THE EFFECT OF END FIXITY ON THE

STABILITY OF STRUCTURES

A THESIS
Presented to

The Faculty of the Division of Graduate

Studies and Research

"by

Takaya Iwamoto

" In Partial Fulfillment

of the Requirements for the Degree

Doctor of Philosophy

in the School of Aerospace Engineering

Georgia Institute of Technology

September, 1970
THE EFFECT OF END FIXITY ON THE
STABILITY OF STRUCTURES

Approved:
J^^L
Chairmaif^ ^

• '' ^ "T

= <fyfl/ay W3.
Date Approved by Chairman:
 ACKNOWLEDGMENTS

The research described herein was part of a general investiga-

tion of structural stability being carried, out by the School of Aero-

space Engineering at the Georgia Institute of Technology; it was

sponsored by the National Aeronautics, and Space Administration under

Grant No. NGL-11-002-096, and by the United States Air Force Office

of Scientific Research, under Grant No. 68-IU76. The support of

these agencies is gratefully acknowledged.

In addition, the author wishes to acknowledge his indebtedness

to Professor Wilfred H. Horton and Doctor Lawrence W. Rehfield, whose

influence on the author's professional development has been profound.

 iii

TABLE OF CONTENTS

Page
ACKNOWLEDGMENT ii

LIST OF TABLES v

LIST OF ILLUSTRATIONS vi

LIST OF SYMBOLS viii

SUMMARY x

Chapter

I. INTRODUCTION 1

II. NOW-DESTRUCTIVE LATERAL LOAD TESTS 5

Introduction of the Uniformly Distributed Load
Search for the Empirical Laws
Generalization to the Case of a Moving Lateral Support
Non-Uniform Cross Section
III. GENERAL INTEGRAL EQUATION FORMULATION 53

Formulation of the Integral Equation

Orthogonality Relations
Some Useful Results Derived, from the Integral Equation
Formulation
Application and Discussions of the Results Obtained in
Section 3
IV. FURTHER CONSIDERATION OF THE BEAM COLUMN 75

V. DESIGN FORMULAE FOR COLUMNS WITH: ENDS PARTIALLY

RESTRAINED AGAINST ROTATION . . . . 85

VI. CONCLUSIONS , 93

APPENDICES

A. ROTATIONALLY RESTRAINED BEAM OF LATERALLY UNYIELDING

END SUPPORTS 95
 iv

TABLE OF CONTENTS (Continued)

Page

Establishment of the Characteristic Equation for

Instability
Extremum Slope Relationship Under Uniformly
Distributed Lateral Load
Establishment of Extremum Slope Relationship
Under Concentrated Lateral Force

B. THE ELASTICALLY SUPPORTED BEAM COLUMN OF UNIFORM

BENDING STIFFNESS 106

Differential Equation and Boundary Conditions

C . APPROXIMATION OF THE UNIFORMLY DISTRIBUTED LOAD

SYSTEM BY A FINITE NUMBER OF EQUALLY SPACE CON-
CENTRATED LATERAL FORCES 110

D. BEAM OF VARIABLE BENDING STIFFNESS Un-

damped-Pinned Beam
Slope Evaluation
Linearly Varying Bending Stiffness: p = 1
Parabolically Varying Bending Stiffness: p = 2
Formulation of the Finite Difference Equation
for Buckling Investigation
Consideration of the Simply Supported Beam
Linear Variation in Stiffness (p = l)
Parabolic Variation in Stiffness (p = 2)
Formulation of Finite Difference Equations for Buckling
Load Determination

E. CANTILEVER WITH TIP LATERAL SPRING 128

Characteristic Equation for Instability

Behavior Under a Concentrated Lateral Force
Behavior Under Uniformly Distributed Lateral Force
Behavior Under a Concentrated Tip Couple

F. COMPUTER PROGRAMS 131^

REFERENCES 1^2

VITA Ikk
 V

LIST OF TABLES

Figure Page

1. Load Deflection Test 7

2. Load-Slope Test 13

3. Distributed Load-Slope Test 19

k. Estimation of the Buckling Load By Eq. (ll) 20

5. Errors Involved in the Estimation Formula Eq. (13) . . . 2*4-

6. Extended Beam Analogies 30

7. Propped Cantilever Beam with Linearly

Varying Bending Stiffness , kO
8. Propped Cantilever with Parabolically
Varying Bending Stiffness kl

9- Simply Supported Beam with Linearly

Varying Bending Stiffness 1+3

10. Simply Supported Beam with Parabolically

Varying Bending Stiffness . . . . . kk
11. Estimation of the Buckling Load by Applying
Three Equally Spaced Concentrated Lateral
Forces on the Rotationally Restrained Beam . . . . . . . k6

12. Sum of the Higher Eigenvalues of Rotationally Restrained

Beams 68

13. Sum of the Eigenvalues for the Cantilever Beam

with a Lateral Tip Spring 69

1*4-. Approximation by Rayleigh Quotient-Rotational Restraint 88

15. Approximation of the Buckling using Rayleigh Quotient-

Modified Lateral Tip Spring Case 91
 vi

LIST OF ILLUSTRATIONS

Figure Page

1. Elastically Restrained Beam Configuration 9

2. Typical Experimental Results of Horton

and Ford 11

3. Geometric Parameters for an Elastically

Restrained Beam , 22

h. Comparative Results for Equal Rotational End

Restraint , 25

5. Beam Analogies for Cantilever with Lateral

Tip Springs 28
6. Non-uniform Beam Unsymmetric about its Center 35

7- Clamped - Pinned Beam of Variable Bend Stiffness

Under Uniform Distributed Load . 38

8. Simply Supported Beam of Variable Bending Stiffness Under

Uniform Distribution Load 38

9. Multiple Concentrated Lateral Load Tests -

a = 0 Case hf
o
10. Multiple Concentrated Lateral Load Tests -
a = 1 Case kQ
o
11. Multiple Concentrated Lateral Load Tests -
a = 5 Case ^9
o
12. Multiple Concentrated Lateral Load Tests -
a = 10 Case 50
o
13. Multiple Concentrated Lateral Load Tests -
a = 100 Case 51
o
Ik. Multiple Concentrated Lateral Load Tests -
a = 1000 Case 52
o
 vii

Page
Figure
15. Elastically Supported Beam Under Combined Loading . . . . ^>k
16. Illustrion of the Product W..M for Two Ideal Cases . . . . 71
17. Illustration of the Figure Mw 7^
18. The Pin Jointed Beam Column with a Central Concentrated
Lateral Load 76

19. Pin-ended Column Qk

20. A Fixed Ended Column 8^

21. Correlation Curve Between the Rayleigh's Quotient and the

Buckling Load for the Lateral Tip Spring Case 92

22. Rotationally Restrained Beam on the Unyielding 96

23. Illustration of x. , x O J I and Z 101

 LIST OF SYMBOLS

constants of integration

flexibility influence function

C/L

modulus of the material

size of division used in finite difference solution

second moment of area

a digit, 1, 2, 3, etc.

a digit, 1, 2, 3, etc.

spring stiffnesses for rotational restraint

spring stiffnesses for lateral restraint

/~P7ET

column or beam length

distance between points of zero shearing force

distance apart of inflexion points

distance of the column end from furthermost inflexion

point

the greater of 1 and 1_

bending moment

number of discrete lateral loads on a beam

» -L p
07 (x) dx
J i
o
number of unknown points within the span of the beam

axial load

critical value of P

positive number which defines taper law,

(El(x) = EI {1-l/T (x/L)5})
 IX

Q, concentrated lateral load

q intensity of distributed lateral loads

_ o
q non-dimensional lateral load parameters = qL /EI

T taper ratio of the bending stiffness

U an energy function

V shearing force

W lateral deflection

W work done by axial force P

w non-dimensional deflection parameter = w/L

X coordinate along beam or column

x non-dimensional coordinate = X/L

a
Z a - Oi-, a parameter defined by 1 + o• ^•••-
l
o 1 A
Z coordinate normal to beam or column

Z a column matrix

<y end rotational fixity parameter = KIJ/EI

Q
|3 end translational fixity parameter == KL /EI
2
A a quantity defined by A = T

A 12+1+ ( ^ Q^) + oyj^

6 lateral deflection of column

6. . kronecker delta
ij
7] w'
. . th
slope in l buckle mode
l

X /PL 2 /EI

cp. orthonormal function, 9. (x)//N.

^ location of concentrated lateral load

a a quantity defined by (T + 2A)

 X

SUMMARY

For a very long time analysts and practicing engineers have

been beset by uncertainty in defining the boundary conditions for real

structures subject to destabilizing loads. Recently, some progress in

the resolution of this question has been made. A new method of boundary

evaluation, which associated the behavior of a structure under a non-

destabilizing force system with the behavior of the same structure

in a destabilizing environment, was proposed. The initial study

demonstrated that in a number of cases the technique had significance

for columns. However, the view point was restricted. It was not

clear from the initial work whether the results were fortuitous or

whether they were a consequence of a general principle. Empiricism,

particularly when applied in a very specialized fashion can often yield

results of interest for a particular or specific example and be value-

less in any expanded problem.. The question - how to determine the

influence of realistic boundaries in a non-destructive fashion - is one

of the most important. Thus, this thesis presents the results of

detailed evaluations which clearly establish that the prior results

are the outcome of a general, relationship basic to the issue .

 1

CHAPTER I

INTRODUCTION

It is generally accepted that the critical load for a relatively

slender, geometrically perfect, centrally compressed column is given by

the formula

Pcr = c TT^I/L2 (1)

where c is the restraint coefficient, a quantity which characterizes

the conditions of end fixity. Nevertheless, there is a major problem

in the practical use of this simple equation. This difficulty was

succinctly stated by Salmon [l] who wrote in 1921:

The most pressing point for future research on the subject

of columns is undoubtedly the question of the degree of
imperfection in practical fixed ends; in short, what value
of K[c]*should be assumed for such ends. A complete answer
to this question is difficult, but, at present, the designer
has no real data whatsoever regarding practical end conditions.

While a satisfactory answer to the above question remains to be

found, the designer is not without some practical information per-

taining to end restraint. Design conventions have emerged which are

usually conservative and which can be applied with confidence to

orthodox types of attachment. For example, Niles and Newell [2]

present the following guidelines for airframe column members:

* K[cJ, end fixity coefficient.

 2

As used in practice, the restraint coefficient is partly rational,

partly empirical. A value less than 1.0 is seldom required,
or a value greater than 2.0 permitted. Aeronautical engineers have
customarily designed compression members in trusses having all-
welded joints on the basis of c = 2.0, if the structure supporting
the member was as stiff as, or stiffer than, the member itself.
If the joints were made with several rivets or bolts, c was taken
equal to 1.5 unless adequate test data were available to justify
a larger coefficient. If the restraining effect of the contiguous
structure was in doubt, c = 1.0 was used.

It is clear that these criteria are of questionable usefulness,

especially in novel and innovative applications, and that more definitive

information certainly is desirable for all cases.

Since Salmon's statement was made, much research on the subject

of columns has been carried out. Great ingenuity has been expended by

experimentalists in simulating the "ideal" end restraint conditions

that are normally considered in analysis. A thorough review of these

efforts has recently been prepared by Horton and Struble [3]• In this

report the authors conclude that the complexity of achieving prescribed

conditions and the fact that such conditions are rarely, if ever, met

in actual structures provide strong justification for seeking a simple

non-destructive testing method. A successful method of this type would

yield data pertinent to the stability of column members with realistic

end restraint without risk of costly damage or loss.

One powerful technique of column evaluation has already been

established. This is the so-called Southwell Method, A review of its

applicability to various types of structures is to be found in Reference

[^] . Despite the power of this technique and its universal applic-

ability in structural stability problems, it does not completely satisfy

 3

the needs. The reason for this is that, before this method can be

applied, data must be generated at destabilizing load levels which

approach the critical. In this sense, the Southwell method can not be

included in the category of the non-destructive test. It is perhaps

more appropriately referred to as an indirect test.

Subsequent to the analysis of Southwell, Fisher [5] made a more

detailed study. In this, he demonstrated that constant lateral load

systems were effectively imperfections. Hence, lateral loading is a

simple means of sensitizing columns to axial load effects . This

means, then, that compressive load level can be reduced, data

acquisition simplified and reliability enhanced by applying a constant

lateral force. The Southwell critical load, being independent of

imperfection level, is unaltered by the procedure.

Another approach to stability evaluation has been based upon

relationships that exist between vibration frequencies and stability.

It consists of measuring reductions in natural, vibration frequencies

as destabilizing load levels are increased. Stability limits may be

estimated from vibration data obtained at subcritical values of load in

this manner. This technique, as the Southwell, method, is indirect

rather than non-destructive in the sense explained earlier and does not

appear to have received much, attention [6] .

Efforts [7,8,951°] directed toward the development of truly

non-destructive testing methods have recently been made. These center

around the establishment of relationships between the behavior of the

member under non-destabilizing conditions and in instability. The

 k

approach adopted in References [7]? [8], and [9] originated from the

observation that buckling loads of columns are raised if end restraint

stiffnesses are increased while the flexibility or compliance of the

member under lateral loads is concomitantly decreased. These early

investigations were concerned primarily with the establishment of

empirical correspondence laws relating critical loads and compliances

under lateral loads. The results obtained in this way are encouraging,

but much additional work remains to be done.

A semi-analytical procedure is presented in Reference [10] which

differs from the above. The author suggests applying lateral loads to

the memberj measuring parameters which vary with the end restraint

stiffnesses and then inserting these measured values into stability

equations in order to predict buckling. Obviously, this idea is both

rational and straight-forward, but it does not permit measurements to

be directly interpreted in terms of buckling. Moreover, since a

precise theoretical calculation is required, extension of the method

to more complex structural elements could present considerable

difficulty.

To date the need for a practical non-destructive stability test

is not yet satisfied. Nevertheless, by correlating buckling loads

with compliances under lateral loads, progress has been made towards

establishing such a test for columns. In the present work the back-

ground of this approach is reviewed in depth, a new correspondence law

for columns is presented and a significant parameter associated with

points of inflection identified. In addition, the practical appli-

cation of the law is discussed and comparisons with previous laws are made.
 5

CHAPTER II

NON-DESTRUCTIVE LATERAL LOAD TESTS

The work reported in References [7]? [8], and [9] is basic to

the current endeavor. The central idea presented in these pioneering

efforts is perhaps best introduced by way of an illustration. Consider

two similar columns of uniform bending stiffness with different con-

ditions of end restraint - one simply supported and the other clamped

at both ends. For these two configurations, the product of the

respective buckling load and the maximum deflection caused by unit

concentrated lateral force applied at the span mid-point is

constant. That is

P
cr <t> = T S <2>

where 6 is the span mid-point deflection, Q is the magnitude of the

concentrated lateral load, and L is the member length.

Equation (l) suggests that an increase (decrease) in restraint

stiffness raises (lowers)

v P , the buckling load of the beam acting as
' cr

a column, and lowers (raises) the compliance (6/Q) in such a way that

the product of the two remains constant. Is this fact simply

fortuitous or is it a consequence of a general principle of mechanics?

In an effort to answer this question, the remaining limiting

cases of restraint were investigated [7]; the results are summarized in

 6

Table 1. The best agreement corresponds to concentrated lateral loads

applied so as to maximize the compliance ( 6 / Q ) . It is seen that

equation (2) is adequate for all but the cantilever beam, in which

case it gives results that are in error by a factor of four.

Struble [9] modified equation (2) to derive a rule which also

applies to the cantilever. He noted that the difference between the

cantilever and the other cases recorded in Table 1 is the number of

inflection (zero moment) points that develop under both lateral and

destabilizing forces. (The cantilever restraint produces a single

inflection point, while the others result in two). Consequently, the

modified law is of the form

P
cr ( l ) = ^ 2 (3)
Cr
^ 12n

where 'n' is the number of inflection points that develop within the

span under the lateral loading. The incorporation of 'n' into the

empirical law is similar to the introduction of the concept of

effective or reduced length in the usual theory of column buckling;

recall that a column's effective length is associated with the

distance between inflection points in the buckled configuration.

Equation (3) can be appropriately called the load-deflection

test. Following Struble [9]? it will usually be abbreviated to

simply "P-delta" test or law.

