Analysis of Multicelled Box Girders by Faucharts Method
1, Introduction
When two box girders are joined by a slab, applied loads will be shared by the girders
according to the position of the load transversely and the relative stiffnesses of the girders
and the connecting slab. The longitudinal moments and shears in the girders will depend
on how the loads are distributed between the girders, and transverse bending in the slab will
be effected by the relative displacements between the girders as well as the transverse
location of the loads. Thus, there is an interaction between the longitudinal and transverse
analysis that is not present in a single box girder.
‘The following is a computerized adaptation of Faucharts method. The span is idealized as
a unitlength slice of the cross section taken at midspan and supported by springs
representing the stiffness of the girders, as shown in Figure 1. The slice is taken at midspan
because this is where the interaction effects in the joining stab will be the greatest. Analysis
both bending in the slab
of this model in a finite element program yields influence lines
and forces in the supports. These influence lines are used for transverse and longitudinal
design, respectively.
w*)
BEBE EaSe:
(0 *
i
fLIE
Figure 1. Idealization of Cross Section
>2, Model Definition
The cross section of the girders is modeled as a framed structure. This structure is
supported on vertical and rotational springs at the center of rotation of the each box girder,
as shown in Figure 2. The box girders are connected to the spring supports by weightless
rigid links which are in turn connected to the joints of the webs and bottom slab. The
connection at the bottom slab is a pin restraint so that the joint is free to rotate
independently of the rigid links. The elements are assembled the same way the span is to
be constructed: usually placing the boxes first, stressing the top slab PT, and subsequently
connecting them with a cast-in-place pour strip. In this, way the dead load will be placed
on the structure in the correct static state and the secondary moments from the post-
tensioning will be accurately accounted for. The top slab is supported horizontally at one
point on each box until the cast-in-place strip is placed. One horizontal support must thea
be removed to allow for axial shortening. Post-tensioning is modeled in the usual way,
providing an equivalent tendon area for the length of the slab being considered.
= see
Pin Restraint ~ typ
Figure 2. Frame Model of Cross Section
Slab Stiffnesses
Since the webs and flanges of a box girder are plates which are effectively restrained from
longitudinal bending, they should be modeled with the elastic plate constant D, instead of
EI. This can be done in the model by dividing I for these elements by 1-v?, Such a
modification is not necessary for a single box girder, since only the relative stiffnesses of
webs and flanges are significant for moment redistribution. It will make a small difference
in this model since the slab stiffnesses relative to the supports springs are important. The
difference will be about 4%.Support Stiffhesses
The elastic supports representing the stiffnesses of the girders must be chosen such that they
preserve the relationship between the girder stiffness and the stiffness of the joining slab;
the slab stiffness is predetermined by virtue of the unit length being considered. This can
be accomplished by considering the elastic curve of a beam subjected to a load in terms of
a Fourier series:
@ we) = Th, wi )
The displacement is then given by:
@sa@-Y, Fel: }
Since the difference in displacement between two girders will also be given in terms of a
Fourier series, the force connecting the girders (that is, the transverse shear force resultant
in the joining slab, Q,, in Figure 3) will be in the same form that was assumed for the
extemal load in Equation 1 above. Therefore, the correct equivalent stiffness for the
girders, considering the first term of the series for a uniform load only, is given by the
following:
In like manner, for the twist, the stiffness is as follows:
44g
we
@) K,= Sec, with c =
ae ys
tFigure 3. Forces on the Girder
3, Influence Lines
Two families of influence lines will be generated. The first is concerned with support
reactions. It is comprised of influence lines used to determine the fraction of the applied
load that is supported by each girder for the longitudinal analysis. The second is for effects
in the slabs and webs. These lines are used for the transverse analysis of the cross section.
Influence Line for the Vertical Reaction of the Support
This influence line gives the vertical reaction of the spring support as a function of the
transverse position of the load. It is determined according to Mueller-Breslau's principle,
by applying a load numerically equal to the support stiffness directly to the support. The
value of the influence coefficient, for vertical force in the support due to a load at any given
node, is then simply equal to the magnitude of the displacement of that node.
Since the spring support is used to represent the longitudinal bending stiffness of a girder,
the force in the support is analogous to the change in shear, or the increment of load, in
that girder. Any load on the system of girders can therefore be apportioned to the
individual girders to calculate longitudinal bending moments.In practice it is not convenient to apply loads directly to the influence lines. A
computationally more direct way is to apply the wheel loads from all of the lanes in their
respective transverse positions and calculate the total force in a support spring. This force,
divided by the total force applied in all of the lanes, will give the equivalent number of lanes
to be carried by the girder associated with that support spring. The longitudinal analysis can
thus be carried out in the normal manner, using the equivalent number of traffic lanes in
place of the acrual number of lanes. (BC does not recognize fractions of lanes, so the load
factors in the SLD block of the service module may have to be changed to get the correct
number).
