition, measurements of the
strain induced by ambient truck traffic were made during steps in which the jacks were
not moving. The complete set of strain data obtained during each individual jacking step
(JS) is provided in Appendix C. The strain measurements have been arranged graphically
versus time, tracing the strain response of each girder chronologically in the jacking
sequence. Each plot shows the strain readings taken from the bottom flanges of girders 9
through 12. Positive gage readings denote tension, and negative readings denote
compression. The pertinent strain gages for this report were installed on the bottom
flange at the span’s midpoint and would register positive tension values under typical
bending. Note that, during jacking, negative strains correspond to a release of strain
when the bridge is lifted upward; positive strains indicate that the bridge is lowering and
tension is building in the girder. The cumulative strain removed from and applied to the
girders is found by adding the endpoints of each steps’ strain measurements.

9
Brandywine River Bridge Forensic Analysis 8/8/03

Problems with Data File JS-26

Almost all of the captured strain data plotted cleanly enough to indicate how
much total strain the girder had sustained in that jacking step. However, problems were
experienced when trying to plot the data file corresponding to JS-26. The data file had
inadvertently overwritten another file when it was downloaded from the laptop retrieving
data from the gauges. Only broken pieces of the original file remained. The last 50
seconds of the trial had been completely lost. However, the first 55 seconds of the trial
were completely recovered (Fig. 1, App. D).
The remainder of JS-26 was reconstructed using three means of averaging. The
strain to jacking pressure (psi) ratios from the other jacking data was extrapolated using
the known pressure that the jacks had reached during JS-26 to find the strain (Fig. 2, App.
D). Linear and logarithmic curves were then fitted to the recovered JS-26 strain curve
(Fig. 3-4, App. D). By plugging the final time into the curve equations, the final strains
for each fit was calculated. These values were averaged with the extrapolated strain
value to yield a reasonably accurate estimation of how much strain the girders absorbed
in JS-26. Figures 5 through 8 in Appendix D show the three strain curve estimations for
girders 9 through 12, respectively.

Jacking Data Results

The strain measurements taken during each step in which the jacks were moving
have been compiled in Table 1. The amount of force applied to the girder, the strain
induced in the girder, and the strain per ton are included. The total strain removed in the
first two steps from girders 9, 10 and 11 are -54.5 µε, -131.0 µε, and -295.7 µε,
respectively. G12’s strain readings during the first two steps are not important because
permanent repairs had not been installed and the bottom flange was severed. The strain
absorbed by girders 9, 10, 11, and 12 during the Step 3 jack release following permanent
repair are 24.8 µε, 66.8 µε, 146.5 µε, and 185.5 µε, respectively.

10
 Brandywine River Bridge Forensic Analysis 8/8/03

Flange Strain (µε) Strain per Ton

Jacking Step Jacking Force (tons) 9 10 11 12 9 10 11 12
JS-1 10.0 -2.6 -9.7 -18.9 -- -0.3 3.7 1.9 --
JS-2 0.0 2.3 2.8 6.5 -- 0.0 0.0 0.0 --
JS-4 0.0 -1.8 -0.6 1.2 -- 0.0 0.0 0.0 --

