You are on page 1of 78

EFFECT OF ABUTMENT SKEW AND HORIZONTALLY CURVED ALIGNMENT ON

BRIDGE REACTION FORCES

A Thesis

Presented to the faculty of the Department of Civil Engineering

California State University, Sacramento

Submitted in partial satisfaction of


the requirements for the degree of

MASTER OF SCIENCE

In

Civil Engineering

(Structural Engineering)

by

Lucas Richard Miner

SPRING
2014
© 2014

Lucas Richard Miner

ALL RIGHTS RESERVED

ii
EFFECT OF ABUTMENT SKEW AND HORIZONTALLY CURVED ALIGNMENT ON

BRIDGE REACTION FORCES

A Thesis

by

Lucas Richard Miner

Approved by:

__________________________________, Committee Chair


Dr. Benjamin Fell

__________________________________, Second Reader


Dr. Matthew Salveson

__________________________________, Third Reader


Dr. Toorak Zokaie

____________________________
Date

iii
Student: Lucas Richard Miner

I certify that this student has met the requirements for format contained in the University format

manual, and that this thesis is suitable for shelving in the Library and credit is to be awarded for

the thesis.

__________________________, Graduate Coordinator ___________________


Dr. Matthew Salveson Date

Department of Civil Engineering

iv
Abstract

Of

EFFECT OF ABUTMENT SKEW AND HORIZONTALLY CURVED ALIGNMENT ON

BRIDGE REACTION FORCES

by

Lucas Richard Miner

This thesis presents analysis results from a parametric study investigating the effects of abutment

skew and horizontal alignment curvature on bridge reaction forces. Over 800 single-span box

girder bridges were modeled using three-dimensional finite element analysis software. The

effects of skew angle, curve angle, and a coupled skew-curve effect on abutment reactions in the

obtuse corners of single-span box girder bridges were found to be significant. These results are

compared with state-of-the-art bridge design practice and LRFD Specification procedures that

provide guidance on designing bridges with skew and horizontal curvature. Interestingly, the

results demonstrate that the obtuse corner reaction forces are greatly influence by the bridge

aspect ratio across a wide variety of skew and curve angles. Furthermore, the horizontal

alignment curvature has a large effect, even at very small alignment central angles. Moreover,

the effect of skew angle is shown to be partially dependent on bridge curvature; although this

coupling of the skew and curve effects is minimal at small skew angles. The bearing stiffness was

also varied, which had a large effect on the reactions of skewed and curved bridges. Empirical

equations for skew and curve correction are proposed and additional research recommended in

the summary chapter.

v
_______________________, Committee Chair
Dr. Benjamin Fell

_______________________
Date

vi
ACKNOWLEDGEMENTS

I am grateful for the formal instruction of Dr. Benjamin Fell, Dr. Matthew Salveson, and

Dr. Toorak Zokaie in my classes at CSUS who clearly expounded the principles of structural

engineering and bridge design in a challenging and stimulating way.

I am also grateful for my bridge design job at Drake Haglan and Associates and my past

internships with the Caltrans Office of Earthquake Engineering and Dr. Toorak Zokaie. Without

these, I would not have received the direction and guidance that was vital to both the scope and

the details of this work.

I am most thankful for the patience and steadfastness of my wife Hannah, the

encouragement and support from my parents Richard and Christina, the prayers of all my friends

and family, and the kindness of my grandmother-in-law Bernice who thought my pursuit of

graduate education in structural engineering important enough to contribute with her own

finances.

vii
TABLE OF CONTENTS

Page

Acknowledgements ................................................................................................................. vii 

List of Tables ............................................................................................................................. x 

List of Figures ...........................................................................................................................xi 

Chapter

1. INTRODUCTION ................................................................................................................. 1 

1.1 Motivation ....................................................................................................................... 1 

1.2 Objectives and Scope....................................................................................................... 8 

1.3 Organization and Outline................................................................................................. 9 

1.4 Definitions ..................................................................................................................... 10 

2. PREVIOUS RESEARCH AND STATE OF PRACTICE ................................................... 14 

2.1 Design Specifications ................................................................................................... 14 

2.2 Bridge Design Practice .................................................................................................. 16 

3. STRUCTURAL ANALYSIS............................................................................................... 18 

3.1 Format............................................................................................................................ 18 

3.1.1 Bridge Configurations ............................................................................................ 18 

3.1.2 Plots ........................................................................................................................ 19 

3.2 Software ........................................................................................................................ 22 

3.3 Modeling........................................................................................................................ 22 

4. ANALYSIS OF THE DATA ............................................................................................... 27 

4.1 Introduction to Results..................................................................................................27 

4.2 Visualization of Bridge Parametric Variation ............................................................... 28 

viii
4.3 Results ........................................................................................................................... 30 

4.3.1 Aspect Ratio = 1.0 .................................................................................................. 30 

4.3.2 Aspect Ratio = 2.0 .................................................................................................. 33 

4.3.3 Aspect Ratio = 4.0 .................................................................................................. 37 

4.3.4 Aspect Ratio = 8.0 .................................................................................................. 39 

4.3.5 General Observations ............................................................................................. 41 

5. FINDINGS AND INTERPRETATIONS ............................................................................ 42 

5.1 General .......................................................................................................................... 42 

5.2 Discussion...................................................................................................................... 42 

6. CONCLUSION .................................................................................................................... 51 

6.1 Summary of Findings ................................................................................................... 51 

6.2 Design Applications ..................................................................................................... 51 

6.2.1 Aspect Ratio Dependency ...................................................................................... 51 

6.2.2 Bearing Stiffness .................................................................................................... 52 

6.3 Further Research ............................................................................................................ 52 

Appendix A: Modeling Details ............................................................................................... 55 

Appendix B: Effect of Softer Bearings ................................................................................... 60 

Appendix C: Effect Of Span Length ....................................................................................... 62 

References ................................................................................................................................ 64 

ix
LIST OF TABLES
Page

Table 2.1: Adapted from AASHTO LRFD Table 4.6.2.2.3c-1................................................ 15 

Table 2.2: Adapted from Caltrans 2014 Amendments to AASHTO LRFD Table

4.6.2.2.3c-1 ............................................................................................................... 15 

Table 3.1- Table showing parametric variation of models analyzed in this study ................... 18 

Table 3.2: The variations used in this study and extent of variation considered ..................... 26 

Table 4.1: A table of graphics that illustrate part of the spectrum of bridge models

considered in this report. All of the bridges in this table have an aspect ratio of

4.0. ............................................................................................................................ 30 

x
LIST OF FIGURES
Page

Figure 1.1: Example of curved bridge alignment and skewed end abutment (left).

Retrieved April 25, 2014 from www.abkj.com. Copyright 2014 ABKJ, Inc ............. 1 

Figure 1.2: Highly skewed railway bridge. Since the railroad above does not cross the

road below at a perpendicular angle, the skew is necessary. Retrieved May 1,

2014 from www.geograph.org.uk .............................................................................. 2 

Figure 1.3: Curved skewed bridge (a) Plan view, (b) Spine model from CBridge, and (c)

Full 3D beam-plate model from CSIBridge of Scales Road Bridge over Slate

Creek in Yuba County, California.............................................................................. 3 

Figure 1.4: Truck driving over left lane of bridge. The left girders will bear most of the

shear force that resists this truck. ............................................................................... 4 

Figure 1.5: Plan view of three dimensional curved box girder bridge analysis model. The

abutment reaction is monitored at the circled node. ................................................... 4 

Figure 1.6: Illustrative example showing the effect of skew angle on abutment reactions

at the obtuse corner for the bridge model shown in Figure 1.5. ................................. 5 

Figure 1.7: Idealization of a simply supported bridge as a beam with pin and roller

boundary conditions. .................................................................................................. 6 

Figure 1.8: Idealization of a simply supported bridge as a plate with edge pin boundary

conditions. In a 3D bridge model, usually the edge pin would be broken up into

multiple conventional pin supports place at the bearing locations. ............................ 6 

Figure 1.9: Curved skewed bridge model showing inside, outside, obtuse, and acute

corners ........................................................................................................................ 7 

Figure 1.10: A (a) curved box-girder bridge model and a straight box-girder bridge model .. 11 
xi
Figure 1.11 A curved skewed bridge model with inside, outside, obtuse, and acute

corners labeled.......................................................................................................... 12 

Figure 1.12 Parallelogram with skew angle labeled. .............................................................. 13 

Figure 3.1: An example of a typical curve for this study. The aspect ratio and bearing

stiffness are shown on the top. The obtuse abutment reaction percentage (Ro) is

plotted against skew from 0 to 60 degree skew. A design curve which

represents the spine model approximation with LRFD modification factors

applied is shown with a dotted line and the rest of the curves correspond to

different central angles. The curve that represents a straight bridge (i.e. central

angle 0) contains markers. ........................................................................................ 20 

Figure 3.2: 3D view of typical box girder model with supports at the end of each girder..... 23 

Figure 3.3: Cross section view of typical box girder with no overhangs. Bridge models

with overhangs would simply have an extra plate element on either side of the

top of the cross section. ............................................................................................ 24 

