You are on page 1of 113

ANALYSIS AND DESIGN OF BOX CULVERTS

A Thesis Presented to The Faculty


of College of Engineering and Technology
Ohio University

In Partial Fulfilment
of the Requirement for the Degree
Master of Science

--
Ali H. Abdel-Haq/

OH10 UNlLJE8SITY
LIBRARY
TABLE OF CONTENTS

SYMBOLS AND NOTATIONS ............................. i

TABLE OF FIGURES ................................... ii


LIST OF TABLES .................................... vi

CHAPTER ONE

ABSTRACT ..........................................
1.1) INTRODUCTION .................................
1.2) OBJECTIVES ...................................
1.3) FINITE ELEMENT MODELING ......................

CHAPTER TWO

2.1) NONLINEAR HYPERBOLIC OF SOIL ................. 10

2.2) HYPERBOLIC VOLUMETRIC CHANGE ................. 14

2.3) 8-NODE ELEMENTS .............................. 16

2.4) 3-NODE BEAM ELEMENT .......................... 22

CHAPTER THREE

HYSTORICAL REVIEW OF CULVERT DESIGN ............... 25

3.1) Axial Load ................................... 26

3.2) Bending Moment ............................... 29

3.3) Box Culverts with PCC Relieving Slabs ........ 30

CHAPTER FOUR

4.1) EFFECT OF CULVERT STIFFNESS ..................


4.2) EFFECT OF SPAN VARIATION ..................... 38

4.3) EFFECT OF CULVERT GEOMETRY ................... 41

CHAPTER FIVE
MODIFYING DUNCAN EQUATION ......................... 51

5.1) For Uniform Sections ......................... 51

5.2) Poposed Design Procedure ..................... 53

5.3) For Non-Uniform Sections ..................... 55

5.4) Live Loads ................................... 57

5.5) Moment Distribution Between the Crown and


the Haunch ................................... 59

CHAPTER SIX

6.1) ANALYSIS OF CULVERTS WITH PCC


RELIEVING SLABS .............................. 78

CHAPTER SEVEN

7.1) EVALUATION STUDY OF THE VARIOUS


SOLUTION TECHNIQUES .......................... 90

7.2) Empirical Solutions Against Finite Element


Solutions .................................... 91

7.3) A comparison Between Empirical Solutions and


Experimental Data ............................ 93

7.4) Finite E1ement:Solutions Against Experimental


Data ......................................... 94

CHAPTER EIGHT

CLOSING DISCUSSION ................................ 102

REFERENCES ........................................ 104


SYNBOLS AND NOTATIONS

Et : tangent modulus.
Vt : tangent poissonts ratio.
GI : major principle stress.
bj : minor principle stress.
Ei : initial modulus.
Pa : atmospheric pressure.
K : modulus number.
n : modulus exponent.
,E, : unloading reloading modulus.
K,, : unloading reloading number.
C : soil's cohesion.
: angle of internal friction.
Rf : failure ratio.
: axial strain.
: radial strain.
Vi : initial void ratio.
d : constant representing the value of the poisson's
ratio with radial strain.
G : value of confining pressure at one atmosphere.
F : reduction in for a ten fold increase in
[J] : jacobian matrix.
6Q : element displacement.
O : lateral displacement of a beam.
: warping factor.
Ac : cross-sectional area.
E : elastic modulus.
I : second moment of inertia per unit width.
H : height of backfill over the crown level.
S : span of culvert.
Y : soil density.
T : thrust.
Es : soil modulus of elasticity.
P, : uniform soil pressure per unit area.
LL : line load.
Mt : total bending moment in the crown and the haunch.
Mtb : total bending moment in the crown and the haunch due
to backfill load only.
CI
I : correction factor for the bending moment calculated
by using the Duncan equation.
Mtl : total bending moment in the crown and the haunch due
to live load only.
Mcl : bending moment in the crown due to live loads.
Mhl : bending moment in the haunch due to live loads.
KIB : a constant (function of span).
KZB : a consatant.
KIB : a constant (function of span).
K 2 :~ a consatant.
Hmin: minimum cover depth.
Rp : moment reduction factor for the slab projection over
the edges of the culvert.
Pp : a : factor which is dependent on the span of culvert.
angle between culvert wall and vertical at
footing.
t : required slab thickness.
tb : basic slab thickness.
Ral : axial load correction factor.
Rc : concrete strength correction factor.
PCC : portland cement concrete relieving slab.
B,B1: constants for modifying the Duncan equation.
I, : second moment of inertia in the crown's vicinity.
Ih : second moment of inertia in the haunch's vicinity.
CL85 : clay with 85% degree of compaction.
CL90 : clay with 90% degree of compaction.
CLlOO : clay with 100% degree of compaction.
SC85 : silty clay with 85% degree of compaction.
SC90 : silty clay with 90% degree of compaction.
SClOO : silty clay with 100% degree of compaction.
CA90 : coarse aggregates with 90% degree of compactions.
CA95 : coarse aggregates with 95% degree of compaction.
SM85 : silty sand with 85% degree of compaction.
SM90 : silty sand with 90% degree of compaction.
SMlOO*: silty sand with 100% degree of compaction.
(Mtb) : total bending moment in the crown and the haunch
of culvert when =130pcf, and H=3.075ft.
ue, ve : horizontal and vertical displacements respectively
at any point in the element with coordinates x,y
U i , V i : horizontal and vertical displacements rspectively
.
at node (i).
vspvm : poissonls ratio of soil and culvert materials
respectively.
TABLE OF FIGURES

general mesh diagram .................


hyperbolic presentation of a stress-
strain curve ..........................
Fig(2-2) : eight nodes element ...................
Fig(2-3) : 3-node beam element ...................
Fig (2-4) : shear curvature in a beam .............
Fig (3-1) : axial load in culvert .................
Fig(3-2)
Fig(3-3)
: symbol diagram ........................
: bending moment distribution ...........
Fig(3-4) : the coefficients P and RHB ............
Fig (4-1) : arching phenomenon ....................
Fig(4-2) : total moment as a function of culvert
stiffness .............................
total moment and culvert stiffness in
transformed axis ......................
Fig (4-4) : total moment as a function of elastic
modulus of culvert ....................
Fig (4-5) : a comparison between Duncan equation and
the F.E.M .............................
Fig(4-6) : moment ratio and span variations ......
Fig (4-7) : relation between the total bending moment
and the span ..........................
relation between the total bending moment
and the height of backfill ............
Fig (4-9) : total moment as a function of culvert
span and stiffness ....................
Fig (4-10) : a comparison between the F.E.M results
and Duncan solution for different spans
and stiffnesses of culverts ...........
Fig (4-11) : measured and designed shape of culvert .
Fig (5-1) : the constant A for CL85 ...............
Fig (5-2) : the constant A for CL90 ...............
Fig (5-3) : the constant A for CLlOO ..............
Fig(5-4) : the constant A for SC85 ...............
Fig(5-5) : the constant A for SC90 ...............
Fig(5-6) : the constant A for SClOO ..............
Fig (5-7) : the constant A for CA90 ...............
Fig (5-8) : the constant A for CA95 ...............
Fig (5-9) : the constant A for CA105 ..............
Fig (5-10) : the constant A for SM85 ...............
Fig (5-11) : the constant A for SM90 ...............
Fig (5-12) : the constant A for SMlOO ..............
Fig (5-13) : modified Duncan solution to F.E.M results
for different stiffness ratios ........
Fig(5-14) : proportional relation between the line
load and its induced moment ........... 75
Fig (5-15) agreement between Duncan solution and
F.E.M results in calculating moments due
Fig(5-16)
to line load ..........................
moment ratio against stiffness ratio ..
76
77
Fig (6-1) moment reduction due to PCC slab ...... 84
Fig (6-2) changes in total moment due to changes
Fig (6-3)
in slab stiffness .....................
various deflected shapes of culverts due
85

to various slab stiffnesses ...........


