6 Stability of Structures

So far in our studies, we have been concerned with the equilibrium of structures.
Once equilibrium is established, the next level involves ascertaining the stability
of the equilibrium state. In this chapter we consider only static stability; dynamic
stability analysis, which takes into account the inertia forces and oscillations, is
also an important topic. A simple way to demonstrate the concept of stability is
to consider a ruler in the vertical position, as shown in Fig. 6.1. Using statics, we

W W

R R
(a) (b) (c) (d) (e)

Figure 6.1 Equilibrium states and perturbed states.

can balance its weight by applying a reaction R = W as seen in (a). However, a

slight perturbation from the ideal vertical position creates a moment, as shown
in (b), and the ruler falls to the side.
In Fig. 6.1, (c), (d), and (e) show stable, unstable, and neutral equilibrium of
frictionless blocks on surfaces. In (c), when the block is moved from the equilib-
rium state at the bottom of the concave surface, it returns. In (d), the block does
not return. In (e) we have neutral stability – the block stays in the perturbed state.
In order to investigate the stability, first we establish possible equilibrium states
and then perturb the system slightly to see if it would come back to the equilibrium
state nearest to it.

6.0.3 Spring-supported Vertical Bar

Consider a vertical rigid bar with a spring support as shown in Fig. 6.2. From
the vertical equilibrium position if we disturb the bar by a small angle θ, the
disturbing moment is
Md = P Lθ, (6.1)
and the restoring moment due to the stretch in the spring is
Mr = k L 2 θ. (6.2)
135

https://doi.org/10.1017/CBO9781139871983.007 Published online by Cambridge University Press

 136 Mechanics of Aero-structures

P P
k

θ
L

O O

Figure 6.2 Spring-supported vertical bar.

Note, for small angles, sin θ ≈ θ and cos θ ≈ 1.

If
Md > Mr , (6.3)
the bar is unstable, and if
Md < Mr , (6.4)
it is stable. The critical load Pcr is found from
Md = Mr , Pcr Lθ = k L 2 θ, (6.5)
which gives
Pcr = k L . (6.6)
In this example the critical load is proportional to the spring stiffness.

6.1 Buckling of a Beam

Consider a beam with bending stiffness E I and of length L, simply supported

at both the ends. If we subject it to an axial compressive load of P, it will be in
equilibrium if P is less than the yield load σ y A, where σ y is the yield stress in
compression and A is the cross-sectional area of the beam. We have to allow one
of the simple supports to move horizontally as the load is increased. Let us assume
the right end is a sliding support. We can disturb the equilibrium by deflecting
the beam in the vertical direction. This is shown in Fig. 6.3. The bending moment
at x due to the eccentricity of the load is
M = −Pv. (6.7)

M
v P
L x

Figure 6.3 Buckling of a beam.

https://doi.org/10.1017/CBO9781139871983.007 Published online by Cambridge University Press

 137 Stability of Structures

Using this in the deflection equation

E I v  = M, (6.8)
we find
P
v  + λ2 v = 0, λ2 = . (6.9)
EI
The boundary conditions are
v(0) = 0, v(L) = 0. (6.10)
The differential equation and the boundary conditions are satisfied by the trivial
solution v(x) = 0, which is the undisturbed state of the beam. Using
v(x) = C cos λx + D sin λx, (6.11)
which satisfies the differential equation, in the boundary conditions, we get
C = 0, D sin λL = 0. (6.12)
As C and D, both, cannot be zero,
nπ
sin λL = 0, λ= , n = 1, 2, . . . . (6.13)
L
The critical loads are given by
n2π 2 E I
Pcr = . (6.14)
L2
The lowest critical load is
π2E I
Pcr = , (6.15)
L2
which corresponds to n = 1. These types of problems, which allow a trivial
solution v(x) = 0 and a sequence of solutions for discrete values of a parameter
λ, are called differential eigenvalue problems, λ being the eigenvalue. For n = 1,
we have
πx
v(x) = D sin , (6.16)
L
and for any value of n
nπ x
v(x) = D sin . (6.17)
L
These shapes are called the buckling modes and the amplitude D remains an
unknown. As our solution is based on small deflections of beams, D has to be
kept small.
The first mode,
v(x) = D sin π x/L , (6.18)
has zero amplitude only at the end points. The second mode,
v(x) = D sin 2π x/L , (6.19)

https://doi.org/10.1017/CBO9781139871983.007 Published online by Cambridge University Press

 138 Mechanics of Aero-structures

is zero at the point x = L/2. This point is called a node. So, we have one node
for the second mode. For the n th mode there will be (n − 1) nodes. Thus, we can
identify the mode based on the number of nodes.
This derivation was done by Euler and it is known as Euler’s theory of buckling
of beams or columns.

6.1.1 Critical Stresses

At critical buckling load Pcr the compressive stress in the beam is
σcr = −Pcr /A, (6.20)
where A is the area of the beam cross-section. Introducing the radius of gyration
ρ as in
I = Aρ 2 , (6.21)
we can write  2
ρ
σcr = −π 2 E . (6.22)
L
The ratio L/ρ is called the slenderness ratio of the beam. The critical stress is
inversely proportional to the square of the slenderness ratio. For short beams (low
slenderness ratio) this formula, which is known as the Euler hyperbola, gives high
values for the critical stress – exceeding the yield stress of the material. We may
plot critical stress against the slenderness ratio as shown in Fig. 6.4.

