You are on page 1of 8

ELASTIC STABILITY OF COMPOSITE COLUMN

By Yaxin Li1
Downloaded from ascelibrary.org by NIRMA INSTITUTE OF TECHNOLOGY on 10/06/17. Copyright ASCE. For personal use only; all rights reserved.

INTRODUCTION

For a vertical column whose cross section, nonzero density (self-weight),


and flexural rigidity are uniformly distributed, the stability due to the com-
bined effect of compression and density was studied by some researcher
(Wang 1988). For the buckling load, the governing differential equation,
which can be derived from the geometrical and statical considerations, is a
second-order nonlinear with variable coefficients. Even upon linearization,
its solution involves linear combinations of fractional-order Bessel functions,
Airy functions, or both. The equation for the eigenvalue problem is exact
but tedious to solve numerically. Thus usefulness of this method is limited.
The present work shows how this complicated problem can be easily
solved by the Rayleigh-Ritz energy approach and derives a simple formula
for predicting buckling load.
A standing heavy composite column is clamped at the bottom end and
is free at the top (Fig. 1). For section 1, i.e., i £ (0, L£), where 0 < £ <
1, its flexural rigidity and uniform density are EJ1 and qu respectively.
Similarly, for section 2, i.e., x E (L£, L), its flexural rigidity and uniform
density are E2I2 and q2. Moreover, there is a concentrated load P at the
top.
We assume that
qx = q2(l - \) ...(la)
E1I1 = T\E2I2 (16)

P = WhL (lc)
and use nondimensional quantities such as

* =f .....(2a)

y=i -<<26>
u= :
S? -----(2c)
2E2I2
W
W
= -n
^ 2qn2L2 •••(2<0

'Ph.D. Student, Dept. of Mathematics, Univ. of Illinois, P.O. Box 4348, Chicago,
IL 60680.
Note. Discussion open until April 1,1993. 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 April 6, 1991.
This paper is part of the Journal of Engineering Mechanics, Vol. 118, No. 11, No-
vember, 1992. ©ASCE, ISSN 0733-9399/92/0011-2320/$1.00 + $.15 per page. Paper
No. 1697.
2320

J. Eng. Mech., 1992, 118(11): 2320-2327


1
Downloaded from ascelibrary.org by NIRMA INSTITUTE OF TECHNOLOGY on 10/06/17. Copyright ASCE. For personal use only; all rights reserved.

1
I 92 £2/2 42

I
I
4- 91
I £1/1

JL. mmwiM
FIG. 1. Composite Fixed-Free Column

«fr.O"^ M
where £7, W, and M(x, £) are the strain energy of bending, work produced
by the forces, and local bending moment, respectively.
And also, we assume that two sections are fixed together rigidly. More
precisely, if y(x) stands for the deflection function, then we have
y(i~) = y($+) (3a)
?&-) = y'in (3b)
FORMULATION

Now, to determine the elastic buckling capacity, we should choose a


suitable deflection function which satisfies the boundary conditions:
y(P) = 0 (4a)
/(0) = 0 (4b)
/'(I) = 0 ..(4c)
The conditions just mentioned are satisfied by the trigonometric function

y(x) = 8, ( l - c o s H j + s 2 ( l - cos-^H) (5)

in which 5;(z = 1,2) are coefficients to be varied.


After guessing y(x), we find the bending moment in section 1 is Mu by
the statical considerations.
2321

J. Eng. Mech., 1992, 118(11): 2320-2327


Mfc, Q = (1 - X)(M>n + 824>12) + S ^ i + 82<t>22
Downloaded from ascelibrary.org by NIRMA INSTITUTE OF TECHNOLOGY on 10/06/17. Copyright ASCE. For personal use only; all rights reserved.

+ u, ( §! cos ^ - + 5 2 cos - — ) = S ^ n + 82i|>12 (6)

where
TTX
t|/ n = (1 - A.)4>u + <f>2i + "• cos - y (7a)

_
3-iu:
«K2 = (1 ^)4>12 + $ 2 2 + ^ COS (76)

TTX 2 / . ir£ . rac


(7c)
4>u(*, 5) = (6 - *)cos Y - - ^sin y - sin
y
3ra 2 / . 3TT£ . 3ra
4>i2(jc, 0 = (S - *)cos sin — sin — - (7d)
2 3ir V 2 2

4>2i(*, 5) = (l - ?)cos Y - \ ( i - sin


Y (7e)

<t>22(*, I) = (i - Ocos - Y + ^ ( x + sin "Y (7/)

Thus, the strain energy of bending in section 1 is readily computed as


(Timoshenko and Gere 1961)

tfi = ~ P M\(x, §dx = Sfo + 28x8213! + 8 ^ (8)


r\ Jo
where

a, = — il/?i dx (9a)

Pi = - P «|»ni|i12 dx (9b)
T| JO

i r* (9c)
7l = - *|»12 ^
T| JO
It should be pointed out that an alternate form to compute strain energy
is involving (v'[x])2 instead of y(x). However, when an assumed deflection
curve is used, it is preferable to use y(x). Since, in general, y(x) is more
accurate than y'(x).
Similarly, the moment in section 2 is

