Studies of Skewed

Concrete Box-Girder
Bridges
Mark R. Wallace, Division of Structures, California Department of
Transportation

Analyses of 51 mathematical models using the finite element analysis pro- portation researchers found the following skew effects
gram CELL were undertaken to describe the behavior of common skewed, (~,!):
concrete box-girder bridges. An assessment was made of effects on struc-
tural behavior of varying superstructure width to span length (aspect) ratio, 1. Significant reductions in longitudinal bending mo-
number of cells, skew angle, type of loading, and depth. Span length and
skew angle were found to be the most significant parameters influencing ments from those of comparable straight or normal
girder shear distribution. Aspect ratio and skew angle affected longitu- structures on orthogonal bearings,
dinal bending moments to the largest extent. Based on the results of this 2. Severe redistribution of reactions and girder
study, methods for shear and bending design of skewed bridges are pre- shears toward obtuse support corners,
sented. 3. Uplift tendencies at acute corners of the longest
span abutments under some loading conditions, and
Bridge engineers have recognized several peculiarities 4. No measurable benefits from including intermedi-
in skewed concrete box-girder bridge behavior. Shear ate diaphragms in the superstructure.
cracks have developed in exterior girders near obtuse
support corners. Transverse cracks have developed at Similar effects of skew have also been reported by others
midspan of certain highly skewed, prestressed bridge (Ei §..'.!.•~·
decks. Tendencies toward uplift at acute support corners
of long span bridges have been noted. OBJECTIVES, CONCEPTS, AND PROCEDURES
Although skewness produces analysis and design com-
plications, it also affords design advantages. Since sup- The objectives of the study reported in this paper were to
port reactions tend to concentrate at obtuse corners, re- describe the behavior of skewed box-girder bridges, and
ductions uccui· in eITective sll'ucture spans. The thick to suggest design criteria. The CELL program was used
end diaphragms act to reduce longitudinal stresses in the to study and analyze 51 mathematical models to assess
superstructure. Hence, possibilities exist for material effects of varying superstructure plan geometry, number
savings by refining structural analysis. of cells, skew angle, type of loading, and depth.
University of California researchers developed a fi-
nite element analysis program CELL for use in perform- Basic Concepts
ing elastic analyses of box-girder bridges having a con-
stant depth and an arbitrary plan geometry (1). They The obtuse corner reactions for any symmetrical loading
further used the program for analyzing skewed bridges. condition will exceed acute corner reactions when a
Results of the study enabled them to devise methods simple span, skewed box-girder superstructure on paral-
whereby the stiffening effects of skewed supports could be lel abutments is supported at its corners. Figure 1 shows
considered for reducing superstructure design moments this concept. A concentrated load P was applied at the
(~. Aluminum models were tested for physical verifi- center of a skewed plate {Figure la). Since the load was
cation (8). Dead-load dzsign moments were reduced applied nearer the obtuse corners, as opposed to the
19 percent for 30-deg skews and 41 percent for 45-deg acute corners, the obtuse reactions were larger. In other
skews. words, a structure will behave more stiffly between its
In an analysis of a highly skewed, two-span, two- obtuse corners than between its acute corners. As a
celled railroad bridge California Department of Trans- commonly accepted design approach, total abutment re-
action R is used to establish a moment diagram for a
structure. having an assumed span length equal to the cen-
ter girder length {Figure lb). In California design prac-
tice, a "whole-width" design concept is employed, in
which the moment of inertia of all girders comprising
Publication of this paper sponsored by Committee on Concrete Bridges. the superstructure is lumped in one unit.