C
σy

L/ρ

Figure 6.4 Domain of safe stresses in the stress versus slenderness ratio plot.

Below the Euler hyperbola and the horizontal line indicating the yield stress,
any state of stress is safe, except near the corner C, where plastic deformation
and buckling occur simultaneously. As our derivation of the Euler buckling load
does not take into account plastic strains in bending, the corner area should not
be considered as safe. There are improved theories that smooth the transition
between buckling and plastic flow.

6.2 Bending under an Eccentric Load

Consider a beam loaded in compression, Fig. 6.5, with an eccentrically applied

load. Unlike the Euler buckling, the state of the beam as a straight line does not
satisfy equilibrium.

https://doi.org/10.1017/CBO9781139871983.007 Published online by Cambridge University Press

 139 Stability of Structures

M
v
L x e
P

Figure 6.5 Eccentrically loaded beam.

The bending moment at x is

M = −P(v + e), (6.23)
and the moment-curvature relation gives
E I v  + Pv = −Pe. (6.24)
The boundary conditions are
v(0) = 0, v(L) = 0. (6.25)
Using a particular solution and a complementary solution of the differential
equation, the general solution is
v(x) = C cos λx + D sin λx − e, (6.26)
where λ2 = P/(E I ). The boundary conditions are satisfied by
1 − cos λL sin λL/2
C = e, D=e =e . (6.27)
sin λL cos λL/2
Then
 
sin λx sin λL/2
v(x) = e cos λx + −1 ,
cos λL/2
 
cos λ(x − L/2)
=e −1 . (6.28)
cos λL/2
The maximum deflection is at x = L/2, with the value
 
1
vmax = e −1 . (6.29)
cos λL/2
As the applied load P is increased from 0, the value of λ increases and so does
the maximum deflection. Finally, when λL/2 = π/2, vmax goes to infinity. The
value of P, when vmax goes to infinity, is exactly the Euler buckling load.
Using Eq. (6.29), we find
 * 
L P
Mmax = Pe sec . (6.30)
2 EI
The maximum compressive stress
P MC
σmax = + ,
A I
  * 
P ec L P
= 1 + 2 sec . (6.31)
A ρ 2ρ E A

https://doi.org/10.1017/CBO9781139871983.007 Published online by Cambridge University Press

 140 Mechanics of Aero-structures

This formula, known as the secant formula, shows the amplification of the stress
P/A by a factor that depends on the eccentricity ratio, (ec/ρ 2 ), the slenderness
ratio, (L/ρ), and the strain, (P/E A).

6.3 Buckling of Imperfect Beams

So far we have assumed the beams in question are perfectly straight. In reality, a
beam may have preexisting imperfections in the form of small deviations of the
neutral axis from the ideal straight line. In other words, we may assume there is a
deflection v0 (x) of the neutral axis before the axial load is imposed. This is shown
in Fig. 6.6. If the elastic deflection v(x) is superposed on the existing, nonelastic
deflection v0 , the change in curvature is v  .

v
P v0 P
x

Figure 6.6 Buckling of an imperfect beam.

The equation of equilibrium is obtained as

E I v  + Pv = −Pv0 . (6.32)

The unknown initial deviation v0 can be expanded in a Fourier series as

πx 2π x 3π x
v0 (x) = A1 sin + A2 sin + A3 sin + ··· , (6.33)
L L L
where the coefficients are kept as unknowns. Assuming a solution for v of the
form
πx 2π x 3π x
v(x) = B1 sin + B2 sin + B3 sin + ··· , (6.34)
L L L
and substituting it in the equilibrium equation, we can equate coefficients of the
sin-terms. This way, we find
   
π2E I 4π 2 E I
B1 − 1 = A1 , B2 − 1 = A2 ,
P L2 P L2
 
9π 2 E I
B3 − 1 = A3 , . . . . (6.35)
P L2

From these,
A1 A2 A3
B1 = , B2 = , B3 = ,..., (6.36)
Pcr
P
−1 4Pcr
P
−1 9Pcr
P
−1
where
π2E I
Pcr = , (6.37)
L2

https://doi.org/10.1017/CBO9781139871983.007 Published online by Cambridge University Press

 141 Stability of Structures

is the Euler buckling load. As in the case of eccentric loading, the amplitudes, Bi ,
of the elastic deflection begin to increase as the load is increased. Finally, when
P = Pcr , it becomes unbounded.
Neglecting the higher order Fourier coefficients, the elastic deflection at the
midpoint of the beam
δ = v(L/2) = B1 = A1 /(Pcr /P − 1). (6.38)
This can be rearranged to get
δ
δ = Pcr
− A1 . (6.39)
P
A plot of δ versus δ/P will have Pcr as its slope. This plot is called the Southwell
plot and it provides a nondestructive means of obtaining the critical load.

6.4 Cantilever Beam

A cantilever beam subjected to a compressive load shown in Fig 6.7, can be

analyzed as follows.

P
v δ
x

Figure 6.7 Buckling of a cantilever beam.

