You are on page 1of 7

Vertical Loads on Concrete Box Culverts

under High Embankments


Richard M. Bennett, M.ASCE1; Scott M. Wood2; Eric C. Drumm, M.ASCE3; and
N. Randy Rainwater, M.ASCE4

Abstract: Vertical loads on concrete box culverts under high embankments are examined, where a high embankment is defined herein as
Downloaded from ascelibrary.org by New York University on 05/12/15. Copyright ASCE. For personal use only; all rights reserved.

one where the height of the fill above the culvert is greater than the width of the culvert. Results from an instrumented culvert are
described, with results from pressure cells, strain gages in the wall, and strain gages in the roof showing reasonable agreement. There was
strong correlation between the height of fill and the pressure and internal forces in the culvert, suggesting that the soil–structure interaction
factor is independent of the H / B ratio. The results of the instrumented culvert were compared to those of other instrumented culverts
under high embankments. The measured roof pressures were significantly greater than the pressure due to the soil overburden, with the
measured pressures averaging 1.5 times the soil overburden pressure. Although the soil–structure interaction factor recommended by the
American Society of Highway and Transportation Officials specification tends to under-predict the loads acting on concrete culverts,
a simplified reliability analysis suggests that the Specification provides a sufficient level of safety.
DOI: 10.1061/共ASCE兲1084-0702共2005兲10:6共643兲
CE Database subject headings: Bridges, viaduct; Culverts; Soil–structure interaction; Vertical loads; Embankments.

Introduction narrow or shallow ditch such that its top is below the natural
ground surface and then is covered with an embankment. Culverts
Cast-in-place concrete box culverts are often used as conduits to can further be classified as rigid or flexible, based on their stiff-
carry water from one side of a highway to the other. Although this ness. Concrete box culverts are typically considered to be rigid
is a simple role, the loadings applied to these structures are rather culverts, and are generally positive projecting culverts. Those are
complex. These structures must resist large vertical and lateral the conditions that are considered herein.
earth pressures, and are often subjected to significant loadings This paper examines the vertical loads from the overburden
during construction of the embankment. Due to soil–structure in- soil for culverts under high embankments. A high embankment is
teraction effects, the state of stress on the culvert depends on the defined in this paper as one where the height of soil fill above the
stiffness of both the structure and the backfill material. Despite top of the culvert H is greater than or equal to the culvert width B.
the complexity of the loadings, a simple approach must be used Results from an instrumented culvert are presented and results
for analysis and design due to the large number of culverts that from other instrumented culverts are summarized. General con-
are being built. clusions are drawn about the vertical pressure on cast-in-place
Culverts can be divided into positive and negative projecting reinforced concrete box culverts under high embankments.
culverts, with a positive projecting culvert having its top project-
ing above the natural ground surface and covered with an em-
bankment. A negative projecting culvert is installed in a relatively Previously Instrumented Culverts
1
Professor, Dept. of Civil and Environmental Engineering, The Univ. The first research into loads on culverts is usually credited to
of Tennessee, Knoxville, TN 37996-2010 共corresponding author兲. E-mail: Marston and Spangler at Iowa State Univ. In 1919, Marston mea-
rmbennett@utk.edu
2 sured the load on a 1.02 m culvert under 6.10 m of fill and found
Consulting Engineer, Engineering Design and Testing Corporation,
Orlando, FL 32714; formerly, Graduate Student, Dept. of Civil and that the load on the culvert was nearly 1.92 times the weight of
Environmental Engineering, The Univ. of Tennessee, Knoxville, TN the soil column directly above the culvert 共Spangler 1968兲. This
37996-2010 was the start of a long research program into both rigid and flex-
3
Professor, Dept. of Civil and Environmental Engineering, The Univ. ible culverts at Iowa State. With respect to rigid culverts, Span-
of Tennessee, Knoxville, TN 37996-2010; E-mail: edrumm@utk.edu gler 共1947兲 gives the results of two circular test culverts that were
4
Faculty Associate, Dept. of Civil and Environmental Engineering, supported on weighing devices. One had “top soil” 共unit weight
The Univ. of Tennessee, Knoxville, TN 37996-2010. E-mail: of 15 kN/ m3兲 fill and an H / B ratio of 6.0. The measured pressure
nrainwat@utk.edu was approximately 1.9 times the overburden pressure. The other
Note. Discussion open until April 1, 2006. Separate discussions must
culvert had sand and gravel fill 共unit weight of 20 kN/ m3兲 and an
be submitted for individual papers. To extend the closing date by one
month, a written request must be filed with the ASCE Managing Editor.
H / B ratio of 4.6. The measured pressure was approximately 1.5
The manuscript for this paper was submitted for review and possible times the overburden pressure. Spangler 共1950兲 reported the load
publication on July 1, 2004; approved on September 23, 2004. This paper measurements on seven positive projecting box culverts ranging
is part of the Journal of Bridge Engineering, Vol. 10, No. 6, November from a nominal size of 0.61 m ⫻ 0.61 m to 2.44 m ⫻ 2.44 m, with
1, 2005. ©ASCE, ISSN 1084-0702/2005/6-643–649/$25.00. most being 1.22 m ⫻ 1.22 m. The measurements were made using

