You are on page 1of 14

CULVERT-SOIL INTERACTION FINITE

ELEMENT ANALYSIS
By Josann Duane, 1 Rees Robinson, 2 and Charles A. Moore, 3 M. ASCE
Downloaded from ascelibrary.org by IMPERIAL COLLEGE LONDON on 05/15/13. Copyright ASCE. For personal use only; all rights reserved.

(Reviewed by the Pipeline Division)

ABSTRACT: A culvert-soil structural model developed using general purpose fi-


nite element codes is described. Features of culvert specific codes such as in-
cremental construction and Duncan's soil model are incorporated in the model
via MARC user coded subroutines. Model evaluation through comparison of
numerical results with results of prior empirical and analytical work establishes
model validity. Our work demonstrates that finite element analysis performed
using general purpose codes models the culvert-soil interaction as effectively
as culvert specific codes. Furthermore, both the interactive graphics and ana-
lytical capabilities of general purpose codes extend beyond those of culvert spe-
cific codes making general purpose codes applicable to a wider range of prob-
lems. This work guides engineers using general purpose codes in evaluating
culvert-soil interaction.

INTRODUCTION

The goal of this work is to develop the analytical tools to evaluate


flexible culvert design and installation procedure. The analytical model
for flexible culvert design is developed using the finite element m e t h o d
by adapting general purpose finite element codes.
Although the finite element m e t h o d is routinely used in aircraft and
automobile design, it is not used extensively in flexible culvert design.
The social and economic consequences of culvert design failure usually
do not justify the cost of culvert analysis to the extent that the conse-
quences of design failure do in the aircraft industry. However, as hard-
ware costs continue to drop, it is conceivable that culvert analysis using
the finite element method soon may become an industry standard.
Presently, most culvert design a n d analysis is done using relations
formulated from extensive test data a n d assumptions about culvert be-
havior. Two empirical formulas have been accepted by the industry for
evaluating flexible pipe design: Spangler's Iowa formula developed in
1938 (21) and the ring compression theory developed by White and Layer
in 1960 (23).
Spangler assumes a uniform vertical pressure distributed across the
top of the pipe equal to the Marston load divided by the pipe diameter.
This pressure is resisted by pressure along the bottom of the pipe dis-
tributed over the supported portion of the pipe defined by the b e d d i n g
angle. The horizontal pressure is a function of the m o d u l u s of soil re-
action, E ' , an empirical quantity d e p e n d e n t on soil type. Spangler as-
sumes a horizontal parabolic pressure distribution acting over the mid-
a
Asst. Prof., Dept. of Engrg. Graphics, Ohio State Univ., Columbus, OH.
2
Struct. Designer, Cannon and Assoc, Consulting Engrs., Tucson, AZ.
3
Prof., Dept. of Civ. Engrg., Ohio State Univ., Columbus, OH.
Note.—Discussion open until October 1, 1986. To extend the closing date one
month, a written request must be filed with the ASCE Manager of Journals. The
manuscript for this paper was submitted for review and possible publication on
February 26, 1985. This paper is part of the Journal of Transportation Engineer-
ing, Vol. 112, No. 3, May, 1986. ©ASCE, ISSN u733-947X/86/0003-0250/$01.00.
Paper No. 20606.

250

J. Transp. Eng. 1986.112:250-263.


die 100° of the pipe equal to the modulus of soil reaction times the
horizontal deflection divided by the pipe diameter. By assuming that the
pipe deforms from a circular to an elliptical shape, Spangler formulated
horizontal' deflection as given by the Iowa formula.
Unlike Spangler, White and Layer assume that the failure of under-
Downloaded from ascelibrary.org by IMPERIAL COLLEGE LONDON on 05/15/13. Copyright ASCE. For personal use only; all rights reserved.