Results for beams of uniform bending stiffness or for limiting

cases of restraint which correspond to either zero or infinite boundary

 7

Table 1. Load-Deflection Test

Ideal Conditions
of End Restraint
© © ©*KD
y- 48
•n* E I
%h^
1.000 0.02083 1.000

0.586L
2.0U5 0.0098*+ 0.968

0.500L
U.000 0.00521 1.000

0.250 0.33333 i+.000

 8

restraint springs are not sufficient to substantiate the soundness of

equation (3). Additional study [7:,9] provided strong evidence that

uniformity of bending stiffness is not essential for the P-delta law

to remain valid. A systematic study [9] of the two elastically

restrained configurations shown in Figures la and lb was also con-

ducted. For beams with elastic rotational restraint at the ends

(Figure la), it was found that the maximum error incurred by using

equation (2) is only 7 percent for all values of the rotational spring

constants.

The P-delta test is no panacea, however. When applied to "beams

possessing the type of boundary restraint depicted in Figure lb the

agreement found is far from satisfactory in certain parameter ranges .

A problem develops as the translational spring stiffness at the tip

increases - the number of inflection points changes abruptly from one to

two producing an unrealistic discontinuity in critical load prediction

based upon equation (2). An error in P of as much as 235 percent was

reported by Struble [9] for this type of constraint; he then proposed an

elaborate modification of the P-delta law in order to improve the

situation, but the modified law was cumbersome to apply and the error

remained unacceptably large for practical purposes.

An attendant experimental study [8] of the P-delta test was

conducted using a test fixture designed to simulate the boundary

conditions shown in Figure la. Maximum compliances were located by

systematically varying lateral load positions along the beam specimens

while critical loads were estimated by the Southwell method. Typical

 9

(a)
Unyielding Supports,, Elastic Rotational Springs

(b)
Cantilever with Elastically Supported Tip

Figure 1. Elastically Restrained Beam

Configuration
 10

test results appear in Figure 2; it is evident that the product of

the critical load and maximum compliance is constant.

While the value of the P-delta test cannot be denied, its

shortcomings when lateral springs are present prompted a further

search for a more comprehensive law. Noting that the transition from

one to two inflection points was at the heart of the problem, Struble

[9] made the following conjecture::

We have thus far emphasized the importance of inflection

points in beam behavior, indicating that there might be
some sensitivity to wave form, as opposed to wave ampli-
tude, that is lacking in the measurement of a single
deflection only. It is noteworthy that an inflection point
is also a point of extremal slope, and it is only natural
to wonder if the measurement of maximum slopes might provide
a better estimation of the buckling load than using the
deflection approach.

He tenaciously employed a trial-and-error approach to finding laws

relating the buckling loads and extreme values of slope or rotation.

Perhaps the most successful of the load-rotation or "P-theta"

laws is

*J<r) - 4 w

where 9 is the difference between the maximum and minimum values of

m
the rotation angle produced by bending under the lateral load Q,. The

load is positioned so as to maximize 9 . This relation is applicable

m
without modification to the four limiting cases of restraint discussed
 1.0

0.8

•O 0 Q0 6 fy
0 6
3
x
o
E

1 \.0.4
0.°

0.2

500 1000 1500 2000 2500 3000

Ifcr (LBS.)

Figure 2 . Typical Experimental Results of Horton and Ford (Reference 8 ) .

 12

earlier. It is illustrated in Table 2. It does not explicitly

contain either material or geometric properties of the beam.

The adequacy of the P-theta test has been explored within the

context of the elastically restrained configurations shown in

Figures la and lb. For beams on unyielding supports with ends

elastically restrained against rotation (Figure la), the use of

equation (h) results in a maximum error in critical load prediction

of 6 percent for all values of end spring stiffnesses. Unfortunately,

such good agreement cannot be achieved if a lateral spring is present

(Figure lb); unacceptably large errors in the range of weak lateral

spring stiffness indicate that this test suffers from a deficiency

comparable to that of the P-deIta law.

In summary, it has been clearly established that both the P-delta

and P-theta tests provide a sound basis for the evaluation of boundary

effects for a wide class of structures subjected to destabilizing loads.

These tests have been empirically developed, however. Their range of

applicability is limited and their underlying principles have yet to

be uncovered. Nevertheless, the potential usefulness of such methods

is beyond question.

Introduction of a Uniformly Distributed Load

In the practical application of either the P-delta or P-theta

tests, maximum flexibility locations are generally unknown at the

outset. It becomes necessary, therefore, to search for them by

applying lateral forces at a sufficient number of stations along the

beam. Such a procedure produces a large quantity of data, of which

 13

Table 2 . Load-Slope T e s t .

Ideal Conditions
of End Restraint
© Q X8
6m E I
8 Per 9m
2
TT^Z EI QL? QTT

1.000 0.1250 1.000

0.612L 2.0^5 0.0609 0.996

0.500L
lj-,000 0.0312 1.000

0.250 0.5000 1.000

• Denotes inflection point.

 1.1+

only a fraction - the extreme values of compliance - are of any

interest if the P-delta or P-theta tests are used. In view of the

inadequacy of these two methods for weak lateral restraints and of the

quantity of unused data available, the question naturally arises

whether this data can be employed effectively to improve stability

predictions in the parameter ranges where improvement is needed.

We are strongly tempted to search for a meaningful statistic

such as some form of weighted average which incorporates all the data

obtained in traversing the beam. We are also cognizant that an

intimate relationship exists between a "moving" concentrated load and a

uniformly distributed load - the effects of the latter are the cumulative

or integrated effects of the former. At this point these two notions

coalesce to suggest a new direction - exploring the possibility of

relationships between a uniformly loaded beam and its stability limit.

We emphasize at the outset that uniformly distributed loading is

not recommended as practicable. It should be regarded as the convenient

limiting mathematical form for use in the search for empirical laws.

The recommended test procedure would consist of suitably superposing

data obtained from the application of a finite number of concentrated

loads along the beam span. We shall consider this matter in more

detail later in the discussion.

Search for Empirical Laws

The development of empirical laws is usually based upon an

intuitive step followed by inductive-deductive reasoning. In this the

experience gained in prior or similar situations is most helpful. So.

 15

it has been in this study. Previous work indicated that extreme

values of beam displacement and slope caused by lateral loads provide

relevant compliance data. Moreover, the method of considering ideal

limiting cases of end restraint first and then proceeding to more

general situations has evolved as a pragmatic method of evaluating

proposed empirical relations. These two notions serve to guide the

course of our investigation.

We note that for a cantilever under uniform lateral load, q

per unit length, the tip slope is given by

9 = qL3/6EI (5)

and that the critical compressive load for an identical column under

axial force is

p c r = jhi/hi? (6)

whence it follows that

P •9 = ^ (7)
cr w
It should be noted at this juncture that the point at which the obser•

vation has been made is the tip of the beam. At this point there are

certain other clearly definable conditions.

(l) the bending moment due to either the side load or the
 16

axial load would be zero.

(2) the deflection in either load condition would be maximum,

(3) the slope at the tip can be regarded as a limiting case

since

6
lim 1 " 62
tx
AK-^O

where 6 is the lateral displacement determined at the tip and 6p is

the lateral displacement determined at a point &x inboard of the tip.

It is also pertinent to note that

(k) the shear at the tip is zero.

(5) that the displacement curve under the action of the side

force has no inflection points.

The introduction of a sideways partial restraint at the tip

causes certain clear changes to take place. If the stiffness of the

restraining spring does not exceed, a certain critical value then the

deflection at the tip will always be the maximum deflection in the beam

but when the value is exceeded it will no longer remain so. Moreover,

no matter how light the spring is, there will always be two zero

moment points along the beam, and there will always be a shear force

at the tip station.

We proceed now to examine the case of a simply supported beam.

Under a uniform load, q_, the maximum slopes are located at the positions

of zero moment, viz. each end, and are given by

 17

1_ £l£ (8)
2k EI

The critical load level in this case is given by

Pcr = # (9)

and so

P • 9 = ^"UT (10)
cr

The similarity -with the previous case is apparent. There is therefore

a strong justification for examining the other limiting cases in the

same fashion. At this stage certain complexities occur. We note now

that if both of the beam ends are clamped, the positions of zero moment

are no longer at the ends of the beam but are moved in from the ends

by 0.211 L and are separated by a distance of 0.578 L. The slopes

at these points are of course much reduced, having a value of

0.193 /'qL3\
"2^4 V EI/'

It is apparent that the prior simple relationship is no longer

applicable.

The discrepancy between the clamped-clamped and the other cases

 18

can be simply resolved by a slight change in the comparison formula

used. We observe that the product

P c r 9max = k x 0.193 TT2 §jf - 0.772 / §5

and that the 9max is positioned at a distance of 0.211 L or O.789 L

from the end. Hence, we conjecture that a 'better quantity to consider

is the least value of P 8max/tmax, where -tmax is the greater of the

cr '
two distances from an end to the inflection point. Thus, we propose

the approximate relation

P Gmax 2
rr ^ q.TT
~ q.n (11)
-tmax ~ ST

The results obtained when this simple rule is used for the four ideal

cases are shown in Table 3» The buckling determinant and the equation

for beam slope under uniform lateral load for the rotationally

restrained body on unyielding lateral supports is derived in Appendix

A. From these derivations we can numerically evaluate the approximation

(ll) in a broader context. The necessary computations were made on a

Univac 1108 and are summarized in Table k. We see from this that

the expression is excellent for all combinations of end rotational

stiffness

A further examination of the case of a beam with one end clamped

and the other simply supported (propped cantilever) shows that the
 19

Table 3. Distributed Load-Slope Test

Ideal Conditions
of End Restrainr & ® x <D
Per L 2 24 ^r^max
TT2EI Jmax'q I? Jma*q TT*

1.000 1.000 1.000

2.0^5 0.500 1.023

0.25L

max
J-I-.000 0.2^5 O.98O

taiax = O.789L

max
0.250 h .000 1.000

• Denotes inflecton point.

 20

Table 4. Estimation of the Buckling Load "by

equation (ll).

Stiffness Parameter Q:

0 •5 1 5 10 50 100 10* 10 5
1.000 1.094 1.175 1.306 1.548 1.730 1.968 2.006 2.042 2ToITc7
1.000 1.077 1.143 1.250 l.Jj-55 1.625 1.893 1.9^3 1.99^ 2.000
0 -1.5 -2.7 -^.3 -6.0 -6.1 -3.8 -3.1 -2.3 -2.2

1.193 I . 2 7 8 l . 4 i 6 1.671 1.864 2 . 1 1 4 2.153 2.190 2.195

1.197 1.267 1.380 1.598 1.780 2.067 2.121 2.175 2.l8l
•3 -.9 - 2 . 5 - 4 . 3 - 4 . 5 -3.3 -.7 -.6
-1.5
1.367 1.512 1.780 1.982 2.243 2.284 2.323 2.327
1.372 1.492 1.721 1.914 2.217 2.332 2.339
.3 - 1 . ^ -3.3 -3.5 -1.2 2.275 .4 .5
-.4
1.668 1.959 2.179 2.461 2.505 2.546 2.552
1.669 1.918 2.129 2.463 2.536 2.590 2.597
0.0 -2.0 -2.3 .0 .8 1.7 1.9

P exact 2.297 2.557 2.891 2.943 2.991 2.997

cr
P~.~
or e s t i m a t e d 2.273 2.520 2.870 2.922 2.970 2.975
% error -1.0 -1.5 -.7 -.7 -.7 -.7
2 . 8 5 4 3*239 3.298 3.353 3.359
2.813 3.187 3.246 3.300 3.306
-1.5 -1.6 -1.6 -1.6 -1.6

3.700 3.7H 3.838 3-845

3.718 3.789 3.855 3.863
•5 .5 .4 .4

3.845 3-91^ 3.921

3.896 3.965 3.973
1.3 1.3 1.3

3.984 3.992
4.077 4.085
2.3 2.3

4.000
4.098 10-
2o5
 21

relation (ll) applies approximately if the slope value at the second

inflection point is used. For this beam under the uniform load q,

the slope at the inflection point remote from the pinned end is

e = o.ikk $L

and this point is located at a distance of 0.75 L from the pinned end

Thus,

V*. /o.3j^L 3 y2WW i \ = o QIK> (-ai.

0.75L V 24EI A T2 Ao.75L/ ^ V 2JT

This observation suggests that equation (11) may be generalized.

If we let Q and 9 p denote the two extreme values of slope that corre-

spond to inflection points located at distances K,. and £p from the

remote ends of the beam (Figure 3 ) 5 then we expect that

( 1 9 | +l Q ]
IJIIJJL! i ^
\ ~ l2 ~ (l2 + l2)

Thus we can derive another equation as a replacement for equation (ll)

2k
(*, + 1 2 )
 22

uniform load
inflection points q (Force/Length)

Figure 3. Geometric Parameters for an Elastically

Restrained Beam.
 23

where I , t p , 0 and 0 are defined in Figure 3. In an effort to

clarify the issue further the behavior of a beam with rotational end

restraint at both ends is considered analytically (Appendix A ) . The

lateral load is uniformly distributed and has a value of q per unit

length. The analysis demonstrates that the sum of the absolute values

of the slopes at the zero moment points is proportional to the cube of

the distance apart of these points. This simple result has important

practical consequences, because the formula (12) can now be written in

the form

y—)(2 • i ext) 2
= TT (13)

< L+ W
In other words to evaluate the critical load under axial com-

pression we do not need to measure slope at a point but merely

distance apart of two points . Practically this is a much simpler

operation.

The numerical calculations appropriate to this formula

demonstrate that for all possible combinations of rotational spring:

stiffness the maximum error is only 2.5 percent. The results are

shown in Table 5. The comparisons between the results by this test and

those of "P-delta" and "P-theta" tests are shown in Figure h.

The calculation was made for the case of equal rotational

restraint, because for all three test methods this is the case in

which the maximum error exists. It is seen that the result for the
 Table 5• Errors Involved in the Estimation Formula
equation (13)•

S t i f f n e s s Pai^ameter OL

0 1 5 10 50 100 10 3 io-
1.000 1.094 1.175 1.306 1.548 1.730 1.9^8" 2.006 2.042 2.046
1.000 1.095 1.176 1.303 1.535 1.715 1.977 2.023 2.069 2.074
0.0 .1 .0 -.3 -.9 .4 .9 1.3 1.4

1.193 1.278 1.4l6 I.671 1.864 2.114 2.153 2.190 2.195

1.197 1.282 lo4l8 1.664 1.856 2.132 2.181 2.229 2.234
•3 .3 .1 -.5 •9 1.3 1.8 1.8

1.367 1.512 1.780 1.982 2.243 2.284 2.323 2.327

1.372 1.51^ 1.77^ 1.975 2.266 2.317 2.367 2.373
•3 .2 -.3 1.0 1.4 1.9 2.0

1.668 1.959 2.179 2.46l 2.505 2.547 2.552

1.669 1.950 2.169 2.484 2.539 2.593 2.600
.0 -.5 •9 1.3 1.8 1.9

2 .297 2 . 57 2.891 2.943 2.991 2.997

2.273 2 . 28 2.895 2.959 3.022 3.030
-1.0 -1.2 .1 .5 1.0 1.1

P exact 2.854 3.239 3.298 3.353 3.359

cr 2.813 3.227 3.300 3.371 3.380
P estimated
a/ cr -1.5 -.^ .I .5 .6
Jo e r r o r
3.700 3.771 3.838 3-845
3.718 3.805 3.891 3.901
.9 1.^ 1.5

3.914 3.924 3.921

3.896 3.985 3.995
1.3 1.8 1.9

3-984 3.992
4.077 4.087
2.3 2.4

4.000
4.098
2.5
 5 10 500 1000
STIFFNESS PARAMETER (|^)
ro
Figure h. Comparative Results for Equal Rotational End Restraint Ul
 26

uniform load test is to all intents and purposes "exact".

Generalization to the Case of a Moving Lateral Support

In the arguments presented so far we have assumed that variation

in rotational constraint is the prime problem of practical concern.

We consider this to be a well founded assumption. Nevertheless, we

realize that problems in which some flexibility exists in the lateral

support system occur. The prior work [7*9] failed for these cases.

The reason would appear to be that a single lumped compliance parameter

cannot accomodate such complex boundary effects. Thus, a second

parameter whose value is determined primarily by the lateral support

flexibility must be introduced.

To search for this parameter we consider the elastically

propped cantilever. To visualize the physics of the problem we begin

with the plain cantilever beam. Under the action of a uniform lateral

load this body deflects in a smooth curve and the maximum deflection and

slope occur at the tip. The introduction of a partial sideways

restraint at the tip causes certain clear changes to take place . If

the stiffness of the spring does not exceed a certain critical value

the deflection at the tip will always be the maximum for the beam.

However, when this critical stiffness is exceeded this condition is

no longer true. Moreover, no matter how light the spring is, there

will always be two zero moment points along the beam, and the zero

shear point will move away from the tip.

