University of Sheffield Department of Mechanical Engineering

MEC 308: Applied Noise and Vibration TUTORIAL 1

The objective of this tutorial is to provide practice in forming equations of motion for MDOF
systems and solving the resulting eigenvalue equations to give natural frequencies and mode shapes.
Both questions are from past examination papers. If a part of a question hasn’t been covered in this
year’s lectures, do a little research (preferred) or skip it.

Question 1 (2009-2010)

The TransPennine express train made up of two cars, each weighing 25 000kg, is connected by
couplings of stiffness 30 000 N/m, as shown in Figure 1.

1. Model the train as a lumped parameter system and determine the equations of the motion of the
system, expressing them in matrix notation. [3 marks]

2. What is the type of the coupling in the analysed system? Is the nature of the coupling dependent
on the coordinates used? [3 marks]

3. Obtain the characteristic equation and calculate the natural frequencies in Hz. [9 marks]

4. Calculate the normal modes of the system and display them graphically. What is the type of
motion of each mode? [7 marks]

5. Name three physical and three modal parameters used to describe vibration behaviour of lumped
parameter systems. [3 marks]

Figure 1: The TransPennine Express.

1
Question 2 (2008-2009)

In the study of earthquakes, a building can be idealised as a rigid body of mass m supported by
three springs, two giving translational stiffness k/2 and the other rotational stiffness kT , as shown in
Figure 2. The translation and rotation of the building from the equilibrium position is described by
the x and α coordinates. The mass moment of inertia about its mass centre G is equal to IG . Assume
small vibrations and neglect damping.

1. Draw the free body diagrams of the building. Derive the equations of motion and express them
in matrix notation. [7 marks]

2. Derive the characteristic equation of the building. [6 marks]

3. Briefly explain how the natural frequencies and normalised mode shapes of the building may be
found. [5 marks]

4. How would you define a semidefinite 2-DOF system? Derive the rotational stiffness for which the
building becomes a semidefinite system. [7 marks]

α x

k/2 kT k/2

Figure 2: Earthquake building model.

2
Solutions 1

1. Draw a free body diagram representing the system by lumped masses.

y1 y2

m m
k

For the system shown above the equations of motion can be written as,

mÿ1 + k(y1 − y2 ) = 0

mÿ2 + k(y2 − y1 ) = 0

These can be assembled into matrix form as,

( )( ) ( )( ) ( )
m 0 ÿ1 k −k y1 0
+ =
0 m ÿ2 k −k y2 0

2. The stiffness matrix is nondiagonal and this represents static coupling. The nature of the coupling
does not depend on the choice of coordinates.

3. Assume harmonic motion,

yi = Yi cos(ωt) i = 1, 2

Substituting into the equation of motion gives,

(−mω 2 + k)Y1 − kY2 = 0

−kY1 + (−mω 2 + k)Y2 = 0

The characteristic equation is then obtained by setting the determinant in the matrix equation
for Y1 and Y2 to zero, that is,

ω 2 (m2 ω 2 − 2km) = 0

The solution of the above equation gives the two natural frequencies,
√
2k
ω1 = 0, ω2 =
m

Substituting the values of m and k given in the question gives f1 = 0 Hz and f2 = 0.25 Hz.

3
4. Substituting ω1 and ω2 into the first and second equations of motion respectively, yields,

kY1 − kY2 = 0
2k
(−m × + k)Y1 − kY2 = 0
m
from which the amplitude ratios in the two modes can be obtained (using the numerical values
from the question) as,

Y2 Y2
= 1, = −1
Y1 Y1

So the modes of vibration are (1, 1)T and (1, −1) as depicted below. As the first mode has
zero natural frequency it represents a rigid body motion, whereas the second mode is a normal
harmonic motion.

