Economical Steel Plate Girder Bridges

RICHARD P. KNIGHT

As a service to the bridge design profession, Bethlehem Steel up to around 200 ft (61 m), representing a majority of the
has conducted hundreds of studies of steel plate girder bridge population. Continuity and composite design are
bridges using its Preliminary Bridge Girder Optimization assumed because these features combine to save 20% or
Program. The program is unique because it optimizes plate more in main girder cost compared to simple span non-
girder designs on the basis of least cost, not least weight. An composite designs. Therefore, only continuous composite
evaluation of many study results leads to a number of designs will be discussed.
guidelines which are offered in this paper. When applied by
the bridge designer, these guidelines should result in PRELIMINARY BRIDGE GIRDER
economical steel plate girder designs for continuous OPTIMIZATION STUDIES
composite bridges with span lengths of up to approximately
Most of the guidelines are based on the hundreds of studies
200 ft (61 m), representing a majority of the bridge
accomplished with Bethlehem Steel's Preliminary Bridge
population.
Girder Optimization Program over more than a decade.
Comparisons of various parameters are made on a cost
Certain material and fabrication costs are built into the
index basis. These subjects are discussed: load factor vs.
program, allowing comparisons to be made on a cost index
working stress design, weathering steel, painted high-strength
basis. The cost indices cited are based on the fabricated main
steel, number of girders in a cross section, optimum web
girder material only. Other components, such as diaphragms,
depth and thickness, transverse vs. longitudinal web
bearings, deck slab, etc. would have to be evaluated before
stiffeners, flanges and flange splices, plus other
selecting a final superstructure design. Also, it may be noted
considerations leading to cost effective steel plate girders for
that the difference between cost indices sometimes seems
bridges.
small. This is because all comparisons are between designs
In the design of steel plate girder bridges, it used to be
which are optimum for their respective parameters.
sufficient to determine a least weight solution and develop it
as a complete set of plans for competitive bidding. That was
DESIGN METHOD
believed to be the most economical design. However, over
the past several years, many improvements have been made Within the 200-ft (61-m) span range, Load Factor Design
in design and analysis methods, materials and construction (LFD) produces savings of 5 to 12% compared to Working
techniques, obliging the designer to consider an increased Stress Design (WSD). For spans longer than 200 ft (61 m),
number of options. Today's heightened competition due to LFD saves from 12 to 20%. This increasing cost advantage
alternate bidding practices, including contractor sponsored of LFD as spans become longer is a function of the increase
alternates, makes it further incumbent upon all concerned in the proportion of dead load to live load as spans increase
with steel bridge design and construction to always strive for in length. Dead loads can be predicted with greater certainty
the most cost-effective solution. than live loads. Therefore, with LFD, dead loads carry a
lower factor of safety than live loads. LFD is the clear
ECONOMICAL GUIDELINES economical choice compared to WSD.
This paper will present some features of economical steel However, some states have used a higher HS live loading
plate girder bridges and offer some guidelines. Application of for LFD than for WSD, or they have increased the beta factor
these guidelines by the bridge designer will, in most cases, above the value in the standard AASHTO formula. The cost
result in economical steel plate girder designs for continuous for using such modified Load Factor Designs is about the
composite bridges with span lengths same as for Working Stress Designs. The LFD method should
not be modified through increases in live loading or beta
factors unless such increases represent a decision to upgrade
Richard P. Knight is an Engineer in the Construction Marketing loadings in general without regard to design method. Using
vision, Bethlehem Steel Corporation, Bethlehem, Pennsylvania. greater loadings for LFD than for WSD effectively nullifies
This paper was presented at the National Bridge Conference, June the cost advantage of LFD which for equivalent conditions is
1, 1983. the clear economical choice over WSD.