ground pipe is dependent on ring compression rather than pipe deflec-


tion. White and Layer's theory assumes a uniform normal pressure dis-
tribution around the pipe equal to the density of the soil times its height
above the pipe. This pressure produces a compressive force in the pipe
equal to the circumferential pressure times the pipe radius.
Current design practice for flexible pipe generally uses a combination
of the ring compression theory and the Iowa formula. Most codes require
a pipe thick enough to resist the ring compressive force. Using thickness
established by ring compression theory, deflection is required to be less
than 5% as calculated using the Iowa formula.
Burns and Richards (12) derived closed form solutions to the partial
differential equations governing soil-culvert interaction for a simplified
two-dimensional elastic model. This model is of interest because it pro-
vides a limiting case check for either empirical models or analytical models
based on numerical methods of solution. Burns and Richards' formulas
substantiate both Spangler's and White and Layer's theories when the
shear stress between the soil and the pipe wall and the bending resis-
tance of the pipe are set to zero and the resistance to circumferential
stress is infinite.
Burns and Richards' simplifying assumption of linear, elastic, homo-
geneous soil infinite in extent made their formulation impractical as a
design tool. These simplifying assumptions need not be made if the gov-
erning equations are solved using the finite element method.
With the increased availability of computers to solve engineering prob-
lems, the finite element method (3,20) has been widely applied to solve
problems complicated by (irregular geometry, nonhomogeneities, non-
linear materials, and space or time dependent boundary conditions.
Using the finite element method, a complex structure is evaluated by
subdividing the space occupied by the structure into regions small enough
that the material properties and boundary conditions can be simply rep-
resented over the subdivision. These subdivisions are called elements.
The governing partial differential equations are solved for each element.
A solution for the entire structure is obtained by forcing the solutions
for the individual elements to match at a finite number of points in space
called nodes.
When applied to flexible pipe analysis, the finite element method has
the potential to accurately analyze the soil-structure interaction (2,4,5,16).
Two culvert specific finite element codes, CANDE (11) and FINLIN (13),
and one soil-structure specific code, SSTIPN (22), were developed to model
soil-structure interaction.
CANDE, FINLIN, and SSTIPN all use a bending element to represent
the pipe. CANDE and SSTIPN use a straight beam/column element
consisting of two nodes with three degrees of freedom at each node.
FINLIN's element has the same properties with the exception of beam
geometry which is curved instead of straight. All three programs use
plane strain elements to represent the soil and an interface element to
251

J. Transp. Eng. 1986.112:250-263.


model the interaction between the pipe and soil.
Each code models the soil differently, but all in some way account for
nonlinear, inelastic soil properties. CANDE has a linear elastic, an over-
burden dependent and two variable modulus soil models. The variable
modulus models are based on the Duncan and Hardin soil models.
Downloaded from ascelibrary.org by IMPERIAL COLLEGE LONDON on 05/15/13. Copyright ASCE. For personal use only; all rights reserved.

FINLIN models the soil material properties by curve fitting test data.
SSTIPN uses Duncan's soil model.
All three codes use an incremental construction solution technique.
The construction simulates pipe installation by adding a layer of soil ele-
ments during each load step. Forces are applied equal to the weight of
the newly added layer. This both simulates the construction practice in
the field and provides the stress-strain increments to be used for non-
linear soil models.

