Vous êtes sur la page 1sur 12

Computers di S~rucfures Vol. 18, No. 6, pp. 103S1046, 1984 Printed in Great Britain.

004>7949/84 $3.00 + .@I Pergamon Press Ltd.

THIN-WALLED BOX BEAM FINITE ELEMENT FOR CURVED BRIDGE ANALYSIS


S. H. ZHANG Research Institute of the Ministry of Communications, China and L. P. R. LYONS Finite Element Analysis Ltd. 15 Holborn Viaduct. London, ECIA ?BP, England
(Received 25 February 1983; received for publication 19 April 1983)

AbstractPractical design of single and multispan curved bridges requires an analysis procedure which is easy and economical to use, and provides a physical insight into structural response under general loading conditions. In the work presented, the thin-walledbeam theory has been directly combined with the finite element technique to provide a new thin-walled box beam element. The beam element includes three extra degrees-of-freedom over the normal six degrees-of-freedom beam formulation, to take into account the warping and distortional effects as well as shear. The beam may be curved in space and variable cross-sections may be included. The performance of the box beam element has been compared favourably against results obtained from full 3D shell element analysis, differential equation solutions and experimental results.

1. INTRODUCTION It is well known that the use of thin-walled box beams is

dominant in modern bridge construction having curved alignments, as it requires appreciably great torsional rigidity. Nevertheless, the advanced box type of structure presents a major problem in the prediction of its specific structural response to a general loading case. In most cases, in addition to the conventional structural actions, cross-sections of thin-walled box beams will warp out of the sectional plane and distort in the sectional plane, as well. Figure 1 shows the linearly independent distortional modes for various box beam cross-sections [16]. These effects may be significant in curved bridges, or in bridges subjected to large live loads with relatively great eccentricity. Indeed, the analysis of such a deformable box beam has been the focus of attention for many investigators in the last two decades, and various theories and analytical methods have been developed. It is obviously not possible in this paper to give a thorough review of the existing methods. However, it is now increasingly accepted that the finite element method is the most powerful and versatile numerical tool of all the methods available. One of the representative finite elements, which is particularly suitable for the analysis of shell boxes, is the flat thin shell box element developed by Lyons [15]. Jirousek et al. have presented a special macro-element, which is formed from a modified Ahmads thick shell element and an assembly element, for practical application to prestressed curved box-girder bridges [ 111. Recent progress also permits a full three-dimensional analysis by using semiloof shell elements with double curvature [ 141. To finite element researchers it would seem that the task of establishing an advanced numerical method for the static analysis of box beams has already been fulfilled, yet in practical design terms engineers have to take account of the computing costs and to deal with the volumes of computing output which to some extent prevent a physical understanding of box beam behaviour. Indeed, in many design offices, despite the rapid progress of computer hardware and software, engineers still prefer to use an approximate grillage analysis [7] or other

simplified methods [ 161for preliminary analysis or where other circumstances permit. In harmony with the relative dimensions of medium and long span box bridges, one of the logical conclusions for finite element researchers is to extend the capability of a conventional beam element and reduce the analysis of a bridge deck to a onedimensional subdivision, whilst retaining the main structural actions. Research work has been carried out assuming a discrete subdivision in the longitudinal direction, i.e. a subdivision into beam elements, by several investigators. Formulations which take account of torsional warping effects have been used for prismatic beams with open sections [l, 5, 12, 13, 211 or beams with undeformable closed sections [6]. Baiant and Nimeiri [2] have contributed a skew ended beam element for box beams curved or straight in space taking both transverse distortion and longitudinal warping into consideration. Mikkolo and Paavola [17] have conducted a somewhat similar approach for the analysis of a rectangular singlecell box girder with side cantilevers. It is observed that the methods given by the latter two papers are available only in the case of single-cell boxes. Since the known functions describing the deformation modes of the crosssection must be chosen in advance for each type of cross-section, difficulties exist in extending these methods for more complicated or more general types of cross-section. Thin-walled beam theory applicable to box beams has been established by Vlasov [19], and elaborated by Dabrowski [4] and numerous other authors. The loaddeflection equations under the combined vertical (qy), lateral (qz), longitudinal (qJ and twisting moment (m,) distributed loads can be written as follows _ EJl& +
+
GJ& _

!$
+

f),

EL + GJ= v,,
R

m =o
x

EZ,

1035

1036

S. H. ZHANGand

L. P.R.LYONS

Fig. 1. Modes

of distortion

(schematic)

for various

box beam cross-sections.

