BUCKLING OF RIGID FRAMES

by

JUNG-ON LOOK

B. S., University of Wyoming, 1961

A MASTER'S REPORT

submitted in partial fulfillment of the

requirements for the degree

MASTER OF SCIENCE

Department of Civil Engineering

KANSAS STATE UNIVERSITY

Manhattan, Kansas

Approved by:

1963

Major Professor
 LO
te
k%
U8i>3
t b p JL
. TABLE OF CJOKTai TS
Decu.-

SINOPSIS 1

INTRODUCTION 2

METHODS OF CALCULATING THE BUCKLING LOAD 4

THE FUNDAMENTAL THEORY FOR THE HRECT ANALYTICAL

PROCEDURE 6

SYMMETRICAL BUCKLING OF FRAMES 13

ANTI-SYMMETRICAL BUCKLING OF FRAMES 15

INDEPENDENTLY BRACED FRAMES 23

EXAMPLES 25

CONCLUSIONS • 33

ACKNOWLEDGMENTS 34

APPENDIX A 35

APPENDIX B 45
 BUCKLING OF RIGID FRAMES

By Jung-On Look,* A. M. ASCE

SYNOPSIS

It is considered that elastic stability is a problem of

great importance in the modern use of steel and high-strength

alloys in engineering structures, especially in tall buildings,

bridges and aircrafts.

This paper presents an analysis of buckling of frames based

on the well-known slope-deflection procedure. The stability of

one-story and multi-story plane frames is studied for the anti-

symmetrical mode of buckling. Other methods for calculating

the buckling load of frames are also discussed. Typical ex-

amples are solved using the direct analytical procedure, and the

results obtained are compared with the results obtained through

the use of the moment-distribution method. It was found that

both solutions give results which are in very good agreement.

It is felt that the slope-deflection method is simpler and

more direct than the moment-distribution method in solving sta-

bility problems.

Graduate Student, Department of Civil Engineering, Kansas

State University, Manhattan, Kansas.
 INTRODUCTION

The modern use of steel and hi^i-strength alloys In engin-

eering structures, especially in tall buildings, bridges, ships,

and aircrafts, has made elastic instability a problem of great

importance. Urgent practical requirements have given rise in

recent years to extensive investigations, both theoretical and

experimental, of the conditions governing the stability of such

structural elements as bars, plates,

The first problem of elastic instability, concerning lateral

buckling of compressed members, was solved about 200 years ago by

L. Euler. Under forces of practical interest, the problem of

lateral buckling of columns, originated by Euler, has been exten-

sively investigated theoretically and experimentally, and the

limits within which the theoretical formulas can be applied have

been established. However, lateral buckling of compressed mem-

bers is only a particular case of elastic instability, which is

very important in the field of structural engineering. B. W.

James adapted the moment-distribution procedure to include the

effects of direct loads of the column; the work of James was ex-

tended by Lundquist^ to determine the elastic collapse loads or

the critical loads within the elastic range of plane frameworks.

L. Euler, "Elastic Curves," translated and annotated by

V. A. Oldfather, C. A. Ellis, and D. M. Brown. 1933.
2B. Vi. James, "Principal Effects of Axial Loads on Moment
Distribution Analysis of Rigid Structures," N.A.C.A . Tech . Note
ilmi 1935.
•*E. E. Lundquist, "Stability of Structure Members Under
Axial Load," N.A.C.A . Tech . Note 617 . 1937.
In what is probably the best existing treatise on critical loads

of elastic structures, Chandler* makes the observation that

Lundquist's work was "....the foundation stone in the concept

of stability, but in regard to the numerical evaluation of

critical loaas left much to be desired. 11

Recently, several authors such as S. Hansbo^ had extended

the moment-distribution method for solving the buckling load of

frames to include the multi-story structures. The energy method

had also been used by some authors to solve the buckling load of

frames, but it has been considered that the result of this method

gives an upper bound of the critical load of the frame which is

too conservative for the purpose of structural design.

This paper presents the slope-deflection method to solve

the buckling load for one-story and multi-story structures. The

fundamental slope-deflection formula relates the bending at one

end of a member, such as a column or a girder, to the end slopes

and relative transverse displacement of these ends. In applying

this formula to problems of transversely loaded frames, the

equation expressing the equilibrium of moments at the joint and

the equilibrium of shear at each story can be obtained. Gener-

ally, it is assumed that the axial strains in the members may be

disregarded, that is, the horizontal or transverse displacements

of all joints in a given frame at a given floor or level will be

*D. B. Chandler, "The Prediction of Critical Loads of Elas-

tic Structures," Ph.D . Thesis . Manchester University 1955.
.

*S. Hansbo, "The Critical Load of Rectangular Frames An-

alyzed by Convergence Methods," Transact ion of Chalmers University
of Te<?hnolQKY, Vol. 164-1*1, 195~
the same. Thus, only one relative transverse displacement need

be defined for all the columns of one story. However, plane

frames usually do not exist singly. A building structure normally

consists of a set of such frames which are connected by floors,

roofs, and some bracing systems. Usually, floors and roofs may

provide an additional rigidity against the lurching mode of

buckling*

Plane frames may buckle in either an anti -symmetric mode or

« symmetric mode viiich Involves or does not involve lateral dis-

placement of the frame, respectively.

METHODS CP CALCULATING THE. BUCKLING LOAD

Several methods" of calculating the buckling load of frames

had been worked out. Among these methods, three essentially

different approaches are discussed:

1. The energy method.

2. The moment-distribution method.

3. The direct analytical solution based on slope-

deflection procedure.
The energy method is based on the condition that if a frame

in stable equilibrium is given a small distortion, it will always

strive to return to its original position. If, on the other

hand, it is in unstable equilibrium, the distortion will increase

to infinity. This fact forms the energy buckling criterion.

Buckling will occur when the vo ric done by the external forces

in a virtual displacement equals the change in strain energy.