different bending moment patterns along
86
culverts due to various slab stiffnesses. 88
Fig (6-5)
Fig (7-1)
a study on the rigidity of slabs ......
different moment distributions obtained
90
from CANDE, SEQ.CON, and experimental
Fig (7-2)
data (backfill load only) .............
different moment distributions obtained
97
from CANDE, SEQ.CON, and experimental
data (live load only) ................. 98
LIST OF TABLES

Table(3-1) : values of the factor K4 ..............


Table(5-1) : numerical values of the constant (b)
34
for various soil types ...............
Table(7-1) : a comparison between the finite element
73
solution and the modified form of the
Duncan equation ......................
Table(7-2) : A comparison between the moment ratio
gg

obtained by using Duncan method, the


modified Duncan method, and the F . E . M
Table(7-3) : a comparison between the experimental
data, the Duncan and the modified
Duncan equation ...................... 100
CHAPTER ONE

ABSTRACT

Box culvert problems are a complicated example of soil


structure interaction where the relative stiffness between
the backfill soil and the culvert materials is a critical
factor in the load carrying capacity of culverts.

Duncan et a1 proposed an equation for the design of


this class of structures. This equAtion doesnlt take into
consideration the soil structure interaction phenomena.
A modified form of the above equation, which is presented
here, with an allowance for the soil properties and
culvert stiffness provides a better agreement with the
finite element solution. Furthermore, the presence of PCC
relieving slabs and their action in transfering live loads
is analyzed and another better agreement with the finite
element method is obtained.

A sophisticated computer program called SEQCON is used


to verify the results obtained from CANDE. The results of
both programs are compared with an experimental data on a
Lane Steel Culvert.
1.1) INTRODUCTION

In the last decade, metal culverts were extensively


used in waterways or as a replacement of highway bridges.
These culverts have different shapes and their design
procedure is, largely, based on field tests and
experience. In some cases when a limited vertical
clearance is available and a large cross-sectional area is
required, like in water conveyence, attention is
focussed on large span culverts such as aluminum box
culverts.

A proliferation in the use of aluminum box culverts has


made it quite necessary to develop a rational basis for
the design of these structures. To accomplish this goal,
an extensive program was undertaken at the University Of
California at Berkley. An important outcome from this
study is attributed to Duncan et all who proposed a design
formula to calculate the bending moment in the crown and
the haunch of aluminum box culverts. This design formula
is applicable to a wide range of culvert spans and is
based on the lowest permissible backfill material. The
effect of soil structure interaction is not included in
this equation and its use is, primarily, intended for
aluminum box culverts. [l]

In their work, Duncan et a1 used the finite element


method to analyze culverts and compare its predicted
solution with the experimental data obtained from the
field. The four nodes (quadratic) or three nodes
(triangular) soil elements are usually utilized in the
finite element formulation . No attempts were made to use
a more sophisticated elements like eight nodes quadratic
elements for soil, or three nodes beam elements for
culverts. These higher order elements are expected to
give a more accurate prediction of the true behavior of
culverts, and ultimately, a better agreement between the
field data and the finite element solution.
1.2 ) OBJECTIVES

The aim of this study is to achieve the following


goals: -

1) To attempt to evaluate the Duncan equation by


studying the effects of its parameters ( span and height
of backfill ) on the bending moment with regard to the
cases of dead load (backfill) and live load (line load
over the crown level) .

2) To perform a parametric study on the effects of

culverts stiffness and soil type and properties on the


bending moment with the hope that an extra variable, which
provides an allowance for culvert stiffness and soil type
and properties, will be introduced to Duncan equation.

3) To extend the analysis of box culverts to include

those with non-uniform cross-sectional properties, (e.g


second moment of area), which would require an evaluation
of the moment distribution in culverts for different
stiffness ratios (Ic/Ih).

4) To analyze and study the influence of PCC (Portland


Cement Concrete) relieving slabs on the induced bending
moment when live loads are applied, paying particular
attention to the live load transfer mechanism through PCC
relieving slabs onto culverts, with any deviation from
the available design form [2] to be proposed when
necessary.

5) To attempt, in the final analysis, to compare an


experimental data obtained from field tests on a 15'-1" x
5' - 8 " Lane culvert [3] against other solutions obtained
from 1) Duncan equation, 2) The modified form of Duncan

equation, and 3) Finite element analysis with both CANDE


and SEQ.CON.

In chapter two, the formulation of the stiffness matrix


for a higher order elements (8-node soil element and 3 -
node beam element) is derived. A brief theoretical
background on the hyperbolic nonlinear model of soil
(Duncan model) is also given.

A historical review on the design of box culverts is


summerized in chapter three. This includes the work of
Duncan et al. which resulted in the Duncan equation for
calculating the bending moment due to backfill load and
live load with and without the presence of PCC relieving
slabs.
A detailed study on the influence of the culvert
parameters on the bending moment carrying capacity is
provided in chapter four. The results of the study on the
effect of culvert stiffness and soil type and properties
is included in this chapter.

Chapter five introduced a modified form of the Duncan


equation which has an extra factor to allow for the soil
type and culvert stiffness. It also includes the total
moment distribution between the crown and the haunch of
culverts which is revised and modified from what was
proposed by Duncan.

A study on the PCC relieving slabs and their action in


transfering live loads is done in chapter six. An
alternative procedure for calculating the bending moment
in culverts in the presence of these slabs is proposed in
the same chapter.

In chapter seven, an overall evaluation study, in a


form of a comparison between various solutions, is carried
out. This includes finite element , empirical, and
experimental solutions of culvert problems.
1.3) F I N I T E ELEMENT MODELING

This investigation will be based on finite element


analysis. The program CANDE (Culvert Analysis and Design)
is basically used in this study. A more sophisticated
program called SEQ.CON (Sequential Construction) which
utilizes 8-node soil elements and 3-nodes beam elements
will be used and the results of this program will be
compared to that of CANDE.

In order to best simulate the appropriate conditions in


the field, the following assumptions and models are
considered :

1) Linear elastic model for culvert materials (steel


and aluminum).

2) Duncan hyperbolic model for the analysis of the


backfill soil materials. The properties of some of the
backfill materials used in this analysis are included in
CANDE. Other soil properties are interpolated from what
is reported in the literature [4].

A general mesh with soil and beam elements is shown in


figure(1-1). In this mesh, its assumed that the culvert
geometry and construction sequence are symmetrical. The
boundary conditions are believed to be as shown on the
same figure, but it may vary at point A when necessary.

