Download as pdf or txt
Download as pdf or txt
You are on page 1of 18

STRUCTURAL EVALUATION OF B O X CULVERTS

By Shad M. Sargand, 1 Associate Member, ASCE, Glenn A. Hazen, 2


Member, ASCE, and John O. Hurd 3

ABSTRACT: Corrugated-metal box culverts are employed as replacements for short-


Downloaded from ascelibrary.org by University Of Ulster AT on 06/22/15. Copyright ASCE. For personal use only; all rights reserved.

span bridges. As a result of increasing use, there is a great demand for establishing
a rational design procedure for these culverts. The available design procedures are
based on experience or mathematical modeling. Only a limited number of attempts
have been made to verify these design techniques with full-scale field tests. Three
corrugated-metal box culverts were fully instrumented and field-tested under similar
backfill conditions and live loads. The first culvert was bulb-angle-rib-reinforced
corrugated aluminum; the second was rib-reinforced corrugated steel; and the third
was corrugated-rib-reinforced corrugated steel. The collected data were evaluated
using a finite element analysis, and the applicability of standard design parameteis
was determined. Maximum bending moments obtained from field tests were com-
pared to plastic moments calculated according to American Association of State
Highway and Transportation Officials (AASHTO) specifications. It was concluded
that the design for these types of structures can be improved if (1) Composite action
is assumed; (2) thrust forces are included in design; and (3) construction procedures
are controlled. The AASHTO specifications were conservative for rib-reinforced,
corrugated-metal culverts.

INTRODUCTION

Rib-reinforced, corrugated-metal box culverts are increasingly being em-


ployed as replacements for short-span bridges. This has resulted in demand
for a rational design procedure to supplant the currently used procedures
that are based on experience or simple mathematical modeling. These flex-
ible structures obtain a large part of their load-carrying capacity from the
surrounding soil. Hence, in developing a rational design procedure for this
type of structure, various factors should be considered, including the me-
chanical properties of the structure and soils, soil-structure interaction phe-
nomena, location and type of load, and construction procedure. Composite
response should also be considered such as when the rib and corrugated
plate act as one member. Otherwise, the rib and plate will respond differ-
ently, i.e., noncompositely.
Several aspects of culvert design have been investigated in recent years.
Duncan et al. (1986) performed a finite element analysis (FEA) to develop
design equations for crown and haunch moment capacities of box culverts.
Analyses were performed on corrugated-aluminum box culverts with dif-
ferent spans and rises subjected to varying cover depths and live loads. To
verify the validity of their FEA, Duncan et al. (1986) performed a full-scale
field test on a rib-stiffened box culvert; and gave four conclusions: (1) Actual
deflections due to live load are likely to be about one-fourth larger than
deflections predicted by FEA; (2) actual bending moments in the crown

'Russ Prof, of Civ. Engrg., Ctr. for Geotech. and Envir. R e s . , Ohio Univ., Athens,
O H 45701.
2
Prof. and Chrmn., Dept. of Civ. Engrg., Ohio Univ., Athens, O H .
3
Hydr. Res. Engr., Ohio Dept. of Transp., 25 S. Front St., Columbus, O H 43215.
Note. Discussion open until May 1, 1993. To extend the closing date one month,
a written request must be filed with the A S C E Manager of Journals. The manuscript
for this paper was submitted for review and possible publication on December 19,
1991. This paper is part of the Journal of Structural Engineering, Vol. 118, N o . 12,
December, 1992. © A S C E , ISSN 0733-9445/92/0012-3297/$1.00 + $.15 per page.
Paper No. 701.

3297

J. Struct. Eng. 1992.118:3297-3314.


sections are likely to be about the same as those predicted by FEA; (3) the
critical location of the live load is always at or near the center span of the
culvert; and (4) comparison of measured and predicted crown deflections
indicated that shear transfer between ribs and plates is less than adequate
to allow full composite action between the two. From these findings, a
simplified design method for determining the bending moments at the haunch
Downloaded from ascelibrary.org by University Of Ulster AT on 06/22/15. Copyright ASCE. For personal use only; all rights reserved.