Loc . cit,
 The stability of a framework with rigid joints can be in-

vestigated by using the moment-distribution method. In the use

of this method, a particular set of values of the external loads

is assumed, and the corresponding axial forces in the bars are

determined, assuming that the frame has pin joints. Then an

arbitrary moment is applied to one of the joints of the frame,

and the moments in the frame are distributed in the usual way.

If the moment-distribution computations converge to finite val-

ues for the final end moments, the frame is, in general, stable.

The entire process is then repeated using increased loads on the

structure but maintaining the loads in the same proportion. If

the loads are above the critical value, the moment-distribution

computations will not converge, in general, to finite values of

the end moments in the columns. Thus by successive applications

of this procedure, the critical load is determined.

The direct analytical solution based on slope-deflection

procedure consists of setting up a system of equations, express-

ing the relations between the joint displacements and the joint

rotations which occur due to distortion of the frame. These re-

lations form a set of linear homogeneous equations, usually

called the stability equations. Generally, the unknowns (dis-

placements, rotations and moments) in the stability equations

are equal to sero, if the determinant of the coefficient is dif-

ferent from sero, which means that no distortion of the frame is

taking place, and the frame is in a stable condition. The sta-

bility criterion is obtained from setting the determinant of the

unknowns equal to zero, and this usually yields the value of the
buckling load.

THE FUNDAMENTAL THEORY FOR THE DIRECT ANALYTICAL

PROCEDURE?

Fig. 1. Applied forces and distortion of a column.

J, E. Goldberg, "Buckling of One-Story Frames and Build-

ings," Journal of the Structural division A.3.C.E . Vol. 86,
. .

Oct. I960, P. 53.

 If a number ab is subjected to an axial compressive load P,

the end b displaces with respect to the other end a, as shown in

Fig. 1. The moments at both ends of the member, by applying the

slope-deflection method, can be found in terms of the angular

and transverse displacements at those ends in the following form:

*ab * *ab jA ab 6 a / B fib ob - (A ab / B ab ) Aab 1 (1)

( L ab )

and

M ba = K ab
j
A ab © / B
b flb fl
- (A
flb
/ B
ab )
A^ ) ( 2)
L ab )

where

K -|L.
ab= »
L ab

E is the appropriate modulus of elasticity;

I denotes the moment of inertia of the cross section of the

member;

L is the length of the member;

6 is the angular displacement at the ends of the member;

A refers to relative transverse displacement of ends.

A and B are constants which depend upon the sign and magni-

tude of the axial load and may be expressed completely as

functions of the ratio of the axial load to the Euler

load of the member. They are given by the following

formulas:

(a) For compressive axial load

A = 3jn pL - pL cos pL ( 3a )
£ \1 - cos pL; - sin pL
pL
 B - pi - sin pL (3b)
2 (1 - cos pL) - sin pL
pL

(b) For tensile axial load

A s pL cosh pL - sinh pL ----------- (4a)

2 (1 - cosh pL) /sinh pL
pL

B - slnhpL - pL
2
——
(1 - cosh pL) / sinh pL
r— — (4b)

pL

in which

pl * 7T /i0i (5a)

and

g« J_ = _j? (5b)

where P is an axial load, denoted as positive while the member is

under compression. The value of A and B may be taken from Ta-

ble 1 or Figs. 2 and 3. It can also be found from Table 1 that

A and B will have the values 4 and 2, respectively, when the ax-

ial load is zero, which agrees with the well-known slope-deflection

equations.

The shear equation may be found by taking the moment about

either end of the column, as shown in Fig. 1.

S ab = SL- (*ab / *ba / P ab Aab ) (6)

L ab

Substitution of Eqs. (1) and (2) into Eq. (6) yields:

Sab = -*ab < A ab / B ab ) (9 a / 9 b - 2 Afe)- P^ &a& (7)

L ab
L ab ^ab
Table i. Slope deflection coefficients A and B for various
values of load ratioO.

G A B e A B

3.9 -73.34 73.56 0.3 3.539 2.109

3.3 -39.05 39.54 0.2 3.730 2.070
3.7 -24.69 25.39
3.6 -17.87 13.79 0.1 3.365 2.033
3.5 -13.73 14.36 4.000 2.000

3.4 -10.91 12.24 4.000 2.000

3.3 - 8.86 10.40 -0.1 4.131 1.968
3.2 - 7.30 9.02 -0.2 4.255 1.933
3.1 - 6.05 7.96 -0.3 4.334 1.910
3.0 - 5.03 7.12 -0.4 4.502 1.883

2.8 - 3.449 5.884 -0.5 4.619 1.357

2.6 - 2.252 5.019 -0.6 4.736 1.334
2.5 - 1.749 4.673 -0.7 4.849 1.811
2.4 - 1.300 4.333 -0.8 4.959 1.739
2.2 - 0.519 3.901 -0.9 5.069 1.769
2.0 0.143 3.521 -1.0 5.175 1.749
1.8 0.717 3.224 -1.2 5.333 1.713
1.6 1.224 2.930 -1.4 5.533 1.681
1.5 1.457 2.873 -1.6 5.777 1.651
1.4 1.673 2.773 -1.8 5.964 1.623

1.2 2.090 2.610 -2.0 6.147 1.593

1.0 2.468 2.468 -2.5 6.580 1.544
0.9 2.645 2.404 -3.0 6.990 1.499
0.8 2.816 2.346 -4.0 7.75 1.43
0.7 2.981 2.291 -5.0 8.42 1.33
0.6 3.140 2.241 -7.0 9.62 1.30
0.5 3.295 2.194 -9.0 10.69 1.26
0.4 3.444 2.150
1
 r

10

!
1
1

!
'

~X,! „„l
™pn Hni
1 1

^j<.. i
i
l

T" '"[K^r

* 1
pSi
J
!