Figure 3.4: Plan view of typical box girder model with 20 degree skew and no curve. .......... 25 

Figure 4.1: A bridge model with obtuse corner circled. This shows the reaction point

under consideration in this section of the report. If this bridge were resting on

bents or abutments, the bearings would push up on the bridge with a force equal

and opposite to the bridge weight. The obtuse corner reaction is the force that

the bearing sitting under the obtuse corner of the bridge must exert to hold the

bridge up................................................................................................................... 27 

Figure 4.2: Obtuse Corner Reaction Percentage (Ro) vs. Abutment Skew for central

angles between -36 and 36 degrees. The curve with the circle markers

xii
represents the non-curved bridge (central angle = 0) and the dashed line

represents the spine model results with Caltrans modification factors applied

(see Section 3.1.2 to clarify plots). ........................................................................... 31 

Figure 4.3: Zoomed in screenshot showing the difference between the curve response of

bridges with positive and negative central angle...................................................... 32 

Figure 4.4: Obtuse Corner Reaction Percentage (Ro) vs. Abutment Skew for central

angles between -36 and 36 degrees. The curve with the circle markers

represents the non-curved bridge (central angle = 0) and the dashed line

represents the spine model results with Caltrans modification factors applied

(see Section 3.1.2 to clarify plots). ........................................................................... 34 

Figure 4.5: Extreme case with central angle 36 degrees and skew 50 degrees to illustrate

the close proximity of the obtuse corner to the bridge center of gravity. ................. 35 

Figure 4.6: Obtuse Corner Reaction Percentage (Ro) vs. Abutment Skew for central

angles between -36 and 36 degrees. The curve with the circle markers

represents the non-curved bridge (central angle = 0) and the dashed line

represents the spine model results with Caltrans modification factors applied

(see Section 3.1.2 to clarify plots). ........................................................................... 37 

Figure 4.7: Obtuse Corner Reaction Percentage (Ro) vs. Abutment Skew for central

angles between -36 and 36 degrees. The curve with the circle markers

represents the non-curved bridge (central angle = 0) and the dashed line

represents the spine model results with LRFD modification factors applied (see

Section 3.1.2 to clarify plots). .................................................................................. 39 

xiii
Figure 5.1: A plot showing the skew effect for multiple aspect ratios (AR in the legend).

The data used to create these curves is the same data shown in Chapter 4. The

design curve is plotted in a dashed line .................................................................... 43 

Figure 5.2: A plot of skew angle vs. percent of abutment reaction at obtuse corner with

dotted lines showing approximation by proposed formula. ..................................... 44 

Figure 5.3: Plot showing the curve effect for various aspect ratios. When the curves

show negative y-axis values, uplift is occurring at the obtuse corner. ..................... 45 

Figure 5.4: Figure showing the obtuse inside and obtuse outside corners of a bridge

model. For any skew angle, a positive central angle will produce exactly the

same curve effect on the outside corner as a negative central angle will produce

on the inside corner. ................................................................................................. 46 

Figure 5.5: Plot comparing recommended formula with analysis results ............................... 47 

Figure 5.6: Comparison of plots of percent abutment reaction at obtuse corner vs.

abutment skew for all four aspect ratios. .................................................................. 49 

Figure 5.7: Extreme case with central angle 36 degrees and skew 50 degrees to illustrate

the close proximity of the inside obtuse corner to .......... the bridge center of gravity. 50 

xiv
1

CHAPTER 1 : INTRODUCTION
1.1 Motivation 

With the increasing rate of urbanization and rapid infrastructure growth, the need for

complex transportation systems has also increased; often leading to road and bridge geometries

with unconventional, non-collinear, configurations. According to the bridge database maintained

by the Caltrans maintenance division, approximately 10,600 of 23,800 bridges in California are

skewed such that the end abutments or bents are not perpendicular to the longitudinal direction of

the bridge deck as shown in Figures 1.1 and 1.2. Many bridges also contain horizontally curved

alignments in addition to the skew. At this time, structural analysis capabilities due to software

development have removed many of the limitations that once existed in curved skewed bridge

analysis and such bridges are becoming much easier to design. This report discusses the

structural analysis issues that arise as a result of skewed and curved bridge geometries, especially

related to the calculation of girder reaction forces.

Figure 1.1: Example of curved bridge alignment and skewed end abutment (left). Retrieved April

25, 2014 from www.abkj.com. Copyright 2014 ABKJ, Inc


2

Figure 1.2: Highly skewed railway bridge. Since the railroad above does not cross the road

below at a perpendicular angle, the skew is necessary. Retrieved May 1, 2014 from

www.geograph.org.uk

Designing the girders and bearings to support a curved skewed bridge can be a complex

problem since the reaction, shear, and moment demands due to gravity loads will vary across the

girders of a skewed bridge deck. Furthermore, it is difficult to predict the response of a curved

and skewed bridge with the analysis techniques commonly used in bridge design. The most

prevalent bridge design software packages such as CTBridge, CONBOX, CONSPAN, and

VBridge rely on a spine model (example shown in Figure 1.3b); utilizing line elements placed

along the center of the bridge. This treatment is accurate for many applications for bridges with

simple alignments but is unable to capture many load distribution effects of curved and skewed

geometries and may generate unconservative results. A more accurate analysis model of a curved

and/or skewed bridge is a 3D beam-plate model as illustrated in Figure 1.3(c).


3

(a)

Node

End
constraint
2D beam element 3D Plate-Beam Model

(b) (c)

Figure 1.3: Curved skewed bridge (a) Plan view, (b) Spine model from CBridge, and (c) Full 3D

beam-plate model from CSIBridge of Scales Road Bridge over Slate Creek in Yuba County,

California

To highlight the differences between results from a spine model and a full 3D plate-beam

analysis, consider the example illustrated in Figure 1.4 featuring a truck driving in the left lane

over a highway bridge (Figure 1.4). In the more accurate beam-plate model, the left girder

would have a larger shear force as compared to the right side. However, in a spine model, there

are no actual girders modeled, so the bridge is assumed to have an equal lateral distribution of the

shear force imposed by the truck.


4

Figure 1.4: Truck driving over left lane of bridge. The left girders will bear most of the shear

force that resists this truck.

For a skewed bridge, a full 3D model shows that the reactions at the abutments vary

widely across the abutment. A spine model ignores these effects. To illustrate, the curved bridge

(Central Angle = 11 degrees) below in Figure 1.5 was analyzed to find the vertical reaction at the

right side of the abutment.

Figure 1.5: Plan view of three dimensional curved box girder bridge analysis model. The

abutment reaction is monitored at the circled node.


5

1200
Abutment Reaction at Obtuse 
1000
800
Corner [kips]

360% error
600 3D Model
400 Spine Model

200
0
0 20 40 60
Skew Angle [degrees]

Figure 1.6: Illustrative example showing the effect of skew angle on abutment reactions at the

obtuse corner for the bridge model shown in Figure 1.5.

As the curves above show, the spine model that meets the LRFD Section 4 requirements

completely ignores the effect of curve and skew and produces very non-conservative results.1

Thinking about a bridge response that varies from girder to girder due to skew is counterintuitive

to someone taking a statics-based approach because they would assume that rigid statically

determinate bridge would behave uniformly no matter how wide or skewed it is (see Figure 1.7).

1
Of course, the LRFD code mandates modifications to the spine model results that attempt to approximate
the true response. These modifications are detailed later in this report. Furthermore, the reason that these
curves do not have a similar reaction for a skew angle of zero is that the 11 degree central angle is ignored
by the spine model that meets the LRFD code requirements. Then, as skew angle increases, the spine
model reactions do not change but the full 3D model reaction increases; capturing the true 3D response.
6

Figure 1.7: Idealization of a simply supported bridge as a beam with pin and roller boundary

conditions.

The problem with this reasoning is that a bridge, in reality, is not statically determinate as long as

it is supported by multiple bearings on each end (transversely along the width). The bridge is

more accurately modeled as a plate with multiple essential boundary conditions at the edge (see

Figure1.8 ). The loads that are borne by each section along the support width can differ greatly.

Figure 1.8: Idealization of a simply supported bridge as a plate with edge pin boundary

conditions. In a 3D bridge model, usually the edge pin would be broken up into multiple

conventional pin supports place at the bearing locations.

As abutment skew increases, the spine beam idealization becomes less accurate because the

reactions of the bridge will vary across the width.


7

A bridge’s skew angle is not the only aspect of its configuration that affects support

reactions. The curved bridge shown in Figure 1.9 will yield a different response than its

equivalent spine model approximation. The abutment bearings supporting the outside exterior

girder will support a higher percentage of the total dead load as compared to the abutment

bearings supporting the inside exterior girder. The torsion produced by the eccentric bridge

center of gravity is not accurately captured in a spine model. Furthermore, the spine model

ignores the shear due to torsion since there is no out-of-plane dimension to allow the torsion to be

transferred into the girders (i.e. resisted by a shear force couple).2

Outside Obtuse Corner

Outside Acute Corner

Inside Acute Corner

Inside Obtuse Corner

Figure 1.9: Curved skewed bridge model showing inside, outside, obtuse, and acute corners

Although a full 3D modeling procedure is necessary for accurately modeling skewed and curved

bridges, full 3D bridge modeling is inefficient and undesirable for most design purposes because

of the lack of available software, and the difficulty of utilizing full 3D analysis results in the

LRFD Specifications. There are currently no bridge design software packages that contain the

option to use a full 3D model.