The propped cantilever for which the tip spring stiffness

exceeds the critical value intuitively seems the easiest to deal with.
 27

Thus, this case is treated first, it is readily apparent that the

lateral spring can be replaced by an appropriate cantilever spring

system. The requirements we must specify for this equivalent system

are:

(1) The cantilever is to be regarded as a continuation of the

beam under consideration. Therefore,

(a) it has the same EI as the beam

(b) it provides the same reaction force at the tip as

does the real spring.

(c) there is continuity of slope and deflection at the

proper point.

(2) To make the equivalent system handleable with the pro-

cedures previously followed the cantilever has an appropriate

rotational spring at its root and an infinitely stiff lateral support.

With these conditions defined we may proceed to the analytical

study of the problem. The system defined is depicted in Figure 5a and

we note that at the tip there are initially four prescribed conditions

viz known values of M , V , 6 , 9 . Two unknowns are to be ascertained

o' o o o

from analysis - namely, the beam length and the end fixity coefficient

a. It can be shown from the usual beam equations that the relationship

between L and the prescribed conditions is

iiiL + kv L3 - km L2 + ^8e L - ^86 = o. (ik)

o o o o

The critical load for the configuration should be as follows:

 28

(a)
Analogy for Gasel: K 3 > 2 4 - E I .

c
Analog y for CaseH: K3 < 24 fi
• Denotes Inflection Point

Figure 5« Beam Analogies for Cantilever with

Lateral Tip Spring
 29

P
cr " i f ¥ • ~^-T • J* (15)
2-t ext (t + L)

where the term

L2
(L + -t)2

is the natural correction for increase in column length from L to

L + t. This semi-empirical formula has been evaluated for values of

3
K L
3
EI

ranging from the critical to virtual infinity. The error is

acceptable over the whole spectrum. The extreme value, 13 percent,

occurs at the lower bound. The average value is, however, much

smaller as is clear from Table 6.

The cases for which the lateral support spring stiffness does

not reach the value which causes the tip to be inclined upwards for a

down loading of the beam cannot be treated by the procedure outlined.

Thus, we are forced to a somewhat more intuitive approach. To deal

with this case we go back to consider the relationship between the

instability loads for a one-end clamped, one-end free column and a

pin-pin column. We note immediately that the critical load of the first

is given by
 30

Table 6. Extended Beam Analogies.

Extended Beam Analogy Case (i)

3 P 2 P e s t . L2
K L exact • L
3 I E r r o r s (fo)
EI Tl2 E I Tl2 E I

24 • 334 1.664 1.897 13

30 • 302 1.731 I.766 3
4o .263 I.85O 1.807 -3
100 .154 1.996 1.833 -8
500 .037 2.037 2.000 -1
0 2.045 2.074 1

Extended Beam Analogy Case (il)

-p 2 v 2
K3L3 exact L
E r r o r s (fo)
L 1
+ est. L
EI Tl2 E I ext Tl2 E I

0 .250 2.0 2.0 .250 0

1 o332 1.8125 1.625 .328 -1.2
2 .414 1.7000 i.4oo .407 -1.7
4 .572 1.571 1.142 .561 -1.9
8 .872 1.455 .910 .830 -4.8
16 1.360 1.368 .736 1.235 -9.2
24 1.664 1.333 .666 1.498 -10.0
 31

P
cr = 7 % ' I ^
L
cf

and the critical load for the second by

Pcr = # i (17)
L
PP

For equality of loads for these two systems we see immediately that

L = 2L p (18)
pp cf

And so5 of course, we focus attention on the use of reflection in the

study of the problem. But reflection of the beam to obtain an equi-

valent beam is not all that is necessary. Some parameter closely

associated with the characteristics of the tip spring is essential.

Instinctively we feel that this may well be the location of the point

at which the shear is zero. We are given some confidence in this

viewpoint by the fact that we can show this point to have significance

in the case we have already treated.

Consider the beam column with uniform lateral loading and suppose

that the zero shear point is positioned at distance Ln from one end

and Lp from the other. Then we can demonstrate arithmetica.ily that for

a given beam
 32

2£3 - • 2L . 2^3 -. ,2L

ext
P x I ! l) - p x i OT , (
- 1-=^) ~ Constant (19)
P
cr X t 2LX + <text J \ L >/ cr l 2L2 + lQJ \ L7

If we assume there are two functions en and gp which make the above
approximation identities, then we can write

C r t 2L
1 + *ext ; U j
" - t 2L
2 +
\xt J
^ bis

ML + L
-P S 1 2 } ^ext + 2L
i e ! + 2 V 2 yiN
cr I 2( L l + L 2 ; + 2^ext Aj

.. 2l3 ,_ .
ex !
•^ i "t
= P -i-r s r + e = Constant
cr L L + </ J
ext

In essence, when e is regarded as the error, this is the formula

previously derived. Hence, we see that the zero shear point is an
important point on the beam but we note that in the previous
investigations this fact was masked.

We proceed now to reexamine the question of the subcritical

propped beam. We begin by reflecting the beam and considering the

equivalent length to be the distance between the zero shear points.

This, of course, differs from the true length and thus we anticipate

some correction will be required. As before, the natural correction

should be the ratio of the squares of the equivalent and real lengths.

With these ideas in mind the analogous expression should be

 33

(20)
^ rff~ ®- *ext

This semi-empirical formula turns out to be a very good approximation,

the maximum error being of the order of 10 percent. The values of

the theoretically exact and approximate P 's together with the

appropriate errors are shown in Table 6.

Now it has become clear that the characteristic distance between

inflection points under a uniformly distributed lateral force is

an important parameter in the mathematical description of the influence

of boundaries on instability load of columns. So, at this stage it is

worthwhile examining the predecessor's work, i.e., "P-deltaM and

"P-theta" methods, to find out whether they can be associated with

the same quantity. To clarify this point the deflection, slope and

moment due to a concentrated lateral force acting on an end rotationally

restrained beam with unyielding lateral supports was derived. The

derivation is given in section (c) of Appendix A. The analysis shows

that when the rotational restraints are equal, the sum of the absolute

values of the extremum slopes, which was used as a parameter in

"P-theta" method is proportional to the. square of the distance between

the inflection points. Therefore, in this case, the "P-theta" method

is equivalent to the following relation

2
2
(-£) • * «t = T ^
 3^

where -L is the distance between the two inflection points when a

exrj
concentrated lateral force is applied at the mid-span of the column

considered.

Won-Uniform Cross Section

So far in the analysis the bending stiffness of the column has

been assumed to be uniform. Now the question is "Is the uniformity of

the bending stiffness essential to the issue presented here?" With

regard to this question, Reference [7] showed one example of the

variable bending stiffness problem for the clamped-pinned boundary

condition and proved the applicability of the so-called "P-delta"

method. In this thesis a simple empirical formula relating the

buckling load of the column of variable bending stiffness to the

extremum slopes due to a uniformly distributed lateral force is

investigated extensively. As a first example, the simply supported

strut depicted in Figure 6 is considered. This strut is non-uniform,

the two halves having different bending stiffness. According to J.

Case [11] the critical load for a strut unsymmetrical about the center

is given by

T- - r+ k ^
cr a b

where P is the buckling load of a strut with two halves like OA, and
3/
P is the buckling load of a strut with two halves like OB. Since the
 35

V—* x

V)
w,. mm
Figure 6. Non-Uniform. Beam Unsyrametric About i t s C e n t e r .
 36

inflection points are always at the ends for this configuration, and the

signs of the slopes at the two ends are opposite, the sum of the

absolute values of the extremum slopes is their algebraic difference.

When the uniformly distributed lateral force of intensity q (lb/in) is

applied to the column considered, the governing differential equations

are

2
E^W" ^ X - l x 0 £X S § (23)

El2 w.. a£ . i (x - if fe| * X £L

Integrating both equations in (23) yields

2
EI^' = ^ X - | x 3 + C1 0 £ X £ | (2k)

EI
2W' = 2
T X
"'6 ( X -| ) 3 +C
2 |* X
* L

The condition of slope continuity at X = L/2 gives

(25)
(> ->) - A - Mfc> MS)
Hence
 37

E(v
'x=L- W 'x=o)=€(i7 + i;) (26)

Equation (22) can be re-written as

P - iGL £ (27)
cr
* (^t
and this in conjunction with equation (26) becomes

P
cr • *> = ^ <28>

Now equation (28) can be re-written as

P • A6 2
^ 2 T - ==2* ™

which is, of course, equation (12) precisely.

The second case considered is the clamped-pinned column as

depicted in Figure 7» The bending stiffness, in general, can be

written as

EI(X) = EIo|l - %ffi (30)

 38

q_=constant

H nnmn X= L
4
X= 0

Figure 7. Clamped-Pinned Beam of V a r i a b l e Bending Stiffness

under the Uniformly D i s t r i b u t e d Load

q=constant

Figure 8. Simply Supported Beara of Variable Bending Stiffness

under the Uniformly Distributed Load

 39

where T, a number greater than 1, represents the taper ratio of the

bending stiffness and EI is the bending stiffness at the root.

o

The slope at the pinned end, which is of prime interest here,

due to a uniform lateral force is obtained by integrating the differ-

ential equation

^ ( E I ( X ) W " ! "q = ° (31)

ax"

and staisfying the boundary conditions

W = w1 = 0 at X = 0 (32)

W = W" = 0 at X = L

The slope as a function of T was calculated for the p = 1 and p = 2

cases. The details are given in Appendix I). The buckling load was

calculated by the finite difference method [12] using Potter's [13]

method and its modification [14]. The set up of the finite difference

equation, the computer program in Fortran language, and the numerical

results for p = 1 and p = 2 cases are also shown in Appendix . The

buckling loads, ratios, the tip slope parameters and the products

of the two quantities are listed in Tables 7 and 8, for different

values of T. The results clearly confirm the excellent accuracy of the

empirical law (ll), viz.

 1+0

Table 7• Propped Cantilever Beam with Linearly Varying

Bending Stiffness

(1) (2) (3) (1+) (5)

Buckling Tip Slope
Load Ratio Parameter

T PL"5 end Product Error ($)

-5 _3 (2)x(3) in Equation
n 2 EI &- (11)

1.053 .810 1.1+66 1.187 18.7

1.071 .81+6 1.387 1.173 17.3
1.111 .911+ 1.259 1.151 15.0
1.125 .931+ 1.221+ 1.11+3 11+.3
1.250 1.085 I.023 1.110 11.1
1.500 1.277 .81+6 1.080 8.0
1.750 1.1+30 .760 1.087 6.7
2 1.1+92 .710 1.059 5-9
3 1.687 .618 1.01+3 1+.3
5 1.831+ .563 1.033 3.2
10 1.9^0 .511+ 0.997 -.3
20 1.993 .508 1.012 .6
100 2.031+ .503 1.023 . 1.1
 Table 8. Propped Cantilever with Parabolically
Varying Bending Stiffness

(1) (2) (3) (10 (5)

Buckling Tip Slope
Load Ratio Parameter

249 Product Error (fo)

end
cr
2
rr EI
<d
EI
(2) x (3)
o

1.053 1.182 1.035 1.223 22. h

1.071 1.208 1.001 1.204 20.9
l.lll 1.258 .9U2 1.185 18.5
1.125 I.27U .926 1.180 17.9
1.1^3 1.293 .906 1.171 17.2
1.167 1.316 .88U 1.163 16.3
1.250 1.381* .82U 1.1U0 1U.0
1.750 I.606 .67U 1.082 8.2
2 1.673 .6^3 1.07^ 7^
3 1.808 .583 1.051* 5.3
5 1.907 .5^5 1.039 3.9
10 1.971* .521 1.028 2.8
20 2.011 .510 1.026 2.5
100 2.035 .505 1.028 l.k
 i+2

^ • {¥) - 4 <*>
bis

The maximum deviation of the product of the critical load and

slope parameter differs from the constant

(4)
by only a few percent for all practically realistic values of T.

The third case considered is that of the simply supported beam,

(Figure 8). The bending stiffness is assumed from the form given in

equation (30) . Linear (p=l) and Parabolic (p=2) cases are considered for

different values of T. The extremum slopes are located at the ends and

are of different signs. Thus, the sum of the absolute values of the

extremum slopes is the difference of the end slopes which are calculated

in Appendix . As before, the buckling load was obtained by a finite

difference scheme . The appropriate computer program is given in

Appendix . The only deviation from the clamped-pinned case lies in

the boundary matrix corresponding to the left end. The empirical law

P J JLJJ Lii = an,.

cr I 2L J ^+

was checked for this configuration. The values of critical load and the

slope parameters and. the product of the twe are listed in Table 9 an(i

10 for different values of T- It is readily seen from these tables

 1+3

Table 9» Simply Supported Beam -with Linearly Varying

Bending Stiffness

(1) (2) (3) w (5)

Buckling Tip Slope
Load R a t i o Parameter

T P L2 12 AG Product E r r o r (%)
cr
L3 (2)x(3) in
TT2EI Equation ( l l '
EI
0

1.053 .1+29 2.1+1+0 1.01+7 h.7

1.071 .1+1+2 2.31+3 1.036 3.6
1.111 .1+72 2.179 1.028 2.8
1.125 .1+82 2.133 1.028 2.8
1.11+3 .1+93 2 .080 1.025 2.5
1.250 .5I+8 1.853 1.015 1.5
1.500 .636 1.581+ 1.007 .7
1.750 .690 1.1+1+8 0.999 -.1
2 .731+ 1.361+ 1.001 .1
3 .827 1.210 1.001 .1
5 .897 1.111+ 0.999 -.1
10 • 9^9 1.053 0.999 -.1
20 .971+ 1.025 0.998 -.1
100 • 995 1.005 1.000 -.0
 kh

Table 10. Simply Supported Beam with Parabolically

Varying Bending Stiffness.

(1) (2) (3) W (5)

Buckling Tip Slope
Load Ratio Parameter

12A9 Product E r r o r \-/ot

3 " (2)x(3) in
TT^I £=~ Equation (11;
iljj.
O

1.053 .6U1 1.660 1.06U 6.U

1.071 .651 1.625 1.058 5.8
1.111 .672 1.563 1.050 5.0
1.125 .678 1.5U5 1.0U8 U.8
I.1U3 .687 1.52U 1.0U7 U.7
1.250 .725 1.U30 1.037 3.7
1.500 .782 1.309 1.02U 2.U
1.750 .818 1.2UU 1.018 1.8
2 .Qkh 1.201 1.01U l.U
3 .900 1.120 1.008 .7
5 .9^1 1.066 1.033 .3
10 • 971 1.032 1.032 .2
20 • 985 1.015 1.000 .0
100 .997 1.003 1.000 .0
 U5

that the formula (ll) is an excellent approximation. The excellence

of the formula (ll) for a uniformly distributed load cannot be denied,

but its practicality can readily be challenged. The characteristics

of a uniformly distributed load are relatively simple to analyze but

relatively difficult to achieve in the laboratory. Admittedly a

uniform load can be regarded as the integrated effect of a series

of discrete loads or as that of a moving load but this clearly

presents practical difficulties. In short, uniformity is a wonderful

mathematical expedient but a poor laboratory tool. Thus, the next

step in making the work fully practical must be to ascertain whether

the uniform load can be replaced by a discrete load system without

inducing unacceptable errors . A detailed analysis of this question has

been made for the beam on unyielding supports. It shows that three

equal forces applied at the l/h9 l/2 and 3 A points of the beam can

be used to determine the characteristic length with reasonable

accuracy. The errors involved in this determination are depicted in

Table 11 while the effects of other multiple discrete load systems are

portrayed in Figures 9 through ll+ . The use of a three force system

simplifies the question of load application, just as characteristic

length used in the analysis improved the ease of determination of the

necessary length parameter.