-10

|
!
\J%L i
Thkr
-
r j ! ! 11 20 n-

-
!
!—V—
ai,
i

i
-
\%%a.\ |

•JO
.5:

;
..V

A i

';';.•:
| !
':::
t ;
•«0 i

1 1 1 •
|

"3:
1
i i
I JO

'
M
HE
-

irTi ::i:
/ Tftr '

^/,M«» ^©» r
i
1

;;i
60
'fc '

6. <H !
i

f|:
L - | !
1

'as ;

! P I?
i
l

; '
j

•o
(o)

-< 4 < i j !0
(6)

2 2 2 4 2 (. Zfi i5 Si i
!

A 3
1

6 10
if 4
t
/

FIG. 2. -VALUES OF A

!'

-i .;;

i
J . 4 J
ii

::!/
~i ;

i
i

p
4
!

! |

'

°i !

\MI
;s
7
n1
1
1 1

I
i i 1 !

i
Mi ; •

t
4-3
|
_jg£
—
|

1 i

1
W^ B

iig
,..,./...,

i
!
1 !
71
\ ,.r.: fHH i

- 3 \_ J^~
UsSf
S
:
.^/....^

-
!
'

Lrt.
..,;
^I tdfigpSG
JrrS y^rr

s~>^ i
M 1 1.,,.!?^- i [*) i -»^^-+ - —«— f-t- .
—— if)

I
* -1 1

FIG. 3.-VALUES OF 3
 11

Because of the various conditions at the column base, the

lower end a, is assumed to be elastically restrained by a rota-

tional spring having a spring rate Qa , then the amount at the

end a, due to this spring will be:

*ab =-Qt ® a <*)

which can be equated to Eq. (1) to yield 6 a or G :

b
-Q
. ». *ab
j
A
ab*. * B ab eb - < A ab / B ) -Ajfc)
ab ... (9)

Thus,

6« s
.
A ab'K£ A
L_ (
- B ab ^ / (A ab / B ab ) ^*kj
L ab >
- - (10)

Substitution of Eq. (10) into Eqs. (2) and (7) yields:

(
2
M ba Kab A
ab - B ab -
}
< )
<h> U«b / B ab )

(1 - B ab ) A ab
<4 L.
^abh ) (11)
:.b

and

s ab =
±& (A ab / B
ab )
(e (i .
(
b bU ]

^ ) - (2 - A ab / B<|b
14 ,
^ab ^^-Qj
,Sfi.
A ab / J^ A ab / f»j

tsI- f
*% ,i2)

For convenience, it is assumed that:

c ab = A ab / B ab (13a)
 12

A ib 8
A
ab -_4_ U3b)
£a_
A
ab * K ab

c «b = 1 - B ab (13c)

4b = 2 - A ab / B ab (13d)
A«b / Qa
*iib

Substitution of Eqs. (13a) to (13d) into Eqs. (11) and (12),

yields the following relations:

Mba * K a b < A
a b eb * c «b <>ab4*b.) IH)
L ab

and

S ab =-KaJi C
^ab
P
ab (c ab © fa - cj b AfibJ - ab
L ab
^ ab
(15)

For the two different conditions of column bases, namely,

pinned end and built-in end, the corresponding value of Q will

be zero and infinity, respectively. Thus, Eqs. (14) and (15)

become:

(a) for pinned end condition:

Mba = Kab MA_,_

< A
ab " *W < 8 b -4klfc)
Lab
(16)

ana

S «*> Z
±& £flb.
(A ab - B
ab ) <6b - Aafc) - P
ab
A^ ( 17 )
Lib *^b »b ab

For this case, Sqs. (13b) to (13d) reduce to:

A ab s " B ab } lda ^
£•£ (Aab <

*ib
 13

Cflb s A ab - B ab (l*b)
A ab

c ab = 2 - A ab / B ab s cab - - -(18c)
A ab

(b) for built-in end condition:

*ba = *ab (A ab % - C ab Aaj,)

Lab
(19)

and

Sab s - K^ C Ob - 2 Aab) - P ab Afib. (20)

^ab L ab ^ ab

For this case, Eqs. (13b) to (13d) become:

Aab = Aab (21a)

c
ab = 1 (21b)

4 : 2 < 2i *>

The above mentioned formulas will be used to evaluate the

buckling load of various frames.

SYMMETRICAL BUCKLING OF FRAMES

The frame buckles in symmetrical mode if lateral displace-

ment is not allowed to occur. The theory is applied to a simple

fixed-end portal frame, as shown in Fig. 4. From symmetry, it

follows that:

Aab » A dc =Abc * °

©b = -e c

The equilibrium condition of joints states that:

2%b = °

Therefore, M ba / M bc = (22a)
 14

or

K ab A
ab e b + 2K bc 6 b " ° (22b)

and

A ab = " 2K bc
(22c)
ab

To find the critical load for the above problem: A at> can be

computed from Eq. (22c) and Q will be obtained from either Table 1
or Figs. 2 and 3. The critical load Per. is found from the Equa-
tion

Per. - £>P e

Fig. 4. One-story single bay rectangular frame.

 15

ANTI-SYMMETRICAL BUCKLING GF FRAMES

The frame, in general, buckles in anti -symmetrical mode if

lateral displacement is allowed to occur.

Fig. 5. Simple portal frame,

I. In case of portal frame as shown in Fig. 5. Since it is

symmetric, then only two equilibrium equations need be written in

order to find the critical load. They are:

(1) ^b = or
s%,
2J,q = °

(2) The fictitious force H must vanish.

For anti-symmetric case, the following relations hold:

©b * ec

A ab = A dc
Abe -
 16

Thus, from

°
Z«s»
Then

« b . / "b. ° (23,)

and

Kab U ab ©b
" c
ab 4*fc) / 6 K bc e b »
(23b)
L ab

or
2 3c)
8D Kab Cab^ab (

L ab ( K ab A ab r oK bc)

Also, the force H must vanish.