+ t
EI,
R

GJ,

e;+q,=O

Details of the notation are given later. Slope-defiection [3.9] and finite difference [10,20] mathematical techniques have been developed by Neins et al. to solve the above equations. A strategy which directly combines the thin-walled beam theory represented by eqns (l)-(4), and the finite element technique, has been adopted in the present theoretical work to give a new thin-walled box beam element. In addition to the usual six degrees-of-freedom at each beam node, three extra degrees-of-freedom have been incorporated in the finite element formulation to take account of longitudinal warping and transverse distortion. These additional degrees-of-freedom are the rate of change of twisting angle &, distortional angle of the cross-section yd, and the rate of change of distortional angle -rd.The thin-walled box beam element, which can be used effectively to cope with the static analysis of single or multicell box beams curved in space and subjected to general loading conditions, may well be regarded as the general representation of a conventional beam element. 2. BASlCASSUMFTIONS The usual assumptions associated with linear elastic small displacement theory have been adopted. These assumptions can be generally stated as being that the structural material is homogeneous, isotropic and linearly elastic, and that the actual deformations are small compared with the structural dimensions. Additional assumptions, which are mainly related to thin-walled structural behaviour, have been considered in this approach, and are as follows: (i) The dimensions of the cross-sections are

significantly less than the span lengths and less than the radii of curvature in the case of curved members. The length/width ratios of the individual component plates from which the boxes are assembled should not be less than 3. (ii) The thickness of the wails are small compared with the dimensions of the cross-section. (iii) For bending action, plane sections remain plane, but not necessarily normal to the beam axis, thus allowing for shear deformation. (iv) For warping torsion analysis, cross-sections are assumed to remain undeformed in their own plane, but may rotate about the flexural axis (locus of the shear centres) and be subject to longitudinal warping. (v) The in-plane longitudinal bending action of an individual component plate is represented by elementary beam theory and the shear deformation caused by distortion is neglected. (vi) The bending action of an individual component plate normal to its plane is represented by the flexural behaviour of an equivalent transverse frame. (vii) Diaphragms are considered to be infinitely or finitely stiff in their own plane, but perfectly flexible in the direction normal to the plane.
3. FINITEEL~NTFO~LATION

3.1 ~~nifjon of element geometry Consider the thin-walled box beam element with variable cross-sections shown in Fig. 2. The element is curved in space but the cross-sections are generated by straight lines. A cross-section is assumed to have a vertical axis of symmetry to simplify the analysis of distortion, however, for bending and torsion analysis this assumption is unnecessary. The element axis is defined as the locus.of the centroids which may be eccentric from but parallel to the flexural axis. The element has two end nodes and a midpoint node situated on the axis. A local rectangular coordinate system (x, y, z) along the axis curve is used in the element formulation. The origin of the Cartesian coordinate system is located at the centroid of the cross-section, and the orientation of

A thin-walled box beam finite element for curved bridge analysis

1037

Fig. 2. Thin-walled box beam element with three nodes. the local axes yz is assumed to coincide with the principal axes of the cross-section. The local x axis is tangential to the element axis in the direction of node 1 towards node 3. The local y axis usually represents the vertical axis of symmetry; the local z axis being defined by a right handed orthogonal system. The gIoba1 Cartesian Coordinates are in terms of a natural coordinate & as well, which varies between - 1 and t 1 on the respective faces of the element. Let r=X.i+Y.jtZrkbethepositionvectorofapointP on the element axis, then a unit tangent vector along the x direction is thin-walled box element has, therefore, nine degrees-offreedom at each node. With reference to the principal local system defined, the relation between the generalised stress resultants and the generalised strain components has been normalised, and can be expressed from the Timoshenko beam theory and the thin-walled beam theory as (Fig. 3) a=De where the generalised stress vector is T u = N,, Q,, Qz.M,,, My, Mz,-!- Br, MA B,, (11) 7 [ ELI in which N, is the axial force, Q, QL are the shear forces, MS, is the pure torsional moment, M,, Mz are the primary bending moments, BI is the torsional warping bimoment, Md is the distortional moment and B,, is the distortional warping bimoment. The total torsional moment is the sum of the pure and warping torsional moments, i&=il&,th& (I?) (10)

where i, j and k are unit vectors in the global X, Y, Z directions respectively, and the Lame factor or the Jacobian factor referring to the Jacobian matrix is