50
 51

The above procedure, however, leads to a conserva- plot of a finite element mesh for a 4-celled box, which is
tive design because the skew itself has the effect of re- typical of all meshes used in the present study.
ducing the maximum moment in the structure. One
means of demonstrating this effect is to superimposetwo Loadings
conditions that are equivalent to the total left support re-
action (Figure le). The true reactions on the left may All model structures were analyzed for dead-load behav-
be described as a component that is normal (which sums ior and several for prestress and live-load behavior.
to R) and a component that is skewed or unbalanced. The
normal component affects longitudinal behavior exactly 1. Structure dead-load was calculated and applied in-
as a normal reaction affects a normal structure. The un- ternally by using the CELL program.
balanced component accounts for the portion of the reac- 2. Live-load behavior was studied through application
tion that is due to skew. It has the effect of applying a of loads appropriate to span lengths. Uniform lane loads
couple to the left support that causes a counterclockwise and concentrated midspan loads (riders) were applied to
rotation about the moment center (y-axis in Figure ld), spans exceeding 48.8 m (160 ft). Short s pan behavior was
opposing the clockwise rotation caused by the load. Thus, studied thro1.1gh application of on e truck (HS 20-44) or the
a resisting couple results from the unbalanced longitudi- maximum number of trucks a particular structure would
nal distribution of support reaction. Its magnitude is cal- be designed to carry.
culated as shown at the bottom of Figure ld. A clockwise 3. Effects of prestress loads were studied to relate
couple of the same magnitude occurs at the other end of results obtained by using the usual analysis and by using
the span. The magnitude of the couple is a function of the the CELL analysis. Skewed bridges having parallel sup-
reaction difference, structure width, and skew angle. ports are currently designed by assuming normal sup-
When the preceding two effects are superimposed, a mo- ports; therefore, working prestressing forces were cal-
ment is obtained at midspan that is reduced from the mo- culated for the normal condition. Figure 5 shows the pro-
ment of the simple span by the magnitude of the couple cedure. The forces were assumed to be applied through
as shown in Figure le and discussed elsewhere (9). cables draped as shown in Figure 5a. Equivalent load-
The above concepts also apply to box-girder bridges; ings, comprising a combination of concentrated loads at
however, they must be modified according to the follow- the girder ends and an upward load acting over the deck
ing indeterminate complexities of box-girder behavior. area, were applied. The equation in Figure 5b was used
to calculate the uniform load to balance the dead load.
1. The end diaphragm contributes measurably to the
stiffness of a skewed structure. As its span lengthens, RESULTS AND OBSERVATIONS
the support rotation increases. When this lengthening
occurs, the end diaphragm provides a good mechanism Moment Reduction
for longitudinal as well as transverse distribution of re-
action, carrying an increasingly greater load to the ob- Results indicate that substantial savings can be realized
tuse corner. Increases in skew angle accentuate this by taking advantage of the reduction of design stresses
redistribution. produced by the effects of skew. Figure 6 shows moment-
2. The unbalanced distribution of support reactions reduction design curves that could be used for arriving at
also produces torsional resisting moments at the sup- more economical designs for simple span, reinforced con-
ports, even under symmetrical loading. If moments are crete, box-girderbridges. These curves represent enve-
taken about the longitudinal centerline (x-axis) of the lopes of data from the models previously mentioned. The
structure, a torsional couple is obtained that depends on dotted lines indicate extrapolation beyond most of the data.
the reaction unbalance and structure width (Figure lf). Initial studies show that similar reductions would exist
for multiple span structures. The relative amounts
Most box-girder bridge sections are designed according carried by various spans depend on several factors, such
to the whole-width concept. However, because skewness as span unbalance and intermediate support conditions.
produces some doubt as to the desirable section for de- Varying superstructure depth-to-span length. (D/L) ratio
sign purposes, the critical moment value had to be de- within the range of 0.0475 to 0.0615 has little influence on
termined between that calculated at a total section, nor- moment reduction values.
mal to the center girder midspan, and that calculated The curves in Figure 6a could be used to reduce the
along a nonnormal section, cut somewhere between lines dead-load design moments in box girders having 6 to 10
1 and 2 in Figure 2. cells and spans ranging from 17.1 to 33.5 m(56to110 ft).
These limits encompass most simple span, reinforced
Models and Program Description concrete structures. Similar reduction curves for struc-
tures having 2 to 5 cells appear in Figure 6b. The curves
The current study was initiated by establishing super- would be used as follows:
structure width to length (aspect) ratios of the longest
spans in each of 395 concrete box-girder structures 1. Calculate the midspan dead-load moment by assum-
being designed in California. Based on those data, a ing that the structure is supported on normal abutments;
range of mathematical models was chosen for assess- 2. If the abutment skews are within 15 deg of one
ment of parameter influence and general behavior. Most another and the cross section is prismatic, proceed to
often, models analyzed were of 4 and 8-celled bridges find a reduction factor;
supported on parallel abutments and containing girders 3. Enter Figure 6a or 6b and use the smaller of the
spaced as shown in Figure 3b. Most models were two skew angles;
skewed 30 and 45 deg, and all webs were assumed to 4. Read a factor from the ordinate axis, using the
be vertical. curve for the appropriate girder (normal) span length;
The support (boundary) conditions used are shown in and
Figure 4a. Each girder reaction point was restrained 5. Reduce the dead-load design moment according to
against vertical movement. To stabilize each model, the formula Mo = F x MN, where M0 = design moment and
transverse movement at center girder reaction points MN = normal moment.
and longitudinal movement at the exact center of the
structure were prevented. Figure 4b shows a layout Aspect ratio was found to be the most significant
52