Therefore,

-S ab -S dc . H . (23d)

Since S ab - Sdc, it follows that

-S ab = « = " {23t)
2

Substitution of fiq. (20) into Kxj. (23e) yields:

K^ C ab <*b * 2 ^lW p ab4^ = ° U3f)

1~
<<
5 »
^ab Lab 2

For this problem, a trial and error procedure for the solu-

tion of Eq. (23) and the determination of the critical load will

be found to be convenient and relatively rapid. Noting that, for

a frame such as shown in Fig. 5, the critical load must be less

than the Euler load computed with an appropriate modulus, a trial

value is selected for P ab and the coefficients evaluated. Assum-

ing A ab to have unit magnitude, the value of efe iB found from

£q. (23c). Substitution of this value of 8 b and the unit value

 17

of ^ab into the Bq. (23f) yields the value of H that would be

required to hold the frame in the deflected position with the pre-

scribed column loads. The sign of t he computed value of H de-

termines whether the axial loads are greater or less than the

critical loads. If H is positive, the assumed column loads ex-

ceed the critical load, since the force H is now supporting the

frame against further deflection. If H is negative, the assumed

column loads are less than the critical loads, since the direc-

tion of H now implies that the frame has "reserve stiffness."

If H vanishes, the assumed column loads are equal to the critical

loads. In the trial-and-error procedure, if H is not zero, its

sign clearly indicates whether the next trial value for the axial

load should be larger or smaller.

II. Portal frame with one fixed-end and one pinned-end col-

umn base, as shown in Fig. 6.

A P ab ^ p cd = p ab

rfir?

Fig. 6. Simple portal frame.

 id

Assuming an anti-symmetric mode of buckling, it follows that:

©b - Qc

A«b =A dc s 1

A bc = o

p F cd P
ab = «

and

then

/ M bc = (24a)
*ba
and

K ab Uab 9 b - C, b Ajfc) / K bc (40,, / 1(H) )

= (24b)

yA
or

^ a (24c)

( K ab A ab^ ° J

Writing the equilibrium of shear in the horizontal direction

yields:

-Sab -S dc » H (24d)

Putting iiqs. (17) and (20) into Eq. (24d) and simplified:

[
Kab c ab * *$& <W
L ab
Ucd " B cd )
J
'
^ - !
(
2
**k C ab / *cdCcd
L cdA c d
^cd *cd **«t>

(A cd- Bed))
)
A / 2 P£k
ib A * H , (24e)
> "= lib

The critical load of this problem can be obtained by using

the procedure described previously.

 19

III. Three-story single bay structural frame.

Fig. 7. Three-story single bay frame.

Assuming an ant i -symmetric mode of buckling for the frame

shown in Fig. 7, it follows that:

9(j m 0g

©b a Of

ea = ee

Pcd = Pgh = p

Acd - A gh
 20

Abe =Afg
Aab = Aef
and

(^jMJat d, c, b, h, g r or f = °

It follows that the expression for the rotations at the joints are:

9d = K cd c cdA
cd (25a)
L C d (*cd A cd r 6 K dh)

A cp
.

Acd
e c = K cd c cd L cd / K cb c cb L cb
(25b)
(K c b A CD / K cd A cd / 6 K cg )

Abe Aab
e b = K bc c bc ""^bc r K ab
Cab
^
£ab
< A ab " B ab) L ab
(25c)

K bc Abc / 6K b f / K ba A ab (A ab - B ab )

and the shearing equations are:

Kcd C c d (8 d - 2 A cd ) / P cd A^ = H_ s
2
(25d)
L cd L cd L cd

^ bc
K be c
L bc
bc < ec " 2
^
^£) / p bc
L bc
= H = ° (25e)

Jab C ab (8 b - 2 Aj|bJ / P ab A^b S 1 H =0 (Iff)

IS ^ab ^ab 2

The critical load of this problem can be obtained by using

trial-and-error procedure as outlined previously.

 21

IV. Three- story three-bay building tram*.

Fig. d. Three-story three-bay frame.

Assuming an anti-symmetric mode of buckling for the frame

shown in Fig. 8, it follows that:

8d - ©
p
ec = e

8b ©n

©1 = *h

oj = ef
 22

Acd s A 6h • A ki A op s

Abe » A fg A jk A no •

Aab A «f Aij-Amn
and

Ym • at all joints of the frame.

the joints
It follows that the expressions for the rotations et

are:
2K hd 8 d - Khg C hg A** = -(26a)
(4*hd / 6Kh! / Khg Ahg) e h /

(4K d h / *dc A dC ) ° d / 2K dh e h . K dc C dc^S. s (26b)

L dc

(K C d A cd / 4K C g / K cb A cb ) © c / 2K cg e g " K cd C C d 4*1
L cd

- K cb C cb Afc = o (26c)
Lcb

(4K gc / *gh A gh / 6K gk / Kgf A gf ) 9 g / 2K gc 6 C - K gh C gh

Ag
- igr c g f Arf
4415
= (26d)
L gf

< K bc A bc f **bf / K ba Aba> S ^ «bf »f " K bc cbc &*

- Kba C ba &&* =0 <26#)
Lba

(K fg A fg / 4Kfb / 6K f j / K f . A fe ) 6 f / 2K fD 9

-KfgCfg Alg - K fe C fe Afe. -o (26f)

L fg L?e

and the shearing equations are:

 23

K cd C
cd (e d
- 2 Atf) / ^ c (e h - 2 Aj&) / p cd A^
Led
**ed tgh
T^h gh cd
Led

/ Pgh A gh = H = (26g)
L gh 2

Kbc c bc < ec - 2 A^) / Kfg Cff (e g - z Afc) / p bc ^Jfe£

L bc Tf§ L bc
L^ L fg

/ Pfg Afg = H - (26h)

Lfg

*ab c ab (e b - 2 Aja) / k^ c. f (e f - 2 Aj*) / p ab ^ak

/ Pef ^«£ « 1H = (26i)

The critical load of this problem can also be obtained as

stated previously.

INDEPENDENTLY BRACED FRAMES*

As it has been stated before, frames cannot exist singly.

