Hydrodynamic Loading on River Bridges

Stefano Malavasi1 and Alberto Guadagnini2

Abstract: During a flood, a bridge may be partially or entirely submerged by the flow and the subsequent loading of fluid plays a major
role in assessing the vulnerability of the structure. We have performed laboratory experiments to quantify the hydrodynamic loading on
a bridge deck with rectangular cross section. We measured the time-varying hydrodynamic forces acting on the obstacle for various
submergences and Deck Froude numbers. The experimental results have been analyzed via dimensional analysis and relationships
between time-averaged force coefficients 共drag, lift, and moment coefficients兲, the Deck Froude number and geometrical parameters of the
problem are discussed and compared against relevant literature. Due to the presence of a free surface, force coefficients can be either
larger 共by more than a factor of 2兲 or lower than the corresponding values of the unbounded domain. The experimental drag coefficients
are then compared with the results obtained by the momentum equation.
DOI: 10.1061/共ASCE兲0733-9429共2003兲129:11共854兲
CE Database subject headings: Hydrodynamic pressure; Bridges; Drag coefficient; Hydraulic models; Lift coefficients.

Introduction structure was positioned in a laboratory flume at an elevation such

that the effects of the channel floor were deemed negligible. The
When a bridge crossing a waterway is partially or entirely sub- force acting on the bridge was calculated by measuring the pres-
merged during a flood event, its deck may be subjected to signifi- sure distributions on the beam located at the center of the flume,
cant hydrodynamic loading. Together with the information of the within a two-dimensional scheme, and disregarding the contribu-
probability of occurrence of flooding, the proper estimate of load- tion of the shear stresses along the bridge surfaces. A model was
ing exerted by the flow on the structure is relevant for designing built to reproduce a full-scale bridge and the results were up-
or evaluating its vulnerability. When considering the specific scaled to the prototype by Froude similarity, assuming that the
problem of river–bridge loading, we expect the presence of the process is valid with respect to the Reynolds number, R, within
channel walls and a water-free surface to cause some deviations the range 1⫻104 ⬍R⬍5⫻106 . Denson 共1982兲 presented an ex-
from the behavior typical of infinite domains. The effects of the perimental study of the drag, lift, and momentum coefficients for
river walls and floor can be neglected in some practical situations partial and total submergence of three different types of girders
for large river cross sections and elevation of the bridge deck. decks. Denson studied the dependence of the force coefficients on
However, a combination of such geometry with large submer- a bridge Froude number, V/ 冑gᐉ, relative inundation depth, h/ᐉ,
gences of the bridge deck, mimicking an unbounded domain, is of and relative thickness of the bridge, s/ᐉ. Here, h is the inundation
limited practical interest. For the design and vulnerability assess- depth 共or the water depth upstream the bridge兲, ᐉ is the total
ment of river bridges, the most interesting situation occurs when bridge length in the flow direction, V is the upstream mean flow
the presence of a free surface prevents one from directly applying velocity, g is gravity, and s is the total bridge thickness. The drag
models of fluid–structure interactions derived for unbounded do- and lift coefficients are evaluated using the parameters s and ᐉ,
mains. Interactions between free surface flows and man-made respectively, as characteristic lengths, while ᐉ 2 was used for ana-
structures are mainly studied with reference to circular cylinders lyzing the momentum coefficient. Although an extensive series of
共Arntsen 1996; Hoyt and Sellin 2000; Zhu et al. 2000兲 and are data are presented, no interpretation of the physical meaning of
mainly focused on the analysis of flow field distortion. The sig- the evidenced dependencies is offered. Both Tainsh 共1965兲 and
nificant influence of the shape of the object renders these results Denson 共1982兲 assumed the parameters to be independent of the
usable only for a limited number of general considerations. Tainsh Reynolds number, R. We comment on this by noting that, while
共1965兲 investigated the force on the beams of a four- and three- such a self-similarity assumption is justified for cylinders of a
beam bridge, for a totally and partially submerged bridge. The rectangular section in unbounded uniform flows 共Blevins 1984兲, it
becomes at least questionable for a different shape of obstacle and
1
Assistant Professor, Dipt. Ingegneria Idraulica, Infrastrutture Viarie, in the presence of a free surface. Naudascher and Medlarz 共1983兲
Ambientale, Rilevamento, Politecnico di Milano, Piazza Leonardo Da directly measured global drag, D, acting on a girder bridge model
Vinci, 32, 20133 Milano, Italy. E-mail: [email protected] by means of a dynamometer. They analyzed the effects of; 共1兲 the
2
Professor, Dipt. Ingegneria Idraulica, Infrastrutture Viarie, Ambien- number and the wheelbase of the girders; 共2兲 the elevation of the
tale, Rilevamento, Politecnico di Milano, Piazza Leonardo Da Vinci, 32, bridge; and 共3兲 the angle between the flow and the bridge axis on
20133 Milano, Italy. E-mail: [email protected] time-averaged hydrodynamic loads for the partially submerged
Note. Discussion open until April 1, 2004. Separate discussions must bridge. They introduced the following functional relationship be-
be submitted for individual papers. To extend the closing date by one
tween the local hydrodynamic load coefficient, C D⬘ , and the con-
month, a written request must be filed with the ASCE Managing Editor.
The manuscript for this paper was submitted for review and possible trolling parameters:
publication on August 23, 2001; approved on April 9, 2003. This paper is
part of the Journal of Hydraulic Engineering, Vol. 129, No. 11, Novem-
ber 1, 2003. ©ASCE, ISSN 0733-9429/2003/11-854 – 861/$18.00.
⬘⫽
CD
1 F 共 x 兲 /h
n ␳␯ m 2
/2
⫽f 冉 ␯m h
, ,
h⫺h b a
冑gh h⫺h b c c
, ,␣,n 冊 (1)