ECOND QUARTER / 1984 89

© 2003 by American Institute of Steel Construction, Inc. All rights reserved. This publication or any part thereof must not be reproduced in any form without the written permission of the publisher.
 STEEL GRADES NUMBER OF GIRDERS IN CROSS SECTION
Unpainted Weathering Steel Substantial savings can be achieved by minimizing the
number of girders in a cross section commensurate with the
The most cost-effective choice of steel grade is unpainted
overall economy of the total superstructure system. For
ASTM A588 weathering steel used in appropriate
example, Fig. 1 represents a structure with 11 girders spaced
environments. The cost advantage of unpainted weathering
at 7 ft-6 in. (2.29 m) supporting a wide multilane divided
steel designs compared to painted high-strength steel designs
roadway. If, as in Fig. 2, three girders are eliminated and the
can range from 2 to 11% for the main girders, even though
spacing of the remaining eight girders is increased to 10 ft-8
the unit cost of A588 is the highest of the three commonly
in. (3.25 m) a savings of approximately 8% to 13% in main
used bridge steels. Unpainted weathering steel is less
girder cost would result. Such savings might be partially
expensive, on a first cost basis, than painted designs without
offset if the deck thickness had to be increased to
consideration of future maintenance painting. These cost
accommodate the wider girder spacing. However, the savings
differences are based on the use of an average paint system.
would be amplified further through use of fewer diaphragms,
If a more expensive, higher quality, longer life paint system
bearings etc. There would also be fewer girders to erect.
were specified, the cost advantage of weathering steel would
Therefore, selection of number of girders (girder
be greater. Therefore, the clear economical choice of steel
spacing) is one of the most important influences on the
grade is unpainted A588 weathering steel.
economy of a plate girder bridge design. It is suggested 10 ft
(3.05 m) be considered the minimum spacing for economical
Painted High Strength Steel
results.
In environments which are inappropriate for unpainted
weathering steel, the most economical painted designs use WEB DESIGNS
high-strength steel. High-strength steel designs are less
expensive than ASTM A36 homogeneous designs by Transverse Stiffeners
approximately 6 to 10% for the main girders. They are less
expensive than mixed designs by approximately 3 to 5%. A In all cases, transverse stiffeners should be placed on only
typical mixed design uses high-strength steel in the negative one side of the web, except at diaphragm connections of
moment regions and A36 in the positive moment regions. interior girders, where they are needed on both sides.
Each field section is homogeneous with respect to yield
strength, but the yield strengths differ from field section to Definitions
field section. In the following discussion, girder webs having transverse
In this discussion, "high-strength steel design" means stiffeners only at diaphragm connections are defined as
hybrid or 50 ksi (345 MPa) homogeneous. Hybrid designs unstiffened. A nominally stiffened web is defined as having a
usually use A36 steel in the webs and A36 or 50 ksi (345 thickness 1/16 in. (1.6 mm) less than the unstiffened web.
MPa) steel in the flanges. The 50 ksi (345 MPa) The thinnest web allowed by AASHTO with a maximum
homogeneous designs are ASTM A572 Gr. 50 throughout, number of transverse stiffeners is defined as fully stiffened.
except A588 is used where the maximum 2-in. (50.8-mm)
thickness of A572 Gr. 50 is exceeded. The switch to A588 Web Optimization
for thicknesses over 2 in. (50.8 mm) will not be necessary
The web designs which the Bethlehem program iterates trade
once AASHTO has adopted the newly revised ASTM A572
off plain material cost vs. the fabrication cost of applying
Gr. 50 specification which maintains 50 ksi (345 MPa) yield
stiffeners. For a typical input, the program investigates
through 4 in. (101.6 mm).
girders with a practical range of web depths in specified
There is usually a slight advantage (1% to 2%) with
increments. Within each depth, it iterates the web thickness
hybrid designs compared to the 50 ksi (345 MPa)
by 1/16-in. (1.6-mm) increments. Web/stiffener designs are
homogeneous designs for span lengths less than 200 ft (61
not generated by themselves, but are an integral part of an
m). For spans of 200 ft (61 m) or more, 50 ksi (345 MPa)
overall girder design which takes into consideration web
homogeneous designs are slightly favored. Again, these
depth and thickness, flange sizes, stiffener location and size,
differences seem small from a percentage standpoint, but it
material cost and fabrication cost. The optimum overall
should be remembered the solutions being compared are
girder design is assigned a cost index of 1.00.
optimum for their respective parameters.
For spans up to approximately 200 ft (61 m) it is
Web Depth
suggested that hybrid be considered the standard painted
design. However, the clear economical choice overall is A plot of the cost index for the optimum girders vs. several
unpainted weathering steel. web depths (Fig. 3) shows that depth variations either side of

90 ENGINEERING JOURNAL / AMERICAN INSTITUTE OF STEEL CONSTRUCTION

© 2003 by American Institute of Steel Construction, Inc. All rights reserved. This publication or any part thereof must not be reproduced in any form without the written permission of the publisher.
 Fig. 1. Note: 1 ft=0.30m