A 1 5 v - 1 0 v vx 5 v - 0 v v culvert with 36.9 inches cover over


the crown level is'used for the study of the effects of
culvert and soil stiffness. Different span box culverts
are used for the study of span variation on the bending
moment. Steel and aluminum culverts are used, the above
investigations are combined and employed to modify the
Duncan equation.
CHAPTER TWO

2.1) CONSTITUTIVE EQUATIONS

I n t h i s chapter, some o f t h e constitutive


equations that are used in finite element
formulations are disscussed. The hyperbolic nonlinear
model of soil is used in both, CANDE and SEQ.CON.
The stiffness matrix of higher order elements (8-node
soil element and 3-node beam element) which are used
in the SEQ.CON are discussed. An allowance for the
bending moment due t o the shear in beam element is
also provided. The soil and beam elements used in
CANDE are discussed somewhere else [12] .

HYPERBOLIC NONLINEAR STRESS STRAIN CURVE:-

The hyperbolic model was developed t o account for


the nonlinear strain-stress relation of soil. In the
finite element analysis, generalized Hook's law is
used t o relate the stress strain relation for each
load increment. It was shown by Kondent et a1 that
the stress strain relation could be approximated,
with a reasonable accuracy, by a hyperbola like the
one shown in fig(1-2) [4]. The equation of the
hyperbola can be written in the following form
11
For all soils (except for the fully saturated ones) tested
under unconsolidated undrained conditions, a steeper
stress-strain and a higher strength will be obtained due to
an increase in the confining pressure and consequently, the
value of Ei and ( a l - ~ 3 ) ,would
~~ also increase.

The variation of Eiwith a3 is represented by the following


equation after Janbu :-

K : the modulus number (dimensionless)


n : the modulus exponent (dimensionless)
Pa: atmospheric pressure

For the unloading conditions, the modulus E U r i s used for

both, the unloading and reloading conditions. The value of Eu r


is related to 03by the equation

The compressive strength ((r ,- a 3) at failure may be

given by the Mohr-Columb strength equation :-


REAL

TRANSFORMED

Fig(2-1) A hyperbolic presentation of a


stress-strain curve. [ 4 ]
The compressive strength at failure ( al - 0 3) may be 13

related to the ultimate compressive strength by the failure

ratio Rf where :-

Differentiating equation (2-1) and substituting (2-2),


(2-3),(2-4)and (2-5) in the resulting expression gives an

expression for Et. That is

The above expression enables us to determine the value of

Et for any stress increment since the parameter K,n,c,+ and

R r a r e p r o p e r t i e s of the soil and may be determined


experimentally.
2.2) HYPERBOLIC VOLUMETRIC CHANGE:-

The v a l u e o f u t c a n b e determined from t h e v o l u m e t r i c change

a n a l y s i s i n t h e t r i a x i a l t e s t . The r a d i a l strain,^,, d u r i n g
t h e test i s c a l c u l a t e d from t h e r e l a t i o n : -

,and , a r e t h e v o l u m e t r i c and a x i a l s t r a i n s r e s p e c t i v e l y .

A plot o f E, a g a i n s t E , c a n b e r e a s o n a b l y a c c u r a t e l y

r e p r e s e n t e d by a h y p e r b o l i c e q u a t i o n o f t h e form [4] :-

rearrange

V, is t h e i n i t i a l v o i d r a t i o ( a t z e r o s t r a i n ) a n d d is a

p a r a m e t e r r e p r e s e n t i n g t h e change i n t h e v a l u e o f P o i s s o n ' s
r a t i o w i t h radial s t r a i n . For most of t h e s o i l , t h e v a l u e of
Vi decreases w i t h c o n f i n i n g pressure (unless t h e s o i l is

s a t u r a t e d and t e s t e d under undrained c o n d i t i o n where vi=1/2

f o r any c o n f i n i n g p r e s s u r e ) by t h e equation:-
In this equation, G i s the value of confining pressure at

oneatmosphere and F i s the reduction in vi for a ten fold

increase in a3.Differentiating equation (2-7a) with respect

to c,,substitutingequation (2-8) and eliminating the strain


using equation (2-1) , (2-2), (2-4) and (2-5) , the tangent
value of poisson's ratio may be expressed as follows [4]:-

Vt '
2
- d(O1 - O3 )
(2-9)
Rf ( I-sin@ )( 0,-03 )
2 C cos @ + 2 0, s i n @
11
Alternatively, the tangent poisson's ratio can be found if
the tangent bulk modulus is known. The tangent bulk modulus
expression is a function of minimum compressive stress given
by:-

Bt = % Pa ( cf3/P,)m (2-10)

I$,: bulk modulus number, dimensionless.

m : bulk modulus exponent, typical range 0.0 to 1.0

But Et = 3Bt (1-2ut)

Therefore u, = 1/2 (1-Et/3Bt)


2.3) 8-NODE SOIL ELEMENTS

Higer order elements are usefull since they utilize higher


order interpolation function. Generally, the high order
elements produce a more accurate solution to the differential
equations. Consider an 8-node element with 2 degrees of
freedom at each node, that is u and v displacements. The
displacement function can be written as :

At the nodes, we have :-

F o r i = 1,2 .........,8

I
+ x
a) Global coordinate s y s t e m b) Local coordinate s y s t e m
Equation (2-13) may be w r i t t e n i n t h e l o c a l coordinates 17

( r , ~ ) with r replacing Xand S replacing y. Writing Equation


(2-13) i n terms o f the shape functions and rearranging g i v e s : -

Where : -
,
N = I /4(1 -r)(l -s)(-1 -r-s) N2= 1/4(1 +r)(1 -s)(-1 +r-s)

The chain r u l e i s used t o t r n s f e r derivatives i n t h e global


coordinates i n t o l o c a l coordinates:-
-
and in matrix form :

Where the matrix [J] is the Jacobian matrix : -


For an 8-node element, the Jacobian matrix [J] can be
written in the form : -
Differentiating equation (2-15) with respect to Xand )/ and
substituting the appropriate terms yield: -

=[Dlm[ul with [ D ] = [J 1-Im[F ]


The same procedure can be used for S. [Dl is called the
derivative matrix. The stiffness matrix can be obtained from
[ 5 , 6 ] :-
[c] i s t h e c o n s t i t u t i v e m a t r i x which, f o r a plane s t r a i n
problem, has the form :
, 2.4) 3-NODE BEAM ELEMENTS

A 3-node element is another example of higher order element


which may be used in finite element analysis in order to
improve accuracy. A t h r e e node element has three nodes, one
at each end and the third is somewhere in between. We here
consider a special case where the third node is in the middle
of the beam [ 7 ] .

F i g (2-3)

Each node (i) has two displacement degrees of freedom


associated with it, They are:-

u : axial displacement.
o : t h e lateral displacement of the beam.

