Contents

Introduction
2.Vibration basics
2.1 Definitions and Classifications.
2.2Harmonic vibrations.
2.2.1Sinusoidal representation. Amplitudes, time and frequency domains.
2.2.2Phasors representation
Beating phenomenon
2.3 Periodic vibrations. Fourier series and harmonic analysis. Fourier transforms.
2.4 Mechanical parameters: mass, spring dumpers.
2.5Degrees of freedom
3. Free vibrations
3.1 Free vibrations in systems SDOF. Natural frequency. Crocker ch 12 p180
Nonlinearity DLI
3.2 Free vibrations in MDOF. Natural frequencies.Modal analysis
3.3Free vibrations in continuum systems B6K
4.Forced vibrations FREQUENCY RESPONSE SYSTEMS TO VIBRATIONS AND SHOCKS B&K
4.1 Forced vibrations in SDOF systems pioneer
4.1.1 SDOF harmonic excitation,
4.1.2SDOF FRF pioneer Applications pioneer
4.1.3Periodic excitation pioneer
4.1.4Non/periodic excitation pioneer
impulse excitations Response models B&K, DLI
4.2Forced vibrations in MDOF systems
4.3Forced vibrations in continuum system
4.4 Role of excitation DLI comments, harmonics, sidebands
What is linearity DLI
L inearity management harmonics in a spectrum DLI Eng. Corp.
Normal, linear Log and VdB vibration Velocity units. DLI Eng. Corp.
Demodulation Phase measurements
Crest Factor
Mobility
Linearity management harmonics in a spectrum DLI Eng. Corp.
Source of excitations

Procedures for stationary random data analysis Harris 222 22.19

 Introduction

Rolling bearings are rated to prevent the initiation of rolling contact fatigue, (RCF).

However, nowadays, due to material and technology improvements, the RCF comprises only a small
fraction of common failure types. Unfortunately, most of failures are caused by bearing operations outside
of recommended practice: bad mounting procedures, misalignment, poor lubrication, contamination, the
rolling bearings can develop prematurely failures. These ahead of time failures are usually accompanied
by an increase in bearing vibration and therefore the condition monitoring was used for many years do
detect degrading bearings before they catastrophically break down.

The sources of bearing vibration are discussed along with the characteristic vibration frequencies that are

Based on the characteristic vibration signatures which rolling bearings exhibit as its rolling surfaces
deteriorate, nowadays vibration monitoring has become a part of many planned maintenance regimes.
However, in most of practical situations bearing vibration cannot be measured directly. The signal provided
by the bearing travels through a mechanical structure with structural resonances which may significantly
alters it before being captured by the measuring transducer. Even worse, the acquisitioned signal
incorporates vibration data from other transmission parts (gears, chains, belts etc)and equipments ( electric
motors, hydraulics). All these make the interpretation data difficult other than by a trained specialist and in
some situations lead to wrong diagnosis
 2. Vibration basics

2.1 Definitions. Classifications

In the absence of any external action, the elements of a mechanical system are positioned in the
reference states. Mechanical vibrations are alternating movements of the component masses of
mechanical systems with respect to their reference states.

Vibration data are acquired by appropriate transducers that generate analog electrical signals
representing instantaneous values of the parameters of motions (accelerations, velocities and
displacements), forces and specific strains, as functions of time. A sample record, representing a single
vibration measurement x(t) over a duration T is called time-history.

A stationary vibration is one whose basic proprieties do not vary with time. Mechanical equipments running
in their normal regimes, with constant speed and loading are accompanied by stationary vibrations.
Stationary vibrations may have a deterministic or a random evolution in time.

A deterministic vibration follows an established pattern so that the value of the vibration at any future time
is completely predictable from the past history.

A random vibrations is one whose future basic proprieties are unpredictable except on the basis of
probability.

A non-stationary vibration is one whose basic proprieties vary with time, but slowly relative to the lowest
component frequency of the vibration. Mechanical equipments running in transient regimes, as speed up
or speed down are accompanied by non-stationary vibrations. Non-stationary vibrations may have a
continuous time evolution or a transient one.

From an energetic point of view vibrations are classified as free vibrations and forced vibrations.

In free vibrations, the vibration movement is the result of an initial disturbance and there is no energy
supplied to system to maintain the vibratory movement. The damping exists in any real system and causes
a fast diminishing to zero of the free vibrations amplitudes.

In forced vibration there is a continuously energy supply toward the vibratory system that compensate the
damping losses and maintain the amplitudes at a energetically balanced value. The values of the basic
proprieties of the forced vibrations depends on the excitation that introduces the energy in mechanical
structure as well as on elastic proprieties of the system, the later being expressed mathematically by
frequency response function.

Shock is a transient vibratory motions induced by an excitation having the form of a pulse or step that acts
theoretically over an infinite short time. Practically are considered shocks any transient excitation acting
over a time shorter than the fundamental period of natural free vibration of the system. The vibration
induced by a shock excitation includes both the frequencies of the excitation and the natural frequencies
of the system. The shorter the shock is the more pregnant are the natural frequencies of the system in the
resultant vibratory movement. The popular SPM testing method is based on this particularity.
2.2 Harmonic vibration.

2.2.1 Sinusoidal representation

The pure sinusoidal movement, (Eq. (2.1) and Figure 2.1), is used to define the basic descriptors of a
vibratory movement in both time and frequency domains.

(2.1)

where A is called the amplitude of displacement, ω is the angular frequency and θ represents the initial
phase. The sinus movement is a continuous periodic motion having the period T. Figure 1. The circular
frequency and angular frequency are defined as:

, and , [rad/s] (2.2)

Fig. 2.1- Basic descriptors of a sinus vibratory movement

In the time domain the sinus vibratory movement has continuous harmonic variations with the same
frequency for the displacement, velocity and accelerations. In the frequency domain the sine movement
has a discrete representation with only one frequency component (Figure 1).

2.2 2 Complex representation of harmonic vibration

The graphical representation of a vibration as a linear time dependent sinusoidal motion has the first
disadvantage that both time and phase are represented along x-axis and the second one that
mathematical concept of negative frequency appears meaningless.

Another way of representing harmonic oscillations is using the complex numbers. Euler’s formulae allow to
write any complex number can as:

(2.3)
and,

(2.5)

The complex number carries amplitude and angle information and is called phasor of the harmonic
motion. Figure 2.2(b) illustrates the sinusoidal motion as a vector sum of two contra-rotating vectors, each
with amplitude A/2, and the same absolute values for the angular frequencies and initial phases but
different signs. From Figure 2.2b it can be seen that when the contra-rotating vectors rotate with time, the
imaginary parts cancel out so the vector sum will always be real and will trace out the harmonic curve
illustrated in Figure 2.2(a).

Fig. 2.2.- Basic descriptors of a sinus vibratory movement by contra-rotating vectors (phasors)

2.2.3 Beating phenomenon.

If the vectors represented asynchronous vibration, vector addition applies, but at any instance of time the
resulting vector has different magnitude because the two vectors rotate with different angular frequency. If
the two harmonic motions, hence phasors, have the same amplitude but slightly different frequencies there
are time instances when the two phasors will have a phase difference of , that gives a
zero value for the resultant amplitude. Using the trigonometric representation we have:

(2.5)

(2.6)

The resultant motion x(t) is a cosine vibration of angular frequency equal to , but having a time
dependent amplitude, Figure 3. Whenever the amplitude reaches a maximum, there is said to be a beat,
and the time evolution is called beating phenomenon. The beat frequency and period are:

(2.7)
Fig. 2.3 – Beating phenomenon

2.3 Periodic vibratory movement

A vibratory movement is periodic if there is a time period T that fulfills the equation:

………………………….….…………………………..(2.8)

Most of real vibrations have periodic evolutions, but very few of them are pure harmonics.