Fig. 2. Note: 1 ft=0.30m

the overall optimum depth are only slightly more costly for Figure 4 shows how girder cost varies with web thickness
both hybrid and homogeneous solutions. This is beneficial in for a given depth. The difference in cost between web
situations where the optimum depth girder cannot be used, thicknesses varied by 1/16-in. (1.6-mm) increments is relatively
such as having to limit structure depth to meet small as the curves are traced in the direction of the thicker webs
underclearance requirements. For whatever reason, for from their lowest, or optimum, points. Figure 4 is intended to
hybrid or homogeneous designs, web depths may be demonstrate that, for spans up to 200 ft, a designer can select a
increased or decreased several inches from the optimum web thickness 1/16 in. thinner than the unstiffened web and rest
without significant cost penalty. assured that he has not incurred undue cost penalty if his
selection does not happen to be the actual cost optimum. For
Web Thickness purposes of this paper, the only way to determine the actual cost
Very seldom does the Bethlehem program determine an optimum is through use of the Bethlehem program.
overall optimum solution having an unstiffened or fully The dissimilarity in the shape of the two curves in Fig. 4
stiffened web. The optimùm usually falls somewhere is linked to tension field action which is not permitted in the
between. For spans up to approximately 150 ft (45.7 m), the load factor design of hybrid girders. This is why Fig. 4
web thickness for the optimum solution is usually 1/16 in. indicates a thicker web for the optimum hybrid girder than
(1.6 mm) thinner than the unstiffened design, requiring for the optimum homogeneous girder. The thinner optimum
nominal stiffening. As span lengths increase, the optimum web thickness occurs on the homogeneous girder because
web thicknesses tend to be thinner than the unstiffened design tension field action may be included when evaluating web
by increments larger than 1/16 in. (1.6 mm). Sometimes the strength. This phenomenon is mentioned in the discussion
increment is as large as ¼ in. (6.4 mm). that follows on longitudinal stiffening.

Fig. 3. Note: 1 in. = 25.4mm Fig. 4. Note: 1/16 in. = 1.6mm

SECOND QUARTER / 1984 91

© 2003 by American Institute of Steel Construction, Inc. All rights reserved. This publication or any part thereof must not be reproduced in any form without the written permission of the publisher.
 Some designs use webs of constant thickness for the bottom flange of field sections up to 130-ft (39.6-m) long
entire length of the girder. Others use webs whose thickness should not exceed three (or the number of shop splices should
varies by field section (i.e., the web thickness is constant not exceed two). In some cases, especially the top flange in
within a field section, but varies from field section to field positive moment regions, a single flange plate size should be
section). The difference between constant and variable web carried through the full length of the field section. As a
thickness designs is small (only 1% or 2%), usually in favor general rule, an average of about 700 lbs. (318 kg) of flange
of variable webs. However, as a general rule, variable material should be saved to justify the introduction of a
thickness web designs are suggested. flange splice.
It generally does not pay to vary flange widths in a field
Longitudinal Stiffening section because of the cost of tapering the ends of the wider
flange plate to feather into the edges of the narrower plate.
Longer spans sometimes require deeper webs stiffened both
Also, many fabricators purchase wide plates, splice them and
transversely and longitudinally. For convenience, such
then strip the required flange plates. This practice results in
girders will be referred to as longitudinally stiffened, even
constant-width flanges. Therefore, flange widths should be
though it is understood they are also transversely stiffened. In
designed as constant for the length of the field section.
such cases, the longitudinal stiffeners should always be
placed on the opposite side of the web from the transverse
HAUNCHES
stiffeners. This minimizes the number of places where
longitudinal and transverse stiffeners intersect, such places Another form of least weight design is the haunched plate
occurring only at diaphragm or bearing stiffeners. girder. Some studies have shown that these designs are not
None of the Bethlehem studies show that longitudinal competitive with constant depth, or parallel flange, designs in
stiffening is economically justified in spans of 200 ft (61 m) spans of 200 ft (61 m) or less. In fact, in recent bidding, it
or less. In fact, longitudinally stiffened designs are not has been shown that they are not generally economical for
usually competitive with transversely stiffened designs until spans up to around 400 ft (122 m). One state invited bids on
span lengths approach 300 ft (91.5 m). The optimum haunched versus parallel flange designs for a four-span
longitudinally stiffened girders are deeper and, within the continuous structure with maximum span of 178 ft (54.3 m).
200-ft (61-m) span range, they weigh 1% to 12% less than All five bidders priced the parallel flange design below the
their optimum transversely stiffened counterparts, but they haunched design. In the longer span range, another state
cost more. The homogeneous designs cost from 1% to 3% advertised a haunched design and invited contractor
more while hybrid designs cost from 4% to 7% more. alternates for a river crossing with a maximum span of 420 ft
The larger cost differential for hybrid girders occurs (128 m). All steel bids were based on an alternate parallel
because AASHTO prohibits use of tension field action in the flange design. There were no bids on the original haunched
Load Factor Design of hybrid girders. The result is an design. The bridge is now under construction using parallel
increase in the number of stiffeners for the longitudinally flange plate girders. Therefore, it is suggested that haunched
stiffened design. This prohibition also causes the transversely designs be considered only when span lengths exceed 400 ft
stiffened hybrid design to have a thicker web than the (122 m).
corresponding homogeneous design. This phenomenon is
further aggravated by the requirement of Load Factor Design PRECAST DECKS
to consider interaction of shear and moment. The
To make steel plate girder bridges even more economical,
specifications as written suggest the designer check for
some consideration is now being given to designs using
maximum shear and maximum moment as occurring at the
precast, prestressed concrete decks which would make it
same point. It is hoped this provision will be clarified in the
possible to use a greatly reduced number of girders with very
near future.
wide spacings and overhangs. Figure 5 shows a four-girder
In any case, for reasons of economy, longitudinal
cross section featuring 16-ft (4.88-m) girder spacings and 9-
stiffeners should not be considered for span lengths less than
ft (2.74-m) overhangs. Comparing this to a conventional
300 ft (91.5 m).
multi-girder cross section using a spacing of 10 or 11 ft (3.05
or 3.35 m), the savings in main girder cost is in the range of
FLANGES
25% to 40%. Of course, the savings might be partially offset
A least weight design would require several changes of by additional deck costs, including attachments to produce
flange size to make the girder section properties closely composite action between the precast panels and the steel
approximate the moment curve. However, because of the cost girders. Precast panels have been used for deck replacement,
of making flange splices, this is not practical. Bethlehem but there is little bidding history for new construction.
studies show that, as a rule, the number of plates in the top or However, on one recently bid project, bids of a conventional-

