See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/245284452

Stability of Beams with Tapered I-Sections

Article  in  Journal of Engineering Mechanics · September 1987

DOI: 10.1061/(ASCE)0733-9399(1987)113:9(1337)

CITATIONS READS

53 580

2 authors, including:

Y.B. Yang
Chongqing University
344 PUBLICATIONS   7,637 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Vehicle-Bridge Interaction View project

vehicle scanning method View project

All content following this page was uploaded by Y.B. Yang on 24 August 2015.

The user has requested enhancement of the downloaded file.

 STABILITY OF BEAMS WITH TAPERED I-SECTIONS

By Yeong-Bin Yang, 1 A. M. ASCE, and Jong-Dar Yau2

ABSTRACT: The purposes of the paper are: (1) To derive differential equations
of equilibrium for a tapered I-beam; and (2) to formulate a finite element for
Downloaded from ascelibrary.org by National Taiwan University on 09/29/13. Copyright ASCE. For personal use only; all rights reserved.

the beam that takes into account the effect of nonuniform torsion. In the virtual
work formulation, the updated Lagrangian approach is adopted, in which the
effect of geometric nonlinearity is considered. The present formulation requires
obtaining a rigorous expression for the strains based on the membrane theory
of shells, through which the effect of tapering is considered. The displacements
of each cross section are determined with Vlasov's thin-walled beam assump-
tions. The derived finite element model, in terms of the linear and geometric
stiffness matrices, is useful in a buckling or an incremental large displacement
analysis. Using the present theory, one is able to investigate various torsional-
flexural instability problems. Examples are prepared and comparisons are made
with existing solutions.

INTRODUCTION

A thin-walled bar is characterized b y the p h e n o m e n o n of nonuniform

torsion along its axis. The bar may buckle by flexure, torsion, or a com-
bination of flexure and torsion, d e p e n d i n g u p o n the loading a n d b o u n d -
ary conditions. The behavior of uniform thin-walled bars (with constant
sections) in bending and torsion has been well k n o w n since the early
works of Vlasov (1961). Unfortunately, the corresponding behavior of
nonuniform thin-walled bars is not as easily determined, and only a lim-
ited n u m b e r of investigations have been m a d e .
Cywinski (1964) used the variational method to study the linear be-
havior of monosymmetric thin-walled bars with nonuniform cross sec-
tions. The behavior of monosymmetric members with variable profiles,
either open or closed, w a s analyzed by Bazant (1965), using a n approx-
imate expression for the longitudinal strains. The torsional response of
tapered I-girders was studied by Lee and Szabo (1967) along with the
stresses calculated. Based on a rigorous elasticity formulation, Wilde (1968)
derived a system of linear differential equations for the b e n d i n g a n d tor-
sion of nonuniform thin-walled members. Of interest are the works by
Kitipornchai a n d Trahair (1972, 1975) on the stability of tapered m o n o -
and bisymmetric I-beams, in which the m e t h o d of finite integrals is
used. Since Kitipornchai a n d Trahair's formulations are based on the
equilibrium considerations, clear physical meanings can be appreciated
for the bending and torsion. Starting from the energy considerations,
critical loads for the torsional-flexural buckling of tapered I-beams w a s
determined b y Brown (1981) using the finite difference scheme. H o w -
ever, Brown's solutions were approximate because the effect of tapering
^ssoc. Prof., Dept. of Civ. Engrg., National Taiwan Univ., Taipei, Taiwan,
Republic of China.
2
Grad. Res. Asst., Dept. of Civ. Engrg., National Taiwan Univ., Taipei, Tai-
wan, Republic of China.
Note.—Discussion open until February 1, 1988. To extend the closing date one
month, a written request must be filed with the ASCE Manager of Journals. The
manuscript for this paper was submitted for review and possible publication on
May 16, 1986. This paper is part of the Journal of Engineering Mechanics, Vol.
113, No. 9, September, 1987. ©ASCE, ISSN 0733-9399/87/0009-1337/$01.00. Pa-
per No. 21798.

1337

J. Eng. Mech. 1987.113:1337-1357.

 was not completely taken into account in his expression for the strains.
Recently, the instability of tapered thin-walled bars was analyzed by
Wekezer (1985) using the finite element method, in which a total La-
grangian formulation was adopted.
In this paper, an attempt is made to study the instability problems of
tapered I-beams. Essential to the analysis of tapered beams is the deri-
Downloaded from ascelibrary.org by National Taiwan University on 09/29/13. Copyright ASCE. For personal use only; all rights reserved.

vation of various strain components based on the membrane theory of

shells and the expression of cross-sectional displacements based on Vla-
sov's thin-walled beam assumptions. Using an updated Lagrangian for-
mulation, the incremental equations of equilibrium are derived from a
fundamental virtual work equation. Solutions of the instability problems
are obtained by the finite element method through the establishment of
linear and geometric stiffness matrices. The present work can be re-
garded as an extension of previous efforts (Yang and McGuire 1986a,b)
on the geometric nonlinear analysis of three-dimensional framed struc-
tures with uniform I-sections.

THEORETICAL ANALYSIS

Elements of the theory of thin-walled beams of variable I-sections will

first be summarized, which are to be invoked later in the variational
formulations.
Assumptions.—For the present purposes, a tapered thin-walled mem-
ber of an I-section will be considered [Figs. l(a-b)]. In analysis, the fol-
lowing assumptions are made: (1) The material is elastic and homoge-
neous; (2) the length of the beam is large compared with the cross-sectional
dimensions; (3) the section is thin-walled; (4) every cross section is rigid
in its own plane; (5) shearing deformation of the middle surface of the
member is negligible; and (6) transverse displacements are much larger
than the longitudinal displacement.
Displacements.—As shown in Figs. l(a-b), a right-handed coordinate
system x, y, and z is chosen such that x coincides with the centroidal
axis, and y and z coincide with the principal axes of the cross section.
The position vector of an arbitrary point, N, located on the midsurface
may be expressed as
r = xi + y) + zk (1)
where i, j , and k denote the unit vectors along the x-, y-, and z-axes,

t. ( ,
T^_t>—H

(a) (b)

FIG. 1.—-Tapered l-Beam (a) Coordinates; (h) Section

1338

J. Eng. Mech. 1987.113:1337-1357.

 respectively. For a tapered beam, both y a n d z are functions of x. There-
fore, the derivatives of r with respect to x, y, and z may be expressed
as follows:
t,x = i + y'\ + z'k (2a)
Downloaded from ascelibrary.org by National Taiwan University on 09/29/13. Copyright ASCE. For personal use only; all rights reserved.