FINITE ELEMENT ANALYSIS USING GENERAL PURPOSE CODES

As an alternative approach to analysis using culvert specific finite ele-


ment codes, culverts can be modeled using general purpose finite ele-
ment codes. Because of the versatility of general purpose codes, large
industrial and government research organizations have invested heavily
in the development of these codes and interactive graphics to assist the
analyst in model preparation and results evaluation.
We used two general purpose finite element codes, SUPERB (9) and
MARC (14,15) to model the culvert-soil system. SUPERB is an integral
patt of General Electric SAE International's I-DEAS program (10). Struc-
tural analysis using I-DEAS progresses in three phases: geometric mod-
eling, finite element analysis, and results evaluation. Color used in all
three phases of analysis increases the density of information displayed
on the screen. In the first phase, a geometric model is built interactively.
Solid modeling algorithms insure that valid and complete geometry is
created (9,18). The geometric model is evaluated by displaying it as a
wire mesh or as a color coded realistically shaded solid or surface. With
the geometric model completed, the analyst interactively subdivides
complex parts and assemblies into a mesh of elements for analysis. A
finite element input data deck is prepared and the model code instruc-
tions are executed, usually in batch mode. Using the OUTPUT DISPLAY
(7) module of I-DEAS, results are evaluated interactively. For example,
the analyst can plot overall stress contours, then zoom to a region of
high stress concentration and plot enlarged details of interest.
General purpose codes provide for optimizing model design to use
computer resources efficiently. For example as each element is passed
into a wavefront solution, the storage space required for the element
nodes is activated and remains activated until all elements having the
same node have passed through the solution. Minimizing the wavefront
requires an element numbering scheme which minimizes the number of
nodes requiring storage at any given time. General purpose codes such
as SUPERB compute the optimum element numbering scheme.
Most general purpose finite element codes have extensive element and
material property libraries. In addition to linear elastic analysis, a variety
of other analytical methods are available. Thus, these codes can model
features of culvert-soil interaction not modeled by culvert specific codes.
252

J. Transp. Eng. 1986.112:250-263.


For example, MARC has viscoelasticity and plasticity models with von
Mises and Mohr-Coulomb yield criteria. MARC also provides creep be-
havior models and models with hydrostatic yield dependence. Both MARC
and SUBERB model culverts where parameters vaiy along the axis of
the pipe using three-dimensional elements.
Downloaded from ascelibrary.org by IMPERIAL COLLEGE LONDON on 05/15/13. Copyright ASCE. For personal use only; all rights reserved.

CULVERT MODEL DEVELOPMENT USING GENERAL PURPOSE FINITE


ELEMENT CODES

In order to accurately model the culvert-soil interaction using general


purpose codes, the mesh dimensions, boundary conditions, type of
analysis, element type, and material properties must be accurately spec-
ified using available code features.
We chose to model a culvert installation symmetrical about the vertical
center line of the pipe, so only half of the system needed evaluation.
For our model the dimension of the culvert-soil system in the longitu-

1/2 Trench Width

Existing Backfill
Soil
Initial Haunch
Backfill

Bed
^ ^

FIG. 1.—Relative Dimensions of Pipe, Trench, and Existing Soil Used in Finite
Element Modeling of Culverts: Case 1 (S < 5 Times Pipe Radius, G = 2R, W =
5R, and D = 1AR); Case 2 (S > 5 Times Pipe Radius, C = S/2 - R/2, W = S/2
+ 5R/2, and D = 1.4R)
253

J. Transp. Eng. 1986.112:250-263.


dinal direction is large compared to the mesh dimension. Thus, we were
able to perform a plane strain analysis without loss of modeling integ-
rity. Fig. 1 shows the pipe installation including boundaries between soil
types filling the trench. The interior boundaries were developed accord-
Downloaded from ascelibrary.org by IMPERIAL COLLEGE LONDON on 05/15/13. Copyright ASCE. For personal use only; all rights reserved.

ing to the SDTM standards for pipe installation (1). The outer bound-
aries were developed according to a study done by McVay at the Univ.
of Massachusetts (16).
Boundary conditions are modeled by constraints on soil movement.
Horizontal movement along the vertical mesh edges and center line is
not permitted. The horizontal edges are far enough from the culvert so
as not to be affected by it. Center line motion is constrained because of
symmetry. Pinned supports are used along the mesh bottom to repre-
sent a boundary sufficiently far from the culvert so as not to affect it.
Our linear elastic model uses 24 in. (61 cm) diameter flexible pipe 0.31
in. (8.0 mm) thick with a Young's modulus of E = 963.6 ksi (6,650 MPa),
Poisson's ratio of y = 0.40 and a density of 127 pcf (20,000 N/m 3 ). The
existing soil has a Young's modulus of E = 174 psi (1,200 kPa), Poisson's
ratio of 7 = 0.35 and a density of 111 pcf (17,500 N/m 3 ). The trench was
filled with bedding soil having a Young's modulus of E = 1,740 psi (12,000
kPa), Poisson's ratio of y = 0.35 and a density of 127 pcf (20,000 N/m 3 ).
The material properties were either selected from the material property
libraries available in SUPERB and MARC or incorporated in the MARC
program through subroutines that we wrote.
As the first step in analyzing the culvert-soil interaction, we generated
closed form elastic solutions for the limiting case of linear, elastic, ho-
mogeneous soil infinite in extent. The check for correspondence be-
tween the SUPERB finite element model and elasticity theory was made
using the normalized bending strain factor. According to elasticity the-
ory of Burns and Richards (12), a pipe buried in an infinite homoge-