854 / JOURNAL OF HYDRAULIC ENGINEERING © ASCE / NOVEMBER 2003

where n⫽number of girders; F(x)⫽time-averaged local hydro-
dynamic load per unit length at 共transverse兲 location x on the
bridge; ␯ m ⫽depth averaged velocity of flow; c⫽wheelbase of
girders; a⫽width of the horizontal girder plates; and the remain-
ing quantities have been reported in the Notation section. Dimen-
sionless quantities in parenthesis include; 共1兲 the local Froude
number, which is computed as the Froude number of the channel
flow; 共2兲 a measurement of the obstruction, expressed by the sub-
mergence of the bridge; and 共3兲 the shape factor of the girders.
After measuring the total dynamic load on the bridge, Tainsh and
Denson calculated C D⬘ as:

⬘ ⫽D nh
CD 冋 1 ␳V 2
sin ␣ 2 冕冉 冊 册
b

0
␯m
V
2
dx
⫺1
(2)

Data were then interpreted by a Karman–Prandtl-type equation

for rough-wall resistance as a function of the submergence of the Fig. 1. Control volume and forces considered in the momentum
bridge while keeping the remaining parameters as constants dur- equation
ing the experiments.
The Federal Highway Administration 共FHWA兲 共1995兲 pro-
posed the following equation for submerged or partially sub-
merged bridge superstructure: F ⬘ ⫽ f 共 b,ᐉ,s,h b ,i,k b ,k c ,␣,h u ,V u ,␳,␮,t,g 兲 (4)

V2 where i⫽slope of the flume; k b and k c ⫽measure of the average

F D ⫽C D
* ␳H (3) surface roughness of the bridge and channel, respectively; ␮
2
⫽flow dynamic viscosity; t⫽time, and the remaining quantities
here F D ⫽time-averaged drag force per unit length of bridge; the have been previously defined or reported in the Notation section.
water density, ␳⫽assumed equal to 1000 kg/m3 ; H represents the When assuming steady-state mean flow, taking ␳, s, and V u as
flow area per unit length obstructed by the deck 共m兲; V⫽mean repeating parameters and limiting our analysis to the time-
flow velocity 共m/s兲, and C D* ⫽drag coefficient per unit length. averaged force, F, the force coefficient is defined as

冉 冊
FHWA suggested a constant value within the range 2⭐C D*
F b ᐉ hb kb kc hu
⭐2.2. This equation has also been quoted in more recent publi- C F⫽ ⫽fI , , ,i, , ,␣, ,RS ,FS (5)
cations 共Hamill 1999兲 without any additional detail. 0.5␳V 2u sb s s s s s s
We present and discuss the results of an experimental study of where Rs ⫽␳V u s/␮, and Fs ⫽V u / 冑gs are respectively, the Rey-
the interaction between a free surface flow and a bridge deck. The nolds and the Froude number referring to the thickness of the
latter is modeled as a cylinder of rectangular cross section, with bridge deck, hereinafter, respectively, called Deck Reynolds num-
no piers in the river, since a simple geometry is more suitable for ber and Deck Froude number, to make a clear distinction from the
the basic understanding of the mechanisms governing the phe- Reynolds and Froude number of the flow. As will be detailed in
nomenon. This nominally restricts our analysis to a particular type the Experimental Setup section, we carried out the experiments
of structure. The study offers interesting practical and phenom- with constant geometry 共i.e., constant s, h b , ᐉ, b), channel slope
enological aspects in that: 共1兲 it gives an original quantification of (i⫽0), average surface roughness (k b ,k c ), and with ␣⬇0.
the relative weight of the various parameters controlling the prob- Under these conditions, Eq. 共5兲 can be expressed as
lem; 共2兲 the final results may be used as a starting point to evalu-
ate the necessity of extending the analysis to other shapes 共to
further address this question, we propose a comparison between
C F ⫽ f II 冉 h u ⫺h b
s
,RS ,FS 冊 (6)
our results and those obtained in literature on girder bridges兲; and
We note that there is a direct proportionality between Rs and Fs
共3兲 our experimental methodology may be used to extend the
when the fluid and the bridge geometry do not change within the
analysis to different situations. First, we set the problem and iden-
experiments. Thus, one could 共in principle兲 eliminate one or the
tify the controlling dimensionless groups. Then, experiments are
other from the functional relationship 共6兲. Furthermore, relying on
performed in a laboratory flume and the hydrodynamic forces
the Deck Froude number produces some experimental advan-
acting on the obstacle are directly measured by means of built-in
tages, since the influence of the depth of the flow and of its mean
dynamometers. The results are given in terms of time-averaged
velocity is well identified in the dimensionless groups. On this
drag, lift, and moment coefficients as a function of the identified
basis, the following functional relationships can be introduced for
dimensionless controlling groups. Finally, experimental drag co-
the drag and lift coefficients, C D and C L , respectively
efficients are compared with the results obtained by the direct
application of the momentum equation. This makes it possible to
assess the relative importance of the different levels of simplify-
ing assumptions usually adopted to solve the latter.
C D ⫽ f VI 冉 h u ⫺h b
s
,FS ; 冊 C L ⫽ f VII 冉 h u ⫺h b
s
,FS 冊 (7)