JOURNAL OF BRIDGE ENGINEERING © ASCE / NOVEMBER/DECEMBER 2005 / 643

J. Bridge Eng. 2005.10:643-649.


friction ribbons and represent an average pressure on the culvert. overburden pressure. The culvert was in a section that was only
Ratios of the measured pressure to the overburden pressure were about half the maximum height of the dam. The writers attribute
共ratios of H / B given in parentheses兲 1.5 共2.6兲, 1.6 共2.8兲, 1.3 共1.8兲, some of the pressure to “transverse arching” or arching in the
1.2 共2.4兲, 1.4 共1.9兲, 1.5 共5.5兲, and 1.5 共4.4兲. direction of the dam or transverse to the culvert. They suggest the
By looking at soil deflections compared to culvert deflections, culvert is acting like an abutment to the arching of the soil over
a method was developed for predicting the load on a culvert, the higher portion of the dam. Because of these conditions, the
which was presented in the form of a chart in Spangler 共1947兲. results will not be considered any further. Trollope et al. 共1963兲
The design pressure was a function of the projection condition, suggest culverts be designed for twice the overburden pressure
various soil properties, and the H / B ratio. Based on typical values with a factor of safety of 3.
for design and formulas developed for the chart by Clarke Höeg 共1968兲 studied the behavior of model steel cylinders
共1967兲, the value of the soil–structure interaction factor, Fe, is in a test box filled with Ottawa sand. There were one to two
determined as cylinder diameters of sand above the cylinder, and then pressure
was applied to the top of the sand with an air bag. Pressures
e0.38共H/B兲 − 1
Downloaded from ascelibrary.org by New York University on 05/12/15. Copyright ASCE. For personal use only; all rights reserved.

