2019-20

Dept. Science et Génie des Matériaux

Mechanics of Materials

chapter 4 - Buckling of columns

Contents
1 Simplified illustration of a mechanical instability 2

2 Buckling of ideal column : Euler’s formula 3

2.1 Column pinned at both ends . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.2 Other linkages : effective length . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.3 example : calculation of a "fixed-pinned" column . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6

3 Elastic imperfect columns 9

3.1 eccentric axial load . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
3.2 Normalised parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
3.3 Column with an initial deflection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14

4 Cross-sections 16

5 Questions of chapter 4 16
INSA-Lyon - 4SGM - Mechanics of materials chapter 4: Buckling of columns

In the framework of mechanics of materials, we have seen that most materials exhibit a linear elastic behaviour under
small loads, followed by a non-linear behaviour at higher loads. Depending on the material, this non-linear behaviour
might be induced by plasticity, cracking or cavitation etc. Under some conditions, rupture can be observed before
reaching the end of the elastic behaviour. This is the case of fatigue and brittle fracture that induce abrupt fractures.
Another example of rupture or collapse of a component is buckling. This phenomenon is very important in civil
engineering for designing large structures, but might appear in any mechanical device.

The simplest illustration of buckling can be observed on slender columns (like a car antenna) submitted to a compressive
load. Increasing the load easily induces some bending of the column so that maintaining a compressive load is difficult.
The bending of such slender columns can also be promoted by adding a small load along a direction perpendicular to
the axis. This type of component ruin is due to a mechanical instability.

The buckling of columns is very different from the bending of beams. The static equilibrium of beams was calculated
considering the beam in the initial position, without load. The assumption was that the bending was small so that it
could be neglected for applying the equilibrium equations. The buckling of columns is based on different hypothesis,
i.e. we consider the bending when applying the equilibrium equations. We are also going to see that shape
imperfection are of importance when calculating the buckling of real components.

1 Simplified illustration of a mechanical instability

We first consider a very simple system to illustrate the principle of instability, but without buckling. Consider a vertical
rigid stem of length L free to rotate on the bottom part and attached to a spring of stiffness k (Figure 1). In the initial
position, i.e. when the back-force F induced by the spring is nil, the stem is vertical.

Figure 1: Vertical bar illustrating the principle of instability, at rest and after a rotation θ.

Let’s consider the case where the stem is rotated of an angle θ and submitted to a vertical load P as shown in the figure
above. If the magnitude of the force is small, such a system defines a stable equilibrium, meaning that when reducing
the load magnitude to 0, the stem is back to its initial position. On the contrary, if the load is too large, the system
collapses and the stem exhibits a large rotation.
In order to describe the mechanics equilibrium of such a system which implies the sum of the moments of the forces
with respect to point O is nil. There are two moments to describe the equilibrium. First that of the load P that induces
a positive rotation of the stem. Secondly that of the spring back-force F that induces a negative rotation of the stem.
As long as the first moment is smaller that the second one, the system is stable. Above, the system is instable, so above
a critical value of the load Pc .
Consider as a first approximation that the rotation angle remains small, therefore both moments contributions can be
written :

(Lθ)P − L(kLθ) = 0

The resolution of this equation allows to find the critical load :

2
INSA-Lyon - 4SGM - Mechanics of materials chapter 4: Buckling of columns

Pc = kL

As long as the load is smaller than this critical value, the system is stable. This means that under a small variation of
the backload dF , the system rotates of dθ, but is back to its equilibrium position θ = 0 when dF is suppressed.
Above the critical value (P > Pc ), the equilibrium is instable, since a small variation dF induces the collapse of the
system.
At the critical value of the load (P = Pc ), the system is at equilibrium for any angle value θ (still under small angle
hypothesis). These different configurations can be represented on a diagram F (θ) as shown in the figure 2. The red
lines represent the locus of the equilibrium (θ = 0 , ∀P and P = Pc , ∀θ). The intersection point between both lines is
call the bifurcation point. At that point a small perturbation makes the system instable, which is the definition of a
chaotic system. We see in the next part that the buckling of an ideal column is closed to such a system.

Figure 2: Instability diagramp = f (θ) diagram exhibiting the instability and the bifurcation point.

2 Buckling of ideal column : Euler’s formula

Ideal column means perfectly straight column without imperfection, loaded by a force applied at the centre of gravity
of a cross section along the column axis.