'.y = J • • (2b)
t,z = k (2c)
where a prime denotes differentiation with respect to x, and a comma
denotes partial differentiation with respect to the variable following.
Likewise, the displacement vector of point N m a y be described as
u = uxi + uy j + w2k , (3)
in which ux = the longitudinal displacement; and uy and uz = the trans-
verse displacements. Conventionally, the transverse displacements are
obtained from the geometrical hypothesis of in-plane rigid cross sec-
tions:
uy = uyC - z%x (4a)

uz = uzC + yQx (4b)

if the angle of twist Qx is reasonably small, where uyC and uzC denote the
transverse displacements of the centroid, C, from the original position.
Strains.—Since the cross sections of a tapered beam vary along the x-
axis, the midsurface coordinates x a n d s are in general curvilinear (Fig.
1). According to the theory of elasticity (Green and Zerna 1968), the Green-
Lagrange strain tensor in its linear form is
1
emn = ~ {t,m • u,* + *,n • u, m ); m,n = x,y,z (5)

Substitution of Eqs. 2a-c a n d 4a-b into Eq. 5 yields

exx = uXiX + y'uytX + z'uz,x (6a)

exy = ~ (ux,y + z'uZiy + UyiX) (6b)

1
e„ = ~ (ux,z + y'UyiZ + uZrX) (6c)
eyy = eyz = ezz = 0 (6rf)
where it should be noted that wyy = uZiZ = 0. From the condition of neg-
ligible shearing strain on the middle surface, one may write

dy dz
exs = exy — + exz — = 0 (7)
ds ds
By substituting the expressions for e^ a n d exz into the preceding equa-
tion, one may solve for the longitudinal displacement ux as
ux = uxC - yu'yc - zuzC - we; (8)

1339

J. Eng. Mech. 1987.113:1337-1357.

 in which the sectorial area to is defined as

<o = J {-zdy + ydz) (9)

Substitution of the cross-sectional displacements (Eqs. Aa-b and 8) into
Downloaded from ascelibrary.org by National Taiwan University on 09/29/13. Copyright ASCE. For personal use only; all rights reserved.

Eqs. 6«-c yields

£xx = u'xC- yuyC - zu"c - we* - i|>e; (io«)
1 / du>\
6xy = z + ( }
~i\ Tr
l / aw\
6xz = (10c)
2V~ ~teH
where ty = w' + y'z — yz' (11)
The nonlinear components of the Green-Lagrange strain tensor may
be written as (Green and Zerna 1968)
1
ilmn = - (u m • u „); m,n = x,y,z (12a)

or T)„ = - (u2x,x + u2yiX + u2ZiX) (12b)

1
1\xy = - (Ux,xUx,y + UZiXUZiy) (12C)

1
•nxz = - {ux,xUx,z + uyiXuyiZ) (12d)

In practice, the term with u%x in the expression of r\xx is small and may
be neglected (Washizu 1975; Argyris, et al. 1979). By substituting the
expressions for ux, uy, and uz (Eqs. Aa-b and 8) in Eqs. Ylb-d and ne-
glecting the higher-order terms, one obtains
1
•n» = - [«£ + "zc - 2zuyC 9X + 2yuzCQx + (f + z2)^2] (13a)

1
%y = ~ [-{u'xc - zulc)uyC + MzcOx] (13b)

1
f\xz = - [-(U'xC - yUyc)u'zC ~ WyC%x] (13c)

for an I-beam with small angles of taper.

Stress Resultants.—The stress resultants on a cross section of the ta-
pered I-beam are defined as

Fx = ixdA (14a)
JA

1340

J. Eng. Mech. 1987.113:1337-1357.

 Fy ~ | CTxydA (14b)

Fz = I <j
Vx.xzdA (14c)
JA
JA
Downloaded from ascelibrary.org by National Taiwan University on 09/29/13. Copyright ASCE. For personal use only; all rights reserved.

Mx = I (vxzy - a^dA (14d)

JA

My•y == \I <r
v*rzdA (14c)
JA
JA

Mz = - | cr
&xxydA (14/)
JA

B = - I <rxwdA (14*)
JA

where A = cross-sectional area; cr* = normal stress; vxy, axz = shear

stresses; Fx = axial force; Fy,Fz = transverse shear forces; Mx = torque;
My, Mz = bending moments; and B = bimoment. The normal stress in
Eqs. 14a-g may be obtained from Hooke's law:
Ox = Eexx = E(u'xC - yu'yc - z«"c - o)6^ - «|<8i) (15)
where E = elastic constant. Substituting for <JX in Eqs. 14a and e-g, one
can express the generalized stresses Fx, My, Mz, and B in terms of the
generalized strains u'xC, u"yC, ulc, 9" , and 9* as

rp "ic
0 A sz
My • = E • U'yC (16)
Mz 9"
iB. 1— J
w hvz '•wy r i-w^i _
8i
in which the beam section properties are defined as

A = <M (17«)

yrfA (17b)

zrfA (17c)

yzdA (17d)
k=
1341

J. Eng. Mech. 1987.113:1337-1357.

 Sw= «>dA (17e)
JA

I*,, = aydA (17/)

JA
Downloaded from ascelibrary.org by National Taiwan University on 09/29/13. Copyright ASCE. For personal use only; all rights reserved.