@i (%)i + @r [ t h e r o t a t ion o f t h e n o r ~ a Il

Thus, the element displacement may be listed inthhe vector

form:- 6 e = [u, w l e l U* o 2 O 2 u3 w 3 e3 ]

Fig(2-4)
Beam c u r v a t u r e

Assumed d e f o
Actual deformatio
The shape functions in local coordinates are: -

Thus, the field displacement and rotation can be defined


in terms of the shape function and the associated nodal
displacement by simple interpolation:-

The Jacobian Matrix [J] for the beam shown in fig (2-3) can
easily be shown to equal L/2 where L is the length of the
element. The strains are defined in terms of the nodal
displacement and shape functions derivatives by the following
expression :
where&/axis a Pseudo-curvature a n d a i s t h e e f f e c t i v e

shear r o t a t i o n . Again, t h e element s t i f f n e s s matrix [Ke] may


becalculated from t h e r e l a t i o n :

o r i n l o c a l coordinate, a t y p i c a l submatrix l i n k i n g nodes


(i) and ( j ) may be given by :

0 0 - det ( J)ds

O Nj

El : flexuval r i g i d i t y
S h e a r m o d u l u s X C r o s s - s e c t i onal area
S: -aA- - Shear r i g i d i t y
factor t o allow f o r warping
CHAPTER THREE

HISTORICAL REVIEW OF CULVERT DESIGN

Since the early part of history, underground


structures like sewers, drains, culverts and others have
been in use. In the past, several decades ago, it has
been possible to design culverts on a rational basis
with a degree of precision comparable with that obtained
in the design of other underground structures.

Although culverts can be built up from various


materials ( concrete , steel and aluminum ) and in various
shapes (pipes, box culverts, arch culverts and long span
box culverts), the response of these structures to loading
is governed by an interaction between a flexible membrane
and a relatively compressible surrounding medium
(compacted soil fill). Analysis of such systems is
difficult due to the complex interaction mechanism
involved. No closed form solution can adequately
approximate the true behavior. Most manufacturers use
design methods based on formulas that assume a grossly
simplified system, but at the same time have the backing
of considerable experience. Others have developed
empirical methods based on small-scale model studies.
26

"The finite element method has now developed to the


extent where models for soil structure interactions
problems may be formulated to provide an adequate means
for analysis of these structures under working load. A
number of various types of metal culverts have been
individually analyzed by this method and the results
appear to show an acceptable correlation with field
measurements [8]It. In fact, in culvert design we are
mainly interested in calculating the axial load, bending
moments and the deflections of the culvert's members.

3.1) Axial Load :-

The axial load in the culvertts members can be


approximated from the ring compression theory (after
White and Layer) [8] : -

Where P : axial load (kips per it.

y : soil density (pcf)

H : depth of cover over crown (ft.)


span
9 : angle between culvert wall and
the vertical at the footing .
The ring compression theory assumes that there will be
no overburden stress redistribution and it neglects all
interface friction and bending stiffness. Basically,
the theory states that the thrust per unit length in the
culvert wall is constant and is equal to the overburden
stress applied at the crown of the culvert which is
multiplied by one-half of the span of the structure .

Duncan found that the finite element analysis gives


30%-40% larger value of axial load than the value
obtained from the above equation. He proposed that the
ring compression forces in arches and closed shape
culvert structures are approximated as [9] :-

in which
P = kpl*Y*s + kp2*Y*~*s+ kp3*LL

P : axial force .
kpl : a coefficient for axial load due to backfill
up to the crown level .
28
kp2 : a coefficient for axial load due to backfill
over the crown level .
kp3 : a coefficient for axial load due to rise/span
ratio.

For aluminum corrugated plates , typical values of


the constants are

kpl=0.2 , kp2=1.3 and kp3= Ac/1.6H

where Ac : cross-sectional area


H :cover depth

Other researchers proposed charts to calculate the


axial force and deflection in arch culverts [ 8 ] . They
proposed that the two key response parameters for these
structures, the relative crown deflection (6) and the
springline thrust (T) are some function of the input
parameters : -

Ac : the cross-sectional area of culvert (per unit


length).
Es,E : elastic moduli of the soil and culvert materials
respectively.
V,,Vm : poisson's ratio of soil and culvert materials
respectively.
P, : uniform soil pressure per unit area .
Hf : height of soil above crown's level .
29
I : moment of inertia of the culvertfs wall material

The numerical value of the dimensionless parameters


(Es/PRt) and (T/PRt) can be determined from proposed
charts and consequently the values of 6 and T can be
estimated .

Symbols and notations


L81

3.2) ending Moment

The bending moment is the most important criterion for


the design of aluminum box culverts. In these structures,
and according to Duncan et a1 [I], two maximum points of
30
bending moments can take place in the crown and in the
haunch of culverts.

A third point of large bending moment may appear in


the foundation of the culvert as seen in fig(3-3).
Sometimes a failure may take place due to the hinge
formation at these points. The design of the aluminum box
culvert is proposed by Duncan et a1 [l].

Mtb total bending moment in crown and haunch

KZB : 0.053
S : span

H : cover depth

Hmin: minimum cover depth (ft), Hmin = 1.4 for all


spans

The above formula is valid only for 8ft ,< S ,< 26ft
The sum of the crown and haunch moments due to live
load is :-
Mt- *
= K 3 ~ LL * S where

K 3 -
-~ o.oa/(a/sf for S less than 20 ft.
K3B = (0.08 - 0.002(S-20 ft.) }/(H/s)
-2

for 20ft. < S 6 26ft.


LL : line load = AL/K4
AL : axial load
K4 : coefficient from table (1) .
The footing load due to backfill materials PB is given
as :-

And due to live load is


P i = RHB * LL

RHB : load distribution factor.

Fig(3-3)
Bending moment distribution
in symmetrical culvert.
3.3) BOX CULVERTS WITH PCC RELIEVING SLABS

Relieving PCC slabs are usually used over the backfill


materials when culverts are used in highways and roads.
These slabs can be considered as one of the backfill
materials which possesses a relatively very high
stiffness. This could have a very large effect on the
transmission of live loads to the crown of culverts.
Specifically, it will tend to reduce the effect of those
loads by distributing them over a large area.

A quantitative evaluation of the induced moments due to


the applied live loads on PCC relieving slabs is suggested
by Duncan, Seed, and Drawsky who proposed that the
bending moment due to this live load (Mtl) is given by:

Pp : a factor that is obtained from fig(3-4).

S : the span of the culvert.


LL : the equivalent line load, values can be obtained
from table (3-1).

The crown and the haunch moments can be calculated once


Mtl is evaluated.

here Rp is a moment reduction factor for slab


projection over the edges of the culvert.
Pp is a factor which is dependent on the span
of the culvert.

Both Pp and Rp are given in fig (3-4) [2].

An alternative method for the design of PCC relieving


slabs can also be applied. The required thickness of PCC
slab may be determined by using the following equation:-

t : required slab thickness.


tb : basic slab thickness (for slabs on soil with
no underlying culvert.
Ral : axle load correction factor.
Rc : concrete strength correction factor.
Rf : 1.2 (for box culverts with spans less than
26ft).
Relieving Slab No Relieving

Cover 2-wheel 4-wheel 8-wheel All


Depth (ft) axle axle axle axle

Table (3-1)
Values of the factor K4

-
Slob ProJeellon (I

Fig(3-4)
The coefficients Pp and Rp [2].
CHAPTER FOUR

4.1) EFFECT OF CULVERT STIFFNESS

In structures like culverts, which are built from


certain materials and surrounded by a backfill of
different materials, the relative stiffness of the two
adjacent materials play a major role in the load transfer
mechanism to the culvert through the surrounding
materials. In fact, the phenomenon of positive and
negative arching (as can be seen in fig(4-1) )is due to
this relative stiffness. Though much research and many
studies have shown the effect of soil stiffness and
culvert stiffness on the bending moment in culverts [l],
very few proposed designs considered these effects.