2.1 Column pinned at both ends

Consider an ideal straight column of length L, perfectly elastic, pinned on the bottom end and on a roller support on
its upper end. It means that any rotation is allowed at both ends, but only a displacement along z axis on the top
end (see Figure 3). This so called pinned-end column is submitted to a load P~ = −P ~z applied on the upper end (see
figure below). We assume that the column bends within the(x, z) plane. When the magnitude of the load remains low,
with this perfect system, the column remains straight and is submitted to a compression stress equal to −P/A where
A is the cross-section area of the column. The system defines a stable equilibrium. Indeed a small force applied at
any point of the column along ~x induces some bending of the column, which is suppressed with the suppression of this
small force. When the magnitude of the vertical load P~ is increased up to a critical value Pc , the equilibrium is neutral.
This means that any fluctuation of the horizontal perturbation bends the column without changing the vertical load.
For loads larger that the critical values, the equilibrium is instable, which means that a small horizontal force abruptly
induces a marked bending.
Let’s calculate the critical value and the shape of the bent column.
We consider the column in the same (x, z) plane defined in the former chapter on beams, so that the axis system is the
same (see figure 3). We have therefore two external forces : the applied load P~ and the reaction in O R~O = −P~ = P · ~z.
Applying the theory of beam mechanics, there is no force that induces bending on that system (no force along z or
moment along y). Therefore we consider the column with a pre-existing bending. This means that the axial force is
not perfectly along the z axis and exhibits some perpendicular component. In such a case the column is similar to a
deflected beam where some bending moment M fy is present.

3
INSA-Lyon - 4SGM - Mechanics of materials chapter 4: Buckling of columns

Figure 3: Straight column pinned at both ends, before and after deflection due to buckling

The calculation of this bending moment at a given point G(x, z) of the column, gives (we consider the resultant of the
forces on the bottom side of the beam) :
−−−→
~ × R~O = P x · ~y
M f G = GO

In the chapter dedicated to the bending of beams, we recall that the beam were initially considered as straight, and the
bending was charaterised by the displacement ux (z). Here we consider the bending as an initial displacement
function x(z). 1 The Bresse equation of the deflected column can therefore be written :

∂2x M fy
2
=−
∂z EIy
and replacing the expression of the bending moment gives :

∂2x P
2
=− ·x
∂z EIy
This equation is traditionally simplified defining the constant k such that
P
k2 =
EIy
substituting in the former equation leads to :

∂2x
+ k2 · x = 0
∂z 2
The general form of this equation can be obtained using the Matlab symbolic toolbox :

clear all
syms k x ( z ) E I L
assume ( k , ’ r e a l ’ )
eq1 = d i f f ( x , z , 2 ) = = - k ^ 2∗ x

eq1 ( z ) =
∂2
x (z) = −k 2 x (z)
∂z 2
x ( z ) = d s o l v e ( eq1 )

1 This point is very important, x(z) therefore represents the displacement of the point located initially at a given z and at x = 0.

4
INSA-Lyon - 4SGM - Mechanics of materials chapter 4: Buckling of columns

x(z) =
C1 cos (kz) + C2 sin (kz)
The general solution is therefore x(z) = A cos kz + B sin kz = 0.
The integration constants might be calculated from the boundary condition of nil displacement in z = 0 and z = L.
• The first constant is obtained for x(0) = 0 ⇐⇒ A = 0.

• The second constant is obtained for x(L) = 0 ⇐⇒ B sin kL = 0. This leads to two possibilities :

1. C2 = 0 which means a global solution x(z) = 0 corresponding to the straight column in its initial position,
without buckling.
2. sin kL = 0 ⇐⇒ kL = nπ with n = 0; 1; 2 . . . The case n = 0 corresponds to a nil load is therefore not
P 2 n2 π 2 EIy
interesting. For n ≥ 1, substituting the value of k leads to L = nπ ⇐⇒ P = . Substituting
EIy L2
this value in the displacement function leads to :
nπ

x(z) = B sin kz ⇐⇒ x(z) = B sin ·z .
L
This function shows a half sine shape for n = 1, a sine-like shape for n = 2 and a sine-shape with more
periods for larger values of n. These shapes also corresponds to an increasing series of values of the load :

π 2 EIy 4π 2 EIy 9π 2 EIy

P = 2
;P = 2
;P = ...
L L L2
as seen in the figure 4.

Figure 4: Plot of the series of solution x(z) for increasing values of loads P .