Figure 1. Total moment at midspan due to skew. Figure 4 . Support conditions for finite element mesh used in models.

cSx
0 L
Figure 5 . Prestress loading.

~;~~:';/ jL-:;:::::::::_
_ __- =-=====:!::L=..~= -~:===-:=l.
MOMENT
• ~
R2 CABLE PATH

~
~
(b) p

R1

Note: 1 cm .. 0.39 in UNIFORM LOAD

kl
NOR MAL COMPONENT UNBALANCED
I W SKEW COMPONENT
TAN e-
e-
(d) Figure 6. Moment F (a) 6 TOJO-CELLED SPANS
10 ..-~.... =:::-.-----.----,,...-----.
reduction curves.
08

x
0 4
i-
(D
-33
-,-
5-m-S4P_A_N_S---+----t--...,,,~,~,'"-I
(f ){ l:·'-R2->R-4 = R-)1_ _ , /
~ 02 @17.1m SPANS - - - > - - - - - + - - - - - 1
t; I
;:: 0 '-----+-----'-----1~----'

( R2-RI )
2 TO 5-CELLED SPANS
T= --- W
2

Figure 2. Location of maximum longitudinal stress. 0 4 - <D

@ 24,4 m SPANS (AND LESS)
02~-----1-----1-----lf-----l

00" Iii" lO' ru

SKEW ANGLE
Notw: 1 m • 3.3 ft.

Figure 7. Dead-load deflection .

(o ) NORMAL BRIDGE GIRDER

~-~
I
j( bl SKEWED BRIDGE GIRDERS
Figure 3. Shear distribution locations and loadings of a typical
I 1~0 ---- ~NTE~- -----
box girder .
~ -.._. 139mm I

(a) I/
I I
- - -· 1
/
1/
I/
/
t_ ____ _'.'.'.'.'..."2_ E~ERIOR .JI
140mm ~
73.B cm
ENO DIAPHRAGMS
15° DEG SKEW

j CENTER
l ,....~---~~
/' 76mm ////

II// ////

RIGHT EXTERIOR //
///

139cm -----------------~:::;;>"

