Vibration Analysis Lectures

Indian Institute of Technology Kanpur
Lecture 2
Basic Vibratory Phenomena
Simple Mechanical System
Physical System Mathematical Model

Can be a Modal Model
WHAT IS A DOF ?
• The ability to move in any one direction is a DOF.

• The number of co-ordinates required to specify the
motion of a system uniquely determines the order of
that system.
1 DOF
6 DOFs
NATURAL FREQUENCY AND TIME RESPONSE
m&x&
Equation of motion : m&x& + kx = 0

Solution : x(t) = A cos ω nt + B sin ω nt = C sin(ω nt + φ )
k
where ω n = is the natural frequency.
m
DISPLACEMENT, VELOCITY & ACCELERATION - I
The constants ( A, B ) or (C , φ ) depend on the initial conditions.

For instance, let us assume that x( 0 ) = x0 and x&( 0 ) = x&0
After some algebra, it can be shown that :
x&0
x(t) = x0 cosω n t + sinω n t
ωn
In the general case :
x(t) = A cos ω nt + B sin ω nt = C sin (ωnt + φ )
x&(t) = Cωn cos (ωnt + φ) → x& max = Cωn
&x&(t) = −Cωn 2 sin (ωnt + φ) → &x& max = Cωn
2
DISPLACEMENT, VELOCITY & ACCELERATION - II
DISPLACEMENT
x(t) = C sin (ωnt + φ )
VELOCITY
x&(t) = Cωn cos (ωnt + φ )
ACCELERATION
&x&(t) = −Cωn 2 sin (ωnt + φ )
SPRINGS - I
FORCE F F=kx
F=kx0
Strain energy=1/2
kx02
x0
DISPLACEMENT x
SPRINGS - II
• What provides the spring (or restoring)

force?
• Gravity
• A spring
• A float mechanism
• Beam flexibility
STIFFNESS FORMULAE
POSITION OF EQUILIBRIUM - I
IN ALL VIBRATION PROBLEMS, THE ORIGIN OF THE

MOTION SHOULD BE TAKEN AS THE STATIC
EQUILIBRIUM POSITION.
POSITION OF EQUILIBRIUM - II
Weight mg
Static deflection: δ = =
stiffnes k
If the vibration amplitudeis x, total deflection will be :
mg
y = x +δ = x +
k
Newton's second law : m&x& = ∑ F
mg
m&x& = mg − ky = mg − k ( x + )
k
m&x& = −kx
TRANSLATIONAL, TORSIONAL & ROTATIONAL SDOF SYSTEMS

PENDULUM WITH SPRING
We will apply : Iθ&& = ∑ Moment

Spring restoring moment :
a Stiffness x displacement x distance = − k (a sin θ )a
Gravity restoring moment :
L Effective weight x distance = −mgsinθ L
θ Re - arranging :
Iθ&& = −k (a sin θ )a − mgsinθ L
m
Re − arranging and assuming small θ
mg &&
Iθ + (ka + mgL)θ = 0
2
(ka 2 + mgL)
ωn =
Note: I=mL2 I
TORSIONAL SYSTEM
kT θ I=0.16 kgm2
Torque T
Torsional stiffness = kT = =
Ang disp θ
D=10 mm
T Gθ πr 4
L= 300 mm Also = where J =
J L 2
G= 80 GPa Combining : kT = GJ / L = 261Nm / rad
kT 261
Iθ&& + kTθ = 0 ωn =
I
=
0.16
= 40.4 rad / s
KINETIC ENERGY vs STRAIN ENERGY
1 2
KE = mv
2
1 2
SE = kx
2
v
ENERGY METHOD TO DETERMINE ωn
MAX KINETIC ENERGY = MAX STRAIN ENERGY
Max velocity, Max displacement,

zero displacement zero velocity
Displacement : x(t ) = x0 sinωnt Max = x0
Velocity: x&(t ) = x0ωn cosωnt Max = x0ωn
x&MAX → x0ωn AND xMAX → x0
FOR THE SDOF MASS-SPRING SYSTEM
1 1 2
m( x0ωn) = kx0 Hence:ωn = k
2
2 2 m
MORE COMPLEX SYSTEMS
GENERAL LAGRANGE EQUATION
d  ∂T  ∂T ∂V ∂D
  − + + = Qi
dt  ∂x&i  ∂xi ∂xi ∂x&i
VERY COMPLEX SYSTEMS Finite element model of a car

Simple Vibratory Phenomena
External Force – Time dependent usually periodic

One harmonic – rotational or multiples is sufficient
Mass times Acceleration

opposite to acceleration
Damping coefficient times

Velocity – dissipates energy
Mean Equilibrium position

under self weight
Static Deflection d = mg/k
Linear System
Inertia Force proportional to accln
Damping force proportional to vel
Stiffness force proportional to displ
System that can be described by one coordinate, say, x

Torsional Vibration
Angular vibrations of any drive train

• A serious problem in reciprocating machines limiting
the speeds, multi cylinder engines are better to even out
the highly pulsating torque
• Usually not a problem in rotating machinery as the drive
torque is fairly uniform
• Torsional vibrations can be very severe under suddenly
applied loads, e.g., rolling mills, electrical short circuit
conditions … Under these conditions, couplings, gear
boxes … are susceptible for failures
• Choose proper coupling to make it work well under d 2θ
normal conditions and act as a fuse under severe loads I 2 + kθ = 0
and protect the machinery dt
• Torsional (Angular) stiffness Nm/rad k
p= rad/s
• Mass moment of inertia Kg-m2 I
Bending Vibration
Bending (Flexural) vibrations of a drive train

• Most common problem in all rotating and
reciprocating machinery
• All heavy duty machinery operate above first
critical speed
• A small unbalance (residual balance or
imbalance) can cause serious problems at
critical speeds
• Bending vibrations effected by
misalignment, loosely mounted parts, bearing
stiffness, gears, asymmetry, instabilities due to
oil film, etc… They cause most machinery
problems
FREE MOTION OF A DAMPED SDOF SYSTEM
Equation of motion : m&x& + cx& + kx = 0

Dividing by m : &x& + 2ζωnx& + ω 2n x = 0
k
where ωn = is the undamped natural frequency,
m
c c
ζ = = is the viscous critical damping ratio.
c0 2 km
SOLUTION OF THE EQUATION OF MOTION
EOM : &x& + 2ζωnx& + ω 2n x = 0

α1t α 2t
Solution is of the form : x(t ) = Ae + Be
A & B are two constants depending on initial
conditions.
αt
To find α1 & α 2 , insert x(t) = Ae into EOM.
αt
Ae (α + 2ζωnα + ω )=0
2 2
n
→ α1, 2 = (−ζ ± ζ − 1)ωn 2

POSSIBILITIES, DEPENDING ON THE VALUE OF ζ
Case 1. ζ < 1. Underdamped case with oscillatory motion.

Both roots are complex : α1 = (−ζ + i 1 − ζ 2 )ωn
α 2 = (−ζ − i 1 − ζ 2 )ωn
The general solution becomes :
− ζωnt
x(t ) = e [ A cos( 1 − ζ 2 ωnt ) + B sin( 1 − ζ 2 ωnt )]
− ζωnt − ζωnt
=e C sin( 1 − ζ 2 ωnt + φ ) = e C sin(ωdt + φ )
where ωd = 1 − ζ 2 ωn is the damped natural frequency.
TIME HISTORY FOR OSCILLATORY MOTION
Exponential term
− ζωnt
e
Oscillatory term
C sin(ωdt + φ )
CASE 2
ζ > 1. Overdamped case with no oscillatory motion.

Both roots are real : α1 = (−ζ + ζ 2 − 1)ωn
α 2 = (−ζ − ζ 2 − 1)ωn
The solution becomes :
(−ζ + ζ 2 − 1)ωnt
x(t ) = Ae
(−ζ − ζ 2 − 1)ωnt
+ Be
EFFECT OF OVERDAMPING
CASE 3
ζ = 1. Critically - damped motion

→ Max rate of decay
Double root : α1 = α 2 = −ωn
The solution becomes :
x(t ) = ( A + Bt )e − ω nt
EXAMPLE: SYSTEM IDENTIFICATION FROM TIME RESPONSE
x(t)
1.0 s 2.5 s 4.0s
4000 kg
t
Find k and c.
Initial conditions :
x& ( 0 ) = 0
T=2π/ωd=3
x (1 ) = 0 , x ( 2 . 5 ) = 0
s
x(4) = 0
ωd= 2π/3
LOGARITHMIC DECREMENT - I
The amplitude ratio between two successive cycles :

− ζωntm
x(tm) e C sin(ωdtm + φ )
=
x(tm + 1) e− ζωntm + 1C sin(ωdtm + 1 + φ )
x(tm) − ζωn(tm − tm + 1)
=e
x(tm + 1) AMP
2π
2π ζωn
ζωn 1 − ζ 2 ωn
ζωnT ωd = e
=e =e
LOGARITHMIC DECREMENT - II
2π
ζ
x(tm) 1−ζ 2
=e
x(tm + 1)
Taking logarithms of both sides :
x(tm) 2πζ
δ = ln = ≈ 2πζ
x(tm + 1) 1−ζ 2
If the two amplitudes are separated by (N - 1) cycles :
1 x(tm)
δ = ln
N x(tm + N)
Viscous Damping Principle
• Damping force is proportional to velocity and

dv
τ =µ = Damping Coefficient C times Velocity dx/dt
dz
– dissipates energy
v
F = (πDt ) µ = cx& • Dashpots can de designed as in shock
δ
absorbers or the equivalent effect of energy
πDtµ
c= N - s/m dissipating capacity determined from tests to
δ
find the value of this coefficient c
DAMPING IS NOT ALWAYS VISCOUS !
• Viscous damping ratio: ζ = c/c0

m&x& +cx& + kx = 0
• Material damping (what the material can dissipateDashpot
in one
cycle)
m &x& + k (1 +iη ) x = 0
•Aerodynamic damping: δ (e. g. gas pressure on a blade)
• Friction damping (energy dissipation via contact mechanism)
AT RESONANCE
ζ =η / 2 = δ / 2π
EXAMPLE OF DAMPED MOTION - PARACHUTE
Find the maximum compression in the