To fix the orientation of the local y axis, the coordinates of the reference nodes 4-6, which in most cases can be the mid-points of the top flanges, are required. The local z axis can then be determined by vector multiplication e, = e, Xey (7)

where the warping torsional moment M, can be obtained from the first differential of the torsional bimoment, Mw= Br. (13)

3.2 Stress-strain relationship The generalised displacements in the local coordinate system are given by 6 = lu, 0, w, %, @yt @,,yd, 8,
ydlT (8)

A similar relationship exists between the distortional moment and the distortional bimoment, Md = Br,. 04)

where u, u and w are the translations along the local x, y, z axes respectively, 0, is the angle of twist, 8, is the rate of twist, Or and 0, are average rotations about the y and 2 axes respectively, yd is the distortional angle, yd is the rate of distortion. The corresponding dispIacements in the global system are given by 6 = [U, v, w 4x, 4,34z, 8x,Ydt Ydl (9)

It should be emphasised that the shear forces and the torsional moment are referred to the shear centre of the cross-section. The generalised strain vector is

axial strain au 6x = z shear strain in y-direction where U, V and W are translations along the XYZ global axes respectively, c&.,(p, and #Pi are rotations about the same axes respectiveiy, whilst the rate of twist 6, and distortional variables -v,+. remain in locals. The Y,+
1-1 ,..

yx

&+aU=aU_,;
ax ay ax

1038

S.

H. ZHANGand L. P. R. LYONS

Fig. 3. Generalisedforces and displacements. shear strain in z-direction where X, Y and Z are the global coordinates of the points on the beam axis curve. It can be seen that a modification has been made to the torsional warping strain to take into account the effect caused by the initial curvature. It states that the torsional warping displacement can be assumed as not only being proportional to the rate of twist but also to the bending rotation. Finally, as a result of the orthogonalisation the rigidity matrix can be shown to be

torsional strain

flexural strain about y-axis

flexural strain about z-axis

torsional warping strain

(17) where: A is cross-sectional area; A,,, A,, are effective shear areas in the y and the z-directions respectively; Jr is torsional moment of inertia; I,, I, are primary bending moments of inertia about the y and z axes respectively; J, is torsional warping moment of inertia; p, is warping shear parameter; Ja is distortional second moment of area: and Jr1 is distortional warping moment of inertia. It should be noted that the bending moments of inertia should be calculated on the basis of an effective flange breadth replacing the actual width to account for the effect of shear lag [18]. The material properties are characterised by the Youngs modulus of elasticity E and the shear modulus G which is expressed as

distortional strain CL = /d distortional warping strain.

The radius of curvature R introduced in the torsional warping strain component can be expressed as 1?=[($)2+($)l+($)2]-~ (16)

G-E--

2(1 t v)

(18)

A thin-walled box beam finite element for curved bridge analysis where v is the Poissons ratio. The conversion modulus of elasticity is

1039

The strain matrix B is then obtained by combining eqns (22) and (24), hence B=LN and L = BS. (26)
nodal

(25)

3.3 Shape functions

define the displacement field Only Co continuity is required for the extensionalflexural effects, and the following quadratic shape functions are used Ni=~([2t&,) for i=land3 (20)

to

3.5 The element sti$ness matrix and the equivalent


force vector

The element stiffness matrix which is of order 27 x 27 is expressed by the usual relationship k = fl +(1/z) B=DBdx= JBT DB d[. I -1 I -(I/Z) (27)

Ni = (1 - .$*) for i = 2

where &,= &. The quadratic shape functions are also used for mapping the element geometry. In contrast to the extensional-flexural effects, it is necessary to satisfy C, continuity for warping torsion and distortion. A series of special fifth order polynomial shape functions have been derived and can be expresses as Ni,=$(4t5[o-2[2-3[d) for i = 1 and 3 (21) Ni2 = $5 (1 t [,,)(1 - 5) Ni, = (l- 5) for i = 2 Ni, = .Ji[(1 - [*) where &,= & and Ji is the Jacobian factor with respect to nodal coordinates. 3.4 Formulation of displacements and strains The generalised displacements in the local coordinate system are 8 = NS = [N1N2N&2, 6 (22)