Crown 45 Springline 45 Invert


Position on Pipe

FIG. 2.—Closed Form Solution, Analytical Solution, and Finite Element Numerical
Solution Equivalence in Region where Both Are Valid: Culvert Surrounded Ho-
mogeneous Linear Elastic Soil Infinite in Extent
i
254

J. Transp. Eng. 1986.112:250-263.


neons medium will deform as an ellipse. Under these conditions the
normalized bending strain factor, Df, is three at the invert crown and
springline of the pipe. Df for a pipe with diameter, D, thickness, t, ver-
tical deflection, Ay, and bending strain, eb, is given by
Downloaded from ascelibrary.org by IMPERIAL COLLEGE LONDON on 05/15/13. Copyright ASCE. For personal use only; all rights reserved.

D D
W
V'Tyl
As can be seen from Fig. 2, our model generates a D/ factor of about 3
at the crown, springline and invert, which is the same as that predicted
by Burns and Richards' closed form solutions.
In modeling the culvert-soil interaction, we began with mesh geom-
etry similar to that used by the CANDI culvert specific finite element
code. A beam element was used for the pipe. The soil was modeled
using both plane strain linear quadratic and plane strain parabolic qua-
dratic elements. We achieved a numerical solution consistent with Burns
and Richards' closed form analytical solution. A model with a mesh of
a mesh of plain strain used computer resources more effectively than a
model with a mesh of plane strain parabolic quadratic elements. The
optimum geometry and element usage is shown in Fig. 3.
Because semi-empirical methods such as White and Layer's ring
compression theory and Spangler's Iowa formula are used extensively
in culvert installation design, we also compared the results our model
with these semi-empirical models. White and Layer's theory states that

FIG. 3.—Finite Element Mesh for Optimum Use of Computer Resources


255

J. Transp. Eng. 1986.112:250-263.


TABLE 1 .—Comparison of White and Layer's Theory with Culvert Model Formu-
lated Using General Purpose Finite Element Codes

Stiffness FEAe,
(psi) Crown Springline Invert White and Layer, e,
Downloaded from ascelibrary.org by IMPERIAL COLLEGE LONDON on 05/15/13. Copyright ASCE. For personal use only; all rights reserved.

(1) (2) (3) (4) (5)


9.73 0.0001697 0.0003731 0.0001898
9.73 0.0001944 0.0004286 0.0001925
9.73 0.0001786 0.0005036 0.0001806 0.000259
9.73 0.0001951 0.0003538 0.0002197
9.73 0.0003538 0.0003519 0.0002165
0.973 0.0003432 0.0007770 0.0003890 0.000559
97.3 0.0000815 0.0001757 0.0000898 0.000120
973.0 0.0000384 0.0000815 0.0000413 . 0.0000558