45
0 - DEFLECTION FROM CELL ANALYS1S
11- DEFLECTION FROM BEAM ANALYSIS
Note: 1 mm = 0.039 in.
 53

parameter influencing overall bending behavior. Al- torsional rigidity (shear flow) by the deck and exterior
though aspect ratio is not referred to in the plots them- girders than by the interior girders (Figure 3b).
selves, it is reflected in the curves in Figure 6. Notice Statics checks indicate a 7 to 9 percent violation of
that the factors decrease with decreasing span length vertical equilibrium between external reactions and in-
and with an increase in the number of cells. As a gen- ternal girder shears. Although this error is tolerable,
eral rule, transverse distribution of girder moments for it could be reduced by mesh refinement. Shear behavior
structures skewed less than 45 deg and having aspect at acute corners is opposite to that at obtuse corners.
ratios of 0.5 and less can be considered the same as that The acute corner exterior girder carries but a small
of comparable structures on normal supports. Six to amount of shear. In some long span and highly skewed
ten-celled structures that have 1:1 aspect ratios and are structures, the shear at the acute corner girder actu-
skewed 45 deg exhibit ratios about 10 percent larger ally reverses in sign, indicating an uplift tendency.
between exterior and center girder moments than do Six separate live loadings were applied to each of four
their normally supported counterparts. To verify the structural models. Figure 9 shows a plan for live loads
analyses, statics checks were made by comparing exter- 1 and 5. Figure Sb shows that live load 5 is more criti-
nal and internal moments at midspan. In most cases, cal for shear in the obtuse exterior and first interior
moments checked to within 1 percent and vertical forces girders than is live load 1. Live load 1 is resisted by a
summed exactly to zero. more uniform shear distribution among the five girders
nearest the obtuse corner. Much of live load 1 travels
Moment Derivation directly to the end diaphragm.
The data indicate that shear distribution depends pri-
Figure 2 shows the migration of the maximum longitudi- marily on span length and that an approximate linear in-
nal stress location in various exterior girders. The crease occurs in the exterior girder shear as skew angle
maximum stress in each individual girder of the struc- increases because of relative support rotation. If other
tural section is indicated near the lines that are desig- parameters are held constant, the longer the span is, the
nated by arrows. The largest moment value obtained by greater is the redistribution of reaction toward the obtuse
summing individual girder moments along any such line corner of a skewed support. [These box-girder study re-
was used for establishing reduction curves. The total sults are in agreement with a previously stated theory
moment for each skewed structure was divided by the (9). ) Therefore , ranges of design shear factors based on
maximum moment value calculated for a normal maxi- span length were developed from the data. For example,
mum structure of the same span and section. The re- the dead-load shear factors for obtuse corners of a re-
sult was the reduction factor F. inforced concrete box girder are shown in Figure 10.
The factors were calculated by taking the ratios of the
Deflections shear results for exterior and first interior girders of
skewed and of comparable straight or normal structures
Dead-load deflection data for a CELL model having a (F = V, / VN). Interpolation within these ranges should
54.9-m (180-ft) span, a 21.9-m (72-ft) width, and 9 gird- lead to acceptable designs, that is, ±5 percent.
ers are presented in Figure 7. This example typifies the The range of shear factors for exterior girder design
study's results. Figure 7a shows the theoretical, elastic, in multiple span structures is likely to be lower than the
deflected shape of a bridge girder with orthogonal bear- comparable range for simple spans. How much lower de-
ings found from simple beam analysis. Results for center pends on various factors such as span unbalance, number
and right exterior girders obtained by using CELL anal- of spans, and number of columns. Preliminary analyses
yses are plotted for 15 and 45-deg skews respectively. of a model having two spans of 30.5 m {100 ft) and 24.4 m
The following information was noted: (80 ft) and skewed 30 deg produced factors for the exte-
rior girder range in the long span about 25 percent lower
1. A significant reduction in deflection due to skew than those for a simple span of the same length. The
is evidenced at 15 deg, short span shears were almost uniformly distributed
2. Deflections are only half as great as normal for across the section.
45-deg skews, The shear factor range for an exterior girder would
3. The discrepancy between exterior and center de- be about 13 percent higher for a combined dead and live
flections is larger at 45 deg than at 15 deg, and loading than for a dead loading alone because live load 5
4. Narrower structures of the same span as that produced a more critical condition in the obtuse exterior
shown in Figure 7 exhibit smaller discrepancies between girder than live load 1. A converse effect occurs at
exterior and center deflections. acute corners and girder shears are also effectively re-
duced. By interpolating within ranges, one may use the
Shear Behavior shear factor curves to obtain a factor for a particular
structure; depth-to-span length ratio (D/ L), span length,
The girder shears were calculated at the CELL mesh and skew ai·e used to obtain a factor to inc1·ease design
node locations nearest the left support. The shear dis- shears from those of identical structural sections sup-
tribution location can be thought of as a section cut par- ported on normal abutments. The following example
allel to the abutment and as close to it as possible (Fig- illustrates the process. For a bridge that has a span
ure 3a). Transverse shear dist1·ibution results under length of 24.4 m (SO ft), a D/ L of 0.052, and abutment
dead load for the simple span box girder depicted in Fig- skews of 30 and 40 deg, find the design shear factors
ure 3 are presented in Figure Sa. The structure is 25.6 for the exterior and first interior girders at the obtuse
m (S4 ft) in span length, 17.1 m (56 ft) wide, and skewed corners. The steps in the solution are as follows:
45 deg at each support. In this case, and for all the dead-
load analyses, it was evident that the exterior girder 1. Enter the exterior girder range in Figure 10 and
framing the obtuse support corner of a structure would inte1·polate between the 0.045 and 0.060 D/ L lines, for
carry a much greater portion of the total structure shear a 33.5-m (110-ft) span to establish a line for the desired
than would any other girdel'. The interior girders carry value of 0.052;
about the same amounts of shear as they would if the 2. Using the span length of 24.4 m {80 ft), interpolate
structure were normally supported. Such behavior might between the new line for a 33.5-m (110-ft) span, D/L of
be expected because a greater contribution is made to 0.052 and the lower limit line [L = 17.1 m (56 ft)) to es-
54