2
Another major issue with the spine model approximation is that it may not give accurate results for bridge
bents and columns. Therefore, in bridge design practice, an entirely separate 2D bent model is developed
in a separate software program such as VBent or LEAP RCPier.
8

If a bridge designer wishes to investigate a bridge’s full 3D response, he or she must

either model it from scratch in an FE program such as SAP2000 or use a bridge modeler such as

CSIBridge; which does not have the capability to produce all of the results needed for bridge

design. The latter is usually the bridge designers’ most efficient option. The problem with this

option is that the bridge designer must also know how to interpret the 3D results and apply the

correct LRFD Specifications to these results.

Therefore, AASHTO LRFD Specifications have produced modification factors and

approximate design procedures that allow the designer to continue using a spine approximation

for bridges that are similar to the ideal straight case. The Live Load Distribution Factors (LRFD

4.6.2), Skew Shear Correction Factors (LRFD 4.6.2), Horizontal Curve Limitations (LRFD

4.6.1), and many other modifiers are an attempt to give a reasonable approximation of the true 3D

behavior of a bridge without performing a full 3D analysis.

1.2 Objectives and Scope 

In many cases, the bridge designer is faced with a choice between spine model

idealization and the longer process of full 3D analysis. In the vast majority of cases, a spine

idealization is used. When a spine idealization is used, and the full 3D response is not

considered, many effects of skew and curve may slip through the design and analysis process

completely unnoticed. The bridge configurations studied below mimic common bridge design

situations, and purpose to provide information to the designer concerning the consequences of

using the analysis approximations at his or her disposal.

The purpose of this study is to provide structural analysis results for vertical abutment

reaction forces for skewed curved bridges. These results shed light on some of the significant

differences between 3D spine idealization and full 3D bridge modeling. In design, this data is
9

directly applicable to the calculation of bridge bearing forces and girder shear forces. The results

shown in this study were compiled from the finite element analysis of over 800 bridge models.

The report results specifically focus on bridge obtuse corner abutment reactions as they vary with

curve and skew of single span box girder bridges.

1.3 Organization and Outline 

Chapter 1 of this report provides an explanation of the motivation for this study. The

beginning of this chapter provides a general discussion of the problem being studied. It also

contains the scope and organization of the report. Relevant terms are defined at the end of the

chapter.

The second chapter discusses the AASHTO LRFD Specification procedure for correcting

the design shear forces in girders of skewed bridges and explains the California Amendments to

the LRFD Specifications. It also presents the LRFD Specification limits for spine model

idealization for horizontally curved bridges and some of the research considered in the

development of these specifications. This chapter also contains a breakdown of some of the main

approximations in current bridge design practice on which this study will comment.

Chapter 3 is a discussion of the details of the structural modeling and analysis that

underlies the conclusions of this report. An explanation, both of the format for the data-

presentation and of the plots used to display the analysis information, is found at the beginning of

the chapter. At the end of this chapter, the software and structural modeling assumptions are

detailed.

The results from this study are presented in Chapter 4. These results consist of plots of

the parametric variation of skew and curve for bridges with differing aspect ratios. There are four

individual plots that show the entire spectrum of skew and curve variation for each of the four
10

aspect ratios analyzed by this report. Important comments and observations are found underneath

each of the result plots. These comments explain what the results indicate about the skew effect,

curve effect, coupled skew-curve effect, and the design curve.

A discussion of the analysis findings and an interpretation of the results are found in

Chapter 5. The comments from the previous chapter are discussed in four categories. The skew

effect category contains a discussion of how the obtuse abutment reaction varies with skew angle.

A formula is proposed which takes the aspect ratio into account for determining skew correction.

The curve effect category also proposes a formula that approximates the variation of obtuse

reaction with aspect ratio. The coupled skew-curve effect category discusses how the curve

effect varies with abutment skew and how the skew effect varies with horizontal curve. Lastly,

this chapter contains a discussion of the limitations of the LRFD Specification procedure for

determining skew correction for shear and reactions at the obtuse corner of skewed abutments.

Chapter 6 contains a summary of the conclusions in this report. A section of this chapter

is devoted to summarizing the major points of this research that are directly relevant to bridge

design. At the end of this chapter, additional research is recommended.

1.4 Definitions 

Below are the theoretical definitions of key terms as they function in this thesis.

Aspect Ratio (AR) – The ratio between the length of a given bridge and its width.

(1.1)

For example, if a bridge is twice as long as it is wide, it will have an aspect ratio of 2.0.

Central Angle - The angle included between two points along the centerline of a curved

bridge measured from the center of the curve as shown in Figure 4.6.1.2.3-1 of the LRFD

Specifications.
11

Coupled Skew-Curve Effect – The change in the skew effect as the curve is varied or the

change in the curve effect as the abutment skew is varied. Since the variation of bridge

skew is dependent on both the skew angle and on the bridge curvature, a coupled skew-

curve effect exists. For example, the reaction at the obtuse corner of a highly curved

bridge (Figure 1.10a) will not increase with skew angle at the same rate as the reaction at

the obtuse corner of a straight bridge (Figure 1.10b).

(a) (b)

Figure 1.10: A (a) curved box-girder bridge model and a straight box-girder bridge model

Curve Effect – The bridge response variation due to varying the horizontal bridge

curvature. In this study, the bridge central angle was varied from -48 degrees to 48

degrees in order to observe the curve effect on obtuse corner abutment reactions.

Full 3D Analysis – The utilization of the beam-plate model that is detailed in this report.

More generally, the utilization of a modeling technique that captures 3D effects that

cause response variation transversely along the bridge width.

Girder – A structural component whose primary function is to resist loads in flexure and

shear. The webs of box girder bridges are also called girders when combined with the

deck and soffit slabs, which are referred to as the girder flanges.
12

Obtuse Corner – The corner of a skewed bridge where the angle between the

superstructure alignment and the abutment direction is greater than 90 degrees. A bridge

with one skewed abutment will have one obtuse corner, one acute corner, and two

orthogonal corners. A bridge with two skewed abutments will have two obtuse corners

and two acute corners. A curved bridge with two skewed abutments will have two

outside corners and two inside corners.

Outside Obtuse Corner

Outside Acute Corner

Inside Acute Corner

Inside Obtuse Corner

Figure 1.11 A curved skewed bridge model with inside, outside, obtuse, and acute corners

labeled.

Skew Angle– The off-normal angle between a line along the length of an abutment or bent

and the bridge alignment at the point of interest.


13

Skew Angle


Figure 1.12 Parallelogram with skew angle labeled.

Skew Effect – The bridge response variation due to changing the abutment skew angle. In

this study, the skew angle was varied between 0 and 60 degrees in order to observe the

skew effect on the obtuse corner abutment reactions.


14

CHAPTER 2 : PREVIOUS RESEARCH AND STATE OF PRACTICE

2.1  Design Specifications 

Currently, the AASHTO LRFD Bridge Design Specifications have adopted some simple

guidelines for the use of spine model idealization in bridge analysis. This provides bridge

designers an alternative to modeling and analyzing every bridge as a full 3D finite element model.

In summary, the LRFD Specification states that bridges with central angles below 12 degrees

may be idealized as straight spine model bridges, and that bridges with central angles less than 34

degrees may be modeled with a curved spine model. For skewed bridges up to 60 degree skew

(45 degrees for curved bridges in California), the skew shear correction factors found in LRFD

Section 4.6.2.2.3c may be applied to shear responses obtained from the bridge analysis model to

approximate the actual skew response. Caltrans has also amended the requirements for box-

girder bridges as explained below. The LRFD Specifications lack guidance on many issues

pertaining to skewed and curved bridges including: abutment reaction calculations, bent reaction

calculations, bearing design, and other related issues.

The section relevant to horizontal bridge curvature states “Horizontally curved concrete

box girders may be designed with straight segments, for central angles up to 12 degrees within

one span unless concerns about other force effects dictate otherwise (4.6.1.2.3)” In other words, a

straight spine model may be used to approximate demand for bridges with central angles below

12 degrees. Furthermore, it says “Horizontally curved nonsegmental concrete box girder bridge

superstructures may be analyzed and designed for global force effects as single-spine beams with

straight segments for central angles up to 34 degrees within one span … unless concerns about

local force effects dictate otherwise.” This conclusion is based on research in NCHRP 620

Development of Design Specifications and Commentary for Horizontally Curved Concrete Box-
15

Girder Bridges by NRV, David Evans and Assoc, and Zocon Consulting Engineers (National

Research Council), and on Live Load Distribution Factors for Concrete Box-Girder Bridges by

Song, Shin-Tai, Y. H. Chai and Hida (Song).