+ve arching f ige (4-1) -ve arching


[I21
36
Here is an extensive study of the effects of
culvert's stiffness and backfill stiffness on the bending
moment in the haunch and the crown of the culvert. These
studies are based on 15.87ft span and 5.0ft rise Lane
culvert with 36.075ft cover over the crown level of the
culvert. Duncan's model is used to simulate the
stress-strain relation in the backfill materials. Uniform
cross sectional properties nbased on the equivalent
uniform cross-sectional properties1' are assumed for the
culvert's beam elements [lo].

Fig(4-2) shows . that the total bending moment in


the culvert is dependent on the culvert stiffness .
Furthermore, the nature of the relation between (EI) and
Mt seems to be hyperbolic. Fig (4-3) confirms the
hyperbolic nature of this relation which appears to have
the form :

EI/Mt = B + Bl*(EI) where B and B1 are constants

In the second part of this study, a thorough


investigation will be undertaken to find a way to
determine, quantitatively, the values of the constants B
and B1. Further investigation of fig(4-2) reveals that
silty clay with an 85% degree of compaction (SC85) ,which
is usually the least stiff permissible backfill for
culvert projects, constitutes an upper boundary for the
soil groups used in this study. A fairly constant bending
moment is obtained for a large portion of the curve and is
independent of (EI) . A more noticeable variation in the
total bending moment due to variation in culvert stiffness
is observed in stiffer soils, but a general tendency
toward asymptotic value is seen in every curve. This
suggests that there is numerically, a relative stiffness
value (ESs3/E1) smaller than which , the bending moment
in the culvert becomes constant and independent of the
culvert stiffness.

A more detailed: study of the effect of (EI) on the


total bending moment is seen in fig(4-4). Here, the same
culvert configuration is used, but with varying value of
culvert stiffness (EI) which covers a wide range of
materials including steel and aluminum. No slippage is
assumed to take place between the culvert plates and their
stiffners.

The early portion of the curve is dominantly aluminum,


while the upper portion of the curve is typically steel.
The total beam bending moment value which, is obtained by
38

using Duncan's equation, is larger than the finite


element solution for aluminum culverts, and hence
justifies its use for the design purposes. A 10% to 20%
larger value of total bending moment (Mt) could be
obtained from the finite element solutions on steel
culverts than those obtained by the Duncan equation for
the span under consideration. This variation between the
finite element solution and other solutions, which are
based on Duncan equation, is not typical. As we will see
later, the Duncan equation tends to give a conservative
estimation of the bending moment, especially due to the
backfill load, whether is used for steel or aluminum and
particularly for large span culverts.

4.2) EFFECT OF SPAN VARIATIONS:-

The effect of culvert stiffness has, by far, been


studied, more, for various soil types, but, also, for
the same culvert span and height of backfill. Naturally,
one would predict that the bending moment will increase
due to an increase in the culvertls span. According to
Duncan's equation, the increase in the total bending
moment should be proportional to the cubic value of the
39

span. Fig (4-5) clearly shows that the relationship


between the span and the total bending moment for culverts
with a span larger than 20ft is not exactly parabolic of
the third degree. In fact, .the total bending moment is
slightly increasing with span increment, but not as
sharply as predicted by Duncan. Perhaps, this could be
the reason why an upper limit for the applicability of
this equation was established. A detailed study of the
finite element solution indicated that a less rapid
increase, and sometimes a decrease in the bending moment
at the crown of the culvert is obtained when the span is
increased . This is due to the fact that a large upward
deflection might occur in long span culverts at the early
stage of backfilling and before reaching the crown level.
This upward deflection and its induced moment will
decrease the bending effect due to further backfill over
the crown level. This may be the explanation why the
Duncan equation overestimates the bending moment in
larger span culverts.

The ratio of the crown bending moment to the total


bending moment was proposed by Duncan to be a function of
the span of the culvert. Fig(4-6) shows that this ratio
is somewhat similar to what was produced by Duncan. This
issue will be discussed and analyzed in more details in
the next chapter.
The change in the total bending moment due to the
culvert span variation is studied for various heights of
backfill over the crown level. The three curves do have a
similar pattern to what was produced by Duncan. Another
similar agreement with Duncan's work is obvious in Fig(4-
8) where a linear relationship is obtained between the

applied height of fill and its induced bending moment.


This linear relation is proved to be valid for various
spans as the curves indicate. In fact, curves similar to
what is obtained in Fig(4-7) constitute the basis for the
Duncan equation.

To combine the effect of relative stiffness and


culvert s span, a plot of the total bending moment
against ( E I ~ / ~ * sis
~ ) shown in Fig(4-9). The obtained
points are best fitted by two straight lines which
intersect at a point that may reasonably be represented by
a span of 21ft and 1=0.65. A more general form of
representing this relation is to put it in a non-
dimensional form by dividing the F.E.M results by its
corresponding Duncan solution. This is done in Fig(4-lo).
In this figure, a family of curves is obtained where each
curve represents a certain culvert span.
4.3) EFFECT OF CULVERT GEOMETRY:-

The culvertls geometry appears to be a crucial factor


in the deflection and the bending moment of the crowns of
culverts. As a matter of fact, the measured culvert shape
or dimension might slightly differ from their
corresponding design value. This slight variation in the
culvertls geometry has been shown to be responsible for a
difference of more than 30% in the deflection or bending
moment. These results were concluded from analysis on
some Kaiser aluminum culverts. In several cases studied,
the obtained crown's moments and deflections when the
design shape of the culvert was used, didn't matched
those results obtained when the measured shape of culvert
was used even though the difference in shape was hardly
noticeable, Fig(4-11). This surprising finding was
confirmed by both CANDE and SEQ.CON.

Itls recommended that further investigation should be


undertaken to quantitatively predict the effect of this
variation and perhaps, to explain why it is so
influential on the culvert's reaction.
'I;
(D
Legend

CULVERT STIFFNESS El
fig(4-3) C u l v e r t s t i f f n e s s and bending moment i n
transformed a x i s f o r d i f f e r e n t s o i l t y p e .
-1
Y O *
Fig(4-10) A comparison between t h e F.E.M and t h e Duncan
e q u a t i o n f o r d i f f e r e n t s p a n s and s t i f f n e s s e s
of c u l v e r t s .
~ i ~ ( 4 - 1 1A) shape d i f f e r e n c e between t h e designed
and t h e measured geometry o f 14.83ft span
and 5.67 r i s e c u l v e r t .
CHAPTER FIVE

MODIFYING DUNCAN EQUATION

5.1) For The Uniform Section:-

In the previous chapter, it was established that the


relationship between the culvert stiffness (EI) and the
total bending moment in the crown and the haunch of
culverts (Mt) is hyperbolic of the form:-

In this chapter, a thorough investigation of culverts


with various spans and sectional stiffnesses will be
made, so as to find a method to determine the values of
the constants (B) and (Bl) for any soil type or culvert
stiffness. This will lead to the introduction of a new
factor to the Duncan equation, which would include the
effect of the culvert stiffness and soil type on the
bending moment carrying capacity of culverts.