If the tip deflection

δ = v(L), (6.40)
the bending moment at a point x is given by
M = P(δ − v), (6.41)
and the equilibrium equation is
E I v  = P(δ − v), (6.42)
with the boundary conditions, v(0) = 0, v  (0) = 0. The solution of the equilib-
rium equation can be found as
P
v(x) = δ + A cos λx + B sin λx, λ2 = . (6.43)
EI
The conditions v(0) = 0 and v  (0) = 0 give
A = −δ, B = 0. (6.44)
So far we have
v(x) = δ(1 − cos λx). (6.45)
To satisfy v(L) = δ, we require
cos λL = 0. (6.46)

https://doi.org/10.1017/CBO9781139871983.007 Published online by Cambridge University Press

 142 Mechanics of Aero-structures

Then
λL = π/2, 3π/2, 5π/2, . . . . (6.47)
The critical load is obtained as
π2E I
Pcr = , (6.48)
4L 2
corresponding to the lowest value of λ.
The mode shapes are given by
v(x) = δ(1 − cos λx/L). (6.49)

6.5 Propped Cantilever Beam

A cantilever beam with a support as shown in Fig. 6.8 is called a propped

cantilever. This can also be referred to as a pinned-clamped beam (a). The forces
acting on this beam are shown in the free-body diagram (b).

(a) P

(b) P
R

Figure 6.8 Buckling of a propped cantilever beam.

With the unknown reaction R at x = L, the bending moment at a section x is

given by
M = R(L − x) − Pv. (6.50)
The equilibrium equation
E I v  + Pv = R(L − x), (6.51)
has the solution
R P
(L − x) + A cos λx + B sin λx,
v(x) = λ2 = . (6.52)
P EI
The boundary conditions are
v(0) = 0, v  (0) = 0, v(L) = 0. (6.53)
From the conditions at x = 0, we get
RL R
A=− , B= . (6.54)
P Pλ
With these constants
 
RL x sin λx
v(x) = 1 − − cos λx + . (6.55)
P L λL

https://doi.org/10.1017/CBO9781139871983.007 Published online by Cambridge University Press

 143 Stability of Structures

The last condition, v(L) = 0, gives the characteristic equation for λ,

tan λL = λL . (6.56)

The smallest root of this equation as shown in Fig. 6.9, is

20.19E I 2.046π 2 E I
λL = 4.4934, Pcr = 2
= . (6.57)
L L2
The mode shapes are given by Eq. (6.55), for different values of λL.

tan x
x

Figure 6.9 A sketch of the location of the roots of the equation tan x = x , vertical lines are spaced at π/2.

6.6 Clamped-clamped Beam

Let us consider a beam with both of its ends clamped. As shown in Fig. 6.10, the
free-body diagram shows a vertical force R and a moment M0 when the clamp is
released at the end x = L. The differential equation of equilibrium is

E I v  + Pv = M0 + R(L − x). (6.58)

(a) P

M0
(b) P

Figure 6.10 A clamped-clamped beam (a) and its free-body diagram (b).

The solution of this equation is

M0 + R(L − x) P
v(x) = + A cos λx + B sin λx, λ2 = . (6.59)
P EI

https://doi.org/10.1017/CBO9781139871983.007 Published online by Cambridge University Press

 144 Mechanics of Aero-structures

Using the boundary conditions, v(0) = 0, and v  (0) = 0, we get

M0 + R L R
A=− , B= . (6.60)
P Pλ
Then
M0 + R L R
v(x) = [1 − cos λx] + [sin λx − λx] . (6.61)
P Pλ
The remaining conditions, v(L) = 0, and v  (L) = 0, give
M0 + R L R
[1 − cos λL] + [sin λL − λL] = 0, (6.62)
P Pλ
M0 + R L R
λ sin λL + [λ cos λL − λ] = 0. (6.63)
P Pλ
Treating (M0 + R L)/P and R/(Pλ) as two unknowns, the determinant of the
system gives
(1 − cos λL)2 + sin λL(sin λL − λL) = 0. (6.64)
Expanding the first term, we get
1 − 2 cos λL + cos2 λL + sin2 λL − λL sin λL = 0, (6.65)
which simplifies to
1 − cos λL − (λL/2) sin λL = 0. (6.66)
Using the identities
1 − cos λL = 2 sin2 (λL/2), sin λL = 2 cos(λL/2) sin(λL/2), (6.67)
we see that the determinant becomes,
2 sin(λL/2) [sin(λL/2) − (λL/2) cos(λL/2)] = 0. (6.68)
Then
λL = 2nπ, n = 1, 2, . . . , (6.69)
makes the first factor zero and the critical load for this solution is
4π 2 E I
Pcr = , (6.70)
L2
which is four times the critical load for a simply supported beam. The mode shape
for these eigenvalues are obtained by noting from Eq. (6.62) that R = 0, and the
deflection has the form
v(x) = D(1 − cos 2nπ x/L). (6.71)
From the second factor in Eq. (6.68), we find
tan(λL/2) = (λL/2). (6.72)
This is the characteristic equation we found for a propped cantilever beam, with
L replaced by L/2. This equation gives a series of eigenvalues located in between

https://doi.org/10.1017/CBO9781139871983.007 Published online by Cambridge University Press

 145 Stability of Structures

the eigenvalues 2nπ. From Eq. (6.62), we find

R M0 + R L 2
= , (6.73)
Pλ P λL
and the mode shapes are given by
 
x sin λx
v(x) = D 1 − 2 − cos λx + 2 . (6.74)
L λL
These are antisymmetric about the midpoint of the beam, resembling two propped
cantilevers pinned at the midpoint.