and crown region for different spans over depths and vehicle loads was
proposed.
Hurd and Sargand (1988) measured the geometry of various sizes and
shapes of corrugated-aluminum and -steel box culverts in service for many
years and analyzed four of these culverts with the Culvert Analysis and
Design (CANDE) program. They concluded that the variation between
design and true geometry has a small but significant effect on deflection,
moment, and thrust.
Beal (1981) instrumented a corrugated aluminum culvert at 16 locations
spaced around the structure's circumference at midspan. Corrugated-alu-
minum plate with reinforcing ribs was also tested in the laboratory. From
the field and laboratory tests and analytical results, Beal (1981) concluded
that for deeply buried box culverts: (1) Backfill placement sequence resulted
in distortion of culvert shape; (2) the maximum compressive stress at the
crown exceeded the nominal yield point value of aluminum; (3) live load
stresses were small compared to dead load stresses; (4) computed design
estimates of thrusts were greater than measured values; and (5) computed
design estimates of moments were less than the measured values .Beal (1986)
further investigated the composite behavior of an aluminum box culvert.
Field data were accumulated to permit the determination of moment, thrust,
and deflection of the structure. Curvature was measured to determine plate
moments since thrust was assumed negligible.

DESCRIPTION OF CULVERTS AND INSTRUMENTATION

For this study, three corrugated-metal box culverts were fully instru-
mented and tested under similar backfill conditions and live loads. Culvert
A was made of bulb-angle-rib-reinforced aluminum; culvert B was made
of angle-rib-reinforced corrugated steel; and culvert C was made of cor-
rugated-rib-reinforced corrugated steel. Instrumented cross sections for
culverts A, B, and C are illustrated in Figs. 1(a)-1(c). Additional details
of construction and geometry of the culverts are given in Table 1. Table 2
presents details of corrugations for each culvert. For each culvert, strain
was measured at a minimum of six locations chosen across the midlength
cross section. At each location, two biaxial electric strain gauges were at-
tached to the inside of the corrugated plate, and two uniaxial gauges were
attached to the outside ribs. The locations of the strain gauges on all three
culverts appear in Table 3. At least five vibrating strain gauges were ce-
mented to the inside plate and four to the outside rib. For culvert C, the
corrugated plate ribs were assumed to act compositely because of the large
number of bolts per unit length connecting the ribs to the plate. For all
structures it was assumed that the stress field in the ribs was uniaxial, while
that of the corrugated structural plates was biaxial. The strain-gauge readings
indicated that this assumption was reasonable. Uniaxial gauges were in-
stalled upon corrugation away from the rib to avoid local effects at the bolt
holes. Bending moments and thrusts were determined using electric-strain-
gauge readings. The vibrating wire gauges were used to supplement electric-
strain-gauge measurements.
3298

J. Struct. Eng. 1992.118:3297-3314.


Downloaded from ascelibrary.org by University Of Ulster AT on 06/22/15. Copyright ASCE. For personal use only; all rights reserved.

£
Reinforcing Angle Rib
(Outside Only)

(Section) 1

Span 14'-10"
CO

CO

Sections

D Vibrating gage 1,2


Bulb Angle Rib + Biaxial gage o Uniaxial gages

J. Struct. Eng. 1992.118:3297-3314.


Uniaxial Gages Uniaxial gage » Biaxial gages
n Vibrating wire
gages

Biaxial Gages
Biaxial Strain Gages

FIG. 1(a). Instrumented Cross Section FIG. 1(b). Instrumented Cross Section FIG. 1(c). Instrumented Cross Section
of Culvert A of Culvert B of Culvert C
TABLE 1. Culvert Construction and Geometry
Backfill
Culvert Ribs Material Span Rise Length height3
(1) (2) (3) (4) (5) (6) (7)
A Bulb angle Corrugated alu- 14 ft 10 in. 4 ft 10 in. 42 ft 7 ft 5 in.
Downloaded from ascelibrary.org by University Of Ulster AT on 06/22/15. Copyright ASCE. For personal use only; all rights reserved.

minum
B Angle Corrugated steel 15 ft 9 in. 5 ft 40 ft (ap- 9 ft
proxi-
mate)
C Corrugated Corrugated steel 15 ft 9 in. 5 ft 40 ft (ap- 8 ft 4 in.
proxi-
mate)
Trorn top of footing to road surface.

TABLE 2. Details of Corrugations

Thickness Corrugation
Culvert Crown plate (in.) Side plate (in.) Pitch (in.) Depth (in.)
(D (2) (3) (4) (5)
A 0.2 0.175 9 2.5
B 0.168 0.168 6 2
C 0.1215 0.1215 6 2

Deflections of the culverts were monitored at 11 points around the inside


surface at midspan. Displacement was measured with a tape extensometer
with respect to two ground reference points. Possible movements of the
reference points were observed with standard surveying instruments. The
backfill (standard Ohio Department of Transportation [ODOT] type 603)
was placed in 6-12-in. lifts, and each lift was compacted to 95% of the
Standard Proctor maximum dry density using a hand-driven tamper. The
compaction of the soil was monitored with a Troxler nuclear-density gauge.
Then, the subbase was placed, leveled, and compacted. Finally, asphalt
pavement was applied.
The electric-strain-gauge and vibrating-wire-gauge readings were re-
corded at the end of each lift. Deflection readings were taken for every two
lifts, since it required more time to obtain these readings.
At the completion of construction, three static live loads of 16, 32, and
42 kips were applied by deflating the tires on the central axle of a gravel-
filled dump truck so that the rear axle was over the loading point. These
loads were measured using a scale.