a uniform compressive pressure equal to the soil density times the height
of fill above the pipe will act normal to the pipe resulting in a com-
pressive "thrust" force equal to this pressure times the pipe radius. Ta-
ble 1 shows the strain due to thrust at the crown, springline and invert
computed by both our finite element model using homogeneous soil with
linear material properties and that computed by White and Layer's the-
ory. In all cases the finite element thrust strains fluctuate about the thrust
strain calculated by White and Layer's theory with the average finite
element results nearly coinciding with White and Layer's predictions.
Due to the use of the empirical modulus of soil reaction, E', Spangler's
Iowa formula is not directly comparable to the finite element computa-
tions. Work has been done by Parmalee and Corotic (18) to correlate the
two. They found, by field testing, that the load acting on the pipe, Wc,
is best represented by the weight of the column of soil above the pipe
divided by 1.38. Using this load and the pipe deflection, the modulus
of soil reaction is calculated using the Iowa formula. The E' calculated
in this way is divided by 1.58 to obtain the elastic modulus, £, needed
for finite element analysis. For the zero trench width case, where all the
soil elements are given an elastic modulus of 174 psi (1,200 kPa), a value
of 183 psi (1,260 kPa) is calculated using the Iowa formula as listed in
Table 2. For the full trench width, a value of 1,724 psi (11,900 kPa) is

TABLE 2.—Comparison of Spangler's Iowa Formula with Culvert Model Formu-


lated Using General Purpose Finite Element Codes
Trench width Stiffness (psi) Deflection (in.) E' (psi) E (psi)
(D (2) (3) (4) (5)
3.0 9.73 0.0806 2,243 1,453
2.0 • 9.73 0.093.7 1,971 1,248
1.5 9.73 0.1624 1,126 713
0.0 9.73 0.5965 290 183
7.0 9.73 0.0680 2,724 1,724
3.0 0.973 0.0766 2,436 1,542
3.0 97.3 0.0730 2,316 1,466
3.0 973 0.0336 3,178 2,011

256

J. Transp. Eng. 1986.112:250-263.


calculated to compare with the input value of 1,740 psi (12,000 kPa).
However, when the pipe reaches a stiffness of 973 psi (125 kPa), the
Iowa formula is no longer accurate since it was developed for flexible
pipe.
Downloaded from ascelibrary.org by IMPERIAL COLLEGE LONDON on 05/15/13. Copyright ASCE. For personal use only; all rights reserved.

This evaluation of model performance shows that general purpose fi-


nite element codes can be used effectively to analyze underground pipe
but care must be taken to represent the culvert installation realistically.
Numerical results from our model built using general purpose finite ele-
ment codes are consistent with predictions of Burns and Richards' closed
form solutions. Our results also agreed with those obtained from two
semi-empirical models. Spangler's Iowa formula and White and Layer's
ring compression theory.

MODEL REFINEMENT

Once we were confident that the numerical results from our model
agreed with semi-empirical formulations and closed form analytical models
over the range of applicability of these models, we refined our model
to more accurately represent soil properties and construction during cul-
vert installation. Two variations from the linear elastic analysis were
studied: simulation of incremental construction using linear elastic ma-
terial properties, and modeling of nonlinear soil behavior using Dun-

Construction
Increment
7

I? / 34
y
M-
Construct on
Increment
0

FIG. 4.—Construction Increments Used in Simulation of Incremental Construction


by MARC General Purpose Finite Element Code

257

J. Transp. Eng. 1986.112:250-263.


Downloaded from ascelibrary.org by IMPERIAL COLLEGE LONDON on 05/15/13. Copyright ASCE. For personal use only; all rights reserved.

Incremental Solution
(ENTER) .4
TP=TEMPERATURE

1° 3
-°-o-o-o-o-o-n- D '
i ,2

Crown Springline Invert

Linear Elastic Solution

I\3
en
00

J. Transp. Eng. 1986.112:250-263.


COMPUTE LINEAR COMPUTE DUNCAN
STRESS-STRAIN STRESS-STRAIN
CONSTITUTIVE CONSTITUTIVE
RELATIONSHIP RELATIONSHIP
Crown Springline Invert
Position on Pipe
4
Bending Strain
(RETURN)
a Thrust Strain

FIG. 5.—Flow Chart of User Defined SUBROUTINE HYPELLA for In- FIG. 6.—Strain Plots Comparing Incremental and Linear Elastic Fi-
corporation of Duncan's Soil Model in MARC nite Element Solutions
can's soil model in combination with the incremental solution technique.
The 27 user subroutines available in MARC (15) provide a wide lati-
tude of operations for solving nonstandard problems such as the culvert-
soil interaction. For example, a user can write his o w n subroutine to
input work hardening slopes and state variable, define anisotropic, plas-
Downloaded from ascelibrary.org by IMPERIAL COLLEGE LONDON on 05/15/13. Copyright ASCE. For personal use only; all rights reserved.