The exact integration of eqn (27) may be achieved by using a three-point Gauss quadrature for the axial and bending contributions and a six-point Gauss quadrature for the torsional and distortional contributions. Since shear deformation has been included in the formulation for which rotations due to bending are interpreted as shear strains, an excess of shear strain energy is stored by the element. This problem can be overcome by using a reduced integration scheme. Thus, the two-point integration procedure, which exactly integrates the bending contribution, tut underintegrates the shear contribution is used with the six-point integration. The nodal force vector equivalent to internal and external forces is written as

(28) where: b is body force vector; q is distributed force vector including patch loads; P is concentrated force vector; l0 is initial strain vector including temperature effects; and u0 is initial stress vector. 3.6 Boundary conditions The application of boundary conditions is generally self-evident. For some conventional support conditions used in bridge construction the following holds true:
0 0 0 0 0 0 0

where 6 are the nodal values of the global displacements and Nii . e, Nii.eY Nii. e, 0 0 0 0 0 0 Nij . e, Nij.e, Nij . e, 0 0 0 0 0 0 Nik . e, Nikse, Nik. e, 0 0 0 0 0 0
0 0 0

Ni =

Nili *e, Nii . eY Nii. e, aNi, -iiee, ax


0 0

Nii *e, Nij *e, Nij . e,


axJ.e 0 0
aNi,.

0 0 Nilk.e, Nik.e, Nik.e,

0 0 Ni, 0 0

0 0 0 0 0
0

0 0 0 0 0 0 Ni2 aNi2

(23)

%k.e,
0

!!!!$
0 0

a2Nil -~

Ni, ax

ax

_I
where i . e,, j . e, . . . ., k 9e, are the direction cosines. With the displacements known within the element the strains are obtained from E=LS (24) (a) If the beam is fixed at the support, no deformation arises in the support cross-section. and therefore the following may be written u=n=w=O U=V=W=O 0, = ey = 8, = 0I I$ & = 9, = 4, = 0 0, = yd = yd = 0. (29)

where L is a linear operator and can be written directly ._.. ~. from the detinition of the generahsed strain vector.

1040

S. H. ZHANC L. P. R. LYONS and in which L is the bending moment of inertia with respect to the entire cross-section. Unfortunately, the primary bending moment M, required for the distortional analysis is generally not known in advance. Thus, following the primary analysis, in which M, is determined, an iterative procedure is necessary to compute the secondary distortional forces md.R. Secondly, in the case of a curved box beam, the interaction of structural effects caused by the deformation of the cross-section can be accounted for within a second iteration. During this analytical stage the angular rate of distortion obtained, can be treated as an initial bending strain, (yd/R). This numerical technique implies the effect of distortion is to reduce the effective rigidity of the cross-section. Normally this effect is not significant in practical curved bridges and can then be ignored. 4. INTERNAL STRESS SYSTEM The normal stresses at the points on the mid-line of the flanges can be obtained by the following expression,

(b) If the support cross-section is connected by a pinned support, braced by a rigid diaphragm, and yet is free to warp, then the independent boundary conditions are

0, =

yd =

(30)

(c) If the cross-section is supported by a linear roller which is orientated perpendicularly to the longitudinal beam axis, and also is braced by a rigid diaphragm, but is free to warp, the boundary conditions may be adjusted to r=w=O 8, = Yd= 0 (31)

(d) If an interior diaphragm is introduced to resist transverse deformation only, thus inhibiting distortional behaviour at that location, then -yd= 0. (32)

From the above conditions, it is obvious in some cases the imposed restrictions at the boundaries are in the local xyz directions. Thus, the necessary transformation of the structural stiffness equations should be made to deal with the boundary conditions. 3.1 Interaction between bending, torsion and distortion for curved box beams Although in the finite element formulation the transverse distortional effect is treated as an equivalent straight box beam with a span equal to the developed length of the axis curve, interacting effects between bending, torsion and distortion need to be taken into consideration in the case of a curved box beam. Firstly, it can be seen from the differential eqn (4) that additional distortional forces occur in a curved box beam due to the radial components, uR. of the longitudinal bending stresses. The system of radial forces can be replaced by a force acting through the shear centre along the local z-direction and a twisting moment. The additional distortional moment per unit length can therefore be expressed as m,,,=pM, R (33)

Nx M. Mz ax=-tfz-~-vt--_rt-co,, A 1, L

B, J,

B,, Jrr

(37)

where o, and orr are the torsional warping and the distortional warping functions respectively. The factor 5 is expressed as 5=4 $ +(3&i-l) ( ,)