correspond to H / B ratios in excess of 400. For a rigid cylinder,


Fe = H/B 艋 2.42 the crown pressure was 1.42 times the applied pressure. For a
0.38共H/B兲
共1兲 cylinder with a diameter/thickness ratio of 40 the crown pressure
0.12 was 0.85 times the applied pressure, and for a cylinder with a
Fe = 1.69 − H/B ⬎ 2.42 diameter/thickness ratio of 80 the crown pressure was 0.69 times
H/B
the applied pressure. These results clearly show the effect of the
in which the vertical pressure on the culvert is determined as the stiffness of the culvert, and that stiff culverts will carry more
pressure due to the weight of the soil column above the culvert pressure than that due to the soil overburden alone.
times the soil–structure interaction factor Fe. The factor Fe Girdler 共1974兲 reported on a 1.73 m wide single cell box
is often referred to as the soil modification factor, because it is culvert under 23.5 m of fill 共H / B = 13.6兲. Roof pressure cells were
numerically equal to the factor by which the soil unit weight can attached 0.25 m 共0.14 B兲 in from the edge of the culvert. One
be multiplied to obtain the equivalent soil unit weight used to pressure cell failed, while the other read 1.48 times the overbur-
calculate the vertical pressure acting on the culvert. den pressure. A 2.03 m wide single cell box culvert under 11.4 m
In Eq. 共1兲, the H / B ratio of 2.42 corresponds to the plane of of fill 共H / B = 5.6兲 had roof pressure cells attached at 0.46 m
equal settlement being at ground surface. The plane of equal 共0.23 B兲 in from the edge. The pressure was very asymmetric,
settlement is defined by Spangler 共1947兲 as the horizontal plane in with one side having 0.90 times the overburden pressure and the
the embankment at which the settlements of the interior soil prism other 1.74 times the overburden pressure, for an average of 1.32
above the culvert and the exterior soil prism outside the culvert times the overburden pressure.
are equal. For embankment heights greater than the plane of equal Penman et al. 共1975兲 report pressure measurements made on a
settlement, the soil–structure interaction factor is essentially con- culvert passing under a 53 m high rockfill dam. The inside of the
stant. In Eq. 共1兲, Fe varies from 1.64 at H / B = 2.42 to 1.69 at culvert was elliptical in shape with the outside following the in-
H / B = ⬁. This suggests a theoretical basis for the soil–structure side shape but being a series of straight surfaces. An earth pres-
interaction factor being a constant for very high embankments. sure cell was placed directly over the crown of the culvert. For fill
Woodbury et al. 共1926兲 report on tests performed by the heights up to an H / B ratio of approximately 9, the measured
American Railway Engineering Association on eight different pressure on the crown of the culvert was approximately 2.1 times
culverts in an embankment on the Illinois Central Railroad. Two the overburden pressure. At increasing fill heights above
of the culverts were rigid culverts, one being 1.07 m extra heavy H / B = 9, the ratio of measured pressure to overburden pressure
cast iron, and one being a 0.61 m ⫻ 0.68 m concrete culvert. The decreased. At the final fill height corresponding to an H / B ratio of
cast iron culvert had 10.4 m of fill 共H / B = 9.7兲 and the concrete 12.9, the crown pressure was 1.76 times the overburden pressure.
culvert had 9.8 m of fill 共H / B = 14.4兲. Pressure readings on both Dasgupta and Sengupta 共1991兲 studied a 1.35 m ⫻ 1.35 m
culverts indicated a crown pressure of 1.58 times the overburden model box culvert backfilled with dry sand to a height of 2.4 m
pressure. above the culvert 共H / B = 1.8兲. Pressures were measured at three
Other early culvert tests were performed at the Univ. of North locations on the roof of the culvert, 0.09, 0.3, and 0.5 B from the
Carolina. Braune et al. 共1929兲 reported on tests performed on outside. The pressure distribution showed a parabolic distribution,
culvert pipe placed on weighing devices. The ratio of the pressure with the ratio of the measured pressure to the overburden pressure
to overburden pressure was 1.27 for a cast iron pipe with a being 1.90, 1.06, and 0.66 at each of the three respective loca-
H / B = 4.5 and an inside diameter to thickness 共d / t兲 ratio of 30, tions. If a parabola is fitted to the measured pressures, an average
1.31 for a concrete pipe with H / B = 4.24 and d / t = 11.6, 1.68 for a pressure can be calculated as the area under the pressure diagram
solid plug with H / B = 4.5, and 1.40 for a solid plug with divided by the culvert width. This results in an average pressure
H / B = 7.2. of 1.32 times the overburden pressure.
Binger 共1947兲 reports on pressure measurements at the center Vaslestad et al. 共1993兲 report on the pressure measurements
of a 2.74 m wide by 3.30 m high box culvert built as part of in the center of a 2.0 m wide by 2.55 m high box culvert under
the Panama Canal construction under approximately 15.2 m of 9.8 m of silty clay fill 共H / B = 4.9兲. The measured pressure was
sandstone fill 共H / B = 5.6兲. The measured vertical pressure at the 1.24 times the overburden pressure. Finite element analyses
center of the culvert was approximately 1.8 times the overburden 共Katona et al. 1981兲 as well as other instrumented culverts
pressure. 共Dasgupta and Sengupta 1991兲 have shown that the pressure is
Trollope et al. 共1963兲 report on pressure measurements on the generally lowest in the center of the culvert. Thus, the 1.24 is
crown of a horseshoe shaped culvert under the embankment of a probably a lower bound to the average pressure on the culvert.
dam. Eight different locations were instrumented, and the Yang 共2000兲 reported the results of pressure measurements on
pressures were 1.6, 2.6, 1.5, 1.7, 1.7, 2.9, 1.9, and 3.2 times the a 9.9 m wide by 3.7 m high double box culvert with an H / B ratio

644 / JOURNAL OF BRIDGE ENGINEERING © ASCE / NOVEMBER/DECEMBER 2005

J. Bridge Eng. 2005.10:643-649.


Fig. 2. Cross section of embankment and culvert

B for an H / B ratio of 1.7. Only one cell of the culvert was


instrumented, as the culvert is perpendicular to the embankment
and the response was assumed to be symmetrical about the cul-
Downloaded from ascelibrary.org by New York University on 05/12/15. Copyright ASCE. For personal use only; all rights reserved.