2.3.1 Time domain description

Because the amplitude A and period T are not sufficient to characterize the time evolution of a non-
harmonic motion along one period, new parameters represented by arithmetic average RA-av, root mean
square average RRMS and the Crest Factor were needed to be considered, Figure 2.1. For the case of pure
harmonic motion the Crest Factor = .

The elastic energy accumulated by the linear spring along one period is:

(2.9)

Divided this energy by one period the average power along one period is found as being proportional with
the square of XRMS revealing an important physical significance which explains the large utilization of X RMS
versus XA-av.

2.3.2 Frequency domain description. Harmonic analysis and frequency spectrum

To evaluate both the effect of vibration on the mechanical structures and the necessary measures to limit
the possible damages it is very useful to have a clear description of the frequencies content of the
vibratory movement. That is achieved by frequency analysis method.
In conditions of Dirichlet’s restrictions the theorem of Fourier establishes that a periodic vibration can be
represented with arbitrary accuracy as a finite or infinite sum of harmonic movements which have their
angular frequencies as multiples of the fundamental one:

(2.10)

(2.11)

The procedure is called the harmonic analysis of the periodic function x(t). The angular frequency Is
called fundamental and the movement x(t) is considered as sum of harmonic movements that have
frequencies equal to the fundamental and its integer multiply. The sum from equations (2.11) is called the
Fourier series, where the constants Ar , Br and Xr are called Fourier coefficients and are mathematically
formulated as:

(2.12)

(2.13)

The motion expressed by the equation (2.11) can be easily written as a sum of sinusoidal motions having
angular frequencies multiplies of the fundamental one (Eq. (2.10)):

(2.14)

In Eq. (2.14) the amplitudes Xr(r) and initial phases are :

, (2.15)

This procedure is called the harmonic analysis of a periodic motion ( or, more generally, periodic functions).

The harmonics can be plotted as vertical line on the amplitude versus frequency diagram called a
frequency spectrum or a spectral diagram.

The spectrum of squared amplitudes is known as the power spectrum, and offers information on how the
vibration power is divided on different harmonics. However, the power spectrum does not contain
information regarding the initial phases.

A reasonable accuracy is obtained even in the sum from Eq. (2.10) the first terms are considered only. This
statement will be sustained by two examples.

Example 2.1 Fourier series analysis of a rectangular wave.

The function x(t) of the hypothetic square motion is expressed as:

The Fourier series coefficients are obtained from Eqs.(2.12) and (2.13):
Fig. 2.4– Frequency spectrum for a rectangular wave

Exemple 2.2 The motion of the piston exemplified in Figure 2.5 is described analytically by the equation:

Only the first two terms attain significant values, and consequently the piston acceleration becomes:
The two components of the sum from Eq.(2.15) represent the frequency spectrum of the piston motion; a
suggestive description is got if amplitudes of acceleration are presented as function of frequency( Figure
2.5).

Fig. 2.5 Periodic non-harmonic motion of a piston and its harmonic components.

2.3.3 Complex form of Fourier series

The complex representation of Fourier series presents any periodic motion as a sum of contra-rotating
vectors at equally spaced frequencies .

……………………………………………….(2.16)

The amplitude of the r component is obtained from the integral:

(2.17)

In Eq. (2.17) the quantity to be integrated is a product between the periodic movement x(t) and the unit
phasor which rotates at the circular frequency . If the periodic movement x(t) contains
component rotating at a circular frequency of then its product with the mentioned phasor annulus the
rotation of this movement component such that it integrates with time at a finite value, Figure 2.6(a). All
components at other frequencies will still rotate even after multiplication by the mentioned unit phasor and
thus integrate to zero over the periodic time, Figure 2.6(b).

Fig. 2.6 – (a) Integration of a non-rotating vector to a finite value; (b)Integration of an phasor to zero

The Eq. (2.10) has the effect of extracting from the movement x(t) the components it contains which rotate
at each frequency , and also “freezes” the phase angle of each as that existing at time zero (when
The actual position of each vector at any other time t can thus be obtained by multiplying its
initial value by the oppositely rotating unit vector . Consequently, the movement x(t) will be the
sum vector of all these vectors in their instantaneous positions. That is the physical meaning of Eq. (2.16).

The series of complex values of represents the complex spectrum components of the vibratory
movement x(t). Because each frequency component contains information relative to amplitude
and phase (equivalent real and imaginary part) the complex spectrum needs a 3D representation, Figure
2.7.

Fig. 2.7 – 3D representation of the complex spectrum of a periodic movement.

As in the trigonometric Fourier series analysis, the complex Fourier series analysis points out that a motion
that is periodic in the time domain has a discrete frequency spectrum that has all its spectrum components
fall at frequencies which are integral multiples of the fundamental frequency. However, the phasor
representation offers an intuitive explanation: the time for one rotation of the phasor at the fundamental
frequency is the one time period. All the other phasors rotate at speeds which are integer multiples of
so each of them rotate its own integer number of turns during the movement period and all have returned
to their starting positions, and the whole process will begin to repeat exactly.

Because the time movement x(t) is a real-valued function, each component at frequency must be
matched by a component at which has equal amplitude but opposite phase. In the complex plane
that means equal real part and opposite imaginary part that represent two complex conjugate complex
numbers:

In this way the imaginary parts of all frequencies will always cancel and the resultant will be always real.

(2.18)

Because the series of imaginary parts (or equivalently phase angles) is anti-symmetric around zero
frequency, the zero frequency (or DC) component has zero (or ) phase angle and is always real.

2.3.4 Power of a time periodic motion. Power spectrum and Parceval’s theorem.

Time domain analysis. The instantaneous power of the motion is equal to . The mean power
over one period is given by integrating the instantaneous value over one period (that represents the
energy along one period) and dividing it by the periodic time:

(2.19)

For a typical harmonic component this results in:

(2.20)

(2.21)

The power content at each frequency is obtained directly by the square of the amplitude of the Fourier
series component. The large of usage of the root mean square value which, directly connected with mean
power, becomes clear. The distribution with frequency of the power content in the vibratory movement
represents the its power spectrum.

Frequency domain analysis. In the frequency domain, except for the DC component the amplitude of any
is , and thus the square of this is . The amplitude spectrum is even and the negative
frequency component (from so the square of its amplitude is also . The total mean power
associated with the frequency will be , the same as obtained in the time domain.

Parceval’s theorem. The total power obtained by integrating the squared instantaneous motion amplitude
with time and dividing by this time are equal with the total power obtained by summing the squared
amplitudes of all frequencies of the frequency component. This is called Parceval’s theorem.

2.3.5 Fourier transform

Letting the period the Fourier series can be extended to non periodic motions. In the case of
the spacing 1/T between the harmonics tends to zero and the amplitudes Cr(f) become a continuous

function of linear frequency . Also, in the assumption of infinite period, the equations (2.17)

and (2.16) tend to:

) (2.22)

(2.23)

The equations (2.22) and (2.23) represent the Fourier Transform Pair.

- the Eq. (2.22), called the forward Fourier transform, converts the motion x(t) from time domain into the
frequency domain, whereas

- the Eq.(2.23), called the inverse Fourier transform, converts the frequency spectrum X(f) from frequency
domain into the time domain.

The Fourier transform decomposes a wave form into harmonics.

FOURIER INTEGRAL TANSFORM TABEL Randal fig. 2,6

Fig. – 2.8 Various forms of the Fourier transform

Experimental harmonic analysis. As result of a measurement the vibration x(t) is obtained as a number of
points x1(t1), x2(t2), x3(t3),…, xN(tN).