7-+@)21
(38a)

for parts between webs, or i=

0 *+Np[l_(p)]
;
c

<

for side cantilevers

(38b)

in which p is the distortional factor due to initial curvature and can be calculated by

where &..i is effective breadth ratio between neighbouring webs; $e,C is effective breadth ratio for side cantilevers: bi is width between neighbouring webs; b, is width of side cantilevers; b, is width from the mid-point of the individual flange plates or the extreme edge of side cantilevers to the point considered. The total shear flow is calculated as

~=++j-

f A

(34)

+++q:-Sf)+$%j:,-S,,) I II

(39)

where b, and b, are the widths of the top and bottom flanges between the side webs respectively. Since I, is usually obtained with reference to an effective flange breadth, eqn (34) can be approximated as (35) where 4 is designated as the shear lag factor and is expressed as (36)

where 4: y, qt. z are statically indeterminate shear flows due to unit shear forces ((Q,/I,)= l,(QJI,)= 1); qg is Bredts statically indeterminate unit shear flow; c$, c?& are torsional and distortional statically indeterminate unit shear flows respectively; si,.y, si,,I are statical moments of area about the centroidal y and z axes respectively; S,, S,, are torsional and distortional statical moments of area respectively. The transverse bending moment of individual component plates due to distortional can be obtained by mdh = vd . Adh ew

A thin-walled box beam finite element for curved bridge analysis

1041

where r&, are the influence values obtained from a Vierendeel frame analysis. Furthermore, the longitudinal bending moment of the plate per unit width can be approximated by multiplying the corresponding transverse bending moment per unit length by Poissons ratio as mlh =
Vmdb

satility and reliability of the thin-walled box beam element, three example problems are given below. Two span straight box beam model solution A two span continuous box beam model subjected to two-eccentric point loads was analysed. The span of each box beam is 3.5 m and the double-cell cross-section is shown in Fig. 4. The modulus of elasticity is 29 kN/mm, and Poissons ratio is 0.18. Sixteen thin-walled box beam elements were used for the present analysis, while 188 flat thin shell box elements [14] were also used to carry out a three-dimensional finite element analysis for comparison. Figures 5 and 6 show the distribution of longitudinal stresses and transverse bending moments at midspan section respectively, in which the values in parentheses are obtained from shell element analysis. Close agreement between the results obtained from the box beam elements and from the shell elements is evident.
Curved box beam model solution

(41)

5. CALCULATION SECTIONAL OF PROPERTIES Although the finite element approach presented in the preceding paragraphs is straightforward, difficulties can arise when calculating the necessary sectional properties required for the box beam element and for the stress calculation. A special program called BOXFRM has been written as a supplement to the main analysis program to resolve this problem [14]. The program applies to box sections with at least one vertical axis of symmetry, and with not more than 4 cells, since above this number the structure is likely to behave as a plate rather than as a beam. The discrete box section is ideahsed, using prismatic beam elements, as a transverse Vierendeel frame with unit length along the beam axis. The analytical scheme is also based on the Timoshenko beam theory and on the thin-walled beam theory, which can be found in detail in Ref. [22]. In addition to the sectional properties required for box beam element data input, the program provides the shear flow functions, the warping functions, the statical moments of area and the influence quantities due to distortion. For multicell box sections, frame analyses are carried out for each different distortional mode and the influence values are obtained by superposition (see Fig. 1).
6. SAMPLE PROBLEMS

The entire theoretical scheme has been incorporated into a finite element stress analysis system LUSAS [14], written in FORTRAN. In order to demonstrate the ver-

A curved box beam model experiment was conducted for the purposes of verification [22]. The radius of curvature to the centre line of the model was 3000mm providing an R/L ratio of 2.0. The model was fabricated from mild steel plate having a thickness of 3.46 mm (Fig. 7). The curved box beam model was welded at one end to a 610 x 610 x 20 mm steel plate which was in turn connected by bolts to a stiff reaction frame. The experimental values of Youngs modulus and Poissons ratio were 2.0 X lo6 kg/cm* (196.2kN/mm*) and 0.27 respectively. A point load of 800 kg (7.848kN) was applied at the tip over the outer web. Seventeen elements including one straight element to represent the end diaphragm were used. The vertical deflections across the top flange at the 8 arc length cross-section from the fixed end, is shown in Fig. 8. The comparison with the experimental results shows close agreement. The longitudinal membrane stresses at the :

E = 29 kN/mm*
U = 0.18
P P

P = 10 kN

(a) J-I

T,ongitudinal

elevation

690 (b) Cross-section

Fig. 4. Two span continuousbox beam model (in millimetres).

