Professional Documents
Culture Documents
Buckling of columns
4 Full lectures.
32
Buckling is one of the most common modes of failure in structural engineering. In this
course we will look at buckling of columns only1 . Buckling is an interesting problem
because it is a stability problem and stability problems are always interesting(!). This is
because they involve solutions to differential equations that infer serious changes to the
system. For example, when we have two adjacent streams of fluid flowing at different
speeds, an unstable shear layer forms and the two streams eventually mix (see figure 6.1).
FAST
SLOW
Figure 6.1: Mixing layer: the top half of the fluid is moving faster than the lower half.
The dye released at the centreline shows that the flow is stable for a while, then the flow
becomes unstable and develops those beautiful patterns. When things go unstable, it is
usually beautiful for a little while then the system can explode and turn ugly!
This is the same for a column. It will go unstable as we will see. When unstable it has
nice sinusoidal buckling modes, but in reality, those modes will last for a brief time before
the column will collapse.
Before we look at real columns, let us try to understand the concept of buckling as a
stability problem. To this end, we will investigate a simplified column consisting of two
rigid beams with simple supports, joined together with a torsion spring as shown in the
figure below. The torsion spring has a spring constant of R which relates the torsion (or
moment) ot the angle of rotation via M = 2R (the 2 in this formula is needed when
is defined as in figure 6.2, i.e., when the spring is loaded equally from both ends). The
free body diagram in figure 6.2 allows us to write the bending moment balance,
X
M = 0 = M P , (6.1)
1
Buckling of thin plates is considered in many textbooks and is also an important design issue
33
P P
L/2
L/2 P
Figure 6.2: Approximation to a loaded column. Two rigid beams (solid black) are joined
with a torsion spring, with rotational spring constant, R .
where is the horizontal displacement of the spring. For small angles, L/2 so that,
L
2R P = 0
2
L
1P = 0. (6.2)
4R
One solution is, of course, = 0. This means the structure is in equilibrium when = 0,
ALWAYS, REGARDLESS of P . The other solution is
4R
Pcr = , (6.3)
L
which gives a value for the load which will cause an angular rotation other than 0. So,
there are three possibilities:
Note that in all cases, the column will not buckle by itself. It will need a perturbation
(like a material imperfection, or a sieways bump) to buckle. This is true of all column
34
bending, although it is of no practical interest, since there will always be a perturbation
in a real system to cause buckling.
Let us know move on to a more realistic case: a continuous beam with length L, elastic
modulus, E, and moment of inertia, I. See figure 6.3. The free body diagram on the right
P
shows the case where the beam has deflected by some amount, y. The moment balance
will show
X
M = 0 = M + Py
M = P y.
35
Now we can see a major difference between column loading and simple bending as shown
in figure 6.4. The bending moment for simple bending is only a function of x. This is the
orthogonal direction to the deflection. Whereas, for the column, the bending moment is a
function of the deflection, y. This gives a circular effect: as deflection increases, bending
L/2
moment increases, which increases the deflection... seems like a pretty good description
of an UNSTABLE process.
Lets keep playing with equation 6.4: we replace P/EI with k 2 , so that our deflection
O.D.E becomes
y + k 2 y = 0. (6.5)
First-year (second?) maths tells us the solution to such an equation is of the form
y = A sin(kx) + B cos(kx), (6.6)
because k 2 > 0 (i.e., eigenvalues of equation 6.5 must be complex). A and B are found
from boundary conditions, which are
y(0) = 0 = B cos 0 B = 0
y(L) = 0 = A sin(kL) A = 0 or kL = n. (6.7)
36
The second result is interesting. If A = 0, then we have the seemingly trivial case where
there is no deflection. However, this is not trivial! It is, in fact, the opposite. It is the most
useful case where the column actually holds the load and doesnt buckle! Unfortunately
there is the other solution that says
n
k = , n = 1, 2, 3, ...
r L
P n
=
EI L
n 2
P = EI (6.8)
L
Note that n 6= 0 otherwise P = 0 now that would be the trivial case. So what does
this last equation mean? P is just a load, right? The values of P given by equation 6.8
must be special. They are. If P is not equal to one of the solutions to equation 6.8, then
A = 0 and the column does not buckle as
nx
y = A sin . (6.9)
L
So, when P is a solution to equation 6.8, A is unknown (could be anything) and, therefore,
in reality would become large so that the column fails. But there are lots (infinite)
solutions to equation 6.8!
Since we now know P as a solution to equation 6.8 causes buckling, let us refer to these
solutions as critical loads,
n2 2
Pcr = 2 EI. (6.10)
L
When P > Pcr the column buckles. These critical loads clearly increase as n increases, so
the lowest critical load is given by n = 1, the lowest possible value of n. This will be
2
Pcr = 2 EI. (6.11)
L
Note that the value of n determines the mode of buckling. Equation 6.11 gives the
critical load for mode 1 buckling and the corresponding buckled shape is
x
y = A sin . (6.12)
L
The mode shape of buckling is shown on the left-hand-side of figure 6.5 There are some
interesting points to raise now:
The critical load depends on EI and L2 . Therefore, making a column longer will
decrease the resistance to buckling dramatically.