6.6.1 Effective Lengths

To include the different boundary conditions and their effect on the critical load,
we may define an effective Euler beam (pinned-pinned) equivalent to any special
case of support conditions.
Let  denote the effective length. Then for an Euler beam, by definition,
 = L. (6.75)
For a cantilever beam
 = L/2. (6.76)
For a propped cantilever
 = 0.7L , (6.77)
and for a clamped-clamped beam
 = 2L . (6.78)
If we sketch the mode shape, the effective lengths can be visualized.
Beams in a vertical position carrying compressive load are often referred to as
columns.

6.7 Energy Method for Buckling of Beams

Energy method provides a means for approximating the critical loads in buckling.
When the beam buckles, the applied compressive load moves. First we want to
obtain the potential energy involved in this movement. We neglect the shortening
of the beam due to compressive stress and attribute the shortening to bending. As
shown in Fig. 6.11, a small triangle at x consisting of the sides x, v, and s

θ
θ0 Δs
Δx Δv δ
P

Figure 6.11 The distance moved by the compressive load during buckling.

https://doi.org/10.1017/CBO9781139871983.007 Published online by Cambridge University Press

 146 Mechanics of Aero-structures

shows
s 2 = x 2 + v 2 . (6.79)
Assuming the curved length of the beam along s is L,

 2
dx dv
= 1− . (6.80)
ds ds
The shortening, δ, can be expressed as
⎡  ⎤
 L    L  2
dx ⎣1 −
dv ⎦ ds.
δ= 1− ds = 1− (6.81)
0 ds 0 ds

Assuming dv/ds to be small, we use the binomial theorem to expand the inte-
grand, to arrive at
  
1 L dv 2
δ= ds. (6.82)
2 0 ds
With this, the potential energy of the force Pcr is
 L  2
1 dv
V = −Pcr δ = − Pcr ds. (6.83)
2 0 ds
Including the strain energy of bending, the total potential energy is

1 L! "
= E I (v  )2 − Pcr (v  )2 ds, (6.84)
2 0
where the primes indicate differentiation with s. The difference between using
s and x as the variable is negligible under the assumption dv/ds is small. The
true buckling mode makes the total potential energy a minimum. Any other mode
shape that satisfies the displacement boundary conditions will give a higher value
for . For example, for Euler buckling, instead of the true mode
v(x) = D sin π x/L , v(s) = D sin πs/L , (6.85)
if we use an approximate mode
v(s) = Ds(L − s), (6.86)
we have
v  = −2D, v  = D(L − 2s). (6.87)
 L
1
= [E I (2D)2 − Pcr D 2 (L − 2s)2 ]ds = 2D 2 L[E I − Pcr L 2 /12].
2 0
(6.88)
Setting this expression to zero, we find
12E I
Pcr = . (6.89)
L2
This approximate critical buckling load is above the exact value of π 2 E I /L 2 . We
may improve the approximation using functions with more degrees of freedom.

https://doi.org/10.1017/CBO9781139871983.007 Published online by Cambridge University Press

 147 Stability of Structures

6.8 Post-buckling Deflections

When the applied load is above the critical load, our small deflection theory
predicts indefinite maximum deflections. However, we have to use large deflection
theory with the exact expression for the curvature of the beam to proceed with
the post-buckling calculations. The exact curvature is given by
dθ
κ= . (6.90)
ds
The geometry of the deflected beam is shown in Fig. 6.11. The tangent to the
neutral axis makes an angle θ with the horizontal at an arbitrary point s. This
angle at s = 0 is denoted by θ0 . We also have
dv dx
= sin θ, = cos θ. (6.91)
ds ds
The equilibrium of the beam under an applied compressive load P requires
dθ
EI + Pv = 0. (6.92)
ds
Here, when v is positive, the curvature is negative. At this point, it is advantageous
to introduce nondimensional quantities

s → s/L , x → x/L , v → v/L , μ2 = P L 2 /(E I ), (6.93)

and rewrite the equilibrium equation, (6.92), in the form

dθ
+ μ2 v = 0. (6.94)
ds
Differentiating this equation with respect to s and using dv/ds = sin θ, we get

d 2θ
+ μ2 sin θ = 0. (6.95)
ds 2
We can convert this to a first-order differential equation by multiplying by dθ/ds
and integrating.

dθ d 2 θ dθ
2
+ μ2 sin θ = 0, (6.96)
ds ds ds
 
1 dθ 2
− μ2 cos θ = C, (6.97)
2 ds
where C is a constant of integration. From Eq. (6.92), at s = 0, we have v = 0
and dθ/ds = 0. Then

C = −μ2 cos θ0 . (6.98)

This leads to
dθ √ +
= − 2μ cos θ − cos θ0 , (6.99)
ds

https://doi.org/10.1017/CBO9781139871983.007 Published online by Cambridge University Press