JS-6 20.0 -7.3 -27.7 -63.5 -- -0.4 -1.4 -3.2 --

JS-7 10.0 -6.4 -19.4 -40.5 -- -0.6 -1.9 -4.0 --
JS-8 10.0 -8.8 -16.1 -47.7 -- -0.9 -1.6 -4.8 --
JS-9 10.0 -3.8 -14.1 -35.9 -- -0.4 -1.4 -3.6 --
JS-10 5.0 -4.4 -12.5 -32.2 -- -0.9 -2.5 -6.4 --
JS-11 0.0 3.5 8.9 20.5 -- 0.0 0.0 0.0 --
JS-12 0.0 -0.6 0.6 -0.6 -- 0.0 0.0 0.0 --
Step 1 Total 55.0 -27.7 -80.3 -199.8 -- -0.6 -1.8 -4.4 --
JS-15 10.0 -5.3 -15.5 -35.3 -- -0.5 -1.6 -3.5 --
JS-16 5.0 -4.7 -10.5 -22.9 -- -0.9 -2.1 -4.6 --
JS-17 10.0 -8.2 -16.1 -33.2 -- -0.8 -1.6 -3.3 --
JS-18 5.0 -9.9 -11.6 -16.7 -- -2.0 -2.3 -3.3 --
JS-19 5.0 -4.1 -8.0 -14.6 -- -0.8 -1.6 -2.9 --
JS-20 0.0 5.3 11.1 26.8 -- 0.0 0.0 0.0 --
Step 2 Total 35.0 -26.8 -50.7 -95.9 -- -1.0 -1.8 -3.5 --
Steps 1+2 100.0 -54.5 -131.0 -295.7 -- -0.8 -1.8 -4.0 --
JS-25 70.0 -4.1 -8.3 -15.8 -39.6 -0.1 -0.1 -0.2 -0.6
JS-26 -70.0 22.2 62.7 111.9 187.8 -0.3 -0.9 -1.6 -2.7
JS-27 0.0 -2.3 -2.2 -2.7 -20.6 0.0 0.0 0.0 0.0
JS-28 35.0 -4.4 -25.8 -28.1 -59.4 -0.1 -0.7 -0.8 -1.7
JS-29 -35.0 13.4 40.5 81.2 117.3 -0.4 -1.2 -2.3 -3.4
Step 3 Total 0.0 24.8 66.8 146.5 185.5 -0.2 -0.7 -1.2 -2.1
Steps 1+2+3 100.0 -29.7 -64.2 -149.2 185.5 -0.6 -1.5 -3.2 -2.1

Table 1. Jacking force and strain

The cumulative amounts of strain removed from and absorbed by the girders in
the Steps 1 through 3 can be simply converted to cumulative stress using σ = εE, where E
equals 29000 ksi. Table 2 shows the stepwise stress and strain in girders 9 through 12
along with final totals. Also shown are the surveyed changes in elevation (deflection) of
G12 for Steps 1 through 3.

G12 Defl. (in) Total Strain (µε) Stress (ksi)

Sequence Start Finish 9 10 11 12 9 10 11 12
Step 1 1.0 2.0 -27.7 -80.3 -199.8 -- -0.8 -2.3 -5.8 --
Step 2 2.0 3.1 -26.8 -50.7 -95.9 -- -0.8 -1.5 -2.8 --
Step 3 2.9 1.1 24.8 66.8 146.5 185.5 0.7 1.9 4.2 5.4
Final Total 1.0 1.1 -29.7 -64.2 -149.2 185.5 -0.9 -1.9 -4.4 5.4

Table 2. Cumulative jacking deflection, strain, and stress

11
Brandywine River Bridge Forensic Analysis 8/8/03

DISCUSSION
The strain and stress values shown in Table 2 illustrate the load distributions
among the girders. The values from Step 1 indicate how loads were distributed when
G12 was cracked. G11, directly adjacent to the damaged girder, appeared to carry nearly
6 ksi beyond its original dead load when G12 was cracked. Considering the accuracy of
the data, G10 absorbed roughly half as much stress, 2.3 ksi, as G11, and G9 likewise
absorbed roughly half a much stress, 0.8 ksi, as G10. The condition of G11 became a
concern because it had assumed a large percentage (about 65%) of G12’s extra dead load.
The values from Step 2 demonstrate the removal of the concrete deck dead load
from directly over G12. Again, the stress release diminished from by roughly one-half
toward the interior girders, starting with 2.8 ksi in G11, 1.5 ksi in G10, and 0.8 ksi in G9.
In theory, the girders would reassume these corresponding stress magnitudes when
released back to the dead load position following Step 3.
In reality, the girders adjacent to G12 absorbed more stress than expected. G11
absorbed an extra 1.4 ksi beyond its previous dead load stress, with G10 absorbing an
extra 0.4 ksi. These values are obtained by either adding Steps 2 and 3 or combining the
values for Step 1 to the final total. This could indicate a slight decrease in load-bearing
capacity in these girders, especially in G11, resulting from G12’s fracture and
transmission of dead load. However, the stress increases were not large enough to
warrant further repairs. The 5.4 ksi absorbed by the repaired G12 was well below the
stress limit of the permanently installed splice plates; therefore, the repairs could be
considered successful.
The surveyed changes in elevation confirmed what the stress and strains had
already suggested. According to the deflection values, G12 sagged about 1” when it
cracked and should have settled 1.1” below its skeleton position when released during
Step 3. Instead, G12 deflected 1.8”, nearly quarter of an inch lower than expected,
possibly indicating some slight losses in structural capacity. Once again, these
indications were small enough to not be considered a problem.