spring if m=20 kg, k=10kN/m, c=540Ns/m
and v=8 m/s.
Compression = static compression + dynamic compression
weight mg 20 x9.81
d1 = = = = 0.0196m
stiffness k 10 x1000
d2
USE INITIAL CONDITIONS TO FIND A & B IN EOM:

t=0
d
1 Datum for motion
xo=-d1=-0.0196 m & v0= 8m/s
− ζωnt
x(t ) = e ( Acosωdt + B sinωdt )
x(0) = A = x0 = −0.0196m
x
− ζωnt
Obtain velocity by differentiating : x(t ) = e ( A cos ωdt + B sin ωdt )
− ζωnt − ζωnt
x& (t ) = −ζωne ( A cos ωdt + B sin ωdt ) + ωde (− A sin ωdt + B sin ωdt )
− ζωnt
=e [( Bωd − Aωnζ ) cos ωdt − ( Aωd + Bωnζ ) sin ωdt ]
→ x& (0) = Bωd − Aωnζ = 8m / s
Substitution for A, ωn , ωd and ζ gives B = 0.434
LAGRANGE’S EQUATION OF MOTION
0
d  ∂T  ∂T ∂V ∂D
  − + + = Qi
dt  ∂x&i  ∂xi ∂xi ∂x&i
T : Kinetic energy x is some general arbitrary co-
ordinate.
V : Potential & spring energy
D : Damping dissipation function
Q : External force
i : Co - ordinate number (i = 1 for SDOFsystems)
Free - vibration : Q = 0 SDOF d  ∂T  ∂V
No damping : D = 0  + =0
dt  ∂x&  ∂x
EXAMPLE 1 – SDOF SYSTEM
FIND THE EQUATION OF MOTION
1 2 ∂T d  ∂T  d
T = mx& → = mx& →   = (mx& ) = m&x&
2 ∂x& dt  ∂x&  dt
1 2 ∂V
V = kx → = kx
2 ∂x
d  ∂T  ∂V
 + = 0 → m&x& + kx = 0
dt  ∂x&  ∂x
EX 2 - PENDULUM WITH SPRING
1 &2
T = Iθ
a 2
D=0
V = ∆SE + ∆PE where ∆ : Change
L
SE : Spring energy, PE : potential energy
θ
1 2
m SE = ky where y = aθ
mgLcosθ 2
mg PE = mgL(1 − cosθ )
L 1
V = k (aθ ) 2 + mgL(1 − cosθ )
2
OBTAIN DERIVATIVES
1 &2
T = Iθ
2
∂T d  ∂T  d &
→ & = Iθ →  &  = ( Iθ ) = Iθ&&
&
∂θ dt  ∂θ  dt
1
V = k (aθ ) 2 + mgL(1 − cosθ )
2
∂V
→ = ka θ + mgL sin θ ≈ (ka + mgL)θ
2 2
∂θ
INSERT INTO LAGRANGE EQUATION
d  ∂T  ∂V
  + = 0
dt  ∂θ&i  ∂θ i
&&
Iθ + (ka + mgL)θ = 0
2
(ka + mgL)
2
ωn =
I
FORCED VIBRATION
EOM : m&x& + cx& + kx = F (t )

(2nd - order ODE with RHS)
Solution : x(t) = CF + PI
xCF(t) xPI(t) Steady-state
Transient
Periodic solution
x(t)=xCF(t)+ xPI(t)
SOLUTION OF m&x& + cx& + kx = F (t )

If we consider ζ < 1 only , the CF is given by :
− ζωnt
x(t ) = e ( A cos ωdt + B sin ωdt )
To obtain the PI, we must know the RHS, F(t).

We will consider one type of excitation only :
m&x& + cx& + kx = F (t ) = F0 sin ωt
We now need to guess a PI.
PROPERTIES OF THE PI
When a linear system is subjected to a harmonic

excitation of the form Fsinωt,
• It will respond harmonically at the same frequency.
• There will be a phase lag between the force and the
response.
Input : F (t ) = F 0sin ωt 0<ω < ∞

Output : xPI (t ) = x0 sin(ωt − φ )
PHASE LAG
φ
100 100
F (t ) = F0 sin ωt
50 50
0 0
x(t ) = x0 sin(ωt − φ )
-50 -50
Tim
-100 -100
0 0.1 e 0.2 0.3
SOLUTION FOR THE STEADY-STATE VIBRATION
The solution for the steady − state vibration can be found by inserting
the PI xPI (t ) = x0 sin(ωt − φ ) into the EOM m&x& + cx& + kx = F0 sin ωt
F0 cω
x0 = & tan φ =
(k − mω 2 ) 2 + (cω ) 2 k − mω 2
ω c
Let r = and ζ =
ωn 2 km
x0 1 2ζr
= & tan φ =
F0 k (1 − r 2 ) 2 + (2ζr ) 2 1− r 2
FREQUENCY RESPONSE FUNCTION (FRF)
Consider the expression :

x0 1 1
H= = =
F0 k (1 − r ) + (2ζr )
2 2 2
(k − mω 2 ) 2 + (cω ) 2
It is of the form :
Output
= Function (k, m, c & ω ) = Function (system properties & ω )
Input
Such a function is called Transfer Function in general

It is called Frequency Response Function (FRF) in
vibration analysis.
FRF PLOT
φ=900 at resonance
I
RESONANCE
EFFECTS OF DAMPING
Reduces
response at
resonance.
Has little effect

elsewhere.
Has relatively
little effect on
resonant
frequency.
Another look at RESONANCE

Another look at RESONANCE……contd…




Q-FACTOR
Resonant amplitude  x0 
Q= = 
Static deflection  d r = 1
F0 F0
We have : x0 = & d=
k (1 − r 2 ) 2 + (2ζr ) 2 k
   
 x0   x0   1 
Q=  =  =
 d  r =1  F0   (1 − r 2 ) 2 + (2ζr ) 2 
 k  r =1   r =1
x  1
→Q = 0  = Inverse of Damping Ratio x 2
 d  r =1 2ζ
HALF-POWER POINTS
Half-power points
X0 /d
Frequencies corresponding to
Q
Q
amplitude =
2
Q/1.414
are called half − power points.
1.0
ω
r =1
DAMPING ESTIMATION FROM AN FRF
H Hres x0 1
Hres H = =
H (ω1) = H (ω 2) = F ( k − mω 2 2
) + ( cω ) 2
2 0
1
ω1 Hres = at ω = ωn = k / m
ω cωn
n
ω2 Determine ω such that H (ω ) = Hres
2
ω
1 1 ω 2
→ = → Yields a quadratic equation in ( )
(k − mω ) + (cω )
2 2 2
2cωn ωn
Hres
There are 2 frequencies ω 1 &ω 2 such that H (ω 1) = H (ω 2) =
2
ω 2 − ω1 c 1
It can be shown that ≈ = 2ζ =
ωn km Q
PHASOR DIAGRAM FOR F=F0sinωt
F0 is rotating with speed ω
50
F 0 sinωt F0 ωt
-50
0 0.04 0.08 0.12
ωt = 2nπ ωt = 2nπ + π ωt = 2nπ + 2π

RELATIVE PHASE BETWEEN DISP, VEL & ACCEL
x = x0 sin(ωt − φ )
π
x& = ωx0 cos(ωt − φ ) = x0ω sin(ωt − φ + )
2
π
ie x& leads x by
2
&x& = −ω 2 x0 sin(ωt − φ ) = −ω 2 x = ω 2 x0 sin(ωt − φ + π )
π
ie &x& leads x& by , and x by π . &x&
2
x&
x ωt-φ
3 CASES TO CONSIDER
mω2 x0
cωx0 ω < ω n -> Kx0 > mω2x0
F0 kx0
Stiffness control
mω 2 x0
cωx0 ω= ω n -> Kx0 = mω2x0
kx0
F0 Damping control
mω2 x0
cωx0 ω> ω n -> Kx0 < mω2x0
kx0
F0 Inertia control
EXCITATION BY ROTATING OUT OF BALANCE
(m + m' ) &x& + cx& + kx = F0 sin ωt

where F0 = m' rω 2
m’rω2cosωt Usually m >> m'

Example : Find x0 when ω >> ωn
m’rω2sinωt
2
F0 m' rω
x0 = →x0 =
(k − mω ) + (cω )
2 2 2
[k − (m + m' )ω 2 ]2 + (cω ) 2
m' rω2 m' rω2 m' r m' r
lim x0 = lim = lim = ≈
ω →∞ (m + m' )ω m + m' m
2
ω →∞ [k − (m + m' )ω ] + (cω )
2 2 2
ω →∞
VIBRATION ISOLATION
CAR ENGINE
• Large forces over a wide

frequency range.
• They arise from the crank-
connecting-rod-piston system
and combustion process.
• If transmitted to the car body,
severe noise and vibration in
the passenger compartment.
• The engine is therefore
mounted on rubber blocks.
TYPICAL ISOLATION MOUNTS
UNDAMPED
SPRING MOUNT
DAMPED SPRING
MOUNT
PNEUMATIC
RUBBER MOUNT
HIGH-SPEED PUNCH PRESS ON RUBBER MOUNTS
MOUNTS
BASIC THEORY
Source of
FT sin(ωt+θ) vibration
The force transmitted to ground is due to the spring and damper :

FT = kx + cx&
Force transmitted FT
What we want to know is the ratio : =
Excitation Force F 0
mω x02
From the phasor diagram :
cωx0F = x0 k 2 + (cω ) 2
φ FT T
F0 ωt-φ From Lecture 4, we know that :

kx0 x0 =
F0
(k − mω 2 ) 2 + (cω ) 2
So, transmissibility, T, is given by :
FT k 2 + (cω ) 2 1 + (2ζ r ) 2
T= = =
F0 (k − mω ) + (cω )
2 2 2
(1 − r 2 ) 2 + (2ζ r ) 2
ω c
where r = and ζ =
ωn 2 km
HOW TO OBTAIN LOW TRANSMISSIBILITY ?
FT k 2 + (cω ) 2
T= =
F0 (k − mω 2 ) 2 + (cω ) 2
• WE WANT T TO BE AS LOW AS POSSIBLE.