Iwz = oizdA (17g)

JA

Iy =• z2dA (17ft)
JA

h = y2dA (17f)
JA

Cw = w2dA (17;)
JA

S* = ifdA (17k)
JA

I* = t\adA (17/)
JA

Ity = tyydA (17m)

JA

4+ = vtyclA (17«)
JA

With the adoption of the y- and z-axes as the principal centroidal co-
ordinates of the I-section and the definition of the sectorial area u in Eq.
9, it can be shown that the following conditions of orthogonality for the
principal generalized coordinates are satisfied:
—
^y ~ ^z lyz ~ ^w ~ >-xm) ~ iwz ~ 0 (1")

In addition, it may be shown that for a doubly symmetric section, as in

the present case, the following conditions are reached:
S* = I*, = I* = 0 (19)
As a result, the expressions for the stress resultants (Eq. 16) can be greatly
simplified as
Fx = EAu'xC , (20a)
My = -Elyiilc (20b)
Mz = EI2u;c (20c)
1342

J. Eng. Mech. 1987.113:1337-1357.

 B = ECWQX + E7^8.; (2CW)
As defined previously, Iy and Iz = the moments of inertia about the y-
and z-axes, respectively; Cw = the warping constant; and 1^ = a section
constant.
Updated Lagrangian Formulation.—In an updated Lagrangian for-
Downloaded from ascelibrary.org by National Taiwan University on 09/29/13. Copyright ASCE. For personal use only; all rights reserved.

mulation, the incremental virtual work equation of equilibrium may be

written for a member between its current (known) deformed state and
its neighboring (desired) state as (Yang and McGuire 1986a):

[Eexxbexx + IGe^he^ + 2Gexzbexz]dV

2 i
+ [vMxx
K 8 T ) „ + 2 ( ^^ 88 ^^ + 2axxMxz¥V
Mx,¥V = <R - RlR
R~ (21)
Jv
Jv
in which V = volume of the member in the current state; G = shear
modulus; 1RI 2R, = virtual work done by the external loads at the current
and the next deformed state; other quantities in Eq. 21 were defined
previously. In the Eq. 21, the stresses ax, cr w , axz denote the Cauchy
stresses in the current deformed state, and exx, e^, exz and T\XX , T)^, r\xz
denote the linear and nonlinear components of the Green-Lagrange strains
(Yang and McGuire 1986a).
For the present purposes, one may assume that there are only con-
centrated loads acting at the ends of a member and that the shear forces
FyA and FyB act at distances aA and aB above the centroids of the cross
sections, respectively. Using the notation ( ) for a row vector and { } for
a column vector, the external virtual work may be expressed as
2
R - >R = <8H> ({2/} - {"/}) .' (22)
for loadings of a conservative type. In Eq. 22, (8M) is a vector that con-
tains the variation of displacements at the nodes, where
(U) = (UxAUyAUzA ^xA^yA^zAUxBUyBUzBQxB%B^zB^'xA^xB) (23)

(Y) is the vector of nodal forces at the beginning of the step, as follows:
(V) = eFxA %A %A lMxA lMyA xMzA %B %B %B
'MxB 'My, 1MtB 'BA %) (24«)
2
and { f) is the vector of nodal forces at the end of the incremental step,
as follows:
<2/> = (2FxA %A 2FzA(2MxA + %AaAQxA) 2MyA 2MzA 2FxB %B
2
FzB(2MxB + %BaBQxB) 2MyB 2MzB 2BA 2BB) (24b)
2
Both Cf) and ( /) are stated in the coordinates of the current deformed
configuration.
The following steps may be performed on Eq. 21: (1) Substitute the
expressions for the strains shown in Eqs. 10 and 13; (2) neglect the higher-
order terms; (3) adopt the condition of Eqs. 18 and 19; and (4) make use
of the expressions for stress resultants (Eqs. 14a-g). One can then obtain
the following equation:
1343

J. Eng. Mech. 1987.113:1337-1357.

 [EAb(u'2) + EIyb(uf) + EIzb(uf) + £Q„8(ef)

+ EQ8(e; 2 ) + 2EI^8(ei9;) + G/8(6i2)]dx + - Fx%{u'2 + u'z2)dx

2 Jo
Downloaded from ascelibrary.org by National Taiwan University on 09/29/13. Copyright ASCE. For personal use only; all rights reserved.

+ -1 Km2)dx Myb(Uyl'%'x)dx- Mz§(uzWx)dx

Jo

Fyb(u'xu'y)dx + Fyb(uzQx)dx -
Jo Jo Jo

Fzb(uyQx)dx - (1 - a) Mxb(u'^uy)dx
o Jo

+ I aMxS(u'y u'z)dx = <8M)({ 2 /} - f/}) (25)

Jo

in which each term is integrated through the length, L, of the member;

and the subscripts C for displacements have been dropped for clarity.
In Eq. 25, / is the torsional constant (Love 1944) defined as
dco\ / do>
7 2 + + dA (26)
fly/ V dz
Q, is a new section property:

Q,= 4?dA. (27)

The term a is a torsional parameter defined as a ratio of the torque re-

sisted by (Txz to the total torque Mx:
f <rxzydA f (-a
(~oyxyz)dA
(28)
JA Mx 1 M,r
M
and K is the Wagner coefficient defined as

K= cr^y2 + z2)<M (29)

For a bisymmetrical section

K = Fxr2; r2 = ^ ^ (30)

using Eqs. 15 and 20. It should be noted that the stress resultants ap-
pearing within the integral signs of Eq. 25 are the forces acting at the
beginning of the incremental step, i.e., Fx = 1FX, Fy = 1Fy, etc.
Differential Equations of Equilibrium.—Integrating Eq. 25 by parts
to obtain the virtual displacements (hux, 8wy, 8M2, 89*) and admitting
1344

J. Eng. Mech. 1987.113:1337-1357.

 the arbitrary nature of virtual displacements, one obtains the differential
equations of equilibrium for the buckling of a tapered I-beam, as the
Euler-Lagrange equations of the functional, i.e.
(Eau'x)' - Fyu'y - Fzul = 0 (31A)
(EIzu'y)" - Fxul' + FZWX + Fyu'x + (MyQ'x)'
Downloaded from ascelibrary.org by National Taiwan University on 09/29/13. Copyright ASCE. For personal use only; all rights reserved.

