VIBRATIONS OF STRUCTURES

Prof. A. Dasgupta
Mechanical Engineering
IIT Kharagpur
 INDEX

[Link] TOPICS [Link]

Week 1

1 Lecture -1 : Transverse Vibrations of Strings - I 4

2 Lecture -1 : Transverse Vibrations of Strings - II 18

3 Lecture 3 : Axial and Torsional Vibrations of Bars 27

Week 2

4 Lecture 4: Variational Formulation - I 37

5 Lecture 5: Variational Formulation - II 46

6 Lecture 6: Modal Analysis - I 52

7 Lecture 7: Modal Analysis - I I 63

Week 3

8 Lec-08 Properties of the Eigenvalue Problem 70

9 Lec-09 Modal Analysis: Approximate Methods - I 78

10 Lec-10 Modal Analysis: Approximate Methods - II 92

Week 4

11 Lecture 11: The Initial Value Problem 103

12 Lecture 12: Forced Vibration Analysis - I 116

13 Lecture 13: Forced Vibration Analysis - II 127

Week 5

14 Lecture 14: Forced Vibration Analysis - III 135

15 Lecture 15: Dumping in Structures - I 147

16 Lecture 16: Damping in Structures - II 153

1
 Week 6

17 Lecture 17 : Beam Models - I 159

18 Lecture 18 : Beam Models - II 167

Week 7

19 Lecture 19 : Modal Analysis of Beams 177

20 Lecture 20 : Application of Modal Solution 185

21 Lecture 21 : Approximate Methods 194

Week 8

22 Lecture 22: Topics in Beam Vibrations - I 203

23 Lecture 23: Topics in Beam Vibrations - II 210

Week 9

24 Lecture 24 : Dynamics of Curved Beams 220

25 Lecture 25 : Vibrations of Rings and Arches - I 227

26 Lecture 26: Vibrations of Rings and Arches - II 232

Week 10

27 Lecture 27 : Dynamics of Membranes 239

28 Lecture 28 : Vibrations of Rectangular Membranes 246

29 Lecture 29 : Vibrations of Circular Membranes - I 255

30 Lecture 30 : Vibrations of Circular Membranes - II 260

Week 11

31 Lecture 31: Dynamics of Plates 266

32 Lecture 32: Vibrations of Rectangular Plates 274

Lecture 33: Vibrations of Circular Plates

33 282

2
 Week 12

34 Lecture 34 : Special Problems in Plate Vibrations - I 291

35 Lecture 35 : Special Problems in Plate Vibrations - II 293

3
 Vibration of Structures
Prof. Anirvan DasGupta
Department of Mechanical Engineering
Indian Institute of Technology – Kharagpur

Lecture - 01
Transverse Vibrations of Strings - I

In this lecture, we will discuss transverse vibrations of strings. Before we start discussing
about vibrations of strings, let us look at the definition of the string. (Refer Slide 1.1).

Slide 1.1: (Refer Slide Time: 00:38)

This definition will be clear once we look at some examples of strings. (Refer Slide 1.2) We
have an ordinary tag 1 which is like a string since it is a one dimensional continuum and it
does not transmit or resist bending in any way. The restoring force comes only when we
make it taut. So, we can say that the restoring force is produced by the tension in the string. In
the absence of the tension force, the string will admit to any shape we give. Another example
of a one dimensional continuum which does not resist bending is a chain 2 or a hanging
chain which also qualities to be a string. Let us now look at a guitar string named as 3 . If
we try to bend it, it restores back as can be seen. It violates the definition of the string, but we
still call it a string.

To understand the reason for this, let us see what happens in a guitar. In a guitar, the string is
under tremendous amount of tension which is the prime reason of the primary restoring force
in the string. The string of course, restores to its original straight shape after it is released
from the bend, but when it is put in a guitar under tremendous amount of tension, the tension

4
becomes the dominant restoring force. And, hence any structural element which is under high
tension qualifies to be analysed, in the first approximation, as a string. The Elements that may
be modelled as taut strings are found in stringed musical instruments such as sitar, guitar,
violin, even in the piano. We have seen all such instruments in which the sound is basically
produced by the vibrations of the string.

Slide 1.2: (Refer Slide Time: 02:41)

Similarly, the cables of a cable state bridge or a cable car are also under tremendous amount
of tension and hence they can also be analysed as strings. The high tension (electric) cables,
which are under very high tension, can also be treated as taut strings. Let us now start with a
mathematical model. In order to model strings, first we will make some assumptions which
are listed in Slide 1.3. Let us first try to understand each of these assumptions (Slide 1.4).

The first assumption says that the motion of the string is planar which implies that the string
vibrates only in one plane, plane of the paper in the case shown. The second assumption says
that the slope of the string is small. When the string deforms, the slope at any point of the
time is small. The third assumption says that the longitudinal motion at any point in the
domain of the string is negligible. If we make a mark on the string (see up-down arrow mark
in Slide 1.4) and trace the motion of this mark, as the string vibrates, we will find that this
mark moves transverse to the string most of the time. We hardly notice any axial motion at
this point. The fourth assumption says that the tension does not change with displacement of

5
the string. So if we put some tension in the string and displace it then the change in tension is
negligible. Having made these assumptions, we can now start modelling the string.

Slide 1.3: (Refer Slide Time: 04:24)

Slide 1.4: (Refer Slide Time: 05:00)

Transverse
motion only
Very small slope

6
Slide 1.5: (Refer Slide Time: 06:22)

We consider a string made of a material of density ρ and an area of cross section A(x) which
maybe a function of the positional coordinate x. It is under a tension T and has a length l. The
transverse motion of the string is measured by the variable w(x, t). Slide 1.5 shows a stretched
or a taut string which has been displaced from its equilibrium position which is the x-axis. In
order to write out the equations of motion, we will consider an infinitesimal element as we do
in Newtonian mechanics. Slide 1.5 shows the free body diagram (FBD) of this infinitesimal
element. This element lies between the coordinate x and x+Δx. On the left end, the element is
under a tension T(x, t) and it makes an angle α with the horizontal direction. On the right end,
the tension is T(x+Δx, t) and it makes an angle α(x+Δx, t) with the horizontal. The stretched
length of this element is Δs. To begin with, we are going to write the equations of
longitudinal dynamics of this infinitesimal element. Since we have assumed that the
longitudinal motion of this element is negligible, we will neglect the inertia force i.e., the
acceleration in the longitudinal direction. Thus, the longitudinal dynamics reduces to just a
force balance equation in the longitudinal direction as,

𝑇(𝑥 + 𝛥𝑥, 𝑡) 𝑐𝑜𝑠[𝛼(𝑥 + 𝛥𝑥, 𝑡)] − 𝑇(𝑥, 𝑡) 𝑐𝑜𝑠[𝛼(𝑥, 𝑡)] + 𝑛(𝑥, 𝑡)𝛥𝑥 = 0 (1.1)

Where, n(x, t) is the external force distribution in longitudinal direction per unit length of the
string. Now, if we divide Eq. 1.1 by Δx and take the limit Δx→0, it will imply

7
 𝜕
[𝑇(𝑥, 𝑡) cos{𝛼(𝑥, 𝑡)}] + 𝑛(𝑥, 𝑡) = 0 (1.2)
𝜕𝑥

Since we have assumed that the angle α made by the string is very-very small. So, we can
safely assume that cos α ≈ 1. Thus, Eq. 1.2 simplifies to
𝜕𝑇
+ 𝑛(𝑥, 𝑡) = 0
𝜕𝑥
Or, in a shorter form we can write
𝑇,𝑥 + 𝑛(𝑥, 𝑡) = 0 (1.3)

In Eq. 1.3, ‘x’ in the subscript would indicate a partial derivative with respect to x and we are
going to follow this notation throughout this course. Thus we have found our equation for the
longitudinal dynamics which is essentially a force balance equation.

Slide 1.6: (Refer Slide Time: 14:40)

Let us now look at the transverse dynamics. Mass of the element of length Δs (Δs ≈ Δx up to a
linear approximation) shown in the FBD (Slide 1.5) can be written as ρAΔx. Since w(x, t) is
the motion in the transverse direction, w,tt represents the acceleration of the element. Using
Newton's 2nd law of motion, we write

8
Where, p(x, t) is the external force distribution in the transverse direction per unit length of
the string. Now, if we once again divide Eq. 1.4 by Δx and take the limit Δx→0, it will imply

In order to simplify Eq. 1.5, once again we bring 2nd assumption from Slide 1.3 into force,
which says that the slope of the string at any position and time is small. Let us see what does
it actually mean. It is shown in Slide 1.6 that up to the linear order approximation sin α ≈ tan
α = w,x. Using this along with some rearrangement, we get the equation of the transverse
dynamics of a string.

Slide 1.7: (Refer Slide Time: 19:52)

We now have the dynamic equations of the string (Refer Slide 1.7). Eq. 1.6 is the linear
second order hyperbolic partial differential equation. Since the equation is of second order
both in space and time, we will need boundary and initial conditions, two in number of each
kind. Let us first try to understand the need of these conditions. The equation of motion, as
we have seen is derived by considering an infinitesimal element of the string. This in no way
tells us how the string is connected to the ground or if at all it is connected to the ground.
Therefore, in order to complete the physical description of the whole system, we will need
the boundary conditions at the two ends of the string. Mathematically, the differential

9
equation has the space derivative of second order which upon integration will generate two
constants. Hence, in order to determine these constants of integration, we need the two
boundary conditions. In a similar manner, when we integrate the time part we will once again
generate two constants of integration which will be solved from the two initial conditions.
The boundary conditions, therefore, complete the description of the system and they are used
for the determination of the constants of the integration (Refer Slide 1.8).

Slide 1.8: (Refer Slide Time: 22:19)

Slide 1.9: (Refer Slide Time: 22:31)

The boundary conditions are of two types (Refer Slide 1.9). The first type is known as the
geometric or essential boundary condition which is fixed by the geometry of the problem.

10
The second type is the dynamic or natural boundary condition which comes because of some
condition on the force or moment. For strings, it is usually the condition on force. Let us now
look at some examples. We consider a uniform taut string (Refer Slide 1.10). In the absence
of the external force in the longitudinal and the transverse direction of the string, Eq. 1.3
becomes T,x = 0 which implies T : Constant. Using constant tension T, the transverse
dynamics Eq. 1.6 simplifies to (Refer Slide 1.11)

From the geometry of the given uniform taut string, one can easily figure out the boundary
conditions (Refer Slide 1.12). We can see that the transverse displacement at both the
boundaries is restrained which implies w(0, t) = 0 and w(l, t) = 0. Thus, in this case we have
geometric boundary conditions.

Let us now consider another example where we have a string with a free or sliding end (Refer
Slide 1.13). Since, there are no external force distribution acting on the string, the equation of
motion will be the same as that in the previous example (Eq. 1.7). Let us now look at the
boundary conditions. There is a geometric boundary condition at the left end which says that
the transverse displacement is constrained at x = 0 i.e., w(0, t) = 0. We now look at the right
boundary in a little detail.

There is a mass-less and frictionless pulley to which the string is attached at the right end. We
will first draw the free body diagram of this connection (Refer Slide 1.11). At the pulley, we
have a normal force N from the frictionless guide. At the string end, we have the tension T
and angle α from horizontal at x = l and at any instant of time t. Now if we write down the
equations of equilibrium for the pulley in the transverse direction, we get

Using the approximation sin α ≈ tan α = w,x, we can write Eq. 1.8 as

11
Slide 1.10: (Refer Slide Time: 23:13)

Slide 1.11: (Refer Slide Time: 23:32)

Eq. 1.8 essentially means that the force in the transverse direction on the pulley is zero. Since
the tension in the string is uniform, Eq. 1.9 can be written as

12
Thus we get the boundary condition at the sliding end of the string. As we have noticed that
the boundary condition has come from a condition on force that the forces on the pulley in
the transverse direction must vanish. Such a boundary condition is known as a natural
boundary condition. Thus, we have seen that a force free boundary gives a natural or a
dynamic boundary condition. We can note here that we can solve for the normal reaction N
from the guide by using the force balance in the longitudinal direction.

Slide 1.12: (Refer Slide Time: 24:49)

Slide 1.13: (Refer Slide Time: 25:28)

13
Let us now look at another example in which we have a hanging string or a chain (Refer Slide
1.14). We have discussed in the beginning that a chain qualifies to be analysed as a string.

Slide 1.14: (Refer Slide Time: 30:28)

Slide 1.15: (Refer Slide Time: 30:55)

Consider a hanging chain made up of a material of density ρ, uniform cross-section area A

and length l. We can easily recognise here that the chain will be under varying tension which
implies T = T(x) where x is the position coordinate along the length of the chain. So, our first

14
objective is to determine T(x). Looking at the force distribution n(x, t) in the longitudinal
direction, one can see that n(x, t) = ρAg where the quantity ρAg is the weight of the chain per
unit length. Now we use the equation of longitudinal dynamics which was in fact just the
force balance equation in the longitudinal direction (Eq. 1.3). We integrate it and use the
boundary condition T(l) = 0 to determine the constant of integration (Refer Slide 1.15). Thus,
we have

We now substitute the Eq. 1.11 in the equation of motion of the transverse dynamics Eq. 1.6
with no external force distribution in the transverse direction [p(x, t) = 0]. Upon
simplification, we get

Slide 1.16: (Refer Slide Time: 36:34)

Thus, we get the equation of motion of the transverse dynamics of a hanging chain in a
uniform gravitational field. Once again, we will need boundary conditions to complete the
description of the problem. Let us look at them (Refer Slide 1.16). At x = 0, we have a
geometric boundary condition. Since the motion is constrained at x = 0, so the boundary
condition would be w(0, t) = 0. At x = l, force free condition prevails which we have seen in
the previous example also (Eq. 1.10). It implies that Tw,x(l, t) = 0. If we consider the last

15
particle of the chain (at x = l), one can notice that it has no restoring force in the transverse
direction since it is the free end of the chain and the tension T is zero at x = l. Therefore, there
exists a possibility, at least theoretically that the displacement might become infinitely large.
But from our experience, we know that a hanging chain or a vibrating chain does not go to
infinity at the free end. The loose end of the chain in fact remains within finite limit. So from
the physical consideration, we must have a finite solution at this end. Thus we write the
boundary condition at the free end as an inequality w(l, t) < 1. We will elaborate on that in
detail when we discuss the solution of vibration of a hanging chain in subsequent lectures.

Slide 1.17: (Refer Slide Time: 42:49)

Let us now see what we have discussed in this lecture so far. We have started with the
equations of motions of a string, the transverse as well as the longitudinal. In the longitudinal
direction, it is just a force balance because we have assumed that the motion in the
longitudinal direction is negligible. Then we have derived the equation of motion in the
transverse direction of the string assuming it to be planar. Then we have looked at the
boundary conditions that come up in the transverse vibrations of strings. There are two kinds
of boundary conditions as we have seen. The first one is the geometric boundary condition
and second one is known as the natural or the dynamic boundary condition. And then we
have seen a few examples of taut strings along with the dynamics of a hanging chain. In each
of these cases we have looked at the boundary conditions that govern the equations.

16
Let us now discuss the initial conditions. As we have discussed earlier that we will need two
initial conditions since the equation of motion is of second order in time. So, these are usually
specified as the initial deformation of the string w(x, 0) = w0(x) and the initial velocity
distribution over the string w,t(x, 0) = v0(x) (Refer Slide 1.17). With these two initial
conditions, we can now completely/uniquely solve the equation of motion of a vibrating
string. With this, we complete the transverse vibrations of strings.

Keywords: string, transverse vibrations, hanging string, string with sliding end, boundary
conditions, initial conditions

17
 Vibration of Structures
Prof. Anirvan DasGupta
Department of Mechanical Engineering
Indian Institute of Technology – Kharagpur

Lecture - 02
Transverse Vibrations of Strings - II

In the second part of our discussions on transverse vibrations of strings, we are going to look
at strings with interaction.

Slide: 2.1 (Refer Slide Time: 00:28)

Pylon

Hangings

Multi-strand cable (acting

Masses as spring and damper)

Taut String

Cable-car
Violin

By the term interaction, we would mean that the string is connected to an actuator or another
system which could be a passive system like an absorber or a damper. Let us now look at
some examples of strings with interaction. The high tension (voltage) cables, for example, are
strings interacting with absorbers or dampers as shown in Slide 2.1. If you look carefully
around the support points of the cable, you might have observed few hangings, enlarged view
of which is shown in Slide 2.1. These hangings are loaded with masses and a multi-strand
cable which acts like a spring and a damper. These hangings are known as Stockbridge
dampers which are used to damp out the wind-induced vibrations. Similar strings with
dampers are also found in the piano.

18
In order to understand the effect of such external devices, we will have to analyze the string
along with these absorbers or dampers. The strings interacting with another system are also
found in the bridge of a violin. The bridge (shown in Slide 2.1) is used to transmit the
vibrations of the string to the sound board or the main body of the violin so that it can be
amplified. Also, we can have strings which are connected to an actuator which may excite
within the span of the string or at the boundary. Further, we can have strings with moving
loads, typical example of which would be a cable car on a taut string as shown in Slide 2.1.
Thus, we have seen that there are various examples in which a string may interact with an
external, discrete or continuous system. In this lecture, we will model and analyze such
strings.

Slide: 2.2 (Refer Slide Time: 08:30)

Equation of motion for oscillator

Let us begin with a simple example. Let us consider a string made of a material of density ρ,
area of cross section A, under a uniform tension T and length l as shown in Slide 2.2. We
consider an oscillator connected to the string at a distance ‘a’ from the origin. Let the mass of
this oscillator be m and the stiffness be k. Let us first look at the interaction force diagram.

For interaction force diagram, I have separated the oscillator from the string and introduced
the interaction force P(t) at the point of separation instead. Let us consider the oscillator part
first. If we consider the coordinate of the mass point as y, then we can write the equation of
motion for the oscillator as shown in Slide 2.2.

19
Now, it can be observed that P(t) is the same force that acts on the string at a location ‘a’.
Since the mass has to be coupled to the string, we have y = w(a, t). If we make this
substitution in the equation of motion of oscillator, we get the force P(t) in terms of string
deflection w, the negative of which, in fact acts on the string.

Slide: 2.3 (Refer Slide Time: 14:18)

Equation of transverse
motion of the string

Force distribution

Interaction force diagram of the string part is redrawn where P(t) is the concentrated force
acting at a location ‘a’ as shown in Slide 2.3. It should be noted here that the string for which
equation of the transverse motion is sought, is uniform with no axial force. Thus, we write the
equation of motion of the string (discussed in lecture 1), where p(x, t) is a forced distribution
or the force per unit length (shown in Slide 2.3). But for the string in question, we have a
concentrated force P(t) acting at x=a. To remove this discrepancy, we have to represent the
concentrated force as a force distribution. Let us look at a force distribution p(x, t) on a string.
We now find the total force fT(t) acting on the string by integrating it over the domain of the
string (0 to l) as shown in Slide 2.3. We can always imagine a situation where the force
distribution gradually becomes narrower and narrower but keeping the same total force. The
force distribution happens to be zero everywhere except on a finite region of the string. Thus,
we can imagine the nonzero force region shrinking to zero, in the case of a concentrated
force. This is what exactly happens in the present problem that the force (total force is equal
to P(t)) is acting only at a particular point.

20
For the same amount of total force, if the region of its application shrinks to zero, it will
cause its amplitude or the magnitude to blow up to infinitely large values. But the area under
such a distribution (integral under such a distribution) is certainly finite since it is equal to the
total force P(t). In order to remove this discrepancy, we write the integral using Dirac delta
function as shown in Slide 2.3.

Now, we can easily derive the properties of the Dirac delta (listed in Slide 2.4). First of all,
the distribution must be zero, everywhere on the string except at x=a. Secondly, the integral
under the Dirac delta distribution over the domain of the string is one which implies that the
area under this distribution is one. It can be further generalized (Property 2) and written in
manner as shown in Slide 2.4. Conclusively we can say that the Dirac delta distribution
achieves a mathematical representation of a physical idealization.

Slide: 2.4 (Refer Slide Time: 23:44)

Let us understand the meaning of the physical idealization? In the macroscopic world, we do
not have concentrated forces. It implies that no force exists which acts over zero area. We
always have forces acting over a distributed area. Let us look at one example. In the violin,
string portion on the bridge is much-much smaller than the overall length of the string or the
damper in a piano makes contact on a very small region as compared to the total length of the
string. In such cases, we can always have a physical idealization that this is almost a point

21
contact. So the Dirac delta distribution gives us a mathematical description of this
idealization. Now using this, we will analyze the string.

Finally, the concentrated force is represented as a force distribution as shown in Slide 2.5.
Thus, with slight rearrangement, we can write the equation of motion as shown in Slide 2.5.
Negative sign with the force distribution is used as per the sign convention that ‘upward’ is
positive. It can be seen here, that we have additional linear mass density (mass/length) and
stiffness terms due to the presence of oscillator.

Slide: 2.5 (Refer Slide Time: 26:06)

Force distribution

Due to oscillator
Mass/length

Thus, this completes the description of a string interacting with an oscillator which is coupled
to it. As an exercise you can try out an example given in Slide 2.5, in which there is a string
made up of material of density ρ, constant area of cross section A, under a tension T and
having a length l and there is an absorber connected to the string. The absorber has a mass m
and stiffness k attached at x = a. Derive the equations of motion. Here you have to note that
there will be two equations, one for the string and the other is for the absorber because the
mass of the absorber is now having a separate coordinate y.

Now, let us look at the cable car problem. In the first approximation, you may consider the
force exerted by the cable-car to be a constant force moving at a constant speed v on the
string. Once again, this constant force F has to be represented as a force distribution p(t)

22
 using the Dirac delta function as shown in Slide 2.6. This force distribution on the string can
be written as F times the distribution, but this distribution is shifting. So I am assuming that at
time t = 0, this force was at the origin, that means at x = 0. It implies p(t) = Fδ(x-vt). Now,
we can write the equation of motion for the cable car problem in the form as shown in Slide
2.6. Boundary conditions, in this case are geometric boundary conditions as mentioned in
Slide 2.6. We can generalize this problem further. Suppose we have a bead of certain mass
moving on a string. We can find out the corresponding interaction force because now the
bead will have the acceleration which comes because of its interaction with the string and that
will produce a varying force because of inertia of the bead. Thus, we can generalize this
problem of cable-car on string to analyze more and more realistic situations.

Slide: 2.6 (Refer Slide Time: 33:11)

Force distribution

Equation of motion

Boundary conditions

Next, let us look at an example of a string with boundary excitation. We have a string which
is free to slide on its left boundary. I will draw a displaced configuration of the string. Here,
we will assume that this string is being driven by a robust actuator which actually gives a
displacement condition on the boundary. So this string is being excited by a system or an
actuator which is robustly driving the left boundary. The equation of motion for the string
which is made up of a material of density ρ, area of cross section A and under tension T can
be written as shown in Slide 2.7. The boundary conditions are as follows. At the left
boundary, the string is given a displacement which is a function of time h(t) and at the right
boundary, the transverse displacement is zero. Thus for both the boundaries, we have a

23
condition on the displacement of the string which implies both of these boundary conditions
are geometric boundary conditions. With non-homogeneous boundary conditions, analysis
may become little difficult. We shall look for the ways to convert the non-homogeneous B.C.
into the homogenous B.C. because that in many ways simplifies the analysis of the problem.

Slide: 2.7 (Refer Slide Time: 37:21)

Slide: 2.8 (Refer Slide Time: 44:55)

24
There are several ways of converting the non homogenous boundary conditions to
homogenous boundary conditions. We will follow the method which makes use of variable
transformation. Therefore, we want to find a transformation of the field variable w(x, t) such
that the resulting [Link]. are homogenous. Introducing the transformation w(x, t) = u(x, t) +
η(x) h(t) as shown in Slide 2.8. Here, u(x, t) is a new field variable and η(x) is an unknown
function which we have to find out. Now, if we substitute this transformation into the
boundary conditions, we get boundary conditions on new field variable u and on function η as
shown in Slide 2.8. We can choose such a form of η(x) which can satisfy these conditions.
The simplest choice would be the one as mentioned in Slide 2.8. Of course, we can have
higher powers of this function. For e.g., η(x) = 1 – (x/l)2 can also satisfy these conditions.
Here, we will make one choice at this point and discuss the implications.

If we put the chosen form of η(x) back into the transformation, and that into the equations of
motion, we get the equation of motion (EOM) in new field variable u and homogeneous
B.C.s as mentioned in Slide 2.9. It can be noticed here that the B.C.s are homogeneous but
the EOM has become non-homogeneous. If we look carefully, the term in the RHS of EOM
is the result of the transformation and this transformation has taken the problem from inertial
frame to a non- inertial frame. Thus, with the choice of η(x), we have shifted to a frame
which is indicated by the green line which is moving as shown in Slide 2.9. Thus, we get the
EOM in non inertial. As we go to a non inertial frame, we have the inertia force which is
essentially coming on the RHS of the equation of motion and making it non homogenous.

Slide: 2.9 (Refer Slide Time: 51:01)

25
Had we used the different form of η(x), let us say η(x) = 1 – (x/l)2, then we would get a
transformation which will take to the quadratic form of non-inertial frame as shown in Slide
2.9. With this we have come to the end of this discussion on transverse vibrations of strings.
In this lecture, we have discussed, strings with interaction with external oscillators and
external systems and we have seen how to represent concentrated forces, how to represent the
interaction and how to convert a problem with non-homogenous boundary conditions to a
problem with homogenous boundary conditions.

Keywords: Transverse Vibrations, String with interaction, Dirac Delta Distribution,

Non-homogenous Boundary Conditions

26
 Vibrations of Structures
Prof. Anirvan DasGupta
Department of Mechanical Engineering
Indian Institute of Technology – Kharagpur

Lecture – 03
Axial and Torsional Vibrations of Bars

In this lecture, we are going to look at two examples on vibrations of one dimensional elastic
structure other then the strings (already covered in lecture – 1 & 2). We are going to look at
axial vibrations of bars and torsional vibrations of circular bars. Let us begin with axial
vibrations of bars. We find bars undergoing axial vibrations in an ultrasonic machine which
consists of a bar called horn, the shape of which is shown in Slide 3.1. The horn of the
ultrasonic generator is connected to an actuator which passes ultrasonic waves through the
bar and because of its shape; we get large amplitude vibratory motions at the work piece as
shown in Slide 3.1. Such a machine is used for machining brittle materials.

Slide: 3.1 (Refer Slide Time: 00:47)

Actuator

↕
Work-piece

Jack Hammer

Further, we find bars under axial vibrations in Pneumatic hammers which are also known as
Jack hammers as shown in Slide 3.1. These hammers are used for drilling or chipping
operations at construction sites. Also, in Piezo-actuators or sensors, we find a bar made of
Piezo-electric material under axial vibrations. Likewise, in various structural elements, we
may find bars in axial vibrations.

27
Slide: 3.2 (Refer Slide Time: 04:57)

In order to model the dynamics of bars in axial vibrations, we begin with some assumptions.
1. All points at any cross section of the bar will have same motion as shown in Slide 3.2.
2. The strain due to axial deformation is small so that we do not have nonlinear effects.
In this lecture, we are going to discuss only linear vibrations of bars.
3. There is no transverse motion of the bar which means that the material points are
vibrating only along the axis of the bar.
4. The material of this bar is linear, homogenous and isotropic.

Slide: 3.3 (Refer Slide Time: 07:56)

Equation of
motion

28
With the above stated four assumptions, we are now going to look into the equation of
motion of a bar in axial vibration. Consider a bar made of material of density ρ, area of cross
section A(x) which may be a function of the spatial coordinate x, Young's modulus of
elasticity E and length l. Now at any location x of the bar, the displacement of the cross
section in the axial direction is measured by the field variable u(x, t) as shown in Slide 3.3.

We now derive the equation of motion of this bar using the Newtonian approach. We
consider an infinitesimal section of the bar, free body diagram of which is shown in Slide 3.3.
The small element considered is of length Δx which lies between x and x + Δx. Thus, stress
on the left and right face can be written as σ(x, t) and σ(x + Δx, t) respectively as shown in
Slide 3.3. Similarly, the area of cross-section on both the faces is defined as A(x) and A(x +
Δx).

Now we write the equation of motion for the infinitesimal element using Newton’s second
law. The mass of the small element may be written as Δm = ρAΔx. The mass times the
acceleration in the longitudinal direction which is u,tt must be equal to the forces in the
longitudinal direction which are σ(x + Δx, t) A(x + Δx) - σ(x, t) A(x) as shown in Slide 3.3. If
we divide the whole equation by Δx and take the limit Δx → 0, we get the equation of motion
for the bar undergoing axial vibrations as shown in Slide 3.3.

In the equation of motion, we can represent the stress in terms of the displacement which is
the field variable u(x, t). In order to do that, we need two things. The first is the material
constitutive relation, which will relate the stress with the strain. We know the Hooke’s law in
one dimension which says that the axial stress is proportional to the strain and the
proportionality constant is the Young’s modulus. Along with this, we need the strain
displacement relation which gives ε = u,x. If we substitute the strain-displacement expression
in the constitutive relation and put that back in the equation of motion, we obtain, on slight
rearrangement of terms the equation of motion for axial vibrations of a bar as shown in Slide
3.3. Note that the equation of motion has been derived by considering a small element of the
bar that in no way tells us or describes to us the full physical picture of the bar.

In order to complete the description of the bar in axial vibrations, we need the boundary
conditions. As we have discussed earlier, these boundary conditions are of 2 types namely
geometric and natural boundary conditions. Let us now look at some examples and identify

29
the boundary conditions. We begin with a uniform bar. Since the area is constant for uniform
bar, it is no longer a function of x. Therefore, it can be taken out of the partial derivative and
the equation can be simplified further to obtain the equation of motion for a uniform bar as
shown in Slide 3.4.

Slide: 3.4 (Refer Slide Time: 17:59)

Equation of motion
for uniform bar in
axial vibrations

For the boundary conditions of this bar, we can see that the leftmost end of the bar is
completely fixed at the wall. Therefore, at x = 0, there cannot be any axial motion of the bar
which implies u(0, t) = 0. Since this boundary condition is fixed by the geometry of the
problem, this is a geometric boundary condition. The right end of the bar is free which means
that there is no axial force at this face of the bar. We know that the force on any cross section
is given by F = σA. Thus, at x = l, the axial force must be zero which implies σA│x=l = 0.
Now, using the Hooke’s law and the strain-displacement relation, we obtain the relation
between the displacement u and the stress σ as σ = E u,x as shown in the Slide 3.4. Thus, the
boundary condition at x = l becomes EA u,x│x=l = 0. Since cross-section area A is uniform,
the boundary condition turns out to be EAu,x (l, t) = 0. This boundary condition at the right
end comes from a force condition. That is why such a boundary condition is a natural
boundary condition. Thus, we finally have the equation of motion of a uniform bar along with
the boundary conditions which completes the description of a fixed free bar in axial
vibrations.

30
Slide: 3.5 (Refer Slide Time: 23:49)

Equation of motion for

axial vibration of bar

Equation of motion
for oscillator

Next we look at another example. We have a fixed-free bar, at the free end of which there is
an oscillator attached in the manner shown in Slide 3.5. There is a point mass m attached with
the spring of stiffness k. The displacement of the mass m from the equilibrium position is
measured by the coordinate y. Let us first write down the equations of motion for this system.
Here we have a field variable u(x, t) which measures the displacement of the material points
of the bar and the variable y(t) which measures the displacement of the discrete point mass m.
The equation of motion of the bar remains the same as before as shown in Slide 3.5. Now, for
the oscillator which is connected at the right end of the bar, we can easily write the equation
of motion by taking it separately as shown in Slide 3.5. One can see that both the equations of
motion are coupled.

Now let us look at the conditions at the boundary. The left end of the bar is completely fixed.
Therefore, we can say that the displacement is zero at this end. On the right end of the bar
where this oscillator is attached, we can expect an interaction force from the oscillator. We
already know the force on any cross section as EAu,x(l, t). Had there been no oscillator, this
force would have been zero. Since there is an oscillator we can write the force due to the
oscillator at this end as k[y – u(l, t)]. This gives us the boundary condition at the right end of
the bar. Thus, we have a geometric boundary condition at the left end whereas on the right
end, we have a natural boundary condition.

31
Slide: 3.6 (Refer Slide Time: 29:48)

With the above mentioned two examples, we now move on to the next example where we are
going to consider torsional vibrations of circular bars. We find such vibrations in the torque
transmitting shafts in rotating machinery like rotors of turbines, crankshafts of engines etc.
Also, we find wires under torsion which are also called drill stings in the dentist’s drill. We
find drill strings in petroleum excavation and mining industries as well. In a mine, we have a
shaft under torsional vibration which transmits the torque to a drill head as shown in Slide
3.6. Though there is some amount of transverse vibrations in the shaft but the torsional
vibrations are quiet dominant in such situations. Thus, we have seen various situations where
we find shafts or circular bars undergoing torsional vibrations.

Once again, for the mathematical model of such a bar, we make some assumptions as
mentioned in Slide 3.7.
1. The bar is circular. In this lecture, we are going to study only the torsional vibrations
of circular bars. (Non-circular bars undergo wrapping at each cross section which
makes analysis quite complicated.)
2. We assume that the strains are small so that the dynamics can be adequately described
by a linear model.
3. We also assume that there is no transverse motion of the bar.

32
Slide: 3.7 (Refer Slide Time: 33:44)

With these assumptions, let us now look at a circular bar. Here we have a circular bar which
is made of a material of density ρ and an area of cross section which may be a function of the
spatial coordinate x as A(x). The shear modulus or the modulus of rigidity is represented by G
and the bar has a length l.

Slide: 3.8 (Refer Slide Time: 35:14)

Before
deformation

Ring After
Axial Lines deformation

Radial Lines
Shear Strain

At any location x, the field variable that measures the torsion or the local torsional
displacement of the bar is represented by φ(x, t) as shown in Slide 3.8. Once again we are
going to derive the equation of motion of such a bar in torsional vibrations using the
Newtonian approach. Consider a small portion of the bar between the spatial coordinates x

33
and x + Δx. On the right end of this portion of the bar, let us consider the moment represented
by M(x + Δx, t). On the left end, we have the moment on the cross section as M(x, t). Now
inside this little element, we consider a ring of certain radius r as shown in red in Slide 3.8.
Let us draw this ring separately for clarity. We now consider, in the undeformed ring of
radius r, two axial lines and the corresponding two radial lines as shown in the Slide 3.8.

When the bar is under torsion, there will be a differential rotation between the left and the
right face of the little element considered due to which the red element made up of two axial
lines which is somewhat like a rectangular element is going to take up the configuration as
shown in Slide 3.8. This small differential rotation between the left and the right face will be
Δφ (angle between the deformed and undeformed radial lines) as shown in Slide 3.8. On the
other hand, the angle between the deformed and undeformed axial lines would be the shear
strain ψ(r, t) experienced by the initially undeformed rectangular red element as shown in
Slide 3.8. It should be noted that the shear strain is a function of the radial position of this
element. Deformed and undeformed rectangular elements are redrawn in the Slide 3.8 for
showing the shear strain in this element clearly. With this kinematics, we can now write r Δφ
= Δx ψ. For Δx→ 0, we can write ψ(r, t) = r φ,x. This is the equation that we obtain from the
kinematics of the torsional deformation of the bar. We can now relate the shear strain at any
radius r, at any time t, in terms of the field variable φ.

Slide: 3.9 (Refer Slide Time: 46:11)

34
It can be noted here that the shear strain is also a function of x as φ is a function of x. Now we
are going to use this kinematic relation further. Let us now look at the constitutive relation of
the material. We know from Hooke’s law that the shear stress is proportional to the shear
strain. If we use the kinematic relation obtained in Slide 3.8, we get the expression of the
shear stress as shown in Slide 3.9. Once we have the shear stress, we can multiply this with
the area and integrate it over the whole area of the face to get the moment/torque in the bar as
shown in Slide 3.9. Here, we have used the definition of polar moment of the area Ip = ∫A(x)
r2dA to simplify the integral.

Thus, we have related the torque at any cross section in terms of the torsional displacement φ.
We can now write down the equation of motion. First, we will write the moment of inertia of
the ring of radius r as ∫A(x) ρdAΔx r2 where ρdAΔx is the mass of the ring. The mass moment of
inertia times the angular acceleration which is the double time derivative of the field variable
φ must be equal to the balance of torques. We already have the expression of moment/torque
at any face which we can use. If we divide this whole equation by Δx and take the limit Δx →
0 and if we use the definition of the polar moment of the area, we get the equation of motion
of torsional dynamics of a circular bar as shown in Slide 3.9.

Slide: 3.10 (Refer Slide Time: 54:01)

Equation of motion

Let us now look at some examples. Say, we have a uniform bar. Then the equation of motion
simplifies to the form as shown in Slide 3.10 as the term GIp is constant. The boundary

35
conditions for this problem are as follows. The bar is connected to the wall at the left end.
Therefore, this end of the bar cannot have any rotation. So this is a geometric boundary
condition as shown in Slide 3.10. On the right end of the bar, the torque is zero which is
therefore given by M(l, t) = GIpφ,x(l, t) = 0 which is a natural boundary condition.

With this, we complete our discussions on axial and torsional vibrations of bars. To
summarise the lecture, we have considered axial vibrations of bars and derived its equations
of motion. We have seen the boundary conditions of two types; namely the geometric
boundary condition and the natural boundary condition. We have also derived the equations
for a bar interacting with discrete system and then finally we have looked at the dynamics of
torsional vibrations of circular bars.

Keywords: Axial Vibrations of Bars, Torsional Vibrations of bar, Natural Boundary

Conditions, Geometric Boundary Conditions

36
 Vibrations of Structures
Prof. Anirvan DasGupta
Department of Mechanical Engineering
Indian Institute of Technology – Kharagpur

Lecture - 04
Variational Formulation - I

In this lecture, we are going to initiate some discussions on an alternate formulation of

dynamics of continuous systems. This formulation is known as the variational formulation.

Slide: 4.1 (Refer Slide Time: 00:27)

Trajectory

Before we get into this variational formulation, we have to understand a few concepts on
which this technique is based upon. Let us understand the concept of configuration space. We
picture a taut string which has been displaced from its equilibrium position which is the x-
axis. Suppose we want to represent or track the configuration of this string. For achieving
this, the simplest thing that we can think of is to track certain material points on the string.
For example, we take seven arbitrary material points. We call the displacement corresponding
to each of these points as w1, w2, w3…w7. To visualize the configuration of the string at the
present instant, we can think of a Euclidian like space.

Here in the Slide 4.1, I have shown only three axes namely w1, w2 and w3. Ideally,
corresponding to each material point, we should have had as many axes. Since we can’t draw

37
a space with seven axes, we have to rely on our imagination to think of a space in which there
are seven such axes. We now mark on these axes of 7-dimensional space, the displacement of
each of the material points chosen as shown in Slide 4.1. Thus, we get a point which
represents the initial configuration of the string (shown with blue in Slide 4.1). The string
now takes a different configuration as shown with red in Slide 4.1. For the changed
configuration, the seven material points share the same locations as that by the points of the
initial configuration as shown in Slide 4.1. Therefore, in a 7-dimensional space, both the
configurations of the string will have the same representative point which is in contrast to the
fact that both the configurations of string are indeed different as shown in Slide 4.1. To
remove this incongruity, we keep on increasing the number of material points to capture all
possible configurations. Thus, we get an infinite dimensional space which can represent all
possible configurations of the string. To represent all possible configurations, we require field
variables, for example w(x, t) in this case where w is the transverse displacement of the string
from the equilibrium. The space in which a point represents any configuration of the string is
known as the configuration space of the string. Therefore, we can clearly see here that for any
continuous system like string, the dimension of the configuration space is infinity.

