Sei sulla pagina 1di 8

FLEXURAL TORSIONAL BUCKLING OF THREE DIMENSIONAL THIN WALLED ELASTIC BEAMS

Katy Saad1, Bernard Espion2 (Member ASCE) and Guy Warze3


ABSTRACT This paper presents an advanced finite element formulation of the torsional behavior of thin walled beams with arbitrary asymmetrical open or closed cross sections. The warping includes two terms to describe the warping along the midline (global warping) and across the thickness (local warping) of the thin walls and is able to model the effects of non-uniform torsion. An updated corotational Lagrangian formulation is presented in order to study the flexural-torsional buckling problem. To solve the nonlinear problem, an incremental-iterative technique combines the Newton Raphson method with the constant arc length of incremental displacements. Numerical examples are presented to show the performance, accuracy and efficiency of the proposed warping function for Eulerian stability of thin walled beams and columns. Keywords: Thin-walled beams, Finite element method, Flexural-Torsional Buckling.

INTRODUCTION The carrying capacity of thin-walled beams and columns is often governed by instability. Thin walled beams may fail in a flexural or/and torsional buckling mode: the beam suddenly deflects laterally or twists out of the plane of loading. In beam geometrical nonlinearities, strains remain small and the large movements are mostly due to rigid body motions. In this paper, an updated corotational lagrangian formulation is thus used. The reference is the last known configuration. The corotational formulation eliminates the rigid body rotations from the incremental solution. High order terms of nodal parameters in the element beam model are neglected and rigid body motions are separated from local deformations. The large displacement/rotation problem is solved step by step with moderate rotations. Results presented in this paper are limited to Eulerian stability analysis. The criterion to determine the buckling state is the singularity of the tangent stiffness matrix of the structure. Geometric buckling is analyzed for thin walled beams with asymmetrical cross sections, where centrod and shear center do not coincide. The study, deriving from Proki work (Proki , uses a single warping function which is valid for either open or closed cross sections.
1

Department of Continuum Mechanics, University of Brussels (ULB), C.P. 194/5, Av. F. D. Roosevelt 50, 1050 Brussels. Email : ksaade@smc.ulb.ac.be 2 Department of Civil Engineering, University of Brussels (ULB). Email : bespion@ulb.ac.be 3 Department of Continuum Mechanics, University of Brussels (ULB). Email : gwarzee@smc.ulb.ac.be

('p5f d0"0SE'0)bi7)7'hADC7"gBAf 0e@X19' d5 8&S7"c  760&b900)(5('4a' H&3#($"2 `"Y 1&6 0)& ('X'W 7V" &G HU%$BA#" !S  %IT8T8I %S  RQPI H'" GF yw u Pxwvtsr q

KINEMATIC MODEL The warping displacement is described by two terms. The first term describes the local warping and depends on the derivative of the torsional rotation angle x,x , on the distance to the shear center hn and on the distance to the midline e (Fig.1).

The second term describes the global warping and is measured by a linear combination of displacement parameters at selected nodes (ui) of the cross section.

z s q hn

e C y G

Figure 1. Cross-section of a thin walled beam The functions i describe the mode of deformation between adjacent transversal nodes of the cross section. The displacements at a point q are:

x,x

i 1


vq wq

0 0

0 w

v 0

( z  z C )T x

( y  y C )T x

The additional parameters (ui) are only related to the out of plane displacement due to the torsional behavior. The cross section remains plane when subjected to normal force or bending. The displacements (ui) are not independent; they cannot take arbitrary values. They represent the warping but they must induce no global elongation of the cross section neither global rotation of the section along flexural axis y and z. So, from the 6+n degrees of freedom, only 6+n-3 are

uq

u0

 ZT

u glob warping

i 1

u loc warping

x,x

where

(y,z)=hn(s).e

( y, z ) u i ( x )

(x,y)=1

( y, z)u i ( x )

(1)

(2)

(3)

independent: three translations (u,v,w), three rotations ( x, y, z) and n-3 relative longitudinal displacements, where n is the number of transversal nodes of a cross section. FINITE ELEMENT MODEL The beam finite element has three longitudinal nodes (Fig. 2). For each end node, there are 6+n degrees of freedom (u,v,w, x, y, z,ui,). For the central node, 5 degrees of freedom (v,w, x, y, z) exist. The beam displacements are expressed in terms of the nodal displacements by using two functions. The transverse displacements (v,w) and the rotations ( x, y, z) are interpolated by a quadratic function N.