+ [(1 - OL)MXUIY + (aMxu'z)" = 0 (31b)

{ElyUl)" - F X - FyWx + Fzu'x + (MX)'
- [(1 - a)Mxuy]" - (oMxu'y)' = 0 (31c)
(EQX + EIaX)" ~ (ECX + EIW^'X)' - (G/0i)'
- (KQ'xy + Myu; + Mzu'^ = 0 (31d)
where the relations Fz = M'y and Fy = -Mz have been used. As a by-
product of the variational procedure, a set of natural boundary condi-
tions can be obtained. Such expressions are very lengthy and are not
presented here.
Eqs. 31a-d state the equilibrium of a deformed tapered I-beam. For a
bisymmetric section, the present equations are more general than those
of Kitipornchai and Trahair (1975) and Brown (1981) in that the effect of
instability due to various initial loads have all been taken into account.
Using the present equations, one may investigate the buckling of a beam
when it is subjected to an axial compression, a shear force, a bending
moment, a torque, or a combination of any of the preceding parameters.
For the special case of a simply supported, tapered I-beam with a mid-
span load, the present equations coincide with those of Kitipornchai and
Trahair. On the other hand, Brown's work (1981) was considered only
approximate since the terms with C^ and 1^ have been omitted from his
expression. Such terms, as may be appreciated from the present study,
represent part of the effect of tapering.
For the special case of a uniform I-beam (C+ = Iwi/ = 0) with a = 1/2
(assuming semitangential torques), the differential equations of equilib-
rium reduce to those of Yang (1986) or the equations that may be derived
from Yang and McGuire (1986a) using standard variational procedures.
Beam Section Properties.—Before the element stiffness matrices can
be derived, the beam section properties, which are functions of coor-
dinate x, must be given. For the present purpose, the following formulas
are used:
A = 2btf+ htw (32a)

/ = ^ (bt3f + htl) (32b)

k = ~ b% (32c)

Iz = - btfh2 (32d)

1345

J. Eng. Mech. 1987.113:1337-1357.

 f y
,1

bh'
fc^
A ^ J ^
Downloaded from ascelibrary.org by National Taiwan University on 09/29/13. Copyright ASCE. For personal use only; all rights reserved.

z •*-

[^
^
£-J
^_j
© ' •••<$)

(a) (b)
FIG. 2.—Diagrams for co and i|>

, 4 +4
r 2 <= -1 (32e)

where tf = flange thickness; b = flange width; tw = web thickness; and

ft = distance between flange centroids (Fig, 1). Determination of the other
quantities, Cw, C^,, and Iw<f, relies upon the calculation of the functions
to and i(i for each section. The sectorial area o> may be obtained from Eq.
9. For the web of an I-section, 2 = 0 and dz = 0; therefore one obtains
a) = 0. For the flanges, dy = 0 and it follows that
» = yz •• •••• • '....•• (33)
On the other hand, the function i|i may be obtained from Eq. 11. For the
web, one has ij; = 0, because w = z = z' = 0. For the flanges, one obtains

+ = 2 - yz (34)

where it is noted that for flanges y - ±h/2. The diagrams for w and \\i
are given in Figs. 2(a-b). Utilizing the diagrams for w and v|/, one obtains
for Cm, Gj,, and /„,,,, from Eqs. 17 ;' and n and 27 the following formulas:
n2
Cw = -Iy....: (35fl)

/2«'\ 2
Q = I — J Cw (35b)

/2«'\
U = I — )CW (35c)

To enhance the finite element formulation that is to be presented later,

it is useful to express all the beam section properties explicitly as func-
tions of x or i (=x/L). Letting b = b0 and ft = h0 at x = 0, and b = bL and
h = hL at x = L, one may write
1346

J. Eng. Mech. 1987.113:1337-1357.

 b = b0qb , (36a)
h = h0qh (36b)
where qb = 1 + $bi (37a)
ft = 1 + P*iV (37b)
Downloaded from ascelibrary.org by National Taiwan University on 09/29/13. Copyright ASCE. For personal use only; all rights reserved.

bL - b0
with p;, = — (38a)

hL- h0
P* = — r — (38b)
"0

Assuming that both the flange thickness, tf, and web thickness, tw, are
constant, one may rewrite Eqs. 32a-e and 35a-c as follows:
A = A0qA (39a)
7 = 7oO; , . (39b)
ly = lyoqiy (39c)
Iz = hQqiz • (39d)
2
r = r\qr (39e)
Cw = Cw0qCw (39/)
Q, = C^qc • (39g)
Iu»li ~ Iwtyollu"!! (39«)

where A0, J0, 1^, 7z0, r\, Ca0, Q, 0 , and Iw¥) denote the beam section
properties that are evaluated at x = 0, i.e., those obtained by setting b
= b0 and h = h0 in Eqs. 32a-e and 35a-c. The qs functions in Eqs. 39a-
h are defined as
qA = 1 + cai (40a)
q] = 1 + Cji (40b)
qiy = 1 + cyli + cy2i2 + cy3i3 (40c)
2 3
<7b = 1 + cAi + cz2i + cai (40rf)
2
qr = 1 + crli + cr2i (40e)
2 3 5
lew = 1 + cwli + cw2i + cw3i + Caii* + ca5i (40/)
1c = % (40$)
2 3
ato = 1 + cni + cai + ci3i + c^i* (40/z)
where all coefficients are given as
1
C = — (Tbatffo + V«,P„) (41a)
A0
1347