The section of the LRFD Specification that pertains to skew and is relevant to this report

is Section 4.6.2.2.3c, which gives skew correction factors for shear, and the California

Amendments to this section. Below are modified tables for shear correction on skewed bridges

between 0 and 60 degree skew.

Type of Superstructure Correction Factor

Cast-In-Place Concrete Multicell Box 12


1.0 0.25 tan 
70

Table 2.1: Adapted from AASHTO LRFD Table 4.6.2.2.3c-1

The variable L is the span length, d is the span depth, and  is the skew angle in degrees. Since

this specification is based on a study from 1975 (Wallace), Caltrans has amended the AASHTO

specification. The Caltrans amendment to AASHTO uses the factors in Table 2.2 below.

Type of Superstructure Correction Factor

Cast-In Place Concrete Multicell Box


1.0
50

Table 2.2: Adapted from Caltrans 2014 Amendments to AASHTO LRFD Table 4.6.2.2.3c-1

It is also noteworthy that there are no skew correction factors for slab bridges. A straight

skewed slab bridge is modeled without skew for design purposes. The reason for this is that slab

bridges are assumed to have ample shear capacity to resist extra shear from the skew and there is
16

no shear check for a slab bridge. Since there is no direct guidance on skewed bent or abutment

reactions, sometimes the skew correction factors are also applied to reactions to calculate bearing

forces. This assumption is neither confirmed or denied by the LRFD Specification, since it

simply recommends that skew be considered in the design of bearings. The LRFD Specification

gives no definite guidance on how skew should be considered in bearing design.

Much study has been conducted pertaining to Skewed Multicell Box Girder Bridges and

this has been utilized in the LRFD Specifications. Much study has also been conducted

pertaining to Curved Multicell Box Girder Bridges. However, to the knowledge of the author, no

study has specifically focused on the relationship between skew and curve for box girder bridges.

This study overlaps the previous curve and skew studies in some respects, but approaches the

variation of curve and skew based on bridge configurations that are purposefully relevant to real

design situations.

2.2 Bridge Design Practice 

In a typical bridge design situation, a curved skewed bridge will most likely be analyzed

only as a spine model. Current bridge design practice relies on the LRFD modification factors to

approximate the skew or curved response. Some of the major approximations that this introduces

are:

1. A curve less than a central angle of 12 degrees is ignored

2. No skew correction will be applied for skewed slab bridges

3. Skew correction for all bridge types is completely independent of bridge aspect

ratio.
17

4. The skew correction factors are usually assumed to also apply to the reactions at

the bearings since the reactions are usually equal and opposite to the shear at the

end of a girder.

The results presented in this report specifically focus on some of the effects that are missed by

spine idealization.
18

CHAPTER 3 : STRUCTURAL ANALYSIS

3.1 Format 

3.1.1 Bridge Configurations

This research looks at the combination between curve and skew in four different major

bridge configurations, i.e. with four different aspect ratios, 1.0, 2.0, 4.0, and 8.0, because the

skew and curve effects were found to vary widely with aspect ratio. First, bridge models were

developed to represent each of these aspect ratios. Then, they were parametrically varied by

changing only the central angle and skew angle.

The bridge configuration variations that were considered are:

Variation Range Discretization

Aspect Ratio 1.0 to 8.0 1.0, 2.0, 4.0, and 8.0

Skew Angle 0o to 60o Increments of 10o

Central Angle -48o to 48o Increments of 6o

Table 3.1- Table showing parametric variation of models analyzed in this study

Single-span 4-cell box girder bridges were analyzed for all three of these variations. The bridge

models with different aspect ratios differed slightly in girder spacing, overhang width, slab

thickness and soffit thickness. For bridge models within a specific aspect ratio, there are no

differences except for the two varied parameters (skew and curve angles) listed above. The most

noteworthy variation between the models of different aspect ratios is the difference in the

overhang length. This is because a larger overhang will distribute a higher percentage of the

reactions to the exterior supports. For this study, it should be expected that approximately 20% of

the reaction will go to each support in the non-curved, non-skewed case in each plot. This is
19

because the overhang lengths were chosen to produce this result. The details of each of these

bridge configurations are found in Appendix A.

3.1.2 Plots

Figure 3.1 is a typical plot for this study. The plot is the percent of the abutment reaction

concentrated at the obtuse Corner (Ro) vs. abutment skew for a range of curve central angles (see

Figure 3.1 for an example curve). The abutment skew is on the x-axis, the percent of abutment

reaction is on the y-axis, and the curve legend on the right shows the central angles that each

curve represents in degrees.


20

Aspect Ratio 4.0, Kbearing = 500,000 k/in
140 Design
Legend (central angle values in degrees) 48
120 o
48 central angle 42
36
100 30
24
80 18
12
60 6
Ro

40 ‐6
‐12
Design Curve ‐18
20 Non-curved case ‐24
(i.e. 0o central angle ‐30
0 ‐36

‐20 0 20 40 60 ‐42
‐48
-48o central angle
‐40
Abutment Skew [deg]

Figure 3.1: An example of a typical curve for this study. The aspect ratio and bearing stiffness

are shown on the top. The obtuse abutment reaction percentage (Ro) is plotted against skew from

0 to 60 degree skew. A design curve which represents the spine model approximation with

LRFD modification factors applied is shown with a dotted line and the rest of the curves

correspond to different central angles. The curve that represents a straight bridge (i.e. central

angle 0) contains markers.

The y-axis value will start near 20% for the non-curved bridge since all models are 4-cell box

girders and are supported in five locations. According to the tributary area method, a straight

non-skewed bridge will distribute approximately 20% of the force to each of the five girders if the

bridge overhang is equal to half the girder spacing.


21

The y-axis value (Ro) is obtained by dividing the reaction force at the obtuse corner by

the sum of all of the reaction forces. These reaction forces are obtained from the bridge model

analysis results. The equation below shows the calculation.

           
100 ∑
(3.1)
       

Ro is the percent of the total abutment reaction that is concentrated at the obtuse corner. The

shape of this curve is the same as the shape of the absolute reaction curve and is similar to the

shape of the skew correction factor curve since they are all related by a constant factor. In order

to compare the y-axis value to the LRFD skew correction factors, one must multiply the y-axis

value by the number of support locations (i.e. the number of bearing pads or the number of

reaction points) and divide by 100.

    #    (3.2)

For this study, the number of supports at Abutment 1 is always five since all models are 4-cell

box-girder bridges.

The plot will also contain a dashed curve for the design response. The design response

represents the results from using the spine model approximation with LRFD modifications

applied. The design curve is calculated by the following equation:

       0             (3.3)

Which, considering the Caltrans equation, is the same as

       0 1.0 (3.4)

This enables the reader to compare the skew effect with the LRFD design equation prediction.
22

3.2  Software 

The modeling for this study was conducted using private software developed by the

author of this report and Zocon Consulting Engineers (Zocon) and owned by Zocon. This

software uses matrix structural analysis based on the stiffness method. Frame elements and plate

elements were used in the 3D models for this study. The frame element is an Euler beam that

accounts for shear deformation. A diagonal boundary spring element was used to model the

boundary conditions. The plate elements defined at the input are broken into constant stress

triangles with membrane and bending degrees of freedom. The only load case that was used was

a self-weight load case. The model results were verified in SAP2000.

3.3 Modeling 

The typical skewed single span box girder for this study is shown below.
23

Figure 3.2: 3D view of typical box girder model with supports at the end of each girder.
24

Figure 3.3: Cross section view of typical box girder with no overhangs. Bridge models with

overhangs would simply have an extra plate element on either side of the top of the cross section.
25

Figure 3.4: Plan view of typical box girder model with 20 degree skew and no curve.

This box girder is pinned (bearing element with 500,000 kip/in vertical stiffness) at the end of

each girder as if it were sitting on a bearing upon an abutment seat. The results in the body of this

report are for bridges that were modeled with stiff bearings (500,000 kips/in) to emulate PTFE

bearings since stiffer bearings are growing rapidly in modern usage and they produce a more

extreme skew response. Fabric bearings are softer and, therefore, produce different curve and

skew responses. Appendix B contains a more detailed discussion of the ramifications of bearing

stiffness as they pertain to this study.

No sloping of the exterior girders was used in this study.


26

Diaphragms (3 ft thick) were modeled at both abutments. These were modeled as beam

elements connecting the span end nodes together along the width of the abutment. In order to

obtain uniformity in the analysis results, the diaphragm material was modeled to be the same as

the rest of the superstructure except for the unit weight. The unit weight was chosen to be 0

pounds per cubic foot since diaphragm length varies with skew. Therefore, the increased weight

of the diaphragm as skew increases has no effect on the results.

The box girder was modeled with longitudinal beam elements for webs and plate

elements at the top and bottom of the bridge to model the flanges (see Figure 3.3). These were

joined by rigid links as shown in the cross section view in Figure 3.3. Realistic lengths, widths,

and girder spacings were chosen for all box girder models in order to make them comparable to

design situations (see Appendix A for complete bridge model input data).