FIELD RESULTS

Figs. 2, 3, and 4 illustrate the maximum crown deflection versus height


of backfill added for culverts A, B, and C, respectively. In all three culverts
the crown moved upwards during early stages of backfill placement. This
upward movement was due to the lateral inward pressure of the compacted
backfill. As the fill began to cover the crown, the crown began to move
downwards in all three culverts. The downward deflection further increased
with the height of the backfill. During the backfilling process culvert A
shifted laterally, but this lateral movement was not observed in culverts B
3300

J. Struct. Eng. 1992.118:3297-3314.


Downloaded from ascelibrary.org by University Of Ulster AT on 06/22/15. Copyright ASCE. For personal use only; all rights reserved.

TABLE 3. Locations of Strain-Gauge installation Sections


Spacing Between (in.)
Section lo- Footing
Number of cated top and section Sections 1 Sections 2 Sections 3 Sections 4 Sections 5 Sections 6 Section 6 Section 7
Culvert section crown 1 and 2 and 3 and 4 and 5 and 6 and 7 and footing and footing
(D (2) (3) (4) (5) (6) (7) (8) 0) (10) (11) (12)
A 6 5 13.0 33.3 48.0 48.0 49.0 36.0 Not available 12.5 Not available
B 7 4 8.0 42.2 39.6 38.5 37.4 39.6 42.7 Not available 8.6
C 7 4 20.9 18.0 29.0 36.5 39.0 26.0 22.0 Not available 17.5

J. Struct. Eng. 1992.118:3297-3314.


— Theoretical
Downloaded from ascelibrary.org by University Of Ulster AT on 06/22/15. Copyright ASCE. For personal use only; all rights reserved.

o Experimental

Backfill Height (inch)

FIG. 2. Vertical Crown Deflection versus Backfill Height for Culvert A

0 20 40 60 80 100 120
Backfill Height (inch)

FIG. 3. Vertical Crown Deflection versus Backfill Height for Culvert B

0 20 40 60 80 100 120
Backfill Height (inch)

FIG. 4. Vertical Crown Deflection versus Backfill Height for Culvert C

3302

J. Struct. Eng. 1992.118:3297-3314.


and C. The lateral shift in culvert A might have been the result of an
imbalance in backfilling procedure and greater flexibility of the aluminum
culvert. The general trends of geometric deformation due to backfill of
culverts B and C are similar. Both culverts B and C behaved as arches
subjected to compressive loading. Culvert A, fabricated from aluminum,
had greater flexibility and responded as a combination of an arch and a
Downloaded from ascelibrary.org by University Of Ulster AT on 06/22/15. Copyright ASCE. For personal use only; all rights reserved.

beam.
Large culvert deflections occurred during paving operations. During this
stage of construction, heavy loads were imposed by asphalt delivery trucks
and compacting rollers. Culvert B had the least crown deflection of the
three culverts. Based on the cross-sectional properties, culvert B had the
greatest stiffness. In the case of culvert B, large gaps between ribs and plate
were observed at the completion of assembly. Therefore, the ribs of culvert
B moved laterally with respect to the plate during backfilling. As a result
of poor fabrication, the larger share of the load was carried by the plate
alone and the stiffeners did not contribute significantly.
The maximum crown deflections in culverts A, B, and C under live loads
are plotted in Figs. 5, 6, and 7, and they are tabulated in Table 4. The
deflections caused by live load alone were considerably lower than those
caused by the construction sequence. For the 42-kip loading, culvert B
deflected the least, indicating that the ribs became effective load-carrying
members, and the ribs and the plate acted as a composite system under
large live loads. Culvert A deflected the most under all loads. Since this
culvert was installed with the least cover, large deflections were probably
the result of a combination of the higher structural flexibility of the alu-
minum material and poorer distribution of the live load. Even though all
three culverts deflected differently, all deflections were within the allowable
design limit of 5%. Based on the configurations of culverts A and B, it was
decided that they should be analyzed as both composite and noncomposite
structures (Beal 1981; Sargand and Hazen 1990), whereas culvert C was
assumed to be composite.
Noncomposite moments and thrusts were calculated from four strain gauges
located across a section. A linear response was assumed across the rib and
across the plate. Thus, the moment was calculated as the sum of moment
in the rib, moment in the plate, and moment resulting from the couple
created by the rib thrust and the plate thrust.
For composite moments and thrusts, the culverts were modeled as Ti-
moshenko beams by assuming that there was no slip between the rib and
plate. Only two strain readings were necessary to calculate the internal
actions since the strain distribution was linear and sectional properties were
calculated for the composite section. As uniaxial gauges were installed on
culvert C, moments and thrusts in the corrugated plate were calculated by
ignoring Poisson's ratio effect on stresses.
The bending moment versus unfolded length, after backfilling to the
haunch, is plotted for each culvert in Figs. 8-10. In agreement with previous
studies, moments were not significant when the backfill was at the haunch
level. In all three culverts, moments were not symmetrical during the con-
struction phase. This may have been due to a shift in the culverts resulting
from a nonsymmetrical placement of backfill. Moments measured in the
sides and crowns of culverts A, B, and C were of approximately the same
magnitude. Maximum moment was observed during asphalt paving.
Although the live load was applied at the center of the culverts, maximum
moment shifted away from the center. Culverts A and B developed moments
3303