J. Eng. Mech. 1987.113:1337-1357.

 Cj = — (2b0tfpb + h0t3wph) (41b)
4/o
cyi = 3p„ (41c)
<V» = 3pg (41d)
Downloaded from ascelibrary.org by National Taiwan University on 09/29/13. Copyright ASCE. For personal use only; all rights reserved.

cy3 = P? (41e)
c2l = ft, + 2p„ (41/)
cz2 = p2 + 2p»fc (41g)

Cz3 = PbP' (41/1)

Cn = — f- ^/P* + M/*opJ (41i)

en = - h b^/p2 + ^ V/koP*J (41;)

(^ = P„ + 2p„ (41*)
2
c«,2 = P + 3 p i + 6p,p A (41/)
2 2
cw3 = 3p b p + 6p p„ + $ (41m)
2 2
cai = 3p p + 2p£p„ (41M)
2
cw5 = p p? ..., (41o)
cn = p„ + 3p(, (41p)
2
ca = 3P(,ph + 3p (410
d = 3p„p£ + p? (41r)
Cn = P/,06 (41s)
where Ipo = fyo + 4o • In the calculation of crl and c r2 , it has been assumed
that pft = Pi,, which is only approximate.
Finite Element Formulation.—In general, the procedure of Yang and
McGuire (1986a) will be followed in the present formulation. The finite
element studied is shown in Fig. 3. A linear displacement field is chosen
for the axial displacement, ux, and a cubic field for other displacements,
i.e.
«* = (ni){ux) (42a)
"y = (n3){uy} (42b)

«z = <"3>{«z} (42c)

0, = («3>{ej (42d)
where, using /' = x/L
(ttj) = ((1 - 0 0 (43a)
2 3 2 3 2 ,3 3 2
(n3) = ((1 - 3/ + 2I )(I - 2z + i )(3i - 2i )(i - i )) (43b)

1348

J. Eng. Mech. 1987.113:1337-1357.

 y
t 9yB
& f^ f ex'B
U
i 4.A| >* ^ K B If
Downloaded from ascelibrary.org by National Taiwan University on 09/29/13. Copyright ASCE. For personal use only; all rights reserved.

L
/* — /
Z

FIG. 3.—Beam Element Model

The nodal degrees of freedom, i.e., the incremental displacements at

ends A and B of the element are (as indicated in Fig. 3)
<«*) = (uxAuxB) (44a)
(uy) = (Uy^LS^MyBL6zB) (44b)
<«z> = (UZA ~ L%AuzB - L6yB> (44c)
<ex) = (exAWxAQxBLQxB) (44d)
The stress resultants at section x of the member can be expressed in
terms of those at the element nodes as
Fx = FxB (45a)

Fy=-(M^+M2B) , (45&)

fz=(A^AV) (45c)

Mx = MxB (45d)
M y = -MyA-(l-i) + MyB-(i) (45e)
Mz = - M M • (1 - 0 + M2B-'(J) (45/)
As mentioned previously, the forces in Eqs. 45a-/ are acting at the be-
ginning of the incremental step, i.e., FxB = 1FxB, MyA = 1MyA, etc.
Now the following procedure may be performed: (1) Substitute the
expressions for the beam section properties (Eqs. 39a-h), those for the
incremental displacements (Eqs. 42a-d), and those for stress resultants
(Eqs. 45a-/) in Eq. 25; and (2) recognize the arbitrary nature of virtual
displacements. One can then obtain the incremental equations of equi-
librium for a tapered I-beam in a matrix form as
IKM + [kgm = m - m (46)
where [ke] and [kg] = the linear and geometric stiffness matrices, re-
spectively; {«} = the element nodal forces at the beginning and the end
1349

J. Eng. Mech. 1987.113:1337-1357.

 of the incremental step (Eqs. 24a-b). The elements of [ke] and [kg] are
given in Appendixes I and II, respectively. Notice that for the special
case of a uniform beam element, i.e., Q = Iw^ = Pi, = $h = 0, both [ke]
and [kg] reduce to those presented by Yang and McGuire (1986a).
The element incremental equations (Eq. 46) derived herein may be used
in an updated Lagrangian, large-displacement analysis. For instance, one
Downloaded from ascelibrary.org by National Taiwan University on 09/29/13. Copyright ASCE. For personal use only; all rights reserved.

may assemble the element equations of a structure to obtain the global

incremental equations of equilibrium as follows:
[Ke){U} + [Kg]{U} = {2F} - m (47)
where [K^] and [Kg] are formed in the usual way by direct stiffness method
procedures; {17} = the global incremental displacement vector; and {:F}
and {2F} denote the external loads acting at the beginning and the end
of the incremental step, respectively.
In a linearized buckling analysis, an attempt is made to detect the in-
stability load without calculating the absolute values of displacements.
Usually, an ideal or perfect member and loading condition is considered,
deformations prior to the occurrence of buckling are ignored, the linear
stiffness matrix is assumed to be constant, and the geometric stiffness
matrix is considered simply as a multiple of its initial value obtained for
a reference load {F}ref. Buckling of a structure is characterized by the fact
that it can deform a finite amount under a particular constant load ({2F}
_
ff} = {0}). At this load, the equations of equilibrium reduce to
[Ke]{U} + \[Kg]{U} = {0} (48)
where [Kg] represents the initial geometric stiffness matrix compared for
the reference load distribution {F}ref; \ = a load factor with respect to
the reference load {F}ref; and {U} = the buckled shape expressed in a
relative sense. Eq. 48 represents one form of the eigenvalue problem,
which may be solved, say, by a combination of Sturm sequence search
and inverse vector iteration with shift (Yang 1984).