Thus, any continuous system is represented as a point in an infinite dimensional configuration

space and the number of coordinates of the configuration space represents the degree of
freedom of the system. Therefore, the degree of freedom of a string or any continuous
system is infinity. Keeping the fact that in the configuration space, any configuration of a
continuous system is represented by a single point, in the mind, we note that as the string
vibrates or as it moves through configurations, there are trajectories in the infinite
dimensional configuration space through which the string moves as shown in Slide 4.1.

While we have understood the concept of configuration space, let us now look at the
variational formulation of dynamics. We imagine an infinite dimensional configuration space.
At time t = t1, we observe that the configuration of the system is represented by a point t = t1
as shown in Slide 4.2. For a system like string, the point in the infinite dimensional
configuration space represents the configuration in the string at time t = t1 as shown in green
in Slide 4.2. Now, if we allow the string to move (Observer remains blind to the process of
transition.), it takes a new configuration at t = t2 as shown in black in Slide 4.2 which is
represented as another point in the configuration space of the string.

38
 Slide: 4.2 (Refer Slide Time: 08:48)

Paths Configuration of string

at different time instants
Configuration of the system
at different time instants in
the infinite-dimensional
configuration space

At this point, the following question arises. How did the string move from configuration at t
= t1 to the configuration at t = t2? We can think of several ways the string could have moved.
Can I say which path the string followed in the configuration space with certainty? Is there a
way of knowing this since we have not seen how it has evolved from one state to another?
This question is answered by something known as the Hamilton’s principle. It says that of
infinitely many available paths for the system to move from configuration at t = t1 to another
configuration at t = t2, the one the system follows will extremize the action which is defined
as an integral from t1 to t2 of a scalar known as the Lagrangian which is defined as a
difference of the kinetic and potential energies of the system.

To summarize, we observed two configurations of the string but we have not observed the
intermediate configurations through which the system might have gone through configuration
1 to configuration 2. The Hamilton’s principle says that the path taken by the string
(intermediate configurations attained by the string) in moving from configuration 1 to
configuration 2 extremizes the action which is defined in the manner as mentioned in Slide
4.2.

Let us now understand the meaning of extremization. Mathematically speaking,

extremization has a connotation of minimization or maximization. In mechanical systems
such as a string or a bar, the extremization actually implies minimization. Here, we have to

39
understand that the action is a function of paths in the configuration space which are the
functions of time. Given the path taken by the string from configuration 1 to configuration 2
which is a function of time, we have defined the action as an integral over the scalar function
L. Thus, the action can be seen as a function of a function, hence called a functional.
Therefore, in this context, extremizing implies minimizing over functions. We have to find a
function which minimizes the Action which is slightly different from minimization of only
functions.

Slide: 4.3 (Refer Slide Time: 19:01)

Extremum points

Taylor Expansion

Suppose we have a function f(x). How do we find out the extremum points for a given
function? If the plot of f(x) versus x looks like as shown in Slide 4.3, we can easily locate the
extremum points. Let x* is an extremum point which is the maxima in the case shown. How
do we detect that the point chosen is a local maximum? For that, we make a test. Suppose x*
is a solution of an extremum point. We disturb or perturb the point x* by a small amount εy
where ε is a small quantity. Now, if we take a difference between the perturbed and
unperturbed quantities [f(x*+εy) and f(x*)] and divide the difference by ε and take the limit
ε→0 as shown in Slide 4.3 and consequently, it turns out to be zero for arbitrary perturbations
y then we say that we have found an extremum. We observe that the mathematical form so
obtained is the standard definition of a derivative df/dx│x=x*.

It should be noted here that x is a variable. The formulation presented here in Slide 4.3 will

40
work even if x is a vector. Let us see this in detail (The independent variable x in the
derivative is now xi). If we Taylor expand the first term and use limit ε→0 for arbitrary
perturbations y, we get the vanishing derivative of function df/dxi at vector x = x* for all i as
shown in Slide 4.3. Let us draw an analogy from here to understand the Hamilton’s principle.

Slide: 4.4 (Refer Slide Time: 25:00)

Condition for
extremization of
Action

Property of Commuting
with partial derivative

Let us imagine that the action has been calculated for different functions as shown in the
figure in Slide 4.4. On the x-axis now, we have different functions w. Let us say w* is the
actual path taken by the string as it moves from configuration 1 to configuration 2. If we
perturb it with another function η and a small quantity ε, take the difference and divide it by ε
and take the limit ε→0 and if this turns out to be zero then we say that we have found a path
that the string has taken to move from configuration 1 to configuration 2. We have seen that
the action has been calculated as an integral from t1 to t2 of the Lagrangian. In the case of
functions, we say df/dx vanishes for extremum, here we say δA vanishes for the extremization
which implies the condition as shown in Slide 4.4. Here the symbol δ is called variational
operator. It is very similar to the total derivative operator except that it does not differentiate
time. Finding the variation of action δA implies perturbing paths while time is held frozen.
Thus, variation operator is like a total derivative operator with time frozen. The second
property of this operator is that the operator commutes with partial derivatives as shown in
Slide 4.4.

41
Slide: 4.5 (Refer Slide Time: 32: 13)

Equilibrium
configuration

Now let us see the application of Hamilton’s principle. Here, we have a string whose
transverse displacement is represented by the field variable w. We assume that the string is
made of a material of density ρ, area of cross section A, length l and is under a tension T as
shown in Slide 4.5. Firstly, we write the Lagrangian and construct the action integral on
which we will apply the variational formulation or Hamilton’s principle. Let us find the
kinetic energy. We consider a small element of the string at a certain location x. The mass of
this infinitesimal element would be ρAdx where dx is the length of the element. If we multiply
the velocity square of this little element and integrate it over 0 to l and multiply by half then
we obtain the kinetic energy of the string. Let us now find the potential energy. If we look at
an infinitesimal portion of the string, we have the length of the equilibrium configuration as
dx which becomes ds after it has been displaced. This change in the length is taking place
under a tension T which we have assumed not changing with displacement. This was one of
the assumptions of our model. In fact tension does change but that change is assumed to be
negligible. Also, there is no axial force on the string. Now, we need to find the length of the
string after deformation i.e., ds. The traverse displacement at the left and right end of the
element is w(x) and w(x + dx, t) respectively. We find the length ds in terms of dx and w,x as
shown in Slide 4.5.

Let us now look at the work done by the constant tension force T as the string stretches. It is

42
 found as follows. We integrate the product T(ds - dx) over the length of the string giving us
the work done which is stored as potential energy in the string. We substitute the form of ds
found earlier and expand it assuming that w,x is small. Finally, we get the expression of
potential energy of the string.

Slide: 4.6 (Refer Slide Time: 41:10)

First Function

First Function

Variation of configuration
over two time instants
must vanish.

Let us now look at the variational principle. We put KE (T) and PE (V) expressions in the
Hamilton’s principle. Now we apply the variation operator on the integrand to obtain the
form as explained in Slide 4.6. Here, δ(w,t2) gives 2w,t δw,t. The variation operator can
commute with the time and space derivative i.e., δ(w,t) = δ(ꝺw/ꝺt) = ꝺ(δw)/ꝺt = (δw),t and
δ(w,x) = δ(ꝺw/ꝺx) = ꝺ(δw)/ꝺx = (δw),x where δw is the small perturbation on the function w.
Since we want to separate out the small arbitrary perturbation over w i.e., δw, we would like
to have something in terms of δw. To obtain that we integrate the term w,t δw,t with respect to
time t and w,x δw,x with respect to spatial coordinate x by parts. If we commute the integral,
we can write the equation as shown in Slide 4.6. It should be noted here that the variation of
the configuration δw has to be calculated over two time instants t1 and t2. We have seen in the
beginning of the formulation of the problem that there cannot be any variation in the
configuration of the string at these two time instants. Here, we are not expecting any variation
at time instants t1 and t2 since there is none whereas we are looking at the variations at the
intermediate times. Therefore, the variation δw must vanish at time t1 and t2 as shown in Slide
4.6.

43
 Finally, we get the form as shown in Slide 4.7 where we have the variation of configuration
in the form Tw,x δw at the boundaries. Also, we have terms at the intermediate positions of
the string. If entire left hand side of the equation has to vanish for arbitrary δw, we get the
equation of motion for the transverse vibrations in string along with the term Tw,x δw
vanishing at the two boundaries which gives us the boundary conditions as mentioned in
Slide 4.7.

Slide: 4.7 (Refer Slide Time: 51:40)

Variation of configuration
at the boundaries

At the intermediate
positions in the
domain of the string
Equation of motion

We can see that the term Tw,x δw can vanish in two ways. At x = 0, vanishing δw means that
w is fixed for all times which implies w(0, t) = 0 (Geometric b. c.) OR Tw,x (0, t) = 0
(Natural b. c.). Similarly, at x = l, we get w(l, t) = 0 (Geometric b. c.) OR Tw,x (l, t) = 0
(Natural b. c.).

We can see here that various combinations are possible at the two boundaries and we can
now easily recognize that the condition on zero displacement (w(0, t) = 0 and w(l, t) = 0)
which is a condition of fixed end is a geometric boundary condition whereas Tw,x (0, t) = 0
and Tw,x (l, t) = 0 is a force condition. Different possibilities at both the ends of the string are
following.
1. One end is fixed while other end is sliding.
2. Both ends may be sliding.
3. Both ends may be fixed.

44
Thus, variational formulation gives us the equation of motion along with all the possible
boundary conditions. This method is in fact a very powerful method for formulating the
equation of motion of very complicated systems. The variational formulation is not just an
approach for finding out equations and boundary conditions but also, as we will see later, a
method which will help us or lead to numerical methods for computational purposes.

To summarize, we have discussed in today’s lecture is the variational formulation of

dynamics of the continuous systems. We have looked at the Hamilton’s principle which
forms the basis of this formulation. We have looked at configuration space of the continuous
system based on which the Hamilton’s principle works. Then we have finally taken an
example of a string and derived its equation of motion and also obtained the possible
boundary conditions for the string.

Keywords: Variational Formulation, Configuration Space, Hamilton’s Principal,

Lagrangian, Extremization, String Vibration

45
 Vibrations of Structures
Prof. Anirvan DasGupta
Department of Mechanical Engineering
Indian Institute of Technology – Kharagpur

Lecture - 05
Variational Formulations II

Let us resume our discussions on the variational formulation that we started in previous
lecture. In this lecture, we are going to look at the transverse vibrations of a hanging chain
which is an inextensible continuum. We are going to look at the equation of motion of such a
chain.

Slide: 5.1 (Refer Slide Time: 01:04)

= ra Use dx = dx i and
(KE)
Deformed ds = rb + dx - ra
configuration

= rb

2
2 Expression for
Axial deflection
(PE)

2
2 2

Let us consider a chain hanging in a uniform gravitational field. We assume that the chain of
constant area of cross section A and length l is made up of material of density ρ. Though it is
an inextensible chain, it does not resemble a string which is a one dimensional elastic
continuum. Yet we will see that the equation of motion for the hanging chain is similar to that
of a string, up to a linear order at least.

For variational formulation, let us write down the energy expressions first. For the field
variable w(x, t), the kinetic energy (KE) T of an infinitesimal element of length dx at any
location x is mentioned in Slide 5.1. The potential energy (PE) of the element is derived in

46
Slide 5.1. Let us understand it a little. Now we want to understand why the chain restores
back to its vertical position. One can easily see that as the chain deflects, the potential energy
of it changes. The PE increases in fact. Now to calculate this PE increase, let us look at a
small element of this chain as shown in Slide 5.1. Here, I have shown the undeformed and
deformed configuration of the chain. The infinitesimal element considered has a length of dx.
Since we have assumed the chain to be inextensible, the length of the infinitesimal element in
the equilibrium configuration will be same as in the deformed configuration. Thus, the length
in both configurations is dx. For the length of the element in the deformed configuration, I
will represent it as a vector ds whose magnitude |ds| = dx. The vectors connecting the
corresponding ends (at x and x + dx) of deformed and undeformed configuration are also
mentioned in Slide 5.1. Now, if we represent the vector ds in terms of the deflection of the
chain w and take the magnitude of the vectors on both sides, we get the relation between dx,
w,x and u,x as shown in Slide 5.1. Using the approximation that u,x << 1, we simplify the
expression as shown in Slide 5.1. Thus, we have the representation of the axial deflection of
the chain.

One can note that the situation regarding the axial deflection is very different in the case of
hanging chain from that in strings. In strings, we had neglected the axial motion, but in a
hanging chain we must consider the axial motion since that is the reason why the potential
energy of the chain is changing as the chain deflects from its equilibrium position. Once we
have the expression for the axial deflection, we can now write down the potential energy (PE)
V of the chain as shown in Slide 5.1. Here, mass of the infinitesimal element of the chain is
taken as ρAdx. The expression for PE is in the form of an integral over the length of the chain
assuming that the potential energy is zero in its equilibrium configuration. Now, we can
integrate the expression of PE by parts where u is taken as the first function and ρAg as the
second function as shown in Slide 5.1. Upon simplifying the boundary term (evaluated at l)
and other integral term by substituting the expression of axial deformation u(l, t), we get the
final form of PE as shown in Slide 5.1. Thus, we have the kinetic and the potential energy
expressions of the chain.

If we look at the expression of kinetic energy, we find that it resembles the kinetic energy of
an elastic string. Similarly, the expression of the potential energy is similar to that of a string
except for the term ρAg(l-x). In the elastic string, we have a tension term instead which
happens to be a constant. On the contrary, in the case of a hanging chain, the tension varies

47
 with the location x.

Slide: 5.2 (Refer Slide Time: 15:38)

Lagrangian

Hamilton’s Principle

Evaluated over
the Boundary

Integrated over
domain 0 to l
Equation of motion

Now the Lagrangian may be expressed in the manner as shown in the Slide 5.2. From
Hamilton’s principle which says that the variation of the action is zero, we integrate the
kinetic and potential energy terms by parts as shown in Slide 5.2. Since we know the
configuration of the chain at the time instance t1 and t2, there cannot be any variation of the
configuration at those time instants. Thus, the variation δw at times t1 and t2 must vanish. It
should be noted here that there is a term which is evaluated only at the boundaries and
another term which is evaluated (Integrated) over the full domain. We can always hold the
boundary variation (the terms evaluated at the boundary only) fixed and change the other
term arbitrarily. Therefore, if the sum δA has to vanish, the terms comprising the sum must
vanish separately. Thus, for arbitrary variation of δw (over 0<x<l) and vanishing the integral
over the domain, we get the equation of motion of the hanging chain as shown in Slide 5.2.
Now, by vanishing the boundary term in the equation, we get the boundary conditions as
shown in Slide 5.3. Let us have a look at the boundary conditions. At x = 0, we get either a
fixed or a free chain boundary condition. At x = l, the tension term ρAg(l-x) vanishes
naturally. Thus, for the boundary condition at this end to vanish, we must have finite
variation of w which implies w(l, t) < ∞. The implications of this boundary condition will be
seen when we discuss this solution procedure for the string.

48
 Slide: 5.3 (Refer Slide Time: 24:08)

Chain fixed at the

Chain free at the top end
top end

Zero at x=l

Next, we are going to slightly generalize the procedure as shown in Slide 5.4. We have been
looking at the action integral in which the Lagrangian may be a function of the velocity w,t,
slope w,x, field variable w and time t. Performing δA = 0 with integration by parts, we get the
equation of motion and boundary conditions as shown in Slide 5.4.

Slide: 5.4 (Refer Slide Time: 27:56)

Equation of motion

Let us now look at an application of this generalized formulation for vibrations of bars as

49
explained in Slide 5.5. We will consider the axial vibrations of a bar first. The field variable
is u(x, t). First, we write down the kinetic energy T. If ρ is the density and A is the cross
section area, then the kinetic energy can be written in the manner as shown in Slide 5.5.

Slide: 5.5 (Refer Slide Time: 36:08)

From theory of elasticity

Fixed boundary at x=0

Equation of motion
Force free
boundary at x=l

The potential energy V of the bar may be written as shown in Slide 5.5, where Adx is the
volume of the small element. Here, we have used the results from the theory of elasticity to
find the potential energy in terms of u,x. Now, defining the Lagrangian or what is more
appropriately called as the Lagrangian density as shown in Slide 5.5. Here, we will make use
of the generalized equation of motion for a given Lagrangian L as derived in Slide 5.4. Here,
you can see that there is no exclusive dependence of the Lagrangian density on u. Therefore,
the term ꝺL/ꝺu is zero. Substituting the Lagrangian in the generalized equation of motion and
the boundary conditions, we get the equation of motion and boundary conditions for the axial
vibration of a bar as shown in Slide 5.5. Here, we have to choose which boundary condition
is applicable for the problem at hand.

Let us discuss the problem of torsional vibration of a circular bar as well. If φ is the field
variable, then we can write the kinetic energy T in the form as shown in Slide 5.6. For
potential energy of the bar, we again make use of the theory of elasticity and get the
expression for potential energy V as shown in Slide 5.6.

50
Slide: 5.6 (Refer Slide Time: 44:07)

From theory of elasticity

Here, we have used the definition of polar moment of inertia as Ip = ∫A r2 dA. Thus, Lagrangian
density is given by the expression as mentioned in Slide 5.6. Now, once again we apply the
generalized equation of motion from Slide 5.4 and find the equation of motion and boundary
conditions as shown in Slide 5.6. At x = 0, there is no twist in the bar and at x = l, the bar is
free. Thus, we have a geometric B. C. at x = 0 and a natural B. C. at x = l as shown in Slide
5.6.

To summarize the lecture, we started with the dynamics of a hanging chain. We derived its
equation of motion and the boundary conditions. Then we generalized the variational
approach to find the equation of motion and the boundary conditions and applied it to two
examples. We will see other examples in the subsequent lectures.

Keywords: Variational Formulation, Configuration Space, Hamilton’s Principal,

Lagrangian, Hanging Chain, Vibrations of Bars

51
 Vibrations of Structures
Prof. Anirvan DasGupta
Department of Mechanical Engineering
Indian Institute of Technology – Kharagpur

Lecture - 06
Modal Analysis - I

In the previous lectures, we had been looking at Newtonian and the variational approach of
deriving the equation of motion of continuous systems. Also, we looked at the boundary
conditions and initial conditions which close the system ensuring unique solutions to our system.
In the next few lectures, we are going to analyze or solve the equation of motion given boundary
and initial conditions. With this information, we can surely find out the solution which represents
mathematically, the behavior of the system. But then with change in initial conditions, for
example, the solution will change. If the solution changes, qualitative feel of the system is lost.

Slide: 6.1 (Refer Slide Time: 01:16)

Form of solution Boundary Conditions

for modal analysis
Field Variable

General solution with real amplitude function

Now, the question arises whether we can find out some characteristic solutions of the system.
By characteristic solutions, I mean the solutions that will not change with change in initial
conditions or certain properties of the system. We know from the study of vibrations of discrete
systems that we can find such a characteristic solution. Therefore, first we solve for the natural

52
frequencies of the system. Natural frequencies indicate how fast the system is going to vibrate
when free to move after initial disturbance. Secondly, we solve for the corresponding mode of
vibration of the system. These two things constitute the characteristic solutions of the system and
finding them out is known as Modal analysis. Thus, when we perform modal analysis, what we
are doing is actually searching for solutions of very special form.

Consider the equation of motion of a string. Say, we have a fixed-fixed string of length l. We
have the equation of motion and the boundary conditions as shown in Slide 6.1. The field
variable here is w(x, t) which is a general function of the spatial coordinate x and the temporal
coordinate t. When we perform the modal analysis, we are looking for the solutions of very
special form which look like as shown in Slide 6.1 where ω is known as the circular frequency
and W(x) is the amplitude function. So, we are searching for solutions of this very special form in
which the spatial function and the temporal function are separated. One should note that the
proposed solution is in the complex form. In that case, the solution is not strictly separated in
space and time. Such cases will be dealt later in this course. Now for the time being, we
introduce this complex form of solution as the actual solution. Since the equation of motion and
boundary conditions are all real, the actual solution is obtained by taking either the real part or
the imaginary part or a linear combination of the real and imaginary parts of the solution form.

The systems that we are going to study here will mostly have real amplitude function. In that
case, we can rewrite the solution in the form as shown in Slide 6.1. Since we are considering
W(x) to be a real function and as I mentioned that you can take the linear combination of the real
and imaginary parts of the complex solution form as the general solution, then the solution looks
like as shown in Slide 6.1. From the general solution structure, we can deduce few of its
properties. Firstly, the solution is strictly separable in space and time. Secondly, the temporal
part of the solution can become zero at a certain time instant. In that instant, the solution is zero
for all x which implies that the motion of the string passes through the equilibrium point at the
same time instant. Thirdly, there can be points at which W(x) is zero. Such points are called
nodes. The fourth property of this solution structure is that the phase difference between any two
points of the system is either 0 or π. If we take any two points on the domain of the system x1 and
x2 and observe the product of the amplitude function and if it turns out to be positive then the

53
 phase difference is 0 and if it turns out to be negative then the phase difference is π. There is
another interesting property of the solution that is the ratio of amplitudes at x1 and x2 is
independent of time as mentioned in Slide 6.1. We will now start with this form of the solution
and try to find out the characteristic motion of the system.

Slide: 6.2 (Refer Slide Time: 14:07)

Solution structure
Eigenvalue problem
Equation of motion and Boundary
conditions after substituting the
solution structure

Let us begin with an example of the uniform taut string whose equation of motion (EOM) and
the boundary conditions (BCs) are mentioned in Slide 6.2. We substitute the solution structure in
the EOM and BCs and with little rearrangement, we see that the equation and the boundary
conditions reduces to the form as shown in Slide 6.2 since eiωt ≠ 0 for all t. Thus, so obtained 2nd
order linear differential equation in W(x) is very familiar and the solution can be written directly
as given in Slide 6.2. Using the boundary conditions of W(x) at x = 0 and l, we write the matrix
form of coefficients D and H as shown in Slide 6.2. If we want to have non-trivial solutions of D
and H, then the determinant of the matrix must vanish which implies sin (ωl/c) = 0 as shown in
Slide 6.2. This equation is known as the characteristic equation because it yields certain
characteristic property of the system such as circular natural frequencies of the system. We note
that there are infinitely many discrete solutions of the characteristic equation. Thus, there are
countable infinite points at which the characteristic equation will be satisfied for special values

54
 of ω. We use an index n and call these points as ωn where the index n goes from 1 to ∞. These
values of ωn are called the circular natural frequency or characteristic frequencies of the system.

Slide: 6.3 (Refer Slide Time: 21:30)

6.3.1
For any time instant
Fourier sine series 6.3.2

Move up

Move down
Intermediate
configuration

If we substitute the value of ω = ωn into the boundary conditions, we can find the non-trivial
solution of D and H as shown in Slide 6.3 with names Dn and Hn. Thus, anything proportional to
these values will be a solution corresponding to each circular natural frequency. Now, the actual
solution can also be obtained as shown in Slide 6.3. The solutions Wn(x) are called the modes of
vibration. They are in fact the eigen-functions corresponding to the natural frequencies of the
system. Therefore, for every natural frequency, there is an eigen-function which describes the
mode of vibration. Thus, the total solution w(x, t) of the vibration problem now looks like as
shown in Slide 6.3. Here also we must have the index n with coefficients C and S. For each n, we
can have a solution and since our system is linear, a superposition/Summation over n = 1 to ∞ of
this solution is also a solution. Therefore, we sum all these solutions and construct the most
general solution of a vibrating string with fixed support.

One can observe that the obtained solution form looks like a Fourier Sine series as shown in
Slide 6.3. Here, the Fourier sine function is periodic whose periodicity is l which is the length of

55
the string. Now, we know from Fourier series theory that any shape between two supports can be
represented as a Fourier sine series. The constant an in the Fourier sine function is the value of
time part of the solution w(x, t) at any time instant. We also know from the Fourier series theory
that the function forms the basis functions which are orthogonal. It implies that if we take any
two eigen-functions with different index n and m, multiply them and integrate the product from 0
to l, we get l/2 δmn where δmn is Kronecker delta function. The Kronecker delta function is
defined as unity if m = n and 0 if m ≠ n. Thus, sin (πx/l) is orthogonal to sin (2πx/l) etc. Let us
look at graphical representation of this solution.

In order to represent the orthogonality conditions, we draw axes which are orthogonal to each
other and name them as sin (πx/l), sin (2πx/l), sin (3πx/l) etc. It is not possible to draw infinitely
many axes, but one can use their imagination to consider this to be an infinite dimensional space.
Now a solution of the form in 6.3.1 or a representation of the same in the form in 6.3.2 for a
particular shape of the string is actually a point in this infinite dimensional space where a1, a2,
a3… are the coefficient an (projection on each axis) of the Fourier series. Thus we have seen that
the configuration of the string at a particular time instant is nothing but a point in the infinite
dimensional space as shown in Slide 6.3. If the string moves, the point in the infinite-
dimensional space also moves as shown in Slide 6.3. If the string executes a periodic motion in
this space, then it would be a perfectly closed curve. Let us now look at the simplest possible
motion of the point. The simplest motion would be, for example when the string moves only
along the axis sin (πx/l) which means all the coefficients of Fourier sine series except a1 are zero.
The solution corresponding to such a string motion is called a modal solution or in this instant,
referred to as the motion of the string in the first mode of vibration which is nothing but only
sin (πx/l). The first mode of vibration of the string with circular frequency ω1 = πc/l is shown in
Slide 6.3. Similarly, if we consider motion only along the second axis sin (2πx/l), then the motion
of the string with circular frequency ω2 = 2πc/l looks like as shown in Slide 6.3. This infinite
dimensional space is sometimes called the modal space or sometimes also can be called as the
configuration space of the string.

To summarize, we say that the configuration of a string at any time instant is represented by a
point and when it moves from one configuration to the other, it is represented as the motion of

56
this point in the infinite-dimensional space. Also, we have seen that the elemental motion or the
motion along each axis represents the modal solution. Now, I am going to demonstrate these
solutions with a small experiment. Before we see the demonstration, let us make an observation.
As you can see in the solution, there is no node within the domain of the string for first mode of
vibration whereas there is one node within the domain of the string for the second mode of
vibration as shown in Slide 6.3. The node is a point in the domain of the system which remains
fixed or the point which does not move from the equilibrium position while other points are
vibrating. In the first mode of vibration, all the points are vibrating in the same phase as evident
from the intermediate configurations as shown with dashed line in Slide 6.3. In the second mode
of vibration, points in the left half and the points in the right half of the domain is out of phase
which means when one half moves up, other half moves down and vice versa. But two points
within left or right half of the domain are moving in phase. Thus, the phase difference in this
case is either 0 (in phase) or π (out of phase). Now we will see a small experiment for different
modes of string vibration.

Slide: 6.4 (Refer Slide Time: 37:14)

First mode of vibration of string

Exciter

Here, we have a string which is made taut by manual pulling at one end and there is an exciter
which is a simple electric shaver on the other end. Now as I pluck the string you see the motion

57
ceases after sometime. This is expected since all real systems have internal damping but in our
model, we have not considered damping. Therefore, in order to cancel the effect of damping, we
must have an exciter which must pump the same amount of energy that is being dissipated due to
the internal damping in the string. Now, you can see the first mode of vibration of the string
(shown in Slide 6.4). Let us try the second mode of vibration. As the frequency is increased,
second mode of vibration of the string becomes visible (shown in Slide 6.5). We can also try the
third mode which may not be visible very clearly though because the amplitudes of vibration in
third mode will be very small.

Slide: 6.5 (Refer Slide Time: 38:30)

Second mode of vibration of string

We should note here that whenever I pluck a string, I actually excite a number of these modes
which makes the solution more complicated. But by special means, we can excite these
individual modes (A particular frequency excitation) and we have seen that in the demonstration.

The next example that we are going to look at is that of a Uniform hanging chain. We have
derived the equation of motion and boundary conditions of a uniform hanging chain previously
(shown in Slide 6.6). Once again, we are going to attempt a solution with the same solution
structure we used earlier. After substituting the solution structure into the equation of motion and

58
 boundary conditions, we get modified EOM and BCs in W(x) as shown in Slide 6.6. This
problem is known as the eigenvalue problem which is a well known Sturm-Liouville problem in
mathematics.

Slide: 6.6 (Refer Slide Time: 40:30)

EOM and BCs after

substituting the
solution structure

(6.6.1)

(6.6.2)

(6.6.3)

Now, we convert the equation into a more familiar form by using transformation of the
independent variable x. Let us consider a variable s which is a function of x (6.6.1).
Transformation of differential terms is given in (6.6.2) and (6.6.3). Now having done this, if we
substitute these back in the eigenvalue problem and convert the equation in terms of s, then we
obtain the differential equation and the boundary conditions as shown in Slide 6.7. The equation
so obtained is in fact a Bessel differential equation whose standard form is given in Slide 6.7. In
this equation, if we put n as zero then it reduces to the differential equation we just obtained
(6.7.1). The solution of the Bessel differential equation with n = 0, which is the solution of
(6.7.1) is mentioned in Slide 6.7 where J0(s) and Y0(s) are known as zeroth order Bessel function
of the first and second kinds respectively. If we look at the Bessel functions of zeroth order, then
we see that the second kind has a logarithmic singularity at s = 0 as evident from the plot of Y0(s)
given in Slide 6.7 (Y0(s) goes to -∞ at s = 0). The presence of Y0(s) in the solution violates the

59
 BC at the free end. Therefore, in order to maintain the finiteness at the free end, we choose E =
0.

Slide: 6.7 (Refer Slide Time: 47:05)

(6.7.1)

To -∞

Thus, the final solution form is obtained as DJ0(s) as shown in Slide 6.7. Let us look at the
boundary condition now. The BC is obtained as J0 (2ω√l/g) = 0. Now we can see from the
profile of J0 in Slide 6.7 that there are infinitely many discrete points at which the function J0 is
zero. These countable infinite solutions in fact give us the circular natural frequency of the
system. If we represent 2ω√l/g by γk then γ1 is approximately 2.4048, γ2 is 5.5201, and γ3 is
8.6537 and so on. From the values of γk, we can find the circular frequency; say for example ω1
as 1.2024√l/g. One can note here that the 1st natural frequency of the hanging chain is 1.2 times
higher than that of the pendulum.

Finally, one can write the general solution using superposition as shown in Slide 6.8. The modes
of vibration of a uniform hanging string are defined by the eigen-functions which look like as
shown in Slide 6.8. Once again, the demonstration of the modes of vibration of a hanging chain
is shown in Slides 6.9 and 6.10. Exciting the higher modes is little difficult.

60
Slide: 6.8 (Refer Slide Time: 53:58)

Eigen-function
(defining the mode
of vibration)

1st mode 2nd mode 3rd mode

Slide: 6.9 (Refer Slide Time: 56:17)

1st mode of
vibration of
hanging chain

61
Slide: 6.10 (Refer Slide Time: 56:37)

2nd mode of
vibration of
hanging chain

Thus, in this lecture, we have studied the modal analysis of continuous systems. We started with
the modal analysis of a taut string and then we looked at the properties of these modal solutions
and finally we looked at the example of a uniform hanging chain. With this, we come to an end
of this lecture.

Keywords: Modal Space, Modal Analysis, Characteristic Solution, String, Hanging Chain,
Bessel Differential Equation, Configuration Space, Mode of Vibrations, Nodes, Eigenvalue
Problem, Eigenfunctions

62
 Vibrations of Structures
Prof. Anirvan DasGupta
Department of Mechanical Engineering
Indian Institute of Technology – Kharagpur

Lecture - 07
Modal Analysis - II

We continue our discussions on Modal Analysis that we had started in the previous lecture. Let
us consider another example of a system consisting of a bar of varying cross section (Slide 7.1).
Consider that A0 and A0/4 is the area of cross section at the fixed end and at the free end
respectively. The field variable is represented by u(x, t) which represents the axial displacement
at any point x and at any time instant t. We assume that ρ is the density of the material of the bar,
area of cross-section is A(x), Young’s modulus is E and the length of the bar is l. The equation of
motion of axial vibrations of a bar of variable cross-section is shown in Slide 7.1. The boundary
conditions for the system are also shown in Slide 7.1. One should note that at the left end, zero
displacement condition while at the right end, no force condition is applied.

Slide: 7.1 (Refer Slide Time: 00:38)

Boundary Conditions
Equation of motion

Solution form:
---- (7.1.1)

---- (7.1.2)

63
Once again, we assume the solution form as shown in Slide 7.1. If we introduce the solution
form in the EOM and the BCs, they take the form (7.1.1) where c2 = E/ρ. Thus, we have the
eigenvalue problem for the axial vibration of bar of varying cross-section which needs to be
solved for circular characteristic frequencies (ω) and the corresponding modes of vibration
(eigenfunctions). We note that (7.1.1) may not be solvable analytically for a general variation of
area A. Thus, we try to find a class of systems or class of variation of cross sectional area for
which (7.1.1) might be solvable analytically.

For this purpose, we make variable transformation. Let us consider a new variable W(x) which is
expressed as a product of some unknown function h(x) and the amplitude function U(x). If we
differentiate U(x) = W(x)/h(x) and slightly rearrange, we get Eq. 7.1.2. If we identify h2 as A(x),
Eq. 7.1.1 modifies as Eq. 7.2.1 as shown in Slide 7.2. The eigenvalue problem along with BCs
can now be written in terms of new variable W using substitution hʹʹ/h=α as shown in Slide 7.2.

Slide: 7.2 (Refer Slide Time: 09:20)

---- (7.2.1)

-- (7.2.2)

Now, we look at a class of systems for which the function hʹʹ/h is a constant α which can be
positive or negative. This class of systems is characterized by variation of h which is hyperbolic
for α > 0, harmonic for α < 0 and quadratic for α = 0. Let us consider a particular case that is
shown in the figure in Slide 7.1. Here the radius is reducing linearly and the area goes from A0 to

64
 A0/4. Therefore, the variation of the cross-sectional area here may be expressed as Eq. 7.2.2 in
Slide 7.2. For such a special form of A(x), one can find that α = 0 which will further simplifies
the eigenvalue problem whose general solution can be written as shown in Slide 7.2. For the
boundary condition at the left end, W(0) = 0 would imply H = 0. The boundary condition at the
free end (x = l) along with the condition H = 0 gives us the characteristic equation of the system
(with the form of variation of cross-sectional area as A(x) in Eq. 7.2.2) as shown in Slide 7.2. It
can be seen that the characteristic equation is a transcendental equation which has to be solved
numerically.

Slide: 7.3 (Refer Slide Time: 20:48)

Characteristic Equation
ω1 ω2
tan(ωl/c)
Solution of Characteristic
equation
-45° line corresponding to –ωl/c

Eigen functions in
terms of W(x)

Eigen functions in 1st Eigen functions/1st

terms of U(x) mode of vibration

2nd Eigen functions/2nd

mode of vibration
Node

A good way to visualize the solution of this transcendental equation is to make a graphical plot
of tan(ωl/c) vs ωl/c as shown in Slide 7.3. A 45° line is plotted corresponding to -ωl/c as shown
in the figure in Slide 7.3. Intersection of tan(ωl/c) with the 45° line gives the solutions of
transcendental equation which are the eigenvalues ω1, ω2 … ωn as shown in Slide 7.3. The
numerical value of each solution is also shown in the Slide 7.3. Here, we should note that there
will be countably infinite number of intersections. Once we have found the eigenvalues or the
circular natural frequencies of the system, we can find out the corresponding Eigen functions
which describe the modes of vibration of the system as shown in Slide 7.3. It should be noted

65
 here that the Eigen functions are in terms of the new variable W which we can write in terms of
our original variable U. These eigenfunctions (1st and 2nd mode only) are drawn approximately in
the Slide 7.3 where the amplitude function represents the axial displacement of the bar. Here, we
note that there is one node in the 2nd mode and no nodes in the fundamental or the 1st Eigen
function. As we have discussed earlier, the node is the point on the bar, which remains stationary
at all times.

Slide: 7.4 (Refer Slide Time: 31:27)

EOM for bar under

axial vibrations
Equation of motion for
discrete mass-spring system
BCs:
Form of solution

Next let us consider a continuous system which is interacting with a discrete system as shown in
Slide 7.4. For an example, we consider a uniform bar fixed at one end and attached to a simple
harmonic oscillator at the other end. Here, we have a discrete mass M and a spring of stiffness K
which is attached to a bar of length l undergoing axial vibrations. These kinds of systems are
quite common when we have to put absorbers for example, on a vibrating continuous system or a
vibrating structure. We will call such systems as hybrid systems because we have both
continuous as well as discrete elements in such systems. Since we have two different kinds of
systems interacting with each other, we will have two equations of motion. For the bar, the
equation of motion can be written directly as shown in Slide 7.4. For the oscillator, the equation
of motion is written in Slide 7.4 where y(t) measures the displacement of the mass M from its

66
equilibrium position. Here, we have two dependent variables; one is the field variable u(x, t) and
the other is y(t). The boundary conditions for the system are also shown in the Slide 7.4. Here,
we have a dynamic boundary condition at x = l. We now represent both the dependent variables
as a vector and search for solutions of the form as written in Slide 7.4. It may be mentioned here
that the vector represents the configuration of the system in a dimension which is infinity plus
one. The dimension is infinity because of the bar and additional one because of the discrete
system attached. Hence, the modal space is of dimension infinity plus one. Now, if we substitute
the assumed solution structure in the equations of motion and the BCs, we obtain the eigenvalue
problem and the boundary conditions as shown in Slide 7.4.

Slide: 7.5 (Refer Slide Time: 43:41)

General solution

The solution of the differential equation in the eigenvalue problem is shown in Slide 7.5. The
boundary condition at x = 0 gives C = 0 which simplifies the solution U(x) as shown in Slide
7.5. Substituting the simplifies form of the solution U(x) in the second boundary condition gives
the characteristic equation of our system as shown in Slide 7.5. Now, once again, this is the
transcendental equation which has to be solved numerically for the eigenvalues ω. We can note
that countably infinite number of solutions exist for this transcendental equation. Corresponding
to each solution or kth circular natural frequency, we have an Eigen functions Uk(x). For each of

67
these eigenfunctions, we can now find the amplitude Yk of the discrete mass as shown in Slide
7.5. Therefore, the general solution may be represented by superposing all the solutions as shown
in Slide 7.5. Once again, we note that the motion of the system is taking place in a modal or
configuration space which is of dimension infinity plus one. Now, we can have two special cases
which fall immediately from the analysis that we have performed.

Slide: 7.6 (Refer Slide Time: 51:23)

The first case is when the stiffness of the spring connecting the bar and the discrete mass tends to
infinity which means that the discrete mass is rigidly attached to the bar. We immediately follow
from the characteristic equation by taking K → ∞, the characteristic equation simplifies to the
form as shown in Slide 7.6. From here, we can find out the circular natural frequencies and the
corresponding eigenfunctions from the characteristic equation. Here, we observe that coordinate
of the discrete mass y(t) becomes equal to that of the bar u(l, t). The second case is when the
mass becomes infinity. In this case, the characteristic equation simplifies to the form as shown in
Slide 7.6. This case is equivalent to the situation where the bar is connected to a spring which is
attached to a rigid wall (diagram shown in Slide 7.6). In this case, the motion of the mass y(t)
vanishes. We should note here that the boundary condition in this problem is dependent on the
circular natural frequency or the eigenvalue itself as shown in Slide 7.4.