In frames which are braced against sidesway, the bracing system

of these frames nay be considered as external to the frame, and

this bracing system may be in the form of sheathing, wall panels,

or diagonal tie-rod in the plane of the frame. By applying the

slope-deflection theory and assuming that the bracing system is rep-

resented by a linear spring, then the critical load of the frame

can also be found. The same procedure will be used as to solve

the anti- symmetrical buckling of frames, except that the spring

Loc . cit.
 24

force must be added into the shearing equilibrium equation.

Spring

77T7

Fig. 9. Representation of an independently braced frame.

For the purpose of illustration, a two-bay single-story frame

as shown in Fig. 9, with the spring at joint b, represents the

additional or external bracing. It is also assumed that the col-

umn bases are fixed. Therefore, the three joint equilibrium equa-

tions will be used as described previously. If ft is the constant

or rate of spring, then K^aD is the spring force which acts in

the direction opposite the deflection^ aD> or it acts in the

positive direction of the fictitious force H. Therefore, this

force, K^aD must be added to the force K in the shearing equilib-

rium equation. Thus, from Fig. 9, the following equations may be

obtained:

-s
ab - S dc - S ef s H / KAab (27«)
 25

or,

C - (aUfc C flb / 2K^ C cd

jUfe °ab *b / hd C
cd 6 C / Kef e f tf
^ab ^cd L«f L z ab LzCd

/ 2K ef C ef - P^ - - P«f / K) A = H « (27b)

ITeT" L ab
Pjcd.

Lcd W
The procedure for determining the critical load from the

above equation is the same as that outlined in the previous sec-

tion, tfiich is based on the trial -and-error procedure.

Example 1

The Portal Frame Shown in Fig. 10. (The same frame is

shown in Fig. 5).

P Pdc
Assuming that •
•

6
£ = 30 x 10 psi

I = 20 ir i>

s e
6b c

Aab A dc = 1

A bc = o

Then,
2
p e s *3,n" - 955,000 lb.
Fig. 10. Simple portal frame.
6
*ab = it& = 15.22 x 10 lb - in.
L ab

K bc = SS3L 15.22 x 10° lb - in.

Ho
 26

The trial-and-error procedure is tabulated as follows:

6b ob-
: : H ob-
Number; •
• : : tained
: tained
of :P- P •
l : :from Eq. from Eq.
Trials: T^ (lb! A B C a A/B:(23c)
: : :

:
<W>(lb)
:

1st 0.7 663,500 : 2.931 2.291 * 5.272 : 0.00372 -670

2nd 0.75 716,000 j 2.399 i

2.3635* 5. 2675 : 0.003755 / 10

:

3rd 0.746 712,000 2.9051 5 i

2.3163! 5.2214:0.003715

Therefore,

Per. = 712,000#
which agrees with S. Hansbo f s "exact" value. From his example,

p. 36 of Reference 5, he obtained

(KL) cr# = 2.7H

where

K - /

or
p cr. a 7.36 (4a)
L2

f> 4 4
Substitution of E « 30 x 10° psi. and I = 20 111 - into the above
equation, yields:

Per. 712,000 #.
 27

Example 2

The Portal Frame Shown in Fig. 11. (The same frame is shown

in Fig. 6.)

Assuming that:

E= 6
30 x 10 psi
H
I s 20 in>

Then,

P e = Z^rn2 = 955,000 lb.

K a b s itEI I 15.22 x 10 6 lb. -in.

Lab

K bc = J& = 1 5.22 x 10 6 lb. -in.

Fig. 11. Simple portal L bc
:rariie,

The trial-and-error procedure is tabulated as follows:

,
r— 9v, ob- H ob-
Number 1
^ :
tained tained
of I
C P=gP e : :
from Eq« from £q.
trials 1
A • B :
C s A/B (24c) (24«)
S

1st 0.45 430,000 : 3. 3695 2.175 5.5445 0.00376 / 1«5

2nd 0.447 427,000 J3.374 2.1706 5.5446 0.003753 / 130

3rd 0.440 420,000 1 3. 3944 2.1676 5.5620 0.003755

-^

Therefore,

p cr. = 420,000 lb.

which is different by 0.942 per cent from 5. Hansbo's "exact"

value for the same frame, p. 42 of Reference 5. He obtained:

 23

(KL) cr# s 2.10

and for EI « 600 x 10 6 lb. -in. 2 , the critical load becomes:

P cr# = 424,000 lb.

Example 3

A Three-story Single Bay Structural Frame aa in Fig. 12.

(The same frame is shown in Fig. 7.)

Assuming that:

£ = 30 x 10 6 pai.
,
£1 I
in>
r

_j I = 100
d h
*
Then,
n £1
at column cd
£1 H
c g P # s 11.93 x 105 lb.

oS
.
at column be
NO £1 £1

£1 Pe = 7.65 x 105 i b .

b f at column ab
e
«0
£1 £1 P« = 21.3 x 10* lb.
a EI e

157.5"

and
Fig. 12. Three-story single
.6
bay frame. Kcd K gh 19.05 x 10 lb. -in.

K dh s *cg - K bf s K ae - 19.05 x 10 6 lb. -in,

K ba = K ef » 25.04 x 10 6 lb. -in.

K bc = K gf = x 5-22 x 10 6 lb. -in.