NUMERICAL STUDIES

An interactive computer graphics program has been prepared to per-

form the linear, buckling, and geometrically nonlinear analyses by using
the linear and geometric stiffness matrices developed in this paper. For
the present purposes, only the buckling analysis results will be pre-
sented and the material is assumed to be elastic. In the following, the
taper constant is denned as the ratio of the minimum to the maximum
value of a particular dimension that varies along the beam axis.
Tapered Beams with Centroidal Loading.—Numerical solutions have
been obtained for a simply supported beam with a central concentrated
load acting at the centroid. The ends of the beam are free to rotate about
the major and minor axes, but are restrained against rotation about the
beam axis. The cross sections at both ends are free to warp. The width
of each flange decreases linearly from the midspan, while the flange
thickness, web depth, and web thickness remain constant. The beam is
of length L = 152 cm (60 in.). The material is aluminum with moduli of
elasticity and rigidity E = 65.31 GPa (9,450 ksi) and G = 25.63 GPa (3,720
1350

J. Eng. Mech. 1987.113:1337-1357.

 Depth Tapered
Downloaded from ascelibrary.org by National Taiwan University on 09/29/13. Copyright ASCE. For personal use only; all rights reserved.

Present -1.52 m-
Kitipornchai Present
— Wekezer
Kitipornchai

0.0 02 0.4 Q6 0.8 10 0.0 0.2 0.4 0.6 0.8 1.0

Taper Constant Taper Constant

FIG. 4.—Critical Loads of Tapered FIG. 5.—Critical Loads of Tapered

Beams with Centroldal Loading Beams with Top Flange Loading

ksi). The following dimensions were adopted for the cross section at the
midspan: b = 31.55 mm (1.242 in.), tf = 3.11 mm (0.1225 in.), h = 72.76
mm (2.865 in.), and tw = 2.13 mm (0.084 in.). Using 10 elements for the
beam, critical loads were calculated for different values of the taper con-
stant and are shown in Fig. 4. Also shown in the figure are the results
predicted by Kitipornchai and Trahair (1972) with the finite integral method
and those of Wekezer (1985) with the finite element method. It can be
seen that good agreement has been made, which confirms the validity
and adequacy of the present formulation.
Tapered Beams with Top Flange Loading.—Numerical solutions have
also been obtained for the critical loads of the same simply supported
I-beam with a central load acting just above the top flange. Linear vari-
ation is assumed either for the flange width or for the web depth from
the center to the ends. The results for both cases are drawn in Fig. 5,
in which the entire beam was divided into 10 elements. As can be seen,
the present solutions agree very well with those of Kitipornchai and Tra-
hair (1972). Again, the validity of the present formulation was verified.
Tapered Columns with Axial Compression.—The performance of the
element model will be investigated for a tapered column subject to an
axial compression. The flange width of the column is assumed to de-
crease from the support to the free end with a taper constant of 0.5. The
dimensions of the cross section at the fixed end are the same as those
at the midspan of the previous examples. The length of the beam is L
= 152 cm (60 in.). Both the present tapered element and the uniform
element of Yang and McGuire (1986a) were used to obtain the critical
loads. The results shown in Fig. 6 indicate that a tapered element con-
verges faster and monotonically, when compared with a uniform ele-
ment.
Beams with Various Loading Positions.—The effect of loading posi-
tions will be studied for a simply supported beam of length L = 6.10 m
(20 ft). Free warping conditions are assumed at both ends. The beam
1351

J. Eng. Mech. 1987.113:1337-1357.

 z o
—ID-o attorn Flange Load
Downloaded from ascelibrary.org by National Taiwan University on 09/29/13. Copyright ASCE. For personal use only; all rights reserved.

T3 n -O t -
o Centroid Load___---v^-
O Ci o
-• o
CM'

'Z o
—o
— 00

Top Flange Load

•*- Present
— Brown
3
0 2 4 6 8 10 0.0 0.2 0.4 0.6 0.8 1.0
Number of Elements Taper Constant

FIG. 6.—Critical Loads of Tapered Col- FIG. 7.—Critical Loads of Tapered

umns with Axial Compression Beams with Various Load Positions

depth is allowed to vary linearly from the midspan to the ends. The
maximum cross section occurs at the midspan, of which the dimensions
are: b = 152 mm (6 in.); tf = 13 mm (0.5 in.); h = 610 mm (24 in.); and
ta = 9.5 mm (0.375 in.) The material properties are: E = 200 GPa (29,000
ksi) and G = 77.2 GPa (11,200 ksi). A central concentrated load is applied
either at the centroid, at the top flange, or at the bottom flange. The
critical loads for the three cases were plotted in Fig. 7. As expected, the
top flange load results in significantly reduced critical loads as compared
with the other load positions.
As can be seen from Fig. 7, substantial discrepancies exist between
the present solutions and those of Brown (1981), which were obtained
by the method of finite differences. Such discrepancies result primarily
from the use of an approximate expression for the axial strain in Brown's
energy approach. For this reason, the present solutions are considered
to be superior. This example demonstrated not only the effect of load
positions but also the effect of taper, through an accurate consideration
of the axial strains in the buckling analysis of tapered I-beams.

CONCLUSIONS

The differential equations of equilibrium for the stability of a tapered

I-beam have been derived from the principle of virtual displacements.
The equations are more general than those derived previously for two
reasons. First, the effect of tapering was consistently taken into account,
due to the use of an accurate expression for the strains based on the
membrane theory of shells. Secondly, the effect of instability due to var-
ious loads has been considered, due to the incorporation of various non-
linear strains in an updated Lagrangian formulation. Such a considera-
tion enables one to investigate the instability of a beam subject to either
an axial force, a bending moment, a shear force, or a torque.
1352

J. Eng. Mech. 1987.113:1337-1357.

 The linear and geometric stiffness matrices have been derived for a
tapered I-beam based on a linear field for the axial displacement and a
cubic field for other displacements. Such matrices are expressed in an
explicit form and may be used to study the instability of a framed struc-
ture with tapered I-sections. The validity and adequacy of the present
theory are verified in the first two examples through comparisons with
Downloaded from ascelibrary.org by National Taiwan University on 09/29/13. Copyright ASCE. For personal use only; all rights reserved.

solutions obtained by other investigators. The third example on the flex-

ural buckling of a tapered column indicates that the speed of conver-
gence of a tapered element is generally better than that of a uniform
element. Both the effect of load positions and the significance of using
a correct expression for the strains were demonstrated in the last nu-
merical example.