If times t1, t2, t3,,…, tN are not arbitrary, but are taken at N+1 equidistant points over the period T, the
constant value X0 and the coefficients Ar and Br can be calculated numerically replacing the integrals in
Eq. (2.11) by summations and with the differential dt with

(2.24)

………………(2.25)

However, practically is quite difficult to determine the beginning and the end of the period itself. In a
situation like this, the samples are taken over a number of periods, and as the period is used the sum of
these periods, that is the true length of the sample. If the vibration is true periodic and the sampling is over
an exact multiple of the period, the first few terms will turn out to be zero, because a periodic motion of
period T cannot have harmonics of period greater than T. For example, if the above formulae are applied
of a period of 5T the first five terms of Ar and Br will be at least A5 and B5. Also, in case of practical
measurements of the values , small harmonics of the period greater than T, called sub-harmonics,
will always be present. The main causes of sub-harmonics are numerical inaccuracies and non-exact
periodicity of the measured vibration, assumed a periodic one.

Sampled time functions case. A sampled time function is represented by a sequence of function’s values of
discrete equi-spaced points in time. The sampled time functions are used in the computer data acquisition.
For sampled time functions the Fourier Transform takes the particular form:
 , and (2.26)

where .

Discrete Fourier Transform. When the sampling operation is achieved in time as well as in frequency
domains, both time motion and frequency spectrum are implicitly periodic. The forward and inverse Fourier
transforms are:

, and (2.27)

Because the infinite continuous integral of Eqs. (2.18) and (2.19) have been replaced by finite sums, the
above forms represents the Discrete Fourier Transform pair or DFT, and is the form used in computer
analyses.

The Nyquist-Shanon sampling theorem. For a Discrete Fourier Transform (DFT) to represent a vibration
accurately, the original vibration must be sampled at a sufficiently high rate.The appropriate rate of
uniformly sampled time series is determined by Nyquist-Shanon theorem. The sampling theorem states that
any continuous baseband signal may be identically reconstructed if the signal bandwidth limited and the
sampling frequency is at least twice the highest frequency of the baseband signal. If a time signal is
sampled uniformly, then the frequency corresponding to one-half rate is called the Nyquist frequency. The
Nyquist frequency describes the high frequency cut-off of the system doing the sampling and therefore is a
propriety of that system. Any frequency exists in the original movement which are at higher frequency than
the Nyquist frequency will be aliased to other lower frequency in the sampled band.

Examples.

The stroboscope is an aliasing device designed to represent high frequencies as low ones, even zero
frequency when the picture is frozen.

The human ear can hear sounds with frequency from 20 Hz up to 20 kHz. Therefore, near –perfect audio
digital recording systems must sample audio signals at at least 40 kHz to be Nyquist sampled. Practically
audio CDs are sampled at 44.1 kHz which allows imperfect low-pass audio filters to filter out higher
frequencies which would otherwise be alised into the audible band.

Power spectra. The conservation energy principle requires that the energies in the time and frequency
domains are equal, that mathematically implies that the integral of the squared modulus of the function
equals the integral power spectrum:

(2.29)

The last equation represent Parceval’s thorem.

2.4 Mechanical components and parameters

A mechanical structure is a combination of masses, springs and dampers. In a linear approach, if any of
these is exposed to a constant force, Figure 2.9, they react, with a constant displacement, a constant
acceleration and a constant velocity, respectively. In rotational motion the displacements, velocities and
accelerations are defined as angular motions, angular velocities and angular accelerations r, and
correspondingly the stiffness and damping coefficients are defined with reference to these parameters,
(Table 1).
During the vibratory motion there is a permanent alternating transfer between the kinematic energy of the
moving masses and potential energy stored in the stressed springs. If dampers were present some energy is
dissipated in each period of the vibration movement so that its level is continuously diminished till vanishes,
unless an equivalent amount of energy was supplied from an external source.

Fig. 2.9 – Effect of a constant force on a spring, a damper and a mass respectively.

Table 1. Analogous quantities in translational and rotational vibrating systems

Translational quantity Rotational quantity

Linear displacement Angular displacement

Linear velocity Angular velocity

Linear acceleration Angular acceleration

Force Torque T

Spring stiffness K Spring stiffness

Damping coefficient C Damping coefficient

Mass M Moment of inertia J

Elastic force Elastic torque

Damping force Damping torque

Inertia force Inertia torque

2.5 Degrees-of-freedom of a mechanical system

The independent coordinates required to locate and orient, at any instant of time, each mass in the
mechanical structure are defined as generalized coordinates. The number of generalized coordinates
represents the number of degrees-of-freedom (DOF) of the respective structure.

A point mass has 3 DOF since the location of the point mass requires knowing the x, y and z translations. A
rigid body mass has 6 DOF since besides the x, y and z translations needed to define the location, the
and rotations are supplementary required to define its orientation at any instant of time.

Any real system is in fact a continuous structure with continuously distributed mass and elasticity with infinite
points whose movements are elastic waves that travel along the system. The waves are mathematically
described by a system of partial differential equations, usually difficult to be solved. However, it is quite
usual to view a general mechanical structure in terms of a finite number of physical points of interest with 1
to 6 DOF for each. In such a discrete system the any element movements are harmonic motions,
mathematically described by a system of ordinary differential equations, much easier to be solved.

In many applications the real mechanical system is modeled as lumped parameter type system:masses are
assumed as rigid bodies where all points within body moves in phase;elastic elements (springs) are
assumed to have no mass. For the automotive car presented in Figure 2.10(a) the modeling depicted in
Figure 2.10(b) allows the mass Mcar to perform only the vertical motion x(t) without any rotation, and
represents the single degree-of-freedom (SDOF) modeling. A more realistic approach considers the masses
of different aggregates as being interconnected by a number of springs and dampers that lead to a multi
degrees-of-freedom modeling.

Fig.2.10 – Modeling of car structure a-single degree of freedom : (a)-car to be modeled, (b) un-damped SDOF modeling,
(c)- 2DOF modeling, (d)-MDOF.

The Figure 2.10(c) exemplifies a two degrees-of-freedom (2DOF) modeling, where the generalized
coordinates x1(t) and x2(t) allow the mass Mcar to perform a vertical translation as well as a rotation around
a transversal axis. The Figure 2.10(d) exemplifies a more complex five degrees-of-freedom (5DOF) car
modeling.While in lumped parameter systems all points within a mass are supposed to move in phase, this is
no longer true for continuous systems which have the mass and elasticity continuous distributed. The
versatile finite element method studies the vibrations of continuous systems by replacing the continuous
mass and elasticity by a large number of discrete elements.
 3.Free vibrations. Natural frequencies
Basic model.

A SDOF system necessary contains two elements: a mass and a spring, but in most of practical SDOF
structures a damper is present, Figure 3.1.

The inertia and elastic parameters are obtained from physical calculations whereas the damping
coefficient is established from measurements. The vibratory motion is the result of an initial disturbance (free
vibration) or of an excitation represented by an external force f(t) acting on the mass or a support
displacement u(t) (forced vibration).

As long as every component element is known the explicit form of vibratory movement is obtained as the
solution of the equation of motion.

Fig. 3.1– SDOF model

For a linear SDOF modeling the equation of motion is represented by an ordinary differential equation
obtained using analytic mechanic methods (Newton’s second law of dynamics, D’Alembert’s principle,
Lagrange’s equation).

(3.1)

The force of inertia, damper reaction force and spring reaction force represented by the three terms in the
left-hand side dynamically balance the external force.

The importance of the role of SDOF systems in vibration theory derives from that any linear system behaves
like:

- a SDOF system near an isolated natural frequency, and

- as a connection of SDOF systems in a wider frequency range.

3.1 Natural frequency of SDOF system

For no external loading the non-homogeneous equation (3.1) transforms in the corresponding
homogeneous form (3.2):

(3.2)
The SDOF system moves free as result of an initial disturbance which mathematically provide the initial
conditions: and needed to particularize the general solution of the free
vibratory movements.