tablish the factors of 1.63 for the 30-deg skew and 1.83 Prestress Behavior
for the 40-deg skew;
3. Establish factors of 1.09 and 1.12 for the first in- Longitudinal stress results indicate that reductions in
terior girder by interpolating within the lower range; and prestressing forces are possible in wide structures hav-
4. Increase the design shears in the exterior and ing skews over 45 or 50 deg. For example, a bridge
first interior girders by the appropriate factor F, ac- skewed 50 deg, and having 13 girders and an aspect ratio
cording to the formula Vo = F x VN, where Vo = design of 0.675, was analyzed by using the CELL program. Re-
shear and VN =normal shear. sults indicate a possible reduction in prestressing forces
of 11 percent. This reduction may have occurred because
High reaction concentrations at the obtuse corner dead-load torque peaks at 45 or 50 deg and tends to re-
cause about 3 to 6 percent of the total structure dead- duce dead-load moments most highly at those skews, as
load shear to be carried by the end diaphragm as struc- discussed by Kollbrunner and Basler (9). However, pres-
ture aspect ratio increases from 0. 5 to 1.0. Most of this ent design procedures for designing a prestressing sys-
dead-load shear is concentrated near the obtuse corner. tem to counteract dead- and live -load stresses in individ-
ual girders appear realistic and effective, as far as the
total structure is concerned. In producing an effective,
uniform, upward load, prestressing has overall effects
Figure 8. Transverse shear distribution. that oppose the dead load. Since end moments produced
by skewed supports resist downward movement under
{a) 25.6m/45°/6 CELLED dead loading, they should also resist upward movement
.sw;. 11•1. "·•J. 14"1., U% 16•1. 2'!W,.
due to prestressing. Analysis results for prestress
PERC ENTAGE OF
TOT AL STRUCTURE loadings support this notion. Distribution of shear in the
SHEAR CARRIED
models under prestress loading was almost identical to
dead-load distribution, though opposite in sign. Thus,
GIRDER NUMBER
draping the cable parabolically with maximum eccentric-
DEAD LOAD ity at each girder midspan, which is current practice, is
(bl
Cf'lo ,.,,. l it ~ 11% 17~.i 19% 27% -t% 2% 6% 10% 15% 24% 44%
the proper general approach. However, a larger torque
is produced in the superstructure by dead loadings than
that produced by the lower uniform load due to prestress-

~ ~
Note: 1 m = 3.3 ft.
ing. Therefore, support design should provide for a
~
7 ~
Q
3 2 ~ 7
larger relative settlement at the obtuse corners resulting
from rotations due to combined loading.
LIVE LOAD I LIVE LOAD 5

Live-Load Bending Behavior

Figure 9. Plan for live loads 1 and 5. Results indicate that live-load reductions of lesser mag-
nitude than those for dead load can be realized in simple
::.:CJ r::::r::::r spans. As a structure widens in relation to its length,
LT:] ·::r:-.:1
r::::r:::1 :: :::r::.~1 the indicated live-load reduction lessens relative to the
r:::r::::r r::::r.:::r dead-load reduction. The smaller reduction is due to a
more localized spanning effect occurring in the individual
LIVE LOAD 5 LIVE LOAD I girder than would exist in a 'corresponding girder of a nar-
row bridge. However, the load case studies were not ex-
tensive enough to provide general behavioral guidelines.
Figure 10. Dead-load shear factors for obtuse corners
of a simple span box girder. SUMMARY
2.6
This study of skew parameters of simple spans provided
~ft; •)
,-9.l the following information.
2.4 -~f'
1. General knowledge regarding the effect of skew on
transverse shear distribution to individual girders near
Q::
2.2 abutments of concrete box-girder bridges;
~ 2. A more accurate method for designing shear rein-
~
forcement than previously used heuristic methods;
~
Q:: 2.0 3. Some insight into the overall structural behavior
~ gained from deflection results;
>;:
,., 4. Evidence that aspect ratio and skew are the signif-
"'~ 1.8 icant parameters influencing bending behavior and that
<:> span and skew are significant in affecting shear behavior
Q::
<;; (depth-to-span length ratio changes within the usual
I
1.6 <ft:I
~ ranges of 0.040 to 0.062 do not appreciably influence
Cl)
overall structural behavior);
5. A proposed method that allows reduction in dead-
I. 4
load design moments in simple span, reinforced concrete
bridges and leads to savings in concrete or steel quantities;
6. Illustration that substantial savings are possible
1.2
in multiple span, reinforced concrete bridges and would
far offset the added design expense involved in finite ele-
ment analyses with the CELL program; and
40 " $()" 60· 7. Evidence that shows that prestress forces can be
No re1 f m • 3.3 fl ANGLE reduced in certain highly skewed structures.
 55