ACKNOWLEDGMENTS

This research is sponsored by the National Science Council of the Re-

public of China under Grant Number NSC-74-0414-P-002-21.

APPENDIX I.—LINEAR STIFFNESS MATRIX

Since the linear stiffness matrix is symmetric, only the nonzero ele-
ments in its upper triangle will be given here.
EA0 ( 1 \
fce(l,l) = -£• \1 + - ca) (49)

ke{\,7) = -k.Q.,1) (50)

ke{7,7) = M l , l ) ' (51)
EIz0 ( 24 21 \
M2,2) = -jj- \\2 + 6czl + — cz2 + - cz3j (52)

ke(2,8) = -ke(2,2) (53)

EW 7 6 \
ke(2,6) = -^\6 + 2czl+-cz2 + -cz3j (54)

M6,8) = ~ke(2,6) (55)

EIz0 ( 8 2 \
ke(6,6) = -^ I 4 + czl + - cz2 + - cz3 I (56)

fce(8,8) = fce(2,2) (57)

fce(2,12) = ^f \6 + 4c2l + y cz2 + 3cz3J (58)

^(8,12) = -M2,12) (59)

E7z0 / 13 4 \
fce(6,12) = — (2 + CA + — cz2 + - cz3J (60)

1353

J. Eng. Mech. 1987.113:1337-1357.

 FT / ^R 11
*,(12,12) = -2 (4 + 3czl + - cz2 + - c23 j (61)

£7 / 24 21
M3,3) = - ^ (12 + 6cyl + y cy2 + - cy3 j (62)
Downloaded from ascelibrary.org by National Taiwan University on 09/29/13. Copyright ASCE. For personal use only; all rights reserved.

M3,9) = -ke(3,3) (63)

fe(3,5) = — ^ (6 + 2cyl + - cy2 + - cy3j (64)

ke{5,9) = -M3,5) • • • (65)

EI / 8 2 \
*,(5,5) = - ^ (4 + cyl + — cy2 + - cy3J (66)

ke(9,9) = ke(3,3) (67)

M 3 , l l ) = ~^f U + 4cyl + j cy2 + 3cy3j (68)

M 9 , l l ) = -ke{3,l\) (69)

M5,H) = ^ f 2 + cyl + ^ cy2 + ^ cy3j (70)

EU I 38 11 \
ke(ll,U) = ~f I 4 + 3cyl + — cy2 + — cy31 (71)

, /A AK ECw0 ( 24 21 132 24
*e(4,4) = -JJ- I 12 + 6cwl + — cw2 + — cm3 + — cw4 +• — cw5 j
GJ0 /6 3 \ EQ 0 /6 3 12 3
+ +
T(5 5 ' j — {5 + 5C*
C + + C 2 +
T5 » C 3
U »
EIa¥3 /6 6 36 6 \
-ir{5C* +
5Ci2 +
3l>C« +
7Ci*) • (72)

, /A ^ EQo / , 7 6 38
fce(4,13) = - j j - 16 + 2 ^ + - c„,2 + - cw3 + — cwi + cw5,

EI^o/1 1 3 1 \
c 1+ c 2 + c 3 + Ci4 (73)
""Tll5 - 5 ' Ii ' i j •
, .„ nox ECaa ( 8 2 12 13
ke{13,13) = —— 4 + cwl + — cw2 + - cw3 + — cai + — cw5i
L \ lo o 35 42
/2 1 \ / 2 1 2 11

-£Uo(l+^C, + lc,, + |c,, + ^C,,) (74)

1354

J. Eng. Mech. 1987.113:1337-1357.

 ECw0 I 17 94 17 \ /1
ke(A,U) = -jj- I 6 + 4c*i + — Cwi + 3cw3 + — cwi + — cw5j + G/0( -

^ /1 1 1 \ EIWM ( 1 3 1 \
+ _ + (75)
Ml5 K*-ii*J — h ^ + i ^ + H
Downloaded from ascelibrary.org by National Taiwan University on 09/29/13. Copyright ASCE. For personal use only; all rights reserved.

ECw0 ( 13 4 26 29
fce(13,14) = —— I 2 + c,! + — cw2 + - cw3 + — cwi + — cwS
(I 1 \ (I 1 1 11
v y y
\30 60 7 \30 60 70 840
„T /l 1 3 11 \
+ KH.* C,l + C,2 H C.-3 + C,4 V(76);
* \30 30 70 210 /
,. ^ EC,*, / „ 38 11 68 73
fce(14,14) = —— 14 + 3cwl + —ca2 + — cw3 + — cai + — cw5

^r r / 2 ! \ „„ / 2 1 3 13
+ GJ0L — H c, + ECML\ 1 c„i +— ; c„2 H cv3
\15 10 7 \15 10 y 35 y 168 y
„ / 13 4 26 29 \
+ EIa^l + - cfl + - c,-2 + - ca + — CiiJ (77)

ke(4,W) = -fce(4,4) (78)

fce(10,13) = -fc(4,13) . (79)
fce(10,10) = fce(4,4) , (80)
fce(10,14) = -M4,14) (81)