J. Struct. Eng. 1992.118:3297-3314.


n Experimental
Downloaded from ascelibrary.org by University Of Ulster AT on 06/22/15. Copyright ASCE. For personal use only; all rights reserved.

—- Theoretical

10 20 30 40 50
Live Load (kip)

FIG. 5. Vertical Crown Deflection during Live-Load Tests for Culvert A

10 20 30 40
Live Load (kip)

FIG. 6. Vertical Crown Deflection during Live-Load Tests for Culvert B

a Experimental
— Theoretical

10 20 30 40 50
Live Load (kip)
FIG. 7. Vertical Crown Deflection during Live-Load Tests for Culvert C

TABLE 4. Maximum Crown Deflections due to Live Load


Maximum Deflection (in.)
Culvert 16 kips 32 kips 42 kips
(1) (2) (3) (4)
A 0.140 0.250 0.370
B 0.035 0.070 0.083
C 0.040 0.080 0.125

3304

J. Struct. Eng. 1992.118:3297-3314.


P Experimental
Downloaded from ascelibrary.org by University Of Ulster AT on 06/22/15. Copyright ASCE. For personal use only; all rights reserved.

_— Theoretical
I

Unfolded Length (inch)

FIG. 8. Bending Moment due to Backfill Up to Haunches as Function of Unfolded


Length for Culvert A

D Experimental
Theoretical

0 20 40 60 80 100 120 140

Unfolded Length (inch)

FIG. 9. Bending Moment due to Backfill Up to Haunches as Function of Unfolded


Length for Culvert B

D Experimental
Theoretical

150 200

Unfolded Length (inch)

FIG. 10. Bending Moment due to Backfill Up to Haunches as Function of Unfolded


Length for Culvert C
3305

J. Struct. Eng. 1992.118:3297-3314.


that were of similar magnitude. Live-load bending moments for each culvert
are shown in Figs. 11-13. For the live load of 42 kips, the magnitudes of
measured moments increased slightly when compared to maximum moments
measured at the completion of backfill for all three culverts. In culvert B,
while the initial moment due to 16-kip load was large, the additional changes
were small.
Downloaded from ascelibrary.org by University Of Ulster AT on 06/22/15. Copyright ASCE. For personal use only; all rights reserved.

Figs. 14-16 illustrate the thrusts in the three culverts under live load.
During the backfilling, culverts A, B, and C developed axial thrust. The
maximum thrust of the top of culvert B is higher than the other three culverts
because of its higher stiffness. The other two culverts had approximately
the same thrust. In all three culverts thrusts were also found for the three
live loads. The culverts gave values of thrust that varied from 3 to 12 kips/
ft under a live load of 42 kips.
The response of an instrumented section was noncomposite or composite
depending on a number of variables. These included the location of the
section around the circumference, the loading condition, the location of
bolts, and the degree of bolt slip. A discrepancy between expected thrust
and measured thrust was found. Since the couple due to thrust predominated
in determining the sectional moment, curvature cannot be related to moment
as was done in previous investigations. In the beginning stages of backfilling,
little agreement between composite and noncomposite action was found for
any of the culverts. The lack of composite response was associated with
poor fit during fabrication. At the final stages of backfilling and during live-
load application, the response was dependent on the location of the section.
Moments measured for the crown and haunches of culvert A are given in
Figs. 17 and 18. These results are typical for all the three culverts. The
difference between noncomposite and composite response at the crown is
evident, whereas the agreement at the haunches is very good. Thus, it is
apparent that shearing load causes relative movement that results in com-
posite action since shear is much more significant at the haunches.
Bending moments were also determined after subjecting the culvert to
live loads of 16, 32, and 42 kips (50%, 100%, and 130% of HS-20 traffic
load). Composite versus noncomposite moments for culvert A are plotted
in Fig. 19 for the 16-kip load. Responses matched well for all live loads.
The plate and rib acted more compositely during live load than during
backfill for all culverts. A corresponding behavior was observed for thrust,
although the agreement between composite and noncomposite thrusts was
not as good.