68
To summarize, we have discussed in this lecture two more examples of modal analysis where we
solved the eigenvalue problems. We have considered a bar with varying cross section and we
have solved a class of problems for which we have obtained analytical solutions which we will
compare against solutions obtained by other methods later in this course. The other thing that we
have discussed here is a continuous system interacting with the discrete vibrating system. So, we
will continue this discussion further in the next lecture.

Keywords: Modal Analysis, Vibrations of Bars, Bars with varying cross-section,

Eigenvalue Problem, Eigenfunctions, Interacting Discrete Vibrating System

69
 Vibrations of Structures
Prof. Anirvan DasGupta
Department of Mechanical Engineering
Indian Institute of Technology – Kharagpur

Lecture - 08
Properties of the Eigenvalue Problem

In the last two lectures, we started discussions on the modal analysis of continuous systems. The
modal analysis essentially means solving an eigenvalue problem. In this lecture, we are going to
look at some properties of a eigenvalue problem that comes while we perform the modal analysis
of continuous systems. Let us start by revisiting the modal analysis problem.

Slide: 8.1 (Refer Slide Time: 01:09)

--- Equation of motion

--- Boundary Conditions

--- Form of solution

Temporal part of the solution

In the last lecture, we have discussed the problem of bar with varying cross section. The equation
of motion of the system and the relevant boundary conditions for the problem is mentioned in
Slide 8.1. In order to do the modal analysis, we were looking for the solutions of a special form
or structure as shown in Slide 8.1. We have seen that in the solution form, field variable is
expressed as a product of an amplitude function, which is a function of x and a harmonically

70
varying time function. We then discussed the properties of this solution and we found that the
solution is actually separable in space and time. If we write the actual solution in its real form, it
appears in the form as shown in Slide 8.1. All points therefore vibrate at the same circular
frequency ω. We also observed that all points of the system pass through the equilibrium point at
the same time instant. The time instant when the temporal part of the solution is 0, the whole
solution is 0, which implies that the bar is in its equilibrium state. Therefore, it can be concluded
that all points will pass through the equilibrium point at the same time. Further, we observed that
the phase difference between any two points on the bar is either 0 or π. Finally, we observed the
existence of nodes, the points at which the amplitude function U(x) is 0. Now, if we substitute a
solution of this structure into the equation of motion, we obtain a differential equation in terms of
the amplitude function U(x) and the corresponding boundary conditions as shown in Slide 8.1.
This forms the Eigenvalue problem for the system as shown in Slide 8.1.

Slide: 8.2 (Refer Slide Time: 07:56)

Let us now represent the above obtained eigenvalue problem in a slightly abstract as shown in
Slide 8.2. Here, λ = ω2 and K[.] is a differential operator. In the case of the tapered bar, μ(x) =

71
ρA(x) and the differential operator K, which is also known as the stiffness operator is K[.] = -
[EA(x)(.)ʹ]ʹ. Thus, we have the structure of the differential equation of our Eigenvalue problem as
shown in Slide 8.2. As I mentioned here that K[.] is known as the stiffness operator because this
term comes from the potential energy in the Lagrangian formulation, while the term μ(x) is the
kinetic energy operator because it comes from the kinetic energy in the Lagrangian formulation.

Slide: 8.3 (Refer Slide Time: 12:19)

-- -- (8.3.1)

---- (8.3.2)

We now write the above differential equation for two modes j and k as shown in Slide 8.3. The
objective of this analysis is to determine certain properties of the Eigenvalue problem. We
multiply the first equation with Wk and the second equation with Wj; subtract one from the other
and integrate it over the domain as explained in Slide 8.3. If the property in Eq. 8.3.2 where
W(x) and W(x) with tilde are the functions that satisfy the boundary conditions of the problem
holds true for the stiffness operator, then the operator K is known as self adjoint operator.

The self adjointness of an operator is connected to symmetry since we know that the stiffness
operator has a corresponding matrix. For example, in vibrations of discrete systems, we have
come across stiffness matrix. Self adjointness of the stiffness operator is nothing but the
symmetry of the corresponding stiffness matrix. Now the question arises, what are the

72
 consequences of this symmetry? As we know, if the matrices are symmetric, the eigenvalues and
the eigenfunctions are real, and the Eigen vectors are orthogonal. In a similar manner, we can
show that the eigenvalues and eigenfunctions are real whenever the stiffness operator is self
adjoint. Also, the eigenfunctions are orthogonal with respect to an inner product that we will find
out in the subsequent part of this lecture.

Slide: 8.4 (Refer Slide Time: 20:33)

---- (8.4.1)

Inner product of
eigenfunctions
Normalization

We will now discuss the orthogonality property. If the stiffness operator K is self adjoint, then
the second term in Eq. 8.3.1 vanishes as shown in Slide 8.4 and it leads to Eq. 8.4.1. If we take
two distinct amplitude functions/eigenfunctions, Wj and Wk, then the integral in Eq. 8.4.1 which
we define as the inner product of these two eigenfunctions, must vanish. In a compact form, we
will write the inner product of two eigenfunctions as shown in Slide 8.4. Now, one may
normalize this property by appropriately scaling the eigenfunctions because we know that any
scaled form of the eigenfunction is also an eigenfunction. Thus, we can scale these appropriately
to have orthonormality of the eigenfunctions with respect to the inner product that we have
defined. In Eq. 8.4.1, we have orthogonality with respect to the inertia operator. If we consider
that μ(x) represents the inertia operator, then the orthogonality is with respect to the inertia
operator. If we write the eigenvalue problem for the jth mode, multiply it with Wk and integrate,

73
 we see that the eigenfunctions are orthogonal with respect to the stiffness operator K also. Now,
we need to understand the physical implication of the orthogonality of eigenfunctions with
respect to inertia and stiffness operators. The physical implication is that there is no exchange of
kinetic or potential energy between the Eigen modes. This orthogonality property is very useful
for solving initial value problems or other problems related to continuous systems as we will see
in the due course.

Slide: 8.5 (Refer Slide Time: 29:08)

Stiffness Operator

(8.5.1) ----
-

Using BCs

Self-adjointness of
stiffness operator

Let us now determine the orthogonality relations for few example systems. Once again, we take
a bar with varying cross section whose eigenvalue problem is found in Slide 8.1 and written
again in Slide 8.5. Now, we will check whether the stiffness operator in the eigenvalue problem
is really self adjoint. For this, we need to show 8.5.1 where Uk and Uj are two eigenfunctions of
the eigenvalue problem. If we integrate the LHS of 8.5.1 by parts, and use the boundary
conditions, we recover the RHS of 8.5.1. We should note here that the eigenfunctions Uk and Uj
satisfy the boundary conditions of the eigenvalue problem also. Thus, we have shown the self
adjointness of the stiffness operator of a tapered bar. We can also write this orthogonality in
terms of the inner product as we have defined for the tapered bar as shown in Slide 8.6.

74
 Slide: 8.6 (Refer Slide Time: 36:36)

Orthogonality in terms
of the inner product

Eigenvalue problem

Eigenfunction

J1=

Next, we consider the example of a hanging string or chain whose eigenvalue problem is shown
in Slide 8.6. The stiffness operator in this case is given by the term K[W] as shown in Slide 8.6.
We can once again check that the self adjointness property holds for the stiffness operator of the
hanging chain also. Also, using this property, we can derive the inner product of the
eigenfunctions of the hanging chain with respect to which the eigenfunctions are orthogonal as
shown in Slide 8.6. The form of eigenfunction for the hanging chain that we have already
derived in a previous lecture is shown in Slide 8.6. The eigenfunctions satisfy the orthogonality
relation which is shown as a inner product of Wj and Wk in Slide 8.6 where αj is also defined.

Now, we consider the example of a uniform bar which is coupled to a harmonic oscillator. The
eigenvalue problem for this system is mentioned in Slide 8.7. We now write the differential
EOM for jth and kth mode and multiply each with the other eigenfunction as shown in Slide 8.7.
If we subtract them, integrate it over the domain of the bar (8.7.1) and rearrange it, we obtain the
form as in 8.8.1. Here, we have used the boundary conditions give in Slide 8.7 to simplify. Now,
if j ≠ k and ωj ≠ ωk which implies that there are no repeated Eigen frequencies, the bracketed
quantity in 8.8.1 must vanish. The bracketed term can also be written as 8.8.2 using the

75
eigenvalue problem form given in Slide 8.7. Thus, the inner product for the hybrid system is now
defined in the form as shown in Slide 8.8.

Slide: 8.7 (Refer Slide Time: 42:36)

--- (8.7.1)

Let us now summarize what we have studied today (Slide 8.9). We revisited the modal analysis
and the eigenvalue problem. Then we looked at the properties of the modal solution. Then we
discussed about self-adjoint operators and the consequence of the stiffness operators being self-
adjoint. Then, we discussed about the Orthogonality property of eigenfunctions. We have
determined the inner product. We outlined the steps to determine the inner product with respect
to which this orthogonality property holds. And also we have looked at the implications of the
orthogonality property of eigenfunctions. Thus, if the eigenfunctions are orthogonal, it implies
that there is no exchange of energy, kinetic or potential between the eigenmodes or
eigenfunctions. With that, we conclude this lecture.

76
Slide: 8.8 (Refer Slide Time: 47:46)

--- (8.8.1)

--- (8.8.2)

Slide: 8.9 (Refer Slide Time: 52:42)

Keywords: Eigenvalue Problem, Modal Analysis, Orthogonality, Eigenfunctions,

Normalization, Stiffness Operator, Self-adjoint Operator, Inner Product

77
 Vibration of Structures
Prof. Anirvan DasGupta
Department of Mechanical Engineering
Indian Institute of Technology – Kharagpur

Lecture – 09
Modal Analysis: Approximate Methods – I

Since last three lectures, we have been discussing about the modal analysis of continuous
systems. We have seen that solving the problem of modal analysis is nothing but solving an
eigenvalue problem analytically. While doing so, we solved for natural frequencies and the
modes of vibration characterized by the eigenfunctions. As a matter of fact, performing an
analytical solution is always preferable because we can easily find out the effects of various
parameters of the system on the modes of vibration and the modal frequency. However, we have
seen that even in very simple systems, the solution of the modal analysis problem requires
solving transcendental equations which might be computationally intensive. Thus, it is of interest
to know if numerical/approximate methods of modal analysis are possible. So in this and
subsequent lecture, we are going to look at few techniques for solving the model analysis
problem or the eigenvalue problem numerically.

Slide 9.1: (Refer Slide Time: 02:20)

To sum up, the motivation for studying the approximate solutions are as shown in Slide 9.1. The
first motivation is that the analytical solutions may be cumbersome. The other motivation is that

78
an approximate method can provide a quick solution to the modal analysis problem which may
be sufficiently accurate for our purposes.
Slide 9.2: (Refer Slide Time: 02:56)

Let us now look at the methods that are available to us for approximate modal analysis. In this
course, we are going to discuss two methods as mentioned in Slide 9.2. One of which is the
energy based methods which will be the topic of discussion in this lecture. The projection
methods will be covered in next lecture. Now, as the name suggests, the Energy based methods
will use the kinetic and potential energy of the system to determine the modes of vibration or the
natural frequencies while the Projection methods will use the governing equation of motion
directly to solve the modal analysis problem.

Slide 9.3: (Refer Slide Time: 04:33)

79
The first energy based method that we are going to look at is the Rayleigh method. The
Rayleigh method is used to determine the fundamental frequency of a continuous conservative
system (Refer Slide 9.3). Let us see, how does this method work?

Slide 9.4: (Refer Slide Time: 05:48)

Let us understand this with the example of a bar (Refer Slide 9.4 and 9.5). Consider a tapered bar
undergoing axial vibration. We write the kinetic (T) and potential (V) energy as shown in Slide
9.5. We know that the bar in axial vibration is a conservative system which means that the total
energy of the bar is a constant. So the total mechanical energy (Ɛ = T + V) is constant.

Slide 9.5: (Refer Slide Time: 05:53)

80
Let us now suppose that the system is vibrating in one of its modes. When a system is vibrating
in one of its modes, the field variable can be written as a separable function of space and time in
the form as shown in Slide 9.5. Substituting the expression of field variable u(x, t) in the total
energy of the bar (Ɛ), we get the expression as shown in Slide 9.6.

Slide 9.6: (Refer Slide Time: 10:01)

Now, if the total mechanical energy (Ɛ) is to be a constant then it would require is that it should
be independent of time and it is possible only when the coefficient of sin2ωt and cos2ωt are
equal. Therefore, we obtain the expression for ω2 as a ration which is defined as the Rayleigh
quotient (R[U(x)]) as shown in Slide 9.7. The Rayleigh quotient is the key concept in Rayleigh’s
method. Sometimes, the Rayleigh quotient is also expressed as maximum potential energy
divided by the maximum kinetic energy. So the maximum potential energy would be the
amplitude of the cos2ωt term whereas the maximum kinetic energy would be the coefficient of
the sin2ωt term. Now, if we know the exact eigenfunction, and if we substitute it in the Rayleigh
quotient, then what we get is the exact circular eigenfrequency corresponding to that mode. Now
the question comes, how do we use the Rayleigh quotient when we do not know the exact
eigenfunction? So, what is usually done is that we try to minimize the Rayleigh quotient to
obtain the fundamental frequency. Thus, the square of the fundamental natural frequency is
obtained by minimizing the Rayleigh quotient over a space of possible eigenfunctions U(x).

81
Slide 9.7: (Refer Slide Time: 16:09)

We have to follow certain rules while choosing a proper eigenfunction U(x) in the minimization
problem? The condition is U(x) must be a member of the set of what are known as Admissible
functions (Refer Slide 9.8 and 9.9). Now what are Admissible functions? These are the functions
which satisfy the following two properties. The first property is that it is differentiable at least up
to the highest order of spatial derivative present in the energy expression. In the example that we
are considering, the highest order of the space derivative is one. Thus, the set of admissible
function should be differentiable up to first order. The second property is that it should satisfy all
the geometric or essential boundary conditions of the problem.

Slide 9.8: (Refer Slide Time: 21:26)

82
Slide 9.9: (Refer Slide Time: 21:38)

The functions that satisfy these two properties are known as Admissible functions. Such
functions can be constructed using polynomials, trigonometric functions and other such
elementary functions. Let us now look at the example of the tapered bar. As one can see, the bar
is fixed at x = 0 and it is free at x = l. Thus, the geometric/essential boundary condition is on the
left boundary where the bar is fixed. So, we must choose admissible functions which satisfy the
boundary condition at the left end. Let us look at such functions.

Slide 9.10: (Refer Slide Time: 26:26)

83
Slide 9.11: (Refer Slide Time: 27:17)

α≠1
α=1

Fundamental natural frequency with α=1

Fundamental natural frequency with α=0.93

Fundamental natural frequency from analytical treatment

We have to choose admissible functions for this problem such that it is zero and x = 0. We
choose such a function U(x) as shown in Slide 9.11. We can have a class of these functions by
raising it to a power α. The profile U(x) versus x/l for different values of α is shown in Slide
9.11. Initially, we will keep α arbitrary. We substitute the assumed form of admissible function
in the Rayleigh quotient which turns out to be a function of α as shown in Slide 9.11. It is to be
noted here that we have used here a class of admissible functions and by adjusting α; we can
minimize the Rayleigh quotient. Thus, the parameter α provides us a handle to solve a
minimization problem as we have formulated it in Slide 9.8.

In Slide 9.11, the term E/ρl in the expression of R[α] has the geometric as well as material
properties of the bar while the remaining part is a function of. We can either put various values
of α or can minimize the Rayleigh quotient with respect to α and determine α. So let us see what
happens if α = 1. We get R[1] = ω2 = 70E/16ρl. Thus, we get an estimate of fundamental circular
frequency as shown in Slide 9.11. Let us see what happens if we minimize R with respect to α.
We get α=0.93 and the corresponding circular frequency found is closer to the exact one (within
3%). Thus, α=0.93 gives a better estimate of the circular natural frequency as shown in Slide
9.11. Thus, we have fairly estimated the fundamental frequency of a tapered bar using a very

84
simple method based on the structure of admissible function that we have chosen. It is to be
noted that at α<1 (True for best estimate), the stress at x = 0 will be infinite. It will give some
unrealistic estimates of stress in the bar. However, the frequency estimate is fairly accurate.
Now, here using Rayleigh method we have estimated the fundamental frequency. Now can we
now go on to find out the higher frequencies?

Slide 9.12: (Refer Slide Time: 37:49)

--- 9.12.1

--- 9.12.2

It is possible using what is known as the Rayleigh-Ritz method. Here, we have Ritz expansion
in addition to what we had with Rayleigh method. Here, we make use of the Ritz expansion of
the amplitude function U(x) in the Rayleigh quotient as explained in Slide 9.12. Let us consider
the Rayleigh quotient for the problem of the tapered bar. As usual, we choose amplitude function
U from the set of admissible functions and minimize the Rayleigh quotient. We now expand the
amplitude function U (with a tilde symbol to distinguish the function from the basis function
Ui(x)). In the Rayleigh’s method, we had kept an unknown parameter α. Here we will expand the
amplitude function in terms of admissible basis functions and a certain unknown coefficient α.
Thus, the amplitude function is a linear combination of any number of admissible functions and
can be written as the dot product of column vector α and U as shown in 9.12.1. We substitute the

85
expansion in the Rayleigh quotient; we get the form as shown in 9.12.2. It is to be noted here that
the vector α is unknown. So we have to minimize the Rayleigh quotient with respect to vector α.

Slide 9.13: (Refer Slide Time: 44:10)

For extremization, the derivative of the Rayleigh quotient which is a function of vector α must
vanish. The derivative has to be taken with respect to each element of the vector α. So, if there
are N elements in the vector α, then there will be N equations in N unknowns. The derivative,
finally gives a discrete eigenvalue problem which can be solved very easily to determine ω and
corresponding values of the eigenvector α (Refer Slide 9.13) which can further be used to
determine the corresponding eigenfunctions using basis function vector U. Thus, using the
Rayleigh-Ritz method, we can find out not only the fundamental but higher modes of vibration.
The accuracy of various modes found through this method will be different. So as a thumb rule,
if we want N modes accurately, we must take 2N terms in the expansion. This is rough estimate
of number of terms in the expansion which may or may not work always but this is a good way
to start.

Now there exists another method which is quite powerful in this class of methods which is
known as the Ritz method. In the Ritz method, we use the idea of Ritz expansion of the field

86
 variable and substitute this expansion directly in the variational formulation. Let us see for the
tapered bar once again. The variational formulation for tapered bar is shown in Slide 9.14.

Slide 9.14: (Refer Slide Time: 49:18)

Lagrangian of a
discretized
system obtained
after spatial
integration

---- 9.14.1

Once again, we use the expansion of field variable u(x, t) in terms of admissible function Hk(x) as
shown in Slide 9.14. Substituting the expansion in the variational form and simplifying it, we
obtain the form as shown in Slide 9.14 where the metrics M and K are given by the expressions
shown in Slide 9.14. The equation of motion obtained from the Lagrangian of the discretized
system is shown in Slide 9.14. Thus, in the Ritz method, we have essentially discretized the
problem. Now once we have discretized, we can search for solutions as we do for discrete
systems. We search for modal solutions of the form shown in 9.14.1 and we solve the eigenvalue
problem. After this, the things are very standard.

Let us look at the axial vibrations of the tapered bar once again as shown in Slide 9.15. Here the
form of the admissible function that we have chosen is Hj(x) where j = 1, 2 as shown in Slide
9.15. So, we will discretize using these two functions. The discretized equation of motion is
shown in slide 9.15. Once we have the discretized equations of motion with us, the standard
procedure follows which means that we assume a solution structure as shown in Slide 9.16.

87
Then, we come to the eigenvalue problem and finally the characteristic equation. Now, if we
solve the characteristic equation shown in Slide 9.16, we will obtain the eigen frequencies of the
system as shown in Slide 9.17.

Slide 9.15: (Refer Slide Time: 55:15)

Slide 9.16: (Refer Slide Time: 56:08)

88
Slide 9.17: (Refer Slide Time: 56:36)

Thus, we obtain the circular eigen-frequencies as ω1R, ω2R calculated using the Ritz. These
frequencies are compared with the exact circular eigen-frequencies. As we can see in Slide 9.17,
fundamental circular frequency matches quite well with the exact while there is some error in the
second circular natural frequency. We should note here that the Ritz and even Rayleigh method
has an upper-bound property, which means that the natural frequency calculated from these
approximate methods is always greater than the exact one. Or, we can say that the actual natural
frequency of the system is lower than what we calculate using these approximate methods (Slide
9.18).

Slide 9.18: (Refer Slide Time: 57:34)

89
Slide 9.19: (Refer Slide Time: 58:15)

In Slide 919, you can see the eigen-vectors K1 and K2 and the corresponding eigen-functions that
have been determined by using the Ritz method. We plot the eigenfunctions U1 and U2 in Slide
9.20.

Slide 9.20: (Refer Slide Time: 58:43)

The plots show the comparison of the eigenfunctions calculated by the Ritz method and those
obtained from the exact solution that we had discussed previously. Once again, we see that the
eigen-function corresponding to fundamental eigen-frequency matches quite well with the exact

90
while that of the second mode, there is an error especially at x/l = 1. Since, we have considered
only those admissible functions which satisfy the geometric boundary condition which is at x = 0
while the natural boundary condition at x = l is not satisfied. We have to take more and more
terms in the admissible function in order to satisfy B.C. at both ends and then, convergence to the
exact solution can be seen.

Slide 9.21: (Refer Slide Time: 59:37)

To summarize the lecture (Slide 9.21), we have considered approximate modal analysis based on
energy methods which uses admissible functions. We have looked at three methods Rayleigh
quotient, Rayleigh-Ritz method and Ritz method. We have seen that these methods have an
Upper-bound property of the eigenvalue estimate. These methods work for conservative systems
with potential forces. With that, we conclude this lecture.

Keywords: Modal Analysis, Admissible Functions, Rayleigh Quotient, Rayleigh-Ritz, Ritz,

Conservative Systems, Eigenvalue Problem, Axial Vibrations, Tapered Bar, Fundamental
Natural Frequency

91
 Vibration of Structures
Prof. Anirvan DasGupta
Department of Mechanical Engineering
Indian Institute of Technology – Kharagpur

Lecture – 10
Modal Analysis: Approximate Methods - II

In the previous lecture, we discussed about the approximate methods of modal analysis which
are one class of methods based on the kinetic and potential energy, the Lagrangian etc. Those
methods are broadly classified as energy based methods for modal analysis. In today’s lecture,
we are going to look at another class of methods which are known as Projection methods.

Slide 10.1: (Refer Slide Time: 00:55)

We have already discussed the motivation for studying the approximate methods (Refer Slide
10.1) in the previous lecture. The analytical methods are more preferable but are quite
cumbersome while the approximate methods can give quick and sufficiently accurate results.
Therefore, we are going to look at these projection methods which work directly with the
governing differential equation of the system. The projection method can be used very easily for
dealing with non-conservative and non-potential forces (Refer Slide 10.2). It is so because this
method uses governing differential equation directly while in the Lagrangian, it is little difficult
to introduce the terms corresponding to non-conservative forces.

92
Slide 10.2: (Refer Slide Time: 01:33)

Slide 10.3: (Refer Slide Time: 02:34)

Let us discuss the projection method in the context of the equation of motion show in Slide 10.3
where μ(x) represents the inertia operator and K[u] is another linear differential operator. So
here, we have a continuous system which is described by a given partial differential equation and
our aim is to discretize the equation of motion. Here, we use the idea of expansion of the field
variable u(x, t) in the form shown in Slide 10.3. One thing that may be mentioned here is that,
even though the expansion of field variable looks like a separable solution but it is not. Had it

93
been only one term then it would be the solution that is separated in space and time. But once
we take the expansion, it is no longer a separable solution. The expansion may be written in
terms of vector multiplication as shown in Slide 10.3. In the projection methods, there is a
restriction on the kinds of functions using which we do the expansion. These functions are
known as comparison functions. These are functions that satisfy two important properties.
Firstly, they must be differentiable at least up to the highest order of space derivative present in
the equation of motion. Secondly, they should satisfy all the boundary conditions of the problem.
It is very important to note here that these functions must satisfy all the boundary conditions of
the problem and this in fact makes this method a little more difficult to apply compared to the
energy based technique where we used admissible functions.

Thus, we expand the field variable in terms of comparison functions Pk(x) and unknown
functions of time pk(t). We now substitute the expansion in the equation of motion. Here, I do not
expect that the assumed form of the solution will satisfy the equation of motion since it is already
an approximate solution. Therefore, what we generate is known as Residue. We do not expect
that the residue e(x, t) will be zero throughout the domain (Refer Slide 10.3). Now, we project
the residue in a certain space. This can be thought of as expanding the solution as a linear
combination of certain functions as shown graphically in Slide 10.3. The point shown in the plot
represents the configuration of the system and as the temporal functions p1(t), p2(t),
p3(t)...change, the point moves in the space. Now, we try to make the residue zero at certain
points. There are various ways of doing it. So, here we first introduce the idea of Projection. We
have suitably defined an inner product of two functions where one is the residue that we have
generated and another is the function H(x) which we can simply define in this case and say that
instead of the residue vanishing identically, we say that the projection of the residue on certain
functions Hj(x) is zero as shown in Slide 10.4. Thus, what we say is instead of having the residue
identically zero, we have a weaker condition which says that the projection of the residue along
certain function directions is zero. We can take N number of functions Hj(x) suitably chosen to
generate N equations. And then, we can attempt to solve for the N unknowns pk(t) that we have
in the expansion in Slide 10.3. Now the choice of the functions Hj(x) decides the method. Now,
what are the different ways of choosing these functions on which we project? The simplest
choice is the Dirac delta function.

94
Slide 10.4: (Refer Slide Time: 11:28)

Once we choose the functions Hj(x) as the Dirac delta functions, the method is known as the
Collocation method. Now, what does it mean to choose Hj(x) as Dirac delta function? If we
substitute the Dirac delta functions in the equation of inner product in Slide 10.4, then we obtain
that the residue is zero at certain points over the domain. For example, for the bar if we choose xj
as shown in the figure e(x) versus x/l in Slide 10.4, the residue must vanish at these points. These
points are known as the Precision points or Accuracy points. There can be various ways of
choosing the precision or accuracy points. They can be uniformly distributed or there can be
other methods of choosing precision points. A good way of choosing precision points is given by
Chebyshev method and they are known as Chebyshev accuracy points.

Now, let us first look at what happens when we substitute the residue in the projection equation.
The residue is rewritten in Slide 10.5 and if we consider the projection of the residue over the
functions of the form of Dirac delta functions, we arrive at the discretized equation of motion
where the matrix elements Mij and Kij are obtained like as shown in Slide 10.5. Thus, we have
seen that when we use collocation method, we obtain the discretized system where the matrix

95
elements are given as shown in Slide 10.5. So let us once again look at the example of the
tapered bar.

Slide 10.5: (Refer Slide Time: 18:51)

We have a fixed-free bar once again and we have the geometric boundary condition at the left
end and the natural or dynamic boundary condition at the right end as shown in Slide 10.6. Now,
we have to choose comparison functions which must all the boundary conditions of the problem
which means they must satisfy the geometric as well as the natural boundary condition. So let us
look at a particular choice as shown in Slide 10.7. One can very easily check that Pj(x) satisfies
the geometric C at the left end as well as the natural boundary condition at the right end. Now,
we choose the accuracy points as we have discussed, as shown in Slide 10.8. We chose them as
uniformly distributed. Thus, if we divide the domain of the bar in n parts then you can take the
uniformly distributed accuracy points or you can also have the Chebyshev accuracy points which
are determined by the expression shown in Slide 10.8. The Chebyshev accuracy points have a
nice geometric visualization as shown in Slide 10.8. In the domain of the bar, if we draw a
semicircle with the length of the domain as the diameter and we fit a regular polygon in it, the
projection of the vertices of the polygon on the domain represents the accuracy points. For
example, if one wants to take three accuracy points, he has to inscribe half hexagon and the

96
projection of the corner points of the hexagon on the domain will give him the accuracy points.
These are the Chebyshev accuracy points.

Slide 10.6: (Refer Slide Time: 21:28)

Slide 10.7: (Refer Slide Time: 22:11)

Slide 10.8: (Refer Slide Time: 22:35)

97
We should note here that the Chebyshev accuracy points will never fall on the ends of the bar.
We will use each one of them, the uniform spacing and the Chebyshev spacing and calculate the
eigenfrequencies.

Slide 10.9: (Refer Slide Time: 28:12)

With uniform and Chebyshev spacing, we obtain eigenfrequencies as shown in Slide 10.9. The
exact ones are also shown in Slide 10.9 for comparison. We should note that the fundamental
circular natural frequency is in some error from the exact circular natural frequency while the
second natural frequency is quite comparable with that obtained from both the methods. In the
collocation method, there lies a disadvantage that it does not have the upper-bound property as
we saw in the energy based methods. If we calculate the eigenfunctions corresponding to these
eigenfrequencies, then what we obtain is shown in the Slide 10.10.

If we look at the fundamental eigenfunctions, the eigenfunction corresponding to the

fundamental mode, the plot is indistinguishable from that obtained from collocation method
(Slide 10.10). Though the natural frequency is in some error, but it can, of course be rectified by
considering more and more terms in the expansion and making it more accurate.

98
Slide 10.10: (Refer Slide Time: 33:18)

On the other hand, if we look at the eigenfunction of the second mode, we see that with the
Chebyshev spacing it appears to be more accurate because the location of the node matches quite
well with that obtained from the exact solution. While that of the uniform spacing, there is some
error. Thus, so far, we have looked at one choice of the projection functions H(x). Now, we will
look at another choice which gives us what is known as the Galerkin method which is another
powerful method making use of the projection technique.

Slide 10.11: (Refer Slide Time: 35:00)

99
Let us recall what we did earlier. We expanded the field variable the way as shown in Slide
10.11 where Pk(x) is the comparison function. After that, we generated the residue and we
projected the residue on certain functions and put it to zero. Now, in the Galerkin method, we
take these projection functions same as the comparison functions, i.e., Pj(x) = Hj(x). Thus, the
projection gives us N equation from which we are going to solve for N temporal function p(t).
Thus, when we choose the projection functions same as the comparison functions in expansion
then we have the Galerkin method and when we do that, we obtain the discretized equation of
motion where the matrices M and K are obtained in the form as shown in Slide 10.11.

Slide 10.12: (Refer Slide Time: 40:04)

Now, let us look at the same example once again. We have a fixed-free tapered bar with the
boundary conditions at the fixed and the free end as shown in Slide 10.12. Once again, we
choose the comparison functions of the form as shown in Slide 10.12. We take two comparison
functions and discretize the equation of motion. We obtain the discretized equation in the form as
shown in Slide 10.12. Then as usual, we do the modal analysis assuming the structure of solution
as shown in Slide 10.12 and we obtain the discretized eigenvalue problem. This gives us first two
circular natural frequencies as shown in Slide 10.12. The superscript ‘G’ indicates that the
frequency is obtained from Galerkin method. Now, if we compare those with the exact ones, we

100
can see that the fundamental frequency compares very well with the exact while the second
modal frequency is in some error. Now here again if one wants to have an accurate modal
solutions for the first N modes, he/she has to use an expansion with 2N terms. Now, we solve the
eigenvalue problem for the eigenvectors k corresponding to each eigenfrequency as shown in
Slide 10.13.

Slide 10.13: (Refer Slide Time: 47:05)

Also, the eigenfunctions are obtained using the eigenvectors. For the first eigenfunction, we use
the first eigenvector in the vector product. Similarly, for the second eigenfunction, we use the
second eigenvector in the vector product as shown in Slide 10.13. Now, we can see a comparison
of these eigenfunctions in Slide 10.14 with the exact solution that we had obtained previously. It
can be observed that, within the accuracy of the plot, the fundamental eigenfunction solved from
the Galerkin method is indistinguishable from the exact one. For the eigenfunction of the second
mode, it is fairly close and we can also see that the boundary condition at the right end (the free
end) is also matching though the location of the node is in slight error. Thus we say that the
Galerkin method is a method of discretizing the equation of motion of a continuous system. We
can use this method also when we have external forcing and that we will discuss later on in this
course.

101
Slide 10.14: (Refer Slide Time: 50:28)

Slide 10.15: (Refer Slide Time: 52:33)

To summarize, we have studied the projection based approximate methods in which the field
variable is expanded in terms of the comparison functions and substituted in the partial
differential equation of motion of the system. Then we generated the residue and found that the
residue will not be zero uniformly over the domain so we use a weaker condition. We project the
residue onto certain functions and the choice of these functions decides the method of projection.
When we use the Dirac delta function, we have the collocation method. When we choose the
comparison function which are used in the expansion of the field variable as the projection
functions then we have the Galerkin method. These projection methods can handle non-
conservative and non-potential forces as well. With this, we conclude this lecture.

Keywords: Projection method, Comparison function, Precision or Accuracy points,

Chebyshev method or spacing, Collocation method, Residue, Galerkin method.

102
 Vibrations of Structures
Prof. Anirvan DasGupta
Department of Mechanical Engineering
Indian Institute of Technology – Kharagpur

Lecture - 11
The Initial Value Problem

In this lecture, we are going to look at what is known as the initial value problem (IVP) in
vibrations of continuous systems. Let us first understand the concept of Initial value problem?
Suppose we have a string and it is given some initial shape or initial velocity distribution over
the string as shown in Slide 11.1. We want to determine the evolution of the system. Given the
equation of motion, boundary conditions and two initial conditions, how will the system evolve
as time progresses? This is the central problem in the initial value problem.

Slide 11.1: (Refer Slide Time: 00:33)

In order to solve this problem, there can be various approaches. First, we are going to look at the
Modal expansion method. The initial value problem can also be solved by Laplace transform
method which we will see slightly later. We have performed modal analysis of such systems
earlier and we have observed that when we have a self-adjoint system which is to say that the
stiffness operator is self-adjoint then the eigenvalue problem has real eigenvalues and real

103
eigenfunctions. The eigenfunctions have an additional property that they form a complete basis.
This is the underlying thing that is used in modal expansion method.

Slide 11.2: (Refer Slide Time: 03:51)

Slide 11.3: (Refer Slide Time: 03:51)

104
Let us first understand what is the meaning of complete basis? Let us consider the string again.
The general solution of the string is expressed in the form as shown in Slide 11.2. The solution
form says that, at a certain time instant t, the solution for the field variable w(x, t) is an infinite
expansion of the eigenfunctions Wk(x) = sin (kπx/l). The statement ‘the solution is a general
solution’ means that any arbitrary shape of the string can be represented by the expansion of type
as shown in Slide 11.2. We have also discussed earlier that the eigenfunctions are orthogonal. If
we want to have a visualization of the orthogonal eigenfunctions, we create a function space with
these eigenfunctions which we call as the basis (Slide 11.3). Of course, there are infinitely many
basis functions but I have drawn only three and with slight stretch of imagination, one can
foresee that there are infinitely many such possible axis which are all orthogonal to one another
to represent the orthogonality property with respect to a certain inner product that we have
discussed earlier. In such a space which is known as the configuration space or the modal space,
a point with coordinates p1, p2, p3, p4 etc. at a time instant represents the configuration of the
string at that time instant. Thus, at any particular time instant, the expansion in Slide 11.2
represents the shape of the string at that time instant. Thus, any shape of the string is a point in
the configuration space. There is no shape that lies outside this space. Therefore, any shape of
the string can be captured by the expansion shown in Slide 11.2. This is known as the modal
expansion theorem. Thus, the Modal expansion theorem says that any shape of the string can be
represented in terms of the eigenfunctions where the eigenfunction form a complete basis which
means any shape can be represented in this basis. We will now try to solve IVP using the model
expansion method.

We now intend to solve the problem whose differential equation, BCs and ICs are shown in Slide
11.4 using the Modal expansion. Now, if we substitute the expansion (Eq. 11.4.1) in the equation
of motion, we get the Eq. 11.4.2 where K is a linear differential operator. The operator K has
only spatial derivatives and it is linear, therefore we can interchange the summation and the
operator in Eq. 11.4.2. Simplification of Eq. 11.4.2 gives Eq. 11.4.3. Now, if we recall the
eigenvalue problem for the system considered shown in Slide 11.4 and if we replace the operator
acting on the kth eigenfunction with ωk2μ(x)Uk(x), we get Eq. 11.4.4. Now, what we obtained is
again a summation in terms of the eigenfunctions Uk(x).

105
Slide 11.4: (Refer Slide Time: 12:32)

--- 11.4.1

--- 11.4.2

--- 11.4.3

--- 11.4.4

Now, if we multiply both sides of Eq. 11.4.4 by the jth eigenfunction (taking inner product with
Uj(x)) and integrate it over the domain of the system, by orthogonality property, we get Eq.
11.5.1 which is valid for j=1, 2, 3…∞. The general solution of Eq. 11.5.1 can be easily obtained
as Eq. 11.5.2. Substituting the general solution in the original expansion, we get the total solution
u(x, t) as shown in Eq. 11.5.3. The initial conditions to be applied are also shown in Slide 11.5.
Now, using the initial conditions, we have to solve for coefficients Cj and Sj. Let us see how we
can do that.

Substituting Eq. 11.5.3 in the initial conditions, we get the form as shown in Slide 11.6. There
are infinitely many coefficients for both Cj and Sj and we have two equations. We know that the
eigenfunctions Uj(x) are orthogonal. Once again, we use the inner product of Uj(x) and Uk(x)
which eventually filters the kth term and we get the integral form of the coefficients Ck and Sk as
shown in Slide 11.6. Substituting the values of Ck and Sk in Eq. 11.5.3, we can get the final
solution. Thus, we have solved the IVP using the modal expansion technique. Now let us look at
some examples.

Consider the example of collapse of a stretched bar. We have a fixed-free bar which is under

106
tension because of the string attached to its free end and it is under a tension T. Now, if we cut
the string, the string snaps. When the string snaps, the bar is going to collapse back. We are
going to study such a collapse of the bar (Slide 11.7).

Slide 11.5: (Refer Slide Time: 18:10)

--- 11.5.1

--- 11.5.2

--- 11.5.3

Slide 11.6: (Refer Slide Time: 21:13)

107
Slide 11.7: (Refer Slide Time: 25:44)

Slide 11.8: (Refer Slide Time: 25:44)

Let us mathematically formulate the problem. We have a uniform bar, redrawn in Slide 11.8. We
are going to formulate the problem at the moment the string snaps. At that moment, the right end
of the bar is force free which gives EAu,x(l, t) = 0. The initial conditions in the bar are
formulated in the Slide 11.8. Using the equation of statics of the bar, integrating it out, we get the
initial condition u(x,0)=Tx/EA (explained the detailed method in Slide 11.8). The initial velocity
at the moment of snapping of the string is zero. Thus, we have the initial value problem for the

108
collapsing bar. Once again, we are going to express the solution of the collapsing bar in terms of
its eigenfunctions as shown in Slide 11.8. The eigenfunctions of fixed-free bar which we already
know are shown in Slide 11.8. Now, we need to determine the coefficients Ck and Sk.