3..1.1 Natural frequency for an un-damped SDOF system

For un-damped SDOF system Figure 3.2, the Eq. (3.2) becomes:

(3.3)

with the solution x(t):

(3.4)

that represents a pure harmonic motion, Figure (3.2). The term:

(rad/sec) (3.5)

represents the angular natural frequency of the SDOF system. The corresponding time period T and linear
natural frequency f are:

(sec) ; (Hz) (3.6)

The angular natural frequency depends on the inertial (m, J) and elasticity (k) parameters of physical
components of the vibratory system, but does not depend on type and value of the initial disturbance,
Figure 2.9. The data referring to initial disturbance establish the values for the integration constants, the
amplitude C and initial phase , respectively.

Fig. 3.2– Free vibrations of an un-damped SDOF system

3.1.2 Natural frequency for a damped SDOF system

A viscous linear damper having the damping constant c is considered additionally to the mass m and the
spring k, Figure 2.10. A critical value for damping coefficient is defined as :

, (3.7)

The position of the actual damping constant relative to the critical value is nominated by the ratio
called the fraction of critical damping. The equation of motion is given by the solution for the Eq. (2.24). The
form of the solution depends upon the value of the damping coefficient c versus the critical value c c

Case 1. Less than critical damping, . The solution is a harmonic motion having amplitudes
decreasing along time:

(3.8)

where the natural angular frequency is depending on the values of mechanical parameters of the
damped SDOF system, but not on the type or value of the disturbance that created the vibratory motion :

= (3.9)

The presence damping alters insignificantly the natural angular frequency, but determines a sharp
diminishing of the amplitudes of free damped vibrations, Figure 3.3. After a small number of oscillations the
free vibration movement of a damped system disappears.

Fig. 3.3 - Free vibrations of a SDOF system with under-critical damping

Case 2 ) and case 3 do not create oscillatory motions. After the initial
disturbance the body is moving exponentially to the reference position, Figure 3.4.

TO BE MODELLED

Fig. 3.4 - Free vibrations of SDOF systems with critical damping and overcritical damping.
3.2 Free vibrations in MDOF systems. Natural frequencies.

Basic proprieties of MDOF vibratory systems are identical to those of SDOF systems:

I. the number of natural frequencies in a MDOF system is equal with the number N of
degrees of freedom;
II. the natural frequencies of a MDOF system depend on the system parameters but not on
the initial disturbances;
III. free vibrations are exponentially decaying harmonic motions carried out on one or more
of natural frequencies.

As consequence of the multi-DOF nature these systems present specific proprieties as:

I. the existence of normal modes of vibration and their orthogonality,

II. the decomposition of arbitrary vibration in normal modes.

3.2.1Equations of motion for a MDOF system

Landau’s uniqueness theorem. If all components of the position vector and all
components of the velocity vector for N particles are given at some instant,
the all components of the acceleration vector at that instant are uniquely
defined. The relations between accelerations, velocities and position vector are called equations of
motion.

Galileo’s relativity principle. The accelerations vectors are the same in all inertial frames, and consequently
the laws of mechanics are the same.

For simplicity we consider a general masses-springs-dampers system that is completely described by N

displacements written as components of one displacement vector:

(3.10)

Using analytic mechanic methods a set of linear ordinary differential equations is obtained as the first form
for the equations of motion,(EOM), of each generalized coordinates,(components of the displacement
vector)

(3.11)

where:

is a symmetric positive defined inertia matrix of order NxN;

is a symmetric non-negative damping matrix of order NxN;

is a symmetric non-negative stiffness matrix of order NxN,

(3.12)

The equations in the system (3.11) are coupled equations.

 3.2.2 Free vibrations in an un-damped MDOF systems

Free vibrations of MDOF systems corresponds to solutions of homogeneous system:

(3.13)

For an un-damped MDOF system all the elements of the matrix C are zeros, and the corresponding system
of differential equations that describes the free vibrations in un-damped MDOF system is:

(3.14)

Generally the coefficients of the equation (3.14) are time independent so we study the feasibility of
harmonic natural vibration of the MDOF system at a single angular frequency :

(3.15)

where is the amplitudes vector. Substitution (3.15) in (3.14) yields a system of linear
algebraic equations:

(3.16)

The linear algebraic system (3.16) is homogeneous and consequently it has a non-trivial solution only if its
determinant is null:

= 0. (3.17)

The equation (3.16) is called the frequency equation and its determinant is called the frequency
determinant. The Eq. (3.17) has N real roots:

(3.18)

meaning that a linear MDOF system can undergo harmonic vibrations at N angular frequencies
representing the spectrum of un-damped natural frequencies of the particular MDOF system. Some of the
frequencies might be multiple. As in the case of SDOF system, the natural frequencies of a MDOF system
depend on the system parameters embedded in M and K, but not on the initial disturbances of each mass.

Eigenvectors. Corresponding to each natural frequency, there is a solution Ak of the algebraic

system (3.16):

(3.19)

The elements of Ak are real valued and equal to the amplitudes of the mass displacements when the
system is vibrating at the frequency

The pair defines the k-th natural or normal mode of vibrations: reveals the angular frequency
of the vibration while Ak sets up the displacement configuration of each mass. The system vibrates
synchronously with angular frequency ; during this synchronous vibration the amplitudes develop a
particular pattern of which its shape does not change throughout the motion. Because the determinant
is singular there are many solutions of Ak. It is common to constrain the solution by setting the
magnitude of Ak to be one, and to have the positive value for the first element of the vector Ak.
The vector xk is termed as eigenvector or mode shape.

Free vibrations. The free vibration is the result of an initial disturbance of the vibrating system. The
disturbance is characterized by the initial conditions referring to displacements
and velocities.

Because of linearity, the sum of the solutions is also a solution, and consequently the general expression of
the free un damped response can be written as:

(3.20)

where the amplitudes and the phase shifts are established from the initial conditions regarding the
displacement and velocity vectors,

3.2.3 Modal analysis

The free vibrations of MDOF systems are governed by the system (3.13) representing a set of ordinary
differential equations, ODEs. The equations incorporated in the system (3.13) are coupled equations, so isn’t
possible to solve each equation independently.

Modal analysis is the procedure for solving the simultaneous system of ODEs. Modal analysis method
transforms the system of coupled ODEs into a set of independent ODEs of which its solution is easily
obtained. The system solution is formed by combining the individual solutions and transform the solution
back to the original system.

The free un-damped vibrations of a linear MDOF system are modeled by the following ODEs system

(3.21)

The equations embedded in (3,21) are coupled equations. The idea is to change the system coordinate
from x to new coordinate system q such a way to obtain un-coupled system of ODEs. In this purpose the
linear transformation x = Aq is applied, with A formed with the modal vectors Ak of the MDOF system,:

(3.22)

Both sides of (3.22) are multiplied by transpose of the matrix A:

(3.23)

where M’, K’ and N are the mass matrix, stiffness matrix and generalized force vector in the new coordinate
system q. Because the modal matrix A was chosen as the transformation matrix the M’ and K’ are diagonal
matrices and the new form (3.23) represents a system of un-coupled ODEs.

If further the matrix A is chosen so , the matrix A is said to be orthonormal with respect to M
and K, and the modal coordinates vector q is called normal coordinate vector. The consequence is that
the stiffness matrix K’ becomes a diagonal matrix having the elements of the angular frequency vector
as diagonal elements:
 (3.24)

The matrix is called the spectral matrix.

The orthogonality propriety is fundamental to modal analysis: it allows to consider the response as a linear
combination of natural modes.

Application. A car is modeled as a 2DOF system with two linear springs and un-damped movements, Figure
3.5. The car parameters have the following values: mass M=1500 kg, moment of inertia IG =2000 kgm2,
kr=36000N/m, kf=40000 N/m, a=1.3 m, b= 1,7 m.

Fig. -3.5 A 2DOF model for un-damped free vibrations of a car.

Derive the differential system of equations of motions.