• T IS SMALL IF ω>>ωn.
• WE WANT TO LOWER ωn
• WE WANT LOW STIFFNESS AND/OR HIGH MASS.
VARIATION OF F0 WITH ω
ω<ω mω 2
x0 ω=ωn mω 2 x0
n
FT cωx0 FT cωx0
F0
F0
kx0 kx0
ω>ωn
mω2 x0 ω >> ω n mω2 x0
cωx0
FT cωx0
F0 FT
kx0 F0
kx0
TRANSMISSIBILTY FT/F0 TRANSMISSIBILITY CURVES
AMPLIFICATION T> 1 ISOLATION T< 1

EFFECT OF DAMPING IN THE ISOLATION REGION
TRANSMISSIBILTY FT/F0
DECREASING
T
LESS ISOLATION WITH INCREASING DAMPING
EXAMPLE: ISOLATION OF RAIL NOISE
RUBBER
PADS
Before isolation
After isolation
Lecture –3
Multi-Degree of Freedom Systems

+ Modal Analysis
MDOF VIBRATING SYSTEMS
• Two complications
– Each node has 6 degrees of freedom
– Many such nodes are needed to describe the geometry of
representative engineering systems.
FOR FE models that use 6 DOF/node elements

Total no of DOFs = No of nodes x 6
EXAMPLES OF SDOF AND 2-DOF SYSTEMS
2-DOF SYSTEMS
SDOF
SYSTEM
Multi Degree of Freedom Systems

• Real life systems are complex, they can bend, twist and elongate in
axial direction, the mass is distributed, not discrete as assumed in
the simple models, similarly, elasticity is distributed, there are no
perfect springs without mass …
• In reality we have infinite degrees of freedom in a system, for

convenience, we can model them as finite degrees of freedom
systems.
• The methods of modeling have been refined over the years

depending on the computational facilities available at respective
times.
• We will illustrate some methods that allowed us to understand the

way real life practical systems behave and derive (rather study)
some properties of significance to practical vibration engineers and
diagnostics.
DETAILED ANALYSIS OF A 2-DOF SYSTEM
STEP 1: SPRING DEFLECTIONS
REFERENC
E
k2 force k2 force
DEFLECTED
k1 force
ELONGATION OF k1: x1 COMPRESSION OF k2: (x1-

x2) DUE TO DISPLACEMENT
OF BOTH ENDS
STEP 2: FREE-BODY DIAGRAMS
k1x1
k2(x1-x2) k2(x1-x2)
m1 m2
x x2
1
STEP 3: OBTAIN THE EQUATIONS OF MOTION
k1x1 k2(x1-x2) k2(x1-x2)

m1 m2
x1 x2
Newton' s 2nd law : m&x& = ∑ F

Mass 1 : Mass 2 :
m1&x&1 = −k 1 x1 − k 2( x1 − x 2) m 2 &x&2 = k 2( x1 − x 2)
m1&x&1 + (k 1 + k 2) x1 − k 2 x 2 = 0 m 2 &x&2 − k 2 x1 + k 2 x 2 = 0
STEP 4: ASSUME SHM WRITE THE MATRIX EOM
We have : m1&x&1 + (k 1 + k 2) x1 − k 2 x 2 = 0
m 2 &x&2 − k 2 x1 + k 2 x 2 = 0
Remembering that &x&1 = −ω 2 x1 & &x&2 = −ω 2 x 2
− m1ω 2 x1 + (k 1 + k 2 )x1 − k 2 x 2 = 0
-m 2ω 2 x 2 − k 2 x1 + k 2 x 2 = 0
m1 0  x1  k 1 + k 2 − k 2  x1   0 
−ω 2
   +     =  
 0 m 2  x 2   − k 2 k 2  x 2   0 
STEP 4: CONTINUED
EIGENVALUE = (NATURAL FREQUENCY)2
([K] − ω [M ]) {x} = {0}

2
k 1 + k 2 − k 2  m1 0 
 − k2   x1 
 k 2   0 m2  
   x2 
STIFFNESS MASS MODE SHAPE
MATRIX MATRIX VECTOR
2-DOF system-> 2 modes -> 2 natural frequencies & 2 mode shapes

STEP 5: CHECK SYMMETRY & POSITIVE MAIN DIAGONAL
THE MASS & STIFFNESS MATRICES MUST BE SYMMETRIC.

THE MAIN DIAGONAL ELEMENTS MUST BE POSITIVE.
k 1 + k 2 − k 2  m1 0 
 − k2 k   0 m2
 2   
STEP 6: OBTAIN THE NATURAL FREQUENCIES
FOR LARGE N, THERE ARE MANY NUMERICAL SOLUTION TECHNIQUES.

USE DET=0 FOR SMALL SYSTEMS.
([K] - ω 2 [ M ]) {x} = {0}

→ det ([K] - ω 2 [ M ]) = 0 or {x} = {0}
For a non − trivial solution :
k 1 + k 2 − ω m12
− k2
det ([K] - ω [ M ]) = 0 →
2
=0
− k2 k 2 − ω m2
2
→ (k 1 + k 2 − ω 2 m1)(k 2 − ω 2 m 2) − k22 = 0
Quadratic in ω2 → 2 natural frequencies : ωn1 & ωn2. (ωn1 ≤ ωn2)
STEP 7: OBTAIN THE MODE SHAPES
Insert ωn1 into ([K]-ω2[M]){x}={0}

By definition, det([K]- ωn1 2 [M])=0
x1 & x2 are linearly dependent, but we can obtain x1/x2
Using the previous result : − m1ω x1 + (k 1 + k 2 )x1 − k 2 x 2 = 0
2
x2 k 1 + k 2 − m1ω n21
Hence : =
x1 ω =ωn1 k2
x2 k 1 + k 2 − m1ω 2
Similarly, for the 2nd mode : = n2
x1 ω =ωn 2 k2
MODE SHAPE INTERPRETATION
Assume that, inserting values for m, k, ω gives :

x2 x2
=1 & = −1
x1 ω =ωn1 x1 ω =ωn 2
MODE 1 MODE 2
The masses move in The masses move out of

phase. X1 and X2 move by phase. X1 moves by +1 unit,
+1 unit each. X2 moves by –1 unit.
GENERAL MDOF ANALYSIS
Ignore damping.
Matrix equation of motion :
[M] NxN {&x&}Nx1 + [K] Nx1{x}Nx1 = { 0 }Nx1
{x} = −ω 2{X} gives :
([K] − ω 2 M ]) Nx1{x}Nx1 = { 0 }Nx1
Eigenvalue problem in ω 2 and {x}
N modes → (ω 2 ) Natural frequency
{x} Mode shape
AN N-DOF SYSTEM HAS N MODES.

3 DOF SYSTEM
TIME FREQUENCY
MODAL SUPERPOSITION
MODE 1 MODE 3
MODE 2
TIME DOMAIN
= + +
FREQUENCY DOMAIN
= + +
Torsional System
One of the earliest fatigue failures experienced
is that of the propeller shaft of a steam engine
driven naval ship during the I world war. It was
reported that the propeller shaft (which has the
lowest torsional stiffness in the system because
of its length) upon its failure was stiffened by
increasing its diameter, however, it failed
earlier. Then it was identified that the excitation
became closer to the new natural frequency
causing fatigue failure in lesser time. From then
onwards, torsional analysis became mandatory
for all reciprocating installations.
• Briefly, we will talk about simple modeling adopted for torsional analysis of a reciprocating diesel engine
driving a generator.
• All the 8 cylinders are considered as discs, whose rotational mass moments of inertia can be determined
and connected by equivalent torsional stiffnesses of the crank shaft.
• The damper connected to cylinder 8 is divided into two separate disks.
• The coupling stiffness is usually the lowest when compared to the stiffness of any of the shaft sections in
the train.
• The generator is modeled as one rotor
• A model thus derived (the details to arrive these values is out of current scope) is given in the next slide.
Reciprocating Engine Installation

The system given here has 13
inertias connected by 12
angular stiffnesses, therefore,
we have a system with 13
degrees of freedom.
• We will attempt to understand
the behavior of such a system
and study some important
vibrational terms that are
regularly used in routine
testing and analysis.
• It is suggested that the
mathematical intricacies in this
process may be ignored by an
engineer in the field – we will
emphasize the physical
concepts that are of concern to
field vibration engineers and
just brush aside the
mathematical stuff (unless, of
course you are otherwise
interested)
I1θ&&1 + k1θ1 − k1θ 2 = 0

• These are governing differential
equations of motion written from I 2θ&&2 − k1θ1 + ( k1 + k 2 )θ 2 − K 3θ 3 = 0
equilibrium conditions, 13 for 13 inertias ...
• These equations are written in a compact I13θ&&13 − k12θ12 + K13θ13 = 0
matrix form
• Mass Matrix [M ]{θ&&}+ [K ]{θ } = 0
• Stiffness Matrix
 I1 
 I2 
[M ] =  I3

 
 I13 

 k1 − k1 
− k k1 − k 2 
[M ] =  1
− k 2 k 2 + k3 − k3 ...

− k12 

 − k12 k12 

Free Vibration - Mode Shapes
Let us assume that the system vibrates at a natural θ1 = 1

frequency p and when it does, each disk has a  θ 
specified amplitude, capital theta with a subscript  2 
denoting the disk number. This assumption gives p = p1 {θ }1 =  θ3 
what is called an eigen value problem, given below.  ... 
θ i = θ i cos pt i = 1,2,...13  
 θ13 1
[[K ] − p [M ]]{θ } = 0
2
θ1 = 1
[[K ] − p [M ]] = 0
2
 θ 
 2 
• On expansion the above gives, a thirteenth degree p = p2 {θ }2 =  θ3 
polynomial equation and therefore, thirteen natural  ... 
frequencies p1, p2, .. p13. This shows, a n degree of  
freedom has n natural frequencies.  θ13  2
• Each frequency gives a specific pattern for the thirteen
amplitudes, with any one amplitude arbitrarily fixed, θ1 = 1
for example one unit for the first disk  θ 
 θ11 θ12 θ13 θ1N   2 
[θ ] =
θ
 21 θ 22 θ 23 θ 2 N 

p = p13 {θ }13 =  θ3 
 θ 31 θ 32 θ 33 θ 3 N   ... 
 