Slide 11.9: (Refer Slide Time: 32:36)

When we substitute the expansion u(x, t) of Slide 11.8 in the initial conditions, we can
immediately see that for all k, the coefficient Sk must vanish. Now, we are left with the equation
containing Ck only. We take the inner product of kth eigenfunction with jth eigenfunction and
perform the integral over the length of the bar; we get the coefficient Ck as shown in Slide 11.9.
Therefore, we get the complete solution of u(x, t) as shown in Slide 11.9 which I have plotted
non-dimensional axial displacement versus non-dimensional free variable x/l at certain time
instances in the Slide 11.10.

At time t = 0, the axial displacement follows a straight line with x since u = Tx/EA is linear
function of x. As time progresses, at t = 0.5l/c, the axial displacement profile is still linear for
certain x since that portion of the bar does not know as yet that the other end has been released.
Therefore, the information of release has progressed to approximately half of the length of the
bar in time t = 0.5l/c. So, there is a propagation of information from the end which was released
at time t = 0 into the bar. At t = l/c, the information has reached the left end and the whole bar
gains an equilibrium configuration which is not shown in these figures. At t = 1.5l/c, there is a

109
compression taking place in the bar. Before t = 1.5l/c, the bar was under tension. When the bar
comes to the equilibrium configuration at t = l/c, the bar is in equilibrium, but then it has
acquired velocity. So, it still goes and hits against the wall and then there is a compression
generating in the bar and that cycle of compression completes at time t = 2l/c. Thus, we can
imagine that there is a propagation of information of the compression and tension wave that
progresses in the bar and it reflects back and forth and that way the bar vibrates (Slide 11.10).

Slide 11.10: (Refer Slide Time: 35:44)

We are going to discuss the propagation of waves in detail later in this course. Now, we have an
animation which shows the same thing (Slide 11.11). Let us now look at another initial value
problem where we are going to have velocity initial condition.

Consider a fixed-fixed string on which we specify a velocity profile as shown in Slide 11.12. The
initial displacement of the string is considered zero but the velocity has a given profile. Let me
mathematically formulate the problem first. The velocity initial condition is provided only over a
small region which is spread over one fifth of the length of the string as shown in Slide 11.12.
For this problem, once again we consider a solution of the form shown in Slide 11.12. Substitute

110
 this solution in the initial conditions; we obtain two equations which need to be solved for
coefficients Ck and Sk. We find that all the Ck will be zero (Slide 11.13) while the coefficients Sk
takes the form for k = 1, 3, 5… 2n-1 (found after taking the inner product with jth eigenfunction
and integrating it over the domain shown in Slide 11.14). The coefficient Sk turns out to be zero
for all the even values of k. Now, substituting Sk in the solution expansion (in Slide 11.12), we
get the final solution of the motion of the string.

Slide 11.11: (Refer Slide Time: 38:51)

Slide 11.12: (Refer Slide Time: 39:51)

Velocity-profile

111
Slide 11.13: (Refer Slide Time: 39:51)

Slide 11.14: (Refer Slide Time: 45:32)

Let us now look at the snap shots of the string at certain time instance. At time t = 0, the string is
in equilibrium position (not shown). At t = 0.05l/c, there is a hump developed in the string. As
time progresses to t = 0.25l/c, this hump spreads over the string. But, we should note that the
portion of the string near ends does not know that a disturbance has been created in the string

112
yet. At time instant t = 0.5l/c, entire string is displaced. Beyond this time, the disturbance reflects
backs from the fixed ends and the hump shrinks and then it comes to the other side of the
equilibrium position of the string.

Slide 11.15: (Refer Slide Time: 47:58)

Slide 11.16: (Refer Slide Time: 47:58)

113
The animation in Slide 11.16 shows this clearly. Here, we can see that the hump develops,
spreads, reflects, collapses and comes on the other side and does the same thing over again. It
should be noted that the animation is a slow motion of what actually happens. The progression of
waves possibly cannot be observed by the human eye as such. One has to see it in slow motion to
observe this kind of propagation of disturbance.

Till now, we have looked at the initial value problem for continuous systems and we have
considered two examples and using the model expansion technique, we have solved the
problems. Now, let us briefly look at how an initial value problem can actually be converted to a
forced vibration problem or a forced dynamics of the continuous system. When we will discuss
the forced vibration analysis (later in the course) then we will understand the analogy very well.
It implies that we can solve the initial value problem as a force vibration problem also. In this
way, we have a unified way of treating both, initial value problem and force dynamics.

Slide 11.17: (Refer Slide Time: 49:57)

Let us now look at the conversion of an initial value problem to a forced vibrations problem as
shown in Slide 11.17. Consider a system with certain boundary conditions and initial conditions

114
as shown in Slide 11.17. Now, we take the Laplace transform of the equation of motion. We
have defined the Laplace transform in the Slide 11.17 where s is the Laplace variable. In the
Laplace domain, we write the equation of motion as shown. Now the same equation in the
Laplace domain can be obtained for the set of equation of motion, BCs and ICs shown in the box
in Slide 11.17. We notice that the system now has zero initial conditions. So, here in instead of
these initial conditions, none zero initial conditions. Also, we have a forcing term, an
inhomogeneity in the equation of motion (Non-zero RHS of Equation of motion). Thus, we have
been able to convert a system with non-zero initial conditions to a system with zero initial
conditions but with forcing. In this way, we can convert an initial condition problem to a forced
vibration or the forced dynamics problem.

To summarize, we have looked at the initial value problem. We have solved IVP using the modal
expansion technique. We have looked at some examples of how the solution behaves and we
have made some interesting observations in the solution how the motion actually takes places as
a propagation of disturbance in the continuous system. And finally we have looked at converting
an initial value problem to a forced vibration problem. With this, we can solve the initial value
problem and the force dynamics problem in a unified manner in subsequent lectures.

Keywords: Initial value problem, Modal expansion method, the basis, Laplace transforms,
Forced vibration, Collapse of bar

115
 Vibrations of Structures
Prof. Anirvan DasGupta
Department of Mechanical Engineering
Indian Institute of Technology – Kharagpur

Lecture – 12
Forced Vibration Analysis - I

We have been looking at the response of one-dimensional continuous system to initial

conditions in the previous lecture. We have also looked at how the initial value problem can
be recast as a forced vibration problem. In this lecture and subsequent lectures, we are going
to concentrate on the forced vibration analysis of continuous systems. Now the question
arises what are the sources of forcing, how or why we should study forcing. There are various
possible reasons. One can have a system with an actuation, say for example, for vibration
control or for some other control. The actuators will excite the modes of the system.
Secondly, one can have a fluid forcing. For example, a bridge or a high rise building that
might be excited by the flowing wind which provides forcing to structures. Earthquakes and
similar natural sources can also cause forcing to structures. Finally, the forcing is also found
as an application in evaluation and testing of materials like to detect flaws or phases or faults
in material or in a structure. From these considerations, it becomes important to analyse
forced vibrations of systems.

Let us briefly look into the ways of forcing a structure. One can attach an actuator to or on the
structure and can force it. Also, in a violin one uses a bow to excite the string; one can hit the
structure with an impact hammer resulting impact or impulsive forcing to the structures. Let
us now start the discussions on forced vibrations of one-dimensional continuous system. Let
us consider a system for which we first formulate the problem mathematically as shown in
Slide 12.1. The kind of systems we have been discussing till now can be put in the
mathematical form as shown in Slide 12.1 where q(x, t) represents a general forcing.
Applicable initial and boundary conditions are also given along with the equation of motion
in Slide 12.1. That forms the complete formulation of the forced dynamics of a system. Now,
the forcing term in the RHS of the differential equation makes the equation of motion
inhomogeneous. Thus, we no longer have w(x, t) = 0 as a solution of this system which is the
trivial solution.

116
Slide 12.1: (Refer Slide Time: 03:09)

Now, there can be various kinds of forcing. We can have harmonic forcing which is the
most common kind of forcing specially when we are evaluating or testing a structure. We
provide harmonic forcing to the structure and try to see whether the output response matches
with our expected one or not. The harmonic forcing is represented as q(x, t) = Q(x)cosΩt
where Ω is the forcing circular frequency and Q(x) is the amplitude function. We can see here
that this kind of forcing is separable in space and time. The amplitude function is the spatial
distribution of the force and the other part is the temporal variation of the force. Also, any
periodic forcing as you know can be represented as a series of harmonic forcing. Thus, if we
know the solution for a harmonic forcing, then we can also find out the response to any
periodic forcing.

When we have the forcing, which is non-separable in space and time, then we call it general
forcing. We have actually looked at one example of general forcing in a previous lecture
when we discussed about recasting the initial value problem as a forced vibration problem.
We will discuss general forcing in the later lectures. In this lecture, we are going to focus on
the harmonic forcing only. The differential equation of motion with harmonic forcing is taken
in Slide 12.2 as Eq. 12.2.1. Here, we represent the force as a real part of the complex forcing
term Q(x)eiΩt. The equation of motion along with the boundary and the initial conditions will
completely specify the forced vibration problem.

117
Slide 12.2: (Refer Slide Time: 09:11)

--- 12.2.1

Now, we write down the general solution of the differential equation (Eq. 12.2.1) as shown in
Slide 12.2. We know that the general solution of such a differential equation can be written as
the homogeneous solution, which means the solution with zero forcing plus the particular
solution which is due to this forcing. This kind of a solution satisfies the inhomogeneous
differential equation since the homogeneous solution will actually go to zero once we
substitute the solution in the equation and the particular solution will equate the right-hand
side. We know the homogeneous solution from last few lectures and it is expanded in terms
of the eigenfunctions Wk(x) of the corresponding eigenvalue problem in (Eq. 12.2.1) as
shown in Slide 12.2. The eigenvalue problem was obtained by considering the homogeneous
part of Eq. 12.2.1 and searching for special solutions which are separated in space and time.
We have a particular solution reading R[X(x)eiΩt] which satisfies the non-homogeneous term
on the right-hand side of the differential equation. Here, X(x) is the amplitude function of the
response. Now, as we know that in an undamped system, the response is proportional to
harmonic time function eiΩt, so, we have taken the response as the real part of the product of
amplitude function X(x) and eiΩt.

Now, when we substitute the solution form in the differential equation of motion, the
homogeneous term vanishes and finally we obtain, with a little bit of simplification, the
differential equation in X(x). Along with this, we get the boundary conditions which the
amplitude function must satisfy (shown in Slide 12.2). This specifies what is known as the

118
boundary value problem. Thus, the boundary value problem corresponding to the amplitude
function of the particular solution is obtained.

Now, we must solve this boundary value problem in order to solve for the amplitude. Now,
there can be various ways of solving this boundary value problem. One is the eigenfunction
expansion which we have been looking at in the past few lectures. This works on the premise
or the fact that for self-adjoint problems, you have the eigenfunctions which are all real and
which form a complete basis for the system. By complete basis, I mean that any configuration
or shape of the system can be represented in terms of the eigenfunctions. The other method is
the green’s function method which we will see in the next lecture. In this lecture, we are
going to focus on the eigenfunction expansion method for solving the boundary value
problem.

Slide 12.3: (Refer Slide Time: 18:45)

--- 12.3.1

--- 12.3.2
--- 12.3.3

--- 12.3.4

We rewrite the differential equation of the boundary value problem in the form as shown in
Slide 12.3. We now expand the general solution of this differential equation in terms of the
eigenfunction Wk(x) of the problem (Eq. 12.3.1). If we substitute the expansion in the
differential equation, we obtain Eq. 12.3.2 where αks are the constants which are to be solved.
Since K is a linear differential operator, we can exchange the summation and the operator.
Recalling the eigenvalue problem, we write the operator acting on the Kth eigenfunction in
the form as shown in Slide 12.3. Substituting the result of eigenvalue problem into Eq. 12.3.3

119
and simplifying, we get the equation where the unknown coefficients αk need to be found
(shown in Slide 12.3). For this, we can use the orthogonality condition for the eigenfunctions.
So, let us see how we can solve it using the orthogonality condition.

Slide 12.4: (Refer Slide Time: 24:45)

--- 12.4.1

The orthogonality condition of the eigenfunctions for the system that we are considering is
shown in Slide 12.4 where j ≠ k. We denote this as the inner product also as shown in Slide
12.4. Multiplying both sides of Eq. 12.3.4 with the jth eigenfunction and integrate over the
domain of the problem. Now, when we do that, due to orthogonality, only the jth term is
filtered out. Thus, we solve for αj where j = 1, 2, 3...∞. One should note here that the solution
of αj is possible only on the condition that the forcing frequency is not equal to any of the
natural frequencies of the system. In case of matching frequencies, αj will go to infinity. This
completes the solution for the non-resonant case. Rewriting Eq. 12.3.1, we now have the
solution X(x), the particular solution and the complete solution as shown in Slide 12.5. The
constants Ck and Sk in the homogeneous solution are to be determined from the initial
conditions which we have covered in previous lecture. This will complete the solution of the
forced vibration problem (refer Slide 12.5).

Let us now consider the case of resonant forcing (refer Slide 12.6). Let us assume that the
forcing frequency is equal to one of the circular natural frequencies of the system, say j th
circular natural frequency. Therefore, the differential equation of motion and particular

120
solution modifies to Eq. 12.6.1 and 12.6.2 respectively. Now, we have been expanding the
amplitude function X(x) in terms of the eigenfunctions Wk(x). Here, we do the same for all k,
except the case where k = j as αj becomes undefined at ωj = Ω. Thus, the same expansion in
terms of eigenfunction works for the entire non-resonant modes. For the resonant mode, we
are going assume that the coefficient αj is now a function of time as we do in Frobenius
method. Thus, total solution for amplitude function looks like Eq. 12.6.3.

Slide 12.5: (Refer Slide Time: 29:40)

Slide 12.6: (Refer Slide Time: 31:59)

--- 12.6.1
--- 12.6.2

--- 12.6.3

--- 12.6.4

--- 12.6.5

121
Now, we substitute the form in Eq. 12.6.3 into the particular solution in Eq. 12.6.2 and then
in that differential equation in Eq. 12.6.1, we obtain a differential equation corresponding to
αj as shown in Eq. 12.6.4. The right-hand side of Eq. 12.6.4 is obtained by first taking the
inner product with Wj, and then using the property of orthogonality. The differential equation
in Eq. 12.6.4 admits a solution of type in Eq. 12.6.5 where βj and γj are constants. The
constant γj vanishes when we substitute the solution form in the differential equation. We
now substitute the solution form in the differential equation in 12.6.4; we obtain the
expression for βj as shown in Slide 12.6.

Slide 12.7: (Refer Slide Time: 38:33)

--- 12.7.1

--- 12.7.2

Substituting all the intermediary results into the particular solution and taking the real part of
it, we obtain the form as shown in Slide 12.7. Thus, this completes the solution of the
particular solution. Now once again, we have to add it with the homogeneous solution and
use the initial conditions to solve the constants in the homogeneous solution.

Here, we see something interesting in this solution form in Eq. 12.7.1. In the numerator Eq.
12.7.1, the amplitude of the resonant mode has an envelope which is linearly increasing with
time. This is what happens for the resonant mode as we are aware of. Now, in a continuous
system like this, we have the integral shown in Eq. 12.7.2 sitting in the numerator of the
resonant solution in Eq. 12.7.1. Now, this integral, in general is non zero but then there are
special instances where this will actually vanish. So, let us look at such situation. Let us

122
consider ωj to be the second natural frequency, i.e., the circular forcing frequency Ω is equal
to the second circular natural frequency of the system. Now, for a taut string, we have the
eigenfunction for the second mode as W2(x) = sin2πx/l. If we have a string which is being
forced by a forcing function of form δ(x - l/2)cosΩt at the mid-span and if we substitute the
expressions Q(x) = δ(x - l/2) and W2(x) = sin2πx/l in Eq. 12.7.2, we see that the integral
vanishes. Thus, the integral vanishes, even though we are exciting the string at its second
natural frequency. It implies that the second mode will not show the resonant behaviour in
this case as the time dependent term in Eq. 12.7.1 vanishes though the response of the system
is still finite. Thus, force like what we just considered, cannot excite this mode since the
forcing is at the node of that mode. So, this is one situation where there won’t be any forcing.

There can be other situations where we will see the similar behaviour. One such example we
are going to look into very shortly. Therefore, just forcing the system at a resonant frequency
does not mean that we will observe a resonant solution. So, the location of the force is also
important in these situations.

Slide 12.8: (Refer Slide Time: 45:21)

--- 12.8.1
--- 12.8.2

Let us now take an example of forced vibrations. We consider a fixed-fixed taut string with
uniform harmonic forcing as shown in Slide 12.8. The differential equation of motion and the
boundary conditions are given in Slide 12.8. Now, we are going to look at the solutions of the
form given in Eq. 12.8.1. If we substitute it in the equation of motion and removing the cosΩt

123
term throughout then we get ordinary differential equation in X(x) along with the boundary
conditions in Eq. 12.8.2. We simplify it a little and finally we get what we call as the
boundary value problem (BVP) of our system. We know that the eigenfunctions of the taut
string are of the form shown in Slide. So, we expand in terms of these eigenfunctions and
when we substitute it in differential equation and take inner product with the jth
eigenfunction, we obtain the solution of αj which then can be substituted in the expansion to
obtain the amplitude function X(x) as shown in Slide 12.8 and 12.9.

Slide 12.9: (Refer Slide Time: 51:02)

--- 12.9.1

--- 12.9.2

The particular solution of the problem is finally obtained as shown in Slide 12.9. The
boundary value problem in Slide 12.8 can also be solved exactly as the given differential
equation is straight forward and the solution of this differential equation can be easily written
as in Eq. 12.9.1 which generates additional constants D and E. We solve for the constants
using boundary conditions as explained in Slide 12.9. One should note that in the
eigenfunction expansion method, we did not have to worry about the boundary conditions
since they already satisfied the boundary conditions. We now obtain the exact particular
solution as shown in Eq. 12.9.2. It can be seen that the particular solution so obtained is in
closed form. Also, we should note that we have the term [cos(kπ)-1] present in the Eq. 12.9.1
which vanishes for even values of k. Therefore, we will have non-zero solutions for odd
values of k only
.

124
Slide 12.10: (Refer Slide Time: 55:12)

In Slide 12.10, I have compared the exact solution closed form solutions with series solution
taking up to three terms. In the first plot, I have taken the forcing frequency very close to the
first natural frequency of the string (Ω = 0.9ω1). For this case, the exact solution and the
series solution matches very well. In fact, I have plotted the series solution with increasing
number of terms, up to the maximum of three. In the first plot, they are indistinguishable with
the exact solution. When we force the system close to the second natural frequency, i.e., Ω =
1.9ω1 (Higher modes are integral multiples of the fundamental frequency), we see that the
dashed curve, which is the series solution with only one term, deviates considerably from the
actual solution. When we take two or three terms in the series, they match quite nicely. When
the forcing is close to the third natural frequency, i.e., Ω = 2.9ω1, we see the one term series
solution is quite off while two or three term series solution matches quite nicely with the
exact solution. When we are forcing the string at frequency close to the fourth natural
frequency, i.e., Ω = 3.9ω1, the one term series solution is quite deviated from the exact one.
The two-term solution (blue solid line) and the three term series solution (red solid line)
approaches the exact solution (black solid line) with increasing accuracy. Thus, as we go to
higher and higher frequencies, we will have to take more and more terms in the series to get
close to the exact solution. Thus, we see that the series solution actually converges on to the
exact solution.

125
To summarise, let us look at what we have studied today. We have discussed the force
vibration analysis of one-dimensional continuous system. We have looked at harmonic
forcing and we have solved the problem using the eigenfunction expansion method. We are
going to continue this discussion in the next lecture. With this, we conclude this lecture.

Keywords: Forced Vibrations, Harmonic forcing, Boundary value problem,

Eigenfunction expansion method, Resonance

126
 Vibrations of Structures
Prof. Anirvan DasGupta
Department of Mechanical Engineering
Indian Institute of Technology – Kharagpur

Lecture – 13
Forced Vibration Analysis - II

In the previous lecture, we had initiated discussions on forced vibration analysis of one
dimensional continuous system. We started with the case of harmonic forcing, separable in
space and time. Let me recapitulate what we started in the earlier lecture before proceeding
with the today’s topic.

Slide 13.1: (Refer Slide Time: 00:53)

So, we were looking at the forced vibration analysis of a system governed by the equation of
motion along with the boundary and initial conditions (BCs and ICs) (Slide 13.1). We
observed that the solution of these dynamical equations can be written as the summation of
the homogeneous solution which we expanded in terms of the eigenfunctions [Wk(x)] of the
unforced problem and the particular solution where, X(x) is the amplitude function (Slide
13.1). When we substitute the shown solution form in the equation of motion and BCs, the
contribution from homogeneous solution is zero while the contribution from the particular
solution along with the boundary conditions gives the Boundary Value Problem (BVP) as
shown in Slide 13.1.

127
Now, we have to solve the boundary value problem in order to calculate the amplitude
function of the particular solution X(x). We also have unknowns, CK and SK which must be
determined from the initial conditions after we have solved the amplitude function of the
particular solution. We have seen the solution of CK and SK in one of our previous lectures. In
today’s lecture, we are going to focus on the solution of the boundary value problem. In the
previous lecture, we discussed about the eigenfunction expansion method. Today we are
going to look at the solution of the boundary value problem using Green’s function method.

Slide 13.2: (Refer Slide Time: 06:39)

--- 13.2.1
--- 13.2.2

What is the Green’s function? The Green’s function G(x, x̅, Ω) is the solution of the
boundary value problem of one dimensional continuous system with a very special form of
Q(x) which is the Dirac delta distribution applied at x = x̅ denoted as δ(x- x̅) as shown in Slide
13.2. Thus, when one applies a harmonic concentrated forcing of frequency Ω at the location
x = x̅, the solution of such a system is the Green’s function. In other words, the Green’s
function is the solution of the vibration amplitude at x due to a harmonic forcing of
frequency Ω applied at x̅.

It should be noted here that whenever we talk about Green’s function, we always use
homogeneous BCs. If it is non-homogeneous, then we have to homogenise the boundary
conditions, the knowhow of which we discussed in previous lectures. Here, we are going to

128
discuss the Green’s function in the context of homogeneous BCs only. Now, the question
arises how is the solution of the Green’s function going to help us in solving a problem with
an arbitrary forced distribution, Q(x)? We claim that the solution of the amplitude function
corresponding to an arbitrary forcing Q(x) is given by the integral in Eq. 13.2.1. Here, we
integrate the product of given force function and Green’s function over x̅, from 0 to l,
yielding a function of x which is the solution of the problem. We, now prove the claim in
13.2.1. Let us rewrite the differential equation in an abbreviated form as shown in Eq. 13.2.2
where L is some operator acting on X(x) giving Q(x).

If Eq. 13.2.1 is the solution of X(x), I substitute it in Eq. 13.2.2. Since L is the linear
differential operator, we can assume that it can commute with the integral. Also note that the
integral is over x̅ and L is an operator on x, so the operator L acts only on Green’s function G.
Now, we already know the result of operator L acting on G, which is δ(x- x̅). When it is
integrated over x̅, we get Q(x) which is the right hand side of Eq. 13.2.2. It means that 13.2.1
must be a solution to the problem 13.2.2. Thus, we only need to solve for the Green’s
function which is for a very special form of force distribution δ(x- x̅) and once this is done,
we can solve for the amplitude function X(x) given any force distribution Q(x) using Eq.
13.2.1. Thus, the solution of the original problem w(x, t) can be written as the summation of
the homogeneous solution and the real part of the argument as shown in Slide 13.3. For
homogeneous part wH(x, t), the unknowns can be found out by using the ICs.

Slide 13.3: (Refer Slide Time: 16:27)

129
Now the problem remains that how to solve for the Green’s Function G. For this, let us take
the example of a string once again (Slide 13.3). Here, we have a concentrated harmonic
forcing δ(x- x̅)eiΩt at x = x̅. The equation of motion (EOM) and BCs of the string are shown in
Slide 13.3. Now, if we substitute the solution form that we have been considering into the
EOM and BCs, we get the boundary value problem (BVP) as shown in Slide 13.3.

Slide 13.4: (Refer Slide Time: 21:28)

--- 13.4.1

In order to solve this, we are going to look at two regimes of the string. Let me call these as
the left regime (0≤x< x̅) and the right regime (x̅<x≤ l). In these two regimes, we rewrite the
differential equation of the BVP as shown in Slide 13.4. At x = x̅, the Dirac delta forcing
function is acting. In other regions, there is no forcing. Now, we can easily write the general
solution for the Green’s functions G in these two regions as shown in Slide 13.4. We have the
boundary conditions at x = 0 & l as shown in Slide 13.4. We also have the continuity
condition at the, junction x = x̅. Here, we have four unknown coefficients namely AL, BL, AR
and BR which are to be solved in order to determine the Green’s function. We will need four
conditions for solving these four unknowns. Two conditions are obtained directly from the
boundary conditions. One further condition has been obtained from the continuity of the
string at x = x̅. The fourth condition must of course come from the force balance at x = x̅. The
force condition for the string can be easily found out by directly integrating the differential
equation of the BVP as shown in Slide 13.5.

130
Slide 13.5: (Refer Slide Time: 27:03)

--- 13.5.1

--- 13.5.2

Now, if we integrate the both sides of the differential equation of the BVP over the domain of
the string, we can easily obtain the fourth condition which is nothing but the force balance
condition at the junction x = x̅ as shown in Slide 13.5. Note that over the interval 0 to l, at
almost all points, the integrand in Eq. 13.5.1 is zero except at x = x̅. It implies that from zero
to x̅ - ε and x̅ + ε to l, the integrand is zero. It is only a small region around x = x̅, the
integrand is non-zero. The term Ω2G in the integrand when integrated over such a narrow
region from x̅ - ε to x̅ + ε, results in zero since G is a continuous function. For a continuous
function G, the integral over diminishing domain goes to zero. Therefore, the term Ω2G does
not contribute anything in the integral. The first integral of other term gives slope of the
string as G,x. Note that the string does not resist bending moments and thereby it can have a
slope discontinuity. Thus we get the fourth condition as Eq. 13.5.2. Now, these four
conditions can be used to solve for unknown coefficients, AL, BL, AR and BR. We finally
substitute the coefficient expressions in Eq. 13.4.1, and we obtain the Green’s function as
shown in Slide 13.6. Now, corresponding to any force distribution Q(x), we solve for the
amplitude function X(x) of the particular solution as shown in Slide 13.6.

As an example, we consider a string with a uniformly distributed harmonic force (Slide 13.7).
We have looked at this example in the previous lecture also. Let us now solve the same
problem with the Green’s function method. The equation of motion with the boundary
conditions is written in Slide 13.7. We rewrite the differential equation of motion in the form

131
so as to match it with that for which Green’s function is derived in Slides 13.3 to 13.6.
Slide 13.6: (Refer Slide Time: 32:35)

--- 13.6.1

--- 13.6.2

--- 13.6.3

The total solution w(x, t) is the summation of homogeneous solution and the particular
solution in the form shown in Slide 13.7. We have already derived the Green’s function for
the taut string and we need to solve the amplitude function X(x) of the particular solution by
performing the integral where Q(x̅)=F0/ρA as shown in Slide 13.7.

Slide 13.7: (Refer Slide Time: 34:59)

132
Note once again that the Green’s function is the response of a system when a concentrated
harmonic force is applied at x = x̅. The domain of the string is from zero to l and we want to
find out the amplitude function X at any arbitrary location x. Therefore, we perform the
integral in Slide 13.7, in two regions of x̅, one from 0 to x and other from x to l. Now, in the
range from 0 to x, we note that x̅ < x. So, the Green’s function corresponding to this region
would be Eq. 13.6.2. Similarly, in the range x to l, we note that x̅ >x, So Eq. 13.6.1 must be
the required Green’s function. Let us now carry out the integration (Slide 13.8).

Slide 13.8: (Refer Slide Time: 41:38)

The integral is quite straight forward and upon simplification, we obtain the amplitude
function of the particular solution as explained in Slide 13.8. Now, one can check the
obtained expression for amplitude with what is obtained in the previous lecture; they are
exactly same. So, when we solve problems of forcing with Green’s function, what we need to
look at is the integral (Eq. 13.6.3) over the domain and to be performed little carefully taking
into account the regions of the problem, as we have done here. Now this Green’s function can
also be determined using the modal expansion technique that we discussed in the last lecture.

Let us briefly look at the method of solving the Green’s function using modal expansion or
the Eigen function expansion method (Slide 13.9). The eigenvalue problem for the Green’s
function is shown in the Slide 13.9. We, now represent the Green’s function as an expansion
in terms of the eigenfunctions of the unforced problem as shown in Eq. 13.9.1.

133
Slide 13.9: (Refer Slide Time: 48:06)

--- 13.9.1

--- 13.9.2

Note that the eigenfunctions already satisfy the boundary conditions; therefore the Green’s
function is also guaranteed to satisfy the boundary conditions of the problem. Substituting the
expansion in the differential equation of the boundary value problem, exchanging the
summation with operator K[.], using Eq. 13.9.2, taking inner product with Wj(x) on both side
of the equation and using the orthogonality property, we obtain the coefficient αj for j = 1, 2,
...infinity as shown in Slide 13.9. Thus, we have the series expansion of the Green’s function
in terms of the Eigen functions of the problem.

Thus, we have solved at the boundary value problem which arises in the force vibration
analysis. The boundary value problem actually gives us a solution of the amplitude function
of the particular solution. In this lecture today, we solved the boundary value problem using
the Green’s function method and we have looked at the same example what we took in the
previous lecture and we compared the solution with that obtained in the previous lecture.
With that, we conclude this lecture.

Keywords: Taut String, Forced Vibrations, Boundary value problem, Green’s Function,
Eigenfunction expansion method

134
 Vibrations of Structures
Prof. Anirvan DasGupta
Department of Mechanical Engineering
Indian Institute of Technology- Kharagpur

Lecture – 14
Forced Vibration Analysis – III

We have been looking at Forced Vibration Analysis of one-dimensional structure in the past few
lectures. Today, we are going to continue our discussions on forced vibration. In the previous
lectures, we have seen systems with harmonic forcing separated in space and time which is
written as the product of a distribution function and a harmonic function of time. In this lecture,
we are going to look at general forcing which need not necessarily be separable in space and
time.

Slide 14.1: (Refer Slide Time: 01:08)

--- 14.1.1

--- 14.1.2

Let us consider some examples of a general forcing. In one of the previous lectures, we have
already seen an initial value problem with certain BCs. An example initial value problem and its
equivalent forced vibration problem are shown in Eqs. 14.1.1 and 14.1.2. Here, δ(t) is the Dirac
delta function and the dotted δ(t) is the time derivative of the Dirac delta function at t = 0. We
should note here that the forcing term in Eq. 14.1.2 is not a form of harmonic forcing. Forcing of

135
such types are encountered in situation like a force travelling on a string say a car (or any load)
travelling at a certain speed over a bridge. Thus, we have a travelling load/mass problem to look
into. In such problems, space and time are coupled; they are not separable as we will presently
see in this lecture. Here, the dynamical problems of this type are considered with general forcing.

Slide 14.2: (Refer Slide Time: 04:58)

Let us now look at how we solve a general forcing problem. For a given differential equation of
motion with certain boundary conditions (Slide 14.2), we will discuss two methods namely the
modal expansion method and the Green’s function method. So, let us start with the modal
expansion method. We have already discussed the knowhow of the modal expansion technique
in previous lectures. We know that the modal space of the eigenfunctions of an EVP of a system
form a complete basis which means that any configuration of the system can be represented by
these eigenfunctions. This is the basis of the modal expansion technique.

Now, applying the same on the differential equation of motion with given BCs as explained in
Slide 14.3. The solution is represented in the form as in Eq. 14.3.1 where Wk(x) is the
eigenfunctions of the corresponding eigenvalue problem (shown in Slide 14.3). We Substitute
the solution form in the equation of motion and BCs. As always, the eigenfunctions satisfy the
BCs. Here, we now have to determine the coefficients pk(t) so that the expansion in Eq. 14.3.1

136
satisfies the equation of motion. Here, K[.] is a linear differential operator and we can commute
it with the summation sign which gives us the Eq. 14.3.2. Recalling the EVP for the system, we
replace the expression of the stiffness operator operating on the eigenfunction K[Wk] with
μωk2Wk which results in Eq. 14.3.3.

Slide 14.3: (Refer Slide Time: 06:54)

--- 14.3.1

--- 14.3.2

--- 14.3.3

Slide 14.4: (Refer Slide Time: 13:04)

--- 14.4.1

--- 14.4.2

137
Now, we use the orthogonality condition by taking the inner product of Eq. 14.3.3 with Wj(x) as
shown in Slide 14.4. We multiply the equation with Wj and integrate it over the domain of the
problem which filters out the jth coefficient in the expansion. Thus, we get the jth equation
where fj(t) takes the form as shown in Eq. 14.4.2. For different Wj(x), we can generate different
equations (Eq. 14.4.1) corresponding to the coordinates of the individual modes pj. Thus, we
have the equations governing the coordinates of the system in the configuration or modal space
(pj) and this represents a forced dynamics problem where the forcing of the jth mode is expressed
in the form shown in Eq. 14.4.2. Thus, for the applied external force q(x, t), one can easily
compute the function fj(t) by performing the integration shown in Eq. 14.4.2. Now, Eq. 14.4.1 is
a second order ordinary differential equation, for solving which we will need two initial
conditions.

Slide 14.5: (Refer Slide Time: 17:16)

--- 14.5.1

We are given the initial conditions on the field variable as shown in Slide 14.5. We once again
use the modal expansion on the given initial conditions as shown in Slide 14.5. Upon
substitution, we get the initial conditions on pk(0) and ṗk(0) which we are now going to solve
from Eq. set 14.5.1. For this, we multiply both sides of the equations with μ(x)Wj(x) and integrate
them over the domain of the problem. Once again, using the orthogonality condition, we filter

138
out the jth term in the expansion. And thus, we get pj(0) and ṗ j(0) for j = 1, 2…. As shown in
Slide 14.5. Now, once we have all these initial conditions, then we can solve the problem in Eq.
14.4.1 easily. Note, that what we have achieved in this process is the discretized equations of
motion (The complete problem statement along with modified ICs in pj is shown in Slide 14.6.).

Slide 14.6: (Refer Slide Time: 23:07)

Slide 14.7: (Refer Slide Time: 24:08)

--- 14.7.1

--- 14.7.2

--- 14.7.3

139
Let us now look at some examples. The first example that we are going to consider is that of a
force travelling on a string (Slide 14.7). We have a taut string on which there is a concentrated
constant force F which is travelling at a speed v. The mathematical formulation of the problem
along with BCs is shown in Eq. 14.7.1. We use the expansion form in Eq. 14.7.2 in terms of the
eigenfunctions of the taut string. If we substitute it in the EOM and follow the procedure that we
just now discussed then for the jth mode, the discretized equation of motion appears to be in the
form in Eq. 14.7.3 where ωj = jπc/l.

Slide 14.8: (Refer Slide Time: 28:28)

--- 14.8.1

--- 14.8.2

--- 14.8.3

The solution pj(t) of the Eq. 14.7.3 can be easily written as shown in Eq. 14.8.1 provided v ≠ c
where c is the speed of transverse waves in the string. We note that when v = c, the forcing term
in Eq. 14.7.3 becomes a resonant forcing of the string in the jth mode and therefore in all modes.
And, for resonant forcing of the string, the solution form obtained is different than that for non-
resonant case. Therefore, the solution in Eq. 14.8.1 is valid as long as v ≠ c. You can also find
out the resonant solution using the standard techniques that you have studied in a course on
discrete vibrations. In the solution form in Eq. 17.8.1, we have two unknown coefficients from
the homogenous solution of the EOM, namely Cj and Sj which have to be found out from the ICs.
So, we consider a stationery string before the force starts travelling on the string which makes

140
the initial displacement and initial velocity as zero and if we solve for these coefficients, they
turn out to be of the form as shown in Eq. 14.8.2. Therefore, the final solution of the problem is
shown in Eq. 14.8.3 which is valid for the time interval for which the force is on the string i.e.,
0 < t < l / v. As soon as the force crosses the string, the solution in Eq. 14.8.3 is no longer valid.
In this case, one has to form the problem of the string with certain initial conditions which are
the final conditions at t = l / v of Eq. 14.8.3. Thus, one has to solve this particular initial value
problem to understand how the string behaves after the force has left the string. Let us now look
at certain configurations of the string.

Slide 14.9: (Refer Slide Time: 34:58)

In Slide 14.9, configuration of the string at certain time instants are shown. In the first snapshot,
at time t = 0.2l/v, the force has travelled one fifth of the total length of the string. We can see that
some portion of the string is as yet undisturbed but when t reaches 0.4l/v, 40% of the distance has
been covered and this information that there is a force travelling on the string has reached the
end x/l = 1. At t = 0.6l/v, we can see kink forming somewhere in the first half of the string which
implies that the information is now reflecting back and forth on the string. At last, at t = 0.8l/v,
the configuration is as shown in Slide 14.9. Further, a simulation/animation of the same is
presented whose snapshot is shown in Slide 14.10. For this, I have taken a number of terms in

141
the expansion which we just discussed. Then, I have plotted the shapes of the string in different
time instants and prepared this animation. Here, you can see that the string appears to move in a
very strange manner. In fact, the certain portions of the string appear static for certain time
instants. In reality, the disturbance which is created by the travelling force travels on the string as
a wave and it reflects back and forth on the string from the boundaries unless the disturbance
reaches a portion of the string which does not know that the force is travelling. Therefore, that
portion of the string remains static or appears as if static.

Slide 14.10: (Refer Slide Time: 35:41)

One should note two things while studying the presented animation (See video in the lecture.).
The first thing is that it is a slow motion of the actual thing that is happening. It is very difficult
to observe this kind of motion by naked eye, so one has to record it and run it in slow motion.
The second thing is that this animation does not take care of the motion of the string after the
force has left. The present animation is just a repeat of itself when the force is travelling on the
string. So here, we do not see the initial value problem solution.

In the first animation (first snapshot in Slide 14.10), the force travels at low speed i.e., v/c = 0.1
which is 10% of the critical speed. In the second animation (second snapshot in Slide 14.10), the
force travels at near resonance speed i.e., v/c = 0.95 which is 95% of the critical speed. Here you
will find that the reflections of the kinks etc. are not so easily observable. Also, the amplitude is
growing as the force moves from left to right which is a typical behavior of near resonant
solution. At resonance we know that the amplitude goes linearly with time which can be inferred

142
from the animation of near resonance speed. The near resonance behavior is distinctively
different from that at far away speed. So, the motion shown in animation is transient motion
which has to be understood in terms of the propagation of waves in the string which we are going
to take up later in this course.

Slide 14.11: (Refer Slide Time: 39:31)

--- 14.11.1

--- 14.11.2

So far, we have looked at the modal expansion solution. There is another method which we are
now going to look at briefly. We have already come across the Green’s function method in our
previous lecture. So, we now solve a problem where the forcing term has a very special form
using Green’s function method. In Slide 14.11, the symbol δ(.) is Dirac delta distribution
function. The forcing term in Eq. 14.11.1 is an impulsive loading at the location x = x̅ and at time
t = t̅ . Now, we are going to look at the solution of the system to such a loading and here the
boundary conditions are homogenous and initial conditions are all zero as shown in Slide 14.11.
As we already know that the solution of such a system is known as the Green’s function.
Therefore, we will represent the solution as w(x, t) = G(x, x̅, t, t̅ ). Remember that the Green’s
function is always calculated with the homogeneous BCs and zero ICs. Also, we have seen in
previous lecture, how non-homogenous BCs can be converted into homogenous BCs.