Computer Modeling of the Brandywine River Bridge

A three dimensional computer model of the Brandywine River Bridge was
constructed using the STAAD structural analysis program. The focus of efforts to model
the bridge was to confirm that the stresses relieved during Step 1 of the jacking procedure
were actually the stresses absorbed by girders 9, 10 and 11 when G12 fractured.
DMJM+Harris consultants had estimated that roughly 75% of G12’s dead load had been
absorbed by G11. Behavior of a model simulating a crack in G12 evaluates the validity
of this estimate (i.e. did G11 assume 75% of G12’s dead load stress or was G12’s dead
load transmitted more evenly to the other girders).
The model mirrored the bridge’s three continuous spans, resting on four pinned
supports. Each girder was modeled using twenty-six tapered beam sections of varying
depth and flange thickness in order to model their haunched shape. STAAD was unable
to model a concrete deck resting on top of the steel girders because it uses one-
dimensional elements running in the same plane; a separate deck would have been
running through the center of the girder cross-section. Therefore, the stiffness provided
by the concrete deck was accounted for by transforming the section of concrete above
each girder’s top flange into an equivalent steel section. This transformed section and the

12
Brandywine River Bridge Forensic Analysis 8/8/03

top flange could then be converted into one steel element of equivalent area with the
same neutral axis. This new cross section could easily be entered into STAAD.
Eight-foot deep steel diaphragms were inserted between the nodes of adjacent
girders to provide transverse stiffness. Lengthwise, weightless, 8-inch thick concrete
plates were attached at either end of the bridge and at the midpoint to accurately model
the ability of the bridge to distribute stresses. The dead load provided by the concrete
deck was concentrated into linear distributed loads running lengthwise on top of the
girders.

Figure 7. STAAD model of the Brandywine River Bridge

The accuracy of the model was validated by subjecting it to similar live load
conditions created during the diagnostic load tests. The strain induced from a 73-kip
vehicle driving in the rightmost lane of the undamaged southbound span had been
documented in the initial field testing report. These measurements were used to calculate
the stress increase in each of the girders before the crack. Table 3 shows the stress
responses of the modeled and actual girders. Because a similarly positioned load
produced a similar increase in stress in the model bridge girders, the STAAD model is
reasonable and can be used to predict damaged bridge behavior.

Stress (ksi)
Load Cases 9 10 11 12
Model Dead Load 6.65 6.66 6.93 7.12
Model Dead + Live Load 7.39 8.21 8.89 8.76
Model Live Load Increase 0.74 1.55 1.96 1.64
Actual Stress Increase 0.67 1.09 1.21 1.02

Table 3. Model verification using live load comparisons

The crack in G12 eliminated the girder’s ability to carry moment and reduced its
shear transfer capacity at its midpoint. Force release commands in STAAD were able to
release the moments and/or shear on either side G12’s center node. Two model scenarios
were tested: (1) G12 carried no moment but still carried shear at its midpoint, and (2)
G12 carried no moment and no shear at its midpoint.
The resulting girder deflections from each case are compared in Table 4 to the
actual deflections of girders 11 and 12 during Step 1. The modeled deflections are
roughly half the magnitude of the actual deflections. This suggests that the models are in

13
Brandywine River Bridge Forensic Analysis 8/8/03

fact behaving stiffer than the real bridge. In Table 5, the deflections have been
normalized by dividing each value by the greatest deflection, at G12. All three deflection
distributions are similar, indicating that the models can relatively mirror the real behavior
of the bridge with some accuracy. The model with both moment and shear released
showed a deflection distribution that was more gradual than the model with only moment
released.