 1+6

Table 11. Estimation of the Buckling Load by Applying Three Equally

Spaced Concentrated Lateral Forces on the Rotationally
Restrained Beam

Stiffness Parameter (KL/EI)

3
0 .5 1 2 5 10 50 100 10J 10"
1.000 1.091+ 1.175 1.306 1.548 1.730 1.968 2 . 0 0 6 2 . 0 4 2 2.01+6
1.000 1.116 1.208 1.345 1.568 1.723 1.837 1.921+ 1 . 9 5 7 1.985
0.0 2.0 2.8 3.6 1.3 -.1+ -6.7 -l+.l -1+.2 -3.0

1.193 I.278 1.1+16 1.671 1.861+ 2.111+ 2.153 2.190 2.195

1.21+1+ 1.347 1.499 1.746 I.916 2.132 2.167 2.201 2.204
4.3 5.4 4.5 2.8 .9 •7 .5 •5 .4

1.367 1.512 I.780 I.982 2.21+3 2.281+ 2.323 2.327

1.457 1.622 1.889 2.072 2.302 2.340 2.375 2.379
6.6 7»3 6.1 4.5 2.6 2.5 2.2 2.2

1.668 1.959 2.179 2.1+61 2.505 2.5I+6 2.552

1.806 2.101+ 2.308 2.563 2.601+ 2.6I+3 2.6I+7
8.3 7.4 5.9 4.1 1+.0 3.8 3.7

2 .297 2.557 2.891 2.943 2.991 2.997

2.1+ 6 2.699 3.003 3.051 3.097 3.102
6.9 5-6 3.9 3.7 3.5 3.5

P exact 2.854 3.239 3.298 3.353 3.359

cr
P estimated 2.971 3-314 3.369 3-421 3.427
/r/cr 1+.1 2.2 2.0
% error 2.3 2.0

3.700 3.711 3.838 3.81+5

3-714 3.779 3.81+0 3.8I+7
.1+ 1.8 5.2 .1

3.81+5 3.914 3.921

3.81+5 3.908 3 . 9 1 6 100
0.0 -.2 -.1

3.981+ 3.992
3.973 3 . 9 8 1 10-
-.3 -.3

1+.000
3 . 9 8 8 10'
-.3
-10
.1 10 50 100 500 1000

Stiffness Parameter <v

Figure 9« Multiple Concentrated Lateral Loads Tests - - a - 0 case 4=-

-<]
 15

N=2

^
10

N=3

^ ^
o
•

t>?
-p -"""" ^ "^^ ~
CD
N=5\^
o ^ « ^ \
CD
FU

-10
.1 5 10 50 100 500 1000
Stiffness Parameter a

Figure 10. Multiple Concentrated Lateral Load Tests -- a = 1 case

o 4=-
CO
 15

v^ N=2

10
N=3

£H
O 5
y
£H
!M

H
,^
fl
<D
O
W
<L)
Qj
n N=J? >

-5

-10
.1 5 10 50 100 500 1000
Stiffness Parameter a.
1

Figure 11. Multiple Concentrated Lateral Loads Tests -- oi = 5 case

-P-
VD
tl
o
*H

£3
-4J

<D
o
<u
PM

-101
.1 .5 5 10 50 100 500 1000
Stiffness Parameter a.
Figure 12. Multiple Concentrated teral Loads Test -- a = 10 cass
o
o
 15

N=2

10

fn
O
fn
fn

N=3
CD
O
fn
CD
Q_,

, N=5

N=4

110
.1 .5 5 10 50 100 500 1000
S t i f f n e s s Parameter a
1
Figure 13. Multiple Concentrated Loads Test -- o- = 100 case
o

H
 15

N=2

10

S 5
in

-P
N=3
(U
o
in
(U
^ 0
\^N=5

- " "

-5 ft=4 \ ^

-10
.1 .5 5 10 50 100 500 1000
Stiffness Parameter a
1
Figure 14. Multiple Concentrated Lateral Loads Test -- o^= 1000 case VJ I

ro
 53

CHAPTER III

GENERAL INTEGRAL EQUATION FORMULATION

Formulation of the Integral Equation

The work in the preceding sections as well as that in

reference [9] has given very clear indication that the instability

behavior of a partially restrained column can be readily associated

with the deformations of the same body under lateral load. We are

led, therefore, to the thought that a clearer understanding of the

relationships established might be derived from an analysis of the

beam column. To this end an integral equation approach to stability

load level determination is developed in this chapter.

In formulating the integral equation for column buckling, it is

convenient to introduce the idea of equivalent buckling force. Con-

sider the elastically supported beam as beam-column shown in Figure

15.

The work done by the axial, force P, is

w e = P J J ( w ) 2 ax (i+i)

v
o

The virtual work of the axial load is

 q(x) Z,W

X ,X

Coordinate

X= 0 X= L

F i g u r e 15 . E l a s t i c a l l y Supported Beam Under Combined Loading.

VJl
-p-
 55

L r L
6W = P f W'6W'dX = PW'6W - P I WM6WdX Ct2)
e J <J
o 0 0

Interpretation of equation (42) becomes clear when the variation of the

external potential is written for a beam under a load distribution q(x)

and shear forces S(0) and S(L) on the boundaries. That is

| L L
6W = S(X) 6W + q(x) 6WdX (43)
e t]
1
o o

Comparison of equation (42) with equation (43) shows that in this

problem there are equivalent force systems. The system with which we

begin can be replaced by the distributed moment -PW" upwards and

additional boundary shear forces, PW'(L) upward at X = L and PW'(O)

downward at X = 0.

Thus the effective distributed load q „„ acting upward is

el i
defined as

q ff = -PW"(X') - PW,(X,)6(X') + PW'(X')6(X,-L) (44)

= -P[W"(X') +W,(X')6(X') - W*(X')6(X,-L)]

where 6(X') is a Dirac delta function.

In general when the loading q(X) is distributed over the beam,

the deflection at X, w(X), can be written

 L
r
w(x) = cCx^OqCxOdX' (^5)

where C(X,X') is the flexibility influence function, interpreted as the

displacement at X due to a unit lateral force at X' .

For the sake of simplicity in the analysis which follows,

non-dimensional quantities defined below are introduced.

v X r XT w ^ PL2
b x
L ' L ' " L ' " EI

i(c) = r . e(x,c) = ^

Equation (M+) and (U5) can now be rewritten as follows

*eff = "^ w "(0 + w'(rj6(c)-w'(c)6(C-D] (U6)

r :L -
w(x) = I C(x,£)q(£)d£ (W
o

Replacing q(^) in equation (k'j) by q ff , the following relation

is obtained.
 57

r 1
w(x) = J C(x,G)qeff(Q)&Q (48)
o

- r1-
P I c(x,c)[wM(c) +w(c)6(c) - w f (c)6(c-i)K

= -p C(x,£)wM(£)d£ - PC(x,0)w'(o) + PC(x,l)w'(l)

o

Integration of the first term in equation (1+8) by parts yields terms

which cancel boundary terms, and when this is done the resulting

equation is,

ac
w(x)=pT ~(*>0 W '(e)dc (U9)
"' o ^

It is convenient for some application to write the previous equation

in terms of the slope. Taking the derivative of equation (49) with

respect to x, an alternative form of this equation is obtained.

1 2-
WW = p r ^4fe^ w , (c)^ (50)
«- ^ OXOL
o •=

Since no specific restriction on the boundary conditions was

imposed in deriving equations (49) and (50), these two equations are

valid for all boundary conditions. Here, for the sake of convenience,

the following notations are introduced.

 58

•<*,C) • % ^ (5D

e(x) = w"(x)

^ 3C

where r(x,£) and d(x,£) are interpreted as the slope and deflection,

respectively, due to a unit moment applied at the location Q . It

should also be noted that r(x,£) is symmetric with respect to x and £,

as can be shown from the reciprocal theorem.

an(
Using these notations, equations (^9) i (50) can now be

written as follows:

- r1
w(x) = P d(x,C)8(C)dC (52)

8(x) = P J r(x,£)e(G)d£ (53)

Orthogonality Relations

For the ideal boundary conditions, namely, simply supported,

built-in or free, the following orthogonality relations are known to

be true [15].
 J w n W m ax = o m
O

and

rL
j Ei(x)w,,mw*,ndx = o (55)
"o

where W and W are the mth and nth buckling mode associated with the
m n
buckling load P and P respectively. Equation (5*0 is, of course,

a general result. The detail of the derivation is given in Appendix

A. The orthogonality relation (equation (10)) in Appendix B can be

rewritten as follows:

r 1
I 9.(x)9.(x) dx = N.6.. (56)
J l 3 l lj

where

, 1 2
N.l = 9 .(x)dx
l • '
o

6.. = Kronecker delta

,=0 when i ^ j

=1 when i = j
 Combining the orthogonality relation (equation (56)) 'with the

integral equation (53)} a different type of orthogonality relation can

be derived as follows. Let 0 , 0 by the slopes of the rth and sth

r s
buckled mode corresponding to the buckling load P and P respectively.
i S

Then from equation (53)}

(x) = P [ r(x,£)9 (£)<!£ (57)

9 (x) = ? r x (58)
s s J ( >G)e s (c)ac
"o

Multiplying (57) by 9 , (58) by 9 , and integrating over the length of

s r
the beam,

1 1
9 9 dx = P r(x,£)e (C)ejx)dxd£ (59)
J r s r
o o o

! e e dx = p Q r(x,c)es(c)er(x)dxdc (6o)
j r s sJo
o

In view of the symmetry of r(x,£), the above integrals are identical

on "both sides of the equation,

lr 1 N
I r(x,c)eTr.(c)efl(x)dcdx = =£ 6 r s (61)
oJo = s
 6l

Equation (6l) yields a very interesting result when r is equated to s

That is

E
K -— ^
i j r(x,c)e (c)e (x)dxd^
"o o r s

'1 2
9 (x)dx
r
o
p 1 r 1
j I r(x,Q)B (Q)Q (x)dxd£
o o

This formula could be used to estimate the buckling load of the

elastically restrained column.

Some Useful Results Derived from the Integral Equation Formulation

With regards to the kernel r(x,£) of the integral equation (53):

the following observations are made:

(1) The kernel is symmetric with respect to x and Q.

(2) The kernel is positive in the following sense.

, 1 r 1
I=| | r(x,£)9(x)9(c)dxd£ ^ 0 (63)
o o

where 9 is an arbitrary function. Equation (63) holds true because the

integral quantity can be shown to be proportional to the strain energy

which is obviously positive. At this stage it is convenient to

 62

introduce orthonormal functions cp. defined by

e (x)
cp^x) = - i (6k)
/ \

For i = j equation (56) can be rewritten as

>1 2
cp. (x) dx = 1
o

Because of the conditions given above for the kernel, it follows from

Mercer's theorem [15] that the kernel can be expanded in terms of the

orthonormal functions and the eigenvalues as follows:

" CD- ( C W M
r(x,£) = ) -i -2= (65)
p
i'i i

Equation (65) yields many interesting relations which connect the

buckling load and the behavior of the beam under the non-destructive

force system. "When it is combined with the orthogonality relation

(equation (6l)) it gives two kinds of infinite sum relationship for

the eigenvalues.

The first is obtained by letting x = £ in equation (65) and

integrating over the span

 r l *± M 1
r(x,x)dx = ; I — dx = ) (—J
-J o P.N. V
.^ V
i=l i' i i=l

the second by integrating r (x,£) with respect to both x and

„ 1 n 1
j" r^Oaxdc-Vy-)/ e^WtoJ 6^(^=7 (
O O 1 1 O
i=l i=l

These two results will be discussed later. Other interesting

can be derived from equation (65) . By definition

g
r(x,r) = a c(x,c)...f 9 i W 9 i^
SxdC ^ pF
i=l i i

Integrating with respect to Q gives

3£l2L£l = ) -i 1 + f (x)
^ ."-; p . N
i=l 1 i

Integrating again with respect to x gives

 V
C(x,C) = I i(x)Vi(C) + F(x) + G(C)
i=l P.N.
1 1

From the reciprocal theorem?

F(x) = G(x)

Then equation (70) c a n ^ rewritten as

w (x)w (C)
+
foO =} ~ F(x) +F(C)
. ,
1=1
P. N.
1 1

Suppose the boundary restraint is such that the lateral movement

allowed at x = 0, then in equation (7l)j>

wi(o) = 0, c(o,c) = 0

Hence,

F(0) + F(C) = 0 for 0 <;, £ <; 1

This means that the unknown function F is identically zero. This

argument holds true also when the lateral stiffness at x = 1 is

infinite. Hence, equations (69) and (70)? with this restriction,

become
 65

ei(x)wi(c)
e(x,C)-2Sg^-y
d
^ - ^ (73)
.--;
1=1
P.N.
i i

c(x,r) = ) w i « w i W (7U)
i=l P.N.
l l

When the end lateral stiffness is finite, the function F represents

the deflection caused by the rigid body rotation. Multiplying equation

(73) "by 0.(x) and integrating with respect to x from 0 to 1 results in

the following simple relation

w
P 1 , (£)
J 9(x,^)9i(x)dx == -i (75)
P.
1

Integration of equation (75) wi-th respect to £ from 0 to 1 yields

1
r
1
P 1T- r*r-_, -|
,_!.,,_ J o w xi ^ ) d £
uu
j L J e(x;,c)dCJ ei(x)dx = ,p. " (76)
o o 1

But,

o 1
e(x) = j e(x,^)d^ (77)
 66

(Note that 9(x) is the slope at x due to a uniformly distributed load.)

Substitution of equation (77) into equation (76) gives

r1
I w. (x)dx

h = 4V^
I e(x)e.(x)dx ^

Since the lowest buckling load Pn (= P ) is of the prime interest,

1 cr
e q u a t i o n (78) can be r e w r i t t e n as

j w 1 (x)dx
f
cr=Jri ™
e(x)e.j(x)dx
o

Application and Discussion of the Results Obtained in Section 3

The main purpose of studying the buckling problem in the integral

equation formulation is to search for the mathematical background for

those empirical formulae derived in Chapter II. The procedure adopted

was as follows.

The instability problem was set up in such a manner that the

critical loads were related to factors which define a non-destabilizing

force system. Now since such force factors systems appear in an integral

quantity, there must be a reasonable probability that a certain para-

 67

meter (e.g., deflection, slope, characteristic length, etc) maybe

associated with this integral. If this is so, we may be able to

derive laws of similar form to the empirical ones.

Following this reasoning infinite sum relations (equations

(66) and (67)) were developed. These equations establish that integral

quantities associated with a couple distributed along the span can be

expressed as the infinite sums of the inverses of eigenvalues or their

squares. Since the empirical law defined earlier is related to the

lowest buckling load P , these equations cannot be the answer to our

cr
problems unless the ratios of the sums of these infinite series to P^

remain nearly constant. To clarify this point beams with equal

rotational restraint and a lateral tip spring are considered. The

summation is carried out over the first eight eigenvalues. The results

are shown in Table 12 and 13. It is clear that these ratios vary

widely. This means that the two equations do not explain the validity

of the empirical law. Since the empirical formula which uses the

characteristic length works extremely well for the case when the

lateral stiffness is infinite., it is worthwhile to investigate

equation (79) with the assumption of" infinite lateral stiffness . In

this case when the denominator is integrated by parts, the boundary

terms vanish and equation (79) becomes

r1
w dx
K.r = 71 (8°)
w w dx
J ,xx n1
o '
H O I ^ - O OOOO 0 O ( ^ 4 V D OOH 0 \ 0 \
VDVD UAUA-4-CVI H O N t ~ - t — V D V O V D LT\ UA
CTNCTNcrNa\Cha\a\cococooococoooco

H CO OO OO LT\ Q \T\r-{ OO OOVQ CO OJ -=f t—

VD ON OO OO CO-3 1 -d" U A U A O N O J V D O O O OO
CTNCTNO H O J l X N t — O 0O-=f t—CO O CO-=f
i—I i—1 i—) i—i i—I r\j OJ OJ OJ OJ OO oo oo

U \ L C3 I I i-1 \ v - > ^ \ ' I '^ SJ 1 I v • / V. VJ *-A_/ U \ \' J 1.1J

LT\ J " -3" OJ O M A O O O O Q t—VO UA UA LT\ UA
VD VD VD vo LA \s\ \x\ ir\ - J ~=t ~=t -=£ -=t ~=t -4

UAOOHHOOOH-4-t~-a\UAOJOJOO\
LT\ 0 - ON J " I-IOOOJ UA CO -d" UA OJ O - d " O
V D V D V D t — C0 0 \ 0 r l 0 | 0 0 4 - UAVO t—CO

O Q C^ OO IT-CO t— t— 0 O 4 M 3 Ol H l A O N
o-3- t— a\vo vo H a\ t— uA-d- OJ OJ -=f a\
OOOHOOVDONOJVDCO H OO UACO Q\
HiHiHHHHHOJOJOJOOOOOOOOOO

rlOl IA
• • •
O H O J O O U A C O O U A O O OUA
H H OJ O O O O
on l>- H o n v o i > - H v o H [>- LTN o ON OJ O LTN [>-
ONCOCO I > - L A O V O _ ^ | - _ ^ | - _ ^ | - V O c O ON H OJ OJ OJ
ONONONCTsChONCOCOCOCOCOCOCO ON ON ON ON

CO t > - L r \ O \ C h L r N L r N H c 0 l>- on-^h O t-\-3- O t>-

J - ( \ i Q ^ H H i ^ o o \ o a i c M ^ - a ] I A C O ON
OJ PO-3--3-V0 ON H o n P O L T W O V O |>-CO CO CO CO
H H H H H H H H H H H

C\l 0\ H [>- ON ON-^t C\l on O CO O N V O on CO VO CO

on cO L r \ H m u - \ a \ l > - v o v Q v o l>- ON H OJ - ^ LTN
CO I>- I>- I>-VO LTN-3- _^" -^" -^" -^" -^" -4" LrNLTNLTNLrN

co OJ o o n i ^ - 3 - o n v o O N O N O N H V O on-ch o c \ i
O V O H L T N O J V O I > - O J V O H I > - O J t^-OJVOH-J1
ojojononJ-LTNVOi>-i>-cocooNONOOHH
H H H H

O OJ OJ OJ CO O N - ^ C O OJ ON t ^ - = f ONVO LTNHVO
L T N o n H Q N - T o v o o n v o I > - I > - O J V O O N H on-4-
OJ on-3" -3" VO O OOLTNVO I>-CO O N O N O N O O O
H H H H H H H H H O J O J O J

O H o j o n L T N O v o o - ^ i - o o o o o o oi>-
H H ai ai M J - U N I ^ O I A O o
H H cnH
 70

Interpretation of the denominator in equation (80) becomes clear when

the bending moment diagram under the uniformly distributed load, the

buckled mode and the product of the two are drawn for two extreme cases,

i.e., simply supported and clamped. It is shown in Figure l6. These

pictures illustrate the fact that to evaluate the integral of the

denominator of equation (80), the necessary range of integral is

approximately (exact when simply supported) between two moment-zero

locations. This indicates that the distance between the inflection

points is an important parameter for the integral

r*
w w dx.
J ,xx n1
o

Although this is not a proof of the validity of the empirical.

formula, it definitely explains why the distance between two inflection

points is an important parameter in the uniform load cases considered.