143
 Let us first look why or how the Green’s function method works. To begin with, we claim that
for any arbitrary forcing of form q(x, t), the solution can be obtained in the form shown in Eq.
14.11.2. Once we have the Green’s function, the solution to an arbitrary forcing can be
determined like this as shown. Let us find out why or how does this happen. We write the
differential equation in Eq. 14.11.1 as Eq. 14.12.1 where the operator P[.] is of the form shown in
Slide 14.12. The problem statement with homogenous BCs and zero ICs looks like as shown in
Eq. 14.12.2. Now, as we know from Eq. 14.11.2, the proposed solution form is substituted. We
now apply the operator P on the double integral as shown in Slide 14.12. Operator P operates
only on G and from Eq. 14.12.2, we obtain that Green’s function is definitely a solution of this
system.

Slide 14.12: (Refer Slide Time: 45:10)

14.12.1 ----
--
14.12.2 ----
--

Let us now calculate the Green’s function for an example string as shown in Slide 14.13. Here,
we have an impulsive loading at time t = t̅ at a location x = x̅ and the solution is a Green’s
function and we are considering homogenous boundary conditions and zero initial conditions. To
determine the Green’s functions, we simplify things by taking the Laplace transform of this
equation. So, we define Laplace transform as in Eq. 14.13.1. The Laplace transform of equation

144
 of motion along with zero initial conditions and the given boundary conditions is shown in Slide
14.13. Thus, we get the boundary value problem (BVP) that we need to solve now.

Slide 14.13: (Refer Slide time: 47:43)

14.13.1-----
-

---- 14.13.2

Slide 14.14: (Refer Slide Time: 51:33)

---- 14.14.1

145
Now, we look for solution of BVP in the form as shown in Eq. 14.13.2. So, substituting the
solution form in the BVP and taking the inner product with sin nπx/l on both sides, we filter out
the nth coefficient as shown in Eq. 14.14.1. Thus, the solution in the Laplace domain is obtained.
Now we have to take inverse Laplace transform which can be done by the residue theorem. The
final solution is shown in Slide 14.14. In the final solution form, there is the Heaviside step
which comes from the causality which means that the solution does not exist before the time
instant the impulse is provided i.e., at t < t̅ , the solution is zero The solution is non zero only
when t > t̅ . Thus, the solution for any arbitrary forcing can be determined using Green’s function.
The Green’s function method is a very powerful method of approaching forced vibration
problems of continuous systems with arbitrary forcing.

To summarize, we have considered the general forcing of one-dimensional continuous system

and we have looked at the modal expansion technique and the Green’s function technique for
solving the forced vibration problem of one-dimensional system. With that, we conclude this
lecture.

Keywords: Forced Vibration with general forcing, Travelling Force, Modal Expansion,
Green’s Function, Homogeneous Boundary Conditions, Boundary Value Problem

146
 Vibrations of Structures
Prof. Anirvan DasGupta
Department of Mechanical Engineering
Indian Institute of Technology -- Kharagpur

Lecture – 15
Damping in Structures - I

When we consider a real-life structure excited by a certain disturbance or an initial condition, we

observe that the structure stops vibrating after an initial response. In our earlier discussions on
the Initial Value Problem, we have seen in the animations that the structure is continuously
vibrating in response to the initial conditions. But, we do not observe this in the real world.
Therefore, we assert that there lies a mechanism inside these structures or some kind of
interaction of the structure with the external world, which drains out the structure's energy. This
mechanism of dissipation of mechanical energy is known as damping. In this lecture, we are
going to look at this mechanism. We are going to make the model of vibrating structures more
realistic by including damping terms.

Slide 15.1: (Refer Slide Time: 02:04)

As, I have said that the damping is a mechanism of energy dissipation. But, the question arises as
to what happens to the dissipated energy? It gets converted into thermal energy, and this

147
conversion is an irreversible process. Most of the damping mechanisms except radiation
damping work on this principle. In radiation damping, the energy drains out of the system in the
form of sound propagating in the fluid medium in which the structure is placed. The energy gets
radiated even from the structure's support points as no support is ideal, i.e., to say rigid. There
will always be some flexibility in the supports, which will cause continuous energy radiation out
of the system. Thus, we have energy dissipation by conversion to thermal energy as a primary
mechanism along with energy lost due to radiation as a secondary mechanism.

Let us now see the different sources of damping. (Slide 15.1)

1. Internal friction or damping: Different layers of the vibrating structure undergo

differentials straining or motion, which dissipates energy. Note that we are considering a
one-dimensional structure, which negates the possibility of any layer in the structure.
However, we will have a phenomenological model based on the rates of stretching or
strain rate to model internal friction or dissipation.
2. External damping: The structure might interact with an external fluid that can provide
damping.
3. Radiation damping: Under this, the energy dissipates due to radiation, as discussed
earlier.

Let us now see the different categories of damping. (Slide 15.1)

1. Distributed damping: Internal friction or damping and external damping due to

surrounding fluid fall under the category of distributed damping.
2. Concentrated/lumped damping: When a lumped dashpot or damper is externally
attached at a point to the vibrating structure, we have concentrated or lumped damping.
An example of such damping is the Stockbridge damper in high tension cables. We can
also have concentrated or lumped damping at connections and joins like riveted
connections etc.

148
Slide 15.2: (Refer Slide Time: 08:09)

--- 15.2.1

--- 15.2.2

--- 15.2.3

Let us begin with a model for internal distributed damping (Slide 15.2). Here, we are going to
discuss the phenomenological model for internal damping. We take a fixed-free bar undergoing
axial vibrations. We recall that when we derived the equation of motion of such a bar, we
considered the constitutive relation σ(x,t) = E ε(x,t). As I mentioned that in order to model
internal damping for a one-dimensional system, we must consider the rate at which the straining
takes place. Thus, we modify the constitutive relation to include the strain rate as in (15.2.1)
where the factor dI = ηE is known as the loss factor, and we call this as the coefficient of internal
damping.

Using ε=u,x, we modify the stress-strain relationship as shown in Slide 15.2. We use uniform
cross-sectional area of the bar while deriving the equation of motion which comes out as (15.2.2)
where the additional term (from free undamped vibration equation) is the damping term.
Similarly, with external distributed damping, we obtain an addition term dE u,t in the equation of
damped vibration of a uniform bar as shown in (15.2.3). The boundary conditions for fixed-free
bar are also shown in Slide 15.2.

149
Slide 15.3: (Refer Slide Time: 13:14)

--- 15.3.1

--- 15.3.2

--- 15.3.3

Now, the equation (15.2.3) may be written in a compact form as (15.3.1), where we define a
damping operator as D[·] exactly the way we have previously defined stiffness operator K[·].
Now, let us check whether the damping terms in the equation of motion really lead to dissipation
of energy. Therefore, we multiple the whole equation with the velocity and integrate it over the
domain of the bar as in (15.3.2). We perform integration by parts with respect to the space
coordinate, and use boundary conditions along with zero velocity initial condition to obtain
(15.3.3).

Thus, we have an integral at the right hand side of (15.3.3) which is always positive provided dE
and η is positive. Since, the integral at RHS carries a negative sign, therefore it is always
negative. The integral at LHS is nothing but the total mechanical energy (E) of the bar. From
(15.3.3), we note that the rate of change of total mechanical energy is always negative which
implies that the energy always dissipates or drains out of the system. Thus, the mechanical
energy is always dissipated by the damping terms that we have considered.

Let us say, we have external forcing q(x,t) also which will give an additional term in (15.3.3) as
shown in (15.4.1). Let us try to understand the implication of a harmonic forcing. The energy of

150
initial disturbances dissipates or drains out of the system in time as indicated by (15.3.3) and we
are left with only a steady state solution due to the external forcing after a sufficiently large time.
Thus, a particular solution is generated because of an external forcing. Further, with periodic
forcing and at steady state, the energy change over one period must vanish. Thus, we can say that
the energy provided by the forcing over one cycle is equal to the energy dissipated by the
damping terms over one cycle as shown in (15.4.2). Thus, from here, one can estimate many
things. For a given harmonic force input, we can record the amount of energy that we are
supplying over one period then we will know how much energy is being dissipated which can be
used to model the internal dissipation and the amplitude of motion in certain cases as well.

Slide 15.4: (Refer Slide Time: 20:28)

--- 15.4.1

--- 15.4.2

Let us once again look at the damped system with certain boundary conditions in Slide 15.5. Let
us try to find a solution of the equation of motion. We discretize the system as we have done
before using the modal expansion. Note that Uk(x) is the kth eigenfunction of the undamped
eigenvalue problem. We substitute the assumed solution form in the equation and take an inner
product with the jth eigenfunction. Thus, we obtain the equation in terms of modal coordinate pj
as shown in (15.5.1) where djk is the matrix as shown in Slide 15.5. In general, there is no
guarantee that djk is diagonal. If this is not diagonal, then all the modes are coupled through this

151
damping term. Only under very special situations, this damping matrix will be completely
diagonal matrix.

Slide 15.5: (Refer Slide Time: 25:46)

--- 15.5.1

--- 15.5.2

--- 15.5.3

Let us look at that condition which will give a diagonal matrix [djk]. If the condition in (15.5.2) is
followed for all eigenfunctions, we see that the damping matrix is completely. It can also be
noted that Uk is an eigenfunction of the damping operator D[·] as well. One special choice of this
damping operator for which the matrix djk becomes diagonalized, is when the damping operator
is a linear combination of the inertia and the stiffness operator as shown in (15.5.3). Such a
damping operator is called the classical or proportional damping operator. Thus, we get the
equations completely decoupled when the damping operator is classical or proportional, which
can be easily solved.

Keywords: Internal dissipation, frictional damping, classical damping, radiation damping,

distributed damping, lumped damping, internal and external damping

152
 Vibrations of Structures
Prof. Anirvan DasGupta
Department of Mechanical Engineering
Indian Institute of Technology -- Kharagpur

Lecture – 16
Damping in Structures – II

In this lecture, we delineate the roles of internal and external damping terms in the equation of
motion. In order to understand this, let us consider a simple example of a fixed-fixed bar
undergoing axial vibrations as shown in Slide 16.1. The equation of motion and boundary
conditions for the given system are written as in Eq. 16.1.1

Slide 16.1: (Refer Slide Time: 00:19)

--- 16.1.1

--- 16.1.2

--- 16.1.3

--- 16.1.4

We substitute the eigenfunction expansion in Eq. 16.1.2 in the equation of motion and take the
inner product with sin kπx/l that filters out the kth coefficient in the expansion. Thus, we obtain a
discretized equation of motion as in Eq. 16.1.3 where the term containing dE is for the external
damping while the term containing η is for the internal damping. We now consider each kind of
damping present one at a time.

Let us say that only external damping is present. In that case, we obtain the equation of motion as
in Eq. 16.1.4 where we note that the damping factor ζk is inversely proportional to k which means

153
that for high values of k, the damping factor is low and vice-versa. It implies that the external
damping is inefficient in damping-out the higher modes

Slide 16.2: (Refer Slide Time: 06:52)

--- 16.2.1

Now, let us consider the internal damping case. In this case, the equation of motion looks like as
in Eq. 16.2.1 where we note that the damping factor ζk is directly proportional to k. Therefore, for
higher modes i.e., for higher values of k, we get better damping because of internal dissipation.
Thus, internal damping is more effective for higher modes while external damping is more
effective for lower modes of vibrations. To understand the reasoning behind this, we have to
look at the expression of, say for example the internal damping in Eq. 15.2.1. We have seen that
the internal damping is dependent on strain rate. As we go to higher modes, for solution form as
in Eq. 16.1.2, we obtain higher values of strain. This implies that for the higher value of k,
higher is the contribution from the strain (u,x) and strain rate term (u,xt). For this reason, higher
modes allow higher effectiveness of the internal damping.

Let us now consider the external damping. We see the first, second and third mode of the bar or
the string which appears as shown in the Slide 16.2. We observe that for higher modes, more are
the number of nodes in the given span and therefore the effective motion of the system in the
ambient fluid is smaller due to reduced effective length within two successive nodes. Thus, the
effective damping because of external damping reduces as we go to higher modes.

154
Slide 16.3: (Refer Slide Time: 12:57)

--- 16.3.1

--- 16.3.2
--- 16.3.3

--- 16.3.4

Let us now look at the case of a discrete or lumped damping. A discrete damping occurs when
we attach for example, an external dashpot to a continuous system at a particular point. For
example, a Stockbridge damper which is typically used in high tension cables. A Stockbridge
damper is attached to a particular point on the cable and that damps the vibrational energy of the
high tension cables. Lumped damping can also be found in riveted connections between two
structural elements. In that case damping is localized at the connection or joint. Let us now look
at an example.

We consider a bar in axial motion, and here we have a dashpot attached to one end of the bar as
shown in the Slide 16.3. We write the equation of motion and boundary conditions of this
problem as in Slide 16.3. Let us search a solution of the form as in Eq. 16.3.1. Note that the
damping is introduced in the boundary condition not in the equation of motion which we can
always do. Now, once we substitute this solution form in the equation of motion, we obtain an
eigenvalue problem as in Eq. 16.3.2.

The solution of the eigenvalue problem is obtained as in Eq. 16.3.3. When we use the boundary
conditions, we obtain Eq. 16.3.4 where parameter γ and a are defined. Note the speed of axial
waves in the bar is indicated by c. For nontrivial solution of B and C, the determinant of the
matrix in Eq. 16.3.4 must be zero which leads us to the characteristic equation in 16.4.1 which
we can solve for γ when a ≠ 1. When a = 1, there are no eigenvalues which implies that there are

155
no solutions of the very special form given in Eq. 16.3.1 but then there are solutions which we
have to understand in terms of wave propagation which we will keep for a later discussions.

Slide 16.4: (Refer Slide Time: 17:50)

--- 16.4.1

--- 16.4.2

--- 16.4.3

Slide 16.5: (Refer Slide Time: 21:28)

Let us now solve the characteristic equation when a ≠ 1. For solving that, we consider γ to be a
complex number of the form γ=α+iβ where α and β are obtained as shown in Eq. 16.4.2 and
16.4.3. We can also make following observations. At d = 0, we get the eigenvalues of a fixed-

156
free bar while at d→∞, we have a fixed-fixed bar which can also be checked from the solutions
of γ. In slide 16.5, a plot of the variations of α and β is shown which is also the locus of the
solution as the parameter a varies. When a = 0, we have zero damping which means it is a fixed
free bar. When a → ∞, d also goes to ∞ and we have a fixed-fixed bar. And when a goes to
unity, we have no solution of eigenvalues as shown in the plot. Thus, you can see that all the
modes have the same decay rate as alpha does not depend on k. And, there is a jump in the
solution when the value of a crosses one. Thus, we have considered one case of a bar with
boundary damping.

Slide 16.6: (Refer Slide Time: 23:03)

--- 16.6.1

--- 16.6.2

--- 16.6.3

Let us look at another example of a discrete damping. We consider a string with a concentrated
damper or a lumped damper at x = a as shown in Slide 16.6. The equation of motion for the
given system is written in the form as in Eq. 16.6.1. If we use the modal expansion of the form
(Eq. 16.6.2), substitute it in the equation of motion and take inner product with the jth
eigenfunction, we obtain the discretized equation as in Eq. 16.6.3. Now if we adjust a such that
the term ja/l is not an integer for all j, then we will find that all the modes are damped. Such a
damping situation is known as pervasive damping. If we put the damper at the middle of the
string, we will find that the second mode is never damped. So, the second mode, the fourth
mode, etc will not be damped. Thus, the damping is called non-pervasive.

157
Thus, in this lecture, we have studied two forms of damping namely internal and external. We
have studied few example problems depicting lumped or concentrated damping.

Keywords: Distributed damping, discrete damping, concentrated or lumped damping,

internal and external damping, pervasive damping

158
 Vibrations of Structures
Prof. Anirvan DasGupta
Department of Mechanical Engineering
Indian Institute of Technology - Kharagpur

Lecture - 17
Beam Models – I

In this lecture, we are going to initiate discussions on the vibrations of beams. We all have
studied beams in the course on mechanics. The beams as a structural element are so ubiquitous
that we find them everywhere. As we have seen in previous lectures, the strings are a one-
dimensional elastic continuum, and so the beams. Thus, a question comes how do we distinguish
between a string and a beam? A beam is a one-dimensional elastic continuum that can resist
or transmit bending moment and shear while a string cannot.

Slide 17.1: (Refer Slide Time: 01:56)

Now, for mathematically modeling the beams, we must make certain assumptions as follows
(Slide 17.1).
1. Regarding the material of the beam, we restrict ourselves to linear elastic material which
is also homogeneous and isotropic.
2. The beam is assumed to be slender that justifies an Euler-Bernoulli beam model.

159
 3. Euler-Bernoulli hypothesis: We take a section of a beam. In the section, there is
something called a neutral axis or neutral fiber. When the beam deflects, it takes up a
shape as shown in Slide 17.1. An Euler-Bernoulli hypothesis states that a plane section of
the beam before deflection which is also perpendicular to the neutral fiber or axis remains
plane and perpendicular to the neutral axis in the deformed configuration.
4. We assume a negligible shear strain in the beam. Another way of saying this is that the
beam is infinitely stiff in shear.
With these four assumptions, we begin modeling the beam as described below.

Slide 17.2: (Refer Slide Time: 07:40)

FBD --- 17.2.1

--- 17.2.2

Once again, let us draw the section of the beam as shown in Slide 17.2. We note that the neutral
axis remains unstrained in the deformed and undeformed configuration. Let us consider an
element which undergoes deformation as shown by the dark blue lines in the beam in Slide 17.2.
Let the depth of the beam be denoted by h. As the beam deforms, it takes the shape of a curve of
radius of the curvature ρ(x,t). Let us now draw the free body diagram (FBD) of this element as
shown in Slide 17.2. The element is inclined at an angle of θ from the vertical and is contained
by an angular measure of dθ as shown in Slide 17.2.

160
First, we find the strain in the fiber of the beam which is at a distance z from the neutral fiber as
in Eq. 17.2.1. The length of the arc in Eq. 17.2.1 is written as the product of radius and subtended
angle. Also, we use the following relation for curvature (1/ρ) and the field variable [which is also
the deflection of the neutral fiber, i.e., w(x,t)].

Since, we have assumed slope of the beam (w,x) to be much smaller than one, we drop higher
powers of the deflection or displacement variable w. It gives us the strain as given by Eq. 17.2.2.
Thus, we have the strain in terms of the deflection of the neutral axis.

Slide 17.3: (Refer Slide Time: 18:22)

--- 17.3.1

--- 17.3.2

--- 17.3.3

Now, once we have the expression of strain, we can bring in the constitutive relation which is the
Hooke's law for determining the stress in the fiber at a location x and at time t. Thus, the axial
stress in the axial fibers of the beam at a location z measured from the neutral axis is obtained as
in Eq. 17.3.1. We note that the stress is linear in z. Let us now consider a cross section of this
beam. The position of a plane z is taken as positive measuring upwards. We want to find out the
moment caused due to the stress distribution on the cross section of the beam as shown in the
figure in Slide 17.3. Thus, the net moment about the neutral axis is written as in Eq. 17.3.2 where
dA is a little elemental area on this cross section. Thus, we obtain the moment acting on the

161
cross-section as in Eq. 17.3.3. Note the definition of second moment of area about neutral axis in
the equation.

Slide 17.4: (Refer Slide Time: 24:42)

--- 17.4.1

--- 17.4.3
--- 17.4.2

Let us now, once again look at the free body diagram of this little element of the beam as shown
in Slide 17.4 where we introduce the interaction forces. We have the shear forces (SF) and
bending moments (BM) at both the faces. The direction of SF and BM in the FBD is taken as
positive. We are going to follow this sign convention throughout this topic. Now, we write down
the equation of transverse dynamics of this little element. Using Newton's second law, we obtain
the Eq. 17.4.1 where ρ is the density of the material and A is the cross-sectional area at the
location x. The smallness of the angle θ is considered while writing Eq. 17.4.1.

Let us now write the equation of rotational dynamics. If ρ is the density of the material and I is
the second moment of the area about the neutral axis, then ρI gives us the moment of inertia of
the element per unit length about the axis which is perpendicular to the plane of the paper.
Multiplying it by dx will give us the mass moment of inertia of the element. Thus, we apply
Newton’s second law for moments which gives us the equation of the rotational dynamics as in
Eq. 17.4.2. The Newton’s second law is applied as follows. The product of mass moment of

162
inertia and angular acceleration (angular inertia) is equated to the sum of all moments about the
center of mass of the element. Here, the higher order term like dVdx/2 is dropped. Using slope of
the beam as w,x = tanθ ≈ sinθ ≈ θ, we write θ,tt = w,xtt. Here, once again, we have used the
smallness of the angle θ. Thus, we obtain the rotational dynamics as in Eq. 17.4.3.

Note that the bending moment M is already represented in terms of the field variable w(x,t) in
Eq. 17.3.2. The shear force V is determined from the equations of equilibrium. Now, we
eliminate the shear force between the Eqs. 17.4.1 and 17.4.3, which gives us the final equation of
motion of the beam as in Eq. 17.5.1. Note the flexure term, the rotary inertia term and the normal
transverse inertia or translation inertia term in the equation of motion. This model of the beam is
known as the Rayleigh beam model.

Slide 17.5: (Refer Slide Time: 33:47)

--- 17.5.1

For very slender beams, the rotary inertia term can be neglected and in that case, this is known as
the Euler-Bernoulli beam model. Now, when we have a forcing present then instead of a zero on
the right hand side of the equation, we have the force distribution function. Next, we are going to
discuss the variational formulation for beam dynamics. As we have seen before that the
variational formulations gives us a very powerful alternative way of deriving the equation, which

163
not only gives an equation of motion but also the possible boundary conditions. The variational
formulation also gives us some very powerful techniques for approximately solving or
discretizing the equations of motion as we have seen in our previous lectures.

Slide 17.6: (Refer Slide Time: 38:35)

--- 17.6.1

--- 17.6.2

Firstly, we write down the kinetic and potential energy expressions as given in Slide 17.6. The
kinetic energy of beam can be written as the sum of translational and rotational kinetic energy in
the beam as shown in Eq. 17.6.1. The potential energy of the beam is obtained as follows. The
potential energy per unit volume of a linearly elastic material is given as the volume integral of
the half of the product of stress and strain which upon simplification and using the definition of
second moment of area, gives as written in Eq. 17.6.2. The expressions of stress and strain are
taken from Eq. 17.2.2 and 17.3.1.

Now, we derive the equation of motion using the Hamilton's principle as shown in Slide 17.7.
We now take the variations after identifying the terms to be integrated with respect to time and
space. Using the method adopted earlier for variational formulation, we obtain the equation of
motion and the boundary terms as shown in Slides 17.7 and 17.8. Let us look at the boundary
terms.

164
Slide 17.7: (Refer Slide Time: 42:52)

Slide 17.8: (Refer Slide Time: 50:44)

Here, we recognize that the bending moment (EIw,xx) is zero or the conjugate of the bending
moment, which is the angle or the slope (w,x), must be zero at any of the end. Further, the shear
force (EIw,xxx) is zero or the conjugate of the shear force, which is the displacement (w) must be
zero at any of the end. Here, the shear force has rotary inertia contribution also, which can be

165
dropped for Euler-Bernoulli beam model. The conjugate of a quantity implies the quantity,
which when multiplied gives us the energy or work. Thus, we obtain the geometric or essential
boundary conditions and dynamic boundary conditions as shown in Slide 17.8.

Slide 17.9: (Refer Slide Time: 54:34)

Simply-supported beam

Cantilever beam

Now, let us quickly look at an example (Slide 17.9). In the first example, we have a simply
supported beam. The boundary conditions for this beam are written in Slide 17.9. Here, bending
moment and deflection at the ends are specified. In the second example, we have a cantilever
beam. Here, the boundary conditions at the fixed end would be on deflection and the slope, while
at the free end, it would be on the shear force and the bending moment as given in Slide 17.9.

In this lecture, we have discussed transverse dynamics of slender beams. We derived equation of
motion for the Rayleigh beam and the Euler-Bernoulli beam and we have looked at the
variational formations from where we derive the equations of motion along with the possible
boundary conditions.

Keywords: Beams, Euler-Bernoulli beam, Rayleigh beam, simply-supported beam,

cantilever beam

166
 Vibrations of Structures
Prof. Anirvan DasGupta
Department of Mechanical Engineering
Indian Institute of Technology - Kharagpur

Lecture - 18
Beam Models – II

In this lecture, we are going to continue our discussions on beam models that we had started in
lecture-17 where we had looked at two beam models under the following assumption. The shear
strain in the beam is negligible or in other words, the beam is infinitely stiff in shear. This
assumption is valid if the beam is slender, that means its transverse dimensions are much smaller
than the length of the beam. For a slender beam, shear may be neglected and thus, the Euler-
Bernoulli hypothesis holds good. We recall the Euler-Bernoulli hypothesis as the one which says
that the plane section perpendicular to the neutral fibers before deformation remains plane and
perpendicular to the neutral fibers after deformation. However, when the beam becomes thick or
its slenderness ratio goes on reducing so that the beam dimension increases in the transverse
direction than that in the longitudinal direction, the shear effect can no more be neglected.

Slide 18.1: (Refer Slide Time: 03:34)

Pure -------------

167
Therefore, in this lecture, we are going to look at a model which incorporates the shear in a very
interesting manner. In the Euler-Bernoulli beam model, we noted that the shear force was
introduced in the equilibrium equations. It was not a quantity derived from the deformation of
the beam in terms of the material properties. In this lecture, we are going to study a more
advanced beam model that goes by the name of Timoshenko beam model. Let us now begin the
modelling of the Timoshenko beam. The assumptions made for the Euler-Bernoulli beam
hypothesis that the slopes are small, the material is linear elastic, homogeneous and isotropic, do
hold for Timoshenko model as well with the only exception of shear in the model.

Let us now see the basic difference between the model that we discussed in the previous lecture
and what we are going to discuss today. As we can see in Slide 18.1, an undeformed element of
the beam undergoes flexure or pure bending. In Euler-Bernoulli beam model, we have discussed
the case of pure bending. In pure bending, the cross section maintains the orthogonality with the
neutral fibers. Also, the slope of the neutral axis ψ(x, t) in the deformed configuration is obtained
in terms of the transverse deflection as mentioned in Slide 18.1. Let us now look at the case of
simple shear as shown in Slide 18.2. The simple shear is a volume preserving deformation where
the angle θ(x, t) is the shear strain. It can be seen here that the slope of the neutral axis is nothing
but this shear strain.

Slide 18.2: (Refer Slide Time: 04:51)

168
When we put ‘pure bending’ and ‘simple shear’ together, we obtain the Timoshenko beam model
as shown in Slide 18.3. Here, we can see that the shear does not change the angle of the section
that is ψ(x, t). The vertical section remains vertical after shear deformation. However, the slope
of the neutral axis does change by angle θ(x, t) due to shear deformation.

Slide 18.3: (Refer Slide Time: 05:53)

Thus, the net slope of the neutral axis becomes the sum of angle due to pure bending ψ(x, t) and
the shear strain θ(x, t). Thus, in this model, we can consider any two of the three variables as our
field variables. Here, we will choose w(x, t) and by ψ(x, t) as our field variables as mentioned in
Slide 18.4. Thus, w(x, t) is the transverse deflection and ψ(x, t) is the flexure angle. Let us now
start representing our deformation in terms of these field variables. First, we obtain the fiber
strain in the longitudinal direction of the beam as explained in Slide 18.4. Furthermore, using the
longitudinal strain, we obtain the longitudinal stress in the fiber using Hooke’s law. Now, we
obtain the bending moment (in the plane of bending) produced due to the axial stress as Eq.
18.5.1. The shear stress and shear force due to the shear strain is also derived in Slide 18.5. Note
the expression of shear strain in Slide 18.5. We recall from mechanics of beams that the shear
stress is not uniformly distributed over the cross section. Thus, multiplying the shear stress with
total area of cross section is actually going to overestimate the shear force. In reality, the shear

169
force will be less than the product of the shear stress and the area of cross section. Therefore, as a
remedy to this problem and to keep the formulation very simple, we introduce what is known as
a shear correction factor. We calculate a corrected area As that is obtained as the actual area of
cross section of the beam A divided by a factor κ where κ is known as the shear correction factor.

Slide 18.4: (Refer Slide Time: 07:24)

Slide 18.5: (Refer Slide Time: 12:34)

----(18.5.1)

(18.5.2) ------

170
The shear correction factor typically takes on values of about 1.2 for rectangular cross section,
1.11 for circular cross section and about 2 - 2.4 for beams with I cross section. Thus, we obtain
the expression of our shear force as in Eq. 18.5.2.

Slide 18.6: (Refer Slide Time: 19:41)

----(18.6.1)

----(18.6.2)

Now, we draw the free body diagram of this beam element, an exaggerated view, in Slide 18.6.
Thus, we write the transverse dynamics for this element as Eq. 18.6.1 where ρ is the density of
the beam material, A is the area of the cross section, then ρA is the mass per unit length of the
beam. Now, we write the equation of rotational dynamics of the beam as in Eq. 18.6.2. Thus, the
equations of motion, so obtained, have two field variables w(x, t) and ψ(x, t). One can easily note
here that these equations are coupled partial differential equations. Now, here using both the
above equations, we obtain a single equation in w(x, t) as in Eq. 18.7.1. Here, we note that the
differential equation of motion has fourth derivative with respect to each, time and space. It
implies that we need four initial conditions and four boundary conditions to solve this equation.
Ironically, we do not have four initial conditions. Therefore, we need to solve this equation with
a little care. Here, we have to substitute the general solution of the Eq. 18.7.1 into Eqs. 18.6.1
and 18.6.2 to find additional constraints. Later, we will use Eq. 18.7.1 to determine the
dispersion relation for the Timoshenko beam.

171
Slide 18.7: (Refer Slide Time: 28:13)

----(18.7.1)

Now, we will discuss the variational approach for deriving the equation of motion of
Timoshenko beam. As we know that the variational approach yields the boundary conditions
directly that we have not discussed as yet.

Slide 18.8: (Refer Slide Time: 34:59)

172
For the variational formulation, we write down the kinetic (T) and potential energy (V)
expressions as in Slide 18.8. Note that the kinetic energy is the sum of translational and
rotational kinetic energy of the beam. Similarly, the potential energy is the sum of PE due to
flexure and shear in the beam.

Slide 18.9: (Refer Slide Time: 39:04)

Now, we apply the Hamilton's principle as explained in Slide 18.9 which gives us the differential
equation of motion and the boundary conditions. The boundary conditions at x=0 and x=l is
tabulated in Slide 18.10. Note the natural and geometric boundary terms in Slide 18.10.

As we have discussed in the beginning of the lecture, the Timoshenko beam assumes that the
cross section which was plane before deflection remains plane after deflection, though it need
not remain perpendicular to the neutral fiber. This is a kind of limitation of Timoshenko beam.
But then, we can very easily relax this by introducing more complex functions for the deflection
field. Therefore, let us review these complex field variables and see what can be done. In Slide
18.11, the difference between the Euler-Bernoulli, the Timoshenko and higher order beam theory
is pictured. In Euler-Bernoulli beam, the plane sections will remain plane as well as
perpendicular to the neutral axis. In the Timoshenko theory, the plane section remains plane but

173
now it need not be perpendicular to the neutral axis. In the higher order theories, the plane
section need not at all remain plane. In this model, we get what is called warping of the cross
section. Thus, if we want to have higher order theories for the vibration of beams, then we have
to relax the flatness condition of the cross section.

Slide 18.10: (Refer Slide Time: 47:13)

Slide 18.11: (Refer Slide Time: 51:29)

174
Slide 18.12: (Refer Slide Time: 52:56)

Slide 18.13: (Refer Slide Time: 54:00)

Thus, for introducing the warping of cross section, we write the kinematics of deformation in the
way as shown in Slide 18.12. Here, U(x, z, t) is the deflection in the axial direction, which is
being represented in terms of functions ψ0, ψ1, ψ2 etc. and expanded in terms of the variable z

175
which is the location of a fiber to be analyzed from the neutral axis. Similarly, the transverse
deflection is also expanded in powers of z, where we introduce the field variables as w0, w1, w2,
etc. Note, that the function ψ0 introduces the stretch of the middle plane.

For this kind of deformation field, we can write down the strains as shown in Slide 18.13. Using
these expressions of strain, we can write down the kinetic and potential energy as well as shown
in Slide 18.13. Now, once we have these expressions, we can follow the variational principle to
derive the equation of motion. Thus, by taking different kinds of expansion for the field variable
U(x, z, t) and W(x, z, t), we can devise higher order theories for beams.

To summarize, in this lecture, we have looked at the Timoshenko model which uses shear in the
beam that is valid for thick beams. Also., we have briefly looked at how we can devise higher
order beam theories.

Keywords: Beams, Euler-Bernoulli beam, Timoshenko beam, Pure bending, Variational

Principle, Higher order beam theories, Thick beams, Shear correction factor

176
 Vibrations of Structures
Prof. Anirvan DasGupta
Department of Mechanical Engineering
Indian Institute of Technology – Kharagpur

Lecture - 19
Modal Analysis of Beams

In this lecture, we are going to discuss the modal analysis of beams. In the previous lectures, we
have been discussing about various mathematical models of beams. But, here, we are going to
discuss the modal analysis that essentially means to determine the eigenfrequencies and the
modes of vibrations of beams. We will begin with the simple case of a uniform Euler–Bernoulli
beam.

Slide 19.1: (Refer Slide Time: 01:11)

---- (19.1.1)
---- (19.1.2)

A uniform Euler–Bernoulli beam is governed by a differential equation as Eq. 19.1.1, along with
the boundary conditions as shown in Slide 19.1. Note that, here, we consider a simply supported
Euler–Bernoulli beam whose boundary conditions are mentioned. Now, we search for a solution
of the type as in Eq. 19.1.2 which is separated in space and time. When, we substitute this in the
differential equation of motion, and after some simplifications, we obtain what is referred to as
Eigenvalue Problem (EVP) as shown in Slide 19.1.

177
Slide 19.2: (Refer Slide Time: 06:38)

---- (19.2.1)
---- (19.2.2)

---- (19.2.3)
---- (19.2.4)
---- (19.2.5)

To solve the EVP, we consider a solution of the form as in Eq. 19.2.1. When we substitute this
solution form in the differential equation of EVP, after some simplifications, we obtain the
constant as in Eq. 19.2.2 that is related to the circular eigenfrequency. Note that the constant β
(with tilde) will have four solutions which are taken as ±β and ±iβ, where β is defined as in Eq.
19.2.3. Thus, the general solution of W(x) is obtained as in Eq. 19.2.4 where A1, A2, A3, and A4
are arbitrary constants, possibly complex numbers.

Slide 19.3: (Refer Slide Time: 06:38)

---- (19.3.1)

178
Now, we can also write the general solution of W(x) in terms of hyperbolic and trigonometric
functions as in Eq. 19.2.5 where B1, B2, B3, and B4 are real constants. Now, the general solution
must satisfy the boundaries conditions as explained in Slides 19.2 and 19.3. For non-trivial
solutions of B2, B4, we obtain the characteristic equation for the simply supported Euler–
Bernoulli beam as shown in Slide 19.3. Now, in the characteristic equation, sinhβl = 0 at β = 0,
that will give us the trivial solution. Therefore, for non-zero values of β, we obtain the condition,
sinβl = 0 that gives indexed values of β as in Eq. 19.3.1. From Eqs. 19.2.2 and 19.3.1, we obtain
the circular natural frequencies of a simply supported Euler–Bernoulli beam as in Eq. 19.4.1.

Slide 19.4: (Refer Slide Time: 19:07)

---- (19.4.1)

---- (19.4.2)

---- (19.4.3)

Now, if we look at the differential equation of the Rayleigh beam model (Slide 19.4), and
follow the steps similar to what we have just now done, then, we will get the circular natural
frequencies of the Rayleigh beam as in Eq. 19.4.2. Note that, we have an additional factor in the
circular natural frequency of Rayleigh beam model as compared to that of the Euler-Bernoulli
beam model. Let us now analyze this. We define a slenderness ratio as length to the Radius of
Gyration which indicates how slender the beam is. A higher slenderness ratio would mean a
larger length as compared to its lateral dimensions. It should be noted that a very slender beam
will have higher slenderness ratio. For a very slender beams (Sr >> 1), the circular natural

179
frequencies of the Rayleigh beam model should match with the Euler–Bernoulli beam as can be
seen in Eqs. 19.4.1 and 19.4.2. But, for higher modes (n >> 1), the frequencies for both the
models are going to be different as shown in Slide 19.5.

Slide 19.5: (Refer Slide Time: 26:10)

In slide 19.5, as we can see that at lower modes, the nondimensional frequencies for both the
models are very close but as we go to higher and higher modes, the difference is appreciable.
Here, one can easily note that the source of the additional term in the eigenfrequency of the
Rayleigh beam model is the rotary inertia. And, the rotary inertia term is effective for higher
modes. We, now obtain the eigenfunction for both the beam models (working on Slide 19.3) as
in Eq. 19.4.3, that look like as shown in Slide 19.6.

In slide 19.6, we note that the third mode has two nodes, second mode has one node and the first
mode does not have any node over the span. Interestingly, it is same as that of a string. We
observe that at higher modes, the beam curvature is also higher which implies more rotation of
the beam element considered. Thus, we get the higher effect of rotational inertia term in the
equation of motion. Moreover, at higher modes (n → ∞), we can easily see that the Rayleigh
beam’s eigenfrequencies become proportional to n, whereas for the Euler–Bernoulli beam, the

180
eigenfrequency increases as n2 as you we note in Eq. 19.4.1.

Slide 19.6: (Refer Slide Time: 28:31)

Slide 19.7: (Refer Slide Time: 33:17)

---- (19.7.1)

---- (19.7.2)

Let us now, discuss a uniform cantilever beam as discussed in Slide 19.7. Note the boundary

181
conditions in this case in Slide 19.7. At the built-in end, we have zero displacement and slope
conditions (Geometric BC), whereas, at the free end, we have the bending moment and the shear
force as zero. The equation of motion remains the same as considered previously. Here, once
again, we search for separable solutions and therefore, we consider a solution form in space and
time as in Eq. 19.7.1. Now, we substitute the solution form in the differential equation of the
EVP as discussed in Slide 19.7. Thus, we obtain the characteristic equation for the cantilever
beam as in Eq. 19.7.2 that is a transcendental equation which we have to solve for β numerically.
We have plotted these functions, appearing in the characteristic equation, after some
rearrangement, in Slide 19.8. The unfilled circles in the plot gives us the eigenfrequency for the
cantilever beam.

Slide 19.8: (Refer Slide Time: 41:42)

In this plot, we have plotted functions namely, cos z and -1/cosh z with z = βl, and wherever they
match, that gives us the solution of z which is the solution of beta l. Here, we observe that as we
go for higher modes, the function -1/cosh z asymptotically reduces to zero. Thus, at higher
modes, the characteristic equation approximates to cosβl ≈ 0. The solution of this approximated
characteristic equation is written in Eq. 19.9.1 where we have added en which indicates the
correction/deviation in the solution due to the approximation/simplification.