1042
-362.531-344.62)

S. H. ZHANGand L. P. R. LYONS -275.67(-257.31) -188.80(-196.77)

362.53C325.56)

275.67(258.60)

188.8OC201.21)

Fig. 5. Comparison of longitudinal stresses at midspan section (N-cm/cm. 1cm = IOmm).

3.-i

Fig. 6. ~istributioo of transverse bending moments at midspan section (N-cm/cm. 1cm = 10mm)

arc length cross-section, and transverse bending stresses for the outer surfaces at the i arc length cross-section have also been plotted in Figs. 9 and 10 respectively. They are in reasonable agreement with those obtained experimentally. Singfespan curved box beam bridge solution Finally, the behaviour of a single span curved box bridge subjected to a truck loading is presented. The loading and geometry of the box beam under study is shown in Fig. 11. Seven internal diaphragms, located at i of the span length, have been introduced, as shown in Fig. II(d). The same bridge was originally analysed by

Heins and Oleinik [IO] using a finite difference method to obtain the solution from eqns (l)-(4). For ~omp~ative purposes the eccentric loads were replaced by a combined vertical and torsional loading, as shown in Fig. II(b) to be consistent with Ref. [lo]. For the same reason, only the primary modulus of elasticity E was used throughout. The results obtained by the present finite element procedure are illustrated in Fig. 12-K Figure 12 shows the bending moments M,, Fig. 13 the distortional angle yd, Fig. 14 the torsional warping stress oIj all with respect to the span length. It is most interesting to note that the results produced by the present study are iden-

800

kg

---I-

1
c
Ln

Fig. 7. Curved box beam model(in millimetres, 1kg = 9.81 N).

A thin-walled box beam finite element for curved bridge analysis 800 kg u

1043

R = 3150 mm Box beam solution

-----

Experimental result

Fig. 8. Transverse distribution of vertical detlections at g arc length section from fixed end (1 kg = 9.81 W.
1000

(kg/cm')

500

0 i

-600

~~
~

-1200t (kg,/cm') R = 3150 mm b Fig. 9. Transverse distribution of longitudinal membrane stresses at i arc length section from fixed end (1 kg/cm*= 0.0981Nlmm3.

lOOO-

-2ooo-

_--)--

Box beam solution Experimental result

Fig. 10. Transverse bending stresses at outer surface at i arc length section from fixed end. CAS Vol. No. 18. 6-G

S.H. ZHANGandL. P. R. LYONS

.169.8"4p

P = 4K
(al

i 16g;,[,

lil

R = 100'

R t-

t-

R
E =

= l-$jSection A-A

(b)

0.947" A
u=

(equiv. steel)

Cd)

Diaphragm locations (node spacing n = 3.03") Fig. Example II. problem given by C. P. Heins [lo]

FE

beam solution shown identical to theoretical solution by Heins [lOI

x H.0

.c,
6.0 u 2 4.0 B y m

11 c 2.0
0

d s m

I
0.25

1
0.50

I
0.75 1.0 X L

Fraction of span length


Fig. 12. Moment diagram

A thin-walled box beam finite element for curved bridge analysis


FE beam solution shown identical to theoretical solution by Heins [lOI

1045

5.0iD , 0 x 4.03.02.0-

Fraction

of span

length

Fig. 13. Angular distortion.

6-

5-

FE beam solution shown identical to theoretical solution by Heins 1101

4_

3-

2/ l-

I
0.25 Fraction

b
length load

I
0.75 1.0x L

0.50 of span

Fig. Torsional 14. warpingstresses.

0.6

v Location 1 Location

of a diaphragn! of a wheel

;1 0.4Y

.-! o-. 2 .z1 .U ti-0.2P 0 -0.4L 0

FE beam solution shown identical to theoretical solution by Heins I 0.56 Fraction of span

[IO] I 1.0 Y L

length

Fig. 15. Distortional warping stresses.