APPENDIX II.—-GEOMETRIC STIFFNESS MATRIX

Only the elements of the geometric stiffness matrix that are related to
the Wagner effect will be given here; the other elements are identical to
those of Yang and McGuire (1986a). Notice that the matrix is symmetric
and the notation K0 = FxBIpoMo is used.
K0 (6 +
3c 12 \
fcg(4,4) = 7 (5 5 * + ^ c"J ~ Fy^A (82)

^(4 / 13) = X 0 ^ + i C r l + l C r 2 ) (83)

fc,(4,14) = K„( ^ - 1 c)j (84)

^(10,10) = j (^ + g crl + ^ cr2J - FvBaB (85)

^(4,10) = -^(4,4) - FyAaA (86)

1355

J. Eng. Mech. 1987.113:1337-1357.

 - (2 1 2 \
Ug 1 3 , 1 3 ) = KoL\ — + — c rt + — cr2 (87)
\15 30 105 /
^(10,13) = - ^ ( 4 , 1 3 ) (88)
Downloaded from ascelibrary.org by National Taiwan University on 09/29/13. Copyright ASCE. For personal use only; all rights reserved.

U13,14) = -K0L {— + — c rl + — cr2) _... (89)

v
* \30 60 70 / '
*i(10,14) = -fc,(4,14) (90)
y i 4 4 4 ) = K 0 L ^2 + ^1c r l + ^ C3 r 2 j (91)

APPENDIX III.—REFERENCES

Argyris, J. H., Hilbert, O., Malejannakis, G. A., and Scharpf, D. W. (1979). "On
the geometrical stiffness of a beam in space—a consistent V.W. approach."
Computer Meth. Appl. Mech. Engrg., 20, 105-131.
Bazant, Z. P. (1965). "Non-uniform torsion of thin-walled bars of variable cross
section." Int. Assoc. Bridge Struct. Engrg., 25, 17-39.
Brown, T. G. (1981). "Lateral-torsional buckling of tapered I-beams." /. Struct.
Div., ASCE, 107(4), 689-697.
Cywinski, Z. (1964). "Theory of torsion of thin-walled bars with variable rigid-
ity." Archiv Inzynieru Ladovej, 10(2), 161-183.
Green, A. E., and Zerna, W: (1968). Theoretical elasticity, 2nd ed., Oxford Uni-
versity Press, Ely House, London, U.K.
Kitipornchai, S., and Trahair, N. S. (1972). "Elastic stability of tapered I-beams."
/. Struct. Div.', ASCE, 98(3), 713-728.
Kitipornchai, S., and Trahair, N. S. (1975). "Elastic behavior of tapered mono-
symmetric I-beams." /. Struct. Div., ASCE, 101(8), 1661-1678.
Lee, G. C , and Szabo, B. A. (1967). "Torsional response of tapered I-girders,"
/. Struct. Div., ASCE, 93(5), 233-252.
Love, A. E. H. (1944). A treatise on the mathematical theory of elasticity. 4th ed.,
Dover Publications, New York, N.Y.
Vlasov, V. Z. (1961). Thin-walled elastic beams. 2nd ed., Israel Prog, for Sci. Trans-
lation, Jerusalem, Israel.
Washizu, K. (1975). Variational methods in elasticity and plasticity. 2nd ed., Perga-
mon Press, New York, N.Y.
Wekezer, J. W. (1985). "Instability of thin-walled bars," /. Engrg. Mech., ASCE,
111(7), 923-935.
Wilde, P. (1968). "The torsion of thin-walled bars with variable cross section."
Archiwum Mechaniki Stosozvanej, 4(20),,431-443.
Yang, Y. B. (1984). "Linear and nonlinear analysis of space frames with non-
uniform torsion using interactive computer graphics," thesis presented to Cor-
nell University, at Ithaca, N.Y., in partial fulfillment of the requirements for
the degree of Doctor of Philosophy.
Yang, Y. B. (1986). "Stability of thin-walled beams—a general theory." Proc, Int.
Symp. Practical Aspects in Computation of Shell and Spatial Struct., Katholieke Univ.,
Leuven, Belgium.
Yang, Y. B., and McGuire, W. (1986a). "Stiffness matrix for geometric nonlinear
analysis," /. Struct. Engrg., ASCE, 112(4), 853-877.
Yang, Y. B., and McGuire, W. (1986b). "Joint rotations and geometric nonlinear
analysis." /. Struct. Engrg., ASCE, 112(4), 879-905.

APPENDIX IV.—NOTATION

The following symbols are used in this paper:

1356

J. Eng. Mech. 1987.113:1337-1357.

 A = cross-sectional area;
aA,aB = heights of points of application of loads above cen
troids;
B = bioment;
b = flange width;
Downloaded from ascelibrary.org by National Taiwan University on 09/29/13. Copyright ASCE. For personal use only; all rights reserved.

r = warping constant;
Cj, = section property defined by Eq. 27 or 35b;
E = modulus of elasticity;
&xx f "xy i &xz = linear strains;
F = force;
{/} = element nodel force vector;
G = shear modulus of elasticity;
=
*Wt|l section property defined by Eq. 35c;
W* == moments of inertia about y- and z-axes;
St. Venant torsion constant;
J
K = Wagner coefficient;
[ke],[kg] = linear and geometric stiffness matrices;
L = length of bar;
MX,MV,MZ = torque and bending moments;
r = polar radius of gyration;
T = position vector;
S = midsurface coordinate;
u = displacement vector;
{«} = element displacement vector;
UX,Uy,UZ = displacement components;
V = volume of element;
x,y,z = coordinates of bar;
a = ratio of torque resisted by a„ to total torque;
fr = (h - b0)/b0;
P* = (h ~ h)/K)
^lxx / f\xy / ^Ixz = nonlinear strains;
6* , 0 y , 0 Z = rotations;
X = buckling load factor;
0"* / O'xy / &xz = normal and shear stresses;
* = parameter defined by Eq. 11; and
(O = normalized sectorial area.

1357

View publication stats J. Eng. Mech. 1987.113:1337-1357.