These different increasing values are solutions of the differential equation. However the smallest value, ie. for n = 1,
when reached induces an abrupt displacement between the column states straight and bent and is therefore the critical
value leading to buckling. It is called the Euler’s load :

π 2 EIy
PC =
L2
It does represent the critical load that induces buckling of a straight elastic slender column. This value only depends
on the geometry of the bar (length L and second moment of area I) and on the elastic modulus E of the bar. In order
to increase the critical load, ie. to limit the buckling, the length must be reduced and the elastic modulus and the
second moment of area must be increased.

5
INSA-Lyon - 4SGM - Mechanics of materials chapter 4: Buckling of columns

In the present case we assumed the column bent in the (x, z) plane. However for a column with a non-circular cross
section the bending occurs along the direction where the second moment of area is minimum.
π

Considering the displacement solution x(z) = B sin · z the maximum value is obtained in z = L/2, substituting
L
that value gives xmax = B, which is not defined. This means that at the critical load the column can exhibit
any deflection value. This means that plotting the graph P = f (x) where x can be considered at any point along the
column gives a graph similar to that given in part 4-1 (see figure 5). The critical point of coordinates (x = 0, P = Pc )
is a so called bifurcation point or instability point and the buckling is said bifurcation buckling. It is important
to notice that the calculation has been made considering the small displacement hypothesis (in the Bresse expression).
Therefore, the graph is only correct for small x values.

Figure 5: Schematic drawing of load-displacement graph of the pinned ends column, exhibiting bifurcation buckling.

2.2 Other linkages : effective length

The critical load can also be calculated for different linkages or support conditions and this results in expression similar
to the Euler’s load of pinned-end columns. Defining the so-called effective length for a column with different linkages
allows to apply the expression for calculating the critical value. The effective length is equal to that of a pinned-end
column that would give the same critical load. This is also the distance between two inflexion points of the x = f (z)
curve.
Let’s consider the case of a column fixed on the bottom and free on the top. Owing to the fact that rotation is not
possible on the bottom, such a column exhibits a displacement curve of a quarter of a sine function, as shown Figure
6. Expending this quarter sine shape, applying a simple symmetry with respect to the x axis leads to the deflection
curve of the pinned-end column, with an effective length Le = L/2. The critical load is given by the expression :

π 2 EIy
Pc =
Le 2
which in that case leads to a critical load for column fixed at the bottom and free at the top 4 times smaller than that
of the pinned end-column. Other examples of columns can be easily calculated and are given in the figure 7.

2.3 example : calculation of a "fixed-pinned" column

Let’s consider a column of length L fixed on its based and pinned at its top (Figure 8). As previously a load P~ = −P · ~z
is applied and when buckling operates a reaction arises in A RxA = −R. In O there are both a reaction force with
components along x and z and a reaction moment along y, verifying the equilibrium equations :

~ O = R~x + P ~z
(
R
M~ O = RL~y

Using Matlab, the calculation of the bending moment of a point M (x, z) gives

6
INSA-Lyon - 4SGM - Mechanics of materials chapter 4: Buckling of columns

Figure 6: Column fixed-pinned, left : buckling, right : illustration of the effective length definition.

Figure 7: Effective lengths of columns with different linkages at the ends.

syms R L x ( z ) z P E I k
R _ o = [R ; 0 ; P ] ; M_ o = - c r o s s ( [ 0 ; 0 ; L ] , [ -R ; 0 ; -P ] )

M_ o =
 
0
 LR 
0

Mf =M_ o + c r o s s ( [ - x ; 0 ; - z ] , [ R _ o ] )

Mf( z 
) = 
0
 P x (z) − Rz + LR 
0

7
INSA-Lyon - 4SGM - Mechanics of materials chapter 4: Buckling of columns

Figure 8: Fixed-pinned column exhibiting buckling.