It should be noted that the live load reduction factor is based on the total umber of lanes
carried by the span, not the equivalent number of lanes on a girder. It is very important,
therefore, to check the girders with every possible combination of lanes loaded (with the
corresponding reduction factor), to determine the critical load case.
It should also be recognized that the influence lines generated are only strictly valid for
loads applied to the longitudinal midspan, since that is the point considered in generating
the spring support stiffness. At other points in the span, the stiffness of the girder relative
the stiffness of the joining slab will be greater, and loads will be shared by adjacent girders
to a lesser extent, This is not a critical fault as loads away from the midspan have a
diminishing effect on longitudinal bending.
Asa final note, it is acknowledged that this development is based on an elastic analysis, and
is therefore appropriate for analysis of service loads. At ultimate, the loads may be shared
by the girders on an equal basis; or each girder may support only the loads directly applied
to it, The actual mechanism depends on the sirength of the girder relative to the strength
of the joining slab.
Influence Line for Torsional Reaction in the Support
This influence line gives the torsional moment reaction of the spring support as a functionof the transverse position of the load. It is determined in the same manner as the influence
line for the vertical reaction: a torque numerically equal to the rotational spring stiffness is
applied directly to the supports. Influence coefficients are then equal to the displacements
of the nodes.
The practical use of the influence lines for torque is different from that of the influence
lines for vertical reaction. Here the idea of an equivalent number of lanes is not
appropriate, so loads must be applied directly to the influence lines. Any longitudinal load
may be applied, as long as it is represented in terms of a Fourier series, consistent with the
development of the model. The model, however, is based on loads applied to a unit length
at midspan. To obtain the effect at the pier, due to loads along the entire length of a span,
the results must be multiplied by L/x.
Influence Lines for Transverse Flexure
The service module of BC will generate influence lines for transverse flexure at a given
point as a function of the transverse position of the load. These influence lines can be
obtained for bending in the webs and the bottom slab, as well as in the top slab. Again, the
lines are only appropriate for describing effects at the longitudinal midspan, unless the
support springs are modified. They can be used directly for uniform longitudinal loads
where every section is loaded identically and there is no appreciable sharing of loads
between the top slabs of adjacent sections (that is, the transverse shear resultant, Q,,=0).
‘The load need only be represented as a Fourier series. The actual force applied to the cross
section model is the intensity of the Fourier series at midspan multiplied by the unit length
of the model.
4, Extension to Variable Loads
‘Since the cross section is modeled with simple beam elements, no plate bending effects are
accounted for, and longitudinal changes in moments in the slab can not be caprured. It is,
therefore, not possible to use the influence lines directly for longitudinally variable loads
(such as truck wheel loads).In order to model these loads, transverse bending is decomposed into two parts: A “local”
part which represents the bending caused by a load applied directly to the top slab, and a
“general” part which consists of the bending caused by differential displacements and pwists
of the girders. These two cases are illustrated in Figure 4. Bending moments are calculated
separately for these two effects, and subsequently superimposed to obtain the total effect.
The local effect depends on the longitudinal variation of the load. For the case of a
constant uniform load, the influence lines generated by BC already include the correct local
effect. If the load is not uniform, however, adjacent sections will no longer be independent
(Q,,*0). A plate-bending influence surface (for example, a Homberg Chart) must then be
used. For these cases the local effect of transverse bending is calculated in exactly the same
manner as live loads are for a single box girder. Moment resultants are obtained from the
influence surface for an infinite plate strip rigidly clamped on both sides. These moments
are then redistributed to account for joint rotations based on a transverse model of the cross
section subjected to unit moments at the joints. In the case of the multi-celled box girder,
the local effect is by definition independent of girder deflections. The original model must
therefore be modified by replacing the springs with rigid supports to suppress the general
effect when computing moment redistribution.
‘The influence lines generated by BC are the summation of the local effect for a uniform
longitudinal load and the general effect. The general effect must therefore be isolated by
removing the local effect. It is necessary, then, to obtain the influence lines for the pure
local effect of a uniform load. This is easily done by running the influence module in BC
using the model with rigid supports. The influence lines created are then subtracted from
the influence lines generated with spring supports to obtain influence lines for the pure
general effect.