An explicit form of the functional relationship between C D 共or

C L ) and related parameters can be obtained by the momentum
Dimensional Analysis
equation approach. Upon considering the control volume around
When the bridge deck can be represented by means of a cylinder the bridge model depicted in Fig. 1, assuming 共1兲 incompressible
of rectangular cross section, the time-dependent hydrodynamic steady state flow in the mean, and 共2兲 zero slope of the channel
force, F ⬘ , acting on the deck under complete or partial submer- bottom, projecting the momentum equation along the direction of
gence is expressed by the functional relationship: the mean flow yields:

JOURNAL OF HYDRAULIC ENGINEERING © ASCE / NOVEMBER 2003 / 855

 D⫽⌸ u ⫹⌫ u ⫺⌸ d ⫺⌫ d ⫺T (8) mean forces were obtained using an acquisition standard period of
20 minuts with the aim of a future use of the same acquisitions for
Introducing the drag coefficient C D leads to higher-order analysis. The frequency analysis of preliminary tests
共 h 2u ⫺h 2d 兲 2 pointed out a resonance frequency of the bridge model of about
C D ⫽g ⫹ 共 ␤ u V 2u bh u ⫺␤ d V 2d bh d ⫺gJW 兲 (9) 28 Hz for the drag and 20 Hz for the lift channels. However,
V 2u s bsV 2u
because of the expected low characteristic frequency of the phe-
Eq. 共9兲 is based on the following additional assumptions: 共1兲 lin- nomenon, the influence of the bridge resonance does not signifi-
ear pressure distributions at sections u and d, 共2兲 mean velocity cantly affect the study of the mean hydrodynamic loads. The drift
normal to sections u and d, 共3兲 flux of turbulent momentum at associated with the acquisition channels was also analyzed. Since
sections u and d proportional to the mean flow momentum-flux it was not possible to identify a systematic trend of the drift for
共coefficients ␤ u and ␤ d include this assumption, as well as the the four channels, we minimized the drift error by directly evalu-
momentum-flux correction coefficients兲; 共4兲 negligible effects of ating the drift associated with each channel in each experiment.
entrapped air and viscous and turbulent stresses over the free The zero-reference value of each channel was calculated as the
surface along the x direction; 共5兲 the integral of the shear stresses mean from a 20 minuts acquisition at the end of each experiment,
acting on the channel walls is proportional to the mean energy with still water in the flume. The mean drag and lift of an experi-
slope, J, multiplied by the weight of the volume, W; the latter is mental run were evaluated by subtracting the corresponding zero-
defined as the total volume between sections u and d, when the reference value. Since the output of the acquisition channels was
bridge deck is substituted by an equivalent volume of fluid. set to zero under the static condition, corresponding to a com-
pletely submerged bridge, the measured data were related only to
the dynamic component of the loading of fluid. In cases of partial
Experimental Setup submergence of the bridge model, the dynamic component of the
loading was obtained after eliminating, from the measured lift, the
The experiments were performed in a 5.0 m long Plexiglas labo- buoyancy force corresponding to the nonsubmerged part of the
ratory flume 共width b⫽0.50 m; height of side walls 0.60 m兲 with bridge deck. In these cases, the volume of the nonsubmerged part
zero bottom slope. Contrary to the techniques used by Naudasher of the deck model was estimated assuming a linear trend of the
and Metz 共1983兲 and Denson 共1982兲, the direct measurement of free surface between the flow depths at the control cross sections
forces was performed using a pair of ad hoc built-in dynamom- u and d. This was necessary due to the nonhydrostatic nature of
eters 共Cigada et al. 2001兲. The dynamometers were covered with the vertical pressure distribution, which prevented the use of pi-
a polyvinyl chloride structure to obtain a rectangular cross section ezometer tubes, and to the high fluctuations of the flow depth very
共length ᐉ⫽0.18 m along the channel axis, thickness s⫽0.06 m) close to the bridge.
and connected to an automatic acquisition system. The PVC cover In this work, we tested a wide range of free surface depths,
is subdivided into three parts and only the middle one 共of width corresponding to both fully 共to approach the more studied un-
⫽0.25 m) is connected to the dynamometer. Each dynamometer bounded flow condition兲 and partially submerged bridges, i.e.,
is connected to two normal measurement channels 共one for hori- h * ⫽0.0– 6.0; the Deck Froude number Fs varied within the range
zontal and one for vertical forces兲 so that time-varying force mea- of 0.26 –0.80, due to limitations of the laboratory apparatus. This
surements are taken by a total of four acquisition channels. The corresponds to a mean flow velocity in the range of 0.20–0.62
box was placed at an elevation h b ⫽0.14 m above the floor of the m/s 共flow rate⫽0.019– 0.082 m3 /s) and to water levels, h u , in the
flume and at a distance of 3.35 m from the beginning of the flume. range of 0.14 –0.50 m, rendering a series of Froude numbers of
The thickness of the bridge deck model was chosen to comply the flow, F, between 0.09 and 0.53.
with restrictions dictated by the optimal size of the dynamometers
共Cigada et al. 2001兲. The remaining geometrical parameters were
fixed on the basis of the capability of the experimental facility and Experimental Results and Discussion
the desire to mimic a significant real scale situation at the labo-
ratory scale. For instance, adopting a geometric reduction scale The experiments on the model of the bridge deck have provided
␭ g ⫽1/33 allows for the reproduction of a typical field scenario in the integral value of the hydrodynamic loading on the sensitive
which hydrodynamic loads might be of some interest. portion of the obstacle. Within the experimental range, we veri-
The flow conditions were controlled by means of flow rate and fied that the velocity distribution within an inner core of the flow
free surface level measurements. The flow rate was measured field, acting directly on the sensitive part of the bridge model, was
through a rectangular sharp-crested weir 共UNI 6871-71P 1971兲. not influenced by the presence of the side walls. As already de-
The mean-free surface level was measured using piezometer scribed, the loading on the obstacle is measured by summing the
tubes at nine crosssections along the flume. The mean level was dynamometers readings along the drag and lift directions. The
obtained from three measurement points along each cross section. moment, M , acting on the bridge is evaluated by multiplying the
Additional water level measurements at the two reference sec- two components of the lift 共one for each dynamometer兲 by the
tions u and d, placed, respectively, at a distance of 0.6 m and 1.5 distance between each dynamometer and the centroid of the trans-
m from the bridge model 共Fig. 1兲, were taken by a micrometer verse section of the bridge deck.
water gauge, to assure more accurate input for the momentum
equation. An impeller flow meter was used to measure the veloc-
Force Coefficients via Direct Force Measurements
ity distribution of the flow.
Preliminary tests allowed identification of the frequency and Here, the results of the direct force acquisitions are presented in
the duration of the acquisition of the forces acting on the struc- terms of force coefficients. We remark that the drag and lift co-
ture. Data acquired with a 100 Hz frequency reached a good efficients are calculated by using the total frontal area of the ob-
stability of mean and variance within 2 minuts from the beginning stacle (s•b/2), which is kept constant even for a partially sub-
of the acquisition. This notwithstanding, all results related to the merged obstacle (0⭐h * ⭐1). We emphasize that the use of the