1 1 1

−1

5. Mass, damping, stiffness (physical); modal mass, modal damping, modal stiffness, natural fre-
quency etc. (modal).

4
Solutions 2

1. Separate free body diagrams are needed for the translational and rotational components of the
motion.

..
IG α

G G
.. ..
m(x + hα)

mg
kT α
kx

The equations of motion for small α are,

m(ẍ + hα̈) = −kx

IG α̈ + (ẍ + hα̈)mh = −kT α + mghα
which become, in matrix form,
( )( ) ( )( ) ( )
m mh ẍ k 0 x 0
+ =
mh IG + mh2 α̈ 0 kT − mgh α 0

2. Assume harmonic motion,

α = A1 sin ωt, x = A2 sin ωt

Substituting into the equations of motion gives,

( )( ) ( )
−mhω 2 k − mω 2 A1 0
=
(mh + IG )ω + mgh − kT
2 2
mhω 2 A2 0

Setting the determinant of the matrix to zero to give a nontrivial solution generates the charac-
teristic equation,

mIG ω 4 − ω 2 (mkh2 + IG k − m2 gh + mkT ) − mghk + kkT = 0

5
3. For the natural frequencies, the solution of the characteristic equation leads to the values ω1 and
ω2 .
For the mode shapes, for each ω calculate the ratio A2 /A1 and normalise accordingly.

4. The system is semidefinite when one of the natural frequencies is equal to zero. This happens for
the analysed system when mghk = kkT .

6
University of Sheffield Department of Mechanical Engineering

MEC 308: Vibrations and Acoustics TUTORIAL 2

The objective of this tutorial is to provide further examples of calculations, but also to introduce
a couple of important ideas in dynamics. The first question shows how one can simplify equations of
motion by scaling of variables. The second question introduces the idea of a type of transmissibility.

Question 1

By introducing scaled versions of the dependent and independent variables for the SDOF system,

mÿ + cẏ + ky = x(t)

show how the equation can be put in a ‘canonical’ form which only contains one physical constant.

Question 2

Figure 1 shows a very basic schematic for a vibration sensor. The essential idea is that, when
the sensor is fixed to a surface moving with a time-dependent displacement x, the internal mass will
vibrate with a displacement response y and the relative displacement z = y − x can be measured. The
measurement is usually made by making the internal mass into a magnet and placing it inside a coil
fixed relative to the casing; the voltage in the coil will then be proportional to the relative motion.
Derive the equation of motion for z in terms of the base displacement x and determine the ’FRF’
between Z(ω) and X(ω). Express the results in terms of the physical constants and in terms of the
damping ratio and natural frequency. Comment on the results.

y k

m
x
c

Figure 1: Schematic for vibration sensor.

1
Solution 1

We start with the standard SDOF equation of motion,

mÿ + cẏ + ky = x(t)

We observe that we can rescale the two dependent variables x and y and the independent variable
t,

u = αy, p = βx, τ = γt

Now, one observes that,

d 1 d
=
dτ γ dt
If one uses dashes to denote differentiation with respect to τ , substituting in all the scaled variables
leads to,

mγ 2 ′′ cγ ′ k 1
u + u + u= p
α α α β
The first constant is easy to remove, one simply sets β = 1, and then,

mγ 2 ′′ cγ ′ k
u + u + u=p
α α α
The next constant is removed by setting α = k, so that,

mγ 2 ′′ cγ ′
u + u +u=p
k k
√
Finally, we can remove the ‘inertia’ constant by setting γ = k/m = ωn ,
cωn ′
u′′ + u +u=p
k
√
cωn c k √c
Now, note that k
= k m
= km
= 2ζ, so finally we have,

u′′ + 2ζu′ + u = p

and only the damping ratio is visible, all other quantities have been removed by the rescaling.

2
Solution 2

A fairly basic free body diagram (I’m not going to draw it), gives,

mÿ = −c(ẏ − ẋ) − k(y − x)

so, in terms of the relative displacement z = y − x, one has,

mz̈ + cż + kz = −mẍ

as the equation of motion for the process x −→ z.

The ‘FRF’ for this process is just the usual one but with Z(ω) taking the role of the output spectrum
and mω 2 X(ω) taking the role of the input spectrum. If we denote the ratio Z(ω)/X(ω), by T (ω), we
immediately see that,

mω 2
T (ω) =
−mω 2 + icω + k
Now, dividing top and bottom by k, gives,
m 2
k
ω
T (ω) =
− k ω + i kc ω
m 2
+1

A little algebra shows that c/k = 2ζ/ωn , so that, also substituting k/m = ωn2 yields,
ω2
2
ωn
T (ω) =
1− ω2
2
ωn
+ 2iζ ωωn

In terms of ‘dimensionless frequency’ n = ω/ωn , one finally has,

n2
T (ω) =
1 − n2 + 2iζn
Note that this makes sense because T is a ratio of quantities with the same dimensions and is thus
dimensionless. Just as in Question One, we have arrived at an expression of the basic physics with
only the damping appearing explicitly. Using the dimensionless frequency is the analogue of scaling the
time variable in Question One. You should explore the relationship between the two questions here.
An ’FRF’ between two response variables in a problem is often called a transmissibility. Note that the
transmissibility here is a little different as the standard definition would be for the ratio Y (ω)/X(ω)
You can consult Thompson for more details about transmissibilities and also about the subtleties
of sensor design.