FINITE ELEMENT ANALYSIS

The CANDE computer program (Katona et al. 1976, 1981) was used in
this study to determine the response of culverts during backfilling process
and live load applications. These results were compared with field values.
The CANDE finite element program was developed specifically for mod-
eling and designing culverts (Katona et al. 1976). It has been widely used
by design and research engineers for modeling culverts of various shapes
and under various installation conditions. The program can perform elastic
analysis of a round pipe, use its own mesh, or work with a user-supplied
mesh. The last option was used in this investigation, by carefully constructing
meshes to adequately simulate the field tests. CANDE allows for simulating
any of several culvert materials, including reinforced concrete, plastic, cor-
rugated aluminum, corrugated steel, or a user-defined material. The cor-
rugated-aluminum option was selected for culvert A, while the other two
3306

J. Struct. Eng. 1992.118:3297-3314.


Downloaded from ascelibrary.org by University Of Ulster AT on 06/22/15. Copyright ASCE. For personal use only; all rights reserved.

T
50 100
Unfolded Length (inch)

FIG. 11. Bending Moment due to Live Loads as Function of Unfolded Length for
Culvert A

10.0- 42 kips
\ 32 kips

5.0-
^/jf 16 kips (
^^^ 16 kips ^A
e "^^s^ J S ^Z^ Experimental
"5,° 1
S -5.0- ^^^=^^0^ n 161d s
P
32 kips \ O 32 kips
-10.0 - 42 kips
M 42 kips
-15.0- 1 1 1 1 1 1 1 1 1 1 1 1
0 10 20 30 40 50 60 70 80 90 100 110 120 130

Unfolded Length (inch)

FIG. 12. Bending Moment due to Live Loads as Function of Unfolded Length for
Culvert B

^ - 4 2 kips
6- / ^ 32 kips

^^•16 kips
4- ^ D

9 / / %
2-
Experimental
t o 16 kips
16 kips n 32 kips
• 42 kips
-2-
- 32 kips
42 kips
-4-
0.0 50 100 150 200

Unfolded Length (inch)

FIG. 13. Bending Moment due to Live Loads as Function of Unfolded Length for
Culvert C

3307

J. Struct. Eng. 1992.118:3297-3314.


D
O
Theoretical
a
0
16 kips

Thrast (kip/ft)
32 kips
Downloaded from ascelibrary.org by University Of Ulster AT on 06/22/15. Copyright ASCE. For personal use only; all rights reserved.

3 42 kips
-10-

Experimental
-20- a • 16 kips
o 32 kips
H
B 42 kips

1
1 I 1 i • 1
0 20 40 60 80 100 120
Unfolded Length (inch)

FIG. 14. Thrust versus Unfolded Length under Live Loads for Culvert A

i
i

40 60
Unfolded Length (inch)
FIG. 15. Thrust versus Unfolded Length under Live Loads for Culvert B

80 100
Unfolded Length (inch)

FIG. 16. Thrust versus Unfolded Length under Live Loads for Culvert C

3308

J. Struct. Eng. 1992.118:3297-3314.


Downloaded from ascelibrary.org by University Of Ulster AT on 06/22/15. Copyright ASCE. For personal use only; all rights reserved.

0 10 20 30 40 50 60 70 80 90

Backfill Height (inch)

FIG. 17. Moment Comparisons at Crown during Backfill for Culvert A

0 10 20 30 40 50 60 70 80 90

Backfill Height (inch)

FIG. 18. Moment Comparisons at Haunch during Backfill for Culvert A

0.2-
f^
is?
a.
0.1 "
/I h o
a
Composite
Nonconiposite
o.o - . 1 \ ll
VV V
Moment

-0.1 -

-0.2- v
1 1 1 1
50 100 150 200 250 300

Unfolded Length (in)