Applying the Bresse equation leads to :

∂2x −P x + R(L − z)
=−
∂z 2 EIy
P
, using the same definition of k 2 = and substituting leads to the equation :
EIy

∂2x R(L − z)
+ k2 x = −
∂z 2 EIy

This equation can be solved using Matlab :

assume ( k , ’ r e a l ’ )
eq2 = d i f f ( x , z , 2 ) + k ^ 2∗ x = = -R/E/ I ∗ (L - z )

eq2 ( z ) =
∂2 R(L − z)
x (z) + x (z) k 2 = −
∂z 2 EI
x ( z ) = d s o l v e ( eq2 )

x(z) =
LR Rz
C3 cos (kz) + C4 sin (kz) − +
EIk 2 EIk 2
The general solution is x(z) = A cos kz + B sin kz − k2 R
EIy (L − z), with three unknown values A, B and R. The three
boundary conditions are nil displacements in z = 0 and z = L and nil rotation in z = 0.
RL


 x(0) = 0 ⇐⇒ A − 2 =0 (1)
k EIy




x(L) = 0 ⇐⇒ A cos kL + B sin kL = 0 (2)


 ∂x R
= 0 ⇐⇒ Bk + 2 =0 (3)


∂z k EIy


There is a trivial solution for A = B = R = 0 corresponding to the initial state without bending.
In order to solve the general solution, we can get rid off R from eqn(1) and eqn(3) which leads to BkL + A = 0.

8
INSA-Lyon - 4SGM - Mechanics of materials chapter 4: Buckling of columns

Rearranging eqn(3) yields A + B tan kL = 0 and substituting A from the former equation in the last one leads to

B(−kL + tan kL) = 0 ⇐⇒ tan kL = kL

. This is a non-explicit equation, and a geometric or numerical calculation must be done to find the solution. Let’s plot
"tan x" and "x" on the same graph (Figure 9) and calculate that solution :

syms x
f p l o t ( tan ( x ) , [ 0 , 6 ] , ’ lineWidth ’ , 2 ) ; h o l d on % p l o t o f tan x
f p l o t ( x , [ 0 , 6 ] , ’ lineWidth ’ , 2 ) ; h o l d o f f % plot of x
x l a b e l ( ’ x ’ , ’ FontSize ’ , 1 4 ) ;
y l a b e l ( ’ tan x , x ’ , ’ FontSize ’ , 1 4 ) ;
eq1 = @( x ) tan ( x ) - x ; x = 4 ;
kL = f z e r o ( eq1 , x ) % s o l u t i o n o f t h e e q u a t i o n

kL = 4 . 4 9 3 4

Figure 9: Plot of the function f (x) = x and f (x) = tan x that gives the graphic solution of the equation tan kL = kL.

The first solution of the equation is therefore kL = 4.493 then replacing k leads to the critical value :

4.4932 EIy
Pc =
L2
or in introducing π 2 to have the same form we obtain :

2.046π 2 EIy
Pc =
L2
Rewriting this equation in order to introduce the effective length leads to Le = 2.046−1/2 L ≈ 0.7L

3 Elastic imperfect columns

The Euler’s critical load is interesting since it stated a maximum load for an ideal column and allows to understand
the idea of buckling. However, real columns present imperfections that modify the instability. For instance even for a
straight column, the loading is never exactly along the axis. In addition real columns are never perfectly straight, but
always present an initial curvature. Finally, except materials presenting an elastic behaviour until failure, like glass or

9
INSA-Lyon - 4SGM - Mechanics of materials chapter 4: Buckling of columns

ceramic at low temperature, the mechanical behaviour is non-linear. In such cases a small deflection is obtained at the
loading onset, therefore the buckling is no-more obtained by bifurcation, but is called divergence buckling.
In order to illustrate this type of buckling we consider two cases, firstly a perfectly straight column with eccentric axial
load, and secondly a pre-deflected column with axial loading.

3.1 eccentric axial load

We again consider the pinned-end column, but with an eccentric axial load, ie. the load P~ = −P ~z is applied at a
distance e from the center of gravity of the normal cross-section. The system is displayed in the figure 10 both on a
"realistic" drawing and its schematic mechanical drawing.

Figure 10: Eccentric axial load before and after buckling.

We again assume that the column bends in the (x, z) plane and that the behaviour is perfectly linear elastic. We assume
the column bent and we calculate the bending moment at a given point z along the column axis.