856 / JOURNAL OF HYDRAULIC ENGINEERING © ASCE / NOVEMBER 2003

 Fig. 3. Comparison between current experiments for Fs ⬇0.7, and
Fig. 2. Dependence of the experimental drag coefficient on h * and 0.8 and data from Tainsh 共1965兲 and Denson 共1982兲. Our model:
Fs , together with the 共constant兲 value of C D ⫽1.3 for unbounded s⫽0.06 m, ᐉ⫽0.18 m, ᐉ/s⫽3, and h b /s⫽2.33; Tainsh: s⫽0.1 m,
flow 共Blevins 1984兲 and the FHWA 共1995兲 expression ᐉ⫽0.56 m, ᐉ/s⫽5.6, and h b /s⫽very high; Denson-M1:
s⫽0.062 m, ᐉ⫽0.3 m, ᐉ/s⫽4.8, and h b /s⫽2.41; Denson-M2:
s⫽0.057 m, ᐉ⫽0.3 m, ᐉ/s⫽5.3, and h b /s⫽2.63; Denson-M3:
total frontal area of the obstacle, also in the case of partially s⫽0.085 m, ᐉ⫽0.3 m, ᐉ/s⫽3.5, and h b /s⫽1.76.
submerged obstacle, is of immediate physical meaning and prac-
tical use. Moreover, as explained earlier, a precise measurement concerns about the safety of the structure might be of secondary
of the free surface level next to the obstacle was extremely diffi- relevance as compared, for example, to the risk of upstream
cult to obtain. flooding. For instance, for Fs ⫽0.80, the value of C D max⫽3.24 is
The moment coefficient, C M , was calculated as reached when the submergence of the bridge is about twice the
M thickness of the bridge deck 共i.e., h * ⫽1.81). This might corre-
CM⫽ (10) spond to a situation where flooding is the predominant source of
1
•␳•V 2u •s• 共 b/2兲 •ᐉ concern for human safety while the stability of the structure be-
2
comes of less importance. Therefore, it seems reasonable to sug-
where the moment, M , acting on the bridge deck was obtained gest the adoption of a value of C D ⫽3.40 as the upper limit of the
summing the moments generated by the two lift components with drag coefficient of a bridge deck of rectangular cross section with
respect to the barycentre of the sensitive part of the obstacle. In a shape factor of ᐉ/s⫽3. We note that C D on cylinders of rect-
the computations, we followed the sign convention of Fig. 1, angular cross section decreases with the increasing of ᐉ/s 共when
where the moment of forces is assumed positive if clockwise. The ᐉ/s⬎0.6) in unbounded flows 共Blevins 1984兲. This behavior is
drag, lift, and moment coefficients are obtained as averages on a also observed in free surface flows, as evidenced by the data of
100 Hz 20 minuts acquisition. Denson 共1982兲 for girder decks 共Fig. 3兲. After the peak and for
Fig. 2 depicts the dependence of the experimental drag coef- h * ⬎1.5 C D generally increases with Fs . For Fs ⫽0.26 the drag
ficient on h * and Fs , together with the 共constant兲 value of C D coefficient reaches the value of unbounded flow (C D ⫽1.3;
⫽1.3 for unbounded flow 共Blevins 1984兲 and the expression of- Blevins 1984兲 at approximately h * ⫽5. We comment on this by
fered by FHWA 共1995兲. To render the latter amenable to direct arguing that in this case, either the effects of the boundary con-
comparison with our results, Eq. 共3兲 was manipulated, upon ditions of free surface and channel floor on C D are negligible or
adopting C D* ⫽2.2, multiplying it by the relative submergence of they compensate for each other in order to attain the unbounded
the bridge, h * , and taking into account the width b of the flume. flow result. On the other hand, increasing the Froude number
This resulted in a linear dependence of C D on the relative sub- renders the effects of the boundaries significant. This is suggested
mergence, regardless of Fs . We note that the drag coefficient by the trend of experimental points related to Fs ⫽0.44, that ap-
increases with the relative submergence of the obstacle for a par- pear to reach a constant value, different from that of unbounded
tially submerged bridge; this behavior persists also for some val- flow, approximately for 4⭐h * ⭐5. This suggests that the effects
ues of h * larger than unity, depending on FS . The influence of the of the different boundary conditions tend to compensate, as ar-
latter is limited for h * ⭐1. Even though the experimental depen- gued for the experiments at Fs ⫽0.26. Additional comments about
dence of C D on h * is not linear for 0⭐h * ⭐1, the behavior pre- the influence of the shape of the bridge deck and the bottom of the
dicted by Eq. 共3兲 is consistent with our results and generally over- flume can be made on the basis of the results presented in Fig. 3
estimates them. Increasing the free surface level over the bridge where our results for Fs ⬇0.7, and 0.8 are superimposed to those
causes C D to reach a peak and then to decrease. Eq. 共3兲 fails to of Tainsh 共1965兲 and Denson 共1982兲. Even though in the presence
predict the presence of the peaks observed for h * ⬎1.0, denoting of different bridge shapes, there is a general qualitative agreement
that it was derived for low submergence values. The peak value, among the predicted behaviors of C D . Specifically, our data for
C D max , of C D is a function of both h * and Fs . While it was not h * ⬍1 compare favorably with those of Tainsh 共1965兲. The latter
possible to identify the format of the functional relationship of are characterized by a very large h b /s and, for the given Fs , tend
C D max , the experiments display an upper bound (C D max⫽3.38 to the limiting value typical of unbounded flows for increasing
when h * ⫽1.2 and Fs ⫽0.44). The identification of a peak value is h * . As previously discussed, we note that the drag coefficient
consistent with what observed by Tainsh 共1965兲 and Denson decreases for increasing the shape factor ᐉ/s. Moreover, for sub-
共1982兲 for different types of bridge decks and might prove useful mergences h * ⬎1, our data consistently overestimate the results
in engineering applications. However, we note that in some cases, of Tainsh 共1965兲 (ᐉ/s⫽5.6) and those Denson 共1982兲 related to
C D max corresponds to values of submergence of the bridge where ᐉ/s⫽5.3