To broaden the basis of study there are two questions which arise.

Is the demonstration restricted to a uniformly distributed load?

Is it possible also to illustrate the mathematical foundation of

"P-delta", "P-theta" and some other empirical formula corresponding

to a different kind of loading? As was shown in Chapter II, the

"P-theta" method can be reduced to the characteristic length relation

equation (21) . Since a concentrated lateral force is used in "P-theta"

and "P-delta" and equation (75) is the relation connecting the buckling

and the behavior of the beam under a concentrated lateral force, this

equation will be examined closely. Letting i = 1 in equation (75)?

 Simply Supported Clamped

w.

w.M

Figure 16. Illustration of the Product w.M for Two Ideal Cases.
 72

the following relation is obtained

P _ = — T ~ (81)
cr
J e(x,c)e1(x)dx

where

P = the lowest buckling

& load
cr
Q(xj^) = slope at x due to a concentrated lateral force at £

9-. (x) ., w.. (x) = slope and deflection of the first buckled mode,

respectively.

When the lateral restraint on the boundaries is infinite, equation (8l)

can be reduced to the following form:

w (£)
p = — r^ (82)
cr .1
-[ M(x^)w.L(x)dx

where M(x,£) is the moment at x due to a concentrated lateral force

applied at Q.

To make the argument simple, consider the beam with equal

rotational restraint. It was shown in reference [6] that to apply

the "P-delta" or "P-theta" method to this configuration a concentrated

lateral force must be applied at the mid-point of the beam. The

bending moment diagram due to a concentrated lateral load at x = 2",

 73

the buckled mode and the product of the two are drawn in Figure 17

for two extreme cases, i.e., simply supported and clamped.

These figures indicate that the denominator of equation (82)

can be reasonably evaluated between the inflection points.

 7h

Figure 17 . Illustration of the Figure Mw.

 75

CHAPTER IV

FURTHER CONSIDERATION OF THE BEAM COLUMN

The analysis of Chapter III has shown beyond doubt that the

inflexion point separation distance for the laterally loaded column is

an important parameter for describing the stability behavior. However,

it has not generated a process which be used to analytically derive

the results obtained by numerical analysis. It does serve, however, to

emphasize that the key to the question may well lie in a study of the

combined load problem.

In this chapter, therefore, we examine the combined load issue

from a somewhat different viewpoint - in essence we shall follow a

stiffness approach. This is,of course, common in studies of structural

frameworks .

We begin with the simplest of problems, viz the pin Jointed beam

column with a central concentrated lateral load (Figure 18) . Taking an

origin at the center, the bending moment at the point whose coordinate

is X is,

M = |Q [| - X] + FW = -EIW" (82 )

Let

k 2 = |r (83 )
 76

Figure 18. Pin Jointed Beam Column with a Central Concentrated

Lateral Load.
 77

thereforej

W M + k 2 W = - 3- r£ - £-1 (8U)
EI lh 2J

The solution is

W = A sin kX + B cos kX - (|) (^ - |) (85)

The "boundary c o n d i t i o n s a r e

at X = 0 V = 0 (86a)

at X = |- W = 0 (86b)

Using t h e s e "boundary c o n d i t i o n s i n e q u a t i o n ( 8 5 ) , we can w r i t e

_ QL_ i" sin kX tan \ kL cos kX /l X\~

W 2P L" kL kL \2 Lj_ (87)

The maximum deflection occurs at X = 0, and is given "by

= QL ftan \ kL ' (88)

max 5 P LlrkL " 1

It can "be seen that as P approaches (TTEI)/(L ) , kL approaches Tt, and

the application of a very small lateral load causes a very large
deflection of the column. The stiffness of the column against lateral
displacement due to a transverse load at its center is

Q ^P jkL . , (89)
W = L ItanfkL-iklJ'
-T-14. — T , - T .1.-T r
u <«.ueiiipivj_i-p.
max ^ *

The variation in stiffness with end load is very nearly linear.

When we treat the somewhat more complex case of the column

with built in ends we find that for this case

Q , kL Q/l kl L", /Q0 N

W = tan + tan (90)
max 2k? • T ]?tS ~ - IT!

or

W = s— - tan —r- - —r-H (QI j

max y ±J
kP I IT UJ y

and so we have that

-3_ = (kLA) . !+P (92

Wma^ tan&L - & L L ^

It is clear that if we change the variable k to k-. such that

k = 2^ (93

Then we may write

 79

^~ " (tan \ L - V ] • T ^>

max

and it is clear that the two stiffness lines are parallel curves, and

the ratios of their intercepts with the axis are thus constant.

That equations (88^ and (9^) are to a close degree linear can

be demonstrated in the following manner. Expand the tangent

tan
~l _tan •;ir _, . i ;PL 2 ; + ^ P L 2 . 2 . IT,'PL2 3+ . . . (95)

~ V§f
Then, the expression for W , equation (88) may be written
max

QL f l /PL \ ^2 /PL2\2 17 /PL2N-3 '; , ..

W
max = UP \3 V&[) 15 \^J 3l? \ E i + • • • J (96)

or

- ^ /•, + 3 . 2 /PL N 3 . 17 f a ', . "L ,„,.,

W 1+ +
U-H!l 1 15 VUEIJ I 315 VHEI; + ' • •J (97)

This is convergent if

2 2 2
PL PL
£ *- 1 • n « nf ^ 5 ^ TT
5 ST ' IffiT 2 T

or if P < P (98)
cr
 80

In these circumstances the series may be replaced by a geometrical

progression.

3 2 2 2
0L PL / PL ^
= +
V* 55EI I1 + ioEi+ Kvm) • • •I (
»)

This summed to infinity, gives

3
W = Qir 1
Lliax
km Pi2
i
10EI

QL3 1
I

U8EI P
H

Pcr

or
P
w = «•! cr
(100)
cr

The pin jointed strut with a uniformly distributed lateral load

of intensity q_ and carrying an axial load of P is a classic problem.

For this the bending moment at any point is given by

M = ^u '1 - coskx sec -~j 0-01)

c
k "

with the convention given in Figure 18 . Thus

W =0 = ^ ; (l-cos kx sec — ) d x + constant (l02)

k
 81

Now the slope is maximum at the ends and zero in the center, so

q q kL . .
= -p x + -o s i n kx sec -r- (103)
k k

Thus

k k

or
.
tan kL
-—
w B hi 1 + — k i — j ^ )
^k —

This is very close to a linear relationship between compressive

force and slope.

As the ends become restrained the expression for slope parameter

becomes increasingly involved. For the encastre strut with uniform

lateral load, the bending moment is given by

M = (Mf - - 2 ) sec - | cos kx + ^ (106 )

k k

where

«• - * 2 f1 - t H ^ ^

Thus the slope expression is

= r[M f - %>] sec ^ sin kx + % x (108)

* kd d
k^
 82

and the maximum value occurs when

-1
2
k
cos kX = - ? — r=— (109)
[W - \ sec £§•
k

The expression for G/L becomes most complex and can only be evaluated

numerically. When this is done we find that the relationship between

G/L and P is again linear and the slope of the line closely approximates

that of the previous case.

We note further that when we examine the case of the column with

an end moment the relationships developed differ from those previously

desired. For the pin-ended column (Figure 18) we obtain

M
^ = (5i) s(i - c2) (no

where

= (1 - kL cot kL)J5kL ,
S* = tan^kL-^kL
fnnilrT _-±-VT. I111 .

and

ru, kL - sin kL ,
C* = —: nip
v±
sinkL-kLcoskL •'

* These quantities have been termed stability functions, see reference

[173-
 83

The variation of stiffness with end load is given in Figure 19.

For the fixed ended column (Figure 20(a)) the relationship is a little

simpler, viz

Q" = (—) 'S (113)

2 °
The graph of M/6 against (P)/(lt EI/L'~) is shown in Figure 20(b). We note

immediately here that the curves are no longer parallel and thus we

should not anticipate being able to generate a viable relationship.

Pure numerical techniques as used in the first part of this thesis

failed to generate any results of value. It seems likely then that the

reason why the simple relationship can be found lies in the fact that

the variation of the pertinent stiffness parameters with respect to the

applied compressive force is essentially linear and the lines have

a slope which is independent of the degree of end fixity. The complexity

of the expressions is however such that at this time we have not been

able to verify this by other than numerical techniques. The reason

is clear. It is not easy to find simple, accurate algebraic approxi-

mations to the various transcendental functions involved. Whether or

not it will be possible in due course to demonstrate these conjectures

is open to question. Approximation analysis, like engineering, is

after all partly science, partly art.

 Qk

P >M
—•V"
(a)

0.25 0.50 O p \L0

^kL=- TT / p / Q
-I 1 1_
1.57 2.22 2.72 TT
(b)

Figure 19. (a) Pin-ended Column, (b) S t i f f n e s s Plot

4.45kL
00

Figure 20. Fixed ended column

 85

CHAPTER V

DESIGN FORMULAE FOR COLUMNS WITH ENDS PARTIALLY RESTRAINED

AGAINST ROTATION

Now that a practical method for determining the fixity which

exists at the ends of columns has been devised it is important to the

practising engineer that simple formulae for critical load in terms

of end restraint be avilable.

The simplest of these laws, clearly, emanates from the P6 law.

It is shown in equation (32) Appendix A that for the column with

equal end constraints that the deflection 65 due to a unit central

lateral load is

a + 8 1 . L£
a + 2 * 192 El'

where a = K/EI; K being the rotational stiffness of the end springs.

Thus, the critical load for such a column is given by

h-n2-EI (a + 2"i , . v
p = . J___, (±1^)
cr 2 la + oj
J_i

and, in the more general case of unequal end fixities a,3 by

 86

This latter result follows by virtue of the fact that the critical

load for a column with unequal end rotational restraints a and (3 can

be shown arithmetically, to a close approximation, to be the geometric

mean of the critical loads of columns with end restraints a and |3,

respectively.

The derivation of an equally simple formula from the relation-

ship between critical load level and the distance apart of the extremum

points is not feasible. The complication arises from the fact that

the expression for this quantity is a surd in the end restraint

parameter.

It is of interest to note that when a Rayleigh quotient

approach is taken to the solution of the same problem the following

formulae can be derived:

2
P Bm..« + i°« + i£ (ll6)
cr _£ d _,, ,-,
L a + 14a + 64

and

2
p J±2EI . a + 14c + 24 (11?)
cr
l/~ & + 18a + 102

The former (n6) is derived by taking the column buckling deflection

function to be identical with tnat due to a concentrated lateral load

applied at the mid-point of the column, the latter on the assumption

that the deflection curve corresponds to the displacement produced

by a uniformly distributed side load.

Comparison of these Rayleigh formulae and the exact results

 87

derived numerically from the characteristic equation is made in

Table ik .

It is clear from this table that the first expression is the

closest approximation. It is interesting to note that the percentage

of error is very nearly constant over a wide range of end stiffnesses.

This suggests that we might divide the constant by a number close

to unity and obtain an adjusted formula of greater accuracy. Indeed,

when the formula is written as

+ 10t
P =^ .% *+ l6
(118)

the maximum-error is reduced to .6 percent as can be deduced from

Table Ik, and of course the extreme values (a •- 0 and a = °°) are
exact.
This is an interesting expression because it can be written

,cr B A . (^§1 (119)

L (of+8) "-2a

2
^TT EI . a+2
L2 ( a +8) 2a

or+8

I
m2EI . g+2 L . Jte X
2 +8 u ;
L" - " ' W
2cv
Wow clearly 0 is small for all values of a between 0 and »,
(a+8) 2
 88

Table 14. Approximation by Rayleigh's Quotient - Rotational Restraints

Concentrated Uniformly
2 L a t e r a l Force D i s t r i b u t e d Load
P
X L
J_l
KL P 2 P 2
EI cr est L „ /tlf\
—s E r r o r (%) —'-75 E r r o r (%)
TT 2 EI 2
TT EI TTEI
0 1.000 1.013 1.3 1.013 .1
0.1 i.o4o 1.054 1.3 1.042 .1
1 1.367 1.385 1.3 1.372 .3
2 1.669 I.689 1.2 I.678 o5
3 1.921 1.938 .9 1.934 .7
5 2.298 2.320 .9 2.334 1.5
8 2.683 2.702 .7 2.745 2.3
10 2.854 2.880 .9 2.941 3.0
15 3.147 3.176 .9 3..272 4.0
20 3.327 3.356 .9 3.475 4.5
30 3.527 3.561 1.0 3.709 5.2
4o 3.633 3.671J- 1.1 3.837 5.6
50 3.702 3.7^5 1.2 3..918 5.8
6o 3.7^ 3-794 1.2 3.973 6.0
80 3.810 3o856 1.2 4.042 6.1
100 3.845 3-894 1.3 4.085 6.1
1000 3.984 4.037 1.3 4.238 6.4
10000 3.998 4.051 1.3 4,254 6.4
 89

Thus, the expression of equation (ll8) is reduced to

^TT EI . a+2 /1lM,.

P = x— * —-?r (114) blS
cr 2. a+b

which is identical to that previously derived from the P6 law.

The approximation formula given in equation (117) can be like-

wise adjusted. To do this we again ensure compliance at the extremes

and hence write

p = ^EI {^lJ^J_2k) (119)

cr 2 2
L of + 18a + 96

which after suitable algebraic manipulation becomes

p
cr = ^ 5 ^ H i + (^fey} d2o)

which for all practical purposes is again

P = iisfsi (S^j (nU) MS

cr Jd. oi+o

2a
since ->—-75-77—,-,oV is
,y very small for all values of a.
(a+o)(o?+12)
Thus we can see that the P6 law and in all probability the

others also would be analytically derivable if we could develop

approximation method for deriving or expressing all relevant stability

and other parameters in simple algebraic terms .

 90

For the cantilever beam with a lateral tip spring neither the

uniformly distributed load nor the concentrated lateral force deflection

curves give a reasonable estimate of the buckling load when used in the

Rayleigh formulation. This is not surprising because neither curve is a

close representation of the deflection shape. The deflection produced

by an end couple is much more similar to that which results from

instability. "When this deflection shape is used a quadratic rational

function can be developed for the critical load viz

g + + 36
cr ° 30 f2 ^ } f (121)
(3 + 15P + 36oJ if

It is clear that this expression is a poor representation because

when p = 0 the value must be (TTEI)/(^L ) whereas the expression in

equation (12l) gives (3El)/(l/~) . Similarly when p == » the critical

2 2 2
load is(2.05rr El)/(L ) whereas the above formula yields (30El)/(L ) or
2 2
(3TT El)/(L ) . The maximum error in the estimation, in fact, amounts

to 50 percent. However, when the approximate expression is plotted

against the exact values on a log-log plot it is found that

+ + 36
P = .740(30 \ ^ f12 (122)
B + 15P + 360

is a very good approximation. The log-log plot is shown in Figure 21

and a comparison between the approximation of equation (122) and the

exact values is given in Table 15.