M fy = P (x + e)

The Bresse equation stating the displacement curve is

∂2x M fy P (x + e)
2
=− =−
∂z EIy EIy
P
stating again that k 2 = we obtain the following differential equation :
EIy

∂2x
+ k 2 x = −k 2 e
∂z 2
Let’s calculate the solution using Matlab (even if we see that it is the same as that obtained in part 4-2-1)

syms k x ( z ) z e
assume ( k , ’ p o s i t i v e ’ ) ;
eq3 = d i f f ( x , z , 2 ) + k ^ 2∗ x + k ^ 2∗ e ;
x ( z ) = d s o l v e ( eq3 )

x(z) =
C1 cos (kz) − e + C2 sin (kz)
So the displacement is expressed as x(z) = A cos kz + B sin kz − e Both constants A and B are obtained considering
the boundary conditions in x = 0 and in x = L

10
INSA-Lyon - 4SGM - Mechanics of materials chapter 4: Buckling of columns


 x(0) = 0 ⇐⇒ A − e = 0 ⇐⇒ A = e
 x(L) = 0 ⇐⇒ e cos kL + B sin kL − e = 0 ⇐⇒ B = e(1 − cos kL) ⇐⇒ B = e tan kL
sin kL 2
The system could have been solved using Matlab :

syms A B L
x ( z ) = A∗ c o s ( k∗ z ) + B∗ s i n ( k∗ z ) - e ;
e1 = x ( 0 ) ;
e2 = x (L) ;
[ A1 B1 ] = e q u a t i o n s T o M a t r i x ( [ e1 , e2 ] , [ A, B ] ) ; mat = i n v (A1) ∗B1 ;
A = mat ( 1 )

A =
e

B = s i m p l i f y ( mat ( 2 ) )

B =
e(cos (Lk) − 1)
−
sin (Lk)
Finally the deflection curve of the column is :
kL

x(z) = e cos kz + tan sin kz − 1
2
This expression shows a relation between the displacement at any point as function of k, therefore as function of the load
P . This is fundamentally different from the ideal buckling seen in part 4-2 where for the critical load the displacement
is either nil or not defined. Here, each value of load gives a displacement that can be calculated.
Owing to the symmetry of the system the maximum displacement δ is in the middle of the column at z = L/2. Lets
calculate the value using Matlab :

x ( z ) = A∗ c o s ( k∗ z ) + B∗ s i n ( k∗ z ) - e ;
s i m p l i f y ( x (L/ 2 ) )

ans =
e cos Lk



2 −1
−
cos Lk


2

e cos kL


2 −1
 
1
δ=− ⇐⇒ δ = e −1
cos kL
2
cos kL/2

From this equation one can plot the load-displacement curve P = f (δ). The displacement is proportional to the
eccentricity. Conversely we observe that the load is not proportional to the displacement, though the system is
linear-elastic. This is at first sight very surprising, since we have seen for beam mechanics that the load-displacement
curves were proportional. This can be explained recalling that the beam is initially loaded by a moment P (e + x). At
the bending onset x = 0 then the moment is P e, so equivalent to a beam, but when the displacement increases the
moment increases. The system is equivalent to a beam where the loading is increased proportionally to the bending
which explains the non linear behaviour.
Let’s plot the graph for a bar of length L = 0.5 m of diameter d = 2 mm, of Young modulus E = 200 GPa, for different
eccentricities from e = 0.005d to e = 0.08d (Figure 11). On the same graph we display the Euler’s critical load which
is in this example equal to P c ≈ 6.2 N. It appears that the curves P = f (δ) exhibits an asymptotic behaviour with a
limit P = P c, which means that the deflection tends to be infinite when approaching P c.

11
INSA-Lyon - 4SGM - Mechanics of materials chapter 4: Buckling of columns

It can also been observed that the initial slopes of the P = f (δ)curve is not infinite and decreases when e increases.
Recalling that it correspond to the onset of the deflection, ie. for a moment equals to P e, it can be calculated considering
that it does correspond to the load-displacement curve of a beam loaded by a constant moment M = P e. In addition
if the eccentricity tends to 0, the initial moment is nil and the curve is vertical (again, case of the ideal buckling).

clear all
L = 0 . 5 ; E = 200 e9 ; d = 2 e - 3 ; I = p i ∗d ^ 4 / 6 4 ; e = 0 . 0 0 5 ∗ d ; Pc = p i ^ 2∗E∗ I /L ^ 2