182
Slide 19.9: (Refer Slide Time: 42:47)

---- (19.9.1)

For different modes, we calculate the correction (deviation) as shown in Slide 19.9. We can see
that the deviation falls rapidly. This e n - the correction is rapidly diminishing. Thus, for higher
modes, it suffices to use en = 0. Thus, the circle natural frequencies are obtained as in Slide 19.9.

Slide 19.10: (Refer Slide Time: 48:56)

183
We have obtained the eigenfunction, as we did for earlier beam kind in Slide 19.10. The
eigenfunctions for different modes are shown graphically in Slide 19.11.

Slide 19.11: (Refer Slide Time: 52:30)

To conclude this lecture, we have discussed the model analysis of beams that is essentially
solving an eigenvalue problem. We have looked at two examples of beam dynamics, simply
supported Euler–Bernoulli beam and a simply supported Rayleigh beam where we have
understood the effect of rotary inertia on the circular natural frequencies of the Rayleigh beam.
And, finally, we have looked at this Euler–Bernoulli Cantilever beam.

Keywords: Beams, Euler-Bernoulli beam, Rayleigh beam, Simply Supported Beam,

Cantilever Beam, Modal Analysis, Eigenfrequency, Eigenfunction

184
 Vibrations of Structures
Prof. Anirvan DasGupta
Department of Mechanical Engineering
Indian Institute of Technology – Kharagpur

Lecture - 20
Application of Modal Solution

In the previous lecture, we have discussed the modal analysis of beams. We have learnt that
the modal analysis is nothing but solving an eigenvalue problem and as a result, we obtained
the circular eigenfrequencies or the natural frequencies of the beam. Furthermore, we
obtained the eigenfunctions that define the modes of vibration of the beam. Regarding the
eigenfunctions, we have discussed it earlier that the eigenfunctions form a complete basis for
the shape of the beam.

Slide 20.1: (Refer Slide Time: 01:22)

---- (20.1.1)

What we, essentially mean by a complete basis? Let us understand it once again. Suppose we
have infinitely many eigenfunctions as we have seen. In slide 20.1, I have drawn only three
of them as a 3-dimensional space and with a little stretch of imagination, one can think of this
as an infinite dimensional space with each axis labelled with one eigenfunction. Any point in
this infinite dimensional space with coordinates α1, along W1(x), α2 along W2(x), α3 along
W3(x), etc describes the shape of the beam. Thus, any shape of the beam may be represented
as a linear combination of these eigenfunctions. Now, when we say that this forms a complete

185
basis, it means that any possible shape of the beam can be represented using these
eigenfunctions.

Thus, the eigenfunctions give us a good way of representing solutions. In general, the
coordinates α1, α2, α3 and so on, may be functions of time and in that case, the motion of the
point in the space of eigenfunctions may be represented through an expansion form as in Eq.
20.1.1, where αk is the functions of time. Thus, they form what are known as modal
coordinates. The space made by eigenfunctions is known as the modal space or the
configuration space. Now, this fact that any shape or any dynamical shape can be
represented through the expansion in Eq. 20.1.1 allows us to solve a number of problems
related to vibrations of beams. We have seen the similar application in case of strings and this
is true in general for any continuous systems. Therefore, let us look at two such examples, in
today’s lecture.

Slide 20.2: (Refer Slide Time: 06:34)

---- (20.2.1)

(20.2.3)
---- (20.2.2)

---- (20.2.4)

As a first example, we look at an initial value problem for a beam. We consider a simply
supported Euler–Bernoulli beam that is initially loaded with a static force, let us say F at the
mid-span as shown in Slide 20.2. Thus, the problem that we are going to address is as
follows. The beam is deflected under the action of a constant force F, applied at the centre of
the beam and at time t = 0, this force is switched off. At the instant, this force is switched off,
the beam is going to spring back. The equation of motion and boundary conditions for the

186
given dynamics problem are written as Eq. 20.2.1. The displacement initial condition for the
given beam would be the deflected shape of this beam due to the applied force that will be
obtained by solving the equation of statics as in Eq. 20.2.2. Thus, by solving this equation
along with the boundary conditions, we determine the initial shape of the beam. Further, we
assume zero initial velocity of the beam.

First, we write the solution of this problem as an expansion in terms of the eigenfunctions as
in Eq. 20.2.3. Using Eqs. 20.2.2 and 20.2.3, we obtain an Eq. 20.2.4 that we are going to
solve for constants Cj. The initial velocity condition will give Sj = 0 for all j. Now to solve for
Cj, we follow the standard procedure of multiplying both sides by a different eigenfunction
sin kπx/l and integrate over the domain of the problem. Thus, we take inner product of Eq.
20.2.4 with eigenfunction sin kπx/l and using the orthogonality property, we filter out the kth
term. Thus, we obtain the value of Ck as in Eq. 20.3.1.

Slide 20.3: (Refer Slide Time: 18:39)

---- (20.3.1)

---- (20.3.2)

Thus, we obtain the final solution of the initial value problem as in Eq. 20.3.2. The obtained
solution form is the function of space and time that one can animate to visualize the response
of the beam. Next, we look at the problem of a travelling force in Slide 20.4. Here, we
consider once again a simply supported Euler–Bernoulli beam carrying a force, which is
travelling with a speed v. Such problems are important in case of, let us say, bridges on which

187
we have travelling loads. The above problem is a simplified version of real-life situation
where we are considering a constant force travelling on Euler–Bernoulli beam.

Slide 20.4: (Refer Slide Time: 21:34)

---- (20.4.1)

---- (20.4.2)

The equation of motion of the chosen system along with the boundary conditions and initial
conditions is written as Eq. 20.4.1. The beam is considered undisturbed before the force
enters the span of the beam. Now, this is a forced vibration problem with a general forcing
and thus, we can write down the solution of this problem as the sum of homogeneous and the
particular solution. The homogeneous solution can be expanded in terms of the
eigenfunctions and represented as in Slide 20.4. Note that the beam is considered as a simply
supported beam, therefore the eigenfunctions are sin jπx/l. The particular solution can be
taken as an expansion in terms of these eigenfunctions along with the modal coordinates pj(t).
Note that the modal coordinates capture the system dynamics because of the forcing present.
Now, when we substitute this solution form in the equation of motion, we obtain Eq. 20.4.2.

We simplify Eq. 20.4.2 by taking inner product on both sides with sin kπx/l, and using the
orthogonality property, we filter out the kth term in this expansion. Thus, on simplification,
we obtain the differential equation in modal coordinate as Eq. 20.5.1 that defines the
dynamics of the kth modal coordinate. Thus, we obtain a forced vibration problem for the
discrete systems. Now, we note that all these modal coordinates (pk) are decoupled that can
be solved independently. We already know the general solution of such a system.

188
Slide 20.5: (Refer Slide Time: 29:18)

---- (20.5.1)

---- (20.5.2)
(20.5.3)

---- (20.5.4)

The general solution of pk is obtained in Slide 20.5. Here, we can note that the forcing term in
Eq. 20.5.1 is a harmonic forcing. The circular frequency of the harmonic forcing is named as
ωk. Now, there can be velocities for which this harmonic forcing equals the natural circular
frequency of the kth mode. We will call such forcing as a resonant forcing that is obtained in
Eq. 20.5.2. Thus, we find out the velocity that will send the kth mode into resonance.

Slide 20.6: (Refer Slide Time: 41:54)

189
Now, for simplicity, let us first assume that the given forcing is not resonant. For non-
resonant case, the solution for the kth modal coordinate is obtained as in Eq. 20.5.3. Now,
using the assumed initial conditions of the beam we obtain the solution for the response of the
beam under a force travelling at a non-resonant speed as in Eq. 20.5.4. Now, if we have the
speed of traveling force equal to one of the resonant speeds in Eq. 20.5.2, that is also called
critical speeds, the solution of the response gets modified. Let us now look at the non-
resonant solution at certain time instance for two speeds, one is a low speed and the other is a
high-speed transport over a beam (Slide 20.6). In slide 20.6, the figure shows snapshots at
certain time instance of a force travelling on a beam. Here, one can note how the beam
deflects at various time instances.

Slide 20.7: (Refer Slide Time: 42:19)

Similarly, the figure in Slide 20.7 shows the snapshots of dynamics when the speed is high.
Now, one can easily spot a difference between both the figures. The deflection of the beam
with low-speed traveling force is completely on the negative side, i.e., below the equilibrium
position. But, for the beam with high-speed traveling force, the beam goes above this
equilibrium line as well. In the simulation snapshot in Slide 20.8, the black spot indicates the
location of the force. The force is of course, acting on the beam and the shown response is the
exaggeration of the real deflection. Now there are two things that has to be remembered with,
when you see this animation. The first one is that this is a slow motion of what is happening.
The second thing is that once this force leaves the beam, the response of the beam is not

190
shown in this animation. This animation is looped and you see only the response of the beam
when the force is on the beam and travelling on it at a constant speed. One more thing to
notice in the response with high-velocity traveling force, which is different from that of the
low-velocity traveling force is that the deflection of the beam for the former case is much
smaller than that of the latter case. Since the force is travelling at a much higher speed, the
beam gets less time for deflection.

Slide 20.8: (Refer Slide Time: 42:55)

Now before we close this discussion, let us quickly look at one of the important properties of
the eigenfunctions, which is orthogonality. We have been using this property in all our
calculations, but let us now formally look at this property right from the equation of motion.
We consider a Rayleigh beam in Slide 20.9 whose equation of motion and boundary
conditions are detailed. Now, when we do modal analysis, we search for solutions of the form
as in Eq. 20.9.1. When we substitute this kind of solution in the equation of motion, we
obtain the description of the eigenvalue problem (EVP) as in Eq. 20.9.2. We can write the
EVP in a compact form as in Eq. 20.9.3, where operator M[.] and the operator K[.] are
defined as mentioned in Slide 20.9. We, now write the EVP for the jth mode that is going to
satisfy the equation of motion. Similarly, the kth mode is going to satisfy the same equation of
motion. Now, we multiply the first equation (jth mode) by Wk and the second equation (kth
mode) by Wj, subtract one from the other, and integrate over the domain of the problem.
Once, we do that and simplify with the boundary conditions, we obtain the orthogonality

191
condition for the beam as in Eq. 20.10.1. Thus, we see that the eigenfunctions are orthogonal
with respect to the inertia operator M[.].

Slide 20.9: (Refer Slide Time: 46:36)

---- (20.9.1)
---- (20.9.2)

---- (20.9.3)

Slide 20.10: (Refer Slide Time: 52:11)

---- (20.10.1)

To summarize the lecture, we have looked at some applications of the modal solution in
solving the initial value problem and a forced vibration problem. We have also looked at the
orthogonality condition of the eigenfunctions.

192
Keywords: Beams, Euler-Bernoulli beam, Rayleigh beam, Simply Supported Beam,
Modal Analysis, Eigenfunction, Orthogonality, Beam with traveling force, Initial value
problem

193
 Vibrations of Structures
Prof. Anirvan DasGupta
Department of Mechanical Engineering
Indian Institute of Technology – Kharagpur

Lecture – 21
Approximate Methods

In the previous lectures, we have been discussing about the modal analysis of beams and we
observed that even for simple configurations of beams or simple beam models, we have fairly
complicated eigenvalue problem at our disposal that we have to solve to perform the modal
analysis. Therefore, one would be interested in knowing if there are any approximate
methods which can quickly give us an estimate of the eigenfrequencies and the modes of
vibration of beams. We have discussed some of these approximate methods performing the
modal analysis for string and bars in previous lectures. In this lecture, we are going to look at
some of these approximate methods applied to beams.

Slide 21.1: (Refer Slide Time: 02:04)

Let me first enumerate the various methods that we have previously discussed in the context
of strings and bars (Slide 21.1). We will use them in the case of beam as well. To
recapitulate, we have used the Ritz method which require admissible functions and
variational formulation. In this method, we expand the field variable in terms of the
admissible functions and substitute it directly in the vibrational formulation of the problem.

194
In Galerkin method, we use comparison functions and work with the equation of motion
instead. These two methods have its own advantages and disadvantages. For example, in the
Ritz method, it is sometime tricky to consider non-potential or non-conservative forces, while
it is much easier with the Galerkin method. We know that the comparison functions have to
satisfy all the boundary conditions of the problem while this admissible function must satisfy
only the geometric boundary conditions. Thus, admissible functions can be very easily
constructed using polynomials or trigonometric functions or other such elementary functions
as opposed to comparison functions. It makes Ritz method easier to use as compared to
Galerkin method.

Slide 21.2: (Refer Slide Time: 07:07)

---- (21.2.1)

---- (21.2.2)

---- (21.2.4) ---- (21.2.3)

To begin with, let us look at an application of the Ritz method for certain problems in beams.
We consider vibrations of a Cantilever beam, whose BCs are discussed in Slide 21.2. Now,
we choose two polynomial functions as in Eq. 21.2.1, which satisfy the geometric boundary
conditions of the problem. Thus, we have the admissible functions ψi(x). The 2-term
expansion of the field variable in terms of the admissible functions is written as in Eq. 21.2.2
where a1 and a2 are the two temporal coordinates. A general expansion form for N admissible
functions is shown in Eq. 21.2.3. Next, we introduce this expansion in the Lagrangian of
Euler-Bernoulli beam (Eq. 21.2.4) which after performing the space integration and applying
the Hamilton’s principle, gives the Euler–Lagrange equations as shown in Slide 21.3.

195
Slide 21.3: (Refer Slide Time: 16:46)

---- (21.3.1)

Thus, we obtain the discretised equation for the Cantilever beam. Now, we perform the
standard modal analysis for this discretised system and we calculate the circular
eigenfrequencies and the modes of the eigenvectors. Let us first look at the eigenfrequencies
(ω1, ω2) as obtained in Slide 21.3. We obtain the exact values of these eigenfrequencies as
well that we have discussed in previous lectures. Now, we can make a comparison. We
observe that the fundamental circular eigenfrequency compares very well with the exact
while the second circular eigenfrequency as obtained from Ritz method is on the higher side.
We have discussed earlier that the Ritz method gives us an upper bound on the
eigenfrequency and thus, the exact eigenfrequency will always be less than the one obtained
from Ritz method.

Let us now look at the eigenfunctions. Corresponding to each eigenfrequency, we obtain the
eigenvectors (A1 and A2) and using these eigenvectors, we construct the eigenfunctions [W1(x)
and W2(x)] using the expansion defined earlier as discussed in Slide 21.3. It would be
interesting to see what happens to the natural BCs. We recall that the admissible functions are
not required to satisfy the natural boundary conditions of the problem. Using the obtained
eigenfunctions, we determine the bending moment and shear force BC at the free end of the
cantilever end as in Eq. 21.3.1. Here, we note that the natural BCs are being satisfied in a
limit. With increasing length of the beam, the term in Eq. 21.3.1 tends to zero quite fast.

196
Slide 21.4: (Refer Slide Time: 25:57)

---- (21.4.1)

We can try out the same problem with a greater number of terms in the expansion. In the
second example in Slide 21.4, we have considered 4 terms in the expansion. The admissible
functions considered for the problem are given in Eq. 21.4.1. Following the similar procedure
that we just discussed, we calculate the eigenfrequencies as shown in Slide 21.4. Once again,
we compare them with the exact ones. Here, we note that the approximate eigenfrequencies
are much closer to the exact solution.

And, in this case as well, if we calculate the first eigenfunction W1(x), we check how far the
natural boundary conditions are satisfied at the free end. Once again, we observe that they are
also going to zero quiet rapidly as indicated in Slide 21.4. Thus, we make the following
conclusion with respect to the Ritz method. As we increase the number of terms in the
expansion, we get more accurate solutions of the eigenfrequencies as well as the
eigenfunctions and they better satisfy the natural boundary conditions.

Let us now look at these eigenfunctions which I have plotted here in Slide 21.5. In the plots,
the solid line indicates the exact value, the chain dotted line indicates the approximate
eigenfunction with 2 term expansion, and the dashed line indicates the same with 4 term
expansion. Here, we see that the eigenfunctions tend to go close to the exact values as
number of terms in the expansion goes up.

197
Slide 21.5: (Refer Slide Time: 31:05)

Slide 21.6: (Refer Slide Time: 31:56)

A B

---- (21.6.1)

---- (21.6.2)

Let us now go over to a second example that is of a plain frame as shown in Slide 21.6. We
have a plain frame constructed out of 2 beams (arranged perpendicular to each other), which
are welded at point B. For simplicity, we consider that the lengths of both these beams is the
same and that is equal to AB = BC = l. Here, we have built in end at A and pinned end at C.
Now, these are essentially 2 beams which have a junction. So, we must treat them

198
accordingly. Let us consider the space coordinate x and the displacement field variable w1(x,
t) for the horizontal beam, and the space coordinate y and the field variable w2(y, t) for the
vertical beam. We intend to determine the eigenfrequencies and modes of vibration of this
frame. We write down the boundary conditions of the given system in Slide 21.6. The
conditions at the junction point B require more elaboration.
1. At point B corresponding to the horizontal beam can not admit any vertical
displacement provided the vertical beam is axially rigid. So, w1(x = l, t) = 0.
2. Using the similar reasoning at point B corresponding to the vertical beam, w2(y = l, t)
= 0.
3. The frame is welded at point B where both the beams maintain an angular relationship
of 90 degrees. Thus, we must have, w1, x (l, t) = w2, y (l, t).
4. From equilibrium considerations at the junction, we have w1, xx (l, t) = -w2, yy (l, t).

Thus, we have all the conditions required for this plain frame. We, now identify the
geometric boundary conditions as shown in red boxes in Slide 21.6 required for defining the
admissible functions. We consider the expansion of field variables as in Eqs. 21.6.1 and
21.6.2 that has been constructed using polynomials. Here, we note that the admissible
functions can be constructed in various other ways as well. Now, we write the Lagrangian for
the problem as in Slide 21.7.

Slide 21.7: (Refer Slide Time: 41:56)

199
We, now substitute the expansions in this Lagrangian and integrate out the space part and
then using the Hamilton’s principle, we obtain the discretised equation in the form as shown
in Slide 21.7 where M is the mass matrix and K is the stiffness matrices. We, once again
perform the modal analysis for this discretised system and obtain the eigenfrequencies. The
circular eigenfrequencies for first 2 modes of the system are mentioned in Slide 21.7. The
slide 21.8 shows the plots of first and second mode of vibration where we note that the
junction angle of 90 degrees has been maintained in both the cases.

Slide 21.8: (Refer Slide Time: 45:18)

Next, we look at the Timoshenko beam which is little more sophisticated model for beam.
We recall that the Timoshenko beam model additionally considers the shear deformation of
the beam. In this example, we will consider a simply supported Timoshenko beam (Slide
21.9). We write the Lagrangian of the Timoshenko beam in Eq. 21.9.1. In order to simplify it,
we use the definition of the shear modulus and the slenderness ratio as given in Eq. 21.9.2.
Thus, we obtain a simplified version of the Lagrangian. We expand the field variables of the
problem as shown in Slide 21.10. Finally, with the application of the variational formulation,
we obtain the discretized equations of motion. Now, if we perform the modal analysis of this
discretised equations, then the first non- dimensionalised circular eigenfrequency is obtained
as 0.612, second one is obtained as 2.087 and so on. The fifth one is obtained as 9.889, and
the sixth is obtained as 11.599. In slides 21.11 and 21.12, the mode of vibration
corresponding to the first, second and fifth, sixth eigenfrequency are shown. In the later

200
cases, we observe that there is hardly any transverse displacement or it is visibly very small,
at least. These are, in fact the shear modes of the Timoshenko beam and these frequencies are
substantially higher than those corresponding to the bending/flexure mode.

Slide 21.9: (Refer Slide Time: 45:18)

---- (21.9.1)

---- (21.9.2)

Slide 21.10: (Refer Slide Time: 50:27)

To summarize this lecture, we have looked at the approximate methods to discretise the
equation of motion of the beam. One can also use the Galerkin method in a similar manner, if

201
a suitable comparison function can be constructed. With the Ritz method, we have seen that
the admissible functions are very easy to compute and if we increase the number of terms in
an expansion, then we can also satisfy the natural boundary conditions that were, in fact
neglected while constructing the admissible functions.

Slide 21.11: (Refer Slide Time: 53:58)

Slide 21.12: (Refer Slide Time: 54:32)

Keywords: Timoshenko beam, Modal Analysis, Ritz Method, Galerkin Method,

Admissible Functions, Comparison Functions, Vibrations of Frames

202
 Vibrations of Structures
Prof. Anirvan DasGupta
Department of Mechanical Engineering
Indian Institute of Technology - Kharagpur

Lecture - 22
Topics in Beam Vibrations - I

Today, we are going to take up some special topics in Vibrations of beams. The first topic that
we are going to discuss is that of a tensioned beam. The question we are looking for the answer
of what happens when a uniform beam is subject to tensile loading? You may recall the
discussions on taut strings that we had in previous lectures. In earlier classes, we talked about the
guitar string which looks like more like a beam under very low tension. And as we make it taut,
it becomes a string. In the present discussion, we are going to look at this transition of behaviour,
from a beam like behaviour to a string like behavior of a string-beam like structure when we put
it under the tensile loading. Let us now look at the following example (Slide 22.1).

Slide 22.1: (Refer Slide Time: 01:30)

---- (22.1.1)

---- (22.1.2)

(22.1.3)

---- (22.1.4)

We consider a Euler Bernoulli beam which is subjected to a tensile loading say T as depicted in
Slide 22.1. The equation of motion for the chosen system is written as in Eq. 22.1.1 where the
first two terms are from the beam equation and the third term containing T is due to the tensile

203
loading present. In the case of a Rayleigh beam, we will have a further additional term due to
rotary inertia as shown in Eq. 22.1.2. Here, we will discuss the Rayleigh beam. Let us first Non-
dimensionalize the equation of motion that we accomplish by using the following parameter
relations as mentioned in Eq. 22.1.3. Thus, we get the non-dimensional equation of motion of the
form as in Eq. 22.1.4. It should be noted that the coefficients of the third and fourth term of Eq.
22.1.4 are recasted in the manner as described in Slide 22.1 where ε = T/EA is the axial strain in
the beam and sr = l/rg is known as the slenderness ratio.

Slide 22.2: (Refer Slide Time: 11:12)

---- (22.2.1)

Thus, we have the equation in non-dimensional form where I have dropped the tilde symbol for
simplicity as shown in Eq. 22.2.1. Let us now look at the third and fourth term of the equation. If
the beam is very-very slender, i.e., its length is much larger than the area property, it implies that
the slenderness ratio sr is very large. For sr >>1, the third and fourth terms of Eq. 22.2.1 can be
dropped in comparison to other terms. Thus, the equation of motion reduces to that of a string.
Thus, in that case, we expect the string like behavior. In other words, we can treat a very slender
beam under tension like a string. On the other hand, if the tension and consequently the axial
strain in the beam is very small, the flexure term in the equation (w,xxxx) might become
significant.

204
To sum up, the flexure (w,xxxx) and rotary inertia term (w,ttxx) is negligible in the case of slender
beams carrying high tension. Thus, the system can be modelled as a string. Whereas if tension is
very small then the flexural term may be significant and the system tend to behave more like a
beam. This is precisely what we observe in the guitar string which are extremely slender. One
can treat the guitar strings under tremendous amount of tension as the string. Another way to
look at the problem of guitar strings is that of the restoring force. The restoring force in the guitar
string is dominantly due to tension then due to its flexural stiffness.

Slide 22.3: (Refer Slide Time: 16:33)

Next, we consider a beam with non-homogeneous boundary condition. In various situations, we

find beams which are excited at the boundary. For example, we may have an absorber attached to
a vibrating structure for the purpose of vibration absorption. Here, we consider a cantilever beam
which is subjected to base excitation where the motion of the base is given by h(t) which is a
specified time varying function. Let us also consider this to be a Euler-Bernoulli beam.

The equation of motion and the boundary conditions for the given system are obtained as in Slide
22.3. Note the displacement of the boundary at x = 0 given by the function h(t). Thus, we have
one kind of non-homogeneous boundary condition in this case. We may have other kinds of non-

205
homogeneity as shown in Slide 22.3. For example, we may have a cantilever beam with the time-
varying moment M(t) at the free end. In this case, we will have non-zero moment BC at the free
end as shown in Slide 22.3. When we have forces or moments at the boundary, it can be easily
included in the equation of motion. Here, we first discuss the kind of system where we have
boundary excitation in terms of displacement. Our objective is to transform this system into a
one with homogeneous boundary conditions. One may ask why do we need this? Can we not
work with non-homogeneous BCs? To answer this question, we recall the modal analysis.
Whenever we have solved the eigenvalue problem, we have considered homogeneous boundary
conditions. To use the modal expansion theorem, we will require homogenized boundary
conditions, in the absence of which, the eigenfunctions won’t satisfy the boundary conditions
with present time-varying functions. Let us now see at a method of homogenizing the boundary
conditions for the chosen problem.

Slide 22.4: (Refer Slide Time: 26:18)

---- (22.4.1)

We try to make transformation of variables as explained in Slide 22.4. We transform the field
variable w(x, t) to another field variable defined as u(x, t) adding to an unknown function η(x)
times h(t) as in Eq. 22.4.1. This functional structure is, of course motivated by the structure of
the boundary conditions. We use this transformation in the equation of motion and the boundary
conditions as detailed in Slides 22.4 and 22.5. Using the homogeneous boundary conditions in

206
new field variable, and a polynomial form of the function η(x), we obtain η(x) = 1 as discussed in
Slide 22.5. Thus, the transformed equation of motion and the homogeneous BCs are obtained.

Slide 22.5: (Refer Slide Time: 26:18)

Thus, we have a forced vibration system along with homogeneous boundary conditions. Now,
we can use the eigenfunctions of the unforced system to solve the forced vibration problem.
Thus, we can easily convert any non-homogeneous BCs problem to a problem with
homogeneous BCs.

Next, we look at a very important phenomenon of damping in the vibrations of continuous

systems. We have discussed this before as well. We have two kinds of damping in continuous
systems. One is the external damping which is more common or very easy to see and then there
is an internal damping which is very hard to model and also very less understood. It is so because
it depends on the material constitution or the internal structure of the material. The source of the
internal damping is usually the friction between the molecular layers inside the material. But,
when we are considering a one-dimensional elastic continuum then there is no question of having
layers. Therefore, we must introduce this internal damping in phenomenological way. Thus, to

207
do that, we follow the Kelvin-Voigt model. Let us now look at this model for internal damping in
Slide 22.6.

Slide 22.6: (Refer Slide Time: 39:54)

---- (22.6.1)
---- (22.6.2)

---- (22.6.3)

In order to implement Kelvin-Voigt model, we modify the constitutive relation of the material as
given in Eq. 22.6.1 where stress due to the mechanical straining is added. Thus, the stress not
only depends on the amount of straining in the fibers but also on the rate at which the fibers are
being strained. Here, η is known as the Loss factor. It can be noted that the second term in the
constitutive relation is written in a special way in order to match it with the first term. If we use
this model to derive the equation of motion then we get an additional term in Eq. 22.6.2 which is
attributed to the internal damping.

Now to get the clear perception that the third term in Eq. 22.6.2 is really a damping term, we
multiply this whole equation with the velocity w,t and integrate it over the domain of the beam.
We integrate it by parts, and use the boundary conditions to simplify. Thus, we obtain the final
form as in Eq. 22.6.3 which indicates the rate of change of total energy. If we observe the nature
of the parameters involved in RHS of Eq. 22.6.3, we find that the mechanical energy reduces for
a positive loss factor.

208
Slide 22.7: (Refer Slide Time: 50:34)

---- (22.7.1)

---- (22.7.2)

Now, let us look at a simply supported Euler-Bernoulli beam with internal damping in Slide
22.7. Now, we have seen that for an undamped beam, the modal vibration is given by the
eigenfunction form as in Eq. 22.7.1. Here, we assume that the beam is vibrating in the nth mode.
If we substitute the modal solution in the damped equation, we obtain the equation of motion in
the form Eq. 22.7.2, where the damping factor ζn is found to be proportional to n2. Thus, for
higher mode of vibration, the damping factor will be much higher. It implies that the higher
modes get effectively damped because of internal damping. One can repeat this analysis with
external damping as well and one can easily show the lower modes are more effectively damped
then the higher modes with the external damping.

To summarize the lecture, here we have considered three topics from beam vibrations. Firstly,
we considered the analysis of a beam under tension followed by the problem of non-
homogeneous boundary conditions. Lastly, we considered the internal damping in beams.

Keywords: Beam under tension, Non-homogeneous boundary conditions, Internal

damping, External damping, Damping factor, Kelvin-Voigt model

209
 Vibrations of Structures
Prof. Anirvan DasGupta
Department of Mechanical Engineering
Indian Institute of Technology - Kharagpur

Lecture - 23
Topics in Beam Vibrations - II

Today we are going to discuss an important concept in beam vibrations. The notion of beam
stability comes when the beam is subjected to an axial load. We find in, for example a column,
that when it is subjected to an axial load/force, the column will fail by buckling at the load
exceeding a limiting value. This is one kind of axial loading that can happen. There is another
kind of axial loading that occurs in case of missiles. In the case of missiles, we have a jet ejecting
out from the rear end resulting in an axial force such that the force direction always maintains the
angular position. In fact, it is observed for a flexible missile system, the force of ejecting jet
maintains the angle. Whereas in the case of columns, the force remains vertical after deflection
that is usually considered small. Thus, there is a difference in these two axial forces that we are
going to discuss in this lecture.

Slide 23.1: (Refer Slide Time: 02:13)

--- 23.1.1

--- 23.1.2

--- 23.1.3

210
Now, when we say a beam with axial force, we are going to consider a column like loading. Let
us consider a simply supported Euler Bernoulli beam with axial force F as shown in Slide 23.1.
The beam configuration before and after deflection is shown in Slide 23.1. Note that the force
retains its direction in this case. The equation of motion and BCs for this beam with the axial
force is obtained as in Eq. 23.1.1 where first two terms are for the Euler-Bernoulli beam and the
third term is due to the axial force which is taken to be compressive. One may recall that in the
case of strings, the force is tensile and thus appears with the negative sign in the equation of
motion.

For this problem, we can now use the modal expansion. We know that the eigenfunctions of a
simply supported beam is given by sin nπx/l. Thus, we construct the solution as an expansion in
the form as shown in Eq. 23.1.2. We substitute the solution form in the equation of motion,
simplify it and use the orthogonality property of the eigenfunctions to obtain the dynamics of the
modal coordinates pj as in Eq. 23.1.3.

Slide 23.2: (Refer Slide Time: 10:40)

23.2.1 ---
-
23.2.2 ---
- --- 23.2.3

Now, we look for the solutions of pk(t) in the form pk(t) = Cest. When substitute it in Eq. 23.1.3,
we obtain the kth solution of s in terms of ωk as sk = ±iωk, where ωk is given by the expression in

211
Eq. 23.2.1 and k = 1, 2, 3.... Now, we make a little bit of simplification for Eq. 23.2.1 by some
redefinition as shown in Slide 23.2. Thus, using these redefinitions, we recast the eigenfrequency
ωk in the non-dimensional form (indicated with an overhead tilde sign) as in Eq. 23.2.2 where S
represents the load/axial force. We can obtain the value of axial force when (non-dimensional)
ωk→0 as shown in Slide 23.2. Thus, we get S = k2π2 at ωk→0 and it implies that the system has
lost stiffness. At k = 1, we obtain the axial force F that we know as the Euler buckling load. To
our amazement, here we note that the Euler Buckling load for a beam is arrived at through the
dynamic analysis.

Slide 23.3: (Refer Slide Time: 19:31)

Let us now see the effect of load S on the eigenfrequency ωk in Eq. 23.2.2. When we increase S
from 0, then ωk comes out to be a real number. Thus, the solution of pk becomes pk (t) ~
exp(±iωkt) which are harmonically varying. Now, when ωk goes to 0 then the function pk loses its
temporal nature and this state is referred to as neutral stability. Thus, we have reached the Euler
buckling load for k = 1, and the beam has lost its stiffness. Now, if the load is increased further
beyond the Euler-Buckling load, then ωk (at k = 1) will become imaginary which gives pk (t) ~
exp(±αt). Thus, we will have the solution containing exponentially increasing and decaying
terms. Now at a finite time, the exponentially increasing terms are going to dominate the total
solution. Therefore, even for a very small initial disturbance, the solution is going to have the

212
exponentially increasing nature as shown in the plot in Slide 23.2. This kind of behavior is
known as divergence instability. Thus, when the axial load exceeds the Euler buckling load for
the first mode (k = 1) than we observe the divergence instability and the beam buckles.

Let us now look at the variation of the non-dimensional eigenfrequency with the non-
dimensional axial load (S) in Slide 23.3. When the load is zero (S = 0), we have the first non-
dimensional circular natural frequency, and as the load increases, the circular natural frequency
goes to zero. The load corresponding to this point (S = Sc) is the buckling load or the critical
load.

Slide 23.4: (Refer Slide Time: 20:10)

+iω (non-dimensional)

-α +α

-iω (non-dimensional)

Now, let us look at the same variation in the root locus diagram in Slide 23.4 where s is the
coefficient of the exponential term in the solution pk(t) = Cest. Here, the real and the imaginary
part of s is plotted. We have already obtained s = ±iω when S < Sc. As we increase the load, the
roots go to zero (S = Sc), and once they reach zero then they move on the real axis. We have seen
that the real solution reads as s = ±α. The root on the positive side of the real axis, i.e., s = +α is
going to lead to instability. That is exactly what we have found in the solution in Eq. 23.2.3 and
the plot in Slide 23.2. Thus, an exponentially increasing solution leads to the divergence

213
instability. Thus, a beam with axial force will show divergence instability if the force exceeds
the Euler buckling load. This method can be used to estimate the buckling load of a column,
without destroying the column itself. If we want to devise a procedure of non-destructive
evaluation of the buckling load of a column experimentally, we can use this dynamic feature.
When we apply an axial force on the column and increase it, we observe the eigenfrequency
variation with the load that can be plotted (as in Slide 23.3). Now, we can extrapolate the curve
to obtain the intersection of the plotted curve with the x-axis (the axis of load) which is going to
give us the buckling load without actually destroying the column.

Slide 23.5: (Refer Slide Time: 23:51)

--- 23.5.1

Next let us look at the example of a beam with a follower force in Slide 23.5. Let us consider a
cantilever beam with a force at the free end which is an axial force, but then the force has this
property that when the beam deflects, the force maintains the same angle with the neutral fiber of
the beam at the free end. Such a force is known as a follower force. Thus, the force direction
follows the tangent to the neutral fiber or the center line of the beam. We have observed such
dynamics in flexible missiles, fluid carrying jets etc.

214
Now, we need to derive the equation of motion of this system. Here, we note that the force has a
transverse component as shown in Eq. 23.5.1 which is a non-potential force. It implies that the
force cannot be derived from a potential or it is not gradient of any potential. Therefore, to derive
the equation of motion of this system, we use the variational principle which appears to be a
safer way for this system since the dynamics of the system is a little tricky as we will see shortly.
Now, since we have a non-potential force, we have to use the Extended Hamilton's principle to
derive the equation of motion for this system as explained in Slides 23.5 and 23.6. We note the
virtual work done by the non-potential forces at the location x = l in Slide 23.5. Thus, we write
the Lagrangian where the potential energy because of the axial straining is considered. Here, the
sign of F is taken appropriately so that it indicates a compressive force.

Slide 23.6: (Refer Slide Time: 33:39)

Now, we work with the Extended Hamilton's principle in Slide 23.6 that, after the standard
integration procedure and some simplification, finally gives the equation of motion and the
boundary conditions. Now, it is interesting to note here that the equation of motion is same as
that of the beam with a normal axial load. Another interesting or surprising thing to note here is
that the boundary conditions are all homogeneous.

215
Slide 23.7: (Refer Slide Time: 43:07)

--- 23.7.1

--- 23.7.2

Slide 23.8: (Refer Slide Time: 47:52)

--- 23.8.1

Let us now look at the eigenvalue problem of the above systems. So, we look for a separable
solution of the form, let us say w(x, t) = W(x)est. Now, when we substitute this in the equation of
motion and BCs and simplify, we obtain the eigenvalue problem as shown in Slide 23.7. Now,
let us make some redefinitions here. Let us search for solutions of W(x) of the form W(x) = Deβx

216
which modifies the eigenvalue problem as Eqs. 23.7.1 and 23.7.2. Now, we are going to solve
Eq. 23.7.2 for β which gives two solutions β1 and β2 as shown in Slide 23.8. Thus, the solution of
W(x) is obtained as in Eq. 23.8.1. The boundary conditions give conditions (in Slide 23.8) that
can be recasted as the product of a matrix A and a vector of unknown coefficients Bi (in Slide
23.9). Note that the matrix A is function of non-dimensional eigenfrequency ω and non-
dimensional axial load S. For non-trivial solutions of the coefficients Bi, the determinant of
matrix A must vanish [Slide 23.9]. Now, one can solve the resulting equation of vanishing
determinant at a given load S, for the non-dimensional circular natural frequencies (ω with tilde
overhead) of the beam numerically.

Slide 23.9: (Refer Slide Time: 53:04)

The results of such a solution is shown in Slide 23.10. When the load is zero, the first two non-
dimensional circular natural frequencies are obtained and as the load is increased, ω2 decreases
while ω1 increases and they coalesce at the certain value of load. This value of load gives the
value of critical load, after which, the eigenfrequencies vanish. Here, once again, we look at the
root locus diagram in Slide 23.11 where the real and the imaginary part of s (exponent in solution
form w(x, t) in Slide 23.7) is plotted. We know that s = ±iω (with tilde over ω). Thus, when we
have the follower force magnitude less than the critical value (S<Sc), then we have pure

217
imaginary solutions as s = ±iω1 and s = ±iω2 as shown in the plot in Slide 23.11. Now, as the
load is increased, both the frequencies come together and coalesce at a certain value which is the
critical value of the load (S = Sc). Subsequently, the solution will also have real parts as well.
Thus, post critical load, the non-dimensional eigenfrequency solutions take the form as ω =
±α±iβ.

Slide 23.10: (Refer Slide Time: 53:57)

Slide 23.11: (Refer Slide Time: 54:34)

+iω2 (non-dimensional)

+iω1 (non-dimensional)

-iω1 (non-dimensional)

-iω2 (non-dimensional)

218
Thus, we have exponential term (e±βt) along with a fluctuating component (e±iαt) in the solution
as detailed in Slide 23.9. Thus, the solution will have an exponentially increasing envelope and
sinusoidally fluctuating displacement/amplitude/response with time as shown in the plot in Slide
23.9. This kind of behavior is known as Flutter Instability in beams with follower force.

Thus, in this lecture, we have discussed the dynamics of beams with normal axial force and
follower force.

Keywords: Beam with axial force, Beam with follower force, Buckling load, Divergence
instability, Flutter instability

219
 Vibrations of Structures
Prof. Anirvan DasGupta
Department of Mechanical Engineering
Indian Institute of Technology – Kharagpur

Lecture – 24
Dynamics of Curved Beams

Till now, we have looked at only straight beams. However, there are various places where we
find curved beams. For example, arches, bridges, and rings. In this lecture, we are going to
initiate some discussions on the dynamics of curved beams. Particularly, we are going to look
at the modelling aspects of a curved beam. Here, the first and the most fundamental
difference between a straight beam and a curved beam that we will note is the coupled
axial/circumferential - transverse/radial motion of the beam due to the curvature in the later
kind. Therefore, the curved beams can no longer be treated using only one field variable.
Instead, we must use two field variables where one will keep track of the circumferential or
the axial motion while the other one will define the radial motion of the beam element.