Downward Deflection (in)

Damaged Load Cases 7 8 9 10 11 12
Moment Release Only -0.19 -0.06 0.08 0.24 0.48 0.88
Moment and Shear Release -0.26 -0.07 0.13 0.36 0.63 0.89
Actual Distribution - - - - 1.12 1.86

Table 4. Downward girder deflection across the bridge’s midpoint cross-section

Downward Deflection as a Percentage of G12

Damaged Load Cases 7 8 9 10 11 12
Moment Release Only -0.22 -0.07 0.09 0.27 0.55 1.00
Moment and Shear Release -0.29 -0.08 0.15 0.40 0.71 1.00
Actual Distribution - - - - 0.60 1.00

Table 5. Downward girder deflection normalized with respect to G12’s deflection

The resulting stress redistributions from each case are compared to the stresses
released during Step 1 in Table 6 below. The first model case and the Step 1 data showed
a similar increase in G11’s stress. G11 absorbed 74% of G12’s dead load stress in the
first model case and absorbed 81% during Step 1, suggesting that DMJM+Harris’
estimates were reasonably accurate. However, the first model case failed to show any
significant stress redistribution to girders 9 and 10 as Step 1. The second model case
showed significantly less stress increase in G11 (taking only 29% of G12’s dead load) but
managed to establish a stress redistribution closer to that of Step1. In both modeled cases,
the magnitudes of stress increase were less than the actual stress increases. The model
bridges were most likely stiffer than the real bridge in these loading cases and were
unable to fully transfer stress.

Stress (ksi)
Damaged Load Cases 9 10 11
Moment Release Only -0.31 0.37 5.28
Moment and Shear Release 0.29 1.77 2.04
Actual Distribution 0.8 2.3 5.8

Table 6. Comparison of damaged load scenarios

Together, the models begin to describe that the girder deflections and the stress
increases experienced during Step 1 were not unreasonable. A better model, however,
would eliminate moment and reduce the shear transfer. Such a model would allow the

14
Brandywine River Bridge Forensic Analysis 8/8/03

fractured G12 to transmit a large amount of its dead load stress adjacently to G11 but also
induce stress increases in girders 9 and 10.

Ambient Traffic Load Analysis

Strain data taken during the “ambient traffic” jacking steps help indicate how the
bridge is responding to the live loads of heavy trucks. The ambient traffic data sets were
perused for sizeable and sudden strain spikes, suggesting the presence of a heavy truck.
The strain values for girders 9 through 12 are recorded for that point in time and plotted
transversely. These transverse strain distribution plots in Appendix E show which girder
was experiencing the most strain and was directly under the moving load for each
recorded strain spike.
These strain distributions can be directly compared with measurements of the live
load strain distributions taken during the initial field testing. Truck passes directly over
G9 and directly over G10 were recorded for both the damaged northbound and
undamaged southbound spans (Chajes et al, 2003), during Steps 1 and 2, and after the
permanent repair of the fracture. Because the trucks were not all the same weight in
every case, the data was normalized by percentages. For truck passes directly over G9,
the strain at G9 was considered 100%, and the strain in girders 10, 11, and 12 were
divided by G9’s value, expressing them as a fraction of the greatest strain. The same was
done for the truck passes directly over G10. Single values for the five stages were
obtained by averaging the percentage distribution from each spike during that stage
describing a similarly positioned truck passing.
Figures 9 and 10 in Appendix E show the transverse percent-strain distributions in
the five stages for truck passes over G9 and G10, respectively. As expected, both plots
indicate that G12 carried more strain in the damaged, unrepaired northbound span than in
the undamaged southbound span. This suggests that the sag and the loss of structural
rigidity in the damaged girder allowed similar loads to cause greater deflection and thus
greater strain. Also, note that during Steps 1 and 2, G12 carried the least strain because it
was supported by the jacking towers and had little room to deflect.
Figure 9 shows that G12 in its repaired condition would absorb an amount of
strain between that of its undamaged and damaged state for truck loads over G9. This
could indicate that some, but not all, of G12’s capacity to resist live load strains had been
restored through repair. However, Figure 10 shows that the repaired G12 displayed the
same strain response as the damaged northbound G12 to truck loads over G10. Though
the strain measurements on which they are founded lack precision, these transverse
distributions suggest at the very least that G12 has permanently lost a small amount of its
original capacity to resist strain under traffic loading.