Pc = 6 . 2 0 1 3

P = 0 : 0 . 1 : 6 ; k = s q r t (P/ (E∗ I ) ) ;
d e l t a = e ∗ ( 1 . / c o s ( k∗L/ 2 ) - 1 ) ;
p l o t ( d e l t a ∗ 1 0 0 0 ,P , ’ lineWidth ’ , 2 ) ; h o l d on
p l o t ( 2 ∗ d e l t a ∗ 1 0 0 0 ,P , ’ lineWidth ’ , 2 )
p l o t ( 4 ∗ d e l t a ∗ 1 0 0 0 ,P , ’ lineWidth ’ , 2 )
p l o t ( 8 ∗ d e l t a ∗ 1 0 0 0 ,P , ’ lineWidth ’ , 2 )
p l o t ( 1 6 ∗ d e l t a ∗ 1 0 0 0 ,P , ’ lineWidth ’ , 2 )
x l a b e l ( ’ \ d e l t a (mm) ’ , ’ FontSize ’ , 1 4 ) ;
y l a b e l ( ’P (N) ’ , ’ FontSize ’ , 1 4 )
g r i d on
p l o t ( d e l t a ∗ 1 0 0 0 , Pc∗ o n e s ( s i z e (P , 2 ) ) , ’ c o l o r ’ , ’ k ’ , ’ lineWidth ’ , 2 )
xlim ( [ 0 0 . 4 ] )
legend ( ’ e = 0.005d ’ , ’ 2 e ’ , ’ 4 e ’ , ’ 8 e ’ , ’16 e ’ )
t e x t ( 0 . 0 2 , Pc + 0 . 2 , ’ Pc Euler ’ , ’ FontSize ’ , 1 6 ) ; h o l d o f f

Figure 11: plot of the Load - maximum deflexion for different eccentricities, The black curve represents the Euler’s
critical load of ideal buckling

One can calculate the maximum stress in this column, noting that the maximum bending moment is in the center
at z = L/2, where x is maximum, which leads to :
Pe
M fmax = P (e + δ) =
cos kL/2
The maximum stress is the sum of the compression component induced by the load P and the bending component.
Considering that the half thickness of the column along direction x (ie. in the bending plane) is ±c, then the expression

12
INSA-Lyon - 4SGM - Mechanics of materials chapter 4: Buckling of columns

of the stress is

max P M fmax P P ec
σzz =− ± c=− ±
A Iy A Iy cos(kL/2)

Since the first component is in compression (negative), the maximum absolute value is maximum on the side of the
compression of the bending component, so when c < 0. Therefore ductile materials that are sensitive to the maximum
shear stress fail from the compression side, while brittle materials sensitive to the maximum tensile stress (larger that
ten times the compression stress) fail from the tensile side.

3.2 Normalised parameters

In order to give more insight in the conditions of divergence buckling, we often define normalised parameters. They are
the :
p
• radius of gyration defined as r = Iy /A. We define also two a-dimensional number that depends on this radius
of gyration r :
• eccentricity ratio ec/r2 and the
• slenderness ratio L/r.
The former equation can therefore be re-written (in order to simplify we consider the absolute value):
 
max P P ec P ec 1
σzz = + 2 = 1+ 2 ·
A r A cos(kL/2) A r cos(kL/2)
q
P
Substituting k = EI y
and Iy = r2 A leads to :

 r −1 !
max P ec L P
σzz = 1+ 2 cos
A r 2r EA

The maximum stress is therefore written as function of the compressive stress defined as P/A. It is interesting to
represent on a unique figure the effect of slenderness and eccentricity on buckling. For this we consider the maximum
stress that a given material can sustain and we plot the maximum load (normalised by the cross section area, P/A) as
function of the slenderness ratio for different eccentricity ratios. An example of such a plot is given in Figure 12. We
max
consider that the maximum stress is equal to σzz = 250 MPA and the elastic modulus equal to E = 200 GPA. Each
point of the graph is calculated considering given values of ec/r2 and L/r,finding the root of the above equation for the
unknown P/A. On the same graph is plotted (dash line) the special case for which the eccentricity is nil, ec/r2 = 0,
then the limit corresponds to the Euler’s relation of an ideal columns, the critical stress is therefore :

Pc π 2 EIy π2 E
σc = = 2
=
A AL (L/r)2

This value depends on the slenderness ratio and it has been plotted in red in the same figure. Therefore the case of a
nil eccentricity is represented by the horizontal line as long as the load due to the former expression is smaller. When
there is a non-nil eccentricity, say for instance ec/r2 = 1 (green curve) in the figure, for a slenderness ratio equal for
max
instance to L/r = 60, then the maximum stress σzz is reached when the apparent compressive stress is equal to
P/A ≈ 110 MPa. Clearly in the figure, when the slenderness is very high (right side of the graph) all the curves are
similar to the Euler’s simplified theory that can be applied. On the opposite for lower slenderness, the initial eccentricity
is of major importance. The curve is calculated using Matlab :