 29

The trial-and-error procedure is tabulated in the following:

=F
No. of trials
1st 2nd 3rd

e 0.407 0.325 0.373

bO ( (
P s (?P e (lb) 0.434 x 10 0.388 x 10 0.446 x 10*
U
O
no A 3.4336 : 3.5528 3.483

B 2.154 2.11925 2.1389

iH C s A / fl i 5.5876 5.67205 : 5.6219

8
p ed from £q.(25a) 0.00376 0.003765 0.00377

H from Eq.(25d) - 6060 - 7320 - 6620

e 0.633 0.507 0.583

(lb) O.484 x 10
6
0.388 x 10* 0.446 x 10 6
u
o
A 3.0876 3.28415 3. 1665
o
JO
I 2.2575 2.19729 2.2322

H C : A / B 5.3451 5.48144 5.3987

8
9 C from Sq.(25b) 0.0048 0.00499 C. 00479
p ;

; H from Eq. (25e) / 39 - 330

rfc

0.182 •0.209
e Since it is :

© unstable at 0.388 x 10 6 0.446 x 10 6

P (lb)
column be for:
o A
the assumed 3.7543 3.7173
jo
column load,
«i : there is no 2. 0633 2.0735
B
need to solve
C s A / B Per. for this 5.8176 5.7908
rH
O column.
U 0.004275 0.00426
6b from Eq.(25c)
S - 8390 :- 7950
H from Eq. (25f)

Thus, P cr . = 446,000 lb. agrees with S. Kansbo's result, p. 31 of

Reference 5. He obtained P C r. 148. 6 x 10" EI. Substitution of

EI 3000 x 10 6 lb. -in. 2 into P C r. T*1611 *

p cr. = 446,000 lb.
 H

30

Example 4

Three-story Three Bay Building Frame as Shown in Fig. 13.

(The same frame is shown in Fig. 8.)

P P P P Assuming that:

I
hi
h
£1
1
B — P
E = 30 x 10 6 psi
c
to I s 100 in>
NO
£1 PI n e, 6d
B n £1 H
c s k o ei * e h
£

£1 £1 £1 B "c ~ ®o

B £1 EI e k = eg
b f J n
Ob s ^n
^1 £1 £1 £1 SI
ej = e f

and all the hori-

168" 240" 168"
zontal joint dis-

Fig. 13. Three- story three bay frame. placements equal

to unity,

then,
6 lb-in.
*cd = K hg = K lk « K po = 17.86 x 10

K cb = K fg « K jk » Kno * 15.62 x 10 6 lb-in.

K ab = K ef = K ij = x mn = 20 »^3 x 10 6 lb-in.
6
K dh = K cg = K bf = K lp = K ko = K jn = 17. «6 x 10 lb-in.

K hl : K gk = K f j : 12 '5 x 1q6 lb-in.

 31

The trial-and-error procedure is tabulated in the following:

r No. of trials 1st 2nd

:

e 0.457 0.496

e 0.48 x 10 6 \
0,52 x 10
to
A 3.3591 l 3.301
U
o
B 2.1751 : 2.1922
x»
o
C s A ^ B 5.5342 t 5.4932
3
r* From Eq. (26a) 9d 0.0039 0.003385
and
Eq. (26b) 9h 0.00217 0.00217
<
H from Eq. (2og) - 9400 : - 3380
0.598 0.648
e
P=9P e 0.48 x 10 6 0.52 x 10 6

A 3.143 3.O636

B 2.240 2.265

C « A / B 5.383 5.3286

From Eq. (26c) 6C 0.004985 0.00505

8 and
•p Eq. (26d) 6 0.003502 0.003308
I
H from Eq. (26h) - 390 /

0.336 0.364

P 9r e 0.48 x 10 6 0.52 x 10 6
o A 3.5368 3.4962

B 2.12375 2.1352

C « A / B 5.66055 5.6314
8
From Eq. (26e) 6b 0.00577 0.00577
•p
and
Eq. (26f) •f 0.C03696 0.003905

H from Eq. (26i) - 5580 - 5050

Since from the second trial H i 0, therefore, P C r. 520,000 lb.
 32

In the above examples the actual loading system consisted

of forces applied at the joints only, and the effect of primary

bending moments was neglectod. This is justified by the results

of E. F. Masur^ who presented a method to solve buckling problem*

including the effect of primary bending moments on the elastic

stability of structure of a portal frame similar to that shown

in Figs. 5 and 6. The loading system is applied on the beam at

a certain distance from the joint. Masur, p. 20 of Reference 9,

states,

....The unbuckled structure is therefore assumed

to be in its virginal state; that is, its members are
straight, and no "primary" bending moments are present.
The replacement of the actual loading system by one that
is (for each member) statically equivalent to it is
usually justified by the assumption that very small
errors are thus introduced.

Therefore, in the elastic range, primary bending moments

affect the stability of structure only very little and can usu-

ally be neglected. The stability of a partially plastic struc-

ture is certain to be intimately related to the presence of

primary bending moments. However, the treatment of this problem

is beyond the scope of this report.

9E. F. Masur, I. C. Chang, and L. H. Donnell, "Stability of

Frames in the Presence of Primary Bending Moments," Proceedings
A.S.C.E ., Vol. 67 No. EM4, August 1961, Part 1.
 33

CONCLUSIONS

This report treats the stability analysis of rectangular

rigid frames through the use of the direct analytical procedure

based on the slope-deflection method. This procedure is fast

and accurate, particularly with the use of digital computers,

because the buckling problem of frames will be reduced to the

solution of a set of simultaneous equations.

The theory of slope-deflection procedure for solving sta-

bility problems of rectangular frames is simple, provided that

buckling occurs within the elastic range. It is felt that the

slope-deflection procedure is more direct and comprehensive than

the moment-distribution method for solving stability problems.

In this report the effect of primary bending moment which

occurs from the application of the actual loading system at

points along the member rather than at joints of the member is

not considered. The replacement of the actual loading system by

one that is statically equivalent to it, is usually justified by

the assumption that very small errors are thus introduced; but

this does not effect the application of this method for engin-

eering purposes.

The comparison of the results obtained using slope-deflec-

tion procedure and the results obtained by S. Hansbo, using- the

moment-distribution method for the same problems, shows that the

difference is smaller than 0.942 per cent. This proves that both
solutions give results which are in very good agreement.

All the results presented in this report are obtained through

the use of a slide rule.

 34

ACKNOWLEDGE HUTS

The author wishes to express his appreciation to his major

professor, Dr. K.N. Jabbour, for suggesting this subject and

for his help end advice throujphout this study.

