Sei sulla pagina 1di 10

25

Stability Of
Structures:
Continuous Models

25–1
Lecture 25: STABILITY OF STRUCTURES: CONTINUOUS MODELS 25–2

TABLE OF CONTENTS

Page
§25.1. Introduction 25–3
§25.2. Example 1: The Euler Column 25–3
§25.2.1. Critical Load Analysis . . . . . . . . . . . . . . 25–4
§25.2.2. Linearization Limitations . . . . . . . . . . . . . 25–5
§25.2.3. Reformulation as Eigenproblem . . . . . . . . . . . 25–6
§25.3. The Fixed-Pinned Column 25–6
§25.3.1. Critical Load Analysis . . . . . . . . . . . . . . 25–6
§25.4. Elastically Restrained Column 25–8
§25.4.1. Problem Description . . . . . . . . . . . . . . . 25–8
§25.4.2. Critical Load Analysis . . . . . . . . . . . . . . 25–8
§25.4.3. Equivalent Spring Constant . . . . . . . . . . . . . 25–10

25–2
25–3 §25.2 EXAMPLE 1: THE EULER COLUMN

§25.1. Introduction

This Lecture focuses on buckling analysis of continuous models of elastic systems. This process
leads to ordinary or partial differential equations. Those can be solved in closed form only for
simple 1D structures, such as prismatic columns treated by linearized prebuckling (LPB).
Despite limitations in terms of obtaining analytical solutions, such problems are important for
displaying key intrisic features of continuous models. For example, the associated eigenproblem
has an infinite (but denumerable) set of eigenvalues, meaning an infinite number of critical loads.
(This should not be surprising, since continuous models have an infinite number of degrees of
freedom.) Of these, the lowest buckling load is of primary interest to designers.
We illustrate the treatment of continuous models using via several examples involving elastic
columns with simple boundary conditions. All of the examples in this Lecture have closed form
solutions.

§25.2. Example 1: The Euler Column

(a) P (b) (c)


P P
A A
A y y y
x x x
X'
X X' X
X
v(x) v(x)
L P M(x) = −P v
elastic

constant EI assumed (+ as drawn)


buckled shape

B B
P

Figure 25.1. The Euler column: (a) problem definition; (b) FBD of whole column assuming a
buckled shape v(x) with its amplitude exaggerated for visibility; (c) FBD at distance x from top.

The Euler column is shown in Figure 25.1(a). It is a homogeneous, prismatric, elastic column
pinned (hinged, simply supported) at both ends A and B, and subjected to axial load P. The elastic
modulus is E. Reference axes are chosen as follows: x is longitudinal, with origin at end A; z is
normal to the plane of the figure and selected so that I = Izz is the minimum second order moment
of inertia of the column cross section about z; finally, y lies in the plane of the figure. For example,
if the cross section is a solid rectangle of width b and thickness t < b, the minimum I = Izz is
b t 3 /12, and y, z will be aligned with the short and long cross-section dimensions, respectively.
It is emphasized that the minimum moment of inertia, herein called I = Izz as noted above, is the

25–3
Lecture 25: STABILITY OF STRUCTURES: CONTINUOUS MODELS 25–4

one that determines the buckling load.* With the axes chosen as shown, the column will buckle in
the x y plane. Figure 25.1(b) pictures a buckled shape that satisfies the end boundary conditions.
Such a curve is called a kinematically admissible buckling mode shape.
The mode shape illustrated by Figure 25.1(b) is mathematically defined by the deflection curve
v(x), whose determination is part of the problem. (The notation choice is not accidental, for v(x)
can be interpreted as a beam deflection.) This deflection is infinitesimal because of the linearized
prebuckling (LPB) assumption. In that Figure it is drawn with exaggerated amplification for
visibility.

§25.2.1. Critical Load Analysis


Figure 25.1(b) also shows reactions at B, which here reduce to just P. There are no end moment
reactions because the column is hinged at A and B; furthermore taking moments with respect to A
and B shows that both lateral reactions (along y) vanish.
Next, do a FBD of a segment AX, with X located at a distance x from A, as shown in Figure 25.1(c).
The displaced X is labeled X’, and the distance from X to X’ is v(x). At this cross section we
will have a bending moment Mz (x), which is positive as drawn in the figure (reason: a +Mz (x)
compresses the “beam top surface”, which by convention lies on the +y side.) Taking moments
with respect to X’ , equilibrium requires Mz (x) + P v(x) = 0. But according to beam deflection
theory, Mz (x) = E Izz v  (x) = E I v  (x). Replacing gives
E I v  (x) + P v(x) = 0. (25.1)
This is a homogeneous, second order, linear ODE in the unknown deflection v(x). For convenience
in reducing (25.1) to standard (canonical) form introduce

def P
λ = + , (25.2)
EI
Dividing (25.1) through by E I and substituting (25.2) we get the canonical form
v  (x) + λ2 v(x) = 0, (25.3)
Its general solution is
v(x) = A cos λx + B sin λx, (25.4)
in which A and B are determined from the kinematic boundary conditions v(0) = 0 and v(L) = 0.
The first one requires that A = 0, whence (25.4) reduces to

v(x) = B sin λx. (25.5)