IO46

S. H. ZHANG L. P. R. LYONS and 4. R. Dabrowski, Curved Thin-Walled Girders. Theory und Annlysis. Springer. New York (1968). 5. M. M. Ettouney and J. B. Kirby, Warping restraint in threedimensional frames. J. Stntct. Diu.. A.S.C.E. 107 (ST8). 1643-1656 (1981). 6. G. A. Gunnlangsson and P. T. Pedersen, A finite element formulation for beams with thin-walled cross-sections. J. Comput. Structures 15, 691-699(1982). 7. E. C. Hambly and E. Pennells, Grillage analysis applied to cellular bridge decks. The Structural Engineer Vol. 53, pp. 367-375,London (1975). 8. C. P. Heins and K. R. Spates, Behaviour of single horizontally curved girder. .I. Struct. Div., A.S.C.E. 96 (ST7). 151l1524(1970). 9. C. P. Heins and L. C. Bell, Curved girder bridge analysis. J. Comput. Structures 2, 785-797(1972). IO. C. P. Heins and J. C. Oleinik, Curved box beam bridge analysis. .I. Comput. Structures 6, 65-73 (1976). I I. J. Jirousek. A. Bouberguig and A. Saygun. A macro-element analysis of prestressed curved box-girder bridges. J. Comput. Structures 10.467-482 (1979). 12. J. L. Krahula, Analysis of bent and twisted bars using the finite element method. AIAA J. 5 (6) 1194-1197 (1967). 13. D. Krajinovic. A consistent discrete elements technique for thin-walled assemblages. ht. J. Solids Structures 5. 639-662 (1969). 14. LUSAS-Finite Element Stress Analysis System, Users Manual. Finite Element Analysis Ltd., London (1980). 15. L. P. R. Lyons, A general finite element system with special reference to the analysis of cellular structures. Ph.D. Thesis, Imperial College of Science and Technology, London (1977). 16. B. I. Maisel, Analysis of concrete box beams using smallcomputer capacity. Development Rep. 5 (publication 44.005). Cement and Concrete Association. London (1982). 17. M. J. Mikkola and J. Paavola. Finite element analysis of box girders. J. Struct. Div.. A.S.C.E. 106,(ST6). 1343-1357 (1980). 18. K. R. Moffatt and P. J. Dowling, Shear lag in steel box girder bridges. The Structural Engineer, London, 53.439-448 (1975). 19. V. Z. Vlasov, Thin-walled efnstic beams. 2nd Edn. National Science Foundation, Washington, D. C. (1961). 20. C. H. Yoo. D. R. Evick and C. P. Heins, Non-prismatic curved girder analysis. J. Comput. Structures 4. 675-698 (1974). 21. C. H. Yoo. Bimoment contribution to stability of thin-walled assemblages. .I. Comput. Structures 11. 465-471(1980). 22. S. H. Zhang, The finite element analysis of thin-walled box spine-beam bridges. Ph.D. Thesis. The City University, London (1982).

tical to the theoretical differential equation solution given by Heins et al. [lo], which can be regarded as an exact solution. 7. CONCLUSIONS box beam element, which has nine degrees-of-freedom at each beam node, has been

A thin-walled

developed to meet the practical requirements arising during the design of single and multi-span curved box beam bridges. The performance of the box beam element has been compared favourably with the results of other methods and with the results of model experiments. The thin-walled box beam element has been incorporated into the LUSAS finite element system. Data preparation in conjunction with the use of the supplementary program BOXFRM is straightforward. The computing time required and output obtained are substantially reduced from three-dimensional finite element solutions. Since the necessary structural actions required for box beam analysis have been taken account of, deformations and stresses determined along the curved box beam bridge, facilitate a physical understanding of the general structural response. The use of the element is particularly advantageous during the preliminary design stage of bridge decks, or in situations where a full three-dimensional finite element analysis is unnecessary. The present approach can also be extended to the analysis of multibox beams by using a finite element-grillage procedure [22].
Acknowledgements-The authors would like to thank Dr. L. F. Boswell of City University for his valuable advice, Dr. D. J. Irving of Finite Element Analysis Ltd and Mr. M. Smith of Kingston Polytechnic for their help on the implementation of the program. RF.FERENCI?S 1. R. S. Barsoum and R. H. Gallagher, Finite element analysis of torsional and torsional-flexural stability problems. Jnt. J. Num. Meth. Engng 2, 335-352(1970). 2. Z. P. Baiant and M. El Nimeiri, Stiffness method for curved box girders at initial stress. J. Struct. Div. A.S.C.E. 100 (STIO),2071-2090(1974). 3. L. C. Bell and C. P. Heins. Analysis of curved girder bridges. J. Struct. Div.. A.S.C.E., 96 (ST8), 1657-1673 (1970).

Vous aimerez peut-être aussi