 148 Mechanics of Aero-structures

where we have taken the negative square root to get negative curvature. We
rearrange this equation to separate the variables in the form
√ dθ
2μds = − √ . (6.100)
cos θ − cos θ0
Similarly, from d x/ds = cos θ and dv/ds = sin θ, we get
√ cos θdθ
2μd x = − √ , (6.101)
cos θ − cos θ0
√ sin θdθ
2μdv = − √ . (6.102)
cos θ − cos θ0
Integrals of the preceding expressions are:
√  θ0
dθ
2μs = √ , (6.103)
θ cos θ − cos θ0
√  θ0
cos θdθ
2μx = √ , (6.104)
θ cos θ − cos θ0
√  θ0
sin θdθ
2μv = √ . (6.105)
θ cos θ − cos θ0
At this juncture, we make a few observations. The first integral gives us an
equation for μ as a function of θ0 , if we assume the deflection is symmetric with
respect to the point s = 1/2 and θ = 0 at this point. Then,
 θ0
μ dθ
√ = √ . (6.106)
2 0 cos θ − cos θ0
The second integral can be written as
√  θ0 + √
2μx = cos θ − cos θ0 dθ + 2μs cos θ0 . (6.107)
θ

The third integral can be explicitly integrated to get

μ +
√ v = cos θ − cos θ0 . (6.108)
2
Traditionally, Eq. (6.106) is expressed using the Elliptic integrals, which are tabu-
lated in mathematical handbooks. However, with our currently available software,
we may do numerical evaluation of the preceding integrals. One difficulty with
the integral in Eq. (6.106) is that the integrand is unbounded at θ = θ0 , that is,
the integral is singular.
Near θ = θ0 , we let
θ0 − θ = φ, (6.109)
which is small, and
cos θ = cos(θ − θ0 + θ0 ) = cos φ cos θ0 + sin φ sin θ0 , (6.110)
cos θ − cos θ0 = sin φ sin θ0 + (cos φ − 1) cos θ0 , (6.111)
= φ sin θ0 , (6.112)

https://doi.org/10.1017/CBO9781139871983.007 Published online by Cambridge University Press

 149 Stability of Structures

Table 6.1. Numerical solutions for P/Pcr , maximum

deflection v (1/2), and the location of the right end of the
beam 2x (1/2) for various values of the angle θ0
θ0 (in degrees) P/Pcr v (1/2) 2x (1/2)
0 1 0 1
10 1.0038 0.0554 0.9984
20 1.0154 0.1097 0.9697
30 1.0351 0.1620 0.9324
40 1.0637 0.2111 0.8812
50 1.1020 0.2563 0.8170
60 1.1517 0.2966 0.7410
70 1.2147 0.3313 0.6546
80 1.2940 0.3597 0.5600
90 1.3932 0.3814 0.4569
110 1.6779 0.4026 0.2372
130 2.1604 0.3925 0.0082
150 3.1054 0.3490 −0.2223
170 5.9505 0.2600 −0.4714
179 15.2183 0.1632 −0.6735

where we have neglected terms of the order of φ 2 . Splitting the interval of

integration (0, θ0 ) into (0, θ0 − ) and (θ0 − , θ0 ) where is chosen to be small,
we write
 θ0
dθ
√ = I1 + I2 ,
0 cos θ − cos θ0
 θ0 −
dθ
I1 = √ ,
0 cos θ − cos θ0

dφ
I2 = √ ,
0 φ sin θ0
√
= 2√ . (6.113)
sin θ0
This way we avoid infinite integrands in our numerical integration. The quantity,

P L2
μ= , (6.114)
EI
can be related to the ratio of the applied load P and the Euler critical load, Pcr ,
through the relation
P μ2
= 2. (6.115)
Pcr π
As can be seen from Table 6.1, when the applied load P is increased the
deflection at s = 1/2 first increases and then decreases. Meanwhile, the point

https://doi.org/10.1017/CBO9781139871983.007 Published online by Cambridge University Press

 150 Mechanics of Aero-structures

of application of the load moves from x = 1 toward the left, eventually moving
past the origin. Along the way, the deflected beam forms a loop. The numerical
calculations were made using the software Mathematica by Wolfram.

6.9 Torsional-bending Buckling of a Compressed Beam

When thin-walled beams are subjected to a compressive load, they have a tendency
to buckle in torsion. We will limit our discussion of this topic to thin-walled open
sections with one axis (the z-axis) of symmetry. Let t denote the thickness of
the section and the s-coordinate describe the mid-plane in the counter-clockwise
sense. As shown in Fig 6.12, an element of length x and cross-sectional area

Δs
σ + Δσ
y σ
σ v
z
x
Δx

Figure 6.12 A stressed element of the beam after deformation.