 91

Table 15. Approximation of the Buckling Load Using Rayleigh's Quotient'

Modified-Lateral Tip Spring Case.

App r oximat ion

3
KL According t o E q . ( l 2 2 )
T1 T
crt
EI EI P L2
est Error(^)
EI

0 .250 .250 .0
0o5 .291 .293 .7
1 .332 .336 1.3
2 .412 ,k22 2.5
5 .648 .673 3.9
8 .870 .895 2.9
10 1.009 1.025 1.6
14 1.257 I.238 -1.5
18 1.457 1.401 -3.9
20 1.538 1.466 -4.6
24 1.662 1.573 -5.3
30 1.779 1.689 -5.1
ko 1.877 1.808 -3.7
50 1.924 1.878 -2.4
100 1.996 1.995 -0.1
500 2.037 2.044 .3
1000 2.042 2.045 .2
 3.CI

P L4
2
7C EI

(Rayleighs Q u o t i e n t ) x
f«EI

Figure 21. Correlation Curve between the Rayleighs Quotient and the
Buckling Load for the Lateral Tip Spring Case
 93

CHAPTER VI

CONCLUSIONS

The analysis recorded in this thesis demonstrates clearly that

simple relationships which couple the stability behavior of a column

under axial compression and the distortions of an identical member under

lateral force can "be developed. These relationships appear to have

importance in the non-destructive evaluation of existing column structures.

It seems likely that they also provide a means whereby designers of

column type structures, e.g. civil engineers might systematically develop

end fixity systems of known performance which could be standardized.

This would, of course, lead to an improvement in design.

Of particular interest in the format of the equation developed is

that in many apparently four variable problems a single parameter,

associated with the transverse loading situation, can be found to

describe the critical load. This condition is not restricted to a

single lateral load distribution but appertains in general. The

parameter, however, distinctly varies with the distribution.

It is also pertinent to point out that for defined end fixity

conditions the equations developed in the thesis are frequently capable

of giving practicing engineers an approximation to critical load values

as expeditiously as the Rayleigh-Ritz procedure and to an equal

accuracy.

An effort to explain the success of the process led to further

 9^

study of the applicability of integral equation in the evaluation of

eigenvalues for the column stability problem. This study showed that

this technique is very applicable to the issue. General relationships

that may be at least as powerful as the normal Rayleigh-Ritz expression

were derived.

Although the empirical laws were not positively shown to have

a solid foundation by analytic processes, the mathematical derivations

clearly indicate that this would be demonstrated if appropriate

approximation techniques could be developed.

 95

APPENDIX A

ROTATIONALLY RESTRAINED BEAM ON LATERALLY UNYIELDING END SUPPORTS

Establishment of the Characteristic Equation for Instability

For the configuration shown in Figure 22a, the non-dimensionalized

buckling equation is

w I V + X2V* = 0 (1)

The appropriate boundary conditions are

w = 0, w" - a w' = 0 at x = 0 (2)

o

w = 0, w" + a w' = 0 at x = 1

The general solution for equation (l) can be written as

w = A sin Xx + B cos \x + Cx + D ' (3)

The solution in conjunction with the boundary conditions yields the

following buckling determinant:

 96

K, K.
Uniform. EI

7777777- 77777777

(a) Buckling

q=constant

K, .niiiuumnnmn K.
•rm

(b) Uniformly Distributed Load

Figure 22. Rotationally Restrained Beam on

the Unyielding End Supports.
 X a? c
sin X cosX-1 1 = 0 (k)
o^XcosX-X sinX -a^XsinX-X cosX a

Expansion of this determinant gives the characteristic equation

^ ^ ( l - c o s X ) - XsinX] - {aQ + a )X(XcosX-sin\) (5)

+ X^sinX = 0

Extremum Slope Relationship Under Uniformly

Distributed Lateral Load
The deflection and slope under a uniformly distributed lateral
load are derived as follows, (see Figure 22) If a non-dimensional
uniform load is defined as

L3
(6)
^ EI

then the non-dimensional differential equation for deflection w is

w - q = 0 (7)

and the appropriate boundary conditions are

 w = 0 and w" - a w' = 0 at x = 0 (8)
o

w = 0 and w" + a-.'W - 0 at x = 1

Successive integrations of the differential equation (7) give

w »'= Jx + A (9)

1-2
w" = ^ qx + Ax + B
1 - "3 A 2
w' = £• qxJ + gx + BX +C

I - I4. A ^ B 2
w = 7% qx + gxJ + -x + Cx + D

where A, B, C, and D are constants. These four constants are obtained

by satisfying the boundary conditions of equation (8). They are thus

determined to be

A = - § (1 + -2^-i) (10)

B =
12A" a o ( o ! l ^

c _ 1
= 35A (V 6)

D = 0

where A = 12 + 4(a + a ) + a a,.

 99

Substitution of (10) into (9) yields the following result

- - o a - OL a (OL + 6)- /nnN

- o a -<*, o a ( a . + 6) en + 6-
«x) = £ L2x3 - 3(1 + -W + -Si x + J^_

- ,•• 1, a - a a (a. + 6) 0 an+ 6 _

V(X) . ^ ^ . 2(1+ ^_^)x3 + JLJ^? + 2 j ^

where w(x), 9(x), M(x) are the deflection, slope and moment, respectively,

The locations of the extremuiri. slope are obtained by solving M(x) = 0.

The result is

*=UZ± W M

where

z = 1 + -S-J,

and

I
ext = / a? -*, 2 2 ^(a-,+6) (13)
V(1 + ^i).| JL-i-
= the distance between two inflection
points
 100

Define

x
l = ¥Z - lext) W

X + l
2= ¥ eJ

The meaning of x., x , i , , and Z are easily understood from Figure

1 2 ext
23 . In equation (12) in the main text, the significant parameter is
the sum of the absolute values of the slopes at the zero moment points.

We seek an alternate method of expressing this quantity. Hence,

2^(9 _ -9 ) is to be evaluated. From (ik), the following
relationships hold:

x± + x 2 = Z (15)

X
2 " X l " ^ext

x1 • X2 = ^ - 4t> = U ^

Thus

2 2 -
x_ - xn = Zl (16)
2 l ext

tt (a
3 3 3 + J o l + 6)
x
2 x
" l =
* ext + lext ~L
 101

x=0 x=l

Figure 23. Illustration of x , x I and Z.

Making use of this result,

•- P ^ (or., + 6) o 12c* (a-, + 6)

2 2
2i(e - e ) = I J hz - - ° i 6z + — 2 i J
- xo x-i extL oA oA
q

p V (or + 6).,
,1 -£Zi^ +, O —1 i =_ -dJL,
n.J
extL 3A J ext

Establishment of Extremum Slope Relationship

Under Concentrated Lateral Force

The rotationally constrained beam as sho-wn in Figure 3

considered under a point lateral load Q, applied at the location

The non-dimensionalized governing differential equation and the

boundary conditions are given by

IV - / %
w - Q6(x - C) = 0

where
2
Q =SL-
* EI

at x = 0 w = 0

w" - or w1 = 0
o

at x = 1 "w = 0

w" + c^w' = 0
The Laplace transform

w(s) = f e~ SX w(x)dx (21)

is the most convenient tool to solve this problem.

Thus we derive

-Cs
s w(s) - [s3w(o) + s2w'(0) + swM(o) + w m (0)] - Q, e =0 (22)

Noting that w(x) = 0 at x = 0, (22) can be transformed back as

follows:

2 ^
w(x) = Ax + Bx + Cx = w (x) for 0 £ x < £ (23)

w(x) = wx(x) + |(x - C ) 3 for £ <; x <; 1

where A, B, and C are constants. The constants A, B, and C are

obtained from the boundary conditions (20). They are:

A = Q £li-ZJL)[ofi(1 . C ) + 2(2 - C )] (2^)

«nC(l " 0
B =Q^ — ^ -[^(1 - C) + 2(2 - 0 ]

C = -5 - - ^ [3(a0 + 2)[ai(l - C)+2}-2(l - C)2(a0+a1-^0«1)]

where A = 12 + ^(aQ + c O + a a-..

Differentiation of equation (23) yields the slope and moment.

w' = A + 2Bx + 3Cx = w ' (x) for 0 <. x < £ (25)

= w 1 , (x) + |(x - Q)2 for Q <, x <; 1

w" = 2B + 6Cx = w "(x) for 0 <; x < £ (26)

= w"(x) + Q(x - Q) for £ «s x ss 1

Define

A = TfQ, B = BQ, C = CQ (27)

In the case where <y = ^ = a, £ is \ . Then 13, tf are given "by

s
= isr^y ( 28 )
and
7* =
? -^
= -
12

The locations of extremum slope x and x n , are given "by:

Xo = T-7-^V (29)
^2 ~ ¥(a+2)
 105

Thus it follows that

*2 " xl = i f ^ T -lext (30)

and

x2 + x = 1

The sum of the absolute values of the extremum slopes is

W
'x=x2 " w'x=xl = ^ W + 3tf X 2
( 2 - Xl2) ^

= * [2B + 3C(x2 + Xl )]

12
extt

The result is pertinent to the empirical law for the single concentrated
load [7]. Substituting l/2 for x in equation (23), deflection in the
middle of the span is given as,

(y+
R - ® 1 L (^o\
U
~ a+2 192 EI '";
 APPENDIX B

THE ELASTICALLY SUPPORTED BEAM COLUMN OF UNIFORM BENDING STIFFNESS

Differential Equation and Boundary Conditions

Consider an elastically supported beam column of uniform bending

stiffness loaded with a distributed lateral force q(X) and an axial

compression force P as depicted in Figure 15 • Let spring stiffnesses

at the boundaries b e K , K for rotational restraint and K , K, for

lateral restraint. Thus we m a y write the strain energy of the system

as

r> L
E I ( W " ) 2 clX + | { K 2 ( W ' ) 2
\- I X==L
K^W')'
X=0
(1)

+ K3W + K,W
4
X=L X=0

and the work done b y the external forces as

L L
• o r

W = £ (W)^ +
q(x)WdX (2)
e 2 o "" o

"When the non-dimensional quantities defined below:

 107

K^L K2L
0i
o " El~ a
l =
EI~
(3)

3 KL-
p =
^o ~ EI i ~ET

X ,2 PL - qlf
x =L
r>: X EI EI

a r e u s e d ; then n o n - d i m e n s i o n a l i z e d s t r a i n energy and work f u n c t i o n s can

he w r i t t e n as

^2. '2
,• = i i (w") dx + ^a w' + 1 ~
I f O x=0 * V x=l w

+ ^ v + ^ 6 w2
x=0 x=l

x r 2 f •'"
We = — j (w 1 ) dx + j q(x) w dx (5)

Total potential energy is

Um
T
= U. - W
l e (6)

Thus the variation of total potential is

 108

1
6U =
T J tW + X w
" " ^ X
H 6wdx +
(w" + a
iwl) 6w
' x=l
(7)
o •L

+ (-w" + a w') 6w' _ n + (-V" - X w 1 + p w)6w

o ' x=0 x=l

+ (w"' + X w' + p w) 6w
o x=0

By equating 6U to zero, the resulting differential equation and the

boundary conditions are developed:

+ X w" - q(x) = 0 (8)

at x = 0: w" - a w 1 = 0 or w,: = 0 (9)

3
o

w"» + x2w' + N = 0 or v = 0
o

at x = 1; w" + OLw1 = 0 or w1 = 0

w"t + x2v» - p w = 0 or v = 0

Orthogonality Relation of the Buckled Mode

Let u, v be the buckled mode corresponding to the different

eigenvalues X and p, respectively. Then from (8)

IV 2
u + X u" = 0 (10a)
 109

v IV + /v" = o (iob)

multiplying (10a) by v, (lOb) by u and integrating the difference of

the products from 0 to 1, the following integral is obtained.

x = j X L(UW + x2u„)v _ ( v OT + ti2v„)u-: d x (ll)

"o
1 1 1
= f (u IV v - u v ^ d x + X2 j u'Vdx - / j vMudx = 0
o o o

Integrating by parts and making use of the boundary conditions of

equation (9)? equation (ll) can be reduced to the following simple
relationship:

I = (X2 - pJ2)[ u'v'dx = 0 (12

o

Therefore
1
u'v'dx = 0 (13)

Since

\/^

This equation is an orthogonality relation of the buckled mode.

 110

APPENDIX C

APPROXIMATION OF THE UNIFORMLY DISTRIBUTED LOAD SYSTEM

BY A FINITE NUMBER OF EQUALLY SPACED CONCENTRATED LATERAL FORCES

According to the result in Appendix A, the distribution of

moment along the constant cross-section beam on the unyielding end

supports subjected to a uniformly distributed lateral force is given by

a +
c^-ar-, n(&n ^)
m(x) = ^£6x t2- A/'l
6(1 4+- -2_i) x + ° x
'"'] (i)
"Av " _ A

The distance between the inflection points t is given by

6XL

£ , =/(l + — — ) - | -JL"i- (2)

ext ^ A ' 3 A

Suppose a beam of uniform EI is loaded with concentrated lateral

forces at the equally spaced intervals. The distance between the

inflection points under this loading system is to be found. According

to Appendix A, the moment distribution due to a lateral concentrated

force applied at loaction £ is given by

m(x) - 2B + 6Cx for 0 <. x < £ (3a)

= 2B + 6Cx + Q(x-g) for £ <; x <: 1 (3b)

 Ill

•where

B
= h-a°ai{z ' 2Q2 + £ 3) + 2 a ° ( 2 £ - 3 c 2 + ^ W

C = - |^tA - 6(0^ + 2)C - 3«0(2 + a i ) £ 2 + 2(aQ + « x + c y * ^ 3 ]

Suppose the number of concentrated lateral forces is N. Then the

location of the forces are

Ci=HTT i=l,2,...,» (5)

From equation (3b), the moment between the i and the (i+l)

location is given by

N i
m = (2B +
i~i+l A i ^x)+QV ( X -^) (6)
i=l r =1

•where B. and C. are obtained by replacing ^ by ^. in B and C of

equation (k). The following identities are used to calculate the

summation in equation (6).

n
) i = |n(n + 1) (7)
Z__i

i=l
 112

V i 2 = | n (n + l)(2n + 1) (8)
i=l

n
^i3 =ln2(n + 1)2 (9)
i=l

Thus

N N N
2 N 2W + 3
Yc = I Yr - < ^ Yr 2 (10)
£ C U ;
A i 2' A i " 6(N + 1) ' L k " 1+(N + 1) °
i=l i=l i=l

The use of equation (10) in equation (6) yields the moment between
•th w . _,Nth _,
l and (l + 1) location.

m := y + pX
i~i+l ^

where

_ «0 ( «1 + 6 ) N(N + 2) _ i(i +!)..- (12)

1
12 A N + 1 2(N + 1)M K
'

P = - 2 AM
TNVT)[N(V*I
+ 5a + 3a + 12) +
o i ^i +6 )( a o +2)] + Qi
 113

If an inflection point exists in this region, it is located at

x = -*
P

Following this scheme, the distance between two inflection points was

calculated. Then this length was used in the empirical formula (13)

obtained in Chapter II to compare with the exact values of the critical

loads. The computer program is shown in Appendix F and the results

are listed in Table 11.

 Ill*

APPENDIX D

BEAM OF VARIABLE BENDING STIFFNESS

Clamped-Pinned Beam

Slope Evaluation

The slope at the pinned end of a beam loaded with a uniformly

distributed load (q per unit length) is calculated. The differential

equation for the lateral displacement W is given by

-2
2-~ (EI(X) W") - q = 0 (1)
3X

Integrating (l) twice gives

El(x)w" = | X2+- AX + B (2)

where A and B are constants. It is assumed that the bending stiffness

is written in the form

P
EI(X) = EI {1 - ^ ) ] (3)

where T is a number greater than 1 and is called the taper ratio and

p is a positive number. When the following non-dimensional quantities and

conditions are introduced

 115

X=
L
q
= m1 w=
L
W

Equation (2) and the boundary conditions can be written as follows:

(! - IxP)wM = |x 2 + Ax + B (5)

where A and B are constants, and

w = 0 at x = 0 (6a)

w'= 0 at x = 0 (6b)

w = 0 at x = 1 (6c)

w"= 0 at x = 1 (61)

When the boundary condition (6d) is used in Equation (5), B can be

expressed in terms of A. Thus,

B = -(A + |) (7)

Hence (5) now is written as

(1 - i * P K = 4 + Ax - (A + i) (8)
 116

For the ease of calculation, only two cases (p = 1 and 2) are treated

in detail.