Slide 24.1: (Refer Slide Time: 01:58)

As with any beam theory, here also we are going to make the following assumptions to
simplify the modelling process (Slide 24.1).
1. The curvature of the beam is constant and the beam is planar.
2. The deflection in the beam is planar.

220
 3. The deflection is small compared to the thickness of the beam.
4. The thickness itself is small compared to the curvature of the curved beam.
5. The Euler-Bernoulli hypothesis holds which implies that there is no shearing effect
on the cross-section of the beam.

Slide 24.2: (Refer Slide Time: 05:09)

--- 24.2.1

--- 24.2.2

--- 24.2.3

Thus, under the above assumptions, we are going to model a curved beam as discussed in
Slide 24.2. For now, we are going to restrict ourselves to the case of a beam with constant
thickness. Let us consider a ring as shown in Slide 24.2. We consider a small element of
angular width dθ of this ring at an angular position θ and designate a radial direction with a
corresponding field variable w(θ, t) and an angular direction with a corresponding field
variable u(θ, t). A dashed circle of radius R is drawn inside the ring that is considered as the
neutral fibre. This neutral ring is assumed to be constant.

We are, now going to look at the deformation kinematics of this little element as discussed in
Slide 24.2. Length of any fiber of the small element before deformation, ds is obtained as in
Eq. 24.2.1 where z is the radial height of the fiber from the neutral fiber. This element
undergoes deformation in 2 steps namely axial and radial elongation. In axial
deformation/elongation (as shown in Slide 24.2), point P1 moves to P2. Note, the deflection in
the circumferential direction is indicated as u(θ, t). In radial/transverse deformation, the point
P2 moves to P3 as shown in the figure in Slide 24.2. which is captured by this field variable

221
w(θ, t). Thus, the angular position after deformation, θ’ is obtained as in Eq. 24.2.2. Thus, the
length of the fiber after deformation, ds’ is obtained as in Equation 24.2.3.

Slide 24.3: (Refer Slide Time: 17:05)

--- 24.3.1

--- 24.3.2

--- 24.3.3

Now, we calculate the strain in the fiber, ε as discussed in Slide 24.3 which is approximated
by considering z/R << 1 as shown in Eq. 24.3.1. Now using this strain, we use Hooke’s law to
obtain the stress, σ as detailed in Slide 24.3 where E is the Young’s Modulus of the beam
material. These stresses can now be integrated over the cross-sectional area of the beam to
obtain the force and moment resultants, N and M respectively.

Now, we draw the free body diagram of the little element of the ring as shown in the figure in
Slide 24.3. where we have the shear force V, the bending moment M, and the circumferential
forces N on both the faces. We now calculate the stress resultants for normal force and
bending moment as detailed in Eqs. 24.3.2 and 24.3.3. We do not consider shear force
resultant as we will see that it will be obtained from the equations of equilibrium.

Let us now start writing the equations of equilibrium as explained in Slide 24.4. We write the
circumferential and radial/transverse equilibrium equation as in Eqs. 24.4.1 and 24.4.2. Next,
we will look at the rotational dynamics of this small element. Here, we will neglect the
rotational inertia of this element, and therefore, this equation boils down to only moment
balance about the centre of mass of this element that is obtained as in Eq. 24.4.3. Now, we

222
substitute the outcome of Eq. 24.4.3 in Eqs. 24.4.1 and 24.4.2, and thus, we eliminate V. We
also replace the circumferential force resultant N from Eq. 24.3.2. Thus, we obtain the
equations of motion as shown in Slide 24.5.

Slide 24.4: (Refer Slide Time: 23:57)

--- 24.4.1

--- 24.4.2

--- 24.4.3

Slide 24.5: (Refer Slide Time: 30:43)

Next, we look at the variational formulation for the same problem. It will be useful when we
do calculations using approximate methods like Ritz method. Thus, we start with by writing

223
the kinetic energy, and potential energy expressions as shown in Slide 24.6. Here, we have
substituted the expression of strain, ε in the expression of potential energy, and in the process
of algebraic simplification and area integration, we obtained terms like second moment of
area as ∫A z2 dA = I. The area integral term containing unit power of z will vanish due to
neutral plane consideration.

Slide 24.6: (Refer Slide Time: 32:24)

--- 24.6.1

--- 24.6.2

Before we move further, let us make some simplifications using certain redefinitions as
shown in Eq. 24.6.1. We redefine and nondimensionalize time, circumferential displacement,
transverse or radial displacement, and the slenderness ratio. The radius of curvature of the
beam divided by the radius of gyration of the cross section reflects the slenderness of the
beam. We now use these expressions into the Lagrangian which eventually becomes as in Eq.
24.6.2. Now, we use the Hamilton's principle to derive the equation of motion as discussed in
slide 24.7.

Performing standard integration by parts, we obtain the equations of motion in Eq. 24.7.1 and
boundary conditions as detailed in Slide 24.8. Now once again, we note the coupling between
the circumferential and the radial displacement field variables in the equations of motion. The
equations in Eq. 24.7.1 are non-dimensionalised that makes it easy to look at the contribution
of various terms in the equation.

224
Slide 24.7: (Refer Slide Time: 40:45)

--- 24.7.1

For example, if the beam is too slender, which means the slenderness ratio is very high, then
the terms containing 1/sr2 will become insignificant. In that case the equations will get
simplify and we will have only three terms in each equation. But if the beam is not so slender,
these terms will significantly contribute in the dynamics. Note the natural and geometric
boundary conditions in Slide 24.8. For a curved beam in the shape of a complete ring, we will
have periodicity conditions in the field variables as mentioned in Slide 24.8.

Slide 24.8: (Refer Slide Time: 48:38)

225
Let us recapitulate what we have discussed in this lecture. Today, we initiated some
discussions on the dynamics of the curved beams, particularly on the beams of constant
curvature. We derived the equations of motion using Newtonian as well as the variational
formulation. In the equations of motion of curved beam dynamics, we observed that the
transverse and circumferential motions are coupled, whereas in the straight beams, we have
transverse motion only. In fact, for straight beams, we can treat the transverse and the axial
motion separately. Thus, we conclude this lecture.

Keywords: Curved beam, rings, coupled dynamics

226
 Vibrations of Structures
Prof. Anirvan DasGupta
Department of Mechanical Engineering
Indian Institute of Technology – Kharagpur

Lecture - 25
Vibrations of Rings and Arches – I

Today, we are going to discuss the vibrations of rings and arches. In the previous lecture, we
initiated some discussions on the dynamics of curved beams and beams with constant curvature
which are in one plane. Before we look into the vibrations of rings and arches, let us recapitulate
briefly what we discussed in the previous lecture.

Slide 25.1: (Refer Slide Time: 01:03)

Curved beams are found in various civil structures like bridges and arches (See Slide 25.1). In
the formulation discussed in the previous lecture, we observed a very important aspect in the
dynamics of curved beams, that is of the coupled axial and the transverse dynamics due to the
curvature of the beam. We made some simplifying assumptions when we formulated the
dynamics. We assumed that the beam is planar despite being curved in a plane. Further, we
assumed that the deflection is much smaller than the thickness of the beam and the thickness in

227
turn is much smaller than the curvature which was assumed to be a constant. We also assumed
that the Euler-Bernoulli hypothesis holds which implies that a cross section of the beam which
was initially flat and perpendicular to the neutral fiber remains flat and perpendicular to the
neutral fiber after deflection. Thus, we neglected shear effects that implies that the beam is
infinitely stiff in shear. Thus, with such considerations or assumptions, we derived the equation
of motion using the variational formulation.

Slide 25.2: (Refer Slide Time: 04:04)

In the variational formulation, we considered the Lagrangian which we wrote as the difference
between the kinetic and the potential energy as detailed once again in Slide 25.2. In the earlier
discussion, we also made simplifications based on certain redefinitions as mentioned in Slide
25.2. To this end, we carried out non-dimensionalization by introducing new terms. And, thus,
we derived the non-dimensionalised form of equation of motion as in Eqs. 25.3.1 and 25.3.2.

Today, we are going to first discuss the vibrations of rings. We consider a uniform ring with
radial and circumferential/tangential direction as shown in the figure in Slide 25.3. The boundary
conditions for a ring, as we discussed in the previous lecture, turns out to be periodicity
conditions on the field variables as in Eq. 25.3.3. Now, we are going to perform the modal

228
analysis of this system. To do that, we search for solutions of a separable form as shown in Eq.
25.3.4 where n can take value 0, 1…∞ to enforce the periodicity condition in Eq. 25.3.3.

Slide 25.3: (Refer Slide Time: 09:40)

--- 25.3.1
--- 25.3.2

--- 25.3.3

--- 25.3.4

If n = 0, the solution becomes independent of θ which implies axisymmetric modes admitted by

the ring. For non-zero values of n, we have non-axisymmetric modes. Now, we substitute the
solution form in the equations of motion. Upon simplification, the equations read as in Eq.
25.4.1. For non-trivial solutions of U and W in Eq. 25.4.1, the determinant of M must vanish as
discussed in Slide 25.4. This leads us to the characteristic equation in Eq. 25.4.2. Now, we solve
the characteristic equation for eigenfrequency ω and subsequently, we substitute the
eigenfrequency in two equations of motion and then solve for the eigenvectors U and W and
thus, we obtain the eigenfunctions. We can easily note here that the eigenfunctions will be
complex in nature. We have discussed already that when we have complex eigenfunctions, then
both the real and the imaginary parts of it can be the eigenfunctions. Thus, what we can conclude
is that for a given eigenfrequency, we can have more than one eigenfunction. This is referred to
as degeneracy. Let us now solve the characteristic equation for n = 0 and try to find out the
eigenfrequencies and the eigenfunctions which will characterize the corresponding mode of
vibration.

229
Slide 25.4: (Refer Slide Time: 16:02)

--- 25.4.1

--- 25.4.2

Slide 25.5: (Refer Slide Time: 12:22)

--- 25.5.1

For n = 0, the characteristic equation for axisymmetric modes reads as in Eq. 25.5.1. It would
imply ω = 0 and ±1. The eigenvectors responding to both the eigenfrequencies are discussed in
Slide 25.5. The eigenfrequency ω = 0 produces a motion along the circumferential direction of
the ring that can be seen as a rigid body mode of vibration that implies that the angular

230
momentum is conserved. Thus, we do not have any vibratory mode at ω = 0.

At ω = ±1, we observe from the corresponding eigenvector that there is no motion in the
circumferential direction. Here, in fact, the motion is in oscillatory mode which is only in the
radial direction. Thus, we have something known as a breathing mode of vibration. In this
mode, the ring expands and contracts alternatively.

Keywords: Curved beam, rings, coupled dynamics, rigid body mode of vibration, breathing
mode of vibration, degenracy

231
 Vibrations of Structures
Prof. Anirvan DasGupta
Department of Mechanical Engineering
Indian Institute of Technology – Kharagpur

Lecture - 26
Vibrations of Rings and Arches – II

This lecture is in continuation of the previous lecture on vibrations of rings and arches. Now, we
consider the characteristic equation (Eq. 25.4.2) when n = 1. Here, we are talking about non-
axisymmetric modes as there is explicit θ dependence.

Slide 26.1: (Refer Slide Time: 00:24)

--- 26.1.1

--- 26.1.2
--- 26.1.3

In that case, once again we get the first solution as ω = 0 which gives the corresponding eigenvector
as in Eq. 26.1.1. Here, once again, we suspect that this is a rigid body mode which it is. To see it
clearly, we expand the eigenvector in the form as shown in Eq. 26.1.1. The deformation in the ring
corresponding to this eigenfunction at θ = 0, π/2, π, 3π/2 is pictured in Slide 26.1. Thus, the ring
executes rigid body modes in the x and y direction which implies linear momentum conservation.
Now, let us consider the other solution (non-dimensional eigenfrequency ω) for n = 1 in Eq. 26.1.2
the eigenvector corresponding to which is shown in Eq. 26.1.3. The ring vibration corresponding
to this mode is shown in Slide 26.2, where the dashed circle is the undeformed position of the ring

232
and the solid circle represents the vibratory mode corresponding to the imaginary part of
eigenfunction in Eq. 26.1.3. Let us now try to understand this motion in detail (as discussed in
Slide 26.3).

Slide 26.2: (Refer Slide Time: 08:28)

Slide 26.3: (Refer Slide Time: 09:27)

233
The imaginary part of eigenfunction in Eq. 26.1.3 leads to the motion as shown in the figure in
Slide 26.3. In this case, the radial and circumferential amplitude at different angular position is
obtained as follows.
𝑢 1 0
𝐴𝑡 𝜃 = 0, { } = { } + 𝑖 { },
𝑤 0 1
𝜋 𝑢 0 1
𝐴𝑡 𝜃 = , { } = { } + 𝑖 { },
2 𝑤 −1 0
𝑢 −1 0
𝐴𝑡 𝜃 = 𝜋, { } = { } + 𝑖 { },
𝑤 0 −1
3𝜋 𝑢 0 −1
𝐴𝑡 𝜃 = , { } = { } + 𝑖 { }.
2 𝑤 1 0

The imaginary part of the above displacement values indicates that the material points on the ring
at θ = 0 will move radially outward (u = 0, w = 1), at θ = π/2 will move circumferentially counter-
clockwise (u = 1, w = 0), at θ = π will move radially inward (u = 0, w = -1), and at θ = 3π/2 will
move circumferentially clockwise (u = -1, w = 0). The motion of different material points on the
ring at this mode is pictured in Slide 26.2 where we can see the clustering or compression of the
material points in the region π/2<θ<3π/2 and rarefaction in the rest of the region. Continuing in
this manner, we can then solve for higher modes where we see something interesting in Slide 26.4.

Slide 26.4: (Refer Slide Time: 13:15)

not a nodal point

234
We have already discussed the breathing mode (ω4 = 1 in Slide 26.4) and the recently discussed
mode ω5 = 1.416 but they are not found to be the lowest modes. The lower and lowest modes
appear to be like (ω1, ω2, ω3) as shown in Slide 26.4 which can be easily obtained (using other
values of n) as we did for other modes. The dashed and solid curve denotes undeformed and
deformed configuration in Slide 26.4.

Now, on a cautionary note, the deceivingly appearing nodal points on the ring may not be the
actual nodal points [See Slide 26.4 for examples.]. In the case of ring vibrations, we have not only
a radial motion but also the circumferential motion. By definition, a node is considered to be a
point at which there is no motion. Whereas, a material point on the ring might not be stationary in
both the modes of motion. Thus, occurrence of actual nodes has to be properly checked by looking
at motion of material points on the ring. So far, we discussed about vibrations of rings. Next, we
will discuss vibrations of arches (Slide 26.5).

Slide 26.5: (Refer Slide Time: 15:58)

--- 26.5.1
--- 26.5.2

A circular arch is nothing but a sector of a ring. In this lecture, we will consider two examples of
arches. One is the pinned-pinned arch and the other is the clamped-clamped arch. Now, as you will
realize that the equations of motion for arches, which are undoubtedly the curved beams, are

235
dynamically coupled and thus, complicated to analyze. Therefore, for the case of arches, we are
going to use approximate methods like the Ritz method. To begin with Ritz method, let us first
look at the boundary conditions of the given arch as we need to choose admissible functions for
the application of the Ritz method.

First, let us consider the pinned-pinned arch as discussed in Slide 26.5. The boundary conditions
for pinned-pinned case (at θ = 0, π) are given in Slide 26.5. We have zero radial (w) and
circumferential (u) motion and zero moment condition at both the ends and thus, we have total six
boundary conditions at our disposal. Out of these six BCs, we have 4 geometric BCs which will
help us in deciding an appropriate admissible function as mentioned in Eqs. 26.5.1 and 26.5.2.
Here, we have taken 3-term expansion for an admissible function for u and w. Now, we substitute
these expansions in the Lagrangian as detailed in Slide 26.6.

Slide 26.6: (Refer Slide Time: 21:56)

--- 26.6.1

--- 26.6.2

Thus, we obtain the discretized equation of motion in terms of coordinates ai in Eq. 26.6.1 where
vector a contains six elements. Now, we can perform the modal analysis and obtain the circular
eigenfrequencies, two of which (ω1, ω2) are shown in Slide 26.6. Note that these are non-
dimensionalised eigenfrequencies. Dimensionalised eigenfrequency can be easily calculated from

236
the expression mentioned in Eq. 26.6.2. Let us now have a look on these first two modes of
vibration of pinned-pinned arch in Slide 26.7. Here, we can see that the first mode is the un-
symmetric mode and the second one is the symmetric vibration mode. In a similar manner, one
can perform for clamped-clamped arch as discussed in Slide 26.8.

Slide 26.7: (Refer Slide Time: 24:06)

Slide 26.8: (Refer Slide Time: 24:38)

--- 26.8.1
--- 26.8.2

237
For a clamped-clamped arch, the boundary conditions are given in Slide 26.8. Once again, using
the geometric boundary conditions, we have the expansions for admissible functions in Eqs. 26.8.1
and 26.8.2. We substitute these expansions in the Lagrangian and obtain the discretized equation
of motion and further calculate the eigenfrequencies which turns out to be in the form (only first
two frequencies are shown) as mentioned in Slide 26.8. The non-dimensional circular
eigenfrequencies in clamped-clamped case are higher than those of the pinned-pinned case as we
expect.

Slide 26.9: (Refer Slide Time: 25:15)

The corresponding modes of vibration for the clamped-clamped case is shown in Slide 26.9 where
the first mode is the un-symmetric mode, and the other is the symmetric mode with higher
frequency. Let us summarize what we have discussed today. In this lecture, we discussed the
vibrations of rings and arches. We have looked at some interesting results in the vibrations of rings
and we have considered two kinds of semicircular arches. Using Ritz method, we have determined
the eigenfrequencies and modes of vibrations for arches. With this, the lecture is concluded.

Keywords: Curved beam, rings, arches, coupled dynamics, Ritz method

238
 Vibrations of Structures
Prof. Anirvan DasGupta
Department of Mechanical Engineering
Indian Institute of Technology - Kharagpur

Lecture - 27
Dynamics of Membranes

In this lecture, we are going to initiate some discussions on Vibrations of membranes. First, we
are going to look at the dynamic modelling of membranes. Let us first define the membranes. A
membrane is a two-dimensional elastic continuum which cannot resist or transmit bending
moment. We have learnt somewhat similar for strings. The string, in fact is a one-dimensional
elastic continuum which does not resist bending moment whereas the membrane is a two-
dimensional elastic continuum. But then a membrane is not just a two-dimensional extension of a
string as we will see in this and subsequent lectures on membranes. Let us first look at the
mathematical modeling of the membrane.

Slide 27.1: (Refer Slide Time: 01:39)

We find membranes in ear drums, drum heads, and in certain kinds of microphones etc. We
study membranes not just to study drum heads or ear drum or microphones, but to understand a
general two-dimensional elastic continuum which does not transmit bending moment. Now, in

239
order to start the dynamic modelling, let us first enumerate the assumptions that we need to make
(See Slide 27.1).
1. We assume that the membrane is flat and the deflection in membrane is purely transverse.
Therefore, when we have, let us say an inflated membrane like a balloon, the dynamics is
more complicated because the deflection can no longer be considered to be just
transverse as there is a coupling between transverse and in-plane modes of deformation.
Thus, we are going to make this assumption that our membrane is flat so that only the
transverse deflection of the membrane dominates.
2. The second assumption that we make is that the slopes are small. The slope is taken as
the ratio of the maximum deflection and the characteristic dimension of the membrane.
For small slope, the deflection should be much much smaller than the characteristic
dimension of the membrane.
3. Thirdly, we assume that the tension in the membrane is uniform and it does not change
with deflection.
4. Finally, we assume that the thickness of the membrane does not change. Thus, under
these assumptions, we are going to now look at the modelling of the membrane.

Slide 27.2: (Refer Slide Time: 05:47)

--- 27.2.3

--- 27.2.2
--- 27.2.1
--- 27.2.4

--- 27.2.5

240
Let me first draw an infinitesimal element of this membrane as shown in Slide 27.2. We have
considered an element of this membrane which was initially flat on the xy-plane. In order to
locate the membrane deflection, we consider a field variable w = w(x, y, t). The free body
diagram of the membrane element is shown in Slide 27.2 where tension force T is shown on all
four edges. Note that T is the force per unit length that we can think in the following way. When
the membrane is being cut, there will be stresses on the cut, and if we integrate the stress over the
thickness then we obtain T, which is some kind of stress resultant as described in Eq. 27.2.1.

Using this, we are now going to write down the equation of motion. The transverse dynamics of
the membrane element is obtained as in Eq. 27.2.2 where µ be the mass per unit area. Here, we
have used Eq. 27.2.3 for slope. Thus, we get the equation of transverse dynamics once it is
divided throughout by dxdy and having used the Taylor series expansion of slopes w,x|(x+dx, y), and
w,y|(x, y+dy). Note that we obtain the Laplacian of w in the equation of motion after the
simplification process as shown in Eq. 27.2.4. The equation of motion is restructured as in Eq.
27.2.5 where c is the speed of transverse waves in the membrane. Note that we have derived the
equation considering the Cartesian coordinate system. We will look at the general situation in the
later part of the lecture.

Slide 27.3: (Refer Slide Time: 19:15)

241
In the equation of motion, we see that there is a double derivative in x and y. Therefore, we
would require boundary conditions on all the four edges. So, let us look at the boundary
conditions as described in Slide 27.3. Let us say we have a rectangular membrane fixed on all
edges. The BCs for fixed edged are mentioned in Slide 27.3. There is another possibility where
we may have sliding edge. The sliding edge in a membrane is very similar to the sliding edge of
a string. The transverse component of the tension force at the sliding edge must be zero as
mentioned in Slide 27.3.

Slide 27.4: (Refer Slide Time: 24:19)

We now derive the equation of motion in polar coordinates. We use polar coordinates when we
have a circular membrane. Suppose we have a circular membrane of radius a that is fixed on the
edge throughout as described in Slide 27.4. Here, we perform a simple co-ordinate
transformation to achieve the equation of motion in Polar coordinates from that in Cartesian
coordinates as done in Slide 27.4. Thus, we get the equation of motion in the polar coordinates
where w = w(r, φ, t) is the field variable. For a fixed boundary, the boundary condition can
simply be obtained as w(a, φ, t) = 0. Here, the choice of polar coordinates is actually guided by
the boundary. For simple boundaries like rectangular membrane or circular number, the choices

242
is very clear but for other shapes, some of which we are going to discuss later on, we will see
how they are dealt with.

Slide 27.5: (Refer Slide Time: 28:54)

--- 27.5.1

--- 27.5.2

We have discussed Newtonian formulation so far. Let us also look at the Variational formation
for the dynamics of membrane as described in Slide 27.5. To reiterate, the variational
formulation is quite powerful technique. Not only this method gives us the equation of motion
but also the boundary conditions. Furthermore, it facilitates the discretization of the dynamics of
continuous systems which paves way for approximate methods like Ritz method etc.

To begin with the variational formulation, we write down the KE expression of membrane as in
Eq. 27.5.1. Next, we find out the potential energy when the membrane deflects in the manner as
explained in Slide 27.5. To determine the potential energy, we consider stress-strain variation as
shown in the diagram in Slide 27.5. In the variation plot, the stress in the membrane is constant
in the same manner the tension in the membrane T is taken to be constant. Thus, the potential
energy per unit volume is given by the area under the stress-strain plot that is written as in Eq.
27.5.2. Finally, when we substitute the expressions for stress integrated over the thickness and
strains in x, y-directions in Eq. 27.5.2, we get the expression for potential energy.

243
Slide 27.6: (Refer Slide Time: 34:52)

Thus, the Lagrangian is obtained as in detailed in Slide 27.6. We now apply the Hamilton's
principle, integrate it appropriately, use Gauss divergence theorem to convert area integral into
line integral, and simplify to give equation of motion and boundary condition in Slide 27.7. Here,
we obtain both natural and geometric BCs in the process.

Slide 27.7: (Refer Slide Time: 41:08)

244
Slide 27.8: (Refer Slide Time: 46:09)

--- 27.8.1

Now, let us consider an example in Slide 27.8. In a rectangular membrane, unit normal at the
boundaries are obtained as shown in the diagram in Slide 27.8. At the boundary x = a, we have

the condition , which gives a dynamic/natural boundary condition as Tw,x (a, y, t) = 0.

If the boundary is fixed, then we will have the geometric boundary condition which says that the
displacement (w) is zero. If we have to formulate Lagrangian for a circular membrane in the
polar coordinates, we write it as in Eq. 27.8.1 which gives the equation of motion and boundary
conditions as detailed in Slide 27.8.

Thus, in this lecture, we have looked at the transverse dynamics of membranes, both from the
Newtonian as well as from the variational perspective. In the following lectures, we are going to
study the vibrations of membranes with circular and rectangular shapes. With that, we conclude
this lecture.

Keywords: Membranes, equation of motion, transverse dynamics, rectangular membrane,

circular membrane

245
 Vibrations of Structures
Prof. Anirvan DasGupta
Department of Mechanical Engineering
Indian Institute of Technology - Kharagpur

Lecture - 28
Vibrations of Rectangular Membranes

In the last lecture, we had discussed the dynamics of a flat membrane and in this lecture, we are
going to carry forward our discussions on the same and look at the vibrations of rectangular
membrane. To recapitulate what we discussed in the last lecture, let me write down the general
mathematical formulation of a flat membrane (Slide 28.1).

Slide 28.1: (Refer Slide Time: 00:53)

--- 28.1.1
--- 28.1.2

We had derived the equation of motion of a membrane having field variable w(x, y, t) or w(r, φ,
t) with boundary conditions as mentioned in Slide 28.1. Note the definition of Laplacian operator
in rectangular and polar coordinate system in Slide 28.1 as well. Thus, for a rectangular
membrane fixed on all sides, the equation of motion and boundary conditions can be written as
in Eqs. 28.1.1 and 28.1.2. Thus, the mathematical formulation for the rectangular membrane is
obtained. We are now going to look at the modal vibrations of the rectangular membrane as
discussed in Slide 28.2.

246
Slide 28.2: (Refer Slide Time: 07:21)

--- 28.2.1

--- 28.2.2

--- 28.2.3

--- 28.2.4

We are interested in a solution form that is separated in space and time as shown in Eq. 28.2.1.
We substitute this solution form in the equation of motion and boundary conditions that upon
simplification gives the eigenvalue problem (EVP) as shown in Slide 28.2. In order to solve the
EVP, let us consider a solution of the form as shown in Eq. 28.2.2. When we substitute this
solution form in differential equation of EVP, we obtain the characteristic equation as in Eq.
28.2.3 where α and β have to be chosen such that this equation is satisfied. The solution form in
Eq. 28.2.2 can very well be written in expanded form as in Eq. 28.2.4 where A1, A2, A3, and A4
are constants to be determined from BCs.

Now, we substitute the general solution of Eq. 28.2.4 in BCs of EVP as discussed in Slide 28.3.
Thus, we obtain the value of constants A1, A2, A3 as zero. Now, we are left with only A4 that we
refer to as only A in subsequent part of formulation. Using left-over boundary conditions, we
obtain the values of αm and βn which when used in Eq. 28.2.3, we get the expression for circular
eigenfrequency ω(m, n) of rectangular membrane as shown in Slide 28.3. Thus, the eigenfunctions
which should also now get indexed is obtained as shown in Slide 28.3. Here, of course, we may
have infinitely many eigenfunctions for natural numbers of m and n. Now, we note that the
fundamental circular eigenfrequency is obtained when m and n are equal to one and the higher

247
circular eigenfrequencies are no longer just integral multiples of the fundamental what we have
seen in strings.

Slide 28.3: (Refer Slide Time: 14:21)

Let us try to visualize the case of strings from the expression of eigenfrequency of membrane in
Slide 28.3. If we remove the term n2/b2 from the expression, we actually obtain the
eigenfrequencies of the string. The higher eigenfrequencies for strings, thus are just integral
multiples of the fundamental frequency. However, in the case of membranes, it is no longer true.
There exists another possibility with eigenfrequencies of membrane that is of two different
modes having same circular eigenfrequency. We will discuss this aspect in the later part of the
lecture.

For now, we discuss about the orthogonality of these eigenfunctions derived in Slide 28.3. To
establish the orthogonality property of eigenfunctions, we write an inner product of (m, n) and (r,
s) eigenfunction that is integrated over the domain of the membrane as given in Eq. 28.4.1. Using
the properties of the trigonometric functions, we simplify Eq. 28.4.1 which gives the
orthogonality property of the eigenfunctions as in Eq. 28.4.2 (Note the Kronecker delta functions

248
in the expression.). Let us now have a look at some of these eigenfunctions, pictorially in Slide
28.5.

Slide 28.4: (Refer Slide Time: 24:11)

--- 28.4.1

--- 28.4.2

--- 28.4.3

--- 28.4.4

Slide 28.5: (Refer Slide Time: 26:44)

249
Here I have plotted four modes of membrane vibrations. The mode (m, n) = (1, 1) is the
fundamental mode where there are no nodal curves. The modes (m, n) = (1, 2) and (m, n) = (2, 1)
have one nodal line along x- and y-axis respectively. The mode (m, n) = (2, 2) contains two
nodal lines as shown in the plot in Slide 28.5. These eigenfunctions can now be used for
constructing a general solution for any problem related to the rectangular membrane as
mentioned in Eq. 28.4.3 where constants C(m, n) and S(m, n) have to be determined from initial
conditions. In order to do that we have to use the orthogonality property. The initial conditions
can be of the form Eq. 28.4.4 where the initial configuration or the displacement of the
membrane and the initial velocity distribution is given. Using initial conditions in general
solution form, we can solve for these infinitely many coefficients C and S using the orthogonality
property that we have just now discussed. We have discussed this procedure in previous lectures
as well. Thus, we can determine these constants and finally construct the solution using the
summation.

Slide 28.6: (Refer Slide Time: 33:37)

--- 28.6.1

We now discuss an interesting property called Modal Degeneracy (Slide 28.6). The expression
of the circular eigenfrequency is derived in previous slides. Suppose, we define a/b = lr, that we
call as the length ratio. If lr is a rational number, we can have two integers a and b such that it is

250
equal to the length ratio lr and eigenfrequency expression is recasted as shown in Slide 28.6. If
for a given (m, n), there exists another mode (r, s) such that Eq. 28.6.1 is satisfied then we have
two identical circular eigenfrequencies ω(m, n) and ω(r, s). Thus, there are two orthogonal
eigenfunctions W(m, n) (x, y) and W(r, s) (x, y) with identical circular eigenfrequencies. Now, this is
an interesting occurrence. An example value for such an event is discussed in Slide 28.6. This
phenomenon is known as Modal degeneracy where we have identical eigenfrequencies but
distinct orthogonal eigenfunctions.

Slide 28.7: (Refer Slide Time: 40:23)

--- 28.7.1

--- 28.7.2

--- 28.7.3

Let us now try to depict this situation graphically (Slide 28.7). In fact, we have been discussing
this visualization in various contexts. Let us consider the modal or configuration space of the
membrane as pictured in Slide 28.7 where the axes are drawn perpendicular to each other to
depict orthogonality. Now, there may be man-many such axes which are orthogonal to one
another. Here, I am only drawing three modes (orthogonal eigenfunction) and one has to stretch
his/her imagination a little bit to visualize more of such modes. In the present modal space, the
orthogonal eigenfunctions corresponding to ω(5, 3), ω(4, 5), and ω(1, 1) are drawn. Note that the
eigenfrequencies ω(5, 3) and ω(4, 5) are equal as discussed in Slide 28.6 which means that the
eigenfunctions W(5, 3) and W(4, 5) correspond to same eigenfrequency say ω(5, 3). Therefore, any

251
linear combination of these two eigenfunctions is also an eigenfunction of the same frequency.
Thus, any vector representing a different eigenfunction that lies in the plane made by these two
eigenfunctions is also an eigenfunction belonging to the same frequency. Thus, we have some
kind of an isotropy in this modal space. This property is referred to as Modal degeneracy that is
not visible in the case of strings.

We have seen how the membrane dynamics is isotropic in the function space which is spanned
by the eigenfunctions W(5, 3) and W(4, 5). It we excite the membrane close to ω(5, 3), we can expect
the solution form as shown in Eq. 28.7.1 that can be written in a concise form as in Eq. 28.7.2.
Note that the frequencies ω(5, 3) and ω(4, 5) are actually one and the same. Since the final solution
in Eq. 28.7.1 is a combination of distinct eigenfunctions W(5, 3) and W(4, 5) corresponding to a
single frequency ω(5, 3), we observe a combination of these two modes appearing in the vibrations
of membrane and it may so appear that the membrane is vibrating in an unsteady manner because
it might be switching from one mode to the other or through a combination of these two modes.
We have this kind of a phenomenon because of isotropy of the modal space. As an example, if
the excitation initial conditions have been set such that the solution takes the form as in Eq.
28.7.3, we can have the following scenarios.

At t = 0, w(x, y, t) = C W(5, 3)
At t = π/2 ω(5, 3), w(x, y, t) = S W(4, 5)

Thus, the deflection pattern of the membrane is going to continuously make transitions from W(5,
3) to W(4, 5) and back. But, since this is the modal solution, we will actually see continuously
changing profile of the membrane and also, we do not see any fixed nodal lines. It the excitation
initial conditions contain cosine terms only then there will be a fixed combination of W(5, 3) and
W(4, 5) as discussed below. For w(x, y, t) = C cos ω(5, 3)t W(5, 3) + S cos ω(5, 3)t W(4, 5) then we will
have the following scenario.

At t = 0, w(x, y, t) = C W(5, 3) + S W(4, 5)

252
Thus, in that case we have fixed kind of a shape of the membrane vibrations. Thus, this is an
interesting feature of modal degeneracy which is present in membranes. In Slide 28.8, I have
shown the phenomenon of modal degeneracy for a square membrane.

Slide 28.8: (Refer Slide Time: 50:51)

Slide 28.9: (Refer Slide Time: 51:01)

--- 28.9.1

253
For the square membrane, eigenfrequency takes the form as shown in Slide 28.9. Here, we find
that ω(m, n) = ω(n, m) with m ≠ n are degenerate modes. Thus, all modes with m ≠ n are degenerate
modes for the square membrane. Let us now have a look at the picture in Slide 28.8. Using the
eigenfunction of the form as in Eq. 28.9.1, we have drawn the deflection pattern of degenerate
modes of a square membrane for m = 3, n = 1 and different values of δ in Slide 28.8. Thus, a
square membrane vibrates in this combination where we see the nodal line patterns changing
continuously through the shapes shown. Here, we do not see a steady nodal line or we can say
that the nodal lines are kind of unstable. Thus, this is something very interesting that one can
observe in an experiment on membrane vibrations.

To summarize, we have studied following things in this lecture. we have looked at the modal
vibrations of rectangular membrane and we have observed very interesting property which is the
modal degeneracy. Some of these aspects, we will discuss further in our next lecture. We also
looked at square membrane as an example case. It concludes this lecture.

Keywords: Membranes, rectangular membranes, square membrane, circular

eigenfrequency, eigenfunction, modal degeneracy

254
 Vibrations of Structures
Prof. Anirvan DasGupta
Department of Mechanical Engineering
Indian Institute of Technology – Kharagpur

Lecture - 29
Vibrations of Circular Membranes – I

We have been discussing about vibrations of membranes for last few lectures. In this lecture, we
are going to look at the vibrations of a circular membrane.

Slide 29.1: (Refer Slide Time: 00:50)

--- 29.1.1
--- 29.1.2

--- 29.1.3

Considering a circular membrane with fixed edge as shown in Slide 29.1. The radius of the
membrane is taken to be a. Any point on this membrane is denoted using the coordinates w(r, φ,
t) where r is the radial, and φ is the angular position. Thus, using w(r, φ, t) as our field variable
which is the transverse displacement of the membrane from its equilibrium position measured at
a coordinate location r, φ, and time t. The equation of motion for circular membrane transverse
vibration dynamics is as written in Eq. 29.1.1 where µ is the areal density and T is the force per
unit length. Note that the equation of motion is written in the polar coordinates. The boundary
condition for the given system is as given in Eq. 29.1.2 which implies zero displacement on the
boundary.

255
 Now, we will be looking at the modal solutions of the circular membrane. Therefore, we are
interested in the solutions of the form in Eq. 29.1.3. If we substitute this solution in the equation
of motion and in the boundary condition and rearrange it by removing the term eiωt, we obtain the
eigenvalue problem (EVP) as detailed in Slide 29.1. Now, we have to solve EVP in order to
determine the circular natural frequencies and the modes of vibration of a circular membrane.
We, now search for separable solutions once again as we did for the rectangular membrane. We
look for solutions which are separable in r and φ as indicated in Eq. 29.2.1.

Slide 29.2: (Refer Slide Time: 06:13)

--- 29.2.1

R = R(r) 1 2 --- 29.2.2

--- 29.2.3

Φ = Φ(φ)
--- 29.2.4

--- 29.2.5

The solution form in Eq. 29.2.1 is motivated by the fact that r and φ are independent coordinates
and therefore, they can be separated out. When we substitute this separable solution form in
EVP, we obtain a differential equation as in Eq. 29.2.2. After some rearrangements, the
differential equation in Eq. 29.2.2 takes the form as in Eq. 29.2.3. Note that Eq. 29.2.3 contains a
function of the coordinate r only (Box 1), a function of the coordinate φ only (Box 2) and a
constant ω2/c2. Therefore, expression in box 2 should be a constant as taken in Eq. 29.2.4 which
takes the solution form as shown in Eq. 29.2.5. Substituting the solution form of Eq. 29.2.5 in
Eq. 29.2.3, we obtain Bessel differential equation in R, the radial function as shown in Slide
29.2. It has the boundary condition R(a) = 0 as well. Thus, we have a second order ordinary

256
differential equation with only one boundary condition. We have come across this kind of
differential equation when we discussed the hanging string if you recall. There also we had the
Bessel differential equation except that the term m2/r2 was zero. So, let us see what is the
solution of the Bessel differential equation?

Slide 29.3: (Refer Slide Time: 16:48)

--- 29.3.1

--- 29.3.2

Slide 29.4: (Refer Slide Time: 18:30)

257
It is a very standard differential equation in mathematics and the solution of it can be written in
the form as shown in Eq. 29.3.1 where Jm, Ym are the Bessel functions of first and second kind of
order m with argument γr, respectively. Let us now have a look at the plots of these functions in
Slide 29.4 that shows both the Bessel functions of zeroth order with argument s. From theory of
Bessel functions, for integral values of the order, the Bessel function of the second kind has a
logarithmic singularity at argument equal to zero. Therefore, for membranes at r = 0, this
function is going to introduce logarithmic singularity that means it is going to go to minus
infinity which is highly undesirable. Therefore, we must drop this function from our solution. In
fact, Bessel function of second kind of all orders have this kind of singularity. Thus, the solution
of radial function R becomes as shown in Eq. 29.3.2. Now we have just one boundary condition
which gives Jm (γa) = 0. Thus, we must choose γ so that this condition is satisfied. Now, let us
look once again at the Bessel function of the first kind in Slide 29.5.