This is called a characteristic stability equation or simply characteristic equation; the names
buckling equation and critical load equation are also used. It has two possible solution types:

B = 0 ⇒ v(x) = 0: the column remains straight
(25.6)
B = 0 ⇒ v(x) = 0: the column buckles with shape sin λx

* If the section happen to have equal bending inertia in all directions, as in the case of a circular or square section, buckling
can occur in any direction. The actual buckling plane is determined by imperfections, a topic not treated in this course.

25–4
25–5 §25.2 EXAMPLE 1: THE EULER COLUMN

These two solution types are designated as trivial and nontrivial, respectively, in the literature.
The critical loads are the values of P at which nontrivial solutions are possible. These are determined
by applying the second end condition: v(L) = 0. Since B = 0 we must have sin λL = 0.
Its solutions are λL = nπ, in which n = 1, 2, . . . is a positive integer. Square both sides for
convenience: λ2 L 2 = n 2 π 2 , replace λ2 by P/(E I ), solve for P and tag those loads as critical:

n2 π 2 E I
Pcr,n = , n = 1, 2, . . . (25.7)
L2

As can be observed, there is an infinite number of critical loads. These are the nontrivial solutions
of the characteristic equation (25.5). The lowest one is associated with n = 1:

π2E I
Pcr,1 = Pcr = . (25.8)
L2

This is called the Euler critical load.

π 2E I
Pcr2 = 4 π 2E I
2
Pcr3 = 9 π 2E I
2
Pcr = Pcr1 =
L2 L L
L/3
L/2

L L/3

L/2
L/3

Mode shape 1 Mode shape 2 Mode shape 3


(n=1) (n=2) (n=3)

Figure 25.2. First three buckling mode shapes for the Euler column.

The buckling mode shapes associated with the set of critical loads (25.7) are
n πx
vcr,n (x) = B sin . (25.9)
L
If n = 1 the mode shape is a half sine wave, similar to that drawn in Figure 25.1(c). If n = 2 we get
a complete sine wave, if n = 3, one-and-a-half sine wave, etc. These buckling shapes are drawn
for n = 1, 2, 3 in Figure 25.2.

§25.2.2. Linearization Limitations


The LPB assumption does yields the critical loads, but also brings about the following inconsisten-
cies, some of which have an air of paradox.

25–5
Lecture 25: STABILITY OF STRUCTURES: CONTINUOUS MODELS 25–6

• The foregoing solution says nothing about the amplitude of the buckling modes (25.8) in terms
of the data. This is a consequence of the LPB assumptions, which led to a simple linear ODE
but filters that information. Determination of this postbuckling behavior is quite involved,
since it requires solving a nonlinear ODE in terms of elliptic functions. Such analysis shows
that the buckled column continues to take up on increasing load, albeit very slowly, as long as
it remains elastic.

• Differentiating (25.7) three times gives a nonzero transverse shear load Vy (x) = E I B vcr 1 =
−E I (π 3 /L 3 )B cos(π x/L). Evaluation at x = 0 or x = L gives a nonzero transverse shear
force whereas Figure 25.1(b) shows zero horizontal reactions there. A similar inconsistency
arises if one differentiates four times. This gives a nonzero lateral load p(x) but no such load
exists. This “paradox” indicates that the sine-wave result must be corrected once the buckling
mode amplitude becomes finite.
• The analysis does not take into account imperfections such an initially crooked column, or
axial load eccentricity. Such an analysis is beyond the level of this course, but one basic result
is used in the justification of the “Southwell plot” experimental procedure described in Lecture
19.