 35

APPENDIX A
 36

Kxample '2. .Mult istory frame.

In (lie three-story
frame according to Fig. IT all
single-hay
members have constant and equal hending rigidity. El tin*. The
columns are centrally loaded hy the loads /'. Assuming /' 0.2.'! /•.'/ t

as a probable value of the critical loafl :

the following data are
obtained:

TABI.K 3 i

Mcml» £i ZP -Xphli
I I

—
:

2 1.44 1.24 .SiiH 1.2I.S .07011 ..1.12 . II I :i

;

(i 1.44 1.24 .fiSH 1.2 IS .07(10 ,:l.".2 -.(1 111

|

-3 2.40 .«.J4
:

.7<>"> 2.14 .1111111 .0403 —.211!

.(I!l2n .OKIIIi
'
2. 4M .(••:) 1 .705 2.14 .1141111 .111:1(1 .21 (1

-4 1.02 .S7(i .C.14 1.48K .:,::, .IIS . :i :. I

—S 1.02 i .1470
J
.(114 1.48S |
.57:, |
.lis
.2:1(1
— .:(.'. I

—5 1
1

(i 1

1
.500

tf//m» A7,,u» !.l „,» #//m»j A7/m»

By applying virtual antisyiniue! rieal moments at the joints 2, G,

3, and 7, adjacent to the critical middle story, it will he necccssary
to study only one half of the frame.
The moment distribution is carried through as in Ex. 1. Each
stage of joint balancing (J/ ) is followed h) a cont'd ion of the shear
(

in the columns of each story through sidesway

(£,).
Table :* c shows that PjEl 0.23;'m s is close to the buckling value.
Another trial with PIE] = 0.2 i/m- will show a rapid divergence! of A'.
Hence P 0.23/?/ t can be accepted as the critical load ]', .
r

This result can be checked by Gkanjiolm's method. The frame 1

)

parts a, b, and c. Fig. 18, buckle simultaneously if the frame

trans-
versals arc split in accordance with Tabic 4.

The frame has a critical load slightly larger than 0.23/i7t which
agrees with the previous result.
 37

TABLE 3 b. Balancing procalure.

V
-
15 1

d '
i
 3*

TABI.K 3 <•. Correction jur xitlp*uiit/.

Member I'll: i r M';l-:i

-.ir,;. - .2:i I 1.50

'

S I
o _
3 l.r,:. 3.50
;i 1
—,1117 ,431 I..1U

i 2 + .022 — .2114 .020 . 1:1

_>
3 :
.lul - l.:tn

:i- -4 •I
.02S - .4 54 . 1(12 .::<;

i
-»
.022 - .2.'.4 — .(1ST .:,i;

*" :i
>_
3 .lli'.l — 2.2« — .:> 1 I.I i>

3 -4 I .021 - .:,('.(•, — .1*2 — .(14

1- _o — .010 — .34 1 — .(I8K -

. 57
-
,<-
4 2 -3 -i .2 17 — 2.7n — .50 - 1.07

3—4 + .03* — .7:1* — .100 + .7(1

1—2 !

+ .003 — .421) — .000 .58

2—3 -.205 —3. 2 11 — .47 + 1.02

3—4 .or,n — .0:17 — .2<>a -;- .71

1—2 ! .015 — .51(1 -- .091 + .58

,V (i .. 2—3 + .300 —3. 70 — .40 — .SI!)

1
3—4 •f .084 -1.14 — .20 + .70

1—2 + .0211 OKI .001 .58

2—3 •: .351 4 ....
.411 .011

3 I .HIS 1 34 .20 .08

1—2 + .042 701 .002

8 S 2—3 + .304 1 OS .45
3-4 i
-.130 1
;',
1 .20

TABLE 4.

Frame jmrt [EI), lit..

"»l («,)= P„IBI
1

.25 1.00 1.5 0.0 3.7o .23

:
_
b .111 . 1 il 0.8 5.« 5.75 .23
<•
;
1.00 .00 4.5 .4 2.25 .24

Multiplier: El !
« 1 1 1 l;,ir
 39

Yr

is: TV *£I 2V
1%
^ *
4 m

Fig. n Fig. 18

Matrix analysis

Example 3. Portal frame.

The simple portal frame shown in Fig. 1'.), fixed at base, is unstable
at a value {kL)„ « 0.805 a = 2.7 2 1 ).

Assuming kL = 2.70 as a probable critical value, Table 5 a U

calculated:

TABLE on.

Member kL •
C Z ! - 7.p\i l'i>
I

1
!

01,23 2.70 2.02 .702 3.30 -.)<2ft

4.00
-
12 « -j .ft 0(1

.Multiplier: A7/i,i j
» » 1
 40

5 EI

4 EI 4 EI

T
1*
^L
Fig. 19

When the frame issymmetric with respect to loading and bending

rigidity as in this case, the value
of p will be equal for the columns
JZ>P.= 1/A where N is the
r
in each particular story. Hence P .

total number of the columns in the story.

and 21. Due to
The matrix elements are calculated in Fig. 20
matrix r will be
the symmetry of loading and bending rigidity, the
symmetric.
Fig. 21 yields

0.331
r,, = 1 — 0.024 0.80

r = - 89 1
.89

I of this matrix is equal to the matrix

The dominant eigenvalue
elements 0.89. Obviously, the frame is stable for kL = 2.70.

M or}o
d£.578=d 21 2 2
1

=.422 d =.422 k =.624 -c 2 ,= -.289

d 23 x
lQ +k,.=+.624
Z1
+.335

c =.334 c =.334
01 32
3 3

Fig. 20 Fig. 21
 .

41

Next st buckling value of kh 2.74 is guessed at. Table 5 b is

calculated:

TAKI.K .>!>.