There are four major bridge configurations used in this study that differ primarily in

aspect ratio. Each of these four bridge configurations are varied parametrically as is show by

Table 3.2.

Parameter Range Increment Size

Abutment Skew 0 – 60 degrees 10 degrees

Span Central Angle -48 – 48 degrees 6 degrees

Table 3.2: The variations used in this study and extent of variation considered

The exact modeling specifications of each of the 4 bridge configurations are listed in the tables in

Appendix A.
27

CHAPTER 4 : ANALYSIS OF THE DATA

4.1  Introduction to Results 

This section provides vertical reaction results for the variation of curve and skew for four

different bridge aspect ratios. The data reported is vertical reaction in the obtuse corner of

skewed curved single-span box-girder bridges as shown in Figure 4.1. After a bridge is

constructed, its ends will sit on a load-bearing device designed to bear the reaction forces of the

bridge. The response on the y-axis of the plots presented below captures the bearing force of the

bearing supporting the obtuse corner of the bridge.

Figure 4.1: A bridge model with obtuse corner circled. This shows the reaction point under

consideration in this section of the report. If this bridge were resting on bents or abutments, the

bearings would push up on the bridge with a force equal and opposite to the bridge weight. The

obtuse corner reaction is the force that the bearing sitting under the obtuse corner of the bridge

must exert to hold the bridge up.

The angle between the bridge alignment and the abutment line is greater than 90 degrees in the

obtuse corner of a bridge. The reaction at the support in the obtuse corner usually increases as the

skew angle increases because this support gets nearer to the bridge center of gravity while the

support at the acute corner is getting further away from the bridge corner. This section presents
28

the responses of bridge models with four different aspect ratios as they are varied in skew angle

and central angle. These results are compared to spine model results that are modified by LRFD

skew correction factors; thus emulating current bridge design practice. The line labeled “Design”

represents this data. For an explanation of how the values on the plots are calculated, see section

3.1.2 of this report; which is titled “Plots”. For interpreting the results correctly, it is helpful to

note that all of the curves below are for 4-cell box girder bridges. Therefore, all of the supports

of a straight bridge will bear about 20% of the reaction force. This will depend on the length of

the bridge overhang.

4.2 Visualization of Bridge Parametric Variation 

In order to provide the reader with some aid in visualizing the meaning of the plots

presented, the following graphics are provided for the third considered bridge aspect ratio. The

third bridge aspect ratio is the model with aspect ratio 4.0. The graphics below illustrate the

parametric variation of curve and skew.


29

Bridge Model – Plan View Description

Aspect Ratio = 4.0

Central Angle = 0o

Skew Angle = 0o

Aspect Ratio = 4.0

Central Angle = -48o

Skew Angle = 0o

Aspect Ratio = 4.0

Central Angle = 0o

Skew Angle = 30o


30

Aspect Ratio = 4.0

Central Angle = 48o

Skew Angle = 30o

Table 4.1: A table of graphics that illustrate part of the spectrum of bridge models considered in

this report. All of the bridges in this table have an aspect ratio of 4.0.

4.3 Results 

The results presented in this section provide graphs of obtuse corner reactions for aspect

ratios 1.0, 2.0, 4.0, and 8.0 from the full 3D analysis and from the 3D spine model modified with

LRFD correction factors.

4.3.1 Aspect Ratio = 1.0

A model of a single-span box girder with a length of 60 feet, a width of 60 feet, a girder

spacing of 12.5 feet and an overhang of 5 feet was parametrically varied between skew angles 0

to 40 degrees and central angles -36 to 36 degrees. The curve below shows the reaction vs. skew

relationship for multiple central angles. The reaction value plotted is the percent of the abutment

reaction that is borne by the obtuse corner. This value is obtained by dividing the obtuse corner

reaction by the sum of the reactions at the abutment.


31

Aspect Ratio 1.0, Kbearing = 500,000 k/in
50 Design

45 36
30
40 24
35 18

30 12
6
Ro

25 0
20 ‐6

15 ‐12
‐18
10 ‐24
5 ‐30
‐36
0
0 10 20 30 40
Abutment Skew [deg]
Figure 4.2: Obtuse Corner Reaction Percentage (Ro) vs. Abutment Skew for central angles

between -36 and 36 degrees. The curve with the circle markers represents the non-curved bridge

(central angle = 0) and the dashed line represents the spine model results with Caltrans

modification factors applied (see Section 3.1.2 to clarify plots).

Skew Effect 

The skew effect is well approximated by the skew correction. This is evidenced by the

fact that the design curve (dashed line) closely follows the curve that represents the non-curved

bridge (circle markers).


32

Curve Effect 

As one would expect, as central angle increases, obtuse corner reaction increases. This

results from torsion due to the change in the center of gravity along the width of the bridge. The

curve effect is not negligible for small curvatures (below 12 degrees). The curve effect is more

pronounced for negative central angles than positive central angles. This can be observed by

noticing that the curves below the 0 degree central angle curve are more spread out than the

curves above the 0 degree central angle curve as is shown below.

The Curve Effect is more


pronounced for negative
central angles than for
positive central angles.

The top curve (36o) starts


at 28%, the middle curve
(0o) starts at 18%, and the
bottom curve (-36o) starts
at 6%.

The distance between the


lower curves (18 – 6 =
12) is greater than the
distance between the
upper curves (28 – 18 =
10).

Figure 4.3: Zoomed in screenshot showing the difference between the curve response of bridges

with positive and negative central angle.

Coupled Skew­Curve Effect 

The coupled curve-skew effect appears to be negligible since the slope of the curves does

not change much as the central angle changes.


33

Design Curve 

The design curve approximates the skew response very well since it follows the skew

response of the non-curved bridge very closely. 

4.3.2 Aspect Ratio = 2.0

A model of a single-span box girder with a length of 80 feet, a width of 40 feet, a girder

spacing of 8 feet, and an overhang of 4 feet was parametrically varied between skew angles 0 to

50 degrees and central angles -48 to 48 degrees. The curve below shows the reaction vs. skew

relationship for multiple central angles. The reaction value plotted is the percent of the abutment

reaction that is borne by the obtuse corner. This value is obtained by dividing the obtuse corner

reaction by the sum of the reactions at the abutment.


34

Aspect Ratio 2.0, Kbearing = 500,000 k/in
80 Design
48
70 42
36
60 30
24
50 18
12
40 6
Ro

0
30 ‐6
‐12

20 ‐18
‐24
‐30
10 ‐36
‐42
0 ‐48

‐10 0 20 40 60
Abutment Skew [deg]
Figure 4.4: Obtuse Corner Reaction Percentage (Ro) vs. Abutment Skew for central angles

between -36 and 36 degrees. The curve with the circle markers represents the non-curved bridge

(central angle = 0) and the dashed line represents the spine model results with Caltrans

modification factors applied (see Section 3.1.2 to clarify plots).

Skew Effect 

The skew effect is more severe than the bridges with 1.0 aspect ratio. This is shown by

the difference in slope between the plots of aspect ratio 1.0 bridges and aspect ratio 2.0. The

obtuse corner reaction of the non-curved model reaches 40% at about 25o skew angle whereas the

non-curved bridge with aspect ratio 1.0 never reached 40% obtuse corner reaction percent. The

skew effect also develops a peak around 40o skew angle


35

Curve Effect 

The curve effect is greatly influenced by the increase in aspect ratio. This is shown by

the fact that uplift occurs in the obtuse corner at -42o central angle. The curve effect is non-

negligible for small central angles. As with the bridges with aspect ratio 1.0, the curve effect is

greater for negative central angles than for positive central angles.

Coupled Skew­Curve Effect 

For bridges with high negative central angle and high skew, the coupled skew-curve

effect is pronounced. This is seen by noticing that the curve for -48 degree central angle crosses

the curves next to it. This is most likely because the obtuse corner at the first abutment becomes

much closer the center of gravity than all other corners as shown in the figure below.

Figure 4.5: Extreme case with central angle 36 degrees and skew 50 degrees to illustrate the close

proximity of the obtuse corner to the bridge center of gravity.

For bridges with positive central angle, the coupled skew-curve effect is minimal since

the slope of the curves does not vary much with changing central angle.

Design Curve 

The skew effect is not well predicted for aspect ratio 2.0. As an important note, the

design curve did not change at all as the aspect ratio changed since skew correction factors are
36

not dependant on aspect ratio. At an aspect ratio of 2.0, the skew response of the non-curved

bridge is higher than the design curve shows. The design curve prediction is non-conservative in

this case.  
37

4.3.3 Aspect Ratio = 4.0

A model of a single-span box girder with a length of 160 feet, a width of 40 feet, a girder

spacing of 8 feet, and an overhang of 4 feet was parametrically varied between skew angles 0 to

60 degrees and central angles -48 to 48 degrees. The curve below shows the reaction vs. skew

relationship for multiple central angles. The reaction value plotted is the percent of the abutment

reaction that is borne by the obtuse corner. This value is obtained by dividing the obtuse corner

reaction by the sum of the reactions at the abutment.