Linearly Varying Bending Stiffness: p = 1. Equation (8) is

now written as follows:

*"- l ( n .x) + _JT_ (9)

where

CT = - (T + 2A) (10)

T = ~ ^ (T + 2A + 1)

Integrating (9) yields

w'= 1, x2
^(<J - x) - Tin ( T - x ) + C (11)
T

When boundary condition (6b) is applied to equation (ll), the constant

C is obtained.

2
C = g- + Tln(T) (12)

Integrating (ll) gives

jg ((G - x^ 3) J +, T_/rp(T -_ x Nj) [ l-i^/m

=
- !L- V
n ( T - x\ ) _n- l l+Tiv
v
+Cx 4-+ D
T» 0-3>)
 117

and boundary condition (6a) is used to evaluate D

D =
' 12 " T T L l n ( T ) _ 1 (lh)

When (6c) is used in equation (13)? the result is

£- 3
3
^ ( a- l ) + T ( T - l)[lnCC-l)-lJ+5- + Tln(T) - 2 - - TT[ln(T)-1 ]=0 (15)

or

T ( T - l)Lln(T - 1) -lj + ^ - i- + T m ( T ) - TT[ln(T)-ll=0 (16)

From equation (10), T can be related to Q as follows:

T - 1 fl N (17)
T = — 5 — (1 - a)

Using (17) in (16) yields

<i-a)(*# ^i)*2LraM- (18)

From ( l l ) , the slope w' at x = 1 i s

V = TL-|(1 - B ) • 1 + ^ ( x . a)lnQ] (19)

When the values of (l -CT)which is calculated from (18) is used in

 118

(l9)? the slope at the pinned end is obtained.

2
Parabolically Varying Bending Stiffness: p = 2. Let A = T

Then (8) is now

2!! = . i + Jo + __j^_ (20)

T 2 A -x A + x

where

= ^ [A + ^ ] (21)

a=
^ ["A+ S"^ (22)

When equation (20) is integrated and the boundary condition (6b) is

used, the resulting equation is

¥' _ X
T = " 2 - "bln(A - x) + ain(A + x) + AlnA (23)

Integration of equation (23) yields

(2k)
f- - - f- + b(A " *){ln(A " *) - 1 }

+ a(A + x)|ln(A + x) - l}+ XAIII(A) + D

 119

where the constant D is evaluated from the boundary condition (6a) as

follows:

D = -(b + a)A[m(A)-l] (25)

When the remaining boundary condition (6c) is used in equation (2k), the

result is

- jt + A[l+ InA] + b [ A l n ( ^ ) - ln(A-l)] (26)

+ a [ A i n ( M i ) + ln(A+l)] = 0

Using (21) and (22) in (26), it is possible to write this equation in

terms of the unknown constant A only. From which it follows that,

I - (\r)[Mn (^|) • in (££)]

A
A = A" ' ~' " (27)

The slope at the pinned end is

-' = T L-i + b HT^T) + aln

H^I (28)

where b and a are as shown in (2l) and (22) .

 120

Formulation of the Finite Difference Equation for Buckling Investigation

The buckling equation and the boundary conditions are

2 2 2
a [EI(X) a^W] + P aw 0 (29)
SK ax^ sx

W = W = 0 at X = 0 (30)

W = W" = 0 at X = L

As b e f o r e , define

X W .2 PL 2 /Q1N
X== W = X = (31)
L> L> ET
o

Then (29) and (30) become

{(1 - - x P K } " + X2w" = 0 (32)

and

•w = w ' = 0 at x. = 0 (33)

w = w" = 0 at x = 1

respectively. Completing the operation in (32) and introducing a new

variable (T| - w " ) , the following set of equations is obtained.

+
\x - £)•? + z (- Y^ i^2 - E i VTJ ^~ 2")-n - o ^

I] - w'T = 0 J
When the central difference is used to represent the derivatives,

equation (34) can be reduced to the following finite difference

equation.

1
- hV)^±+1- ^ +
O - - - 1 P " 1 (^ 1 + 1 - ^ i_1 ) (35a)
2
+h ( x2 . sLp*lx-p-zy __ 0

and, -hY + (w1+1 - 2W1 + w 1 " 1 ) =o (35b)

or in matrix form,

I*] z i_1 + LB] /+ W z

i+1
-o (36)

where

Z1 = i ^ }- (37)
1
v1 J

and
 l-i(x;P-phx*-r
A T\ i

-2(l-ix/) + h 2 ( x 2 - ^ x.P-2)

-h

n 1/ p P~l^
1 ix. + phx. )

i C

and h = size of division = l/(number of division)

i denotes the i-th location,

for 1 <> i <> N-l. Boundary matrices are formulated as follows

at x = 0;

w ' = s^V1*-^ = °' •"• w i = w-i

then

^o = ~2 (w l " 2 w o +W
-l } =
h Wl
h h
 123

and
w =0 (hi)
o

Replacing Z by

2 Wl
h
}

the matrix equation for i = 1 in (36) becomes

P-2
2 (te)
-2 (i - £) + n ^ - p(p-«V-} Mi - T
k^-v^
\ T / L ' h
-2
-h

! - I(hP ph P ) , 0

c
+
+
0 , 1
i-

Since v n and ^ are both zero (w" = 0, v = 0 at x = l), Z n is a zero

matrix. Then
 22k

1 - f{(n-l)V-nh (n-D^V"1}^ 0
,n-2 (^3)
1

-2 (l - ±(n-l)V)+ h2|x2 - £^i)( n .l) n - 2 h n - 2 ^

,n-l =
-h ,-2
_J

Since the matrix equation obtained is banded along the diagonal, it can

be conveniently treated by Potter's method (Reference [13]), but it

should be noted here that to remove the singular behavior of the

determinant, a slight modification is necessary (Reference [lU]).

Calculation of the buckling load according to the finite

difference formulation derived was done on UNIVAC 1108 computer. The

computer program is given in Appendix F .

Consideration of the Simply Supported Beam

Determination of Absolute Values of End Slopes

The simply supported beam shown in Figure 8 is considered. The

variation of the bending stiffness is of the same form as (3)- First

the sum of the absolute values of the slope at both ends under a

uniformly distributed load is treated.

Linear Variation in Stiffness (y - 1 ) . Since w" = 0 at x = 0

 125

and 1, constants A and B in equation (5) can be determined easily.

Thus for this case

2
(T - x ) =r = | ( x -x) m

Hence

2 2l = - (T - 1) - x + T(T - 1) ^ (U5)

Integration of (45) yields,

9 v2
£ w» = _ ±- - (T _ i ) x _ T ( T .. i)in(T - x ) + C (46)

where C is a constant. From (46) the algebraic difference of the end

slopes is obtained. It is

(w
'x=l "W'x=0) = I Ll " T + T(T-D1H( T 4T)] ^7)
\ 2
=
Parabolic Variation in Stiffness (p =• 2) . Define A T ? then

(A2 - x2) f- = |(x2 - x) (48)

Equation (48) can be rewritten as

 126

2 „" = . ! + A+l J_ + A^i _L. (1,9)

v
T 2 A+X 2 A-x ^;

I n t e g r a t i o n of (2+9) g i v e s

| . w« = . x + A 4 i i n ( A + x ) . A z i ln(A-x) + C
1 C- £1

where C is a constant. Hence

Tv x=l x=Cr 2 A-l 2 v

T

or

«W-^=^ + ^'(&K)] (50

Formulation of Finite Difference Equations for Buckling Load Determination

The finite difference formulation and the boundary matrix at

x = 1 are the same as for the clamped-pinned case. The only change re-

quired is the boundary matrix at x = 0. Equation (^2) is to be replaced

by the following matrix equation:

 127

p-2
- 2 ( l - 5 ! ) + h ^ -p(p-l)^ \}

-h > -2

1 -T-(hP + phP),
z2=,
(51)
1
 128

APPENDIX E

CANTILEVER WITH TIP LATERAL SPRING

Characteristic Equation for Instability

The buckling load of a cantilever beam with a lateral tip spring

(Figure 5) is to be calculated from the characteristic equation.

The non-dimensionalized differential equation is

w I V + X 2 w" = 0 (1)

And the appropriate boundary conditions are

w = w' = 0 at x = 0 (2)

w" = w,,T + X 2 w T - p^w = 0 at x = 1

Thus the general solution is obtained as

w = A sin Xx + B cos \x + Cx + D (3)

Fulfillment of the boundary conditions (2) by the general solution

yields the following equation:

 129

sin X cos X
W
3
-X -3 1 (sin X-X) -PjCcos \-l) B) 10.

By equating the determinant value of coefficient matrix in

equation (k) to zero, the buckling equation is obtained; viz

sin X cos X
= 0 (5a)
-X -P-^sinX-x) -31(cosX-l)

Expanding the determinant, the folio-wing characteristic equation

is obtained:

J
P (sin X - X cos X) + X cos X = 0 (5b)

From equation (k),

= - A tan X

Using this result and boundary conditions at x = 0 of equation (2) in

equation (3), the buckled mode shape is:

•w (x) = A(sin Xx - X x ~ t a n X(cosXx-l)} (6)

Behavior Under a Concentrated Lateral Force

A concentrated lateral force is applied to the cantilever beam

 130

with a lateral spring (Figure ) . The deflection under this loading is

to be derived.

Non-dimensionalized differential equation is

w IV - Q8(x - 0 = 0 (7)

where 6(x) is a Dirac delta function. Boundary conditions are:

at x = 0 w = w' = 0
at x = 1 w" = w'" - p w = 0

Let the operator of Laplace transformation be denoted by L( ).

Defining w(s) = L(w(x)), equation (7) can be transformed in the

following form:

—rs
s^w(s) - {s3w(0) + s2w'(0) + sw"(0) + w'"(0)3 - Qe =0 (8)

Noting that w(o)and w T (0) are zero, equation (8) can be transformed

back as follows:

w = Ax + Bx = w (x) 0 <; x <; Q (9)

J
w = w. to + g (x - £ ) C <. x £ 1

Two constants A, B are evaluated from the boundary conditions. They are,
 131

A
= - 5ip^y tpx(i-c) - r d-e)3 + 2] do)

B = i ^ 3 j [^(l-C) - Pxd-C)3 + 6C]

Behavior Under Uniformly Distributed Lateral Force

A uniformly distributed lateral load is applied to a cantilever

beam with a tip spring as shown in Figure 5. The distribution of shear

force, bending moment, in addition to slope and deflection are to be

derived.

The non-dimensionalized differential equation is

w I V - c[ = 0 (11)

Boundary conditions are the same as the previous case. Integrating (ll) in

succession and satisfying the boundary conditions yields the following

results:

5B 1 + 2k
W = 5 [X (12a)
'" " 8(Pl + 3)]

5p + 2h + 12
_ i rk 2 i h
w ] (i2b)
s^-p^r3« +tr^r
- , 3 ( 5 P , + 2U) , p + 12
3 +
w - fe [8x - -j±rr- *• 6 PTTT- ^ ^C>
 132

q k ^ 1+ 2k
3 h + J 2
2

From (12b), the locations of zero-moment are given by

P 5pn + 2U B-, + 12

_ p 1 + 12
x x =
l - !» 2 4(p.+3)

Hence, the distance between inflection points is

3P
x (l4)
*ext = »2 " l = UTp^37

and the location of zero shearing force is

5(3n + 2k
x = 1
o sT^ri) d5)

Behavior Under a Concentrated Tip Couple

A concentrated couple is applied to a cantilever beam with a

lateral tip spring (Figure 5 ) .

The deflection due to a couple can be obtained by differ-

entiating the deflection due to a unit lateral load with respect to the

coordinate Q where the lateral load is applied. Thus, differentiating

equation (9) with respect to Q gives

w = Cx J + Dx = w(x) 0 ^ x < £

w = w(x) - | ( x - C) 2 C * x <; 1

where

c
=" M ^ 3) L A + p i ( 1 - £)2_

D
= Mp/ + 3) L-^i + 3\(1 - C)2 + 6]

Espec ially when a couple is applied at the tip, the deflect

w(x)
= ^T3l K x +(6 " p l } }

Slope at the tip due to an end couple is

",(1) = ^ + 3) L ^ + 2(6-^)] - ^ g ^ y i
 13^

APPENDIX F

COMPUTER PROGRAMS

C BEAM WITH ROTATIONAL SPRINGS AT BOTH ENDS. FIND PORT. FIND RATIO
C OF LENGTH UPTO THE INFLEXION POINT FROM THE OPPOSITE END OF THE
C BEAM TO THE SLOPE AT THAT POINT .FIND PRODUCT OF THIS RATIO AND
C PCRT .
DIMENSION BTA(^O), PEX(^O),AL3(^0),P3(^0),ERR0R3(^0)
REAL L,LL0W,LUP,LNEW
DATA( BTA (i),I=l,27)/0.,.25,.5,75,1.,1-5,2.,2.5,3.,3.5,k.,4.5,5.
1,5-5,6.,6.5,7.,7.5,10.,20.,50, , 1 0 0 . , 5 0 0 , 1 0 0 0 . , 5 0 0 0 . , 1 0 0 0 0 . , 1 0 0 0 0 0
1.0/
C CYCIxE FOR BETA1
DO 10 1=1,27
WRITE (6,1^0)
lUO FORMAT(IHIJIOH BETA1,10H BETA2,10H PCRT,10H L l / 2 4 * 1
1,10H ERR0R1,10H L2/24* 2,10H ERR0R2,10H LEXT,10H
1PEXT510H ERROR//)
DO 15 M=l,27
C CYCLE FOR BETA2
C=I2.O+U.C^(BTA(I)+BTA(M)>I-BTA(I)^BTA(M)
BI=BTA(I)+BTA(M)
B2=BTA(l)*BTA(M)
C CYCLE FOR FINDING L(=PEX(l)) ,WHEN CHARACTERISTIC EQJtf. IS SATISFIED
30 DO UO K=l,600
L=FL0AT(K)-0.9
DET=BI*(SIN(L)-L*COS(L))/(L*L)
DET=DET+B2*(2.0-2.0*COS(L)-L*SIN(L))/(L*L*L)+SIN(L)
200 IF(K.EQ.l) GO TO 60
IF(D1*DET)50,50,60
50 DL0W=D1
DUP=DET
LUP=L
C B I S E C T I O N SCHEME STARTS .
7 0 LNEW=LLOW+DLOW* (LUP-LLOW)/(DLOW-DUP)
DIF1=LNEW-LL0W
DIF2=LUP-LNEW
I F ( D I F 1 . L T . 0 . 0 0 0 5 - O R . D I F 2 . L T . 0 . 0 0 0 5 ) GO TO 1 0 0
D E T = B I * ( S I N ( L N E W ) -LNEW^COS(LNEW) )/(LNI^W^LNEW)+SIN(LNEW)
DET=DET+B2*(2 .0-2 ,0*COS(LNEW) -LNEW*SIN(LNEW) )/(LNEW)**3
IF(DET*DLOW) 8 0 , 8 0 , 9 0
80 DUP=DET
LUP=LNEW
GO TO 70
 135

90 DL0W=DET
LL0W=LNEW
GO TO 70
60 D1=DET
LL0W=L
1+0 CONTINUE
C BISECTION SCHEME ENDS .
100 PEX(l)=LNEW^LNEW/9.8696
C TO OBTAIN THE INFLEXION POINTS X1,X2 .
A=6.0
B=-6.O*(I.O+(BTA(I)-BTA(M))/C)
D=BTA(l)*(BTA(M)+6.0)/C
E =SQRT (B*B -k. 0*A*D)
Xl=-(B+E)/l2.0
X2=(-B+E)/l2.0
C AL1=LENGTH BETWEEN LEFT END AND 2ND INFLEXION POINT.
AL1=X2
C AL2=LENGTH BETWEEN RIGHT END AND 1ST INFLEXION POINT.
AL2=1.0-X1
C THETA1=SL0PE AT A L 1 .
THETA1=(2.0^X2^3+X2^X2^B/2.0+D^X2+(BTA(M)+6.0)/C)/12.0
C THETA2=SL0PE AT AL2 .
THETA2=(2 .0*Xl**3+Xl*Xl*B/2 .0+D*Xl+( BTA (M) + 6 . 0 ) / C ) / l 2 . 0
P l = A L l / ( 2 k . 0*THETA1)
P2=AL2/(2k.0*THETA2)
Pl=ABS(Pl)
P2=ABS(P2)
ERR0R1=(PI-PEX(I)*100.O/PEX(I)
ERR0R2=(P2-PEX(l)*100.0/PEX(l)
AL3(M)=X2-X1
P3(M)=(l.0+AL3(M))/(2.O^AL3(M)^3)
ERR0R3 (M) = (P3 (M) -PEX(I) )*100 .0/PEX(l)
15 WRITE (6,51) BTA(I),BTA(M),PEX'(I),P1,ERRDR1,P2,EREDR2,AL3(M),P3(M),
1ERR0R3(M)
51 F0RMAT(2F10.2,8F10.6/)
10 CONTINUE
STOP
END
 136