Calculate the natural modes of the system and write an expression for the free response.
Determine the response if the system is subjected to the initial configuration x0=0.5m.

The differential equations of motions are obtained by writing dynamic equilibrium equations for the two
admitted degrees of freedom.

The matrix form of the system of differential equation is:

,
The elements and of spectral frequencies vector are determined from frequency determinant
equation (3.17), = 0.

and ;

=91.8571 and .

The elements of matrix A are determined from substituting each angular frequency into the frequency
equation (3.16), , that is solved for A.

For :

, .

For :

The first natural vibration mode is and the second natural mode is .

The general free vibrations response is a linear combination of the two modes:

The constants are determined considering the initial conditions referring to displacements
and velocities.

The two natural modes , and

are schematically represented in Figure 3.6.
Fig. 3.6 Natural modes of vibration for the 2DOF car modeling.

This is not a harmonic motion, but it is a linear combination of two oscillations.

Application. pioneer.netserv.chula.acFRF-AAA-Examples
 3.4 Free vibrations in systems with continuum distributed mass and elasticity

Beams, plates shells, rings are examples of continuous vibratory systems. In such systems the inertia, elastic
and damping parameters are continuously distributed, and the number of degrees of freedom is infinite
( ) even the system is size limited.

The main motion phenomena in continuously vibratory systems are wave motions. Any free or forced
vibrations of such a system can be expanded in terms of elementary wave motions.
4. Response of Mechanical Systems to External Excitations.

4.1 Response of linear SDOF systems

When a SDOF mechanical system is subjected to the external forces, Figure 4.1, the motion x(t) of its mass
is termed the response of the system to the particular forces.

Fig. 4.1 The displacement x(t) of a SDOF system to an external force f(t).

As previously presented, the motion of mass in a SDOF linear system is described by an ordinary linear
differential equation, (ODE):

(4.1, 3.1)

An extremely useful tool in obtaining the exact solution for the displacement x(t) is the superposition
principle, which is applicable to linear systems and their differential equations provided by linear modeling.
The principle of superposition states that the effect of simultaneously super-imposed actions is equal to the
sum of the effects of each individual action.

4.1.1 Response to a harmonic excitation. Frequency response function

Harmonic excitation is common to many system involving rotating and reciprocating motions. Moreover,
periodic excitations can be represented as an infinite series of harmonic functions and by the principle of
superposition, the response is the sum of the individual harmonic response.

Let us to consider the excitation as: = . Introducing it in Eq. (4.1) and after
normalization the Eq. (4.1) becomes:

+2 + (4.2

where is the natural circular frequency of the un-damped free vibration, and is the
fraction from critical damping, both being defined previously, (Eq (3.6) and (3.7)).

If z(t) is the complex solution of the differential equation :

+2 + (4.3)

then the solution x(t) is the real part of z(t); x(t) =Re[z(t)].
We assume that the solution z(t) will have the same form as the forcing function: harmonic function with the
same angular frequency, but different amplitude and phase: . Introducing the assumed
solution in Eq. (4.8) we have:

(4.4)

The quantity:

(4.5)

Is called frequency response function. The frequency response function is a complex function that has the
modulus and the argumernt :

, (4.6)

. (4.7)

The solution for x(t) is:

(4.8)

The frequency response function modulates the harmonic input by the magnitude and phase ,
Figure 4.2. Graphical representations of the functions mentioned in Eqs. (4.7) and (4.8) formed the Bode
diagrams, Figure 4.3.

Fig.4.2 – Frequency response function concept as a black box

Fig. 4.3 – Bode diagrams representing the modulus and phase of FRF for a SDOFsystem.

Particular notes:

- when the ratio is small, and consequently ,

-when the ratio is large and ,

-when the ratio the amplitude might have a sharp increase, that is called as the
resonance regime.

Resonance phenomenon. Resonance is defined to be the vibration response when the frequency of the
excitation equals the natural frequency of the un-damped SDOF system, ( ). At this point, the phase
shift of the response is – radians. The resonant frequency will give the peak amplitude for the response
only when For , the peak amplitude will be at , that is slightly before .
For there is no peak.

4.1.2 Response to an excitation forces created by rotating unbalanced mass

The rotating unbalance masses create centrifugal forces which represent a very frequent excitation force.
A calculation model for SDOF systems is presented in Figure 4.4. The vertical component of the centrifugal
force that creates the excitation force has the particularity that its amplitude is proportional with the square
of the angular frequency:

(4.9

The frequency response function of the system represented in Figure 4.4 is obtained in a similar way as
presented before for the SDOF subjected to a harmonic excitation force.
Fig. 4.4– (a) SDOF excited by a centrifugal force; (b) modulus of the corresponding frequency response function

The differential equation of motion is written considering the force (4.14) in the general form (4.1):

(4.10)

(4.11)

The quantity:
 (4.12)

Is the frequency response function with the modulus and the argumernt :

, (4.13)

. (4.14)

The solution for x(t) is:

(4.15)

The plot of frequency response function is presented in Figure 4.7(b). Particular notes:

- when the ratio is small, and consequently ,

-when the ratio is large and ,

-for sufficiently low damping, the maximum occurs when , and the

maximum value is .

4.1.3 Response to the excitation induced by a harmonic base movement

This type of excitation is met at: vehicles suspensions, sensitive equipment placed on a vibrating
foundation, buildings subjected to earthquakes. A model for the damped SDOF system having the
excitation due to the base movement, y(t), is presented in Figure 4.5.

The equation of motion is written taking care to consider the relative values and for
computation of the elastic and damping forces, respectively.

(4.16)

The Eq. (4.21) is subjected to a normalizing operation and dimensionless parameters are introduced:

(4.17)

The excitation is considered as a harmonic movement:

(4.18)

Introducing the complex form of y(t) into the differential equation (4.21) we have the equation (4.23) where
z(t) is a complex variable:

(4.19)
Fig. 4.5– (a) SDOF excited by a basemovement; (b) modulus of the corresponding frequency response function.

The solution z(t) for the differential equation (4.23) is search as being also harmonic with the angular
frequency equals with that of the excitation, but a different amplitude Z

, and

(4.20

In Eq. (4.24) represents the frequency response function of the SDOF system with the excitation given
by movement of the base. The corresponding modulus for this form of frequency response function is

(4.25)

The graphical dependency on the dimensionless ratio is depicted in Figure 4.5(b).

Displacement transmissibility, is defined as the ratio between the output displacement and input
displacement, and represents how motion is transmitted from the base to the mass at various driving
frequencies.

(4.26)

Force transmissibility, , is defined as the ratio between the force transmitted to the mass through both
the spring and damper, and the force that loads the spring at amplitude Y. The forces transmitted by the
spring and damper are balanced by the inertia force:
 (4.27)

Application to measuring devices: accelerometers and seismographs.

The computation model is presented in Figure 4.6.

Fig.4.6- SDOF computation model for measuring devices (accelerometers and seismometers).

If is the relative displacement then the differential equation of motion is written as:

If then

For the last equation provides:

The graphic presentation of transmissibility is presented in Figure 4.6(b). Particular notes:

I. the value of displacement transmissibility starts from T(0) =0 and remain close to
as long as attains small values comparing to natural frequency,
(accelerometer region),
II. the value of displacement transmissibility approaches unity when the angular
frequency is large comparing to natural frequency , (seismometer region).

Accelerometer. The exciting frequencies met in most of mechanical equipments are situated inside the
audio range, (0…20 kHz). The accelerometer must have a sufficiently high natural frequency to assure the
acquisition data operation in the recommended (, ) frequency domain.The ratio between
the displacements amplitudes of the accelerometer mass and base is:

(4.28)

From Figure 4.7 results that if the ratio , then the quantity .

Fig.4.7- SDOF computation model for accelerometers .