TO; . , ->\ 3-ITX


M2 = 8X <f>n(x, 1) + |JL cos + 82 4> (x, 1) + u, cos — . . . (10)
12

and the strain energy of bending in this section is


2322

J. Eng. Mech., 1992, 118(11): 2320-2327


U2 = I Ml{x, 0 dx = 8fc2 + 28i82p2 4- S ^ (11)
Downloaded from ascelibrary.org by NIRMA INSTITUTE OF TECHNOLOGY on 10/06/17. Copyright ASCE. For personal use only; all rights reserved.

where

tt2 = 4>ll(*, 1) +
f TTX
M- COS y dx (12a)

f (j>u(x, 1) + a cos y 4>12(x, 1) + |x cos — dx (126)

3-rrx
72 = (|>12(x, 1) + (x cos dx (12c)

Finally, the strain energy of bending of the composite column is given by


U = U1 + U2 = 8?a + 28182|3 + 8 h (13)
where
a = ax + a 2 (14a)
P = Pi + P2 • (146)
7 = 7i + 7 2 (14c)
During the process, the total work produced by both the concentrated
axial load and column self-weights is computed as

W =
\ f/o WW1 - Wt -x) + l-Z]dx + £ (y') 2 (l " x) dx

+ (x £ (j') 2 dx 1 = 8f4 + 2 8 A 5 + 81C (15)

where
I f * I f 6
4 = - (1 - 0 J (1 - cos TTX) dx + - (1 - X) JQ (1 - cos TTX)

cos dx
(*• - x) dx + - I (1 - cos TTX)(1 - x) dx + £
§/:<- TTX)

(16a)

3 f6 3
B = - (1 - £) (cos TTX - cos 2TTX) dx + - (1 - X)

f* 3 f1
(cos TTX - cos 2TTX)(£; - x) dx + - I (cos TTX - cos 2-rrx)
3a P
• (1 - x) dx + — (cos TTX - cos 2TTX) dx (166)

2323

J. Eng. Mech., 1992, 118(11): 2320-2327


9 f6 9 f6
C = - (1 - 0 J (1 - cos 3TTX) dx + - (1 - X.) Jo (1 - cos 3TTX)
Downloaded from ascelibrary.org by NIRMA INSTITUTE OF TECHNOLOGY on 10/06/17. Copyright ASCE. For personal use only; all rights reserved.

1 1
9 f 9(x f
• (£ - x) dx + - I (1 - cos 3TTX)(1 - x) dx + — j (1 - cos 3TOC) <£t
' ° (16c)
Thus the critical value of the load q2 is found from the equation
U = W (17)
which represents the condition when the equilibrium configuration changes
from stable to unstable.
Substituting (13) and (15) into (17), we obtain
<2(82a + 28182p + 8I7) = 8?A + 28J82.B + 8|C (18)
in which the nondimensional quantity Q is

Q = zMj •••••(u)
TT £, 2 7 2
The conditions for Q to be a minimum are

g= ° •-«
| =° »
or
S^ga - A) + 82(Qp - B) = 0 (21a)
S^QP - B) + 8 2 (2 7 - C) = 0 (216)
In order to avoid a trivial solution, we require that the determinant of
aforementioned equations must be equal to zero, i.e.,
Q2(ay - p2) + <2(2p5 - yA - aC) + AC - B2 = 0 (22)
the real roots of (22) give the buckling capacity of the two buckled modes.
The lower feasible root corresponds to the critical buckling load. The com-
plex roots simply mean that the buckling is impossible for the corresponding
various loads.
The preceding analysis is also applicable to composite column of more
than two different flexural rigidities or variable cross section, with slight
modifications in computing U and W.
So far, the discussion has been focused on fixed-free composite column.
For other end conditions, we should assume the appropriate function for
the deflection curve y(x), which satisfies the boundary conditions. For ex-
ample, if the column is fixed-hinged one, y(x) may be approximated by
v = 81(91^: - tan 0^) + 82(62x - tan 62x) (23)
where 6,-(i = 1,2) are the two smallest roots of following transcendental
equation
6 = tan 6 (24)
2324

J. Eng. Mech., 1992, 118(11): 2320-2327


RESULTS AND COMPARISONS

By using the symbolic computation technique such as MAPLE on a com-


puter, we find the variable coefficient functions are:
Downloaded from ascelibrary.org by NIRMA INSTITUTE OF TECHNOLOGY on 10/06/17. Copyright ASCE. For personal use only; all rights reserved.