tic, viscoelastic, and hypoelastic material properties, a n d input ultimate


stress for cracking analysis. Employing user subroutines in analysis en-
tails writing subroutines which give values to the parameters n e e d e d for
nonstandard analyses.
Incremental construction simulates the installation procedure used for
culverts. As shown in Fig. 4, 6-12 in. layers of soil are placed in the
trench and compacted until the trench has been filled. To model this
procedure, we added the elements comprising a soil layer a n d the load
from the soil gravity load a layer at a time. To d o this we wrote two
subroutines, SUBROUTINE CREDE and SUBROUTINE HYPELA, and
used the TYING CHANGE option in MARC. SUBROUTINE CREDE gives
an initial state variable to each element in the mesh. SUBROUTINE
HYPELA defines elastic constants as functions of elastic strain a n d state
variables and gives constitutive relations of the elements. These two sub-
routines are used together to give a low stiffness to the elements that
have not yet been added to the system and regular soil stiffness to those
elements already in the system. The TYING CHANGE option forces the
node displacements to be the same during each load step.
To implement incremental construction we used SUBROUTINE CREDE
to compute the initial stresses for each successive soil layer a n d add soil
layers to the model. Using SUBROUTINE CREDE the lowest soil layer
is given a d u m m y variable TP = 1.0. Each successive layer is given TP
= TP- 0.1. Soil is added one layer at a time by incrementing TP by 0.1
at each load step and requiring elements to have TP greater than or
equal to 1.0 to be included in the solution.
Duncan's soil model was i m p l e m e n t e d using SUBROUTINE HYPELA
and SUBROUTINE CREDE. During each load increment, the behavior
of the soil is assumed to be linear and the relationship between stress
and strain is assumed to be governed b y generalized Hooke's Law of
elastic deformations. As shown in Fig. 5, SUBROUTINE HYPELA com-
putes values for Young's modulus, E, and the bulk modulus, B, for each
load increment using Duncan's soil model. The change in the value cal-
culated for E represents the nonlinear, stress dependent stress-axial strain
behavior of the soil. The change in the value calculated for B represents
the nonlinear volumetric strain behavior of the soil. The parameters on
which E and B depend are found from standard triaxial tests as ex-
plained by Duncan et al. (6).

MODEL EVALUATION

Strain plots comparing the incremental solution with the linear elastic
solution are shown in Fig. 6. The incremental case gives slightly higher
strains for the pipe than does the linear elastic case. The thrust strains
are markedly less compressive for the incremental case with the strains
becoming tensile near the crown. Deflections increase significantly w h e n
an incremental solution is used. Horizontal a n d vertical deflections for

259

J. Transp. Eng. 1986.112:250-263.


Duncan's Soil Model
-•*.
.4 -

/ \ / \
.2 -
Downloaded from ascelibrary.org by IMPERIAL COLLEGE LONDON on 05/15/13. Copyright ASCE. For personal use only; all rights reserved.