4.1.4 Laplace Transform Method

The Laplace transform of a function x(t) is defined as:

= (4.29)

where s (a complex function) is a subsidiary variable : (rad/s).T o solve an equation of motion

(EOM)

each of its element has to be transformed into the Laplace space.

Similarly , and

Generaly:

The EOM becomes:

Assuming that at steady- state the corresponding solution is :

(4.30)

is called the Transfer Function

(4.31)

If the complex function s is admitted as

(4.32)

the Transfer Function (TF) becomes the Frequency Response Function, (FRF)

Laplace Transform: Example Free vibration in un-damped SDOF system-

EOM : , and the normalized form:

Laplace transform: , and

The Laplace transform table are used to get x(t) from X(s)

i.e ,

4.1.5 Response to the unit impulse excitation

A unit impulse excitation(called Dirac function), is characterized by a force which at time has an
infinite magnitude but encloses unit area, whereas for any time , f(t) = 0.

(4.33)

The impulse excitation acting at time induces a response of the vibratory system which is denoted
, Figure 4.8. It is evident that if , .

Fig. 4.8 The time domain response of the vibratory system to unit impulse excitation

A large force of magnitude that is acting over a short period, can be written as:

(4.34)

The sampling propriety. Multiplying a continuous function f(t) by the quantity , and integrating over
the entire time axis the value of function f(t) at the time t=a is obtained.

(4.35)

Use of Fourier Transforms

The superposition principle together with Fourier transform operation represents a very efficient tool used to
study the response of the linear system to external forces.

The Fourier transform F(f) of the excitation f(t) is first determined and the response of the system to each
Fourier component is determined separately.
Applying the forward Fourier transform to impulse response function transforms it into a complex
frequency response function H(f).

The time response x(t) is obtained as the inverse Fourier transform of X(f) :

Introducing , expanding the regions of integration and considering the statement (4.3) the forward
Fourier transform X(f) can further be written:

The final form:

(4.36)

reveals a very important and useful statement: the forward Fourier transform of the response is the product
between the frequency response function and forward Fourier transform of the excitation force. The
spectrum of the output is obtained very simply by multiplying the input spectrum by the frequency response
function at each frequency. The phase of response is the sum between the phases of frequency response
function and excitation.

(4.37)

(4.38)

Note, that in this formulation the solution is a superposition of steady state response to simple harmonic
excitations. The relation defining the power spectrum is obtained by squaring the equation between
amplitudes:

(4.39)

Note. If the excitation applied at the time t=0 is the unit impulse (Dirac function)then the response is
depicted as h(t) and called the response function to the unit impulse excitation. In this case the Eq. (2.6)
becomes:

The response in the time domain is obtained applying the inverse Fourier transform to X(f), that in this
particular case is H(f):

(4.40
Hence, the response function to the unit impulse excitation and frequency response function formed a pair
of Fourier Transform functions. On basis of the convolution theorem: , and the response in the
time domain is:

(4.41)

The last equations reveals that a time analysis of the response x(t) is not able to provide any direct
information regarding the excitation f(t).

Application. The complex frequency response function H(f) of a SDOF system

A complex frequency response function represents a vibrating system response which gives information on
both the absolute value of the response and the lag between the response and the excitation.

Fig. 4.9 The displacement x(t) of a SDOF system to an external force f(t).

For generality, the excitation is considered an arbitrary Fourier series component, :

Applying the forward Fourier transform to both sides and using Fourier Transform proprieties we have:

(4.42)

where

is the natural circular frequency of the un-damped free vibration, and is the fraction

from critical damping, both previously defined (Eq (3.6) and (3.7)).
The modulus and phase of the frequency response are as follows:

(4.43)

(4.44)

4.1.6 Response of a SDOF system to periodic excitations

A periodic excitation repeats itself in time, called period T:

(4.45)

By using the Fourier series any periodic excitation can be decomposed in a sum of many harmonic
excitations whose frequencies are multiples of the fundamental frequency (see chapter 2.3.4). In the case
of linear systems (linear springs and linear dampers) the response can be obtained by the superposition the
responses of individual harmonics.

Plot of the amplitude of each harmonic component versus its frequency represents the frequency
spectrum. For periodic excitations that is a discrete frequency spectrum. Harmonic decomposition and
discrete frequency spectrum is exemplified in Figure 4.10 for the case of a square wave excitation.

To be updated Fig. 4.10 – Harmonic decomposition of periodic square wave excitation – (a) time domain; (b) frequency
domain.

(4.46)

In the base of superposition principle, Figure 4.10 and Figure 4.11, the response is written as the sum of
responses to each Fourier component of excitation:

(4.47)

where is the frequency response function of the linear SDOF system:

(4.48)
To be updated Fig. 4.11 – Response of the linear system to components of a periodic square wave
excitation.

To be updated Fig. 4.12 – Response of the linear system to components of a periodic square wave
excitation.

4.1.7 Response to an arbitrary external force f(t)

Time superposition approach. Duhamel’s integral The idea is to regard the arbitrary excitation as a
superposition of impulses of varying magnitude and applied at different times. The vibratory system is also
admitted as a Linear Time Invariant (LTI) system that means that the shape of the response will not be
influenced by the time the input is applied to the system. The external function f(t) can be considered as
consisting of an infinite number of impulses, each with an infinitesimal width and a height , Figure 4.3.
Based on superposition principle the response of the system to the external function f(t) can be obtained
superimposing the responses produced by the action of each of these impulses. Mathematically this
statement can be written in the form:

(4.49)
The superposition integral represents a convolution, and the excitation f is said to be convolved with h to
produce the response x. Because the convolution operation is commutative we can write the equivalent
form:

(4.50)

where .

The equation (4.50) points out that the response to external force f(t) can be obtained by superposition of
free vibration solutions.

Fig. 4.13 Time domain superposition

4.1.8 Response of SDOF systems to shocks

Shocks may be described in terms of force, acceleration, velocity or displacement. A simple shock may be
defined as a transmission of kinetic energy to a system which takes place in a relatively short time
compared with the natural period of oscillation of the system. The Fourier analysis proved to be helpful to
estimate the effect that the corresponding shock, f(t), would have on a certain mechanical system:

Various shock time functions and the amplitude of their Fourier spectra are presented in Figure 4.14.

Broch 2.13

Fig. 4.14 Example of shock time functions and their amplitude spectra as obtained by Fourier transforms.

a rectangular shock pulse, (b) a peak sawtooth shock pulse; (c) a half-sine shock pulse.

The shocks spectra exemplified in Figure 4.14 point out two importantparticularities:

I. the spectra are continuous with no discrete frequency components;

 II. shock pulse contains energy spread over all frequencies from zero to infinity.

In the expressions for F(t) in the Figure 4.14 all the expressions within the parallel brackets approach unity as
frequency f goes to zero. That means that at very low frequencies the magnitude of the spectrum
component is equal to the area (amplitude-time integral) of the shock pulse, irrespective of the shock
shape. It means that the severity of the shock is determined by the area of the shock pulse alone.

4.1.8 Response of SDOF systems to transient phenomena

Transient phenomena may last for one single period being called “shock wave”, or several periods being
an “oscillating phenomenon”. Transient phenomena are termed also complex shocks. A Fourier spectrum
function of a typical transient excitation is presented in Figure 4.15. Particularities:

I. the magnitude of the spectrum components tends to zero as the frequency goes to zero,
that is a rather different evolution comparing with that corresponding for simple shocks
(Figure 4.14);
II. maximum magnitude of the spectrum is reached around frequency f0 which corresponds
roughly to the frequency of oscillation of the transient;
III. the maximum is broader the quicker the transient phenomenon ceases;
IV. when the transient acts as a harmonic vibration the frequency degenerates into a discrete
spectral line.

Broch 2.14
Fig.4.15 – Oscillating transient and its Fourier spectrum function
 4.2 Response of linear MDOF systems

Fig. 4.16 Response of 2DOFl linear system

Fig. 4.17 Response of a MDOF linear system.