JOURNAL OF HYDRAULIC ENGINEERING © ASCE / NOVEMBER 2003 / 857

Fig. 4. Proportional uncertainty, uC Dmeas% associated with drag Fig. 6. Proportional uncertainty, uC Lmeas%, associated with lift
coefficient obtained via direct force measurements coefficient obtained via direct force measurements

and 4.8. Conversely, our data considerably underestimate the re-

value of the lift coefficient, C L max , is not significantly affected by
sults of Denson 共1982兲 for ᐉ/s⫽3.5. However, in this latter case,
Fs and h * . The scenario corresponding to unbounded flow (C L
we should consider the possible influence of the channel floor on
⫽0) is reached for Fs ⫽0.26 when h * ⫽2.5, while the corre-
the results (h b /s⫽1.76). Thus, our C D values can reasonably
sponding unbounded counterpart of C D was attained for larger
represent an upper boundary for hydrodynamic loading on cylin-
relative submergences. After that, C L departs from zero, increas-
ders of rectangular cross section 共and for the same type of girder
ing 共negatively兲 with h * . On the other hand, the experiment cor-
bridge like those considered兲 with shape factor ᐉ/s⬎3, which are
responding to Fs ⫽0.44 does not reach the unbounded flow con-
relevant in practical engineering applications. We expect that this
dition 共consistently with what was observed for the drag
conclusion holds at least until: 共1兲 shear stresses on the bridge
coefficient兲 and does not appear to tend toward an asymptotic
prevail on 共or are of the same order of magnitude of兲 the effect of
limit, in the range investigated. We argue that, similar to what has
pressure drag, as in the case of large ᐉ/s, or 共2兲 h b /s is low,
been observed for the drag coefficient, the channel floor influ-
rendering the effects induced by the presence of the channel floor
ences the results to some extent. This appears more evident for
more significant for a given Fs , as seen, for example, for the
the lift coefficient, which is much more sensitive to the asymmet-
experiments of Denson characterized by h b /s⫽1.76. To gauge
ric flow conditions than the drag coefficient. From a practical
the accuracy of our experimental results, Fig. 4 depicts the uncer-
standpoint, it is interesting to that in our experiments, the module
tainty associated with the drag coefficient. Methods for evaluating
of the dynamic contribution to the lift exceeds that of the buoy-
the uncertainty of measurements are reported in Appendix I. We
ancy force only when Fs ⫽0.8, so that the resulting force acting
note that, with the exception of a limited number of points, the
on the bridge deck is directed downward. This result follows from
performance of the acquisition system can be rated as excellent.
Fig. 7, depicting the static component of the lift acting on the
Fig. 5 depicts the dependence of the time-averaged dynamic
structure 共buoyancy force兲 together with the lift force computed
component of lift coefficient, C L , on h * and Fs . The uncertainty
by adding to it the hydrodynamic lift for various h * and Fs . It
related to the corresponding measurements is reported in Fig. 6.
appears that the scenario corresponding to total submergence
The qualitative behavior of the experimental points is opposite to
coupled with very low velocity of the flow is the worst hydraulic
that observed for the drag coefficient, with a decreasing limb and
condition for the bridge stability along the vertical direction.
a minimum, which appears to be attained for h * ⬎1, followed by
Here, we do not show additional comparisons of C L with results
a rising limb. It is noted that the absolute values of C L are larger
from other authors, mainly because the distinction between static
than those of C D and always negative, denoting that the hydro-
and dynamic components of the lift was not reported in relevant
dynamic lift is directed downward. The maximum 共negative兲
literature and different results can be obtained for different bridge