§25.2.3. Reformulation as Eigenproblem

Lecture 17 emphasized that a discrete stability problem handled through the LPB assumptions may
be presented as a matrix eigenproblem. This is also the case for the linearized continuous problem,
but it requires a reformulation. For the Euler column, the two end boundary conditions v(0) = 0
and v(L) = 0 may be written conjointly as
    
1 0 A 0
= . (25.10)
cos λL sin λL B 0


This is a matrix eigenproblem in λ = P/(E I ), but note that (unlike the discrete case) this variable
does not appear linearly. For a nontrivial solution, meaning that A and B are not both zero, the
determinant of the matrix on the left side must vanish. This leads to the condition

sin λL = 0 (25.11)

which agrees with that previously found from the characteristic equation (25.5). This reformulation
becomes useful in more complicated cases.

§25.3. The Fixed-Pinned Column

As second example we consider the configuration shown in Figure 25.3(a). The buckled column
is drawn in Figure 25.3(b) in a deflected position. The most notable difference with respect tp the
Euler column of Figure 25.1(a) is the presence of additional reaction forces: the lateral reactions
R A and R B , and the fixed end moment M B . These are positive as shown in Figure 25.3(b).

25–6
25–7 §25.3 THE FIXED-PINNED COLUMN

§25.3.1. Critical Load Analysis


Equilibrium of y forces in Figure 25.3(b) gives R A = R B . Equilibrium of moments taken with
respect to either A or B yields M B = R A /L. Next consider the FBD of the portion AX shown in
Figure 25.3(c). Equilibrium with respect to X’ gives the non-homogeneous ODE
MB x
E I v  (x) + Pv(x) = R A x = . (25.12)
L
The LHS is the same as in (25.1), but the RHS is no longer zero. Dividing through by E I and
setting λ2 = −P/(E I ) gives after some manipulations

 λ2 M B x
v (x) + λ v(x) =
2
. (25.13)
PL
The general solution is the sum of the homogeneous solution v H (x) = A sin λx + B cos λx and
the particular solution v P (x) = M B x/(P L):
MB x
v(x) = A sin λx + B cos λx + . (25.14)
PL
The three kinemtic BCs are v A = v(0) = 0, v B (0) = v(L) = 0 and v B = v  (L) = 0. These
provide three equations:
MB MB
B+ = 0, A sin λL + B cos λL + 0, A λl − = 0. (25.15)
L PL
Solving these equations simultaneously one obtains the characteristic equation

tan λL = λL . (25.16)

The samllest roots of this transcendental equation, to 4 places, is λL = 4.493. The corresponding
critical load is
20.19 E I
Pcr = . (25.17)
L2
Since 20.19 ≈ 2.05 π 2 this critical load is approximately twice that of the Euler column. Thus
fixing one end has substantially increased the critical load.

25–7
Lecture 25: STABILITY OF STRUCTURES: CONTINUOUS MODELS 25–8

(a) P (b) (c)


P P
RA A RA A
A y y y
x x x
X'
X X' X RA
X
v(x) v(x)
L elastic P M(x)
constant EI assumed (+ as drawn)
buckled shape

MB
B
B RB = R A
P

Figure 25.3. Column fixed at one end and simply supported (hinged, pinned) at the other.

§25.4. Elastically Restrained Column

The title configuration is of interest because it covers the three boundary condition cases tested in
the Column Buckling Lab.† as special cases.

§25.4.1. Problem Description

Consider the modification of the classical Euler column pictured in Figure 25.4(a). The column
is simply supported (pinned) at both ends A and B, and axially loaded by P. The rotation at B is
further restrained by a torsional spring with stiffness k.

Column AB has length L and constant flexural rigidity E I , in which I ≡ Izz is the minimum
moment of inertia of the cross section that controls buckling. (For a rectangular cross section of
width b and thickness t < b, I = Izz = b t 3 /12.) The column is simply supported (pinned) at
both ends A and B, and axially loaded by P. The rotation at B is further restrained by a torsional
spring with stiffness k. Select x as shown and consider the buckling shape v(x) sketched in Figure
A.1(b). The end rotation at B is θ B = v  (L), positive CCW. The spring at B applies a restoring end
moment M B = −k θ B . For convenience in obtaining dimensionless equations we define

EI
k=β (25.18)
L

where β is a numerical coefficient. If β = 0 the problem reduces to that of the classical Euler
column, whereas if β → ∞ we obtain the case of a column simply supported at A and fixed
(clamped) at B.

† Prior to Fall 2010, this used to be Experimental Lab 3. It is now a “Lab-Homework”, meaning that the results are
presented as part of an assigned Homework rather than a formal Lab Report.

25–8
25–9 §25.4 ELASTICALLY RESTRAINED COLUMN

(a) (b) (c) (d) P


P P P
RA RA
A y y,v A
x x M(x)= −P v+RA x
X
RA beam-column
L P
v(x)

k
MB = − k θB C elastic beam
B
B
RA
x P
−θB

Figure 25.4. Column with elastic restraint: a torsional spring, at end B.