‘The bending moments for the general effect are calculated by applying the Fourier series
representations of the loads to these influence lines. The total bending is then simply the
summation of the two effects.“Local!” Effect
“General” Effect
Figure 4. Decompostion of Transverse Bending
‘Transverse moments generated in this manner occur, in the slab at midspan, where the
general effect will be a maximum, The general effect may, of course, reduce the total
moment. The slab must therefore also be designed for the local effect acting alone, with
the general effect taken as zero, as would be the case at a pier.
5, Extension to Non-Prismatic and Multiple Span Bridges
The above development is valid only for prismatic bridges, since in the integration of the
load function, EI was taken as constant (Appendix 1). Likewise, the assumed boundary
conditions applied were for a simply-supported beam. Obviously the method can be
extended to the general case by including a function for the variation of EI in the
integrations, and using boundary conditions appropriate for other geometries. In practice
this would produce difficult integrals and awkward boundary conditions, resulting in much
more computational effort and possibly a requirement for numerical integration. Great
simplification can be achieved by recognizing that the method only considers one point in
the span (usually the midspan), and that the girder is considered to be a single spring.
Because of this, a complete description of the stiffness characteristics of the girder is notEE IEE —£?-?WSKEE—wCWDW'DIDE:*=C— EEE EE EE ee ee ee a eee eee
needed; only the equivalent stiffness at midspan is required. In order to use the method in
the form presented above, the girder stiffness is represented by an equivalent simply-
supported beam with the midspan cross-section properties. This is easily done by calculating
the deflection of the real girder and solving for the length of a simply-supported prismatic
girder that would produce the same deflection under the same load. Note that the same
load (for example, point or uniform), that the system is being designed for, should be used
to determine the equivalent length.
Transverse moments derived in this manner can be used directly in design. The longitudinal
bending moments, however, must be taken from a longitudinal model of the ceal structure
carrying the number of equivalent lanes computed in the transverse analysis. The prismatic
model is used only to determine the equivalent number of lanes to apply to each girder.
‘The equivalent span length for torsion is determined in the same way, The owist at midspan
of the real structure is equated to the twist at midspan of a prismatic beam fixed at both
ends.
6. Fourier Series Representation of Loads
As outlined above, all ioads applied to the mode! must be given in terms of a Fourier series.
In fact, the variation of the load along the length of the beam is not important; it is only
necessary to know the intensity of the Fourier series at midspan (the point where the
transverse slice is taken), Tae actual force applied to the transverse cross-section model is
equal to the intensity of the Fourier series at midspan multiplied by the unit length of the
model. Below are Fourier series representations of some typical loads:
1. Uniform Load:SSS SS ee ae a a
—_ on -o- -e- - ees - - ee ee eee
Since the value of the sign term at midspan is unity, the expression becomes, taking
the first term only, simply:
2. Point Load:
we = Do, Fad]
‘This expression can be summed over the number of axles of a truck N (Figure 5) to
give the total effect of a line of wheels in a longitudinal plane:
Mz)
‘Again, evaluating this expression at midspan and taking the first term gives the
2,
s@) = Th Fil 3
2P,
wa = 2 Dea
following:
Figure 5. Multiple Truck-Axle Loads3. General Load:
wa) = OT, =)
wea) = a fe mse fe
Where w,(x) is the is the actual load function (linear, for example) that is being
reproduced with the Fourier series.“~~ ae a so —- =!
a — a — eo ee - ees ee —
Appendix |
Delection oF 2 Beam Subjected. +
a Loud Giver interms of a Fourier Series
Equilibeiams
Med
VedVv V+ wdx -(VedV/) = 0
2
wer Hot
t+ ;
* babe wd Ee (Ugdlfde ~ Me dire
Geometry: ees
ds
a - dw
$ dx NV" oi
Materiel:
&
o* it MELT g = ELT
dM
Vr ox = Ret FE) = er gs
d
Ee 3
-df . (er Spy «er Ss
let wx) +5 wm sin (ZX)
+
Wr sin(Z2x) = EL 1 oe = a od sin 42x)ee en ee ee
2s aa a a a eae
\= {ooo dx = - We & cos (4Ex) +,
Mz (Wordx = ~ Wa (EY sin (BE x) + CK ln
G7 [tonda “Bt (mex) dex
E> EE Wn (a#)? cos (Ex) + OE #CgX + Cg
As [eer > oE We EY sin (SE Fx) CR eM Gk Hey
Boundary Conditions:
Adoy=0 = C420
Mlo)=0 =7 C270 Ay? = EN sin(4™ x)
MiL)=O =7 Cr =0
A(L)20 @ C20
k = FL. Twn sin (48x)
A Leyes War sin iit x’)
ELnt
Ki it