c _ 1 K/
1
o
t; o —

-.2 - l ^ °-°-D-p- D . D . D -a-D- a A1 / \


/ ' \
/ \ '' *
-•4 ,;a / \! \
i i ii
Crown Springline Invert

Linear Elastic Soil Model

Crown Springline Invert


A Position on Pipe
Bending Strain
D
Thrust Strain

FIG. 7.—Strain Plots Comparing Duncan's Soil Model and Linear Elastic Soil Model

the incremental case are 0.1262 in. (3.207 mm) and 0.1453 in. (3.691 mm),
respectively. For the linear elastic case, they are 0.0791 in. (2.010 mm)
and 0.1034 in. (2.627 mm). The Df curves are almost identical for both
cases.
Strain plots comparing the solution using the Duncan soil model with
the linear elastic solution are shown in Fig. 7. In evaluating the Duncan
soil model we held the trench width constant at 3 pipe diameters and
used Duncan soil parameters (6,19) found from standard triaxial tests.
Duncan's model produces an increase in bending strain with dips in the
curve at the springline, invert and crown. The dip at the springline is
typical when using Duncan's model. Dips at the invert and crown, how-
ever, are not. They are a result of tension existing in the soil elements
adjacent to the pipe in the area where this is occurring. The Duncan
model as it is currently implemented in MARC does not disallow tension
in the soil. Therefore, this condition can not be corrected without mod-
ifying the code. The Duncan model thrust strain in the pipe is less than
that computed using linear elastic soil properties. The Duncan model
260

J. Transp. Eng. 1986.112:250-263.


deflections, 0.2118 in. (5.381 mm) horizontal and 0.2267 in. (5.757 mm)
vertical, are higher than those computed using linear elastic theory, 0.1335
in. (3.391 mm) horizontal and 0.1515 in. (3.848 mm) vertical.
Downloaded from ascelibrary.org by IMPERIAL COLLEGE LONDON on 05/15/13. Copyright ASCE. For personal use only; all rights reserved.

CONCLUSIONS

We have demonstrated that culverts can be modeled using general


purpose finite element codes. Numerical results computed using our
model agree with those obtained from the semi-empirical formulations
of White and Layer and the Iowa formula over the region w h e r e these
formulations have been s h o w n to be valid. Our results are also consis-
tent with the closed form solutions to culvert models obtained by Burns
and Richards. We duplicated capabilities of culvert specific finite ele-
ment codes by incorporating in our model features of these codes such
as Duncan's soil model and incremental construction.
The real strength of general purpose codes lies not in their capability
to duplicate the culvert specific code model of the culvert-soil interaction
but in their capability for modeling p h e n o m e n a not described by the
culvert specific codes and in their use of graphics to dramatically reduce
model preparation time and aid in result evaluation.
For those w h o have access to general purpose finite element codes
installed on a local computer, the use of these codes for modeling the
culvert soil interaction is an economical and effective w a y to approach
this engineering problem.

ACKNOWLEDGMENTS

The writers acknowledge O w e n s Corning Fiberglas Corporation for


their support of this work.

}
APPENDIX I.—REFERENCES

1. ASTM, "Standard Recommended Practice for Underground Installation of


Flexible Thermoplastic Sewer Pipe," ASTM Part 34: Plastic Pipe and Building
Products, No. D2321, 1980, pp. 170-175.
2. Clough, R. W., and Woodard, R. J., "Analysis of Embankment Stresses and
Deformation," Journal of the Soil Mechanics and Foundation Engineering Division,
ASCE, Vol. 93, No. SM4, July, 1967, pp. 529-549.
3. Desai, C. S., Elementary Finite Element Method, Prentice-Hall, Inc., Engelwood
Cliffs, N.J., 1979.
4. Duncan, J. M., "Finite Element Analysis of Buried Flexible Metal Culvert
Structures," Laurits Bjerrum Memorial Volume, N. Jambu, F. Jorstad, and B.
Kfoernsli, Eds., Norwegian Geotechnical Institute, Oslo, Norway, 1976, pp.
213-222.
5. Duncan, J. ML, "Behavior and Design of Long Span Metal Culverts," ASCE
Journal of the Geotechnical Engineering Division, Vol. 105, Mar., 1979, pp. 399-
418.
6. Duncan, J. M., Byrne, P., Wong, K. S., and Mabry, P., "Strength, Stress-
Strain and Bulk Modulus Parameters for Finite Element Analysis of Stresses
and Movements in Soil Masses," Univ. of California, Department of Civil
Engineering, Report No. UCB/GT/80-01, Aug., 1980.
7. General Electric, "User Manual for MODEL CREATION, CROSS-SECTION
ANALYSIS, and OUTPUT DISPLAY," General Electric CAE International,
1982.