θ N 1 θ N 2 θ N 3 θ NN   
 θ13 13
Arranging each shape in corresponding columns, we get the modal matrix.
Bending Vibration
α11m1 &x&1 + α12 m2 &x&2 + ... + α1n mn &x&n + x1 = 0

α 21m1 &x&1 + α 22 m2 &x&2 + ... + α 2 n mn &x&n + x2 = 0
...
α n1m1 &x&1 + α n 2 m2 &x&2 + ... + α1n mn &x&n + xn = 0
[α ][M ]{&x&} + [I ]{x} = 0
[M ]{&x&} + [α ]−1 [I ]{x} = 0
[M ]{&x&} + [K ]{x} = 0
• The conclusions in the previous slide are not restricted to
torsional vibrations alone, they are true for all kinds of
vibration, bending, torsion, axial, combined bending and
torsion etc.
• Here, in this slide we show how the same equations can be
derived for bending, by using influence coefficient approach.
• In all vibration problems, the first thing is to set up a
workable
mathematical model, write the eigen value formulation,
determine the natural frequencies and mode shapes.
• Modern FE codes, ANSYS, NASTRAN …use finite
elements, make a CAD model, mesh and ask for the natural
frequencies and mode shapes.
FORCED RESPONSE
THE RESPONSE OF THE SYSTEM TO SOME GIVEN

HARMONIC EXCITATION CAN BE FOUND USING A
TRANSFER FUNCTION APPROACH:
OUTPUT = SYSTEM FUNCTION x INPUT
X = H(system properties, ω) x Force
For forced response, we have : ([K]-ω2 [M]) {X} = {F}
→ { X } = ([K] - ω 2 [ M ]) −1{F } = [ H ]{F }
WE WANT THE NORMALIZED RESPONSE TO A SINGLE
EXCITATION, APPLIED TO EACH CO-ORDINATE IN
TURN SO THAT WE CAN OBTAIN THE TOTAL
RESPONSE BY SUMMATION.
Orthogonality Conditions
• The mode shapes of a system have special

and useful properties that explain significant
physical observations. We will just state {u } [M ]{u }= 0 r ≠ s
s T r
them.
• Let us take two different modes, say rth and
{u } [M ]{u }= m r = s
s T r
r
{u } [K ]{u }= 0 r ≠ s
sth modes. When they are multiplied with
mass matrix or stiffness matrix, the result is s T r
zero. This multiplication is orthogonalization
as written in a compact matrix form here.
• Generalized Mass – However, if we choose
{u } [M ]{u }= k r = s
s T r
the mode shapes r and s to be the same, say,

r, we get rth mode generalized mass, similarly, {u~ } [M ]{u~ }= 1
r T r
sth mode generalized stiffness.
{u~ }= {u }
• Remember, the mode shapes are T
proportional, therefore the generalized mass r 1 r
and stiffness depend upon the original shapes
that you choose – a unique way of choosing mr
the shapes is such that the generalized mass is
one unit and the stiffness is p2. They are then
called orthogonolized mode shapes.
Modal Analysis
We will write the multi-degree system equations

first.
• Recall, we stated that the natural frequencies and [M ]{q&&}+ [K ]{q} = 0
mode shapes are special solutions to the system. We
haven’t yet been able to say what happens, if one just {q} = [U~ ]{η }
bumps, or raps the system, how does it respond? The
system knows its mode shapes and try to respond in [M ][U~ ]{η&&}+ [K ][U~ ]{η } = 0
all modes, and adjust them in such a way they satisfy
the type of displacement given in the test. That is it [U~ ] [M ][U~ ]{η&&}+ [U~ ] [K ][U~ ]{η} = 0
T T
weighs how much amount in each mode it should

respond, as decided by the generalized coordinates [I ]{η&&}+ [λ2 ]{η } = 0
(weighing terms) and the result that we can obtain is
called Modal Analysis. This expansion is called η&&(t ) + p 2η (t ) = 0 r = 1,2, K N
modal expansion.
• We pre-multiply the equations with transpose of the η&& + 2ξpη + p 2η = 0 r = 1,2,K N
modal matrix and use the orthogonal properties
derived before to get the result given here.
• All practical systems, therefore reduce to simple
single degree of freedom models for any given mode
of vibration.
• We can introduce damping in the system for each
mode and get the damped modal equations.
Proportional Damping
[C ] = α [M ] + β [K ]
• Damping matrix does not obey the
[I ]{η&&}+ [U~ ] (α [M ] + β [K ])[U~ ]{η&}+ [p 2 ]{η } = 0
T
orthogonality properties stated earlier, hence, the
concept of a proportional damping is evolved,
wherein, the damping matrix is taken to be the
[I ]{η&&}+ (α [I ] + β [p 2 ]){η&}+ [p 2 ]{η } = 0
sum of a mass matrix and stiffness matrix with
appropriate proportionality constants, a and b. 2ξ r pr = α + β pr2
• If we know a and b we can find the
viscous damping ratio in the corresponding α β pr
ξr = +
mode and thus use experimental value to 2 pr 2
write damped modal equations.
• General finite element codes such as η&&r + 2ξ r prη&r + pr2η r = 0 r = 1,2, K
ANSYS, NASTRAN … adopt beta
damping in place of viscous damping..
Modal Analysis
Modal Analysis
Modal Analysis
Modal Analysis - example

Modal Analysis – example.. contd…






VIBRATION ABSORBER
VIBRATION ABSORBER …contd…

LECTURE 4
Continuous Systems Approach

CONTINUOUS SYSTEMS Systems governed by Wave Equation

Strings under tension
CONTINUOUS SYSTEMS WAVE EQUATION contd….

Axial vibration of Bars
CONTINUOUS SYSTEMS WAVE EQUATION contd….

Torsional vibration of RODS
Solution of WAVE EQUATION

Solution of WAVE EQUATION contd….


Free Vibration of Strings

Free Vibration of Bars

Free Vibration of Bars contd….
BEAMS Bending Vibrations

BEAMS Bending Vibrations contd….






RAYLEIGH’s Method
RAYLEIGH’s Method contd….

Example………..RAYLEIGH’s Method
Example………..RAYLEIGH’s Method …. contd…



RAYLEIGH – Ritz Method
undetermined
RAYLEIGH – Ritz Method …. contd….

RAYLEIGH – Ritz Method contd….

Example………..RAYLEIGH – RITZ Method

Example………..RAYLEIGH – RITZ Method …. contd…

Example………..RAYLEIGH – RITZ Method …. contd…

GALERKIN Method
GALERKIN Method …. contd…



Lecture 5
Stability Considerations
STABILITY
STABILITY Considerations Phase-Plane

STABILITY Considerations Phase-Plane ..contd..




• other criterion to be discussed during Rotor Dynamics

Lecture 6
FINITE ELEMENT FORMULATION

&
COMPONENT MODE SYNTHESIS
Finite Element Formulation – Axial Vibration of Beams

Reference Systems
Example
Guyan Reduction Scheme

Component Mode Synthesis

Example
Stiffness and Mass Matrices for substructures 1 and 2
Reduced Equation
Lecture 7
Rotor Dynamics
Rotor Dynamics
Rotor Dynamics is different from Structural

Dynamics, as we deal with a rotating
structure. Basically, all the vibration
phenomena will be valid, however, there are
several differences and we have to set up
procedures on handling the rotors and their
vibratory phenomena.
• Rankine is attributed to have mentioned the
existence of a critical speed of a rotor in
1869. He defined this as a limit of speed for
centrifugal whirling.
• There were many doubts whether a rotor can
cross such a critical speed? It was presumed
that it will be unstable after crossing the
critical speed. This is somewhat similar to
Speed of sound and whether one can cross
this barrier in flying.
• We have to wait for nearly 50 years to have
a clear understanding on this topic.
William John Macquorn Rankine (1820-1872)

Ingenious Flexible Shaft
• Though, we have not understood clearly about the

basics of rotor dynamics, Laval built the first
turbine in 1883 which ran successfully at 40000
rpm!!!
• From simple equilibrium conditions, he derived a
correct relation for the whirl radius (though whirl
and spin have not been clearly differentiated), in his
own notation, it is
ω 2δ eccentricity
y=
Fg
−ω2
stiffness W
• When the denominator is zero, the whirl radius is
infinity, defining the critical speed.
• Laval proved that one can cross the critical speed
and is stable at high speeds as y approaches -d
Jeffcott’s fundamental contribution
• Jeffcott in 1919 treated the problem as

forced vibration and identified the basic
principles of rotor dynamics. He showed
that the shaft did not rotate about rest
position but about its own center line, which
is spinning. The spinning rotor whirled
about the mean bearing center line.
• He identified the unbalance in the system
to be the driving force setting the whirl in
forced vibration. He derived simple
differential equations of the system and
solved them.
• Jeffcott through this model has shown that
one can cross the critical speed without any
instability. • In this model, the shaft is assumed mass less and
• In English speaking countries, a rotor the disk to be rigid. The total mass M of the rotor is
such as the one shown is named after put as disk and the stiffness K is represented as
Jeffcott, though in Germany, Scandinavia, shaft. The eccentricity is denoted by EG = a. The
Holland, they prefer to call it as Laval rotor, rotor spins about its own axis with an angular velocity
because, it was Laval who derived the w and whirls with angular velocity n. Jeffcott
expression in the previous page. considered synchronous whirl, i.e., n=w.
Equations of Motion - Solution
d2 dz
M 2 ( z + a cos ω t ) + C + Kz = 0
dt dt
d2 dy
M 2 ( z + a sin ω t ) + C + Ky = 0
dt dt
r = z + iy
d 2r dr
M 2 +C + Kr = Maω 2 e iω t
dt dt
O is the bearing center line, E is the disk geometric • Solution for the amplitude R of r, in
center, G is the mass center, EG=a is the eccentricity, terms of a and W = w/p
OE=R is the whirl radius about the bearing center line.
Disk rotates/spins about E with angular velocity w ccw
R Ω2
direction and the whirl is assumed synchronous with R= =
spin. The whirl is lagging in phase by an angle f from
the unbalance force vector in direction of EG.
a (1 − Ω ) + (2ξ Ω )
2 2 2
• Write down the inertia forces, stiffness forces and 2ξ Ω

damping forces in the respective directions … φ=
1 − Ω2
Jeffcott Rotor - Solution
• When the rotor is stationary initially, the rotor

speed is zero and the response is also zero for all
damping values.
• As the speed increases, the whirl radius
increases with the phase angle less than 90 deg
until resonance. At resonance, the phase angle is
90 deg irrespective of damping. The peak value
occurs slightly beyond resonance.
• Damping limits the resonant response and the
quality factor (magnification) is 1 over twice the
damping ratio, 1/2x.
• At high speeds beyond resonance, the system
runs smoothly with response equal to eccentricity
and a phase 180 deg, thus keeping the cg steady
at the bearing center line.
Translatory and Conical Whirl
•Rigid rotor on flexible bearings is same as

a flexible rotor on rigid bearings - however
it provides for translatory and conical whirl
modes. The first flexure of the rotor comes
after these whirls.
Bowed Rotor - Bow r0 at α0
d 2r dr
M 2 + C + Kr = Maω 2 e iωt + Kre i ( ax + a0 )
dt dt
R
R=
a
Ω2 R0 e ia0
= e i (ω t −φ )
+ e i (ω t −φb )
(1 − Ω ) + (2ξ Ω)
2 2 2
(1 − Ω ) + (2ξ Ω)
2 2 2
r0
R0 = bow factor
a
Rotors get statically bent or bowed or warped  2ξ Ω 