Fig. 5. Dependence of the experimental dynamic lift coefficient on

h * and Fs ; 共constant兲 value of C L ⫽0 for unbounded flow is also Fig. 7. Comparison between lift on structure and its static
reported component 共buoyancy force兲 plotted as function of h * and Fs

858 / JOURNAL OF HYDRAULIC ENGINEERING © ASCE / NOVEMBER 2003

Fig. 8. Dependence of the experimental moment coefficient on h * Fig. 9. Dependence of C D on h * for Fs ⫽0.44 and 0.70 as obtained
and Fs . Data in the shaded area are related to partially submerged by: 共a兲 Current experimental data relying on direct force measure-
obstacle and were obtained under the assumption that the buoyancy ments; 共b兲 application of Eq. 共9兲 with J⫽0 and ␤ u ⫽␤ d ⫽1; and 共c兲
force associated to the nonsubmerged part of the bridge be equally application of 共9兲 with J⫽0 and ␤ u ⫽␤ d ⫽1.
distributed on the two dynamometers.

shear stresses acting over the external surfaces of control volume,

shapes when only the total lift is evaluated. Upon inspection of T, in Eq. 共9兲 is evaluated by calculating the mean energy slope J
Fig. 6 and considering that the main contribution to the uncer- on the basis of the average flow condition between the sections u
tainty of the data is due to the drift of the acquisition channels and d of Fig. 1 as J⫽V 2av/(Ks 2 R 4/3av ); here, V av and R av are ob-
共Appendix I兲, we note that such uncertainty decreases with Fs and tained, respectively, as an average between the mean velocities
h * . The same behavior can be observed for the uncertainty re- and hydraulic radii at sections u and d, and Ks is the flume
lated to C D 共Fig. 4兲. The dependence of the time-averaged mo- Strickler roughness coefficient estimated by means of the experi-
ment coefficient, C M , on h * and Fs is depicted in Fig. 8. The C M mental flow profile in the absence of the obstacle. On the basis of
is not very large and its values fall within the range ⫺0.3⬍C M proper experimental tests, we assumed in our experiments the
⬍0.7, which is in agreement 共not shown兲 with the data of Denson value Ks⫽90. Fig. 9 also shows that the largest C D values are
共1982兲. Accurate evaluations of the behavior of the momentum obtained by applying the equation 关Eq. 共9兲兴 without introducing
coefficient are, however, difficult because of the uncertainty asso- the effects of energy losses and for unit momentum flux coeffi-
ciated with this quantity. Data in the shaded area in Fig. 8 are, cients. Taking into account the energy loss between the two con-
indeed, associated with the partial submergence of the obstacle; trol sections, even though by means of an approximate procedure,
these were obtained under the assumption that the buoyancy force causes C D to more closely resemble its measured counterparts.
associated with the nonsubmerged part of the bridge can be This occurs especially when the contribution of the shear stresses
equally distributed between the two dynamometers. is not significant, for example, for low values of Fs , or when our
integral evaluation gives a better approximation of the actual dis-
tribution of the shear stresses. This happens, for example, for very
Comparison with Results from Momentum Equation
low or very large values of h * , and small A i /A u , which is the
The nature of the dependence of hydrodynamic loads on signifi- ratio between the area A i of the obstacle impacted by the flow and
cant controlling parameters is, in principle, encapsulated in the the area A u of the flow at the upstream reference section. In this
momentum equation, together with a given set of assumptions. case, the flow field is less influenced by the presence of the ob-
For instance Eq. 共9兲 is an explicit relationship between C D and a stacle; thus, the mean velocities used for evaluating J are more
series of dimensionless parameters, under some given assump- suitable to represent the flow field around the structure. Con-
tions. Calculation of force coefficients via the momentum equa- versely, the maximum difference between the measured force co-
tion is appealing when it allows one to obtain a result of imme- efficients and the results obtained via the momentum equation
diate engineering application by means of some simple occurs for 1⬍h * ⬍2, associated with relatively large values of
measurements and reasonable assumptions. With reference to the A i /A u and significant velocities. A maximum overestimate of
assumptions and the format of Eq. 共9兲 and limiting our reasoning 30% can be observed when shear stresses along the channel walls
to the drag coefficient, we note that the most problematic input are neglected and unit momentum flux coefficients are adopted.
quantities are the momentum coefficients, ␤ u and ␤ d , respec- The uncertainty associated with the momentum equation method
tively, in the reference sections upstream and downstream of the is evaluated by considering the contributions of the involved pa-
bridge, and the mean energy slope J. An accurate evaluation of rameters. Fig. 10 shows the uncertainty associated with the drag
the momentum coefficients needs a large number of velocity mea- as obtained via direct force measurement and momentum equa-
surements on the reference sections while the assumption of a tion. It is interesting to note that there is a difference of an order
value of J is a rough but commonly used method to bypass the of magnitude between the uncertainties of the two methods for
evaluation of shear stress distribution on the channel walls when small Fs and large h * . As detailed in Appendix I, each contribu-
using global equilibrium arguments. tion is accounted for by means of the related sensitivity coeffi-
Fig. 9 depicts the dependence of C D on h * for Fs ⫽0.44 and cient. The values of sensitivity coefficients of the free surface
0.70 as obtained by 关Fig. 9共a兲兴 current experimental data relying level measurements are about three orders of magnitude larger
on direct force measurements, 关Fig. 9共b兲兴 application of Eq. 共9兲 than the remaining sensitivity coefficients. Thus, the water depth
with J⫽0 and ␤ u ⫽␤ d ⫽1, and 关Fig. 9共c兲兴 application of Eq. 共9兲 measurements significantly contribute to the global uncertainty.
with J⫽0 and ␤ u ⫽␤ d ⫽1. In the latter case, the integral of the Since the dynamic contribution to the structure loading is a small