1 + (x - | (1 + X) +
! ( i
~ x ) 2 + - [(1 + M. - 0 3
2-n
_ + (l - n)(i - g + M.)2 sin ^ + J_ [?(9 _ 10x + 5 , 2 )
ZTTT) T|ir

ir£
isin- 2 2sin
2
+ 9r,(i - o - 2m] + — [(i + M- - 9Cn - i) - fr] + —
TJTT T|TT
5T1 (1 X )(5 + 11X)
• [ 2 ^ + (1 + p. - 0 ( 3 + X - 3^)] + " " 3 sin ^
T)Tr

32 <
(1 - X ) ( c o s ^ + X s i n ^ - 1 T\ COS (25a)
T|Tr3

2T)
1 + ix - | (1 + X) + §(l-^)2l +
^[(l + M.-€) 3
(x3] + ^ ^ (1 + |x - 02 sin 3ir? + ^ ^ 2 [£(9 - 10X + 5X2)
6TTT| 9-niT

8 sin —
9T,(1 - g) - 2 ^ ] + [(1 - ,,)(! + ii - 0 + &]
9-niT 2

: :2
2 sin _M
+ ^ - [2£X2 + (1 + |x - 0 ( 3 + X - 3T0] 1
9ryn- 27TITT3

32
• [5t] - (1 - X)(5 + HX)]sin 3TT£ + (1 - X) ( x sin ^
27T!TT 3

3IT£ .\ 3TT£
- COS — h 1 I + T| COS —— (256)

= 32(1 - X)
+ [27(en - 6 - -n) + 2(1 + |x)(13 - 5X)]
9T|TT3 12T\TC'

+ TJ—, 2 [X£(23X - 20) + 14|XTI] + ^ ^ _ _ ^ ^ sin ir£


12T]IT 2T|TT

sin TT^ C
+ 3 [3(1 - T,) + X(19X - 22)] + j^ [(n - 1 + 5X)(1 - 5
9T]Tr:

x „,v,i 1 A99X2 19 71X , 19iA . . y sin2ir£

2325

J. Eng. Mech., 1992, 118(11): 2320-2327


C0S
n -L tw/i \ , M i (11(1 - in) + 5X)(£ - |x - 1)
2
(l + [x - o (i - -n) + -^r
Downloaded from ascelibrary.org by NIRMA INSTITUTE OF TECHNOLOGY on 10/06/17. Copyright ASCE. For personal use only; all rights reserved.

4 sin
2
> • [((x + 1 - od - . ) + m + 1 6 X9-niT
(1 x)
" 3

3ir€

1 r x
1 - - (1 - cos TT£)
4 ir
(26a)
3X 3X
*-£ Y (1 - cos 2<IT£) + 3 - — (1 - cos irO
(266)
C = ? ( l - X e + 2M,)-^ 1 - | ( 1 cos 3TT|)

(26c)
The Table 1 shows the comparisons between exact values by Wang (1988)
and approximations by this approach. Since Wang considered the uniform
flexural rigidity and self-weight, so we set X = 0, i) = 1 in the coefficient
functions. And also, according to Wang's (1988) paper, the definitions of
p„, and a„, are
q2L3
P„. = (21a)
£,2*2

PL2
{21b)

TABLE 1. C o m p a r i s o n s with Exact Results

K a„ (A = a
J$" P = QTT 2 /4 100(0,,, - p)/p w
(1) (2) (3) (4) (5)
80.0 -36.7551 -0.4594 79.5098 0.6127
70.0 -30.5413 -0.4362 70.1520 -0.2170
60.0 -24.5577 -0.4092 60.3110 -0.5183
50.0 -18.8694 -0.3773 50.2496 -0.4992
40.0 -13.5612 -0.3390 40.1295 -0.3238
30.0 -8.7285 -0.2909 30.0416 -0.1385
20.0 -4.4467 -0.2223 20.0057 -0.0284
10.0 -0.7327 -0.0732 10.0001 -0.0012
-10.0 5.2486 -0.5248 -10.0008 -0.0079
-20.0 7.7082 -0.3854 -20.0008 -0.0040
-30.0 9.9253 -0.3308 -29.9739 0.0870
-40.0 11.9582 -0.2989 -39.8419 0.3952

2326

J. Eng. Mech., 1992, 118(11): 2320-2327


In Table 1, columns one and two are exact values from Wang's paper,
column three is our input u., and column four gives the approximations to
the exact $w. It is clear that the errors are less than 0.7% from column five.
Downloaded from ascelibrary.org by NIRMA INSTITUTE OF TECHNOLOGY on 10/06/17. Copyright ASCE. For personal use only; all rights reserved.

APPENDIX I. REFERENCES
Timoshenko, S. P., and Gere, J. M. (1961). Theory of elastic stability, McGraw-Hill,
New York, N.Y., 90.
Wang, C. Y. (1988). "Approximate formulas for buckling of heavy column with end
load." J. Struct. Engrg., ASCE, 2316-2320.

APPENDIX II. NOTATION

The following symbols are used in this paper:

EJx = flexural rigidity of section 1;


=
£2/2 flexural rigidity of section 2;
L = total height of column;
L£ = height of section 1 of column;
P = concentrated load;
q1 = uniformly distributed load on section 1;
q2 = uniformly distributed load on section 2;
x - coordinate axis in direction of height; and
y = coordinate axis in direction of minimum flexural rigidity of cross
section.

2327

J. Eng. Mech., 1992, 118(11): 2320-2327

You might also like