3
University of Sheffield Department of Mechanical Engineering

MEC 308: Applied Noise and Vibration TUTORIAL 3

The objective of this tutorial is to provide another example of using Rayleigh’s method for estimating
the fundamental natural frequency of a structure; the second question shows how the Rayleigh-Ritz
method extends Rayleigh’s method.

Question 1

Use Rayleigh’s method to estimate the first natural frequency of the tapered beam shown below.
Assume unit width for the beam.

O h
x

Figure 1: Tapered cantilever beam.

Assume a trial mode shape of the form,

( )2
x
W (x) = 1 −
L
Question 2

Use the Rayleigh-Ritz method to estimate the first two natural frequencies of the tapered beam
considered in Question 1. You may assume basis functions of the form,
( )2
x
W1 (x) = 1 −
L
( )2
x x
W1 (x) = 1−
L L

1
Solution 1

The proposed trial function for the mode shape is,

( )2
x
W (x) = 1 −
L
First of all, you check that the function satisfies the appropriate boundary conditions e.g. W (L) = 0,
dW/dx(L) = 0.
Rayleigh’s quotient requires one to specify the cross-sectional area A(x) and second moment of area
I(x). You should establish that,

hx
A(x) =
L
( )3
1 hx
I(x) =
12 L
Then, Rayleigh’s quotient is,
∫L ( )( )2
h3 x3 2
0 E 12L3
dx Eh2
L2
ω2 = ∫ ( ) ( )4 = 2.5 4
L hx
1 − Lx dx ρL
0 ρ L

or,
√
Eh2
ω = 1.5811
ρL4
In this case, the exact value for the frequency is known to be,
√
Eh2
ω = 1.5343
ρL4

2
Solution 2

The trial basis is,

( )2
x
W1 (x) = 1 −
L
( )
x x 2
W1 (x) = 1−
L L
So a two-term approximation to the beam mode shape is,
( )2 ( )2
x x x
W( x) = c1 1− + c2 1−
L L L
Rayleigh’s quotient is,
Vmax
ω2 = ∗
Tmax
where,
∫ ( )2
L d2 W (x)
Vmax = EI(x) dx
0 dx2
∫ 2
∗
Tmax = ρA(x)W (x)2 dx
0
where A(x) and I(x) are defined as for Question 1.
Subsituting the mode shape expansion W (x) into the equations above gives,
( )
Eh3 c21 c2 c1 c2
Vmax = + 2 +
3L3 4 10 5
( )
∗ c2 c2 2c1 c2
Tmax = ρhL 1 + 2 +
30 280 105
so the Rayleigh quotient is clearly a function of c1 and c2 .
The conditions that make the quotient stationary are,
∂ω 2
=0
∂c1
∂ω 2
=0
∂c2
As discussed in the notes, these lead to,
∂Vmax ∂T ∗
− ω 2 max = 0
∂c1 ∂c1
∂Vmax ∂T ∗
− ω 2 max = 0
∂c2 ∂c2

3
 ∗
Substitution of the expressions for Vmax and Tmax leads to the matrix equation,
 ( ) ( ) ( ) ( )
1
− ω̂ 2 15
1 1
− ω̂ 2 105
2
c1 0
 (2 ) (5 )  =
1
2
− ω̂ 2 2
105
1
5
− ω̂ 2 140
1 c2 0

where,
( )
3ρL4
ω̂ 2 = ω2
Eh2

Setting the determinant of the matrix to zero gives the characteristic equation for the natural
frequencies,
1 4 13 2 3
ω̂ − ω̂ + =0
8820 1400 50
The roots of the above equation are found to be ω̂1 = 2.6599 and ω̂2 = 8.6492 and these lead to
estimates of the natural frequencies,
√
Eh2
ω1 = 1.5367
ρL4
√
Eh2
ω2 = 4.9936
ρL4
Note that the first of these is a better approximation to the exact first natural frequency than the
solution found by Rayleigh’s method to Question 1.