The study did not provide comprehensive live-load anal- vated Support Effects on Curved Bridges. Journal
yses, general multiple span design criteria, or refined of Structural Division, ASCE, July 1974, p. 1379.
criteria for prestressing forces required in skewed 7. Suite of Bridge Design and Analysis Programs: Vol-
bridges. ume 2- MOT/CIRIA Finite Element Package for the
Analysis of Reinforced Concrete Slab Bridge Decks.
RECOMMENDATIONS New Zealand Ministry of Transport, May 1969.
8. W. G. Godden and M. Aslam. Model Studies of
In the design of skewed, simple span, concrete box- Skewed Box-Girder Bridges. Univ. of California,
girder bridges, the shears in the exterior and first in- Berkeley, UC-SESM Rept. 71-26, Dec. 1971.
terior girders should be modified according to factors 9. C. F. Kollbrunner and K .. Basler. Torsion in Struc-
obtained from design charts such as those shown in Fig- tures. Springer-Verlag, Berlin and Heidelberg,
ure 10. In the design of abutment bearing pads and end 1969, pp. 62-95.
diaphragms for shear, the effects of the high obtuse
corner reactions should be considered. Simple span,
reinforced concrete bridges should be designed for longi-
tudinal bending according to the following procedure:

1. Analyze each structure for dead loading by assum-

ing that it behaves like a beam on normal supports;
2. If the abutment skews arebothgreaterthan 15 deg
and within 15 deg of each other and the section contains
2 to 10 cells, proceed to reduce dead-load design mo-
ments according to Figure 6;
3. Apply live load to the structural section by ana-
lyzing the superstructure as if it were on normal supports
(this conservative approach is required because informa-
tion is limited regarding live-load behavior);
4. Use established methods for designing longitudinal
reinforcement but base the design on the reduced dead-
load moment obtained in step 1 {the bars should be ex-
tended farther toward the obtuse corners than is current
practice); and
5. Accept the CELL program for general use in de-
sign practice.

ACKNOWLEDGMENTS

This research was performed and financed by the Cali-

fornia Department of Transportation. Assistance was
provided by A. C. Scordelis of the University of Cali-
fornia, Berkeley, and is gratefully acknowledged. The
contents reflect the views of the author, who is respon-
sible for the facts and the accuracy of the data presented.
The contents do not necessarily reflect the official views
or policies of the California Department of Transporta-
tion. This report does not constitute a standard, speci-
fication, or regulation.

REFERENCES

1. K. J. Willam and A. C. Scordelis. Computer Pro-

gram for Cellular Structures of Arbitrary Plan Ge-
ometry. Univ. of California, Berkeley, UC-SESM
Rept. 70-10, Sept. 1970.
2. C. D. Comartin and A. C. Scordelis. Analysis and
Design of Skew Box-Girder Bridges. Univ. of
California, Berkeley, UC-SESM Rept. 72-14, Dec.
1972.
3. H. D. Nix and E. E. Evans. Analysis of a Skewed
Concrete Box-Girder Railroad Bridge {Floral Park
Underpass). Bridge Department, California Division
of Highways, Research and Development Rept.
4152-73-8.
4. M. R. Wallace. Floral Park Underpass. Office of
Structures Branch, California Department of Trans-
portation, Research and Development Supplemental
Rept. 0002-1-73-10.
5. R. G. Sisodiya, A. Ghali, and Y. K. Cheung. Di-
aphragms in Single and Double-Cell Box-Girder
Bridges With Varying Angle of Skew. ACI Journal,
Proc., Vol. 69, 1972, p. 415.
6. D. W. Funkhouser and C. P. Heins. Skew and Ele·