φ = tan -1  2 
phase angle of unbalance
due to sudden thermal loads, or leaving the rotor 1− Ω 
unattended for long periods without barring.
Sometimes, the rotor is balanced in a tunnel and φb = φ + a0 phase angle of bow
left for long periods in a crate without adequate
support to avoid gravity sag. In all these cases,
the rotor comes to a halt at the heavy spot, with •Response due to conventional unbalance.
the rotor sag and eccentricity in one line.
•In the above, OO’=R is whirl radius, O’E=r0 is
the bow, EG=a is the eccentricity, a0 is the bow Response due to bow unbalance.
location with the eccentricity.
•If the rotor is dropped from a height, the bow
location angle is 180 deg, which is in a direction
opposite to the mass center.
Response - Self Balancing Speed
[ ]
R = Ae − iφ + Be − i (φ + a0 ) eiω t a0 = 00
Ω 2 + R0
= [A + Be ]e (
R
)
− ia0 i ω t −ψ R= =
a (1 − Ω ) + (2ξ Ω )
2 2 2
Ω 2 a0 = 180 0
A= Ω 2 − R0
(1 − Ω ) + (2ξ Ω)
R
2 2 2 R= =
a (1 − Ω ) + (2ξ Ω )
2 2 2
R0 Ω s = R0
B=
(1 − Ω ) + (2ξ Ω)
2 2 2
•A is response due to conventional unbalance and
B due to the bow unbalance.
•Usually, the bow gives a0 = 0, then these two
 A sin φ + B sin φb  responses get added, i.e., the unbalance increases.
ψ = tan −1
 •For a dropped rotor the bow usually is a0 = 180,
 A cos φ + B cos φb  then these two responses oppose.
•The above condition leads to a self balancing
speed, Ws, speed at which the response becomes
zero.
Response of a bowed rotor
Phase relationship of a bowed rotor

Whirl Amplitude of a bowed rotor
Rotor with bow phase 1800
Phase relationship of a bowed rotor

Whirl Amplitude of a bowed rotor
Rigid Rotors in Flexible Anisotropic Bearings

Rigid Rotors in Flexible Anisotropic Bearings ..contd..



Rigid Rotors in Flexible Anisotropic Bearings

with Cross-Coupling & Damping
Rigid Rotors in Flexible Anisotropic Bearings .contd..

Rigid Rotors in Flexible Anisotropic Bearings .contd..

.contd..
UNEQUAL MOMENTS OF INERTIA

UNEQUAL MOMENTS OF INERTIA …. contd…




Instability in Torsional Systems

Instability in Torsional Systems …. contd…

Hill’s Equation; Mathieu‘s Equation

Floquet’s Theory; Strutt Diagram

Floquet’s Theory determines the behaviour of the Mathieu’s
Equation and describes the Strutt Diagram
GRAVITATIONAL Effect
GRAVITATIONAL Effect …. contd…




OIL WHIRL
ROTOR STABILITY IN FLUID FILM BEARINGS contd..

ROUTH-HURWITZ CRITERION
ROUTH-HURWITZ CRITERION contd..

Example ROUTH-HURWITZ CRITERION

Example ROUTH-HURWITZ CRITERION …. contd…

Holzer’s Method for Torsional Systems

Holzer’s Method for Torsional Systems …. contd…

Holzer’s Method for Torsional Systems …. contd…

Example …… Holzer’s Method

Myklestad-Prohl Method for Beams

Myklestad-Prohl Method for Beams ..contd…



GYROSCOPIC EFFECTS
GYROSCOPIC EFFECTS Freely Spinning Disc

GYROSCOPIC EFFECTS Freely Spinning Disc ..contd..

GYROSCOPIC EFFECTS Disc on Shaft
Additional term in place of 0

GYROSCOPIC EFFECTS Disc on Shaft .. contd…

Rigid Rotor Imbalance Classification
•The imbalance in a rotor can be static – i.e., the

principal inertia axis is shifted parallel to the shaft axis,
a single correction mass is sufficient to balance the
rotor in a plane containing cg.
•A special case of imbalance can be a coupled
unbalance, in which the principal inertia axis passes
through the center of gravity, then a couple, two equal
and opposite masses are needed for correcting the
imbalance.
•In a general dynamic unbalance case, sum of the
above two special cases, we need two correction
masses in any two convenient planes – we will discuss
this further.
Dynamic Unbalance
Single Plane Balancing
•Balancing of single discs is a fairly simple task

as
all the unbalance can be considered as confined
to be in one plane.
•If you run the rotor with the residual unbalance
as it existed and which is to be corrected, a
vibration pickup on the bearing will sense a
response due to this unbalance. In order to relate
the vibration signal to the rotor, one can have a
phasor placed on the shaft, relative to which the
vibration signal is measured, that gives the
magnitude and phase angle of the response with
the residual unbalance. Let that be a vector Ov as
shown. Note the 0, 90, 180 and 270o positions
taken for the given rotation with 0 as the phasor
location.
•Though we know a phasor location marked on
the shaft, we have no idea where the unbalance
is, now we place a known mass at a known
location as per the chart and measure the
response which is O+T. The effect of trial mass is
now given by (O+T) – Ov = Tv.
•The correction mass is therefore to be placed at
35o in the direction of rotation from the location of
trial mass, the magnitude is decided by the
lengths of the vectors Tv and Ov.
Rigid Rotor Balancing
Distributed Unbalance
•Very few rotors can be considered as single plane rotors, therefore we

have to devise a method of balancing rotors with distributed unbalance.
•When the rotors do not deform and retain the original shape, the rotor can
be considered as rigid – in practice rotors running well below the critical
speed are rigid rotors, then we can devise the following procedure.
•Divide the rotor into several disks, as many as you want, say n in number.
Let us consider the ith rotor at a distance zi from a plane which we will
denote left plane marked L. In this plane, the unbalance is taken as miai at
an angle fi from x axis. There are n such masses in these n disks chosen.
•We also choose another convenient plane and denote this as plane R as
shown.
Rigid Rotor – Two Plane Balancing
•Let Fi be the unbalance force in ith plane. Introduce

two equal and opposite forces Fi in plane L. Since
the rotor is rigid these two equal and opposite forces
have no effect on the equilibrium of the system. We
have three forces now, two of them form a couple
ML as shown.
•Now split the couple into two equal and opposite
forces in the planes L and R separated by distance a.
•In plane L we have now two forces, combine them
to form into one and let this be Fi L and let the force
in the right plane be denoted as Fi R.
•Repeat this to all the n disks and form a set of n
concurrent forces in each of the planes L and R.
Find the resultants of these concurrent forces and
denote them FL and FR.
•Thus we reduced the original distributed unbalance
in the system to two unbalance forces in any two
convenient planes – remember this is valid only for
rigid rotors and a process of removing these two
unbalance forces is called Two Plane Balancing or
Dynamic Balancing, in place of simple statics used
in Single Plane Balancing or Static Balancing.
Influence Coefficients
•As in single plane balancing fix the

angular locations and use a phasor to
measure the phase. Choose the
measurement planes to be, say left
bearing and right bearing. Let the
responses without any trial mass be as
shown.
•Now introduce a trial mass in right plane

and let the responses be as denoted here.
•Similarly, let a trial mass in left plane

give the responses as shown here.
•Using these six responses, we derive the

influence coefficients, measures that tell us
how the responses will be arising out of
known forces. The example in the
following pages illustrates how to achieve
the balancing of a rigid rotor.
Example
Run 1 : No Trial Mass
Left Plane L :
9.144 microns at 90o L1 = 9.144i
Right Plane R :
10.16 microns at 45o R1 = 7.188 + 7.188i
Run 2 : Trial Mass 6.8g at 22.5o in Right Plane

T R = 6.294 + 2.608i
Left Plane L :
5.08 microns at 27 o L 2 = 4.521 + 2.311i
Right Plane R :
6.35 microns at 99 o R 2 = −0.9906 + 8.274i
Example – contd.
R 2 − R1
a bR =
TR
•Influence coefficient denoting the

=
(− 0.9906 + 6.274i ) − (7.188 + 7.188i )
response in plane b due to a unit force in 6.294 + 2.608i
plane R
= −1.160 + 0.336i
L 2 − L1
a aR =
TR
•Influence coefficient denoting the
response in plane a due to a unit force in
plane R
=
(4.521 + 2.311i ) − (9.144i )
6.294 + 2.608i
= 0.229 + 1.181i
Example – contd.
Final Run 3 : 6.8g at 36 o

T L = 5.5 + 3.997i
Left Plane L : 9.4 microns at 0 o
L 3 = 9.4 + 0i
•Influence coefficient denoting the response in
Right Plane R : 30.5 microns at 90 o
plane b due to a unit force in plane L
R 3 = −4.775 + 30.1i
R 3 − R1
a bL =
TL
a bL =
(− 4.775 + 30.1i ) − (7.188 + 7.188i )
5.5 + 3.997i
= −0.558 + 3.76i
•Influence coefficient denoting the response in L 3 − L1

a aL =
plane a due to a unit force in plane L TL
9.4 − 9.144i
a aL =
5.5 + 3.997i
= 0.327 − 1.9i
Example – contd.
•We need to eliminate the original response vectors

by using correction masses in planes L and R, to
− R1 = m R a bR + m L a bL
estimate these masses, we form the set of
equations given here using the influence
coefficients determined earlier.
− L1 = m R a aR + m L a aL
L1 a bL − R1 a aL
mR =
•Solving the above two equations, we get
a bR a aL − a aR a bL
R1 a aR − L1 a bR
the expressions for correction masses.
mL =
•When these correction masses are added, (we may a bR a aL − a aR a bL
have to do this in two masses each by placing them
in the nearest locations) the response theoretically
should be zero. This does not happen as there are
m R = 10.94 + 1.56i
several assumptions made in this analysis, the first
correction should bring the responses to be low, one m L = −2.81 − 1.9i
or two additional balancing runs may be needed to
achieve the desired grade quality.
Shop Balancing set up