261

J. Transp. Eng. 1986.112:250-263.


8. General Electric, "User Manual for SUPERB 6.0," General Electric CAE In-
ternational, 1983.
9. General Electric, "User Manual for GEOMOD," General Electric CAE Inter-
national, 1983.
10. General Electric, "I-DEAS Technical Overview," General Electric CAE Inter-
Downloaded from ascelibrary.org by IMPERIAL COLLEGE LONDON on 05/15/13. Copyright ASCE. For personal use only; all rights reserved.

national, 1984.
11. Katona, M. G„ Smith, J. M., Odello, R. J., and Allgood, J. R., "CANDE: A
Modern Approach for the Structural Design and Analysis of Buried Cul-
verts," Civil Engineering Laboratoy, Naval Construction Battalion Center,
Oct., 1976.
12. Krizek, R. J., Parmalee, R. A., Day, J. N., and Enlagger, H. A., "Structural
Analysis and Design of Pipe Culverts," National Cooperative Highway Re-
search Program, Report 116, HRB, 1971.
13. Leonards, G. A., and Roy, M. B., "Predicting Performance of Pipe Culverts
Buried in Soil," Joint Highway Research Project No. C-36-62F, File No. 9-8-
6, Purdue Univ., 1976.
14. MARC Analysis Research Corporation, "MARC Purpose Finite Element Pro-
gram Manual," Part 1, Volumes A and B, MARC Analysis Research Cor-
poration, 1980.
15. MARC Analysis Research Corporation, "MARC General Purpose Finite Ele-
ment Program Manual," Part 2, Volumes C, D, and E, MARC Analysis Re-
search Corporation, 1980.
16. McVay, M. C , "Evaluation of Numerical Modeling of Buried Conduits," the-
sis presented to the Univ. of Massachusetts, at Amherst, Mass., in 1982, in
partial fulfillment of the requirements for the degree of Doctor of Philosophy.
17. Parmalee, R. A., and Corotic, R. B., "Analytical and Experimental Evaluation
of the Modulus of Soil Reaction," Transportation Research Record, No. 518,
1974, pp. 29-38. (
18. Requicha, A. A. G., and Voelcker, H. B., "Solid Modeling: Current Status
and Research Directions," IEEE Computer Graphics and Applications, Vol. 3,
No. 8, Oct., 1982, pp. 25-37.
19. Robinson, R. A., "General Purpose Finite Element Codes Applied to Culvert
Analysis," The Ohio State Univ., 1983.
20. Smith, I. M., Programming The Finite Element Method with Applications to Geo-
mechanics, John Wiley and Sons Ltd., New York, N.Y., 1982.
21. Spangler, M. G., "The Structural Design of Flexible Pipe Culverts," Public
Roads, Vol. 18, No. 12, 1938.
22. Univ. of California, "Organization and Description of Input Data for Soil
Structure Interaction Program (SSTIPN)," Univ. of California, Berkeley, Calif.,
Jan. 1979.
23. White, H. L., and Layer, J. P., "The Corrugated Metal Conduit as a
Compression Ring," HBR Proceedings, Vol. 39, 1960, pp. 389-397.

APPENDIX II.—NOTATION

The following symbols are used in this paper:

B = soil bulk modulus;


D = culvert diameter;
Df = normalized bending strain;
dy = active load on culvert;
E = Young's modulus;
E' = modulus of soil reaction;
G = distance from finite element mesh lower extent to culvert in-
vert;
H = height of haunch;
I = moment of inertia;

262

J. Transp. Eng. 1986.112:250-263.


R = culvert radius;
S = distance from soil surface to culvert crown;
TP = temperature dummy variable;
t = thickness of culvert;
W = distance from culvert center to finite element mesh vertical edge;
Downloaded from ascelibrary.org by IMPERIAL COLLEGE LONDON on 05/15/13. Copyright ASCE. For personal use only; all rights reserved.

€j, = culvert bending strain; and


cr3 = minimum principal stress.

263

J. Transp. Eng. 1986.112:250-263.

You might also like