vert centerline.
Fig. 1. Soil–structure interaction factor versus H / B for instrumented Vibrating wire strain gages and pressure cells were used to
culverts: 共1兲 Spangler 1947; 共2兲 Spangler 1950; 共3兲 Woodbury determine the internal forces and pressures on the culvert due to
et al. 1926; 共4兲 Braune et al. 1929; 共5兲 Binger 1947; 共6兲 Girdler the backfill 共Fig. 2兲. Six pairs of vibrating wire strain gages were
1974; 共7兲 Penman et al. 1975; 共8兲 Dasgupta and Sengupta 1991; used at each location, three pairs in the wall and three in the roof.
共9兲 Vaslestad et al. 1993; and 共10兲 this study At each location one strain gage was mounted towards the inside
of the culvert and one towards the outside so that axial and bend-
ing strains could be separated. Six pressure cells were used at
of 1.18. This culvert was a replacement for a failed culvert. The each location, three on the roof and three on the wall correspond-
original culvert was installed on level ground. The replacement ing to the locations of the strain gages.
was installed in the bottom of an excavated slope on the original A gage length of 150 mm was chosen for measuring the strain.
bottom slab. In order to minimize vertical and lateral pressures on This gage length is long enough to cross several interfaces be-
the culvert, backfill within 2 m of the culvert was not compacted. tween the aggregate and the cement paste, and thus measure an
The recorded pressure was 1.26 times the soil overburden. Be- average strain at the macro level. The gage length is small enough
cause of the compaction conditions, this culvert is not considered to have a nearly constant strain over the gage length; the change
any further, but it is interesting to note that even with the large in bending moment or axial force over 150 mm would be small. If
layer of uncompacted fill, the soil–structure interaction factor was the concrete were to crack in flexure this gage length would prob-
still greater than 1. ably not be sufficient to span several cracks, as is usually desired
A summary of the results reported in the literature of instru- when measuring strain in reinforced concrete. Nonetheless, it was
mented culverts is given in Fig. 1. The soil–structure interaction felt that the 150 mm gage length was best suited for the culvert.
factor is given as a function of the H / B ratio of the culvert. In all The strain gages were held in place by tying them to the rebar
cases, the soil–structure interaction factor is greater than 1.2. with plastic ties.
The vibrating wire pressure cells were 230 mm in diameter
and 6 mm thick. The pressure cells were fixed to the culvert roof
with concrete anchors through four mounting lugs around the
Description of Instrumented Culvert
edge of the plate. A quick setting high strength grout pad was
used to assure uniform contact between the plate and concrete.
The instrumented culvert results reported in this paper are from a
Medium sand was used to cover the cell and transducer housing
culvert located in Greene County, Tenn. It is a 99 m long cast-in-
to protect the cell from possible point loads or other stress distor-
place double-cell reinforced concrete box culvert. Typical inside
tions induced by large size particles in the crushed gravel. A
cell dimensions are 2.4 m high and 3.0 m wide. The thickness of
geosynthetic cover was attached to the concrete with adhesive to
the center wall is 0.28 m. The thickness of the side walls and the
separate the gravel and the sand.
top and bottom slabs vary according to changes in the overburden
height. Typical thicknesses are 0.78 m for the top and bottom
slab, and 0.41 m for the side walls. Thus, the outside dimensions
of the culvert are 4.0 m high and 7.0 m wide at the sections that Determination of Forces and Moments
were instrumented.
The culvert overlays silty clay with shallow outcrops of lime- Strain gage readings were converted to axial force and bending
stone rock. A 300 mm layer of well-graded crushed gravel was moments at each cross section. It was assumed that plane sections
used immediately below the culvert to level the ground and adjust remain plane and the strains are small enough so that the materi-
for the changes in thickness of the bottom slab. Gravel backfill als are linear elastic. The modulus of elasticity of the concrete
was placed to a height of 0.6 m above the top of the culvert roof. was estimated from the equations 共Mirza et al. 1979兲
Silty clay from a nearby cut was used for the rest of the embank- Ec = 5020冑 f cr
ment. The embankment has a slope of 2 horizontal to 1 vertical.
共2兲
Two sections along the length of the culvert were instrumented
f cr = 0.89f ⬘c 共1.17 + 0.08 log10 R兲
共Fig. 2兲. Section A is located approximately in the center of the
culvert length under full embankment height, with the final eleva- in which Ec is the modulus of elasticity of the concrete 共MPa兲;
tion of the pavement 18.9 m above the culvert roof, for an H / B f cr⫽corrected strength of the concrete corrected for speed of load-
ratio of 2.7. Section B is located under a sloping side of the ing 共MPa兲; f ⬘c ⫽28 day compressive strength of the concrete
embankment. The embankment has a height of 11.7 m at Section in MPa loaded at the nominal testing speed for cylinder tests

JOURNAL OF BRIDGE ENGINEERING © ASCE / NOVEMBER/DECEMBER 2005 / 645

J. Bridge Eng. 2005.10:643-649.


Table 1. Results of Linear Regression for Measured Roof Pressure
versus Fill Height
Soil–structure
Slope interaction
Pressure 共kPa per meter Intercept factor,
cell fill height兲 共kPa兲 r2 Fe
A4 39.3 −41 0.99 2.09
A5 29.5 −8 0.99 1.57
A6 27.9 −11 0.98 1.48
B4 26.7 −3 0.98 1.42
B5 17.8 13 0.96 0.95
B6 19.4 −4 0.97 1.03
Fig. 3. Roof pressure versus fill height above culvert for Section A
Downloaded from ascelibrary.org by New York University on 05/12/15. Copyright ASCE. For personal use only; all rights reserved.