ts subjected to a stress σ , in the deformed shape, produces a force in the

y-direction,
 f y x = σ ts[−v  + (v  + v  x)],  f y = σ tsv  . (6.116)
Integrating d f y over the cross-section of the beam, we get

f y = v  σ tds = N v  , (6.117)

where N is the net tension in the cross-section. Comparing the load-deflection

relation,
d 4v
= fy, EI (6.118)
dx4
and the second derivative of our buckling relation,
d 4v
= −Pv  ,
EI (6.119)
dx4
we see that a lateral loading of (−Pv  ) arises due to the curvature of the com-
pressed beam.
When there is a rotation of the cross-section superposed on the displacement
due to bending, the local displacement is s-dependent and so is the curvature.
In Fig. 6.13 C represents the centroid, S the shear center. Due to bending we
have displacements (v, w). The displacement due to rotation about S, u T , can be
expressed in vector form as u T = iθ × r s . Using
r = y j + zk = ek + r s , r s = y j + (z − e)k, (6.120)

u T = θ i × [y j + (z − e)k] = θ[yk − (z − e) j ]. (6.121)

https://doi.org/10.1017/CBO9781139871983.007 Published online by Cambridge University Press

 151 Stability of Structures

dfy
rs θ
dfz
rs y
θ
z
S e C

Figure 6.13 Rotation of a section about the shear center.

Then, the total displacements are v − (z − e)θ in the y-direction and w + yθ in

the z-direction. On an area element tds with axial stress σ , the lateral forces are
d f y = σ tds[v  − (z − e)θ  ], d f z = σ tds[w + yθ  ]. (6.122)
' '
Integrating over the area using yd A = zd A = 0, we find
f y = σ A[v  + eθ  ], f z = σ Aw . (6.123)
The forces d f y and d f z also create an incremental torque per unit of length. About
the shear center S, this torque due to bending, dTB /d x, is obtained as

dTB
= − [d f z y − d f y (z − e)],
dx A
  
 
= −σ θ (y + z + e )tds + Aev ,
2 2 2
A

= −σ [θ I S + Aev  ],

(6.124)
where I S is the polar second moment of the area about the shear center,
I S = I yy + Izz + Ae2 = A(ρ 2 + e2 ), (6.125)
with ρ being the radius of gyration. For a compressive applied load P,
σ = −P/A, (6.126)
and
dTB IS
f y = −P[v  + eθ  ], f z = −Pw ,
= P θ  + Pev  . (6.127)
dx A
Using the distributed forces in the deflection equations, we get
E Izz v  = −P[v  + eθ  ], E I yy w = −Pw . (6.128)
From Chapter 4, Eq. (4.55), we have the torque relation
G J θ  − Eθ  = TB , (6.129)
where  is the Wagner torsional-bending constant for the cross section, which
arises from the warping constraint. We introduce dTB /d x, by differentiating this
equation, to get
G J θ  − Eθ  = P(ρ 2 + e2 )θ  + Pev  . (6.130)

https://doi.org/10.1017/CBO9781139871983.007 Published online by Cambridge University Press

 152 Mechanics of Aero-structures

Although various combinations of boundary conditions are possible to complete

our formulation, let us restrict them to the simply supported case, where
v = w = θ = 0, v  = w = θ  , (6.131)
at the ends x = 0 and x = L, and to the clamped-clamped case, where
v = w = θ = 0, v  = w = θ  = 0, (6.132)
at the ends. For the symmetric section we had in mind, the deflection in the
z-direction is not coupled with the other variables, and
π 2 E I yy
Pw = (6.133)
L2
is a critical load for the simply supported case.

6.9.1 Torsional Buckling

As a special case of the differential equation (6.130), by letting v = 0, we get
G J θ  − Eθ  = P(ρ 2 + e2 )θ  . (6.134)
We may consider two types of boundary conditions at either end:
θ = 0, θ  = 0, (6.135)
when warping is prevented and
θ = 0, θ  = 0, (6.136)
when warping is allowed. In comparison to beams we refer to these two cases as
“fixed” end or “simply supported” end, respectively. If we have simple support at
both ends, using
θ = C sin π z/L , (6.137)
we get the critical load
 
1 π 2 E
Pθ = 2 + GJ . (6.138)
ρ + e2 L2

6.9.2 General Case

For the general case with one axis of symmetry, using the mode shapes
v = B sin π x/L , θ = C sin π x/L , (6.139)
and the buckling loads for the uncoupled case,
 
π 2 E Izz 1 π 2 E
Pv = , Pθ = 2 + GJ , (6.140)
L2 ρ + e2 L2
the remaining differential equations give
(P − Pv )B + PeC = 0, (6.141)
PeB + (ρ 2 + e2 )(P − Pθ )]C = 0. (6.142)

https://doi.org/10.1017/CBO9781139871983.007 Published online by Cambridge University Press

 153 Stability of Structures

For a nontrivial solution of this homogeneous system, we require the determi-

nant to vanish, that is,

ρ2
P2 − P[Pv + Pθ ] + Pv Pθ = 0. (6.143)
ρ 2 + e2

The solutions of this quadratic, P1 and P2 , give two more critical loads. Of course,
we have to use the lowest value of these critical loads as the buckling load. Note
that when e = 0, Pv and Pθ are the solutions as the system becomes uncoupled.