5.Source of excitation
6 Sources of vibrations in rolling bearings

6.1. Bearing characteristic frequencies

6.1.1 Kinematics of rolling bearings for low speed applications

In a first evaluation, in a rolling bearing, mounted on a rotating shaft and a otating housing, exist:

- the inner ring rotation at angular speed of the shaft, ;

- the outer ring rotation at angular speed of the housing, ;
- the rolling elements orbital movement around the bearing axis, at the cage angular speed, ;
- the rolling elements rotation around their own axes with a particular angular speed, .

Fig.6.1 Basic angular speeds in a rolling bearing

When a load occurs between a rolling element and raceway, a contact surface is formed. When the
rolling element rotates relative to the deformed surface, the simple rolling motion does not occur; rather, a
combination o rolling and sliding motions occur. Hence, a system of complex equations needs to be
developed to calculate the rolling element speeds. Also, for angular-contact bearings, if the rolling motion
does not occur on a line exactly parallel to the raceway, a parasitic motion called spinning occurs. Such a
motion is pure sliding contributing significantly to bearing friction power loss. Finally, motions between roller
ends and ring flanges in roller bearings are also pure sliding and can result in substantial power loss.

In most applications, particularly those operating at relatively slow shaft or outer-ring speeds, these internal
speeds can be calculated with sufficient accuracy using simple kinematical relationships; that is, the balls
or rollers are assumed to roll on the raceways without sliding.

In the case of slow-speed rotation or an applied load of large magnitude, rolling bearings can be analyzed
while neglecting dynamic effects. As a general case, it will be assumed that both inner and outer rings
rotate with non-zero angular speeds: ωe or the outer ring, and ωi for the inner ring. It will be further assumed
that a common contact angle exists on both raceways.

Conditions of pure rolling motions of the ball on both outer raceway, (point E), and inner raceway, (point I),
provide the equations to obtain the angular speeds:

 the angular speeds of the cage,

 the angular speed of the ball.

Fig. 6.2 Low speed kinematics of a angular contact ball bearing

If only inner ring rotates:

6.1.2 Bearing characteristic frequencies

We consider the following notations:

= inner ring rotational frequency;

outer ring rotational frequency,

= rolling element spin frequency

= fundamental train (cage) frequency relative to outer ring

 = fundamental train (cage) frequency relative to inner ring

rolling element pass frequency of outer ring

rolling element pass frequency of inner ring

For stationary outer ring and rotating inner ring, the fundamental frequencies are derived from the bearing
geometry and kinematics.

A single defect on a ball or roller would contact both raceways in one ball or roller revolution so that the
defect frequency is . In addition, the defect could contact one or both sides of the cage pocket, but
this will have little influence on vibration measured external to the bearing.

The presented equations assume the rolling elements roll over the raceways surfaces and no sliding is
present. However in practice this is rarely the case and due to a number of factors the rolling elements
undergo a combination of rolling and sliding. As a consequence, the actual characteristic defect
frequency may differ slightly from those predicted, but this is very dependent on the type of bearing,
operating conditions and running internal clearance. This sliding can be taken into account by multiplying
the theoretical frequencies with a sliding factor “e” that usually takes value between 0.9 and 1.0. Generally
the bearing characteristic frequencies will not be integer multiples of the inner ring rotational frequency.
That notice helps to distinguish the bearing characteristic frequencies from other sources of vibration.

Since most vibration frequencies are proportional to speed it is important when comparing vibration
signatures that data are obtained at identical speeds. Speed changes will cause shifts in the frequency
spectrum causing inaccuracies in both the amplitude and frequency measurement.

The rolling element pass frequency divided by the shaft rotational frequency is called the bearing speed
ratio. The actual value of the bearing speed ratio is a function of the bearing loads, internal clearance,
condition of lubrication. If the bearing speed ratio is below predicted values it may indicate insufficient
loading, excessive lubrication or insufficient radial internal clearance, which could result in increased
friction generating higher operating temperature and premature failure. A higher than predicted bearing
speed ratio may indicate excessive loading, excessive internal clearance or insufficient lubrication.

Rolling bearings are a mechanical system whose parts rolling elements, inner and outer rings, and cage
interact to generate complex vibration signatures. Like any other manufactured part rolling bearings have
degrees of imperfections and generate vibration as the surfaces interact during the rolling motion.
Nowadays, also the amplitudes of the surface imperfections are in the order of nanometers, significant
vibrations can still be produced in the entire audible frequency range( 20 Hz – 20 kHz). If a defect was
present on the active raceways a more or less significant change in the vibration pattern of the particular
bearing is detected.

A number of harmonics and sum and difference frequencies are also evident in the spectrum.

In the following the main sources of rolling bearing vibration are discussed with a special attention on
vibration induced by bearing failures.

6.2 Structural elements and variable compliance

Under the applied load the concentrated contacts implied in a rolling bearing manifest local deflection,
and consequently the rolling bearing manifests like a non-linear spring. Figure 6.3. However for each
particular operating conditions a bearing spring constant may be determined by taking the slope of the
force-deflection curve.

Fig. 6.3 The load-deflection relationship in a radial rolling bearing

Rolling bearing present a hardening non-linear characteristic that means the stiffness increases with
increasing load or built-in preload, and consequently increased bearing stiffness raises the value of
resonant frequency associated with this spring. The radial stiffness decreases with increasing contact angle,
whereas the axial stiffness decreases.

Variable compliance

Rolling bearings carry loads with a finite number of rolling elements whose angular positions, with respect to
the line of action of the load, continually changes in time, Figure 6.4. This mere change of position causes
the inner and outer raceways to undergo periodic relative motion. The greater the number of loaded
elements less the movement. For radially loaded or misaligned bearings “running clearance” determines
the extent of the loaded zone, and hence variable compliance increases with clearance.

Fig. 6.4 Bearing model for variable compliance

The movement is periodic with base frequency equal to the rate at which the rolling elements pass through
the load zone. The non-linear character of the deflection-load relationship determines a non-sinusoidal
character of vibrations derived from variable compliance so that the frequency analysis of the movement
yields the base frequency and a series of harmonics. These kind of vibrations are an inherent feature of
rolling bearings, even the bearing is geometrically perfect, but do not indicate a poor quality, and explain
why bearing failure detection is best performed by monitoring frequencies other than the fundamental
bearing frequencies.

Variable compliance levels can be higher than those produce by roughness and waviness of the rolling
surfaces. It can be reduced to a negligible level by using rolling bearings with a sufficiently high axial pre-
load.

6.3 Geometric imperfections as source of vibration

It is convenient to consider geometrical imperfections in terms of wavelength compared to the width of

the contact area achieved between the most loaded rolling element and raceway:

- surface features of wavelength of the order of the contact width or less are termed roughness,

- surface features of wavelength longer than the contact width are termed waviness, (Figure 6.5).

Fig. 6.5 – Waviness and roughness of rolling surfaces versus width of the contact area and thickness of lubricating film

6.3.1 Surface roughness as source of vibration.

When the roughness heights are high comparing with the thickness of the lubricating film, Figure 6.5, the
asperities of contacting surfaces interact randomly resulting a random sequence of small impulses which
excite all the natural modes of bearing and supporting structures. Surface roughness induces vibration
predominantly at frequency above sixty times the rotation frequency of the bearing, that means that the
high frequency part of the frequency spectrum represent a series of resonances.

Local film variation as a function of local surface roughness is usually characterized by so called ‘lambda
parameter’ and defined as the ratio of the minimum film thickness ‘h0’ to the composite roughness of two
surfaces in contact:

where are RMS surface roughness of the raceway and rolling element, respectively.

To evaluate the effect of ratio on film thickness and lubrication quality of common rolling bearings, the
diagram presented in Figure 6.6 is largely used.
Fig. 6.6 – Percent film versus lubrication parameterer ratio.