MciuImt kl. ZppZp

2.7 2.8.- .soc. 3..v.> .898

01,23 4

12 4.HO ..Mill

Dimension: l/m

C
.12=-??! =C,21 M w01 =l
d
'3=.581=<fo 2
*
^O*
d =.419 d =.419 k =.680 -c-.--.291
10 23 x
+ k2j=+. 680
+ .389

c =.338 c =.338
01 32

Fig. 22 Fig. 23

Fig. 1'2 to i>:{ vield

0.380
Tim 1.21
t's
1 — 0.GS0
The matrix

r = r o i.2i"i

[1.21
J

yields /. = ] . 2 1

Thus IcL • 2.7-1 is above the critical value. Interpolation between

these two values of the, dominant eigenvalue gives (l:L) cr 2.714. The =
value obtained by the matrix method is in agreement with the
»exact» value.
 42

Example ft. Portal frame li.iving one column fixed at base and (he
other hinged.
Finally, the matrix method will lie used to find the critical load
of the portal frame, Fig. 30. As first approximation of the critical

load the mean value may be chosen between the critical load for
the frame fixed at base and for the frame hinged at base. This
give.- us a probable value kL — 2.04. 1 )
Now assuming as the first trial kL = 2.io, table 7 is calculated.

FABLE 7.

Mc'mlicr I kL • ' « .-
:
Zp —ZplZp

(il 2.10 3.:iS .043 1.800 .0 104 7 —2.00

.00000
2:s 2.10 1.08 o .81) — .00381 — .405

12 4.00 .500

Jin 1
! i.ilier: , Kljm 1 1 JB//m» Eljm*
l
»

The matrix elements are ovulated in Fig. 31 to 33.

Fig. '.i'2 yields

rM = 0.07 9/(1 — 0.978) = 3.60

•Likewise, Fig. 33 yields

T n == 0.095/(1 — 0.154) = 0.112

Thu. \he matrix

~0
3. go'

0.112

The dominant eigenvalue / of this matrix is

;. = ]/3Tgo • 0.112 = 0.04

Hence KL = 2.10 is below the critical value.

N< a i/alue kL = 2.11 is guessed at. Table S is calculated.

 43

5EI
"T

4E1 4EI 4.0 m

5.0 m
Fig. 30

c =.335 c
91
=.271
12
19 =.542
d•12 d j=.669
2
M 01 =l
d =.458 d =.331 =- .271
1Q 23 -c
21
k 1= . 978 k 01 = .350
Z1
7079~

C
01
=.295 (

Fig. 32
A
Fig. 31

C 12 =.335 c 91 =.272
M 02 =1 dj =.543
2
d
2 J=.670

-c =- .335 d =.457 d =.330

12 J0 23
k = .430 k = 154
12 2
.

c
01 =.295_K2

M 02 =1 M01 =1
-c.,= - .335 .272
k i9= -440 k = . 157 k 1= 1. 002 .356
12 2
"7105 .084

A.
Fig. 35 Fig. 3G
 44

Table 8.

- Zp.
Member KL e c Z :
p : s- p
: : : j

£P

'2.11; 0.645 1.671; 0.01036 -2.67

01 3.37; ;
;

;
0.00648

-0.792; -0.00388 -0.474

23 |2.H; 1.97; ° ;

~~~~"
12 : o ;
4.00; 0.500 ____-
Multiplier EI/M; 1 i EI/M 3 ;
EI/M 3 1 1
; : :

The matrix elements are evaluated in Figs. 34 to 36.

Fig. 36 yields K x >1 , which proves that KL - 2.11 is above the

critical value. Inspection shows that KL = 2.10 is closer to

the critical value than is KL - 2.11. Hence, KL = 2.10 can be

accepted as the critical value.

 45

APPENDIX B

BIBLIOGRAPHY ON BUCKLING OF RIGID FRAMES

 46

Chilver, A. H. "Buckling of Simple Portal Frame," Journal of

Mechanics and Physics &Solid s. Vol. 5, Nov. 1956. ". 13-25.

Goldberg, John E. "Buckling of One-Story Frames and Buildings,"

Vol. 86, Oct. I960. P. 53.

Low, M« V. "Some Model Tests on uiti-Story Rigid Steel Frames,"

I

Proceedings, Institution Civil engineers, Vol. July 1959.

Lundquist, 1. E. "Stability of Structural Members Under Axial

Load," N.A.C.A .. Tech. Note 617, 1937.

Merchant, V. "Critical Loads of Tall Building Frames," The.

Structural Engineer . August, 1956.

Masur, £. F., and L. H. Donnell. "Stability of Frames in the

Presence of Primary Bending Moments," Proceedings A.S.C.E ,
.

Journal of the Engineering Mechanics Division Vol. 87.

,

P. 19, August 1961, Part 1.

Hansbo, Sven. "The Critical Load of Rectangular Frames Analysed

by Convergence Methods," Transactionc of Chalmers University
.

of Technology, Vol. 164-181, 1956.

Timoshenko, S. P., and J. M. Qere. "Theory of Elastic Stability,"

McGraw-Hill Book Co., Inc., New York, N.Y., 1961.
 BUCKLING OF RIGID FRAMES

by

JUNG-ON LOOK

B. S., University of Wyoming, 1961

AN ABSTRACT OF A MASTER'S REPORT

submitted in partial fulfillment of the

requirements for the degree

MASTER OF SCIENCE

Department of Civil Engineering

KANSAS STATE UNIVERSITY

Manhattan, Kansas

1963
 It is considered that elastic stability is a problem of

great importance in the modern use of steel and high-strength

alloys in engineering structures, especially in tall buildings,

bridges and aircrafts.

This paper presents an analysis of buckling of frames based

on the well-known slope-deflection procedure. The stability of

one-story and multi-story plane frames is studied for the anti-

symmetrical mode of buckling. Other methods for calculating

the buckling load of frames are also discussed. Typical ex-

amples are solved using the direct analytical procedure, and the

results obtained are compared with the results obtained through

the use of the moment-distribution method. It was found that

both solutions give results which are in very good agreement.

It is felt that the slope-deflection method is simpler and

more direct than the moment-distribution method in solving sta-

bility problems.