For the longitudinal displacements (u0, ui), a linear interpolation function N is used.
 ! '

z,w
Tz Ty

y,v

O,C u 01 T x1
x2
T y2 T z2

u 03

v1 w1

T y1 T z1

v2 w2

v3

T y3

u11u 21u n1
Fig 2. Finite element model

w 3 T z3 u13 u23 u n3

BUCKLING Using the updated lagrangian formulation, the incremental virtual work is given by:
kl ij ij d

where the volume V corresponds to the last known configuration which is taken as reference. dijkl

e *ij )dV
(

0 1

3 42

(d ijkl d

u i df i dV

&

% $ "#

x3

avec

 

   

1 3 2 2 4 (1 ) (1 2 )

x l

Tx

     

(4)

(5)

x,u

(6)

is an element of the Hooke matrix; ij is the linear part of the Green-Lagrange strain tensor; e*ij is the non linear part of it. ij is the Cauchy stress tensor; fi is the external force vector. The equation (6) can be developed and transformed into a matrix formulation: [KT] {q} = {F} where [KT] =[KL ] + [KNL]

(7)

An incremental iterative method is used to solve the nonlinear system of equations (7). It assumes that the solution is known at the initial discrete step (t), and iterations are performed to calculate the (t+1) equilibrium configuration by considering the equilibrium between the exterior load forces and the nodal interior forces (equivalent to stresses in the element). Critical loads are calculated by taking into consideration that the structure, already in equilibrium, reaches instability if there is more than one equilibrium position for the same loading level. The criterion to determine this buckling state is the singularity of the tangent stiffness matrix [KT] of the structure. EXAMPLES The above-described finite element model has been implemented into a software. In the following examples, critical loads for columns, beams and frames are compared with analytical values obtained with other kinematic formulations. Example 1: Plane frame buckling The first example illustrates the plane flexural buckling of a portal frame. The columns and the beam of the frame are identical and the closed cross section is given below (Fig. 3). E=210GPa, G=80GPa. The difference between the finite element results and the solution given by Timoshenko (1961) is shown in Fig.3. The numerical value of the first critical load given by the software converges to the value given the analytical solution of Timoshenko (Pcr=652N).

100.0%
P

Difference with reference solution for the first critical load

P t=0.001m

10.0%
0.04m

5m 5m

1.0% 0.1%

0.04m

10 Number of elements

100

Fig.3 Stability analysis of a frame consisting of members with closed cross section Example 2: Flexural torsional buckling of a column A column with open monosymmetric cross section is submitted to an axial load (Fig.4),

L=20m, E=210GPa, G=80GPa. The non coincidence of the centrod(G) and the shear center(C) leads to a coupling between the flexural and the torsional buckling. P tfs=0.02m tw=0.01m 0.2m
L

0.1m

Difference with reference solution for the first critical load

100.0% 10.0% 1.0% 0.1% 1


0.25m

G Tfi=0.014m C

10 Number of elements

100

Fig. 4 Stability analysis of a column with open monosymmetric cross section Two analytical studies are used to compare the results of the software. The first is based on Vlassovs theory and the second is based on the warping function described in this paper. The buckling non linear equations are developed from a state of combined torsion, bending and axial compression. They are obtained from general equilibrium equations written for the deformed beam or column. The solution is given by taking into consideration the boundary conditions. The difference between the numerical value of the first critical load obtained by the software and the reference value based on Vlassovs theory (Pcr=102223N) is illustrated in Fig.4. The analytical value based on the above described warping function converge to this value with a difference of less than 0.0001%.