Flexible Rotor Balancing

Influence Coefficients
Balancing Masses
Trial Mass in Plane 2

v 1
− v 1
ai12 = i 2 i
T
Trial Mass in Plane j
vij1 − vi1
aij1 =
T
...
k th Speed Tests
vij1 − vik
aijk =
T
{U }= [a ] {v} −1
Classification of Rotors
•Class 1 – Rigid Rotors: Rotors that can be corrected in any two arbitrary planes and after
correction, its unbalance does not significantly change at any speed upto the maximum
operating speed and when running under conditions which approximate closely to the final
supporting system. Rotors which do not satisfy this condition are classified as flexible rotors.
•Class 2 – Quasi Flexible Rotors: Rotors that cannot be considered rigid but can be balanced
adequately in a low speed balancing machine. These are rotors, (1) where the axial distribution
of unbalance is known – e.g., 2A – shaft with a grinding wheel; 2B – Shaft with a grinding wheel
and pulley; 2C – Jet engine compressor rotor; 2D – Printing roller; 2E – rotors with a long rigid
mass supported by a flexible shaft, whose unbalance can be neglected, such as computer
memory drum; (2) where axial distribution is not known – e.g., 2F – Symmetrical rotors with two
end correction planes, whose maximum speed does not significantly approach II critical speed
and whose service speed does not contain I critical speed and with controlled initial unbalance;
2G – same as 2F but with an additional central correction plane and that it may have its service
speed in I critical speed range; 2H – Same as 2F rotors but unsymmetrical.
•Class 3 – Flexible Rotors: Rotors that cannot be balanced in a low speed balancing machine
and that require some special flexible rotor balancing technique – e.g., Generator rotors•
Class 4 – Flexible Attachment Rotors: Rotors that could fall in categories 1 or 2 but have in
addition one or two more components that are themselves flexible or flexibly attached – e.g., a
rotor with centrifugal switch
•Class 5 – Single Speed Flexible Rotors: Rotors that could fall into category 3, but for some
reason, e.g., economy, are balanced only for one speed of operation – e.g., high speed motor
ISO 1940: Rotor Classification

Balance Quality Grade G

L-8
Vibration Measurement
Pickups, Analyzers, Modulation,
Cepstrum Analysis,
Digital Measurement
Why Vibration Measurements?

• We studied basic principles of structural vibrations, and special applications to rotors and their
behavior under dynamic conditions.
• While all this information and much more is basic to the understanding of machine behavior, the
current course is concerned primarily with the health of a machi ne. Every machine deteriorates in
condition, however well it has been designed/ The rotating machinery is very expensive and they
should be available for their operation for long uninterrupted intervals (years) and without any failures
and unexpected shut downs.
• Asset management is an important aspect to any heavy industry with rotating machinery. We find
several indications reflecting the condition of a machine in its life time; e.g., we know that an
automobile engine needs a change of lubricating oil as it becomes contaminated with dirt over a
period of time, so if we can find a measure of the state of the lubricating oil, an appropriate action can
be planned – or if one finds the bearing temperatures are going up, there is some rubbing and
inadequacy of lubricating oil, so on and so forth. Some of these measures, however, do not tell us the
problem in a machine sufficiently well in time for a proper asset management.
• Over years of asset management by maintenance people, we know that the earliest indications of any
fault in a rotating machine are detected by an increase in vibration and sound levels, that is why every
asset management practice adopts vibration measurement as the first and foremost step that may be
further assisted by other measurements such as lub oil particle counting, bearing pressure and
temperature measurement, process parameter variations … Basically it is the vibration level, its
signature in time and frequency domains, orbits, and trends over short and long durations that help us
in understanding the health of the machine and predicting any impending problem so that timely
action is taken. Therefore vibration measurement and its analysis is important first step in asset
management of rotating machinery.
Vibration Units of Measure
AMPLITUDE
Displacement in microns, mils, below 10 Hz

Is a measure of the distance the object moves
Velocity in mm/sec, inch/sec, between 10-1000 Hz

Is a measure of how fast it moves - (Speed) - most destructive
energy
Acceleration in mm/sec², Spike Energy - g above all

The force imparted on the vibration object as it changes its velocity.
TIME (FREQUENCY) Cycles / Sec, Hertz, RPM
PHASE Radian, Degree

Displacement, Velocity and Acceleration
• Displacement, velocity
and acceleration for a
given frequency are all
related through
• V = Xp
• A = Vp = Xp2
• Usually velocity is taken

as a standard as it
represents the energy
associated in the system;
for a given velocity and
frequency, we can find
the associated amplitudes
of displacement and
acceleration.
Transducers Selection
Type of Measurement:
1) Contact (Seismic)
2) Non Non-Contact (Relative) Contact (Relative)
•Displacement: Eddy Current Proximity Pickup

•Velocity (Seismic): Electro Dynamic Transducer
•Acceleration: Piezoelectric Pickup
Direction & Location of Measurement: Direction & Location of Measurement:

• Horizontal, Vertical and Axial
• Bearing Pedestal, Shaft, Journal housing, Structure
Eddy Probe – Velocity Transducer - Accelerometer
• Measure Relative Distance Between Two Surfaces

• Accurate Low Frequency Response
• Limited High Frequency Sensitivity
• Require External Power Source
• Sensitive to observed material
• Often Measure Bearing Housings or Machinery Casing Vibration

• Effective in Low to Mid Frequency Range (10 Hz to around 1,500 Hz)
• Self Generating Devices
• Are Electro-Mechanical Devices With Moving Parts That Can Stick or Fail
• Rugged Devices
• Operate in Wide Frequency Range (Near 0 to above 400 kHz)
• Good High Frequency Response
• Some Models Suitable For High Temperature
• Require Additional Electronics
Sensor Relationships
Amplitude
(mils, in/sec, g’s)
• The adjoining figure gives the

displacement and acceleration for a
velocity of 0.628 in/sec
• At 1 Hz, displacement x = v/p, i.e.,
mils
• Corresponding acceleration at 1 Hz
is a = vp
• Common machinery operating

range is shown by the rectangular
area, displacement and acceleration
at any frequency can be obtained in
terms of 0.628 in/sec velocity
Vibration Sensor Principle
• A vibration sensor utilizes the basic single degree system response.

• The sensor when placed on a vibrating member experiences x(t) that is to to be measured
• The mass responds with absolute displacement y(t)
• Relative displacement z(t) = y(t) - x(t) and Z/X is shown above used as a measure of x(t)
• The sensor range and application is depends on how we proportion its mass and spring,
Seismometer / Velocity Transducer
• A Seismometer or Velocity Transducer measures rate of • Electromagnetic Principle

change of relative displacement z(t) = y(t) - x(t) using Variation of Permeance of magnetic
the electromagnetic principle through the induced circuit causes a change in the flux FX
voltage as illustrated in right side, as a measure of x(t)
• The sensor range for Z=X is r >> w/p, therefore p dφ
should as low as possible, i.e., the mass is relatively V = −n × 10 −8
heavier to the spring stiffness, then we measure the dt
velocity of the support and hence Velocity Transducer. dφ dx
= −n × 10 −8 V
dt dt
Accelerometer Principle
• An Accelerometer uses Piezo electric principle

given in next slide to measure the acceleration
directly
• The sensor range for Z to represent Xw2 the
acceleration of the support is r << w/p, therefore p
should as high as possible, i.e., the mass is
relatively lighter
• Since p is a constant, we need the factor f to be as
close to unity as possible, then we measure
acceleration directly.
• When the damping ratio is 0.7, the factor f equals
unity to a large possible range of r, say up to 0.2.
Damping is therefore very crucial to an
accelerometer. Z r2
• Piezoelectric crystals have very low mass and =
very
X (1 − r ) + (2ξ r )
2 2 2
high frequencies, therefore are very good

ω2
candidates for measurement of acceleration. They = 2
f
can also be made very light and hence have useful p
applications for measurements on small 1
components without effecting the basic structure. f =
(1 − r ) + (2ξ r )
2 2 2
Accelerometer Elements
Output impedance from accelerometer is very high - problems

matching, noise, cable length Charge amplifier is therefore used.
Elements of a measurement system

Proximity Probe – Non contact Measurement
Probe mounted in the proximity of the moving object -

High frequency field setup in the gap.
•Can act as a key phasor
•Two probes can give the orbit
•Mounted in the bearing - relative displacement
Sinusoid Signal - Characteristics
• The vibration signal in general is somewhat periodic with

several harmonics, the basic signal is therefore a sinusoid.
• A sinusoid, its squared sinusoid are shown above. Since the
energy in the system is proportional to velocity squared, we
prefer a root mean square value of the amplitude, rms value, to
judge the condition of the machine.
Calculation of RMS Value - Sinusoid
Let the pk to pk value of a sinusoid be 240 mils/s.