FIG. 19. Moment Comparison versus Unfolded Length under 16-kip Live Load for
Culvert A

were modeled as a user-defined material to avoid some problems in the


sawtooth approximation used in the steel option (Katona et al. 1976; Rauch
1990).
The original soil models available in CANDE were modified to include
the Duncan hyperbolic model (Katona et al. 1981), which has proven to be
3309

J. Struct. Eng. 1992.118:3297-3314.


a fairly popular choice and has been used in several studies. The in situ soil
was modeled as a clay under 90% compaction, using standard parameters
from the CANDE parameter library. The backfill sand was taken to the
laboratory and subjected to the conventional triaxial and the multiaxial tests.
In a multiaxial device, a cubical specimen is loaded on six orthogonal
faces by six flexible pressurized membranes. Any of the principal stresses
Downloaded from ascelibrary.org by University Of Ulster AT on 06/22/15. Copyright ASCE. For personal use only; all rights reserved.

can be varied independently, allowing for the simulation of complex stress


histories. The three applied stresses in the direction of opposite surfaces
are considered to be principal stresses. This loading will ensure that a prop-
erly prepared specimen will float between unconstrained deformations. A
constant stress state is maintained in the horizontal planes by applying air
or hydraulic oil pressure to two axes to simulate a backfill loading. The
pressure is then applied to the remaining axis. Deformations are monitored
along with axis of loading with six linear variable differential transformers
(LVDTs). There are three LVDTs located on each face.
A multiaxial apparatus allows several principal stress paths, including
hydrostatic compression (HC) and conventional triaxial compression (CTC),
to be conducted. Bulk modulus parameters were obtained from the HC
tests, while the other parameters were obtained from the CTC tests and
from an entire series of stress-path tests, with these latter parameters known
as the general path parameters. Both sets of parameters were tested in
simulations, as well as a set from the CANDE library, 100% compaction
silty sand, and a combination of the two sets of laboratory parameters using
separate regions of the mesh, as indicated in Fig. 20. The computed results
from the trials did not vary significantly, indicating that CANDE is not
overly sensitive to the parameters used in the Duncan hyperbolic model
(Duncan 1980). The combination of CTC and general path parameters in
Table 5, as distributed according to Fig. 20, was selected as the most logical
method of modeling the culvert backfill.
The crushed limestone aggregate was modeled as coarse-aggregate at

Asphalt
Limestone subbase
/ o
Backfill sand, CTC path parameters

/ !
/ s
~~7~~ ^~~i
Corrugated steel plate 1 / 8*
1 Backfill sand,
I general path
$
\ parameters
\ a
Covcrete

In situ clay, 90% compaction

FIG. 20. Material Zones Used in CANDE Half-Mesh


3310

J. Struct. Eng. 1992.118:3297-3314.


TABLE S. Parameters Used for Modeling Culvert Backfill Sand with Duncan's
Hyperbolic Soil Model (Sargand and Hazen 1990)
Soil model parameter CTC path parameters General path parameters
(1) (2) (3)
4>o 39° 39°
Downloaded from ascelibrary.org by University Of Ulster AT on 06/22/15. Copyright ASCE. For personal use only; all rights reserved.

A<)> 5.5° 3.0°


C 0.0 0.0
K 450 1200
n 0.35 1.1
Rf 0.60 0.70
Kb 300 350
m 0.30 0.25

105% compaction and using the Duncan hyperbolic model parameters in


the CANDE library. The fresh asphalt was modeled in the construction
sequence as an elastic solid with an elastic modulus of 7 ksi and a Poisson's
ratio of 0.2. For live-load simulation, the asphalt was assumed cured, with
a Poisson's ratio of about 0.4 and an elastic modulus between 300 and 500
ksi, with the exact value to match the recorded temperature at the test
determined from a table (Kelly 1986). Parameters for concrete footing were
obtained from the literature.
All CANDE solutions are based on the assumptions of plane strain con-
ditions, small displacements, and quasistatic responses. Hence, longitudinal
variations or time-dependent effects cannot be modeled directly. Since it is
impossible to simulate a load spread over a small area, as under a truck
tire, it is necessary to devise an equivalent linear loading. Katona et al.
(1976), and later Duncan (1979), devised methods for finding the appro-
priate loading. Because of the deficiencies in these models (Rauch 1990),
the equivalent linear loading was determined using the pavement simulation
program ILLISLAB. The procedure was to try different loadings along a
l-in.-wide strip over the length of the paved surface over the culvert until
one was obtained that had a computed deflection that matched the one
computed by the program using actual truck loading.
A 241-node, 226-element mesh modeling half of the cross section of the
culvert, called a half mesh, was used in most computations. For this reason,
nonsymmetrical construction and live loadings were not analyzed. The mesh
was divided into several zones representing the different materials involved
in the culvert construction, as shown in Fig. 20. Interface elements were
not used in order to avoid the convergence problems associated with them
(Rauch 1990). Also, since small variations in the geometry of a box culvert
can have significant effects on predicted values of moments and thrusts
(Hurd and Sargand 1988), the true initial shape of each culvert was measured
in the field before the placement of backfill material.
Each culvert was modeled with a construction sequence, consisting of in
situ loads followed by a series of backfill loads, with an additional 5-psi
surface pressure to simulate equipment loading, and paving loads. To sim-
ulate live loading, half of the equivalent line load was applied to the half
mesh, and the total response minus the construction response was compared
to the incremental response measured in the field.
In culvert A, the final deflection was calculated to be only one-half as
large as measured, as shown in the theoretical curve in Fig. 2. Crown
deflection for culvert B was calculated to be much less than the field mea-
3311