E = 200 e3 ; smax = 2 5 0 ;
x = [ 0 9 0 ] ; y = smax ∗ [ 1 1 ] ; % plot eccentricity = 0
mecc = [ 0 . 1 ; 0 . 2 ; 0 . 4 ; 0 . 6 ; 1 ; 1 . 5 ] ; mLr = 1 0 : 9 : 1 6 0 ;
mPa = z e r o s ( s i z e ( mecc , 1 ) , s i z e (mLr , 2 ) ) ;
s e e d = [ 5 3 ; 4 0 ; 4 0 ; 4 0 ; 4 0 ; 1 0 ] ∗ 1 . 5 ; % s e e d must be adapted t o s o l v e
f o r i = 1 : s i z e ( mecc , 1 )

13
INSA-Lyon - 4SGM - Mechanics of materials chapter 4: Buckling of columns

f o r j = 1 : s i z e (mLr , 2 )
e c c = mecc ( i ) ; Lr = mLr( j ) ;
eq = @( Pa ) smax - Pa . ∗ ( 1 + e c c . / ( c o s ( Lr / 2 . ∗ s q r t ( Pa/E) ) ) ) ;
Pa = s e e d ( i ) ;
mPa( i , j ) = f z e r o ( eq , Pa ) ;
end
end
E u l e r = p i ^ 2∗E . / mLr . ^ 2 ; % Euler ’ s p l o t
p l o t (mLr , Euler , ’ - r ’ , ’ lineWidth ’ , 2 )
h o l d on
p l o t ( x , y , ’ - - k ’ , ’ lineWidth ’ , 2 )
f o r j = 1 : s i z e ( mecc , 1 )
p l o t (mLr , mPa( j , : ) , ’ lineWidth ’ , 2 ) % plot of d i f f e c c e n t r i c i t i e s
end
y l a b e l ( ’P/a ’ , ’ FontSize ’ , 1 4 ) ;
x l a b e l ( ’ L/ r ’ , ’ FontSize ’ , 1 4 ) ;
ylim ( [ 0 3 0 0 ] ) ; g r i d on
l e g e n d ( ’ Euler ’ , ’ e c / r ^ 2 = 0 ’ , ’ 0 . 1 ’ , ’ 0 . 2 ’ , ’ 0 . 4 ’ , ’ 0 . 6 ’ , ’ 1 ’ , ’ 1 . 5 ’ )
t e x t ( 2 0 , smax + 1 5 , ’ $ \ sigma _ {max}$ ’ , ’ FontSize ’ , 1 8 )
set (0 , ’ defaulttextinterpreter ’ , ’ latex ’ )
t e x t ( 1 1 2 , 1 6 2 , ’ $ \ l e f t a r r o w $ E u l e r : $ \ f r a c {\ p i ^ 2 E} { (L/ r ) ^ 2} $ ’ , ’ FontSize ’ , 1 8 )
hold o f f ;

Figure 12: Maximum value of P/A as function of slenderness, L/r for different eccentricity ratios, case of a material
having a limit strength σmax = 250 MPa. On the same graph is plotted the ideal Euler’s case.

3.3 Column with an initial deflection

A second method to consider a defect in the column is to consider an initial deflection. Such a hypothesis is relevant
since, even if low, real bars always exhibit an initial curvature. In order to illustrate such an example, consider again
the pinned end column of length L, as seen in the figure 13.
We consider as a first approximation that the initial deflection x0 (z) can be described by a sine function as:
πz
x0 (z) = a sin
L

14
INSA-Lyon - 4SGM - Mechanics of materials chapter 4: Buckling of columns

Figure 13: Pinned-pinned column exhibiting an initial bending x0 = f (z), i.e. without load.

It is important to note that this hypothesis implies that the curvature is not constant, but maximum in the center
of the column. A constant curvature would require to consider an arc of a circle, but the calculation would be more
difficult. The calculation is obviously similar, but x this time represents the displacement from the initial value x0 . The
bending moment at a given point M(x0 + x , z) is :

M fy = P (x + x0 )

The Bresse equation is therefore :

∂2x P (x + x0 )
=−
∂z 2 EIy
p
substituting the constant k = P/EIy in the former expression leads to :

∂2x πz
+ k 2 x = −k 2 a sin
∂z 2 L
The general solution is the same :

x(z) = A cos kz + B sin kz

πz
and an obvious particular solution is : x0 (z) = C sin . Substituting this solution and its second derivative in the
L
differential equations leads to :