SUMMARY AND CONCLUSION

This report has presented the details of a forensic analysis of a fractured steel
girder in the northbound span of the I-95 Brandywine River Bridge near Wilmington, DE.
The fracture was put into context by discussing the fundamentals of steel fracture
mechanics, providing a description of the bridge, and detailing the crack characteristics in
G12. Previous analysis established that the crack had resulted from incomplete fusion in
the weld connecting the tension zone longitudinal stiffener to the web. Following retrofit
and temporary repair, initial field testing evaluated the effectiveness of the temporary

15
 Brandywine River Bridge Forensic Analysis 8/8/03

repairs. These studies provided the groundwork for current research regarding the
permanently repaired bridge condition.
Analysis of strain data collected throughout the repair process indicated how
much stress the other girders absorbed when G12 fractured. G11 assumed the majority of
G12’s dead load stress, and girders 10 and 9 assumed proportionally less of the remaining
released stress. Computer modeling attempted to confirm the field results. The STAAD
models offered some insight into the load redistribution, but the models were unable to
precisely mirror cracked behavior. Analysis of stresses caused by ambient traffic loads,
compared to initial field test results, indicated that the bridge may have a somewhat
different load distribution following permanent repair.
Results of this research contribute to knowledge of bridge behavior in a damaged
condition. Hopefully, this discussion will help other researchers and consultants in their
analysis of other damaged bridges. Future research on this bridge could focus on
determining how stresses were actually distributed throughout the girders when G12
fractured. Development of a more realistic computer model would make possible more
accurate predictions of the stress response to structural damage. Future monitoring of the
bridge will indicate how effectively the repairs will resist dead and live loads, as well as
the effects of fatigue, in the long term.

ACKNOWLEDGEMENTS
This material is based on work supported by the National Science Foundation
under Grant No. EEC-0139017, “Research Experiences for Undergraduates in Bridge
Engineering,” at the University of Delaware. I would like to acknowledge Dr. Michael J.
Chajes, Chairman of the Department of Civil Engineering, for his guidance and support
throughout the program. Special thanks also go to Gary Wenczel for his help with the
field testing and to Diane Kukich for organizing the program and aiding the creation of
this report. I would like to thank the Delaware Department of Transportation for
providing me with the plans and information for the bridge. Also, I am grateful to John
Milius and David Hanly from DMJM+Harris for their correspondence on the details of
the fracture.

REFERENCES
Askeland, D. R., and Phule, P. P. (2003). The Science and Engineering of Materials:
Fourth Edition, Brooks/Cole and Thomson Learning, Pacific Grove, Calif.

Chajes, M. J., Mertz, D. R., Shenton, H. W., III, Richardson, D., and Wenczel, G.
(2003). “Field Test of the I-95 Bridge Over the Brandywine River.” Report to
the Delaware Department of Transportation, University of Delaware, Newark, Del.

Fisher, J. W. (1984). Fatigue and Fracture in Steel Bridges: Case Studies, John Wiley
& Sons, New York.

Gurney, T. R. (1968). Fatigue of Welded Structures, Cambridge University Press,

London.

16

View publication stats