Aspect Ratio 4.0, Kbearing = 500,000 k/in
140 Design
48
120 42
36
100 30
24
80 18
12
60 6
Ro

40 ‐6
‐12
‐18
20 ‐24
‐30
0 ‐36

‐20 0 20 40 60 ‐42
‐48

‐40
Abutment Skew [deg]

Figure 4.6: Obtuse Corner Reaction Percentage (Ro) vs. Abutment Skew for central angles

between -36 and 36 degrees. The curve with the circle markers represents the non-curved bridge

(central angle = 0) and the dashed line represents the spine model results with Caltrans

modification factors applied (see Section 3.1.2 to clarify plots).


38

Skew Effect 

At an aspect ratio of 4.0, the skew effect is much more severe than at lower aspect ratios.

This can be seen by the fact that percent of abutment reaction in the obtuse corner reaches 40% at

a skew angle of 15o. The skew effect reaches a peak at about 40 degree skew.

Curve Effect 

The curve effect is also more severe than at lower aspect ratios. This is seen in the fact

that uplift occurs at a central angle of 30o in the non-skewed case. The curve effect is not

negligible at small central angles. As with the bridges with aspect ratios 1.0 and 2.0, the curve

effect is greater for negative central angles than for positive central angles.

Coupled Skew­Curve Effect 

For bridges with high negative central angle and high skew, the coupled skew-curve

effect is more pronounced than in lower central angles. This is the same effect observed and

discussed in the previous pages for aspect ratio 2.0. For bridges with a positive central angle, the

coupled skew-curve effect is minimal. This is evidenced by the fact that the slope of the curves

does not vary much with the changing central angle.

Design Curve 

The design curve does not accurately approximate the 3D response for aspect ratio 4.0.

At only 25 degree skew, the non-curved model response is 50% higher than the design

approximation. The design curve is highly non-conservative for most cases in this aspect ratio.

 
39

4.3.4 Aspect Ratio = 8.0

A model of a single-span box girder with a length of 200 feet, a width of 25 feet, a girder

spacing of 5.5 feet, and an overhang of 1.5 feet was parametrically varied between skew angles 0

to 60 degrees and central angles -48 to 48 degrees. Below is a screenshot of an example model.

The curve below shows the reaction vs. skew relationship for multiple central angles. The

reaction value plotted is the percent of the abutment reaction that is borne by the obtuse corner.

This value is obtained by dividing the obtuse corner reaction by the sum of the reactions at the

abutment.

Aspect Ratio 8.0, Kbearing = 500,000 k/in
250 Design
48
42
200 36
30
150 24
18
12
100 6
Ro

0
‐6
50 ‐12
‐18
0 ‐24
‐30
0 20 40 60 ‐36
‐50 ‐42
‐48

‐100
Abutment Skew [deg]

Figure 4.7: Obtuse Corner Reaction Percentage (Ro) vs. Abutment Skew for central angles

between -36 and 36 degrees. The curve with the circle markers represents the non-curved bridge
40

(central angle = 0) and the dashed line represents the spine model results with LRFD modification

factors applied (see Section 3.1.2 to clarify plots).

Skew Effect 

At this aspect ratio, the skew effect observed is more severe than in lower aspect ratios.

The non-curved case reaches 40% obtuse corner reaction at about 8o skew angle. At 40 degree

skew, the obtuse corner is bearing over 100% of the load, which means that the sum of the other

reactions is a negative number. This suggests that a large amount of uplift is occurring in the

acute corner and at the inside obtuse support. The skew effect peaks at about 45o skew.

Curve Effect 

The curve effect is much more severe than it was at lower aspect ratios and is not

negligible for all central angles. This can be seen by observing that uplift occurs in the obtuse

corner at only -24o central angle. As with the bridges with aspect ratios 1.0, 2.0, and 4.0, the

curve effect is greater for negative central angles than for positive central angles.

Coupled Skew­Curve Effect 

The coupled skew-curve effect is pronounced for all central angles. This is can be seen

by noticing the varying slopes of the curves they fan away from each other and draw near to each

other depending on the skew and central angle. For positive central angles with skew between 0

and 40 degrees, the bridge curvature augments the response. This is evidenced by the slight

spreading out of the plot curves for positive central angle from 0 to 40 degree skew. After 40

degree skew, as the plot curves slope downwards, the bridge curvature causes the curves to come

together. For negative central angles, the coupled skew-curve response is similar to the response

at aspect ratio 4.0. The coupled response has decreased in severity from the response at aspect

ratio 4.0 since the curve for central angle -48 degrees intersects the 0 central angle curve at about

60 degree skew, whereas it crossed around 57 degree skew for the bridge with aspect ratio 4.0.
41

Design Curve 

The design curve does not provide a good approximation for the response at this aspect

ratio. Since the skew correction factors are not dependant on aspect ratio, they do not capture the

increase in the skew effect at higher aspect ratios.

4.3.5 General Observations

Skew Effect 

The skew effect increases as aspect ratio increases. For large aspect ratios, the reaction

vs. skew curves peak somewhere between 35 and 45 degrees.

Curve Effect 

The curve effect increases as aspect ratio increases. The curve effect is very significant at

all aspect ratios. The curve effect is not negligible for central angles less than 12o

Coupled Skew­Curve Effect 

The coupled skew-curve effect is noticeable in both low and high aspect ratios but is not

very significant in models with skew angles less than 35 degrees. It is almost unnoticeable in

models with a positive central angle but is very noticeable in models with negative central angles.

The coupled skew-curve effect does not necessarily increase with aspect ratio.

Design Curve 

The design curve, which reflects the spine model results with an applied skew correction

factor, nearly approximates the skew effect for aspect ratios near 1.0. However, for aspect ratios

of 2.0 and higher, the design curve predictions are non-conservative. For high aspect ratios, the

design curve does not reasonably approximate the obtuse corner reaction for any skew angle

above 10 degrees.
42

CHAPTER 5 : FINDINGS AND INTERPRETATIONS

5.1 General 

According to the analysis of the bridge configurations above, aspect ratio plays a

prominent role in the behavior of the skew effect, the curve effect, the skew-curve effect, and the

accuracy of the design curve. For bridges with aspect ratios greater than 1.0, the 3D spine

analysis with LRFD modifications may yield non-conservative results since the full 3D response

is difficult to predict. Stiffer bearings such as PTFE bearings will most likely give a similar

bridge response to the curves above whereas the skew and curve responses will change if less-

stiff bearings are used. The coupled skew-curve effect depends on aspect ratio and is not

necessarily increased as aspect ratio is increased. For this study, the coupled skew-curve effect

was observed most prominently when aspect ratio was 4.0. However, this effect is very small at

low skew angles (below 35 degrees) and low central angles (below 30 degrees).

5.2 Discussion 

The values of the reactions at the obtuse corners of the bridges analyzed above are clearly

affected by skew angle, central angle, and some coupling that occurs when the two are combined

together.

Skew Effect 

The skew effect increases significantly as aspect ratio increases. Based on the results

presented in Chapter 4, the skew effect can be isolated by only considering the non-curved data

for all aspect ratios. The plot below shows this comparison.
43

Skew Effect for Varying Aspect Ratios
140
120 Design
AR = 1.0
100 AR = 2.0
AR = 4.0
80
Ro

AR = 8.0

60
40
20
0
0 20 40 60
Abutment Skew [deg]
Figure 5.1: A plot showing the skew effect for multiple aspect ratios (AR in the legend). The data

used to create these curves is the same data shown in Chapter 4. The design curve is plotted in a

dashed line

As is shown in Figure 5.1, the skew effect is clearly dependent on aspect ratio and the design

curve does not provide a sound prediction at high aspect ratios.

Neither the LRFD Specifications nor the Caltrans Amendments provide equations

specifically for approximating reactions. In general, the skew shear correction factors are thought

to apply to reactions as well as to shear. The LRFD equation that applies to the box-girder bridges

analyzed in this report is the Caltrans correction factor shown below.

  1.0 (5.1)
44

The variable  is the skew angle in degrees. This study has found that the skew effect cannot be

accurately predicted based solely on the skew angle itself. The results of this study suggest that

the skew correction equation grows steadily less accurate as aspect ratio increases.

A simple equation that would better approximate this effect by adding dependence on

aspect ratio is:

.
    1.0 (5.2)

where  is the skew angle and AR is the aspect ratio (Length/Width). Below is a plot showing

how this equation approximates the true response for three of the four aspect ratios considered in

this study.

Proposed Correction Formulas 
Compared to 3D Model Response
120 AR = 1.0

AR = 2.0
100 AR = 4.0

80 Pred: AR = 1.0

Pred: AR = 2.0
Ro

60 Pred: AR = 4.0

40
20
0
0 10 20 30 40
Abutment Skew [deg]
Figure 5.2: A plot of skew angle vs. percent of abutment reaction at obtuse corner with dotted

lines showing approximation by proposed formula.