Cπ 2 πz πz πz a
− sin + k 2 C sin = −k 2 a sin ⇐⇒ C =
L2 L L L (π/kL)2 − 1

The general solution of the equation is therefore :

a πz
x(z) = A cos kz + B sin kz + sin
(π/kL)2 − 1 L

The constant can again been obtained from the boundary conditions :
(
x(0) = 0 ⇐⇒ A = 0
x(L) = 0 ⇐⇒ B sin kL = 0 ⇐⇒ B = 0 or sin kL = 0

15
INSA-Lyon - 4SGM - Mechanics of materials chapter 4: Buckling of columns

Figure 14: Example of different cross-sections

The first solution B = 0 gives the displacement curve :

a πz
x(z) = sin
(π/kL)2 − 1 L
The maximum displacement is at the center in z = L/2 which leads to :
a
xmax =
(π/kL)2 − 1
This maximum displacement is infinite when (π/kL)2 − 1 = 0, i.e. when substituting for k 2 = P/EIy leads to the
critical loads
π 2 EIy
Pc = .
L2
The total deflection of the column (considering the initial bow and the displacement) is :
a 1
δ = xmax + a = +a=a·
1 π 2 EIy 1 − P/Pc
P L2 −1
We obtain an expression of the maximum deflection as function of the applied load and the Euler’s critical load. It can
also be noted that the second solution sin kL = 0 gives again the critical Euler’s load, which represents an asymptote
of the load deflection curve.
The maximum stress (absolute value) is given by the sum of the compressive average stress and of the maximum bending
stress :
max P M fy P Pc 1
σzz = + c= + 2
·a·
A Iy A Ar 1 − P/Pc
This relation can be compared to that defined for the eccentricity loaded column, defining the parameter ac/r2 equivalent
of the eccentricity ratio, which represents the column imperfection.
 
max P ac 1
σzz = 1+ 2 ·
A r 1 − P/Pc

4 Cross-sections
All along this chapter we assume the columns to bend in the (x, z) plane. Except for circular cross-section that owing
to the symmetry can bend in any plane, the deflection operates in the direction of the minimum second moment of
area. Considering the following cross-sections (Figure 14) exhibiting two axis of symmetry (x and y), Ix < Iy and the
bending plane is therefore (y, z).

5 Questions of chapter 4
4-1 Buckling definition
- Define what is buckling.
- Give different situations where buckling may occur.
- Explain why buckling might be dangerous or critical
- Explain the difference between buckling of "ideal columns" and "imperfect columns".

16
INSA-Lyon - 4SGM - Mechanics of materials chapter 4: Buckling of columns

4-2 buckling of ideal columns

- The Euler’s formula is :
π 2 EIy
PC =
L2
detail each variable and precisely explains to which system it may apply.
- Establish the Euler’s formula given above.
- Define the effective length.

- Draw a diagram showing the bifurcation point.

- Consider a perfect column of length L = 2 m, of square cross-section of side b = 0.5 cm clamped at both ends,
and made of steel. Calculate the critical load.

4-3 buckling of imperfect columns

- Specify the two types of imperfection considered in this lecture, and explain if they are realistic and may be
applied for real columns.
- The resolution of buckling on a pinned-pinned column of length L and diameter d submitted to a force P with
an eccentricity e shows that the maximum displacement in the middle of the column is given by :
 
1 P
δ=e − 1 with k 2 =
cos kL/2 EI

(a) Considering realistic numerical values, plot the graph P = f (δ),

(b) How can we explain the non-linear P = f (δ) curve though the behaviour is elastic ?
(c) Describe the shape of this plot for large values of P, specifying what is "large value of P",
(d) How is the slope at the curve origin for P ≈ 0,
(e) Calculate the value of that slope. For this you could consider a beam with a given constant moment (specify
the value of that moment).
- Consider a pinned-pinned column of axis z made of steel, of length L = 1 m, diameter d = 4 mm, submitted to a
force P = 0.8Pc with an eccentricity e = 0.03 mm.

(a) Where is located the maximum stress ?,

(b) Specifying the component, write the expression of the maximum stress.
(c) Plot the graph σzz = f (x) at z = 0.5L.
- Consider the graph of Figure 12,

(a) explain what do the axes represent,

(b) describe the different curves,
(c) try to understand the Matlab code used to calculate the curves.

17