• The amplitude is 240/2 = 120 mils/s
• Square of the signal varies from 0 to 1202 = 14400 mils/s
• Mean of the square = 7200 mils/s
• RMS = (7200)1/2 =120/1.414 = 84.85 mils/s
• Pk to Pk RMS = 169.7 mils/s
Sum of Two Sinusoids in Phase
Both sinusoids amplitude = 1 The amplitudes of both sinusoids that make

Peak to peak of sum = 3.5203 up the sum is obtained in frequency domain.
Mean = 0
Root Mean Square of sum = 1
Sum of Two Sinusoids out of Phase
Peak to peak = 3.2506 • Frequency Domain remains same

Mean = 0 • Phase Information lost
Root Mean Square = 1
Calculation of RMS with two components
Consider a periodic signal with a sum (-0.1 sin 81.1t +0.05

sin 173.52t). Let us find the means, mean squares and root
RMS1 = 0.005 = 0.0707
mean squares of the components and the total signal.
1. Let the two components be denoted by 1 and 2. The mean RMS 2 = 0.00125 = 0.0354
values of individual components as well as the sum, being RMS total = 0.005 + 0.00125 = 0.0791
harmonic signals, are zero RMS pk to pk of over all signal = 0.1582 units
M1 = 0, M2 = 0, Mean of sum = 0
2. The time period of the lower frequency is large of the two
and is given by 2p /81.1 = 0.0775 s. We now follow the steps
given here to find MS (mean squares) and RMS [Integrate Note
and average over the period 0.0775s]
1 − cos 2 pt
sin 2 pt =
2
∫ [(0.1sin 81.1t ) + (0.05 sin 173.52t ) ]dt

0.0775
1 2 2
MS =
0.0775 0 and integral of cosine term becomes zero
∫ [(0.1sin )]
0.0775
=
1
0.0775
2
) (
81.1t + 0.0025 sin 2 173.52t dt
0
1  0.01   0.0025 
=  × 0.0775  +  × 0.0775 
0.0775  2   2 
= 0.005 + 0.00125
Fourier Analysis
3D View
View in time domain
View in frequency domain

Fourier Analysis
∞
1
f (t ) = a 0 + ∑ (a n cos nωt + bn sin ωt )
2 n =1
• Fourier analysis is a mathematical tool T
2
that helps us in identifying the frequency a n = ∫ F (t ) cos nω tdt
components of a periodic vibration signal T 0
which is composed of several harmonics. T
2
• The periodic signal is assumed to consist bn = ∫ F (t )sin nω tdt
of several harmonics of the fundamental T 0
frequency and an infinite series gives 1 ∞
accurate results. In practice, first few f (t ) = a 0 + ∑ An cos(nωt − φ n )
components are considered. The harmonic 2 n =1
cosine and sine components are An = a n2 + bn2

• The amplitude in each harmonic, n = 1, 2,
is given by bn
φ n = tan −1
• The phase angle is an
Fourier Analysis - Complex Form

• Fourier analysis is written in complex
form so that faster numerical tools can be cos nωt =
2
(
1 inω t
e + e −inω t )
developed. The cosine and harmonic
terms are written in exponential form as sin nωt =
2i
(
1 inω t
e − e −inω t )
given here ∞
1 1 1 
• The harmonics can be written as one f (t ) = a0 + ∑  (an − ibn )e inω t + (an + ibn )e −inω t 
2 n =1  2 2 
sided series 0, 1, 2, … infinity +∞
• Or as two sided series, - infinity, …, -2, - Two sided series f (t ) = ∑ Cn e inω t

−∞
1, 0, 1, 2, … infinity. 1
• In this form we loose phase information, Cn = (an − ibn ) Cs = 1 As
2 2
the harmonics in two sided series have a Harmonic 1
+ T
2
1
magnitude equal to half the values. Cn Components Cn = ∫ f (t )e
inω t
T
• A finite time signal is considered in this Complex 1
− T
2
process. 1
T0
We will get back to this again
∫ {x(t )} dt
2
• Average power of the signal over a Pav =
T0 0
period of time, T0 is 2
∞ 2
a  A
• Complex harmonic components directly = 0  +∑ n
 2  n=1 2
represent the average power in the
= ∑ Cn
2
corresponding frequency term, this is
Parseval theorem.
Saw Tooth Signal
4t
x(t ) = 2 − 0 < t < T0
T0
4  2πt 1 4πt 1 6πt 
= sin + sin + sin 
π  T0 2 T0 3 T0 
4 2πkt
= sin k = 1,2,3,...
kπ T0
• To illustrate the usefulness of Fourier series, consider a periodic

saw tooth signal given above with its components.
• The top figure shows addition of terms one by one up to six.
• The figure below shows the net result of sum of first six terms,
which is getting closer to the original saw tooth form.
• We can increase the number of terms in the summation to get more accurate result.
• Sixty terms sum is given in the figure below.
• In practice, we are interested in the lower harmonics, or those harmonics which can be
excited to resonance from any per rev or nozzle excitations.
• Here, the sum up to 25 terms is given above with the harmonics in

frequency domain given below.
• The first harmonic is one unit, followed by the second at ½ unit and
the higher ones decrease in magnitude rapidly.
Rectangular Signal
x(t ) = 1 0 < t < T0 / 2

= −1 0 < t < T0 / 2
4  2πt 1 2πt 1 2πt 
= sin + sin 3 + sin 5 + ...
π  T0 3 T0 5 T0 
4 2πkt
= sin k = 1,2,3,...
kπ T0
• This and the next slide illustrate a rectangular periodic signal and its Fourier components.
Octave Band Analysis
• 4-5 decades ago, a Fourier analysis was a tedious job, the signal is first recorded on a
recorder, e.g., a UV recorder, it was then enlarged and digitized manually to obtain the signal
as a function of time. A hand calculation or a main frame computer was then used to determine
the Fourier components in any diagnostics and trouble shooting exercise.
• Dedicated analog instruments are then developed using filter circuits, which are expensive –
the accuracy was limited octave bands, e.g., 11 filters are common with center frequencies
beginning from 31.5 and doubling consecutively.
• A vibration chart thus developed is called octave band analysis, which gave the relative
energy levels in these bands to make a diagnostics analysis.
Sequential analysis
One filter at a time
Analysis time
between each analysis
Significant transient
data may be lost
• Subsequently, we had 1/3 octave band analyzers which helped in

narrowing the frequency zones for a better diagnostics.
Fourier Transform – time to frequency domain
•Not limited to periodic functions alone

•Achieved through Fourier Integral
•Aperiodic function repeats itself
after a large (infinite) time
•Associated frequency becomes negligibly small
•Replace w by w’, nw’ becomes continuous
•T = 2 / pw’ approaches infinity
nω → ω
+∞
F (ω ) = (TC n )T →∞ = ∫ f (t )e −iω t dt
Forward Transform – Fourier Integral −∞
+∞
F (ω ) = Ff (t ) = ∫ f (t )e −iω t dt
−∞
+∞
1
f (t ) = F F (ω ) =
−1
∫ f (ω )e −iω t dω
•Inverse Transform – Conversion 2π −∞
from frequency to time domain
Discrete Fourier Transform (DFT)

T
t k = 0,
2T
,K ,K
(N − 1)T
N N N
• f(t) is analog should be converted kT
= k = 0,1,2, K ( N − 1)
to a digital text file N
N −1 2πkr
1 −i
• N intervals of a vibration record in Fr = ∑f k e N
r = 0,1,2, K ( N − 1)
N
time T, time step T/N k =0
1
{F }N ×1 = [A]N × N {F }N ×1
N
• Frequency of the entire period = 2p /T

• Discrete frequencies 2p /0, 2pN /T , 2pN /2T , 2pN /3T, … 2pN /(N-1)T
Example:
• Let 5000 samples be taken in 1 sec.
• Let us use N = 4096 out of these 5000 data points.
• T = 4096/5000
• Smallest discrete frequency = 2pN /(N-1)T = 2p(4096) /
[4095(4096/5000)] rad/s
= 5000/4095 = 1.22 Hz.
Fast Fourier Transform - FFT
[A] factored into log 2 N matrices [A ]i
1 only two non zero elements in each row

{F }N ×1 = [A]N × N {F }N ×1
N For N = 4
2πkr [A] = [P][A]1 [A]2

−i
a kr = e N
1 0 0 0 ↑ ↑ 0 0  ↑ 0 ↑ 0
0   
0 1 0 ↑ ↓ 0 0 0 ↑ 0 ↑
=
0 1 0 0  0 0 ↑ → ↑ 0 ↓ 0
   
0 0 0 1  0 0 ↑ ←  0 ↑ 0 ↓
0 ↑ ↑ ↑ ↑ 
 
1 ↑ → ↓ ←
[A] = 
2 ↑ ↓ ↑ ↓
 
3 ↑ ← ↓ → [P] Permutation matrix
Comparison - DFT/FFT Computation Requirements
• On the present day Desk top or lap top

computers, FFT can be performed in milli
secs, thus making the conversion of time
domain data to frequency domain almost
instantaneous and in real time.
• With an A/D converter card, a lap top has

thus become a good diagnostics
instrument and tool.
Amplitude Modulated Signal
x(t ) = x + x a cos ω a t  cos ω t

Modulation Amplitude x a
Amplitude (x - x a ) to (x + x a )
Modulation Frequency ω a
x(t ) = x + x a cos ω a t  cos ω t

1 1
= x cos ω t + x a cos(ω + ω a )t + x a cos(ω − ω a )t
2 2
Eccentrically mounted gear

ω c gear mesh frequency
= carrier frequency
ω a rotational speed
ω c + ω a andω c - ω a
= side bands
Amplitude Modulated Signal
Mean = 0
Mean square = 0.56
Root Mean Square = 0.75
Side bands in Frequency domain
Frequency or Phase Modulated Signal
x (t ) = X cos (ω + ω ' cos ω a t )ω t

x (t ) = Re Xe i φ (t )
ω ' and ω a - magnitude and frequency of modulation

ω ' / ω a - modulation index
Frequency varies (ω − ω ' ) to (ω + ω ' )
Phasor of constant magnitude
Variable angular speed
In torsion al vibration , the angular frequency
changes over the mean value correspond ing to the
rotational speed and the resultig signal is a
Frequency Modulated Signal.
Frequency or Phase Modulated Signal
Mean = 0 Modulation Magnitude = 0.5