C FINITE DIFFERENCE METHOD FOR BUCKLING OF A CLAMPED-PINNED BEAM

C OF VARIABLE E I .
DIMENSION A ( 2 , 2 ) , B ( 2 , 2 )
COMMON N,H,X,A,B,TAPER,POWER,HSQR
TAPER=3
POWER = 2 .
DATA((A(I,J),I=1,2),J=1,2)/1.0,0.,0.,1.0/
DATA(N15N2JN3JL1,L2,L3)/10,30,5,1J51,10/
DO 30 N=N1,N2,N3
H=I.O/N
HSQR=H*H
DO 1+0 L=L1,L2,L3
X=FLOAT(L)
I F ( L . G T . ( L 2 - l ) ) GO TO 20
CALL DETERM(DET0,DET2)
I F ( L . E Q . L l ) GO TO 50
IF(DET1*DET2) 6 0 , 6 0 , 7 0
50 DET1=DET0
GO TO kO
70 DET1=DET2
GO TO 1+0
60 DETL0W=DET1
DETUP=DET2
XL0W=X-L3
XUP=X
GO TO 1+00
8o IF(DETLOW-*DET2) 85,85,90
85 DETUP=DET2
XUP=X
GO TO 1+00
90 DETL0W=DET2
XLOW=X
GO TO 1+00
20 WRITE(6,100)
100 FORMAT(1H1,12H NO SOLUTION)
GO TO 30
C SUBPROGRAM FOR BISECTION .
1+00 X=XLOW+DETLOW/(DETLOW-DETUP)^ (XUP-XLOW)
DIF1=ABS(X-XL0W)
DIF2=ABS(X-XUP)
I F ( D I F 1 . L T . 0 . 0 1 . 0 R . D I F 2 . L T . 0 . 0 1 ) GO TO 200
CALL DETERM(DET0,DET2)
GO TO 80
C END OF BISECTION PROGRAM.
kO CONTINUE
200 x=x/9.8696
WRITE(6,3OO)N,X
300 FORMAT( I5,2F15.6/)
30 CONTINUE
STOP
END
 137

SUBROUTINE DETERM(DETO.DET2)
DIMENSION A ( 2 . 2 ) , B ( 2 J 2 ) J P ( 2 , 2 ) , A P ( 2 , 2 ) , B A P ( 2 , 2 ) J V ( 2 ) ,JC(i+)
DIMENSION L B l ( 2 , 2 ) , L B 2 ( 2 , 2 ) , R B l ( 2 , 2 ) , R B 2 ( 2 , 2 )
COMMON N,H,X,A,B,TAPER,POWER,HSQR
REAL LB1,LB2,JC
LB1(1,1)=-2.0*(1.0-(H**POWER)/TAPER)+X*HSQR
im(l,2)=2.0*(l.0-(H**P0WER)/TAPER)/HSQR
LBl(2,l)=-HSQR
LBl(2,2)=-2.0
A(1 5 1)=1.0-(H**P0WER)/TAPER
C=LBl(l,l)*LBl(252)-LBl(l52)*LBl(25l)
DO 500 1=1,2
DO 500 J=l,2
500 LB2(I,J)=A(I,J)
B(l#2)=0.
B(2,l)=-HSQR
B(2,2)=-2.0
V(l)=1.0
CALL GJR(LB1,2,2,2,2,$350,,IC,V)
CALL MXMLT(LB1,LB2,P,2,2,2,2,2)
DSIGN=C/ABS(C)
N^-=N-3
DO 280 M=1,N^
A ( i a ) = ( i . o - ( ( M + i . o ) * H ) * * POWER/TAPER)
B(I,I)=-2.O*A(I,I)+X*HSQR
CALL MXMLT(A,P,.AP,2,2,2,2,2)
CALL MXSUB(B,AP,BAP,2,2,2)
C=BAP(l,l)*BAP(2,2)-BAP(l,2)*BAP(2,l)
V(l)=1.0
CALL GJR(BAP,2,2,2,2,$350,JC,V)
CALL M X M L T ( B A P , A , P , 2 , 2 , 2 , 2 , 2 )
U=C/ABS(C)
2 8 0 DSICN=DSIGITO
A(1,1)=1.0-((N-1.0)*H)*^F0WER/TAPER
B(I 5 I)=-2.O*(I.O-((N-I.O)*H)-^POWER/TAEER)+X*HSQR
DO 550 1=1,2
DO 550 J =1,2
RB1(I,J)=A(I,J)
550 RB2(I,J)=B(I,J)
CALL MXMLT(RB1,P,AP,2,2,2,2,2)
CALL MXSUB(RB2,AP,BAP,2,2,2)
Y=BAP(l,l)*BAP(2,2)-BAP(l,2)*BAP(2,l)
DETO=DSIGN*Y
DET2=DET0
350 RETURN
END
C FINITE DIFFERENCE METHOD FOR BUCKLING OF A S . S .BEAM
C OF VARIABLE E I .
DIMENSION A ( 2 , 2 ) , B ( 2 , 2 )
COMMON N ,H,X,A,B,TAPER,POWER,HSQR
TAPER=3•
P0WER=2.
DATA((A(l,j),I=l,2),J=l,2)/l.O,0.,0.,1.0/
DATA(Nl,N2,N3,Ll,L2,L3)/lO,30,5,l,51,10/
DO 30 N=N1,N2,N3
H=1.0/N
HSQR=H*H
DO UO L=L1,L2,L3
X=FLOAT(L)
IF(L.GT.(L2-1)) GO TO 20
CALL DETERM (DET0,DET2)
I F ( L . E Q . L l ) GO TO 50
IF(DET1*DET2) 6 0 , 6 0 , 7 0
50 DET1=DET0
GO TO kO
70 DET1=DET2
GO TO UO
60 DETL0W=DET1
DETUP=DET2
XL0W=X-L3
XUP=X
GO TO UOO
8o IF(DETLOW*DET2) 85,85,90
85 DETUP=DET2
XUP=X
GO TO UOO
90 DETL0W=DET2
XL0W=X
GO TO UOO
20 WRITE(6,100)
100 FORMAT(IH1,12H NO SOLUTION)
GO TO 30
C SUBPROGRAM FOR BISECTION .
UOO X=XLOW+DETLOW/(DETLOW-DETUP)^(XUP-XLOW)
DIF1=ABS(X-XL0W)
DIF2=ABS(X-XUP)
I F ( D I F l . L T . 0 . 0 1 . 0 R . D I F 2 . L T . 0 . 0 l ) G 0 TO 200
CALL DETERM(DET0,DET2)
GO TO 80
C END OF BISECTION PROGRAM.
UO CONTINUE
200 x=x/9.8696
WRITE (6,300) N,X
300 FORMAT( I5,2F15.6/)
30 CONTINUE
STOP
END
 139

SUBROUTINE DETERM(DET0,DET2)
DIMENSION A(2,2),B(252),p(2,2),AP(2,2),BAP(292),V(2),JC(l+)
DIMENSION LBl(2,2),LB2(2,2),RBl(252),RB2(252)
COMMON N,H,X,A,B,TAH;R,IOWER,HSQR
REAL LB1,LB2,JC
L B 1 ( 1 , 1 ) = - 2 . 0* (1.0-(H**FOWER)/TAPER)+X*HSQR
LB1(1,2)=0.0
LBl(2,l)=-HSQR
LBl(2,2)=-2.0
A(1,1)=1.0-(H**FOWER)/TAPER
C=im(l, l)-*LBl(2,2)-LBl(l,2)*LBl(2,l)
DO 500 1=1,2
DO 500 J=l,2
500 LB2(l,j)=A(l,j)
B(l,2)=0.
B(2,l)=-HSQR
B(2,2)=-2.0
V(l)=1.0
CALL GJR(LB1,2,2,2,2,$350,JC,V)
CALL MXMLT(LB1,LB2,P,2,2,2,2,2)
DSIGN=C/ABS(C)
N^=N-3
DO 280 M=1,NU
A(I,I)=(I.O-((M+I.O)*H)**POWER/TAPER)
BCI^L^^.O^-ACIJIJ+X^HSQR
CALL MXMLT(A,P,AP,2,2,2,2,2)
CALL MXSUB(B,AP,BAP,2,2,2)
C=BAP(l,l)-*BAP(2,2)-BAP(l,2)*BAP(2,l)
V(l)=1.0
CALL GJR(BAP,2,2,2,2,$350,JC,V)
CALL MXMLT(BAP,A,P,2,2,2,2,2)
U=C/ABS(C)
280 DSIGN=DSIGN*U
A(1,1)=1.0-((N-1.0)-*H)*-*P0WER/TAPER
B (1,1) =-2 .0* (1.0 - ((N -1.0 )*H )**POWER/TAPER) +X*HSQR
DO 550 1=1,2
DO 550 J=l,2
RBl(l,j)=A(l,j)
550 RB2(l,j)=B(l,j)
CALL MXMLT(RB1,P,AP,2,2,2,2,2)
CALL MXSUB(RB2,AP,BAP,2,2,2)
Y=BAP(l,l)*BAP(2,2)-BAP(l,2)*BAP(2,l)
DETO=DSIGN*Y
DET2=DET0
350 RETURN
END
 1^0

C FINITE NUMBER OF CONCENTRATED LOAD AT EQUALLY SPACED LOCATIONS

C TO APPROXIMATE THE UNIFORMLY DISTRIBUTED LOAD ON THE COLUMN
C ROTATIONALLY RESTRAINED AT BOTH ENDS. THE DISTANCE BETWEEN TWO
C INFLECTION POINTS IS TO BE CALCULATED. 2k JUNE 1970
DIMENSION BETA(l^)
REAL JO,NO,LEXT
INTEGER COUNT
DATA(BETA(l),I=l,lU)/0.0,.1,.5,1.0,2.0,5.0,10.0,20.0,50.0,
1100.0,500.0,1000.0,10000.0,100000.0/
DO 20 1 = 1 , I k
DO 25 K=l,lU
BETAl=BETA(l)
BETA2=BETA(K)
BETA26=BETA2 +6.0
BDIF=BETA1-BETA2
BSUM=BETA1 +BETA2
BPR0=BETA1 *BETA2
DELTA=12.0 +U.O*BSUM +BPRO
LEXT=SQRT((l .0+BDIF/DELTA)**2-2.0*BETAl*BETA26/3.O/DELTA)
PCRUNI=(1.0 +LEXT)/(2.0*LEXT**3)
Cl=BETAl*BETA26/l2.O/DELTA
C2=(BPRO+5 .0*BETAl+3 .0*BETA2+12 .0)/DELTA
C3=BETA26*(BETA1 + 2.0)/DELTA
WRITE(6,200) BETA1,BETA2
WRITE(6,250)
DO 10 N=l,20
N0=FLOAT(N)
X0=0.0
J1=0
COUNT=1
60 DO 30 J=J1,N
J0=FL0AT(J)
XFRONT =J0/(NO +1.0)
XAFT=(J0+1.0)/(N0+1.0)
CC=CI*NO*(NO+2.O)/(NO+I.O)-.5*JO-KJO+I.O)/(NO+I.O)
CX=-.5*NO/(NO+1.0)*(NO*C2+C3)+JO
x=-cc/cx
IF((XFRONT-X)*(XAFT-X))k),^0,30
kO Y=X-XO
XO=Y
IF(C0UNT.EQ.2) GO TO 100
COUNT =C0UNT+1
J1=J+1
GO TO 60
30 CONTINUE
100 PCRFIN=(l.O+Y)/(2.0*Y**3)
ERROR=(PCRFIN-PCRUNl)/PCRUNr*100.0
WRITE(6,300) N,LEXT,PCRUNI,PCRFIN,ERROR
10 CONTINUE
25 CONTINUE
 lin

20 CONTINUE
200 FORMAT(1H1,12H BETA1=,F10.2,12H BETA2= ? F10.2//)
250 FORMAT(kE TTOjlIH LUNI,15H PCRUNI,
120H LFIN,20H PCRFIN,10H ERROR//)
300 F0RMAT(I5,5F15.6)
STOP
END
 lk2

REFERENCES

Salmon, E. H., "Columns, A Treatise on the Strength and Design of

Compressive Members", Henry Frowde and Hodder and Stoughton,
London, 1921.

Niles and Newell, Airplane Structures, Vol. 1, John Wiley and Sons,
New York, 1953, p- 350.

Horton, W. H., and Struble, D. E., "End Fixity of Columns",

USAAVLABS TR 70-10, July 1970, United States Army Aviation Material
Laboratories, Fort Eustis, Virginia.

Horton, W. H., Cundari, F. L., and Johnson, R. W., "The Analysis of

Experimental Data Obtained From Stability Structures on Elastic
Column and Plate Structures", Israel Journal of Technology, Vol. 5,
Wo. 1-2, 1967, pp. 104-113-

Fisher, H. R., "An Extension of Southwells Method of Analyzing

Experimental Observations in Problems of Elastic Stability",
Proceedings of the Royal Society of London, Series A, ikk, p. 609-
630, I934.

Massonet, C , "Les Relations entre les Modes Normaux de Vibration

et la Stabilite des Systemes E*lastiques"_, Goemaere, Bruxelles, 1Q^0.

Horton, W. H., Craig, J. I., and Struble, D. E., "A Simple,

Practical Method for the Experimental Determination of the End
Fixity of a Column", Proceedings of the 8th International Symposium
on Space Technology and Science, Aug. 1969, Tokyo, Japan.

Horton, W. H., and Ford, J. S., II, "Some Experimental Studies on

the Relationship Between End Fixity and Critical Load Level for
Struts", USAAVLABS TR 70-21, July 1970, United States Army Aviation
Material Laboratories, Fort Eustis, Virginia.

Struble, D. E., "Boundary Effects in Structural Stability: A New

Approach", Ph.D. Thesis, January 1970, Georgia Institute of
Technology, Atlanta, Georgia.

Baruch, M., "On Undestmotive Determination of the Buckling Load of

an Elastic Bar", GITAER 70-1, March 1970, Georgia Institute of
Technology, Atlanta^, Georgia.

Pippard, A. J. S., and Pritchard, J. L., "Aerosplane Structures",

Longmans Green and Company, London, 193^, p. 103.
 1^3

12. Collatz, L., "The Numerical Treatment of Differential Equation",

Springer Verlag, Berlin, i960.

13- Potters, M. L., "A Matrix Method for the Solution of a Second
Order Differential Equation in Two Variables", Mathematisch Centrum,
Amsterdam, Holland, Report MR 19,1955.

Ik. Robert, E. B.; and Robert, E. F.; "A Modification of Potters Method
for Solving Eigenvalue Problems Involving Triagonal Matrices",
AIAA Journal, December 1966, p. 2231-2232.

15. Wang, C. T., "Applied Elasticity", McGraw-Hill, New York, 1953,

p. 269.

lo. Courant, R., and Hilbert, D., "Methods of Mathematical Physics -

Volume 1", Interscience, New York, 1953*

17* Livesley, R. F., and Chandler, D. B., "Stability Functions for

Structural Frameworks," Manchester Univ. Press, 1956.
 1^

VITA

Takaya Iwamoto was born March 25, 19^-2 in Eishokodo, Manchuria.

He is a Japanese citizen and completed his secondary education in

Nobeoka, Japan in i960.

Mr. Iwamoto studied aeronautics at the University of Tokyo and

graduated in 196^-. Subsequently, he was employed by Kawasaki Aircraft

Company, Ltd. as a research engineer. After working in many diversified

areas he decided to pursue a graduate degree in the field of Aerospace

Structures, and entered graduate school at the Georgia Institute of

Technology in September 1967. He was a research assistant there until

June 1971.

He is married and presently resides in Kakamigahara, Gifu, Japan

where he is employed in research and development by Kawasaki Heavy

Industry, Ltd.