Fig (5-la) .. .to Fig (5-12a) insure that the relation


between EI/Mt against EI is linear. Each figure is for
52

different soil type and contains curves for various span


culverts. It could be noted that all straight lines
(which represent different spans) do have approximately
similar points of intersection with the y-axis which is
the value of the constant (B). In fact, the points which
are used to construct those straigt lines are fed to a
computer program from which the numerical values of the
constants (B) and (Bl) were determined. It was found that
the values of the constant (B), for the various spans but
the same soil type, are different by a very small amount.
Considering the fact that a small variation in the value
of (B) would have little influence on the value of the
total bending moment (Mt) therefore one can conclude
that an average value of (B) for that soil type may be
used and the slight variation of the value of (B) due to
span variation may be ignored. In other words, it is
assumed that the constant (B) is a function of only the
soil type, and is independent of the culvert span. Table
(5-1) contains the values of (B) for a wide practical
range of soil types.

The constant (Bl) is a function of both, the soil


type and the culvert span. Stiff soils tend to have a
slightly lesser value of the constant (Bl) than those less
stiff. On the other hand, the constant (Bl) is heavily
dependent on the span of the culvert. The value of (Bl)
53

seems to be related to the culvert stiffness by a


hyperbolic relation. Fig(5-lb) .... to Fig(5-12b) enable
the value of the constant (Bl) to be determined for any
kind of soil and culvert span .

5.2) Proposed Design Procedure

Having established a way to calculate the constants (B)


and (Bl) for any kind of soil and culvert stiffness, it
is now possible to include the effect of soil and culvert
siffness on the bending moment carrying capacity of
culverts. The finding will lead to the introduction of a
new factor to the Duncan equation. The following
procedure summarizes how to include this factor :

1) Calculate the bending moment due to the backfill


materials (Mtb) by using Duncan equation for the culvert
under cosideration.

2 Calculate the value of (Mtb) by using Duncan


equation for the culvert span under consideration but for
H=3.075 ft. and Y =130 pcf. Call this value M ~ ~ * .
54

3) From table (5-1) obtain the corresponding value of

the constant (B) for the used backfill materials.

4) For the culvert span and soil type used, obtain


the value of the constant (Bl) from the corresponding
curve of (Bl) against span.

5) Use the obtained values of (B) and (Bl) to


calculate the correction factor from the following
equation:-

6) The modified bending moment value of Duncan equation


is:-
(Mtblf = Cf * Mtb
5.3)Non-Uniform Cross Section:-

Up to this point, all studies have been based on


uniform cross- sectional properties. The culvert
stiffness is assumed to be constant in all culvert
elements. This may not always be true due to the
non-uniform distribution of stiffeners on the culvert
plate. These stiffeners are assumed to fully adhere to
the main culvert plate and hence form a composite
section. In order to account for the variation in
sectional properties, various studies are made on
culverts with a wide range of (ICI) (the culvert
stiffness in the vicinity of the crown to the culvert
stiffness in the vicinity of the haunch).

Since the sectional properties are not the same, the


larger value of the moment of ineria (I) is used in
predicting the total bending moment. These predicted
values are divided by their corresponding finite element
solution so as to visualize the applicability of the
modified form of the Duncan equation to the non-uniform
culvert sections. Fig (5-13) shows that unless (Ic/Ih) is
larger than five, safe and reasonable values of the
bending moment, with regard to the finite element
solution, can be obtained. Since in practice, the ratio
of I, to Ih is most properly less than five; therefore,
the modified form of the Duncan equation can safely be
applied to culverts with non-uniform sectional properties.
5.4) L i v e Loads:-

The effect of live loads is studied when the load is


applied directly above, or near the crown level of the
culvert. These are the worst loading position in which
the live load may be located. Live loads, which are
usually due to vehicular traffic, are converted to their
equivalent line loads by using table (3-1) [I].

A proportional relation between line loads and their


induced bending moments, are obtained from the Duncan
equation, as well as from the finite element analysis,
as seen in Fig(5-14). Furthermore, a very good
agreement, with a reasonable difference, was obtained
from the finite element solution and the Duncan solution.
Fig(5-15) shows a .comparison between Duncan predicted
values and the finite element solution for various span
culverts.

The general agreement between Duncan prediction of the


live load moment and the finite element analysis is
restricted to the lowest permissible quality of backfill.
When a better quality soil is used, a less severe effect
of live load is reflected in the form of less induced
bending moment. ~nvestigations on different quality
58

backfill indicated that a difference of approximately 50%


in the live load moment could be obtained between very
good quality backfill materials (SMlOO), and poor quality
backfill (CL90) .

Its concluded that the Duncan solution, for live


loads, would provide a reasonable conservative basis for
predicting the bending moment and hence can be used
without reservations. Such a conservative approach is
believed to be justified from the practical point of view,
since it provides a safeguard against improper conditions
in the field, like poor compaction for example. If a
good quality backfill is guranteed, then its reasonable
to apply a reduction factor, which shouldn't exceed 0.7,
for the calculated bending moment due to live loads.
Other modifications in calculating the bending moment due
to the backfill materials should be carried out as
mentioned before.
5.5)Moment Distribution Between The Crown
And The Haunch:-

Until now, all reference was made to the total bending


moment in the crown and the haunch of culverts. For
design purposes, we are interested in determining the
bending moment in the crown and the haunch separately.
Duncan et a1 used the span of culverts as the only
criteria for distributing the total bending moment between
the crown and the haunch. A plot of the ratio of the
moment in the crown, to the total moment (Mc/Mt) against
the culvert span, was shown for various soil types in fig
(4-6). This curve is similar to what was obtained by
Duncan et a1 from the finite element analysis. However,
a more regular and systematic variation between the
stiffness ratio (Ic/Ih) and the moment ratio (Mc/Mt) is
obtained in fig (5-16) . Fig(5-16) indicates that various
spans and soil types do affect the moment ratio, but not
as strongly as does the stiffness ratio ( I C I ) It is
suggested that the upper boundary curve should be used for
all culvert span and soil types. This will tend to
slightly overestimate the bending moment in the crown
and slightly underestimate the bending moment in the
haunch. This approach seems to be acceptable since the
experimental data [l] showed that comparable results are
obtained from the finite element analysis, regarding the
60

bending moment in the crown of the culvert. A fairly


conservative estimation is obtained from the finite
element analysis of the bending moment in the haunch,
when it is compared to what was actually obtained in the
field.
(la) lNVlSN03 3Hl
(re) lNVlSN03 3HI
h
-P
n r l
P .rl
u U)

(la) INVlSNLX) 3HI


&
4;t;;
z3
cd m
P a,
mi-'
c cd
0 w
0 a,
n
-rl 0

2 z4-l g
cd

- c d o

OOL x (VI/m
Soil Type The Constant (B)

Table (5-1)
Numerical Values Of The Constant (B)
For Various Soil Types
Span
ft.-in.

~ i ~ ( 5 - 1 3The
) ratio of the modified Duncan equation
to the F.E.M is greater than 1 when the
stiffness ratio 'is less than five, for
different cul~~rrtspans.
a,.
F: m
2 g
h
-I= -P
m
CHAPTER SIX

6.1) ANALYSIS OF CULVERTS WITH


PCC RELIEVING SLABS

It has been indicated in chapter three that one of the


major reasons for using PCC relieving slabs is their
abilites to effectively reduce the severe effects of the
live loads that are applied on, or near, the crown level
by distributing those loads over a large area.