Mean square = 0.50 Modulation Freq = 4 Hz
Root Mean Square = 0.71 Carrier Freq = 15 Hz
Torsional Vibration Measurement
ω rotational speed
Ω frequency of torsional oscillation
θ amplitude of torsional vibration
Instantaneous shaft position
φ = ω t + θ cosΩt
Angular velocity φ& = ω - θΩ sin Ωt
Angular speed range
ω min = ∞ − Ωθ to ω max = ∞ + Ωθ
ω − ω min
θ = max
2Ω
2π 2π
ω min = , ω max =
Tmax Tmin
ω max − ω min
θ=
2Ω
Power (Auto) Spectrum - Cross-Spectra
Amplitude Spectrum
• Amplitude spectrum is given by the magnitude
and frequency of each harmonic component. F (ω )An , φ n
• Power spectrum is obtained by squaring each Discrete
amplitude and halving them and plotted in Power Spectrum
discrete form at each frequency.
• Auto spectrum is same as power spectrum when
1 2
An
the frequency domain is expressed continuously, 2
this is obtained by taking the product of F(w) Auto Spectrum
and its conjugate marked by * Continuous
• Auto or Power spectrum is a measure of power S ff (ω ) = F * (ω )F (ω )
associated with corresponding frequency
component and hence important for vibration Excitation f(t)
Cross spectra
engineers. Response x(t)
• Cross spectrum involves two different functions.
Let us take the excitation and response functions S fx (ω ) = F * (ω )X (ω )
and their cross spectra are defined by S xf (ω ) = X * (ω )F (ω )
• Cross spectra are used in defining coherence,
which is a degree of linear dependence between
two signals, see next slide.
Coherence
Frequency Response Function

• Consider the frequency response function.
• Express this function in two different ways, first X (ω )
using the conjugate of amplitude spectrum of H (ω ) =
F (ω )
forcing function and next by using the conjugate
of amplitude spectrum of response function. X (ω )
H 2 (ω ) = X (ω )
• Coherence is defined by the ratio of these two
F (ω ) H (ω ) =
response functions. F (ω )
X (ω )X * (ω )
• If the measurement of forcing function and = , X (ω )F * (ω )
F (ω )X * (ω ) = ,
response function are free from noise error and F (ω )F * (ω )
S (ω )
that there is a perfect linear relationship between = xx S xf (ω )
S fx (ω ) =
them, Coherence will be unity. S ff (ω )
• Coherence can be used to determine whether two
different signals are coming from the same H 1 (ω )
y 2 (ω ) =
machine or if any of the signals are lost in H 2 (ω )
measurement due to some fault, coherence S xf (ω ) S fx (ω )
between a good and a bad signal will be poor. =
S ff (ω ) S xx (ω )
This property can be used in identifying any
faulty sensor or problem in the transmitting path.
Cross and Auto Correlations
• Correlation is a measure of how well two

functions relate each other, cross correlation is
given by the average of the product x(t ) f (t + τ )
where t is the delay period.
• Autocorrelation function is a measure of a T /2
1
function correlates with itself. R fx (τ ) = lim T →∞ ∫ x(t ) f (t + τ )dt
• Autocorrelation has a maximum value when the T −T / 2
delay period is zero, and it reduces to mean 1
T /2
− R ff (τ ) = lim T →∞ f (t ) f (t + τ )dt
T −T∫/ 2
square value of the function.
• Autocorrelation decreases as the delay period is
increased.
• When a signal is random, its mean square value
will be negligible.
• If a weak periodic signal is buried in a truly
random noise, the spectrum reveals very little,
however, autocorrelation reveals this weak
periodic signal, since the random part has zero
autocorrelation.
Cepstrum - (Time Domain)
• Definitions of Cepstrum Power spectrum of a logarithmic power spectrum

C xx (τ ) = F [log S xx (ω )]
2
• Terminology
Cepstrum for Spectrum
Quefrency (sec) for Frequency Inverse Fourier transform of logarithmic power spectrum
Rahmonics for Harmonics C xx (τ ) = F −1 [log S xx (ω )]
Lifter for Filter
Gamnitude (Hz) for Magnitude Inverse Fourier transform of amplitude spectrum
Saphe for Phase C xx (τ ) = F −1 [log X (ω )]
• Applications
• Side bands of a multi stage gear box,
they appear as distinct frequencies in Identify periodic structures in a spectrum
cepstrum domain. Appear as distinct peaks in cepstrum
Two families of side bands

frequency differences ∆f 1 and ∆f 2 Hz
Two distinct peaks at frequency

1 / ∆f 1 and 1 / ∆f 2 sec
Two Speed Gearbox with Defects

Advantage of Cepstrum Analysis

A/D Converter
• Analog to Digital converters and

modern high speed desk top computers
have revolutionized the way we record
and analyze vibration and other
process parameter time domain
signals. The analog tape recorders
have disappeared and the storage has
become very simple and reliable. The
analog instruments have also been
replaced by software and we can have
today a sensor and its conditioner
directly connected to a computer for
all the required analyses in real time.
• A 3 bit parallel A/D converter is
illustrated here. Analog voltage
compared with each node’s voltage
• Output voltage high (on) when the
analog voltage is above ref. voltage
and low (off) when it is below.
• Binary encoder compares and gives a
3 bit binary output.
3-bit encoder binary output
An encoder circuit which reads the comparator

outputs, high or low and produces a 3-bit binary
output corresponding to one of the eight possible
on/off conditions of the inputs 1 through 7
Examples
3 bit comparator Resolution with

16 bit comparator
4 5 Typical range for output voltage : - 10 to + 10 V

Input signal between E ref and E ref
8 8
comparators 1 - 4 read high and 5 - 7 read low
Converter saturates if the input signal
Input state is 4
exceeds the upper or lower limit
corresponding to binary output
Appropriate signal conditioning viz.,
(
100 2 = 2 2 × 1 + 21 × 0 + 2 0 × 0 = 4 ) amplitude attenuation or dc offset removal
All comparators read high, the output is 1112 (= 7)

1
3 - bit parallel encoder' s output E ref
23
All comparators read low, the output is 000 2 (= 0)
1
Voltage resolution per bit E v = ∆V fs
2n
∆V fs full scale range, n number of bits
16 - bit A/D converter card

Range divided into 216 = 65,536 increments
Voltage resolution 0.3mV for - 10 to + 10V range
Signal Sampling
• Signal sampling is an important
criterion while acquiring the data
1
Sampling rate f s =
∆t
• The same signal above shows
inadequate sampling rate, number of
points captured in a unit time is too
low and the original signal character
is lost.
• Decreasing sampling time or

increasing sampling rate improves
the digitization process.
Aliasing – Nyquist Criterion
1
f Nyq = fs
2
Sampling Example
Frequencies resolved 0, ∆f, 2∆∆fK n∆∆fK N∆∆= (f Nyq )

1
2
Highest frequency component - 2000 Hz (120,000 rpm)
Minimum sampling rage > 4000 Hz
Number of samples in FFT 4096
Vibratory signal f s = 5000 samples per second
N = 4096 of these samples ∆f = 5000/4096 = 1.22 Hz
Multiplexer – De-multiplexer
• A multiplexer is an IC chip (Transistor

Transistor Logic TTL 74150) which
performs like a selectable commutator, in
this case a 16 position – single pole
switch, 1 of 16 selector. A desired input is
chosen by the 4 digit binary number dcba
(d is Most Significant Bit and a is Least
Significant Bit). Saves costs of wiring
while acquiring a large number of signals.
The signals are monitored in sequence,
one by one by a computer separated by a
few seconds.
• De-multiplexer or a 1 of 16 data
distributor is similar but with reversed
action (TTL 74154). Here a multiplexer –
de-multiplexer is shown where a single
line is used rather than separate lines to
connect two positions. For the purpose of
binary control the ports d, c, b and a
should be connected.
• Only 5 circuits instead of 16 to connect
between two points

Vibration Analysis Lectures

Uploaded by

Copyright:

Available Formats

Vibration Analysis Lectures

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Vibration Analysis Lectures

Uploaded by

Copyright:

Available Formats

Indian Institute of Technology Kanpur

Simple Mechanical System

Physical System Mathematical Model

• The ability to move in any one direction is a DOF.

NATURAL FREQUENCY AND TIME RESPONSE

Equation of motion : m&x& + kx = 0

DISPLACEMENT, VELOCITY & ACCELERATION - I

The constants ( A, B ) or (C , φ ) depend on the initial conditions.

DISPLACEMENT, VELOCITY & ACCELERATION - II

• What provides the spring (or restoring)

IN ALL VIBRATION PROBLEMS, THE ORIGIN OF THE

TRANSLATIONAL, TORSIONAL & ROTATIONAL SDOF SYSTEMS

PENDULUM WITH SPRING

 We will apply : Iθ&& = ∑ Moment

KINETIC ENERGY vs STRAIN ENERGY

ENERGY METHOD TO DETERMINE ωn

MAX KINETIC ENERGY = MAX STRAIN ENERGY

Max velocity, Max displacement,

MORE COMPLEX SYSTEMS

GENERAL LAGRANGE EQUATION

VERY COMPLEX SYSTEMS Finite element model of a car

Simple Vibratory Phenomena

External Force – Time dependent usually periodic

Mass times Acceleration

Damping coefficient times

Mean Equilibrium position

System that can be described by one coordinate, say, x

Angular vibrations of any drive train

Bending (Flexural) vibrations of a drive train

FREE MOTION OF A DAMPED SDOF SYSTEM

Equation of motion : m&x& + cx& + kx = 0

SOLUTION OF THE EQUATION OF MOTION

EOM : &x& + 2ζωnx& + ω 2n x = 0

→ α1, 2 = (−ζ ± ζ − 1)ωn 2

POSSIBILITIES, DEPENDING ON THE VALUE OF ζ

Case 1. ζ < 1. Underdamped case with oscillatory motion.

TIME HISTORY FOR OSCILLATORY MOTION

ζ > 1. Overdamped case with no oscillatory motion.

ζ = 1. Critically - damped motion

EXAMPLE: SYSTEM IDENTIFICATION FROM TIME RESPONSE

The amplitude ratio between two successive cycles :

Viscous Damping Principle

• Damping force is proportional to velocity and

DAMPING IS NOT ALWAYS VISCOUS !

• Viscous damping ratio: ζ = c/c0

• Friction damping (energy dissipation via contact mechanism)

EXAMPLE OF DAMPED MOTION - PARACHUTE

Find the maximum compression in the

Compression = static compression + dynamic compression

USE INITIAL CONDITIONS TO FIND A & B IN EOM:

LAGRANGE’S EQUATION OF MOTION

EXAMPLE 1 – SDOF SYSTEM

FIND THE EQUATION OF MOTION

EX 2 - PENDULUM WITH SPRING

INSERT INTO LAGRANGE EQUATION

EOM : m&x& + cx& + kx = F (t )

SOLUTION OF m&x& + cx& + kx = F (t )

To obtain the PI, we must know the RHS, F(t).

When a linear system is subjected to a harmonic

We will apply : Iθ&& = ∑ Moment