6.10 Example: Split Circular Cross-section Column under Compression

A split circular cross column with radius R, thickness t, and length L is shown
in Fig. 6.14. The ends of the column are built into two rigid plates, preventing

Figure 6.14 A split circular cross-section column under a load P.

warping. A compressive load P acts on the column. To simplify the calculations,

we set R 2 = t L.
To find the buckling load we use

π R3t
I yy = Ix x = I = , e = 2R,
2
2 2
J= π Rt 3 ,  = R 5 t[π 3 − 6π]. (6.144)
3 3

For the built-in end conditions, we take

     
2π z 2π z 2π z
w = A 1 − cos , v = B 1 − cos , θ = C 1 − cos .
L L L
(6.145)

The differential equation for w is not coupled to the other variables and the
buckling load is
 2
4π 2 E I 2π 3 R 3 t E R
Pw = = = πEA , (6.146)
L2 L2 L

https://doi.org/10.1017/CBO9781139871983.007 Published online by Cambridge University Press

 154 Mechanics of Aero-structures

where A = 2π Rt, the cross-sectional area. The constants Pv and Pθ are obtained
as
 2
R
Pv = π E A , (6.147)
L
 
1 4π 2 E
Pθ = 2 + GJ ,
ρ + e2 L2
   
G EA R 2
= 4π(π − 6π) +
3
, (6.148)
E 15 L
where ρ = R. Using the further assumption E/G = 2(1 + ν) = 2.6, we get the
numerical values
 2  2
R R
Pv = Pw = 3.14E A , Pθ = 10.21E A . (6.149)
L L
For the critical load for the coupled bending-torsional instability we have to solve
the quadratic
ρ2
P2 − P[Pv + Pθ ] + Pv Pθ = 0. (6.150)
ρ 2 + e2
The roots of this equation are:
 2  2
R R
P1 = 2.495E A , P2 = 64.255E A . (6.151)
L L
From this, we conclude that the minimum load for buckling is 2.495E A(R/L)2 ,
with a coupled bending-torsional mode of instability.

6.11 Dynamic Stability of an Airfoil

Let us consider a simplified model of a symmetric airfoil in a steady flow of air

or water with speed V . The airfoil is given two degrees of freedom: displacement
v perpendicular to the flow direction and a rotation θ. Restorative forces and
torques are provided by two springs. It is simpler to keep the motion v in the
horizontal plane and the airfoil span vertical. This arrangement is shown in
Fig. 6.15.

k1 k2
θ
C G v

V c1
c2
k1 k2
b

Figure 6.15 A spring-supported airfoil in a steady flow.

https://doi.org/10.1017/CBO9781139871983.007 Published online by Cambridge University Press

 155 Stability of Structures

We may assume this setup as a long wing section suspended in a wind or water
tunnel. Let C denote the aerodynamic center at quarter chord and G the center
of mass.
With a lift force L acting through C, the equations of motion are:
m v̈ = L − 2k1 (v + c1 θ) − 2k2 (v − c2 θ), (6.152)
I θ̈ = −M + Lb − 2k1 c1 (v + c1 θ) + 2k2 c2 (v − c2 θ), (6.153)
where the dots indicate time derivatives and m is the mass of the wing, I is the
moment of inertia about the point G, and
1 1
L=ρV 2 SC Lα θ, M = ρV 2 ScC Mα θ, (6.154)
2 2
with ρ the fluid density, V the flow velocity, S the plan-form area of the wing,
C Lα , and C Mα the lift and moment curve slopes, and c the wing chord. We may
shorten these expressions for the lift and the moment to have
1
ρV 2 SC Lα .
L = L θ θ, Lθ = (6.155)
2
1
M = Mθ θ, Mθ = ρV 2 ScC Mα . (6.156)
2
Using harmonic solutions of the form
v = Beiωt , θ = Ceiωt , (6.157)
the equations of motion reduce to
    
mω2 − 2(k1 + k2 ) L θ − 2(k1 c1 − k2 c2 ) B 0
= .
−2(k1 c1 − k2 c2 ) I ω + Mθ + L θ b − 2(k1 c1 + k2 c2 )
2 2 2
C 0
(6.158)
For a nontrivial solution of this system the determinant of the matrix has to be
zero. This gives a quadratic equation for the square of the angular frequency ω.
If ω2 is positive, we have oscillatory motions for v and θ. If ω2 is negative or
complex, we may have unstable solutions, which may be grouped as (a) growing
in time without oscillations and (b) growing in time with oscillations. The group
(a) behavior is called divergence and the group (b) behavior is called flutter.
The model we presented here is overly simplistic. In the case of a real wing
attached to a plane, bending and torsional stiffnesses of the cantilevered structure
replace our springs. We also have to distinguish locations of the neutral axis,
shear center, center of mass, and aerodynamic center (usually the forward quarter
chord). Additionally, the lift L and the moment M are not just functions of the
angle of attack θ; they both depend on v, v̇, v̈, θ̇ , and θ̈.

FURTHER READING
Curtis, H. D., Fundamentals of Aircraft Structural Analysis, Irwin (1997).
Fung, Y. C., An Introduction to the Theory of Aeroelasticity, Dover (1969).
Megson, T. H. G., Aircraft Structures for Engineering Students, Butterworth-
Heinemann (2007).
Rivello, R. M., Theory and Analysis of Flight Structures, McGraw-Hill (1969).

https://doi.org/10.1017/CBO9781139871983.007 Published online by Cambridge University Press

 156 Mechanics of Aero-structures

Sun, C. T., Mechanics of Aircraft Structures, John Wiley (2006).

Timoshenko, S. P., Theory of Elastic Stability, McGraw-Hill (1936).