92 ENGINEERING JOURNAL / AMERICAN INSTITUTE OF STEEL CONSTRUCTION

© 2003 by American Institute of Steel Construction, Inc. All rights reserved. This publication or any part thereof must not be reproduced in any form without the written permission of the publisher.
 Fig. 5. Note: 1 ft=0.30mm

deck/multi-girder design vs. a precast-deck/two-girder design 4. Designs should use the fewest number of girders
were within 3%, even though there were site constraints compatible with deck design and other factors. It is
working against the latter. This is encouraging for the future suggested a girder spacing of 10 ft (3.05 m) be
of precast decks and wide-girder spacings in new considered the minimum for economical results.
construction. Certainly this concept deserves serious 5. Transverse stiffeners (except diaphragm connections)
consideration. should be placed on only one side of the web.
6. Web depth may be varied several inches greater or
OTHER CONSIDERATIONS
lesser than the optimum without significant cost penalty.
Within the present state of the art there are possibilities for 7. A nominally stiffened web (1/16 in. or 1.6 mm thinner
improving economy, such as: (1) omission of bottom lateral than unstiffened) will be the cost optimum or very close
bracing as covered by new AASHTO empirical methods, (2) to it.
use of elastomeric bearings or pot bearings instead of
8. Designs with web thickness which varies by field
expensive, custom-fabricated steel rocker bearings and (3)
section are suggested.
use of composite construction in negative moment regions
over the piers. Even if negative moment regions are designed 9. Longitudinally stiffened designs should not be
as non-composite, there is a statement of AASHTO allowing considered for spans less than 300 ft (91.5 m).
inclusion of the whole deck slab in the section properties of 10. Use no more than three plates (two shop splices) in the
negative moment regions when calculating maximum live top or bottom flange of field sections up to 130-ft (39.6-
load deflections. It is possible that steel plate girder designs m) long. In some cases, a single-flange plate size should
have been made unnecessarily conservative because this be carried through the full length of the field section.
provision was overlooked. 11. An average of about 700 lbs. (318 kg) of flange
material should be saved to justify the introduction of a
SUMMARY flange splice.
Most of the guidelines developed in this paper are derived 12. Use constant flange widths within field sections.
from the many studies accomplished with Bethlehem Steel's 13. Haunched girder designs should not be considered for
Preliminary Bridge Girder Optimization Program. These most conventional cross sections until spans exceed 400
guidelines apply to spans up to 200 ft (61 m) in length, a ft (122 m).
majority of the bridge population, although some also apply
to longer spans. 14. Omit bottom lateral bracing where permitted by
AASHTO.
1. Load Factor Design (LFD) is more economical than
Working Stress Design (WSD). Modifying LFD by 15. Use elastomeric bearings or pot bearings in lieu of
imposing higher loads than on WSD nullifies the usual custom-fabricated steel bearings.
cost advantage of LFD. 16. Consider use of composite construction in negative
2. Unpainted A588 weathering steel is the most economical moment regions.
design. Properly designed in the appropriate Probably the most influential of these guidelines are use
environment, weathering steel bridges are more of (1) Load Factor Design, (2) unpainted weathering steel
economical than those requiring painting of the whole and (3) minimum number of girders in the cross section.
structure. These three should always be the first consideration.
3. The most economical painted design is hybrid. Painted 50 Application of all these guidelines in the design office
ksi (345 MPa) homogeneous designs are a close second. should lead to economical steel plate girder bridges.

SECOND QUARTER / 1984 93

© 2003 by American Institute of Steel Construction, Inc. All rights reserved. This publication or any part thereof must not be reproduced in any form without the written permission of the publisher.