45

The equation fits the data very closely for aspect ratios between 1.0 and 4.0 from skew angles 0 to

40 degrees. The equation approximation is too conservative for the aspect ratio 8.0 and for skew

angles above 40 degrees. Therefore, this equation is recommended to approximate the skew

effect for aspect ratios between 1.0 and 4.0 from skew angles 0o to 40 o.

Curve Effect 

The curve effect increases as aspect ratio increases. Based on the results presented in

Chapter 4, the curve effect can be isolated by only considering the non-skewed data for all aspect

ratios. Below is a plot that shows the effect of aspect ratio for non-skewed box girder bridges.

Curve Effect for Varying Aspect Ratios
80
AR = 1.0
60
AR = 2.0

AR = 4.0 40
AR = 8.0
20
Ro

0
‐50 ‐30 ‐10 10 30 50
‐20

‐40

‐60
Central Angle [deg]
Figure 5.3: Plot showing the curve effect for various aspect ratios. When the curves show

negative y-axis values, uplift is occurring at the obtuse corner.


46

This plot shows that the curve effect is greatly dependant on aspect ratio.

The LRFD Specification allows curved bridges with central angles below 12 degrees to

be idealized as straight spine models. The results from this study show that, at aspect ratios of 1.0

and above, small central angles can produce an increase in the reaction in the obtuse corner that is

not negligible.

Furthermore, the effect of horizontal alignment curvature is different for negative and

positive central angles. This is easier to think about in terms of “inside corner” and “outside

corner.”

Figure 5.4: Figure showing the obtuse inside and obtuse outside corners of a bridge model. For

any skew angle, a positive central angle will produce exactly the same curve effect on the outside

corner as a negative central angle will produce on the inside corner.

The higher the central angle, the more the reaction forces will shift to the outside corner and the

less reaction force will be resisted at the inside corner. This happens because the center of mass of

the bridge is closer to the outside corner. A bridge with this configuration requires a force couple

at each end to provide torsion resistance. This accounts for the difference in the curve effect

between negative and positive central angles.


47

Therefore, two equations are proposed to approximate this effect.

          1.0 (5.3)

.
          1.0   (5.4)

The variable  is the central angle and AR is the aspect ratio. Below is a plot showing the

correlation between the proposed formula and the data from this study.

Proposed Correction Formulas 
Compared to 3D Model Response
AR = 1.0
60
AR = 2.0 50
AR = 4.0
40
Pred: AR = 1.0

Pred: AR = 2.0
30
Pred: AR = 4.0 20
Ro

10
0
‐35 ‐15 ‐10 5 25
‐20
‐30
Central Angle [deg]
Figure 5.5: Plot comparing recommended formula with analysis results

The variation of the curve effect with aspect ratio is more difficult to capture with a simple

empirical formula as compared to the skew effect. This formula was calibrated for the inside and
48

outside girders of bridges with central angles between 0 and 30 degrees and aspect ratios between

1.0 and 4.0.

Coupled Skew­Curve Effect 

The influence of the coupled skew-curve effect on reactions in the obtuse corner is

negligible at skew angles less than 35 degrees. Furthermore, it is negligible for all skew angles

below 60 degrees at the outside corner of curved bridges with aspect ratios below 4.0. Therefore,

no equation is developed in this report that relates these effects. For aspect ratios between 1.0

and 4.0, skew angles between 0o and 40o, and central angles between 0o and 30o, the additive

formula below is proposed to predict the response of a skewed curved box girder bridge from 3D

spine model results.

.
    1   (5.5)

. .
    1 (5.6)

The variable  is the central angle,  is the skew angle, and AR is the aspect ratio.

The primary situation where the coupled skew-curve effect is significant is at the inside

obtuse corner. The plots from Chapter 4 are shown side by side in the figure below. The lower

curves (negative central angle) at high skews show a significant coupled skew-curve effect.
49

Aspect Ratio 1.0, Kbearing = 500,000 k/in Aspect Ratio 2.0, Kbearing = 500,000 k/in


50 Design 80 Design
48
45 36
70 42
30
40 24 60
36
30
35 18 24
50 18
30 12
12
6
40 6
Ro

25

Ro
0 0

20 ‐6 30 ‐6
‐12
15 ‐12
20 ‐18
‐18 ‐24
10 ‐24 10 ‐30
‐36
5 ‐30
‐42
‐36 0 ‐48
0
0 10 20 30 40 ‐10 0 20 40 60
Abutment Skew [deg] Abutment Skew [deg]

Aspect Ratio 4.0, Kbearing = 500,000 k/in Aspect Ratio 8.0, Kbearing = 500,000 k/in


140 Design 250 Design
48 48
120 42 42
36 200 36
100 30 30
24
80 18
150 24
18
12
12
60 6
100 6
Ro

0
Ro

0
40 ‐6
‐6
‐12 50 ‐12
‐18
20 ‐24
‐18

‐30 0 ‐24
0 ‐36
‐30
0 20 40 60 ‐36
‐20 0 20 40 60 ‐42
‐48 ‐50 ‐42
‐48
‐40 ‐100
Abutment Skew [deg] Abutment Skew [deg]

Figure 5.6: Comparison of plots of percent abutment reaction at obtuse corner vs. abutment skew

for all four aspect ratios.

The reason for the significant coupled skew-curve effect at negative central angles is that

the inside obtuse corner becomes much closer the center of gravity than all other corners as

shown in the figure below.


50

Figure 5.7: Extreme case with central angle 36 degrees and skew 50 degrees to illustrate the close

proximity of the inside obtuse corner to the bridge center of gravity.

The bridge in Figure 5.7 is an extreme case that was developed purely to exemplify the case

when the inside obtuse corner is near the bridge center. As this shows, the skew angle brings the

obtuse corner near to the center of mass of a curved bridge. Therefore, a coupling of the skew

and curve effects is noticeable in the inside obtuse corner.

Design Curve 

The design curve shown in the plots was an application of the LRFD skew correction

factors with Caltrans Amendment modifications. The skew correction factors were developed

from previous research that was specifically for shear response. These skew correction factors

depend solely on the bridge type and the skew angle and are not affected by aspect ratio.

For low aspect ratios, the design curve approximates the reactions fairly well for a non-

curved case. However, for aspect ratios greater than 1.0, these results show that the design curve

does not accurately predict the reactions and gives a non-conservative estimate. This report

proposes that aspect ratio be considered in the design of skewed curved bridges.
51

CHAPTER 6 : CONCLUSION

6.1  Summary of Findings 

The most noteworthy conclusion of this report is that as bridge aspect ratio (length

divided by width) increases, the skew effect and the curve effect are significantly increased as

well. For high aspect ratio bridges, a 3D spine model analysis with LRFD skew correction

factors applied yielded non-conservative results for obtuse corner abutment reaction forces.

Furthermore, for aspect ratios between 2.0 and 8.0, the coupled skew-curve effect is significant in

inside obtuse corners of curved skewed bridges. Moreover, a reduction in bearing stiffness

caused the skew effect to decrease. Lastly, curved box girder bridges with small central angles

(less than 12o) still exhibited significant increases in reactions at the outside corners and

significant decreases in reactions at the inside corners.

6.2  Design Applications 

6.2.1 Aspect Ratio Dependency

The data found in this thesis suggests that skewed or curved bridges being designed with

aspect ratios greater than 1.0 should be investigated for skew and curve effects on shear demand

and bearing demand. A simple 3D spine model analysis with LRFD shear correction factors

applied may yield highly non-conservative results. Therefore, a full 3D model such as a beam-

plate model should be built to accurately predict shear and reaction forces in bridge members.

The formulas for skew and curve correction proposed in this report highlight the importance of

aspect ratio in determining the curve and skew effects. These may serve as a basic estimate of the

affect of aspect ratio on skew and curve effects. However, since these equations were empirically

developed for a limited range of bridge models, they are not recommended for design.
52

6.2.2 Bearing Stiffness

In bridge design, bearing stiffness is highly dependent on bearing type. Plain elastomeric

bearing pads are much softer than PTFE bearings and the skew and curve responses of bridges

designed with these two bearing types should be expected to vary significantly. The study of the

effect of softer bearings found in Appendix B yields the following conclusion. Bridges with

softer bearing pads exhibit a less severe skew effect. See Appendix B and Section 3.2.1 of this

report for details.

6.3 Further Research 

This study can be expanded to further quantify other issues not extensively studied here.

These include:

 Other bridge configurations such as multi-span bridges

 Moment response in girders of curved bridges

 Laboratory and field studies

 Bearing stiffness and arrangement

This thesis highlights the major role that aspect ratio plays in determining the skew

effect, curve effect, and coupled skew-curve effect on vertical abutment reaction forces. This

study, however, used a limited range of bridge configurations and types. The effect of skew and

horizontal curvature on abutment reactions of multi-span bridges should be explored. This would

provide relevant data for bridges with bents. Single-column and multi-column bents should be

studied since the transverse response of a single-column bent is different from a multi column

bent. The data presented in this study focuses on 4-cell box girder bridges. The conclusions

would be strengthened by the analysis of 1-cell and 2-cell box girder bridges as well as wide

bridges with cell numbers greater than six. Since this study focused on aspect ratios between 1.0
53

and 8.0, the data could be filled out if aspect ratios of 0.3 to 1.0 were investigated. Other bridge

types including slab bridges and beam-slab bridges should also be investigated.