EXERCISES
6.1 Two rigid bars, each of length L, P
are under a compressive load of P,
as shown in Fig. 6.16. There are L
three pin connections – two at the k
top and bottom supports and one
L
in the middle. A spring of stiff-
ness k resists the sideways deflec-
tion. Sketch the perturbed state for
Rigid bars under
this configuration. Balancing the Figure 6.16
compression.
moment, obtain the critical load for
the instability of this structure.
6.2 A simply supported beam has a 2EI
EI
bending stiffness of E I on the left L
P
L
half of its length 2L and a stiffness
of 2E I on the right half. This is A beam with varying
Figure 6.17
shown in Fig. 6.17. Find the critical stiffness.
load for the buckling of this beam.
6.3 A cantilever beam with bending P
L
stiffness E I has a tip support with k
spring stiffness k as shown in
Fig. 6.18. Find an equation for the
A cantilever beam with a
critical load for the buckling of this Figure 6.18
spring support.
beam.
6.4 For the beam and spring arrangement shown in Fig. 6.18, obtain the spring
stiffness k in terms of E I and L if its critical buckling load is three times
larger than that for the cantilever beam without any spring support.
6.5 A built-in beam has a bending P
stiffness of E I and a central sup- L L
k
port with a spring stiffness of k (see
Fig. 6.19). Find an equation for the
critical load for the buckling of this A built-in beam with a
Figure 6.19
central support.
beam.
6.6 The simply supported beam in Fig. 6.20 is subjected to a compressive load
P. It has a bending stiffness E I and length L. The left end of the beam
has a torsional spring with spring
k P
constant k = 2E I /L. The function
of this spring is to exert a resistive L
moment proportional to the slope.
Beam with a torsional
Obtain a transcendental equation Figure 6.20
spring.
for the constant μ = λL where
λ2 = Pcr /E I . Find the smallest root of this equation and compare it to
the Euler value.

https://doi.org/10.1017/CBO9781139871983.007 Published online by Cambridge University Press

 157 Stability of Structures

6.7 A column of length L has a y

2a
symmetric cross-section (shown in
2a
Fig. 6.21) in the shape of an I , with x
web depth 2a and flange width 2a.
The web and flanges have a thick- An I-beam under
Figure 6.21
ness of t. Calculate , I yy , Izz , and compression.
J for this section. Obtain the tor-
sional buckling load in terms of E A(a/L)2 , where A = 6at, if the ends are
simply supported and warping is prevented.
6.8 A simply supported column of length L has a solid circular cross-section
with radius R. The column is made from a material with Young’s modulus
E and density ρ. If the column is spinning with an angular velocity 
while subjected to a compressive force of P, obtain the critical value of the
frequency, , for instability.
6.9 A column has bending stiffness
4E I in its middle section of length
EI 4EI EI
2a and E I for the first and third
sections of length a. This is shown a 2a a
in Fig. 6.22. If it has simple sup-
ports at the ends with a compres- A variable stiffness
Figure 6.22
column.
sive load P, obtain the critical load
for buckling.
6.10 A column consists of two beams
with bending stiffness E I pin-
connected at the middle (Fig. k P
6.23). A torsional spring with stiff- L L
ness k is attached between them.
The spring exerts a moment resist- Figure 6.23
A column with a torsion
ing change in slope. It is free to spring.
move vertically. Obtain the criti-
cal load for this setup. Also, find
its limiting cases for k → 0 and
k → ∞.
6.11 The determinant of Eq. (6.158) is zero for nontrivial solutions. This can
be rewritten in a simpler form, by first dividing all the terms by m and
redefining B and C to obtain
 
 2 k + k2 L θ k1 c1 − k2 c2 
ω − 2 1 −2 
 
 m mc mc 
  = 0.
 k c − k c I 1 k c 2
+ k c 2 
 −2 1 1 2 2
ω2 + (Mθ + L θ b) − 2
1 1 2 2 
 
mc mc2 mc2 mc2
Rewrite this equation by introducing

ω12 = 2k1 /m, ω22 = 2k2 /m, γ1 = c1 /c, γ2 = c2 /c,

1 ρV 2 SC Lα I b
λ= , μ= , ν= , Mθ = 0.
2 mc mc2 c

https://doi.org/10.1017/CBO9781139871983.007 Published online by Cambridge University Press

 158 Mechanics of Aero-structures

6.12 Consider an NACA0012 airfoil with a chord length of 10 cm and span of

20 cm, vertically placed in a water tunnel. The lift curve slope, C Lα , for
this airfoil is 0.1 per degree of angle of attack. The airfoil is made from
balsa wood which has a density of 200 kg/m3 . The cross-sectional area of
the airfoil is 8.17 cm2 and its center of mass is at 4.2 cm from the nose.
It has a polar moment of area 44.9 cm4 about the center of mass. For this
symmetric airfoil, you may neglect C Mα with the aerodynamic center at
quarter chord. Assume both springs have k1 = k2 = 10 kN/cm. They are
at c1 = 2.2 cm and c2 = 5.0 cm. If water has a density of 1000 kg/m3
and the flow velocity is 10 m/s, obtain the frequency of oscillations of the
airfoil.

https://doi.org/10.1017/CBO9781139871983.007 Published online by Cambridge University Press