37
P P
L/2
n=1 n=2
Figure 6.5: Mode 1 and 2 buckled columns. These will be the initial shapes the column
takes before failure (by cracking or yielding).
Stronger materials buckle the same as weaker materials. Only stiffness and shape
are important. i.e., a material that has a higher yield stress will not help strengthen
against buckling.
In reality, for a simply supported beam with no other constraints except those at the ends,
this will be the one and only mode of buckling. This is because once this load has been
exceeded (in reality) the column has failed and will not hold any more load in a useful
way. However, theoretically, if the load is higher than the critical load for first mode
buckling, the column will not buckle unless the load reaches the second mode, n = 2:
2 2x
Pcr = 4 EI and y = A sin . (6.13)
L2 L
This critical load is 4 times higher than the first mode, so it would be good if we could force
our column to choose mode 2, rather than mode 1. Is this possible? Yes! If we physically
inhibit the column so mode 1 cannot occur (e.g. figure 6.6). Adding these supports at the
centre of the column is a simple way to increase the resistance to buckling. This is not
really surprising, however. All we have really achieved is to create two columns on top
38
6.1. COLUMNS WITH MORE COMPLEX SUPPORTS
L/2
n=2
Figure 6.6: A column that is supported at its centre so that it cannot buckle with mode
1 shape. Thus, mode 2 is the first possible mode of buckling and the buckling load is 4
times higher than it would be without supports.
of each other, each with half the length of the original. i.e., we have effectively have two
half-length columns and we already knew that halving the length increases the critical
buckling load by 4 times. Interesting, nonetheless.
The previous section gave us the general steps to solving column problems:
39
6.1. COLUMNS WITH MORE COMPLEX SUPPORTS
6.1.1 Fixedfree
Let us first consider the case of a fixed support with free top essentially an axially
loaded cantilever as shown in figure 6.7. The free body diagram allows us to write down
P P
P
M
L L L
Figure 6.7: A column that is supported at its centre so that it cannot buckle with mode
1 shape. Thus, mode 2 is the first possible mode of buckling and the buckling load is 4
times higher than it would be without supports.
y + k 2 y = k 2 . (6.14)
This is still a linear O.D.E, but it is no longer homogeneous. This is not really a problem
for Melbourne Uni engineers who are blessed with the innate ability to solve such equa-
tions. We simply add the homogeneous solution (equation 6.6) to the particular solution,
which is (you should be able to prove this easily) and end up with
40
6.1. COLUMNS WITH MORE COMPLEX SUPPORTS
where is the deflection at the top of the beam as shown in figure 6.7. The boundary
conditions are a little more detailed as compared with the simple supports,
y(0) = B + = 0 B =
y (0) = 0 A = 0
y(L) = = (1 cos(kL)) cos(kL) = 0
The last condition is the one that will tell us what the critical loads are. Clearly the
solutions to this equation are
(2n 1)
= 0 and kL = , n = 1, 2, 3, ... (6.17)
2
The second solution reveals the critical load because k 2 = P/EI,
r
P (2n 1)
L =
EI 2
(2n 1)
corresponding mode shape y = 1 cos x . (6.18)
2L
2
Pcr = EI. (6.20)
4L2
This load is 4 times less than the critical load for a simply supported column. That
means that a fixed-free column is much weaker than the simply supported case. That is
pretty obvious, right? Now, there is something interesting here. If we look at figure 6.8, it
appears that mode 1 buckling of a fixed-free column of length L, is identical to the mode
1 buckling of a simply supported column of length, 2L. In fact, if replace L with 2L in
the critical load for a simply supported column (n = 1),
2 2
Pcr = EI = EI, (6.21)
(2L)2 4L2
41
6.1. COLUMNS WITH MORE COMPLEX SUPPORTS
P P
2L
Figure 6.8: Mode 1 buckling of a fixedfree column. The right-hand figure shows this
column behaves like a simply supported beam of length 2L.
This brings us to the concept of effective length of a column, Le . For complex supports,
we can define an Le such that we can use the simple support buckling load prediction
with L = Le , i.e.,
n2 2
Pcr = 2 EI, (6.22)
Le
Furthermore, we define a constant K such that Le = KL for the complex supported
column. In the case of fixed-free supports, K = 2.
Again, there are more modes to consider. With the fixed-free supports, the mode 2
buckling is quite interesting and is illustrated in figure 6.9. In this case, the mode 2
buckling load is,
2
Pcr = EI, (6.23)
16L2
The next example is a statically indeterminate load case. We do not need to know what
the reactions are if we are interested in buckling, so we will leave them as variables.