§25.4.2. Critical Load Analysis


The FBD of the complete column is shown in Figure 25.4(b). Taking moments with respect to B
one finds that
R A = M B /L = −k θ B /L = −β E I θ B /L 2 . (25.19)
Now cut the column at section X at distance x as sketched in Figure 25.4(c). Moment equilibrium
with respect to X yields the second order differential equation
x E I θB x
E I v  + P v = R A x = −kθ B
= −β , (25.20)
L L2
which results from equating the bending moment M(x) = E I v  to −Pv + R A x. Dividing through
by E I and calling λ2 = P/E I produces the canonical form
θB x
v  + λ2 v = −β
. (25.21)
L2
The general solution of (25.21) is the sum of the homogeneous and particular solutions:
θB x
v(x) = C1 sin λx + C2 cos λx − β . (25.22)
λ2 L 2
Since v(0) = 0, C2 = 0. The rotation is θ(x) = v  (x) = C1 λ cos λx − βθ B /(λ2 L 2 ). Evaluating
this at x = L yields θ B = θ(L) = C1 λ cos λL − βθ B /(λ2 L 2 ), from which we can solve for the end
rotation:
λ3 L 2
θ B = C1 cos λL . (25.23)
β + λ2 L 2
Inserting this into the solution (25.22) with C2 = 0 gives v(x) = C1 sin λx − C1 βλx cos λL/(β +
λ2 L 2 ). Applying now the second boundary condition: v(L) = 0, furnishes the stability equation
 
βλL
v(L) = C1 sin λL − cos λL = 0. (25.24)
β + λ2 L 2

25–9
Lecture 25: STABILITY OF STRUCTURES: CONTINUOUS MODELS 25–10

For buckling to occur, C1 = 0, and the expression in parentheses in (25.24) must vanish. Calling
α = λL, which is also a dimensionless variable, we obtain the trascendental equation

αβ
tan α = (25.25)
α2 +β

For a given β ≥ 0 we seek the solution αcr > 0 of (25.25) closest to zero. If β = 0 there is no
closed form solution and it is better to proceed numerically, using for example a Newton solver. It
is easily shown that for β = [0, ∞], π ≤ αcr < 4.5, so α ≈ 4 is a good start point for a Newton
iteration. Such a solver is implemented in Mathematica in the built-in function FindRoot. For
example, the statements

beta=100; Print["alphacr=",alpha/.FindRoot[Tan[alpha]==
alpha*beta/(alpha^2+beta), {alpha,4}]];
return αcr = 4.4494 as the desired numerical solution of (25.25) for β = 100. Here is a table for
selected values of β:

β 0 1 3 10 100 1000 10000 ∞


αcr π 3.4056 3.7264 4.1323 4.4494 4.4889 4.4930 4.4934
L e /L 1.0000 0.9224 0.8431 0.7602 0.7061 0.7000 0.6992 0.6991

The critical load Pcr and the effective length L e (tabulated above) are

αcr
2
EI π2E I π
Pcr = = , Le = L (25.26)
L2 L 2e αcr

This solution also applies to the buckling of a column AB rigidly connected at B to an elastic
beam BC, as illustrated in Figure 1(d). This is actually one of the configurations to be tested in
Lab 3. All that is needed is to work out the appropriate value of k = M BBC /θ BBC obtained from a
moment-deflection analysis of beam BC subjected to an applied end moment M BBC = M B .

§25.4.3. Equivalent Spring Constant


Suppose the restraining elastic beam BC in Figure A.1 has length L r = L BC , modulus Er and
moment of inertia Ir about z. Assume a simply support condition at C. Under an end moment
M B applied as pictured in Figure 25.4(d), the restraining beam deflects by† vr = −M B xr (L r2 −
xr2 )/(6Er Ir L r ), in which xr is the distance from C. The end rotation is θ B = (dvr /d x)|xr →L r =
M B L r /(3Er Ir ), whence the equivalent torsional spring stiffness is k = M B /θ B is 3Er Ir /L r .
Comparing with (1) shows that β = 3(L/L r )(Er Ir )/(E I ). If the restraining beam is fabricated of
the same material and has the same cross section dimensions of the beam-column, as is the case in
Lab 3, Er = E, Ir = I , and β = 3L/L r .

† E. M. Popov, Engineering Mechanics of Solids, Prentice Hall, NJ, 1999, case 5 of Table 10, p. 853.

25–10

Potrebbero piacerti anche