the soil modification factor or soil–structure interaction factor, Fe


did not change with embankment height, or remained a constant.
which is approximately 0.24 MPa/ s, and R⫽loading rate The unit weight of the backfill was obtained from numerous
共0.00069艋 R MPa/ s 艋 69兲. measurements from nuclear density gage readings obtained as
The average concrete cylinder strength from five cylinders was part of the onsite inspection 共average 18.7 kN/ m3兲, and from tube
33 MPa. The rate of loading of the culvert was quite slow, with R samples 共average 18.8 kN/ m3兲. In Table 1, this measured soil unit
assumed to be 0.00069 MPa/ s. This results in a corrected strength weight of 18.8 kN/ m3 is used to obtain the soil–structure interac-
of the concrete, f cr, of 27 MPa. The modulus of elasticity, Ec, was tion factor, Fe for each of the sets of pressure cell readings. At
taken to be 25,200 MPa. The extreme fiber tensile strain was both Sections A and B, the pressure 共and Fe兲 was significantly
compared to the cracking strain of the concrete to determine higher over the stiff outside walls than over either the middle wall
if the concrete had cracked. The cracking strain of 153␮␧ was or the middle of the box.
calculated by dividing the modulus of rupture by the modulus of An equivalent uniform design pressure can be obtained for a
elasticity. None of the recorded strains exceeded this value, so all particular structural action 共e.g., end moment, midspan moment,
sections were treated as uncracked. end shear, etc.兲 by determining the uniform pressure that would
A strain distribution was obtained for a particular cross section give the same value of the structural action as that obtained
using the strain from the strain gages and the dimensions from under the actual load. The variation of the equivalent uniform
the cross section. The strain was converted to stress, which was design pressure with respect to different structural actions is small
converted to compressive and tensile forces in the concrete and 共Wood 2000兲 and can be approximated as the integral of the pres-
reinforcement. The axial force and bending moment were sure diagram divided by the region over which the integration is
obtained by summing forces and moments about the mid-point of performed. By fitting a parabola to the recorded pressure data,
the cross section. Compressive forces were taken as positive. performing the integration, and converting to an equivalent unit
Bending moments were considered positive when the tensile weight by dividing the pressure by the height of the fill, an
stresses were on the inside of the culvert. equivalent uniform weight of 30.8 kN/ m3 was obtained at Sec-
tion A and 19.6 kN/ m3 at Section B. These results correspond to
a soil–structure interaction factor of 1.64 and 1.04, respectively.
Measured Results
Results from Strain Gages in Wall
Three methods were used to determine the vertical pressures
acting on the culvert, and to compute the soil–structure interac- The strain gages readings from the wall of the culvert were used
tion factor, Fe. The pressure cell readings give a direct indication to determine the axial force in the wall. Results of the linear
of the pressure on the roof. Strain gages in the wall were used to regression for the axial forces per meter length of the wall from
determine axial forces in the wall, and a pressure was back each culvert section are shown in Table 2. The results for Section
calculated based on an assumed tributary width. Finally, strain B were quite erratic and thus less reliable than those from Section
gages in the roof were used to obtain bending moments in the A. An equivalent soil unit weight can be obtained by dividing
roof, and a pressure was obtained based on the load being the
second derivative of the bending moment.
Table 2. Results of Linear Regression for Wall Axial Forces versus Fill
Height
Results from Pressure Cells in Roof
Equivalent Soil–structure
Pressure cell readings versus fill height above the top of the Slope unit interaction
culvert are shown in Fig. 3 for Section A. A linear regression 共kN/ m per meter Intercept weight factor,
was performed between the pressure and the fill height for each Section fill height兲 共kN/ m兲 r2 共kN/ m3兲 Fe
location, with the slope giving the change in pressure per unit A1 67.4 344 0.90 33.7 1.79
increase in backfill height, or an equivalent backfill unit weight. A2 67.9 547 0.92 33.9 1.80
The intercept was ignored as it was considered to be from lack of A3 83.1 710 0.98 41.5 2.21
initial balance of the instruments. With this method, the effect of
B1 45.0 687 0.96 22.5 1.20
random errors in the readings was reduced. Results of the linear
B2 22.1 867 0.65 11.0 0.58
regression are given in Table 1. The high coefficients of determi-
B3 9.6 976 0.50 4.8 0.25
nation, r2, indicate a strong linear relationship. This implies that

646 / JOURNAL OF BRIDGE ENGINEERING © ASCE / NOVEMBER/DECEMBER 2005

J. Bridge Eng. 2005.10:643-649.


Table 3. Equivalent Unit Weights from Roof Bending Moments Table 4. Summary of Equivalent Unit Weights from Various
Measurements
Equivalent Soil–structure
Assumed unit interaction Section A Section B
pressure weight factor,
Section distribution 共kN/ m3兲 Fe Soil–
Equivalent structure Equivalent Soil–
A Uniform 34.5 1.84 unit interaction unit structure
Parabolic 36.1 1.92 Measurement weight factor, weight interaction
B Uniform 11.8 0.63 method 共kN/ m3兲 Fe 共kN/ m3兲 factor, Fe
Parabolic 13.1 0.70 Pressure cells 30.9 1.64 19.6 1.04
共parabolic
pressure distribution兲
the axial force by a tributary width and the height of the fill. Wall axial loads
A tributary width of 7 / 16 the width of one cell was chosen for Location 1 33.7 1.79 22.5 1.20
Downloaded from ascelibrary.org by New York University on 05/12/15. Copyright ASCE. For personal use only; all rights reserved.