J. Struct. Eng. 1992.118:3297-3314.


surement; the FEA computation is shown as the theoretical curve in Fig.
3. The discrepancy may have been the result of loose connections between
plate and rib. Culvert C, for which the FEA results are shown in Fig. 4,
was stiffened with corrugated plate and so did not exhibit the flexibility of
culverts A and B. No peaking was indicated by CANDE. Maximum de-
flection was predicted to be the smallest. Overall, the CANDE finite element
Downloaded from ascelibrary.org by University Of Ulster AT on 06/22/15. Copyright ASCE. For personal use only; all rights reserved.

program predicted the deflection of the box-type culverts tested with rea-
sonable accuracy when only load increases, but the accuracy of simulation
of loading and unloading conditions is questionable.
Overall, the deflections correlated better when simulating live loading
than when simulating backfilling and particularly when simulating paving.
Deflections in all the cases were found to be reasonable. Culvert A, as
expected, and as shown in Fig. 5, gave the maximum deflection because of
its high flexibility and less cover. Deflections due to live-load application
were very small compared with deflections due to backfill. Theoretical de-
flections were much larger than measured values for culverts B and C, as
shown in Figs. 6 and 7. Theoretically, the displacement model will predict
deflections less than actual values. Soil materials all stiffened under com-
paction. The modeling of the two-dimensional roadway simulation of load
needs further refinement.
The culverts experienced permanent deformation during construction and
use, causing variation in the response of each culvert during the application
of live load. Consequently, there is a negative impact on the load capacity
of these soil structures. The FEA provided a symmetric result, so a shifting
of the flexible culvert during backfill, which occurred in the field, was not
duplicated due to the use of a half mesh. Maximum moment and thrusts
were recorded at different places than where the finite element solution
predicted.
Finite element solutions for bending moment compared very favorably
to the experimental results during later stages of fill and when subjected to
live load, as shown in the theoretical curves of Figs. 8-10 and 11-13,
respectively. Because each culvert responded noncompositely in the early
stages, it is difficult to make an accurate comparison between experimental
and composite responses. Due to insufficient shear transfer to force a com-
posite response, the moment in culvert B does not compare well.
Measured thrusts, when compared to calculated values, were very incon-
sistent, as shown in Figs. 14-16. This inconsistency results from the tendency
of ribs to be primarily compressive members and the plate to be in tension
while resisting moment with a couple action. Thus moment is only piecewise
continuous between bolts. In addition, the soil frictional forces acted to
resist thrusts. Duncan et al. (1986) did not consider thrust to be important

TABLE 6. Design Moments


Experimental Plastic Moment Plastic Moment
Moment (AASHTO) (Actual)
(kip-ft/ft) (kip-ft/ft) (kip-ft/ft)
Culverts Crown Haunch Crown Haunch Crown Haunch
0) (2) (3) (4) (5) (6) (7)
A 0.7 -1.86 16.85 6.39 17.8 12.82
B 2.22 -4.12 11.15 11.60 35.86 39.87
C 3.75 -2.88 11.59 11.34 36.3 14.52

3312

J. Struct. Eng. 1992.118:3297-3314.


1

because of the shape of the box culvert, the difficulty in making measure-
ments, and the method used to calculate thrusts. In this study, thrusts were
an important load-carrying mechanism for all three culverts. The American
Association of State Highway and Transportation Officials (AASHTO) re-
quired minimum plastic moments for the crown and haunch of each culvert
are tabulated in Table 6. These were computed from AASHTO guidelines
Downloaded from ascelibrary.org by University Of Ulster AT on 06/22/15. Copyright ASCE. For personal use only; all rights reserved.

using the average value of the crown-to-haunch moment proportioning val-


ues. The actual plastic moments are also tabulated. These computations
were based on the actual geometry of the culverts. The actual plastic mo-
ments were very conservative compared to the AASHTO specifications.