JOURNAL OF HYDRAULIC ENGINEERING © ASCE / NOVEMBER 2003 / 859

 共Fig. 7兲. Coupling these results with that observed for the
drag coefficient makes it possible 共in principle兲 to identify
favorable conditions for bridge stability. It appears that the
scenario corresponding to total submergence coupled with
very low velocity of the flow is the worst hydraulic condition
for the bridge stability along the vertical direction.
4. We offer an estimate of the reliability of a simplified ap-
proach based on the momentum equation for evaluating the
drag coefficient. When compared to experimental data, use
of the momentum equation with some simplifying assump-
tions renders overestimated drag coefficients 共Fig. 9兲. A
maximum overestimate of 30% can be observed when shear
stresses along the channel walls are neglected and unit mo-
mentum flux coefficients are adopted.
Fig. 10. Comparison between drag uncertainties due to direct force 5. The uncertainty analysis on experimental and computational
measurement, uC Dcalc , and momentum equation, uC Dmeas results has given useful information for the interpretation of
data and has demonstrated the reliability of our results. The
large influence of the free surface level on the uncertainty of
percentage of the total loading for small Fs , it is easy to under- results obtained by the momentum equation is highlighted.
stand that, in this case, the results are very sensitive to the uncer- 6. Comparison against relevant data from literature allows
tainty of the water level measurement. Moreover, the pressure commenting upon the influence on the results of the bridge
contribution to the structure loading depends on the differences deck shape and the channel floor. The latter tends to increase
between the areas of the sections upstream and downstream of the the dynamic loading induced on the structure.
obstacle and these differences also change with h * ; thus, the 7. More experimental evidence about the nature of the flow
results of the momentum equation are less reliable for small Fs field around the obstacle is needed to fully explain the be-
and very small or large h * . The practical applicability of these havior of C D and C L . This can be accomplished, for in-
results is limited by the experimental conditions and the need to stance, with the aid of flow field measurements and/or visu-
measure the free surface level at two sections up- and down- alization techniques, or numerical simulation. We think the
stream from the bridge. Nevertheless, they offer useful informa- functional dependencies evidenced in this work may serve as
tion about the reliability of such a simplified approach to the reliable tools for advancing the knowledge of the phenom-
estimate of the drag on a bridge deck. enon and setting a new basis for designing such additional
experiments.

Conclusions
Acknowledgments
Our experimental work and analysis lead to the following major
conclusions: The work was supported by the Italian Ministero dell’Universita’
1. The force and moment coefficients are significantly influ- e della Ricerca Scientifica e Tecnologica 共MIUR-research title:
enced by h * 共which summarizes the relative importance of Ricerca di base sulle vibrazioni indotte da fluidi su strutture
the depth of the free surface flow, the thickness of bridge snelle兲. The writers thank A. Deantoni and M. Lanfranconi for
deck, and its elevation from the floor of the channel兲 and Fs , contributions to experimental work and Professor Francesco Bal-
the Deck Froude number. lio for useful discussions and comments.
2. In the fluid-dynamic range tested, the dependence of the
forces and moment coefficients against h * displays a maxi-
mum 共either positive or negative兲, which changes mildly Appendix I. Force Coefficient Uncertainty
with Fs 共Figs. 2, 5, and 8兲. We evidenced a value of C D
⫽3.40 as an upper limit for the drag coefficient under a The uncertainty determination of the force coefficients was pro-
range of flow conditions of practical interest for river–bridge vided in respecting to the International Organization of
decks. Since the adopted shape factor ᐉ/s⫽3 can be consid- Standardization-GUM 共1995兲. If X is a quantity depending on y i
ered as a lower boundary for real scale cases, our values can parameters, the uncertainty of X can be evaluated as
reasonably represent an upper boundary for hydrodynamic
loading on cylinders of rectangular cross section with shape
factor ᐉ/s⬎3. The validity and practical usefulness of this uX⫽ 冑兺 冉 冊
i
⳵X
⳵y i
2
uy i2 (11)
conclusion are strengthened by comparison with results from
literature, obtained for different bridge shapes. It is also Upon considering Eq. 共5兲, the uncertainty of the force coefficient
demonstrated that the widely used FHWA 共1995兲 recommen- was evaluated by considering the contributions of each term
dations, which are valid for h * ⬍1, are consistent with our weighting its influence by the appropriate sensitivity coefficients.
results and generally overestimate them. For each measurement input, these coefficients were obtained by
3. In the experimental range, the module of the dynamic con- calculating the partial derivative of Eq. 共5兲 with respect to all
tribution to the lift exceeds that of the buoyancy force only parameters.
when Fs ⫽0.8, so that the resulting force acting on the bridge For the quantities that were treated as constant during the ex-
deck is directed downward 共Fig. 7兲. Our study shows that the periments as the specific weight, ␥, the flow density, ␳, the chan-
dynamic component of the vertical force tends to equilibrate nel length, b, the bridge thickness, s, and the weir height, p, we
buoyancy forces acting on the bridge deck for increasing Fs assumed the following extended uncertainties: u␥⫽3%; u␳