Figure 6.10 illustrates the problem. From a moment balance it is possible to show that
42
6.1. COLUMNS WITH MORE COMPLEX SUPPORTS
P P
L/3
n=1 n=2
43
6.1. COLUMNS WITH MORE COMPLEX SUPPORTS
P P
L L
R A
Mo
Figure 6.10: The fixedpinned column. The right-hand figure shows the deflection curve
as well as the reaction forces and moments.
2.046 2
Pcr = EI. (6.30)
L2
Thus, this support type is stronger than the simple support case. Again, this is pretty
obvious it is more supported. Finally, simple algebra gives us the effective length for
this support type:
1
Le = L K = 0.699
2.046
The final support type we will consider is the fixedfixed case, where the column is rigidly
held at both ends. In this case, both supports support a moment as well as a reaction
force (although the reaction force at each support is zero owing to the symmetry of the
problem). It can be shown that the bending moment in the column is
M = P y + Mo , (6.31)
44
6.2. COLUMNS WITH ECCENTRIC LOADS
P P
L L
Figure 6.11: The fixedfixed column. The right-hand figure shows the deflection curve.
It is, therefore, evident that Le = 0.5L for the fixedfixed configuration. That is, K = 0.5.
If K < 1, system is stronger than the simply suppported case. That is, it will buckle
under a higher load than the simply supported case.
We will now look at the case where a load is not centralised on the column. That is, the
load is offset from the neutral axis of the column by an amount, e. If you think about it, it
is very likely that real columns will experience such a loading case. Figure 6.12 illustrates
the problem. It is shown that an eccentric load, P , may be replaced with a centralised
45
6.2. COLUMNS WITH ECCENTRIC LOADS
P P
e Mo
y
L L
Mo
P
Figure 6.12: The eccentrically loaded column. The right-hand figure shows the deflection
curve and end conditions.
load of the same magnitude with the addition of a moment, Mo = P e. With this new
configuration, we can easily find the bending moment in the column:
X
M = 0 = Mo + P (y) M
M = P (e y)
From EIy = M and k 2 = P/EI, you can show that
y + k 2 y = k 2 e. (6.33)
The solution to this equation is simply,
y = A sin(kx) + B cos(kx) + e, (6.34)
which we can reduce using our boundary conditions:
y(0) = 0 = B + e B = e
kL
y(L) = 0 = A sin(kL) e cos(kL) + e A = e tan . (6.35)
2
Since we all remember our half-angle identities, there is hardly any point in reminding (!)
that
1 cos x 1
= tan x. (6.36)
sin x 2
46
6.2. COLUMNS WITH ECCENTRIC LOADS
From equation 6.37 we can determine the maximum deflection, which is simply
kL
= y(L/2) = e sec 1 . (6.38)
2
Recall that k 2 = P/EI and that, if e = 0, the load that would buckle the column is
Pcr /EI = 2 /L2 . So we could rewrite k 2 as
2 P
k2 = , (6.39)
L2 Pcr
which allows us to write the deflection in terms of load
" r ! #
P
= e sec 1 . (6.40)
2 Pcr
This is nice because the deflection is a function of the load as a fraction of Pcr and if
P = Pcr then the eccentrically loaded column has surely failed.
This little section is added for completeness, since we looked at other supports for the
centrally loaded column.
Fixedfixed: There can be no eccentric load in this case as both supports are rigid.
No matter where the load is applied, the rigid support will simply push evenly on
the whole column!
Fixedpinned: This is a little complicated... you need to go back to the bending
moments and find deflection O.D.E. again. You could show that the end result is,
kL kL
y = e tan sin kx + cos kx 1 + k tan (L x) . (6.41)
2 2
47
6.3. MAXIMUM STRESS IN A COLUMN
Fixedfree: The situation is similar to the centrally loaded column. There we found
we could just replace L with 2L in all equations. You do the same for eccentric
loading.
When a column is eccentrically loaded, there are two components making up the maximum
stress,
P Mmax c
max = + . (6.42)
A I
The first is simply due to the axial load, while the second is due to the bending moment
in the column. Note that c is the distance from the neutral axis to the extreme edge of
the cross-section of the beam. Mmax will obviously occur where the deflection is greatest,
which, for the simply supported column is at x = L/2 (the following applies to simply
supported column only). This maximum deflection, is given by equation 6.40 and
Mmax = P + P e
" r ! #
L P
= P e sec 1 + Pe
2 EI
r !
L P
= P e sec
2 EI
r !
L P
= P e sec ,
2r EA
When max exceeds the yield stress, the column fails (hopefully trivial).
48
6.3. MAXIMUM STRESS IN A COLUMN
P 2E
max = = . (6.44)
A (L/r 2 )
With these points in mind, figure 6.13 roughly illustrates the secant formula.
49