this research, which is the average of 1 / 2 共the tributary width Location 2 33.9 1.80 11.0 0.58
for a fixed-fixed roof兲 and 3 / 8 共the tributary width for a fixed- Location 3 41.5 2.21 4.8 0.25
pinned roof兲 the width of one cell. Soil–structure interaction Roof moments
factors obtained from the axial load measurements, Table 2, are in Uniform 34.5 1.84 11.8 0.63
reasonable agreement with those obtained from the pressure cells. distribution
Parabolic 36.1 1.92 13.1 0.70
distribution
Results from Bending Moments Calculated in Roof
A linear regression was performed on bending moments in the
Table 4, although each set of measurements has its own potential
roof versus fill height, with r2 being greater than 0.9 for all bend-
sources of errors, they yield relatively consistent results. Vertical
ing moments at Section A and r2 being greater than 0.7 for all
pressures on the roof of the culvert were close to two times the
bending moments at Section B. The moments were used to back
soil unit weight at Section A, but were approximately the soil unit
calculate pressures. If a parabolic pressure distribution is as-
weight at Section B 共under the sloping embankment兲. The soil–
sumed, an equivalent uniform pressure can be obtained as the
structure interaction factor Fe = 1.64 for this culvert 共H / B = 2.7兲 is
pressure integrated over the length of the beam divided by the
plotted in Fig. 1 based on an equivalent uniform pressure obtained
length. The equivalent uniform pressure is calculated as
from the measured parabolic pressure distribution.

w=
96
L 共21 − 4.5k1 − 4.5k2兲
2 冉
M cl −
M1 + M2
2
冊 共3兲
The results shown in Table 4 for Section B are based on the
height of the embankment directly above B. The measured
vertical pressure continued to increase at Section B as the fill
where w⫽equivalent uniform pressure; L⫽length of the beam; continued to be placed over the center portion of the culvert with
M 1⫽moment at one end; M 2⫽moment at the other end; no change in the fill height at B. From axial force measurements
M cl⫽moment at the centerline; k1⫽ratio of the pressure at end 1 in the wall at the completion of the fill at Section A, the soil–
to the centerline pressure; and k2⫽ratio of the pressure at end 2 to structure interaction factor Fe was 1.53, 0.94, and 0.70 for B1,
the centerline pressure. Since the force on the beam segment is B2, and B3, respectively. Based on the pressure cell readings,
the second derivative of the bending moment with respect to the factor Fe at the time of completion of the embankment at
length and the bending moment was only obtained at three loca- Section A was 1.59, 1.11, and 1.77 for pressure Cells B4, B5, and
tions, the values of kA and kB have to be assumed. For a uniform B6, respectively. An equivalent uniform pressure distribution
pressure, k1 = k2 = 1. If a parabolic pressure distribution is assumed would result in a soil–structure interaction factor 1.30. Problems
based on the pressure cell readings at completion of construction, with the strain gages in the roof at the completion of the embank-
k1 = 1.32 and k2 = 0.94 at Section A and k1 = 1.50 and k2 = 1.09 at ment prevented a similar examination of the bending moments.
Section B. Regardless, the results from Section B suggest that portions of
If the moments in Eq. 共3兲 are taken as the moments per unit fill the culvert under the backfill side slope may experience vertical
height, as obtained from the slope of the linear regression of pressures greater than those determined from the height of the soil
moment versus fill height, then w will be the equivalent pressure column at the point in question.
per unit fill height, or the equivalent soil unit weight. Table 3
gives a summary of the back-calculated equivalent soil unit
weights from the roof bending moments, and the resulting soil– Comparison to American Society of Highway
structure interaction factors, Fe. There is little difference in the and Transportation Officials Specification
results between the assumption of a uniform pressure distribution
and a parabolic pressure distribution. The implication is that the The American Society of Highway and Transportation Officials
use of an equivalent uniform pressure distribution for design is 共AASHTO兲 Standard Specification for Bridges 共AASHTO 1996兲
sufficient. uses a soil–structure interaction factor based on the installation
conditions and the compaction of the sides. For embankment
conditions, the soil–structure interaction factor is determined as
Summary of Measurements
and Effect of Backfill Slope H
Fe = 1 + 0.20 共4兲
B
Table 4 is a summary of the measured equivalent soil unit weights
and the values of the soil–structure interaction factor, Fe, obtained where Fe need not be taken greater than 1.15 for installations with
from the three independent sets of measurements. As seen from compacted fill at the sides, and need not be taken greater than 1.4