Although Duncan, Seed and Drawsky [2] treated the


presence of PCC slabs as a reduction factor in the
distribution of live loads, they recognized that its
quite unrealistic to merely treat the live load as a line
load in the presence of these slabs.

In this chapter , the effect of PCC slabs, as a


function of their stiffness, in distributing live load
over culverts is studied. It could be argued that the
rigidity of the slab is a dominant factor in transfering
the live load to culverts. If the stiffness of the slab
is increased, a general tendency toward uniformly
distributing the line load is observed. When the
stiffness becomes very high, the slab is said to be
79

rigid. That is, it tranfers the line load into a


uniformly distributing load. The rigidity of one foot of
concrete slab is examined for different cases of culvert
spans and heights of backfill.

In fig (6-l), the effect of replacing the last layer of


soil backfill by a concrete slab of one foot thickness and
a projection of one foot over the edge of the culvert, is
quite noticeable. A reduction factor of more than 45% in
the total moment is obtained when these slabs are
introduced. Furthermore, this reduction factor seems to
be, relatively, independent of the span of the culvert
since different span culverts tend to have similar
reduction factors.

In road design, a PCC relieving slab is not the only


finishing layer used over the backfill materials. In many
cases a less stiff finishing layer, like pavement, is
frequently used. The variation in the material stiffness
of the relieving slab is expected to be critically
important in transfering the live load onto culverts. To
account for these variations, hypothetical values of slab
modulus (Eslab) are used. Those moduli have a range of
values between an extremely rigid materials on one hand,
and an ordinary backfill soil layer on the other. A
summary of the results is presented in fig(6-2) which
80

depicts a considerable reduction in the total bending


moment, both in the crown and the haunch of the culvert,
as a result of increasing the slab stiffness. This
increase in (Mt) is not indifinite, but actually, it
approaches a certain value after which it becomes quite
independent of the slab stiffness. This value of the
constant bending moment is achieved when the relieving
slab becomes so rigid that it can act as a rigid body.
That is because of the small relative deflection between
various points within the slab. Such a case would imply
that the line load applied over the slab is transmitted
into a uniformly distributed load underneath it. If the
above argument is true, then a line load applied on a
very rigid slab and its equivalent uniformly distriuted
load should be expected to yield the same deflection and
bending moment pattern along the culvert plate. A plot of
these patterns for slabs, with various elastic moduli,
is shown in fig(6-3) and fig(6-4) . It is clearly seen
that the deflection pattern and the moment distribution
along culvert plate is identical for both cases, the case
of a line load over a rigid slab and the case of its
equivalent uniformly distributed load.

For the above conclusion to be practically applicable,


it should be valid for a practical range of backfill
height and culvert spans. The same loading conditions are
81

applied on a 141-10uculvert, but with different backfill


materials between the crown of the culvert and the slab.
Fig(6-5) shows that a concrete slab would behave as a
rigid body provided that the height of the backfill under
these slabs is greater than the minimum which is 1.4ft.
Similar results are obtained from a 25'- 3" span culvert
with different heights of backfill.

The important conclusion from the above argument is


that for a practical range of the culvert's spans and the
heights of backfill, a one foot PCC with one foot
projection over the edge of the culvert is rigid enough to
transmit live loads into uniformly distributed loads. In
design procedures, itls only required to calculate the
equivalent uniformly distributed load for the applied live
load and carry on calculations as for backfill materials.
This could be done by treating the applied line load as
an increase in the density of the bacfill layers over the
crown's level. In such a case, the method described in
the last chapter may be used to calculate the bending
moment without any reference to live load. If a slab of
less than one foot thickness or is made from other
materials, other than concrete, a correction factor
depending on the geometrical and the material stiffness
(EI)slab may be obtained from Fgi (6-6) and applied to the
calculated bending moment. Under severe loading
82

conditions like small height of backfill or very poor


soil, the Duncan, Seed, and Drawsky approach may safely
be used to predict the bending moment due to line loads.
-a F
.d
0
3
-a a
t
c
.d
-a R -P
P k
ul

cd a,
r i
Cc
>
r i

-a hr
5
0

Y)
u E
"- m
c
k .ti
0 a
-' Pa,
o e
a,
k a,
c

k ,G
rTl P
-9 h
2 xx00000
Y X Y X Y

+\
:.I~~~@I;EE~EE~
~ ~ 0 0 0 0 0 0 0 ~

OWW
r'd
cd
o h
-4 ri
Ti f!
F4 0
cd k
l a b stiffness
E=IOK

e !
!K --
E=lOOK
UB - -
0 E500K
A E=lOOOK
X E=1500K
V EaOOOK
4- ES500K
0 E=3000K
E=6000K
0 - -
El U-D-L

Fig(6-ba) Different bending moment distribution in culvert


due to different slab stiffness.
-
--1
I -
I--
40 - 60 80 100 120
DISTANCE ALONG CULVERT
~ i g ( 6 - 4 . b )A n i d e n t i c a l b e n d i n g moment p a t t e r n s d u e t o t h e
a p p l i c a t i o n o f a 3ine l o a d and i t s e q u i v a l e n t
uniformly d i s t r i b u t e d load.
CHAPTER SEVEN

7.1) EVALUATION STUDY OF THE VARIOUS


SOLUTIONS TECHNIQUES:-

This study is intended to carry out some comparison


investigation on solutions to culvert problems obtained
from empirical equations (the Duncan equation and the
modified Duncan equation.), finite element solutions
(CANDE and SEQ.CON), and an experimental data (Lane Steel
Culvert). It is well admitted that a comparison of only
one field test with theoretical solutions doesn't provide,
by any means, enough facts to ascertain the suitability
of a particular solution. Relying, though, on the facts
of what was reported in the literature about the behavior
of culverts in the, field, a general judgment on the
different theoretical methods (whether finite element
method or empirical equations) could be obtained.
7.2) Empirical Solution Against
Finite Element Solutions:-

A comparison between the finite element solution, the


Duncan equation and the modified form of Duncan equation
for various culvert parameters and live loads has been
carried out. Some of the obtained results are included
in Table(7-2) and Table(7-2). For all culvert parameters
and line loads used, the modified form of the Duncan
equation showed a better agreement with the finite element
solutions more than the original form. In fact, the
modified form of the Duncan equation is more powerful in
predicting the bending moment due to backfill loads or
other forms of dead loads. The bending moment due to live
loads is calculated according to the Duncan equation,
without any modifications, since it was decided that no
correction should be applied.

Another good prediction of the bending moment is


obtained when a concrete relieving slab of one foot in
thickness is applied over a 14'-1011 span culvert as shown
in the table (7-1). The PCC relieving slab did,
successfully, transfer the line load into a uniformly
distributed load.
92

Satisfactory prediction of the moment ratio was also


obtained. The predicted values of the moment ratio are in
better agreement with the finite element solution, and
they provide a better basis for design than the Duncan
method.