Slide 29.5: (Refer Slide Time: 21:11)

The plot in Slide 29.5 uses m = 0 which implies zeroth order of the Bessel function. This
function goes to zero at the points shown in the list inside the plot. We have shown only three but
this function has infinitely many roots. The first root is at s = γa = 2.405 [Indicated with
subscript (m=0, ith root) where i=0, 1, 2…]. The second root is at s = γa = 5.52 and the third root

258
is at s = γa = 8.654. Thus, the circular eigenfrequency corresponding to the mode 0 and 1 are
found as shown in Slides 29.5 and 29.6. Note that when m = 0, we have what are known as the
axisymmetric modes. And, for m ≠ 0, we have the unsymmetric modes.

Slide 29.6: (Refer Slide Time: 23:40)

Keywords: Membranes, circular membranes, axisymmetric modes, Bessel functions of first

kind, Bessel functions of second kind, logarithmic singularity

259
 Vibrations of Structures
Prof. Anirvan DasGupta
Department of Mechanical Engineering
Indian Institute of Technology – Kharagpur

Lecture - 30
Vibrations of Circular Membranes – II

This lecture is in continuation of our discussions on modal analysis of circular membranes that
we began in previous lecture. So far, we have found the solution of the radial function as the
Bessel function of first kind. We have already found the eigenfrequencies from numerical results
of Bessel function. In this lecture, we will explore an approximate analytical solution of the
Bessel function as it would be easier to use it in the absence of any numerical solution that may
too difficult to determine.

Slide 30.1: (Refer Slide Time: 00:20)

--- 30.1.1

--- 30.1.2

The approximate analytical form for the Bessel function of first kind, mth order can be written as
shown in Eq. 30.1.1 when the argument, x is very large. Let us see now how good this
approximation is. In Slide 30.2, a plot shows the actual and approximate Bessel function of the
first kind and of order 0, 1, and 2. The dashed line is the exact Bessel function and the solid line
is the approximate one. For 0th order, the solid line and the dashed are almost indistinguishable

260
except in the beginning. As we are interested in this approximation only to find out the roots of
the Bessel function, the zeroth order approximation gives very good match with the exact ones as
can be seen in the first plot in Slide 30.2.

Slide 30.2: (Refer Slide Time: 01:38)

For first order function, there is slight error in the first root but then from the second root
onwards, there is good match. For second order function, approximation is not so good for the
first root, there is some error in the second root, third root is quite close and from the fourth root
onwards, it is a good match. Thus, we have seen that for large arguments, we generally have a
good match between approximate and exact functions.

Now, we apply approximate definition of Bessel function on the boundary condition Jm (γa) = 0
as we have seen in the previous lecture. This gives us circular eigen frequencies as given by Eq.
30.1.2. Now, let us see how good is this approximation. Thus, we can estimate the approximate
eigenfrequency for different order (m) and mode (n). We have listed approximate and exact
values of eigenfrequencies for various order and mode for comparison in Slide 30.1. Thus, we
have seen that as we go to higher modes in each case, we approach the exact solution. Let us

261
now look at the eigenfunctions.

Slide 30.3: (Refer Slide Time: 9:09)

--- 30.3.1

--- 30.3.2

--- 30.3.3

The eigenfunction is already defined in the form as shown in Eq. 29.2.1 and once again, written
in Slide 30.3 where radial function is proportional to the Bessel function of first kind. The
eigenfunction is a complex function of the form cosmφ + i sinmφ. Now, as we have discussed
before, we can take the real part or we can take the imaginary part or we can take a linear
combination of these two parts. Here, we write the eigenfunction as a linear combination of
cosine and the sine term as in Eq. 30.3.1 where the coefficients G and H are indexed by (m, n).
Here, I am introducing another notation with superscript c and s to express Eq. 30.3.1 in concise
form. The superscript c and s indicate cosine and sine function, respectively.

The orthogonality property applied on Ith and Jth eigenfunction, that could very well be the cosine
or the sine mode as obtained in Eq. 30.3.2, is given by Eq. 30.3.3. It can be easily seen that the
cosine and sine mode are orthogonal, that corresponds to a single circular eigenfrequency ω(m, n)
which happens only when m ≠ 0 (unsymmetric modes). If m = 0 then we have only the cosine
mode, we do not have the sine mode as sinmφ = 0 at m = 0. Thus, at m ≠ 0, we have modal
degeneracy. So, for circular membrane, all unsymmetric modes are modally degenerate. Let us

262
see how the vibration modes look like in a circular membrane (Slide 30.4).

Slide 30.4: (Refer Slide Time: 18:30)

In the figure in Slide 30.4, I have plotted out two symmetric modes (m = 0) and two
unsymmetric modes (m ≠ 0). For axisymmetric ones, modes corresponding to the first and
second root (n = 1, 2) are plotted where we note the presence of a nodal circle (n-1 number of
nodal circles) as n increases. Now, as we increase m, we generate a nodal line or a nodal
diameter as we can see in third and fourth plot (for m = 1) in Slide 30.4. For these unsymmetric
modes, n = 2 gives both, one nodal circle and a nodal diameter. Note that the unsymmetric
modes are degenerate modes. For example, corresponding to eigenfrequency ω(1, 1), there are two
eigenfunctions, one is the cosine the other is the sine that can be seen by the different orientation
of the nodal diameter. Now, the question may arise why do we have degeneracy?

We have already discussed about the isotropy in the modal space of rectangular membrane. Now,
we can see this isotropy in the physical space as well in Slide 30.4. The reference line from
where φ is measured is arbitrary. We can have the sine or cosine mode depending on where the
reference line is chosen. In this manner, as there is arbitrariness in the position of reference line,
there is isotropy in the rotational direction of the circular membrane. Therefore, we have modal

263
degeneracy, that is there are two independent orthogonal eigenfunctions corresponding to a
single eigenfrequency. Any arbitrary combination of these eigenfunctions is also an
eigenfunction. This, we refer to as an isotropy in the modal space or the configuration space of
the membrane. In the circular membrane, this is also visible in the physical space as we have
seen in Slide 30.4. Let us now write down the general solution of the vibration of the circular
membrane.

Slide 30.5: (Refer Slide Time: 20:30)

The general solution for the circular membrane may be written as shown in Slide 30.5 where we
note that m starts from zero which indicates the axisymmetric mode, and A(m, n), B(m, n), G(m, n) and
H(m, n) there are constants which are to be determined from the initial conditions. Recall that we
can determine these constants using the orthogonality property of the eigenfunctions.

Let us now look at an interesting property that arises because of modal degeneracy which is
called mode splitting or frequency splitting (Slide 30.6). We consider a circular membrane and
put an external interaction, for example a spring as shown in Slide 30.6. We have a reference line
at φ = 0 and the position of the spring is at r = r0. The equation of motion for such a system can
be written as in Eq. 30.6.1where an additional term due to spring stiffness at r = r0 and at φ = 0 is
added. To understand the behavior of its solution, let us just look at a single mode expansion as
in Eq. 30.6.2. Here, we have taken a single mode, and corresponding to a single frequency, there

264
are two degenerate modes (sine and cosine). If we substitute the expansion in Eq. 30.6.2 in the
equation of motion in Eq. 30.6.1, take the inner product, and discretized the equation, we get the
dynamics of the modal coordinates g(m, n), and h(m, n) as shown in Slide 30.6. We can note that the
equation corresponding to the sine mode remains unaffected because of the spring whereas the
one corresponding to the cosine mode is affected as there is an additional term with ω2(m, n). It
implies that the frequency of these two modes is now separated and this is called the mode
splitting or the frequency splitting. This feature takes place because of an external interaction
which as you can understand, breaks the symmetry.

Slide 30.6: (Refer Slide Time: 22:52)

--- 30.6.1

--- 30.6.2

Let us summarize this lecture. Here, we discussed the vibrations of a circular membrane and
modal degeneracy. We have seen, when we break the isotropy or the symmetry of the system
then the frequency splits. This concludes the lecture.

Keywords: Membranes, circular membranes, axisymmetric modes, Bessel functions of first

kind, Bessel functions of second kind, modal degeneracy, mode splitting, frequency
splitting, isotropy of modal space

265
 Vibrations of Structures
Prof. Anirvan DasGupta
Department of Mechanical Engineering
Indian Institute of Technology – Kharagpur

Lecture - 31
Dynamics of Plates

In this lecture, and in subsequent lectures, we are going to discuss the dynamics of plates.
First, let us try to find answer to this question: what is a plate? We have seen that a membrane
is a two-dimensional continuum which does not transmit any bending moment. Conversely,
the plate transmits or resists the bending moment. Thus, a plate is a two-dimensional elastic
continuum which resists or transmits bending moment. In applications, the plates are found in
various machines, and civil structures etc.

Slide 31.1: (Refer Slide Time: 02:03)

First, we are interested in modelling the dynamics of plates or to simply put it, in setting up
the equations of motion for plate vibrations. To study the plate dynamics, we make following
assumptions to simplify the model (Slide 31.1).
1. There is a plane or fibers which are unstressed that are also called as neutral fibers.
Thus, we assume the presence of neutral fibers which occurs if we do not have in-
plane forces on the plate. When a plate undergoes small transverse vibrations then
there are fibers which remain unstrained.

266
 2. There is no shear deformation. This is also known as the Kirchhoff hypothesis which
corresponds to the Euler-Bernoulli hypothesis for beams.
3. The slopes in the plate due to transverse deflections are small.
4. The plate thickness is constant.

Slide 31.2: (Refer Slide Time: 04:49)

We consider a plate lying in the xy-plane and a little element in this plate as shown in Slide
31.2. The deflection in the plate is transverse to xy-plane. The non-zero stresses acting on the
plate element are normal stresses σxx, σyy and shear stresses σxy, σxz, and σyz. Now, there can be
a distributed force on the plate as well but at present, we are going to drop that. Now, we
define what are known as the force and moment resultants as explained in Slide 31.3. Thus,
the in-plane stress resultants Nx and Ny are due to the normal stresses, and Nxy is because of
the shear stress as in Eq. 31.3.1. The out of plane shear stresses are denoted by Qx and Qy as
obtained in as in Eq. 31.3.2 and we also obtain the in-plane moment resultants Mx, My, and
Mxy due to the normal and the shear stresses as shown in Eq. 31.3.3. Note that the moment
resultant Mx, obtained as the integral of zσxx is along the y-axis, but we still call it as Mx
because it is the moment obtained with σxx. Also note that the above resultants have the units
of force per unit length or moments per unit length. These force and moment resultants on
shown on the infinitesimal element in Slide 31.4. As an example, Nx acts at face x = 0, along
the negative x-direction and Nx + Nx, x dx is along positive x-direction. Similarly, other normal
force resultant can be seen on corresponding faces.

267
Slide 31.3: (Refer Slide Time: 09:38)

--- 31.3.1

--- 31.3.2

--- 31.3.3

Slide 31.4: (Refer Slide Time: 13:26)

Likewise, the shear force and moment resultants are shown on corresponding faces in Slide
31.4. Let us now write down the constitutive relation for the material as detailed in Slide
31.5. We use Hooke’s law to relate stresses to the strains. As the plate is considered infinitely
stiff in shear, there won’t be any shear deformation, which implies εxz = 0 and εyz = 0. Now
we look at the geometry of deformation to calculate the strain in terms of deflection of the
plate as shown in Slide 31.5. We consider the plate initially undeformed represented by

268
dashed curve. The plate deflects in the manner as pictured in Slide 31.5 where a line which is
initially along the z-axis deflects to a new configuration. Thus, the displacement is in the
direction of x that can be written approximately as u = - zw,x as indicted in Slide 31.5. Thus,
we obtain εxx = u,x = - zw,xx.

Slide 31.5: (Refer Slide Time: 20:30)

Slide 31.6: (Refer Slide Time: 28:37)

Similarly, we obtain the strain field εxx and εxy as shown in Slide 31.5. We note that the strain

269
field, and so the stress field is linear in z. Therefore, the force resultants Nx, Ny, and Nxy are all
zero. We will have only non-zero resultants as Q and M. The moment resultants are thus,
obtained as in Slide 31.6 while the force resultants Q will ultimately be obtained from the
equations of motion. Now we can write down the equations of motion.

Slide 31.7: (Refer Slide Time: 31:00)

We write down the equation of motion in the transverse direction as in Slide 31.7.

Slide 31.8: (Refer Slide Time: 32:47)

--- 31.8.1

--- 31.8.2

--- 31.8.3

270
Let us now look at the rotational dynamics as detailed in Slide 31.8. We write equations for
angular acceleration in x and y-direction as in Eq. 31.8.2 where I = ρh3/12 is the moment of
inertia per unit area. We now eliminate Qx and Qy in Eq. 31.8.1 using the expressions in Eq.
31.8.2 that, on simplification gives the equation of motion of the plate a shown in Eq. 31.8.3.
Note that the term ∇4 is the square of Laplacian. This model is known as the Kirchhoff-
Rayleigh plate model where we can note the rotary inertia term in the equation of motion. If
we drop the rotary inertia term assuming that the moment of inertia is very small, we will
have the Kirchhoff plate model.

Slide 31.9: (Refer Slide Time: 41:40)

We now discuss about the boundary conditions in Slide 31.9. Let us consider a plate having a
clamped boundary. At clamped edge, the displacement and slope must be zero which are
geometric boundary conditions. If we have a simply supported edge at x = a, we have the
BCs as displacement and moment equal to zero which are natural boundary conditions. The
zero-moment condition can be written as in Slide 31.9. Now, we come to an interesting case
of a free boundary as discussed in Slide 31.10. If a plate has a free boundary then intuitively,
we may guess that the shear stress resultant Qx which is along the z-direction is going to be
zero. The moment resultant Mx is equal to zero and in addition we have another moment
because of the in-plane stresses that also has to be put to zero. But we note that the
differential equation of the plate is of 4th order and it cannot satisfy three boundary
conditions at an edge. Thus, there seems to be a discrepancy that has be corrected. Intuitively

271
speaking, there must be a combination of BCs that satisfies the free edge condition. To check
that, let us look at the moment resultant Mxy.

Slide 31.10: (Refer Slide Time: 45:50)

--- 31.10.1

--- 31.10.2

The moment resultant Mxy on the face x = a is a moment because of the in-plane shear stress
as can be seen in figure in Slide 31.10. This moment resultant can change as we move in the
y-direction. We consider two locations separated by a small distance ε. Thus, as we move in
the y-direction, the moment at the distance epsilon from Mxy can be written as Mxy + εMxy,y.
Now, these moment resultants can be equivalently represented as a couple where Mxy and Mxy
+ εMxy,y are equivalently written as forces Mxy/ε and (Mxy + εMxy,y)/ε along positive and
negative z-directions as shown in the figure in Slide 31.10. Thus, the net transverse force at
the edge x = a is obtained as Mxy,y as shown in Eq. 31.10.1 which is an additional edge
resultant force in the transverse direction, same as the shear resultant Qx. Therefore, we
combine these two forces so as to give an Edge Force in Eq. 31.10.2. An edge force is a
combination of the force because of the out of plane shear stress and the moment because of
in-plane shear stress. For free edge, the edge force must vanish. The vanishing moment
resultant and an edge force BCs in terms of the field variables, thus can be written easily as
done in Slide 31.10.

Let us summarize the lecture. In this lecture, we have looked at the equation of motion of
small amplitude, small slope vibrations of a flat plate along with BCs. The boundary

272
conditions for the clamped and the simply supported the boundary conditions are quite simple
whereas for the free boundary, an edge force needs to be determined. It concludes the lecture.

Keywords: Plate dynamics, Flat plate, Kirchhoff hypothesis, Kirchhoff-Rayleigh plate,

Edge Force

273
 Vibrations of Structures
Prof. Anirvan DasGupta
Department of Mechanical Engineering
Indian Institute of Technology – Kharagpur

Lecture - 32
Vibrations of Rectangular Plates

In the last lecture, we had seen the mathematical modelling of plate vibrations under small
transverse displacements. In this lecture, we are going to discuss the vibrations of rectangular
plates.

Slide 32.1: (Refer Slide Time: 00:44)

--- 32.1.1

--- 32.1.2

Let us recall the equation of motion for a Kirchhoff plate with differential operator and
parameter definitions involved as shown in Slide 32.1. We are interested in the modal
analysis and therefore, we are looking for the solutions in the Cartesian coordinate of the
form separable in space-time as in Eq. 32.1.1. Note that the plate is considered to lie in the
xy-plane and the displacement or the field variable w(x, y, t) is measured in the direction
transverse to the plane that is perpendicular to the plane of the paper. If we substitute this
solution form in the equation of motion, we obtain the differential equation of the eigenvalue
problem as in Eq. 32.1.2 where γ is defined as shown. The complete eigenvalue problem
description will also have the boundary conditions along with this differential equation that
we will look at later. Let us look at this differential equation first.

274
Slide 32.2: (Refer Slide Time: 06:45)

--- 32.2.1

--- 32.2.2

Slide 32.3: (Refer Slide Time: 11:10)

--- 32.3.1

--- 32.3.2

--- 32.3.3

--- 32.3.4

We can write Eq. 32.1.2 as the operator (∇4- γ4) operating on W(x, y, t) resulting zero as
shown in Eq. 32.2.1 in Slide 32.2. We can factorize the operator in the form shown and thus,
the factored operator can operate on W in any order. Now, we consider two functions W1(x,
y) and W2(x, y) operated by the factored parts as shown in Slide 32.2. Therefore, we say that
the total solution W(x, y) can be written as a combination of functions W1 and W2. One can
easily check that a solution like in Eq. 32.2.2 satisfies our original differential equation of the

275
eigenvalue problem in Eq. 32.2.1. However, this is not the most general solution structure
that is possible. In fact, this solution is valid or good for a class of problems by which I mean
that class of boundary conditions.

Let us now search for solutions of these two differential equations. First, we take the
differential equation as shown in Slide 32.3, that is also known as the Helmholtz equation
which we have come across in the dynamics of membranes as well. We can also recall the
general solution of the Helmholtz equation as shown in Eq. 32.3.1 where α2 + β2 = γ2. Let us
now look at the other differential equation for which we look for solutions separable in x and
y as shown in Eq. 32.3.2. Substituting this solution form in the corresponding differential
equation and simplifying by dividing throughout by XY, we obtain the equation as in Eq.
32.2.3 where we note that the first term is solely a function of X, the second term is only a
function of Y, and the third term is a constant. Therefore, for arbitrary XY if this equation has
to be satisfied, then each of the terms must be constants which gives equations of the form as
shown in Eq. 32.2.4. Let us now look at the solution of these differential equations for X and
Y in Slide 32.4.

Slide 32.4: (Refer Slide Time: 17:50)

--- 32.4.1

The solution X(x) and Y(y) of these differential equations may be written as shown in Slide
32.4 and thus, W2(x, y) is obtained as W2(x, y) = X(x)Y(y) where product of two constant is
named as a different constant. For example, product of C1 and C3 is written as A5 and so on.

276
Thus, the total solution for the eigenfunction is obtained as W(x, y) = W1(x, y) + W2(x, y) as
in Eq. 32.4.1. Note that the above eigenfunction solution contains product of the
trigonometric and hyperbolic functions in x- and y-directions while there is no cross product
of trigonometric and the hyperbolic functions present. Therefore, intuitively speaking, this
solution does not seem to be general enough. Nevertheless, this solution can solve a class of
problems if not all. Thus, one can use this solution for a class of boundary conditions. Let us
now look at a certain example.

Slide 32.5: (Refer Slide Time: 23:37)

We consider a rectangular plate which is simply-supported on all four boundaries. The plate
has sides x = 0, a and y = 0, b. We have discussed the boundary conditions of such a plate in
the last lecture. We have mentioned the BCs in slide 32.5 once again. The corresponding BCs
of the eigenvalue problem can be easily determined from BCs on the field variables and their
derivatives as detailed in Slide 32.5. We use these boundary conditions in the solution form
given in Eq. 32.4.1 that gives the final outcome as obtained in Eq. 32.6.1. The final solution
for eigenfunction in Eq. 32.6.1 further gives conditions on α and β for (m, n)-mode of
vibration that can be used to determine the eigenfrequency and the eigenfunction of the
corresponding mode as discussed in Slide 32.6. Here, we can note that these eigenfunctions
satisfy the orthogonality conditions as mentioned in Slide 32.7. Thus, the general solution of
the plate using these eigenfunctions can be obtained as in Slide 32.7. Here, one can note the
similarity at the eigenfunctions for membranes and the simply-supported plate.

277
Slide 32.6: (Refer Slide Time: 29:40)

--- 32.6.1

Slide 32.7: (Refer Slide Time: 37:13)

Let us now look at another example of the plate with mixed kind of support (Slide 32.8).
Here, we consider two edged as simply supported while the other two opposite edges as
clamped. The boundary conditions for the EVP of such a plate can be written as in Slide 32.8.
Here, we note that the solution form in Eq. 32.4.1 cannot satisfy the boundary conditions in
Slide 32.8. As we discussed earlier about the generality (rather lack of generality) of the
solution form of Eq. 32.4.1, these boundary conditions cannot be satisfied by this class of

278
solution. Therefore, we have to start afresh for this set of boundary conditions. Now, we look
at the eigenvalue problem afresh in Slide 32.9.

Slide 32.8: (Refer Slide Time: 40:40)

Slide 32.9: (Refer Slide Time: 45:47)

--- 32.9.1

--- 32.9.2

The differential equation of the eigenvalue problem is once again taken. Let us now try a
solution form as in Eq. 32.9.1 which already satisfies the two boundary conditions, i.e., at x =
0 and x = a in Slide 32.8. For the y-coordinate, we consider an unknown function Y(y). If we

279
substitute the solution form in the differential equation, and make some simplifications, we
obtain the fourth order ODE in function Y(y) as shown in Slide 32.9. If you try a solution of
the form as in Eq. 32.9.2 for the ODE in Eq. 32.9.1, the differential equation reduces to a
polynomial equation of 4th degree in p as mentioned in Slide 32.9 that can be factorized as
shown. Thus, we have two solutions of p as p1 = α and p2 = β as shown in Slide 32.10.

Slide 32.10: (Refer Slide Time: 50:20)

--- 32.10.1

Slide 32.11: (Refer Slide Time: 56:20)

280
Thus, the solution Y(y) and W(x, y) is obtained as shown. Note that β can have both real and
imaginary solution form and thus, contributes the solution Y(y) in terms of trigonometric
functions. Now if we substitute this solution form in the boundary conditions of Slide 32.8,
we obtain the characteristic equation for non-trivial solutions of C1, C2, C3, and C4 as in Eq.
32.10.1. We can solve this characteristic equation numerically for γ, and thus the natural
frequencies can be determined. We have written out the eigenfrequencies of first three modes
in Slide 32.10. A figure in Slide 32.11 shows the first three modes of vibration of this plate.
In the first mode, there are no nodal lines whereas the second and third modes have these
nodal lines. We can also note the clamped edges having zero slope in the figure.

Let us summarize this lecture. Here, we discussed the vibrations of rectangular plates. We
have looked at two classes of boundary conditions with their solutions. We determined the
eigenfrequencies and the modes of vibration for each class of problem. It concludes this
lecture.

Keywords: Plate dynamics, Rectangular plate, Helmholtz equation

281
 Vibrations of Structures
Prof. Anirvan DasGupta
Department of Mechanical Engineering
Indian Institute of Technology – Kharagpur

Lecture - 33
Vibrations of Circular Plates

In this lecture, we are going to discuss the vibrations of circular plates, the plates with polar
geometry.

Slide 33.1: (Refer Slide Time: 00:42)

--- 33.1.1

We consider a circular plate with field variable represented as w(r, φ, t) [Slide 33.1]. The
equation of motion for a Kirchhoff plate reads as shown in Eq. 33.1.1 where ∇4 in polar
coordinate is as shown. Here, essentially, we are going to solve the eigenvalue problem for
various kinds of boundary conditions for the vibration of circular plate. Let us now have a look at
the boundary conditions in the case of rectangular plates for a simply supported edge that we will
use to obtain the similar condition for the circular plate. The BC for the rectangular plate is as
shown in Slide 33.2. When we come to the simply supported edge condition for the circular
plate, what we can do is to use the transformation from the rectangular to polar coordinates as

282
shown. Thus, the BC for simply supported edge of circular plate is obtained as discussed in Slide
33.2.

Slide 33.2: (Refer Slide Time: 04:25)

Slide 33.3: (Refer Slide Time: 09:57)

283
Likewise, for the Clamped edge of a circular plate, the boundary conditions would be the
vanishing displacement and slope at r = R. For the free edge, one can derive the BCs from the
boundary conditions of the rectangular plate using the transformation as discussed in Slide 33.2.
For the free edge, the stress resultant because out of plane shear and the edge force at r = R must
vanish.

Slide 33.4: (Refer Slide Time: 13:21)

--- 33.4.1

Let us now look at the eigenvalue problem that we obtained while performing the modal
analysis. Thus, we look at the solutions of the form, separable in space and time as shown in
Slide 33.4. When we substitute the assumed solution form in the equation of motion, we obtain
the equation as in Eq. 33.4.1 that can be decomposed as shown. We once again try a solution
form for W(r, φ), separable in r and φ as shown. Now, if we substitute this solution form in
decomposed equation, we obtain the form as shown in Slide 33.4. Note that the function R(r)
[named as Rm(r) for mth solution] is being operated by two differential operators. Now, we
consider the function R(r) to be constructed out of two functions A(r) and B(r) such that A(r) is a
solution for the first bracketed operator and B(r) is a solution for another bracketed differential
operator as shown in Slide 33.4.

284
Slide 33.5: (Refer Slide Time: 18:21)

Since A(r) and B(r) satisfies the first and second bracketed differential operator in Side 33.4, we
obtain, respectively, the Bessel differential equation and the Modified Bessel differential
equation as shown in Slide 33.5. The solution of the Bessel differential equation can be written
as shown where C1 and C2 are constants and Jm and Ym are the Bessel function of first and second
kind of order m, respectively. Similarly, the solution of the Modified Bessel differential equation
is written in terms of the Modified Bessel functions as shown in Slide 33.5. Now, we know that
the Bessel function Y has a logarithmic singularity at zero argument that is at r = 0. Therefore,
the function Y cannot appear in the solution of Am for a complete circular plate. If we have an
annular plate, then the function Y can be present, though.

Let us now look at the behavior of the Modified Bessel functions appearing in the solution of Bm.
In the figure in Slide 33.6, the Modified Bessel functions are plotted for m = 0. Here, once again,
we see that the Modified Bessel functions of the second kind K0 has a singularity again at the
argument zero. Therefore, in a complete circular plate, function K cannot appear. However, for
annular plate, it can appear. Thus, the total effective solution in R becomes as shown in Slide
33.5.

285
Slide 33.6: (Refer Slide Time: 24:30)

Slide 33.7: (Refer Slide Time: 26:29)

For boundary conditions, we consider a clamped circular plate for which the displacement and
slope (derivative of field variable with respect to radius r) at r = a will be zero. Now, if we use
these conditions on the solution of Rm(r), for non-trivial solutions of C1 and C3, we obtain the

286
characteristic equation (by vanishing the determinant of matrix of coefficients) as shown in Slide
33.7, from where we will solve for γ. Since γ is in terms of the frequencies, knowing the γ gives
us the frequency. The characteristic equation has been solved numerically and first few values of
γa is shown in Slide 33.7.

Slide 33.8: (Refer Slide Time: 32:48)

Now, the above characteristic equation affords an approximation as we have seen earlier. The
Bessel function of the first kind Jm(x) and Modified Bessel function of the first kind Im(x) has an
approximation for large arguments as shown in Slide 33.8. The function Im(x) is observed to be
proportional to exponential x which is increasing in x. The derivative of this function, therefore is
also some function of x times the exponential function which is also increasing in x. Therefore,
for large arguments of this Modified Bessel function, they can be dropped from the equation.
Thus, the characteristic equation simplifies as shown when γ is very large. Using this
approximation of the characteristic equation, we can approximate the circular
eigenfrequencies of circular plate as shown in Slide 33.8.

Now, let us look at the eigenfunctions in Slide 33.9. The radial eigenfunction R(m, n)(r), having
two indices m and n can be written as shown. Now, our original eigenfunction W(m, n)(r, φ) can

287
be written as shown that has the cosine and the sine component or any linear combination of
these two. These two eigenfunctions are of course orthogonal while they correspond to the same
eigenfrequency ω(m, n). Therefore, these are degenerate modes for m ≠ 0. The mode
corresponding to m = 0 is the axisymmetric mode which implies that the function is dependent
on radial coordinate only and there is no dependence on φ. For m ≠ 0, we have the unsymmetric
modes.

Slide 33.9: (Refer Slide Time: 37:21)

Slide 33.10: (Refer Slide Time: 42:36)

288
In Slide 33.10, different modes of vibrations of a clamped circular plate can be visualized. Note
the absence or presence of nodal curves, nodal diameter or nodal circle in different modes. The
modes (m, n) = (1, 1) and (m, n) = (1, 2) are unsymmetric modes which have modal degeneracy.
From the geometric symmetry or isotropy of the circular plate, we can note that there is no
particular choice of any reference line from where the angular coordinate φ can be measured.
The arbitrariness of the reference line for a perfectly circular plate gives rise to degeneracy, that
is two different eigenfunctions for a same eigenfrequency.

Slide 33.11: (Refer Slide Time: 45:20)

Now, using these eigenfunctions, we can write down the general solution where we can have
linear combinations of cosine and sine terms in the solution as shown in Slide 33.11. The general
solution can also be written in the complex form as shown. Here, one can note that the modal
degeneracy gives a travelling waves solution in the circumferential direction as shown in Slide
33.11. Modal degeneracy leads to seemingly unsteady modal solutions as can be seen from the
general solution form.

Let us now summarize this lecture. In this session, we have discussed the vibrations of circular
plates. Further, we have looked at some of the boundary conditions for the circular plate. We

289
found that the analysis of BCs in the case of circular plates is more complicated than that of the
rectangular plates. In the end, we discussed the modal analysis of clamped circular plate.

Keywords: Plate dynamics, Circular plates, Modal degeneracy, Kirchhoff plate

290
 Vibrations of Structures
Prof. Anirvan DasGupta
Department of Mechanical Engineering
Indian Institute of Technology- Kharagpur

Lecture – 34
Special Problems in Plate Vibrations - I

For last few lectures, we have been discussing about the vibrations of plates. We have already
looked at the vibrations of rectangular and the circular plates. In this lecture, we are going to
discuss some special problems in vibrations of plates.

We have seen in the previous lectures that even for some simple problems, the analytical
solution becomes very complex and sometimes with increasing complexity of the geometry
or the boundary conditions, the analytical solution becomes intractable. Therefore, in such
situations, we would like to have approximate methods for discretizing the dynamics of plates
and performing the modal analysis that will generate approximate eigenfrequencies. But then,
there is always a scope of improving the accuracy of these approximate methods. We have
already seen several approximate methods to solve the problems of vibrational dynamics in
this course. In this lecture, we are going to look at two examples that we are going to solve
using the approximate methods.

Before we start discussing the examples, let us look at the variational formulation of plate
dynamics. The variational formulation for plates is little cumbersome. Nonetheless, we
attempt it here. As we have learnt previously that the variational formulation generates the
boundary conditions along with the governing equations. One can verify that the boundary
conditions that we have discussed based on the Newtonian formulation and based on Born-
Von Karman’s boundary condition for free edge, we can obtain them directly from the
variational formulation.

Let us now consider a plate as shown in Slide 34.1. The kinetic energy of the plate can be
expressed as in Eq. 34.1.1 where z is the coordinate perpendicular to the plane of the shown
figure and I is the moment of inertia per unit area of the plate. Here, we have considered the
rotary inertia term while writing for the kinetic energy. Next, we write down the potential
energy expression as explained in Slides 34.1 and 34.2 where the term D is defined as shown,
E is Young’s modulus and ν is Poisson’s ratio.

291
Slide 34.1: (Refer Slide Time: 03:04)

--- 34.1.1

Slide 34.2: (Refer Slide Time: 12:22)

Having obtained the expressions of kinetic and potential energy, we can now write use the
Hamilton’s principle to derive the equation of motion (Slide 34.2). We can also use the
variation formulation for discretization using Ritz method. Thus, the Ritz expansion along
with the Hamilton’s principle lead us to the discretization of the dynamics.

Keywords: Plate dynamics, Variational formulation

292
 Vibrations of Structures
Prof. Anirvan DasGupta
Department of Mechanical Engineering
Indian Institute of Technology- Kharagpur

Lecture – 35
Special Problems in Plate Vibrations - II

In this lecture, we look at two examples on plate vibrations. The first example that we are
going to discuss is that of a square plate supported on a circular boundary as shown in Slide
35.1. Let us look at the geometry of the problem as shown.

Slide 35.1: (Refer Slide Time: 00:38)

--- 35.1.1

We consider a square plate of uniform thickness and side length 2l which is simply supported
on a circular boundary as shown. The radius of the support circular boundary cantered at the
geometric centre of the square is a. We already have the kinetic and potential energy
expressions of a plate as derived in lecture 34 but we will make some non-dimensionalization
that will simplify the equations as shown in Slide 35.1. To list a few, x-coordinate is non-
dimensionalised with respect to the half-length of the plate. Similarly, the field variable w
which tracks the transverse displacement of the neutral plane is also non-dimensionalized
with l and we define non-dimensionalised half-length of the plate. Now, we consider the
Kirchhoff plate.

293
In the Kirchhoff plate, we do not consider the rotary inertia terms which simplifies the kinetic
energy expression. One can drop the tilde symbol from the non-dimensionalised term for
convenience of writing and that is what we have followed here. The limits on x and y
coordinates go from -1 to +1 in the integration as shown. Similarly, the potential energy
expression is written as shown in Slide 35.1. Let us now look at the support conditions.

For this problem, we will be using the Ritz method that requires only admissible functions.
Since the plate is simply supported, the only condition at this boundary is of zero
displacement. Therefore, we parameterize this circle using angle φ and the boundary
condition is written as shown in Eq. 35.1.1. The same BC can also be written in terms of x
and y coordinates as shown. Now, we write the solution of the field variable as an expansion
as shown in Slide 35.2 where W(m, n) (x, y) is admissible function that need to satisfy only the
geometric boundary conditions of the problem

Slide 35.2: (Refer Slide Time: 08:19)

--- 35.2.1

A good choice of an admissible functions for this problem can be of the form as shown in Eq.
35.2.1. One can always think of such construction of admissible functions but before we go
for the general solution, let us look at only one term expansion and determine the Rayleigh
quotient that will give us the fundamental frequency as discussed in Slide 35.2. The
fundamental frequency (non-dimensional) shown is kept as a function of radius of circular
support, r = a. Now, if we want to know what should be a good radius of circular support so

294
that the plate is well-supported. In that case, what essentially, we are asking is what radius
will give a very high natural frequency as high natural frequency would mean the support is
quiet stiff. Therefore, we can make the natural frequency stationary with respect to r = a and
that gives a value of optimal radius (r = aopt) as shown in Slide 35.2 and corresponding to this
optimal radius, we get maximum frequency that can be checked by taking the second
derivative. Let us now go back to our Ritz discretization problem and now we can take this
optimal radius and discretize the problem.

Slide 35.3: (Refer Slide Time: 17:12)

We substitute the expansion form in the Hamilton’s principle as shown in Slide 35.3. Thus,
the discretized equation is obtained as shown. If we perform the modal analysis of the
discretized equation, then we obtain the natural frequencies as shown. Note that there are two
frequencies that are the same. Thus, there is a modal degeneracy. One can also check that the
eigenvectors corresponding to these equal eigenfrequencies are distinct. The figures in Slide
35.4 show the modes of vibration of the plate where the non-dimensional optimal radius of
circular support which happens to be around aopt = 0.8 is shown. The first plot is the
fundamental mode and plot number 2 and 3 are the degenerate modes. In degenerate modes,
one can see that plot 3 is a clockwise rotated (by π/2) version of plot 2. Now, it may be of an
interest to see what happens when radius of circular support changes.

295
Slide 35.4: (Refer Slide Time: 21:11)

1 2

3 4

Slide 35.5: (Refer Slide Time: 22:20)

In the figure shown in Slide 35.5, the variation of the first few non-dimensional circular
eigenfrequencies with radius of circular support (r = a) is shown. Here, you can see a
maximum value of frequency at certain value of radius. Thus, we have seen how we can
optimally support a square plate. Now, one can attempt any other form of plate with a circular
support and try to find out the optimal support radius that gives the maximum stiffness to the
structure. This, in fact is a very important problem in structural engineering. The next

296
problem that we are going to look at is of the vibrations of a plate having beam stiffeners at
the edges (Slide 35.6).

Slide 35.6: (Refer Slide Time: 24:00)

We consider a stiffened square plate undergoing transverse vibrations. We consider that the
plate is clamped on two opposite edges while the other two edges are free but stiffened by
beam stiffeners as shown. The side length of the square plate is 2a. Let us now look at the
kinetic energy of the plate that is composed of the kinetic energy of the plate and the kinetic
energy of two beams. We assume this plate to be a Kirchhoff plate while the beams to be
Euler-Bernoulli beams. The kinetic energy of the plate is obtained as shown. The potential
energy, once again will be the sum of the potential energies in the plate and in the beams as
shown in Slide 35.6.

Now, once again to simplify these expressions, we use a non-dimensionalization scheme as

discussed in Slide 35.7. The Lagrangian now reads as shown. Here, we once again use the
Ritz discretization. We use the expansion form and admissible functions that needs to satisfy
only the geometric BCs as shown in Slide 35.7. In the present problem, we have the
geometric boundary conditions only on the clamped edges where the displacement and the
slope are zero. Note the form of admissible function satisfying the geometric boundary
conditions. Now, we use this expansion in the Lagrangian, discretize it and finally solve for

297
the eigenfrequencies, which are obtained as the non- dimensional numbers as shown in Slide
35.8.

Slide 35.7: (Refer Slide Time: 31:18)

Slide 35.8: (Refer Slide Time: 36:05)

Now, we can obtain the eigenfunctions from the expansion by calculating the eigenvectors of
the discretized system. In the figures shown in Slide 35.9, first four modes of the plate are
plotted. Here, one can note the clamped edges shown as straight lines. Also, one can note the

298
motion of stiffened edge where the displacement is not very high as can be seen at the centre
of the plate. Had there been no stiffeners then this would have vibrated like a beam and the
displacements here also would have been as high. Due to the stiffening effect of the beam, we
have lower displacement at the beam stiffened edges. The second mode is an asymmetric
mode with one nodal line at the centre. The next higher mode has two nodal lines and in the
next, we have one nodal line parallel to the support where the displacement of the stiffened
edge is seen to be quite low.

Slide 35.9: (Refer Slide Time: 37:05)

In summary of this lecture, we have looked at two special examples of the plate vibrations.
We started off with the variation formulation of plate dynamics. Then we have looked at a
square plate on a circular support and subsequently, an edge-stiffened square plate. It
concludes this lecture and the course as well.

Keywords: Plate dynamics, Variational formulation, Rayleigh quotient, Square plate,

Circular support, Beam stiffeners, Edge stiffened plate

299
 THIS BOOK
IS NOT FOR
SALE
NOR COMMERCIAL USE

(044) 2257 5905/08

[Link]
[Link]