JOURNAL OF BRIDGE ENGINEERING © ASCE / NOVEMBER/DECEMBER 2005 / 647

J. Bridge Eng. 2005.10:643-649.


for uncompacted fill at the sides. The soil–structure interaction average pressure. Finally, inspections have in general not revealed
factor recommended by AASHTO 关Eq. 共4兲兴 is shown in Fig. 1. problems with culverts due to excessive overburden pressure.
The compaction conditions at the sides for many of the instru- Therefore, although AASHTO underestimates the overburden
mented culverts from the literature were impossible to determine. pressure on a culvert, the safety level may be adequate.
For the Tennessee culvert described in this paper, the backfill The AASHTO LRFD bridge design specification 共AASHTO
was compacted at the sides. All of the measured soil-structure 1998兲 requires buried structures to be considered as nonredundant
interaction factors in Fig. 1 are greater than 1.15, with the small- under earth fill. This requires a 5% increase in the load factor,
est being 1.21. This suggests that the factor Fe in the AASHTO resulting in an increase in the reliability index to 2.22. Culverts
specification underestimates the vertical earth pressures on cul- could also be considered to fall under the Strength-IV load com-
verts. Subsequent analyses will only consider the case of uncom- bination, which is the load combination relating to very high dead
pacted fill at the sides, since this yields the greater value of Fe, load to live load force effects, which would further increase the
even though this may not represent actual conditions. load factor and the reliability index. The writers feel that the
An analysis of the Fe obtained from Spangler’s theory and the intent of the Strength-IV load combination was for long span
AASHTO specification was conducted for all the culverts shown bridges, and not culverts, and the Strength-IV load combination
Downloaded from ascelibrary.org by New York University on 05/12/15. Copyright ASCE. For personal use only; all rights reserved.

in Fig. 1. Both Spangler’s theory, Eq. 共1兲, and the AASHTO should not be used for culverts under high fill.
code are evaluated on the basis of the ratio of the experimentally
measured soil–structure interaction factor to the value calculated
by either Spangler’s theory or the AASHTO specification. For Conclusions
Spangler’s theory, the mean of the ratio is 0.92, with a coefficient
of variation of 14%. For the AASHTO formula for uncompacted Vertical pressures on concrete box culverts under high embank-
sides, the mean of the ratio is 1.08, with a coefficient of variation ments are examined, where a high embankment is one where the
of 15%. Both methods have some bias, with Spangler often height of the fill above the culvert is greater than the width of the
overestimating the measured pressure and AASHTO consistently culvert. Results from an instrumented culvert are described, with
underestimating the pressure. The random error, as measured by results from pressure cells, strain gages in the wall, and strain
the coefficient of variation is reasonable, being on the same order gages in the roof showing reasonable agreement. There was
as coefficients of variation for the strength of bridge beams which strong correlation between the height of fill and the pressure and
are 12–16% 共Nowak 1995兲. internal forces in the culvert, suggesting that the soil-structure
The safety level in box culverts is evaluated based on a interaction factor is independent of the H / B ratio.
simplified analysis to determine the reliability index The results of the instrumented culvert were compared to in-
strumented culverts under high embankments from the literature.
ln R − ln S In all cases, the soil-structure interaction factor Fe was greater
␤= 共5兲
冑V2R + V2S than 1, with a mean of 1.51. The roof pressure on box culverts is
significantly greater than the pressure due to the soil overburden.
in which ␤⫽reliability index; R⫽resistance; S⫽applied load; Although the measured pressures are greater than those obtained
V⫽coefficient of variation; and an overbar indicates a mean using the Fe given in the AASHTO specification, a simplified
value. The resistance is assumed to follow the same statistics as reliability analysis suggests that there may be a sufficient safety
the moment of a reinforced concrete T-beam, or a bias 共ratio of level in the AASHTO specifications.
actual strength to calculated strength兲 of 1.14 and a coefficient
of variation of 13% 共Nowak 1995兲. The coefficient of variation of
the load is a function of the variability of the soil unit weight and Acknowledgments
the variability of the soil-structure interaction factor. The soil unit
weight is assumed to have a bias of 1.00 and a coefficient of This work was sponsored by the Tennessee Department of
variation of 7% 共Duncan 2000兲. The statistics used for the soil– Transportation. This support is gratefully acknowledged.
structure interaction factor are those for AASHTO with uncom-
pacted sides, or a mean ratio of experimental to calculated of 1.08
and a coefficient of variation of 15%. The coefficient of variation References
of the load is determined as the square-root-of-the-sum-of-the-
squares of the coefficient of variation of the soil unit weight and American Association of State Highway and Transportation Officials
the soil-structure interaction factor, or 17%. 共AASHTO兲. 1996兲. Standard specifications for highway bridges,
For load factor design, a load factor of 1.3 is used for the dead 16th Ed., AASHTO, Washington, D.C.
load, and a strength-reduction factor of 0.9 is used for flexure. American Association of State Highway and Transportation Officials
This results in a reliability index of 1.96. This reliability index is 共AASHTO兲. 共1998兲. AASHTO LRFD bridge design specifications,
significantly below a target reliability index of 3.5 chosen for the 2nd Ed., AASHTO, Washington, D.C.
development of the AASHTO LRFD code 共Nowak 1995兲. Binger, W. V. 共1947兲. “Discussion to ‘Underground conduits—
There are several reasons to justify a lower reliability index An appraisal of modern research’ .” Proc. Am. Soc. Civ. Eng., 73,
for culverts, particularly under high fill. Nowak 共1995兲 reported 1543–1545.
reliability indices as low as 2.0 for certain bridge beams. Thus, Braune, G. M., Cain, W., and Janda, H. F. 共1929兲. “Earth pressure
experiments on culvert pipe.” Public Roads, 10共9兲, 153–176.
this safety level may be adequate. A failure of a culvert would
Clarke, N. W. B. 共1967兲. “The loads imposed on conduits laid under
generally not be a life-safety issue, although replacement costs embankments or valley fills.” Proc. Inst. Civ. Eng., Struct. Build., 36,
could be significant. Culverts under high fill essentially experi- 63–98.
ence their maximum load upon completion, or are proof loaded. Dasgupta, A., and Sengupta, B. 共1991兲. “Large-scale model test on square
Monitoring of the instrumented culvert described herein over an box culvert backfilled with sand.” J. Geotech. Eng. Div., Am. Soc. Civ.
approximate 4 year period after completion showed no increase in Eng., 117共1兲, 156–161.