CONCLUSIONS

In this study, three box culverts with approximately the same rise and
span were studied. Culvert A was bulb-angle-rib-reinforced corrugated
aluminum; culvert B was angle-rib-reinforced corrugated steel; and culvert
C was corrugated-rib-reinforced corrugated steel. The culverts were in-
stalled under similar site conditions. Design comparisons and relative per-
formance measurements were made. Additional comparisons were made
on three different rib and corrugated-rib constructions. Conclusions are
drawn about the effects of backfill materials, construction sequences, and
permanent deformation of culverts. The following conclusions result from
field measurements, analytical analyses, and correlation of this study with
published results of previous culvert investigations.

• Culverts A and B behaved noncompositely at low levels of backfill.


As backfilling progressed as well as during live-load application, the
response of each culvert became composite. Culvert C could always
be modeled as a composite structure.
• Under the well-controlled construction procedures used in this study,
the deflections in all three culverts were reasonable, and the mo-
ments were within design limits. However, the thrusts must be con-
sidered in the design procedure. This finding contradicts previous
investigations.
• This investigation and previous work indicate that the permanent
distortion of the culvert during installation and service should be
incorporated in design.
• Culvert B was well overdesigned on the basis of deflection and
moment. All three culverts satisfied AASHTO design criteria.
• A large portion of the permanent deformation took place during
paving operations.
• CANDE predicted moments and deflections during backfill and live
loads with reasonable accuracy. During paving, the culverts expe-
rienced loading and unloading. CANDE does not include a material
model that can simulate the loading and unloading that took place
during pavement placement operations, and it specifically cannot
efficiently model asymmetric loading conditions.
3313

J. Struct. Eng. 1992.118:3297-3314.


APPENDIX I. CONVERSION TO SI UNITS
To convert To Multiply by
ft m 0.3048
in. mm 25.4
kip kN 4.448
Downloaded from ascelibrary.org by University Of Ulster AT on 06/22/15. Copyright ASCE. For personal use only; all rights reserved.

ksi MPa 6.895


lb-ft N-m 1.356
lb/ft N/m 14.593

APPENDIX II. REFERENCES


Beal, D. B. (1981). "Behavior of an aluminum structural plate culvert." Report RR-
81-90, Federal Highway Admin., U.S. Dept. of Transp., Washington, D.C.
Beal, D. B. (1986). "Behavior of a corrugated metal box culvert." Report Number
FHWA-NY-RR-86-133, Federal Highway Adm., U.S. Dept. of Transp., Wash-
ington, D.C.
Duncan, J. M. (1979). "Behavior and design of long-span metal culverts."/. Geotech.
Engrg. Div., ASCE, 105(3), 399-418.
Duncan, J. M. (1980). "Hyperbolic stress-strain relationships." Proc. Workshop on
Limit Equilibrium, Plasticity and Generalized Stress-Strain in Geotech. Engrg.,
ASCE, 443-460.
Duncan, J. M., Seed, R. B., and Drawsky, R. H. (1986). "Design of corrugated
metal box culverts." Transp. Res. Record, 1008, 33-41.
Hurd, J. O., and Sargand, S. M. (1988). "Field performance of corrugated metal
box culverts." Transp. Res. Record, 1191, 39-45.
Katona, M. G., Smith, J. M., Odello, R. S., and Allgood, J. R. (1976). "CANDE—
A modern approach for the structural design and analysis of buried culverts."
Report FHWA-RD-77-5, Federal Highway Admin., U.S. Dept. of Transp., Wash-
ington, D.C.
Katona, M. G., Vittes, P. D., Lee, C. H., and Ho, H. T. (1981). "CANDE-1980:
Box culverts and soil models." Report FHWA-RD-80-172, Federal Highway Ad-
min., U.S. Dept. of Transp., Washington, D.C.
Kelly, H. F., IV. (1986). "Development of mechanistic flexible pavement design
concepts for the heavyweight F-15 aircraft," PhD dissertation, University of Illi-
nois, Urbana, 111.
Rauch, A. (1990). "Experimental and numerical investigation of a deep-corrugated
steel, box-type culvert," MS thesis, Ohio University, Athens, Ohio.
Sargand, S. M., and Hazen, G. A. (1990). "Structural analysis of corrugated metal
box-type culverts." Report FHWA/OH-901002, Ohio Dept. of Transp., Columbus,
Ohio.

3314

J. Struct. Eng. 1992.118:3297-3314.

You might also like