860 / JOURNAL OF HYDRAULIC ENGINEERING © ASCE / NOVEMBER 2003

⫽3%; ub⫽⫾0.001 m; us⫽⫾0.001 m; and up⫽⫾0.001 m. Un- T ⫽ integral of shear stresses acting along lateral surface
certainties related to the remaining parameters are individually of control volume of Fig. 1;
discussed next. V ⫽ mean velocity of flow;
The uncertainty, uF, associated with direct force measure- ␣ ⫽ angle between flow direction and bridge axis;
ments was evaluated under the hypothesis of linear drift behavior ␤ ⫽ momentum coefficient;
on each channel of acquisition. The drift of drag and lift compo- ⌫ ⫽ hydrodynamic force component in momentum
nents was evaluated by summing the time-average value of the equation;
unloading tests of the two channels acting on drag and lift direc- ␯ ⫽ local velocity of flow;
tions, respectively. On this basis, the force uncertainty was evalu- ⌸ ⫽ pressure force component in momentum equation;
ated as uF⫽⫾⌬/2, where ⌬ was the value of the linear drift of and
the drag and lift signals during the loading test. When the force ␳ ⫽ density.
was evaluated via the momentum equation, the uncertainties were Subscripts
always evaluated by considering the individual contribution of u ⫽ at upstream reference section;
each term involved. Starting from Eq. 共8兲 where the contribution d ⫽ at downstream reference section; and
of shear stress was considered negligible and under the assump- s ⫽ referred to thickness of bridge.
tions of the Dimensional Analysis section and using mass conser-
vation to eliminate the dependence by the mean velocity at the
downstream reference section. The uncertainty of the upstream References
mean velocity, uV u , was calculated starting from the definition of
V u as the ratio between the flow rate, measured by sharp-crested
Arntsen, O. 共1996兲. ‘‘Disturbances, lift, and drag forces due to translation
weir, and the area of the upstream reference section. The flow rate of a horizontal circular cylinder in stratified water.’’ Exp. Fluids,
uncertainty was then evaluated on the basis of the standard UNI 21共5兲, 387– 400.
6871-71P 共1971兲, while the uncertainties on the free surface level Blevins, R. D. 共1984兲. Applied fluid dynamics handbook, Van Nostrand
measurements were calculated considering the maximum ampli- Reinhold, New York, 309–347.
tude (⌬h⫽h max⫺hmin) of the recorded value during a test and Cigada, A., Malavasi, S., and Vanali, M. 共2001兲. ‘‘Direct forces measure-
assuming a rectangular probability distribution within this range. ment on a submerged bridge model.’’ Fluid structure interaction, WIT
Press, Boston, 305–314.
Denson, K. H. 共1982兲. ‘‘Steady-state drag, lift, and rolling-moment coef-
Notation ficients for inundated inland bridges,’’ Rep. No. MSHD-RD-82-077,
reproduced by National Technical Information Service, Springfield,
The following symbol are used in this paper: Virg., 1–23.
Federal Highway Administration 共FHWA兲. 共1995兲. ‘‘Stream stability at
b ⫽ channel length;
highway structures.’’ Rep. No. FHWA-HI-96-032, FHWA, Washing-
C D ⫽ drag coefficient;
ton, D.C., 56 –59.
C F ⫽ dynamic force coefficient; Hamill, L. 共1999兲. Bridge Hydraulics, E & FN Spon, London, 344 –346.
C L ⫽ lift coefficient; Hoyt, J. W., and Sellin, R. H. J. 共2000兲. ‘‘A comparison of tracer and PIV
C M ⫽ moment coefficient; results in visualizing water flow around a cylinder close to the free
D ⫽ drag component of average force on bridge deck; surface.’’ Exp. Fluids, 28共3兲, 261–265.
F ⫽ average force on bridge deck; International Organization of Standardization-GUM. 共1995兲. Guide to the
F ⫽ froude number of flow; expression of uncertainty in measurements, Geneva, Switzerland,
g ⫽ gravitational constant; 1–15.
h ⫽ water depth; Naudascher, E., and Medlarz, H. J. 共1983兲. ‘‘Hydrodynamic loading and
h * ⫽ (h u ⫺h b )/s; backwater effect of partially submerged bridges.’’ J. Hydraul. Res.,
h b ⫽ elevation of bridge deck from channel floor; 21共3兲, 213–232.
Tainsh, J. 共1965兲. ‘‘Investigation of forces on submerged bridge beams.’’
L ⫽ lift component of average force on bridge deck;
Rep. No.108, Dept. of Public Works, New South Wales Univ., Sydney,
ᐉ ⫽ bridge length; Australia, 1–25.
M ⫽ moment acting on bridge deck; UNI 6871-71P, SEPT. 共1971兲. Pompe—Metodi di prova e condizioni di
P ⫽ pressure force acting on bottom of control volume of accettazione, Ente Italiano di Unificazione, Milano, 1–13 共in Italian兲.
Fig. 1; Zhu, Q., Lin, J. C., Unal, M. F., and Rockwell, D. 共2000兲. ‘‘Motion of a
R ⫽ Reynolds number; cylinder adjacent to a free-surface: flow patterns and loading.’’ Exp.
s ⫽ bridge thickness; Fluids, 28共6兲, 559–575.

JOURNAL OF HYDRAULIC ENGINEERING © ASCE / NOVEMBER 2003 / 861