648 / JOURNAL OF BRIDGE ENGINEERING © ASCE / NOVEMBER/DECEMBER 2005

J. Bridge Eng. 2005.10:643-649.


Duncan, J. M. 共2000兲. “Factors of safety and reliability in geotechnical Spangler, M. G. 共1968兲. “Discussion to ‘The modification of the
engineering.” J. Geotech. Geoenviron. Eng., 126共4兲, 307–316. pressures on rigid culverts with fill procedures’ .” Highway Research
Girdler, H. F. 共1974兲. “Loads on culverts under high embankments.” Record, 249, Highway Research Board, Washington, D.C., 41–43.
Research Rep. No. 386, Interim Rep. No. KYHPR-72-68; HPR-1(9), Trollope, D. H., Speedie, M. G., and Lee, I. K. 共1963兲. “Pressure
Part II, Department of Transportation, Lexington, Ky.
measurements on Tullaroop dam culvert.” Proc., 4th Australia–
Höeg, K. 共1968兲. “Stresses against underground structural cylinders.”
J. Soil Mech. Found. Div., 94共4兲, 833–858. New Zealand Conf. on Soil Mechanics and Foundation Engineering,
Katona, M. G., Vittes, P. D., Lee, C. H., and Ho, H. T. 共1981兲. 81–92.
“CANDE-1980: Box culverts and soil models.” Rep. No. FHWA/RD- Vaslestad, J., Johansen, T. H., and Holm, W. 共1993兲. “Load reduction on
80/172, Federal Highway Administration, Washington, D.C. rigid culverts beneath high fills: Long-term behavior.” Transportation
Mirza, S. A., Hatzinikolas, M., and MacGregor, J. G. 共1979兲. “Statistical Research Record. 1415, Transportation Research Board, Washington,
descriptions of strength of concrete.” J. Struct. Div. ASCE, 105共6兲, D.C., 58–68.
1021–1037. Wood, S. M. 共2000兲. “Internal forces in a reinforced concrete box
Nowak, A. S. 共1995兲. “Calibration of LRFD bridge code.” J. Struct. Eng., culvert.” Master’s thesis, The Univ. of Tennessee, Knoxville, Tenn.
121共8兲, 1245–1251.
Downloaded from ascelibrary.org by New York University on 05/12/15. Copyright ASCE. For personal use only; all rights reserved.

Woodbury, W. H., Bayer, E. J., Botts, A. E., Daley, C. A., Pettus, J. K.,
Penman, A. D. M., Charles, J. A., Nash, J. K. T. L., and Humphreys, J. D.
and Wilks, J. R. 共1926兲. “Corrugated metal culverts for railroad
共1975兲. “Performance of culvert under Winscar dam.” Geotechnique,
purposes. Preparing specifications, with assistance of committee on
25共4兲, 713–730.
Spangler, M. G. 共1947兲. “Underground conduits—An appraisal of iron and steel structures.” Bull. Am. Railway Eng. Assoc., 27共284兲,
modern research.” Proc. Am. Soc. Civ. Eng., 73, 855–884. 794–828.
Spangler, M. G. 共1950兲. “Field measurements of the settlement ratios Yang, M. Z. 共2000兲. “Evaluation of factors affecting earth pressures on
of various highway culverts.” Iowa Engineering Experiment Station buried box culverts.” PhD dissertation, Univ. of Tennessee, Knoxville,
Bulletin 170, Ames, Iowa. Tenn.

JOURNAL OF BRIDGE ENGINEERING © ASCE / NOVEMBER/DECEMBER 2005 / 649

J. Bridge Eng. 2005.10:643-649.

You might also like