7.3) A Comparison Between Empirical


Solution And Experimental Data:-

Experimental results obtained from tests on lane steel


culverts [3] were used to compare the total bending moment
in culverts with those obtained by using the modified and
the original form of the Duncan equation. Standard
methods for converting the measured stresses into bending
moments were used [ll]. It was decided to compare data
obtained from the September test (with good soil) only,
this was because of the presence of some erroneous data
points in the crown area that were reported from the May
test (with the poor quality soil).

Table (7-3) summerizes the results of this comparison


and shows that the modified form of the Duncan equation
provided an underestimation of the total bending moment.
On the other hand, the Duncan equation provided a
slightly more conservative value of the bending moment.
This shortcome in the modified form of the Duncan equation
was due to the unexpected high experimental value of the
bending moment in the haunch of the culvert. In the crown
of culverts, a general agreement was obtained between the
experimental values and the predicted values of the
bending moment. Actually, the modified form of Duncan
equation was based on the finite element solution obtained
from the CANDE. As we will see soon, CANDE was unable to
anticipate the high moment value in the haunch.
7.4) Finite Element Solutions Against
Experimental.Solution:-

The finite element analysis is carried out by using


both programs, CANDE and SEQ.CON, which basically differ
in the order of their elements. The backfilling process
is simulated by assigning different construction sequences
numbers for the soil elements in CANDE, and by using the
embankment option in SEQ.CON. The dense liquid process
used in SEC.CON to simulate the backfilling operation is
quite representative of the actual conditions in the
field. In this process, the modulus of the newly added
element is reduced by a factor of one hundred so to
simulate the looss backfill layer (befor compaction) which
possesses weight but not stiffness. A normal value of
this modulus is assigned to the backfill element before
the addition of subsequent layers. The nodal forces
resulted from the addition of a new layer is calculated
from the following eqation [13] :

{F) : nodal forces.


y : soil density.

[N] : interpolation function


v : volume of lift incerement.

CANDE adopts the incremental costruction techniques


where the stiffness.matrix for the first soil increment
and associated loads is computed and consequently,
displacements, stresses, and strains are calculated.
When the second increment is added, the combined
stiffness matrix for the first and second increment is
computed. However, the slight variation in the
simulation of the backfilling process, mentioned above,
is not expected to yield much difference in the computed
bending moment which is our main concern.

In both computer programs, and when using a nonlinear


model of soil, the state of stress is calculted for each
soil increment while soil modulus and poissonls ratio are
updated acordingly. A hyperbolic nonlinear model for soil
(Duncan model) is provided in CANDE and SEQ.CON, but it
was noticed that SEQ.CON uses different techniques to
calculte the poissonls ratio. this required the subroutin
in SEQmCON, which is responsible for calculating the soil
parameters for the hyperbolic model, to be modified to a
form similar to that of CANDE. In this case, any
difference in solution which could be attributed to the
soil model is avoided and a better basis for comparison
will be provided.
Finite element solutions obtained from both, CANDE and
SEQ.CON., were compared with the experimental data on the
Lane Steel Culvert. Fig(7-1) and fig(7-2) compare the
bending moment due to CANDE solution, SEQ.CON. solution,
and experimental data obtained from the application of
aj ri-
F= c
a,a 0
a c
3 cdd
P (d
a, - 0
P W d
n
C Z d
04f+
KG 0 . 4
.rl h
k -4:
Span H Ic LL Soil Moment Ratio (M,/Mt)

(ft) (ft) Ih lb/in type Duncan M-Duncan F,E.M

Table ( 7 - 2 )
A comparison between the moment ratios
obtained by using Duncan method,the
modify Duncan method and the finite
element method.
Total bending moment Mt (Lb.in/in)

Duncan Modified Duncan Experimental

Bending moment ratio (Mc/Mt)

0.58 0.50 0.41

Table (7-3)
A Comparison between the
experimental data, the Duncan
and the modified Duncan equations
CHAPTER EIGHT

CLOSING DISCUSSION

Throughout this study the following points are believed


to be particularly important in the analysis and design of
culverts:-

1) Introducing a new variable to the Duncan equation


would provide a better agreement with the finte element
solution. Generally speaking, the finite element
solution provides a conservative basis for culvert design
espicially in the haunch of culverts. The modified form
of the Duncan equation should be checked against some
experimental data to assess its applicability to predict
the culvert moment. However, it's believed that the
modified Duncan equation provides a more realistic
approach to culvert design for the case of backfill,
particularly when large span culverts are considered. The
bending moment due to the live load is believed to be
adequately covered by the Duncan equation.

2) A fairly rigid concrete slab has a substantial usage


103

due to its influence on the loading mode. A live load


could be transmitted through those slabs into a uniformly
distributed load. This finding, if ascertained
experimentally, would lead to a practical approach to
encounter the severe effects of live loads on culverts.

3) The finite element provides a rational conservative

ground for the design of box culverts. A very good


prediction of the bending moment in the crown's area is
usually obtained from the finite element method. In the
haunch area, a fairly conservative values are obtained
from the that method. However, a more sophisticated
finite element formulation, which utilizes higher order
elements like eight node elements and three node beam
elements with more realistic techniques (like SEQ.CON),
is expected to yield more accurate prediction of the
bending moments and the deflections of culverts.
REFERENCES

1- J.M.Duncan, and R.H.Drawsky, "Design and Behavior of


Aluminum Box culvert^^^ . Report NO. UCB/GT/83-04, May
1983.

2- J.M.Duncan, R,B.Seed, and R.H.Drawsky, IfDesign of


Corrugated Metal Box culvert^^^. Transportation Research
Record 1008, 1985.

3- C.D.Gorman, llA comparison of Field Test Results with


Theoretical Analysis of Lane Metal Products, Low Profile
Box culvert^^^ . Report No. 79-67-1, Bethlehem Steel
Corporation, April 1981.
4- Kai S. Wong, and J.M.Duncan, "Hyperbolic Stress
Strain Parameters for Non-Linear Finite Element Stress
Movement in Soil Masses, Report No. TE-74-3, University
Of California-Berkley, July 1974.

5- Paul E. Allair, "Basic of the Finite Element Methodt1.


W.m.c.Brow Publishers, College Division, Dubuques-Iowa,
1985.

6- G.T.F.Ross, ItFinite Element Method in Structural


Mechanics1I. Ellis Horwood Limited, 1985.
7-E.Hinton, and D.R.J.Owens, "Finite Element Programingt1.
Academic Press Inc. (London) LTD 1977.

8- R.C.L.Flint, and J.N.Kay, IfResponse of Corrugated


Metal Arches to Soil LoadIt, Transportation Research
Record 878.

9- J .M.Duncan, ItBehavior and Design of Long-Span Metal


Culvertu, Journal of the Geotechnical Engineering, March
1985.

10- Wei-Wen Yu, "Cold Form Steel Designtt, A Willey-


Interscience Publications, 1985.

11- David B. Beal, I1Behavior of a Corrugated-Metal Box


Culverttt. Research Report No. 90, Engineering Research
and Development Bereau, New York State Depatment of
Transportation, June 1981.

12-M.G. Katona, J .M.Smith, and R.S .Oello, ItCANDE- A


Modern Approach for: the Structural Design and Analysis of
Buried Culvertstt. Report No. FHWA-RD-5, Dederal Highways
Administration, Office of Research and Developments,
Washington D.C 20590, October 1976.

You might also like