Moment response in some girders of skewed bridges is generally thought to be less

severe than in non-skewed bridges. All girders are usually designed for maximum moment. This

practice is considered to be conservative. This, however, is not the case for curved bridges. The

outside exterior girder will carry higher moment than the inside exterior girder since it is longer.

The moment in the outside exterior girder will also be increased due to the torsion produced by

the curve effect as presented in this thesis. Therefore, investigation of the effects of skew and

horizontal curvature on girder moment would be useful. Furthermore, the effect of differential

prestressing to counteract differing girder moments is a topic which deserves additional

consideration because of the many complications that arise due to constructability, camber, and

elastic shortening.

Since all of the data from this study was gathered from computer models, field testing

and laboratory verification would be helpful. A significant body of data currently exists, so a

study of existing curved and skewed bridges which highlights the effects of aspect ratio on

abutment reactions would refine the conclusions presented in this study.

This study did not consider a large range of bearing stiffnesses. The data presented in

Chapter 4 reports results for bridges modeled with stiff bearing elements (500,000 kips/in) and

the data presented in Appendix B compares these results to results for bridges modeled with soft

bearing elements (5,000 kips/in). This range of bearing stiffnesses is not sufficient to formulate

strong conclusions on the effects of bearing stiffness. Furthermore, the effects of bearing

arrangement should also be investigated. Caltrans Memo to Designers 7-1 proposes that

abutment skew be considered in bearing design and that bearing spacing may be varied to

accommodate the distribution of vertical abutment reactions along the bridge width. This,
54

however, introduces differential stiffness across the bridge width and may cause an increase in

the skew effect since the some of the shear force will shift to the stiffer side of the bridge. In

design, abutment bearings may also be shifted toward the obtuse corner so that there is a wider

distance between the edge bearing and the acute corner than the distance between the opposite

edge bearing and the obtuse corner. This is not a common design practice, but it may be

beneficial in preventing a skewed bridge from lifting off the bearing as a result of acute corner

uplift.
55

APPENDIX A: MODELING DETAILS


This appendix contains the details for the modeling of the four major bridge

configurations used in this report. The analysis results obtained above can be verified

independently with the information below. The bridge configurations fall into four major

categories based on aspect ratio. The skew angles, central angles, and bearing stiffnesses were

varied to produce the analysis results presented in this report.


56

Model 1: Aspect Ratio 1.0


57

Model 2: Aspect Ratio 2.0


58

Model 3: Aspect Ratio 4.0


59

Model 4: Aspect Ratio 8.0

 
60

APPENDIX B: EFFECT OF SOFTER BEARINGS

This appendix presents the same bridge data that was presented in Chapter 4 except that

the bridge bearings were modeled to be softer. In Chapter 4, the bearings were modeled with a

stiffness of 500,000 kips/in. This was taken to correspond to the stiffness of stiff bearings used in

bridge design such as PTFE bearings. All the bridges in Chapter 4 were also analyzed with soft

bearings with a stiffness of 5,000 kips/in. This stiffness was taken to correspond to the stiffness

of soft bearings such as plain elastomeric bearings.

Data From Chapter 4 Soft Bearings

Aspect Ratio 1.0, Kbearing = 500,000 k/in Aspect Ratio 1.0, Kbearing = 5,000 k/in


50 45
% Abut. Rxn. @ Obtuse Corner

% Abut. Rxn. @ Obtuse Corner
Design Design

45 36 40 36
30 30
40 24 35 24
35 18
30 18

30 12 12
6 25 6
25 0
20 0
20 ‐6 ‐6

15 ‐12 15 ‐12
‐18 ‐18
10 10
‐24 ‐24

5 ‐30 5 ‐30
‐36 ‐36
0 0
0 10 20 30 40 0 10 20 30 40
Abutment Skew [deg] Abutment Skew [deg]

Aspect Ratio 2.0, Kbearing = 500,000 k/in Aspect Ratio 2.0, Kbearing = 5,000 k/in


80 Design 70 Design
% Abut.  Rxn. @ Obtuse Corner
% Abut.  Rxn. @ Obtuse Corner

48 48
70 42 60 42
36 36
60 30
50 30
24 24
50 18 18
12 40 12
40 6 6
0 30 0

30 ‐6
‐12
‐6
‐12
20 ‐18
20 ‐18
‐24 ‐24

10 ‐30 10 ‐30
‐36 ‐36

0 ‐42 0 ‐42
‐48
‐48

‐10 0 20 40 60 ‐10 0 20 40 60
Abutment Skew [deg] Abutment Skew [deg]
61

Aspect Ratio 4.0, Kbearing = 500,000 k/in Aspect Ratio 4.0, Kbearing = 5,000 k/in


140 Design 100 Design

% Abut. Rxn. @ Obtuse Corner
% Abut. Rxn. @ Obtuse Corner 48
48
120 42 42
36 80 36
100 30 30
24
24
80 18 60 18
12 12
6
60 6
0 40 0
‐6
40 ‐6
‐12
‐12
‐18
20 ‐18 20 ‐24
‐24
‐30
‐30
0 ‐36 0
‐36
‐42
‐20 0 20 40 60 ‐42
‐48
‐48 0 20 40 60
‐40 ‐20
Abutment Skew [deg] Abutment Skew [deg]

Aspect Ratio 8.0, Kbearing = 500,000 k/in Aspect Ratio 8.0, Kbearing = 5,000 k/in


250 Design 140 Design
% of Abut. Rxn. @ Obtuse Corner

% of Abut. Rxn. @ Obtuse Corner
48 48
42 120 42
200 36 36
30 100 30
150 24
80 24
18 18

100
12
6
60 12
6
0
‐6
40 0
‐6
50 ‐12 20 ‐12
‐18 ‐18
0 ‐24 0 ‐24
‐30 ‐30
0 20 40 60 ‐36 ‐20 0 20 40 60 ‐36
‐50 ‐42 ‐42
‐48 ‐40 ‐48

‐100 ‐60
Abutment Skew [deg] Abutment Skew [deg]

As can be easily discerned from the plots above, the skew effect is decreased with softer bearings.
62

APPENDIX C: EFFECT OF SPAN LENGTH

Some studies have suggested that span length alone (not aspect ratio) determines the

skew effect. A cursory investigation was performed to explore the effect of span length.

Below is a comparison of results from this variation.

L = 80’, W = 40’, D/S = 0.05 L = 80’, W = 80’, D/S = 0.05

D/S = 0.05, AR = 2.0 D/S=0.05 , AR = 1.0
80 Design 45

% Abut. Rxn. @ Obtuse Corner
Design
% Abut.  Rxn. @ Obtuse Corner

48
70 42 40 36
30
36
60 30 35 24
24
50 18
30 18
12
12
25 6
40 6
0
20 0

30 ‐6 ‐6
‐12 15 ‐12
20 ‐18
‐18
‐24 10
‐30 ‐24
10 ‐36 5 ‐30
‐42
0 ‐48 0 ‐36

‐10 0 20 40 60 0 10 20 30 40
Abutment Skew [deg] Abutment Skew [deg]

L=160’, W = 80’ D/S = 0.025

D/S = 0.025, AR = 2.0
70 Design
% Abut.  Rxn. @ Obtuse Corner

48
60 42
36
30
50 24
18
40 12
6
30 0
‐6
‐12
20 ‐18
‐24
10 ‐30
‐36
0 ‐42
‐48

‐10 0 20 40 60
Abutment Skew [deg]

In these curves, D/S represents cross section depth divided by span length. From the data above,

we see that aspect ratio has more effect than span length on determining the severity of the skew

effect, curve effect, and coupled skew-curve effect. It is also noteworthy that the effect of width
63

(with span length held constant) and depth to span ratio do not affect the skew effect as much as

aspect ratio.
64

REFERENCES

AASHTO (2012), AASHTO LRFD Bridge Design Specifications, 6th Edition with Interims,

American Association of State Highway and Transportation Officials, Washington, D.C.

National Research Council. NCHRP Report 620: Development of Design Specifications and

Commentary for Horizontally Curved Concrete Box-Girder Bridges. Washington, DC:

The National Academies Press, 2009.

Song, Shin-Tai, Y. H. Chai and Susan E. Hida (2003), Live Load Distribution Factors for

Concrete Box-Girder Bridges, Journal of Bridge Engineering, ASCE, Vol. 8, No. 5, pp.

273-280.

Wallace, Mark (1975), Skewed Concrete Box-Girder Parameter Studies, California Department

of Transportation.

Zokaie, Toorak, Mish, K. D., and Imbsen, R. A. (1993) Distribution of Wheel Loads on Highway

Bridges, Phase 3, Final Report to NCHRP Project 12-26 (2).

Zokaie, Toorak, M.ASCE. (2000) AASHTO-LRFD Live Load Distribution Specifications, Journal

of Bridge Engineering, Vol. 5, No. 2, May 2000, pp. 131-138

You might also like