Critical loads [N] 0 Number of elements 5 10 15 20 25 30 0.E+00 2.E+06 4.E+06 6.E+06 8.E+06 1.E+07

Fig. 5 Critical loads for the centrally loaded column (Fig. 4)

In a general flexural torsional buckling of a beam-column, the (xy) bending modes, the (xz) bending modes and the twisting modes are coupled. The first-order theory gives three homogeneous equations and represents an eigenvalue problem. When, the shear center (C) and the centrod (G) coincide, the equations are uncoupled and the solution gives a discrete set of buckling modes. The lowest critical load is, in general, of practical significance. When C and G do not coincide, buckling involves simultaneously torsion and bending, and the critical load is lower than if torsional effects are prevented. Numerically, when the number of finite elements increases, the number of detected critical loads increases. The number of detected buckling modes increases because it depends on the discretization and on the interpolation functions. Fig. 5 shows the critical values obtained for the same example (fig. 4) up to 10MN. The squares in fig. 5 represented along the horizontal axis are the reference values of the buckling loads (based on Vlassovs theory). The other sets of values corresponds to buckling loads detected by finite element analyses with increasing number of elements (1, 2, 5, 10, 16 and 20 elements). For one element, there are only three critical values when the applied load P increases from zero to 10MN. For two elements, the numerical values of buckling loads are improved and other critical loads appear and so on... So, each critical value converges to the reference solution with the increasing number of elements. Example 3: Lateral torsional beam buckling An I beam is loaded by two couples at its ends and is therefore subjected to uniform bending (Fig.6) L=20m, E=300GPa, G=99.5GPa. The critical value corresponding to the lateral torsional buckling is Mcr=6262.26Nm. Fig. 6 shows how the numerical solution converges to the reference analytical solution (Timoshenko, 1961). 0.08m

Difference with reference solution for the first critical load 100% 10%

tf=0.01m 0.38m tw=0.00375m

1% 0% 1
0.08m

10 Number of elements

100

Fig. 6 Lateral torsional buckling of an I beam Example 4: Torsional buckling of a column The last example illustrates a case of pure torsional buckling. A column with a cruciform section submitted to an axial load is considered (Fig. 7). The thickness of the walls is t=4mm. L=1m, G=80.8GPa, E=210GPa. According to Vlassovs theory, this kind of cross section does not warp. The theoretical torsional Eulerian buckling load is 258 398N. To initiate the torsional buckling of the column, a perturbation is needed in the finite element analysis; this is introduced by applying a small torsional moment Mx at mid height of the column. Fig. 7 gives the

relationship between the axial load and the angle of twist at mid height for increasing values of P and Mx. The horizontal line is the critical load 258398N. The curves represent the geometrical non-linear variation of the angle of twist at mid height for different values of the ratio Mx/P: Mx/P=1.5 10-7 m for the curve , 5 10-7 for , 1.5 10-6 for , 1.5 10-5 for and 1.5 10-4 for . The load-angle of twist relationship are obviously influenced by the magnitude of the applied torsional perturbation, but all curves reach asymptotically the level of the elastic buckling load. P
8cm 8cm

8cm

8cm

3.E+5

2.E+5 P[N]

1.E+5

0.E+0 0.000 0.001 0.002 0.003

0.004 [rad]

0.005

0.006

0.007

Fig. 7 Torsional buckling of a column with cruciform cross section SUMMARY AND CONCLUSIONS A finite element formulation has been used to study the flexural torsional buckling of elastic structures composed of arbitrary thin walled cross sections. Some numerical examples for open and closed sections are presented. The numerical results are in excellent agreement with existing analytical solutions. This model, which can be easily applied to an assembly of beams and columns with all kind of profiles, is currently used to study the flexural torsional buckling of structures combining beams with different cross sections (open or/and closed).

REFERENCES Criesfield M. A. (1997), "Non-linear Finite Element Analysis of Solids and Structures", Wiley, Chichester, England. De Ville De Goyet V. (1989), "L'analyse statique non linaire par la mthode des lments finis des poutres section non symtrique", Ph.D. thesis, Universit de Lige, Belgium (in French). Murray N. (1986), "Introduction to the theory of thin-walled structures", Oxford University Press, New York, USA. Proki A. (1990), "Thin walled beams with open and closed cross section", Ph.D. thesis, Univ. of Belgrade, Yugoslavia (in Serbian). Shakourzadeh H., Guo Y.Ch. and Batoz J. L. (1996), "On the large displacement and instability of 3D elasto-plastic thin walled beam structure", Wiley, Eccomas 1996, Volume 2. Timoshenko S., Gere J. (1961), Theory of elastic stability, McGraw-Hill, New York, USA. Trahair N. S. (1993), Flexural-Torsional Buckling of Structures, E & FN SPON, London. Bleich F. (1952), Buckling strength of metal structures , McGraw-Hill, New York, USA.

Potrebbero piacerti anche