Evaluation of the roughness values for a particular ehl regime

The simulation has been performed in the following general conditions:

- an EHD regime with the minimum film thickness:

- the roughness of the rolling element has been considered to be finer than that of the contacting
raceway:

The lambda ratio becomes:

Further, admitting that , results:

Minimum Lubrication parameter

EHD Film 1 2 3 4

Thickness

0.2 0.1 0.06 0.04

6.3.2 Waviness as source of vibration.

Waviness can occur in the machining process.A ring type element is compresses at the points of contact in
the chuck , three jaw or five jaw, causing stresses in the ring. The ring is turned or ground perfectly circular;
however when it is discharged from the chuck, the stresses are released and the ring becomes lobed. The
number of lobes per circumference is called waviness. Careful attention is required to the form and
precision of all associated bearing components.Any geometric errors on the outside diameter of the shaft
or bore of the housing can be reflected on the bearing raceways leading to the increase in vibration level.

Waviness produces vibration that are usually predominant at frequencies below sixty times rotational
frequency, but can induce vibration at frequencies up to three hundred times the rotational frequency.

6.4 Discrete defects as source of vibration.

Discrete defects can take a variety of forms: identations, scratches along and across the roling surfaces,
pits, debris and particle in the lubricant. Bearing manufacturers have adopted simple vibration
measurements on the finished product to detect such defects. However this type of measurement provides
vibration data which are valid for that type and size of bearing.

A number of harmonics and sum and difference frequencies are also evident in the spectrum.

Rolling element pass frequency can be generated as a result of elastic proprieties of the raceway material
due to variable compliance or as the rolling element pass over a defect on the raceway. The frequency
generated at the inner and outer ring raceway can be estimated roughly as and ,
respectively. Imperfections on the surface of raceways and rolling elements, as a result of the
manufacturing elements, interact to induce other discrete frequencies and sidebands, Table 1.

Table 1. Frequencies related to surface imperfections

Surface defect Frequency

Component Imperfection

Inner raceway Excentricity

Waviness

Discrete defect

Outer raceway Waviness

Discrete defect

Rolling element Diameter variation

Waviness

Discrete defect

Analysis of bearing vibration signale usually complex and the frequencies generated will add and subtract
and are present in bearing vibratrion spectra. Over the years a number of diagnostic algorithms have been
developed to detect bearings faults by measuring the vibration signatures on the bearing housing. These
methods take advantage of both the characteristics frequencies and the natural frequencies of the
bearing.
 6.4.1Discrete fault on outer raceway

A discrete fault on outer raceway will generate a series of high pulses at a rate equal to the ball pass
frequency relative to the outer ring. Usually the outer ring is stationary and the amplitude of the pulse will
remain theoretically the same and will appear as a single discrete peak within the frequency domain,
Figure 6.7.

Fig. 6.7 -

6.4.2 Discrete defect on the inner raceway

A discrete defect on the inner raceway will generate a series of high pulses at a rate equal to the ball
frequency relativ to the inner raceway. Because the inner ring is rotating, the defect will enter and leave
the load zone causing a variation in the roling element-raceway contact force, hence deflections. While in
the load zone the amplitudes of the pulses will be highest but then reduce as the defect leaves the zone,
resulting in a signal which is amplitude modulated. In the frequency domain this not only gives rise to a
discrete peak at the carrier frequency (ball pass frequency) but also creates a pair of sidebands spaced
either side of the carrier frequency by an amount equal to the modulating frequency (inner ring rotational
frequency), Figure 6.7 and Figure 6.8.

Fig. 6.7 – Amplitude modulation

Notes about the side bands evolution:

I. as the level of the amplitude modulation increases so will the sidebands;

II. as the defect increases in size more sidebands are generatedand at some pointthe ball pass
frequency may no longer be generated, but instead a series of peaks will be generated spread at
the inner ring rotational frequency.

Although defects on the inner and outer raceways tend to behave in a similar manner, for the same
size defect the amplitude of the spectrum of an inner raceway defect is much less.This is because
adefect on the inner raceway comes into the loaded zone once per revolution and the generated
signal has to travel through more structural interfaces before reaching the transducer location.
Consequently a foult on the outer raceway tends to be easier to detect.
Likewise the rolling element pass frequency can also be modulated at the fundamental train
frequency. If a rolling element has a defect it will enter and leave the load zone at the fundamental
train frequency causing amplitude modulation and result in sidebands around the ball pass frequency.

Amplitude modulation at the fundamental train frequency can also occur if the cage is located
radially on the inner or outer ring.

6.4.3 Rolling element defect as source of vibration

Defects on the rollingelements can generate: a frequency twice ball spin frequency (when the defect
strikes both raceways), harmonics andthe fundamental train frequency.

Sometimes the generated frequency may not be so high because the rolling element is not always in
the loaded zone whenthe defect strikes. Also, when a defect on ball is orientated in the axial direction
it not always contact the inner and outer raceway. When more rolling elements are defective, sum of
the ball spin frequency can be generated and if these defects are severe then vibration at the
fundamental train frequency can be generated.

6.4.4 Cage defect as source of vibration

The cage tends to rotate at 0.4 times inner ring speed and has a low mass. Unlike the raceway defects,
cage failures donot usually excite specific natural frequencies. In the case of cage failures the
signature is likely to have a random bursts of vibration as the rolling elements slide and the cage starts
to wear or deform and a wide band of frequencies is likely to occur. As a cage starts to deteriorate
wear can start to occur on the sliding surfaces ( in the cage pocket, or on the cage guiding surfaces in
the case of ring guided cage). This kind of cage wear leads to a less stable rotation of the cage or a
greater excursion of the rolling elements, resulting an increased sideband presence around the other
bearing fundamental frequencies, like the ball spin frequency.

Excessive clearance can cause vibration at the fundamental train frequency as the rolling elements
accelerate and deccelerate through the loaded zone. Consequently large impact forces result
between the roling elements and the cage pockets, Figure 6.8. Also, outer raceway defects and roller
defects can be modulated with the same fundamental train frequency (FTF).

Fig. 6.8- Impact forces between the rolling elements and cage pockets

6.4.5 Cotamination as source of vibration

Contaminants can cause wear and damage to the rolling contact surfacesand generate vibration
across a broad frequency range. In the early stages the crest factor of the time signal increases,but it is
unlikely that this will be detected in the presence of other sources of vibration.

7. VIBRATION MEASUREMENTS

7.1 Introduction

Comparing with other parts of a mechanical transmission ( gears, transmission chains, reciprocating
elements etc) rolling bearings produce very little vibration when they are fault free.

Where machine speeds are very low, the bearings generate low energy signals that may be difficult to
detect. Also, bearings located within a gear box can be difficult to monitorbecause of the high
energyat the gear meshing frequencies, which can mask the bearing defect frequencies.

When inside a rolling bearing a fault develops it provides distinctive characteristic frequencies. A fault
represented by an incipient spall on the raceway is normally dominated by impulsive events at the
raceway pass frequency resulting in a narrow band frequency spectrum. As the damage worsens
there is likely to be an increase in the characteristic defect frequencies and sidebands. Usually this
evolution is accompanied by a drop in the amplitufdes and an increase in the broadband noise with
the considerable vibrationat shaft rotational frequency.

Vibration measurements can be generally characterised as falling into one of categories: detection,
diagnosis and prognosis as basic steps in a general condition monitoring scheme presented in Figure
7.1 and Figure 7.2

Fig. 7.1 Condition monitoring levels

 Fig. 7.2 – Vibration measurements

The time domain measurements provides in fact little information: a fault may be identified, but it may
not give a relible information of where the fault is, i.e. misalignment, unbalance, nature of bearing
deterioration ( inner or outer raceway, rolling element, cage) etc.

When an improved diagnostic capability is required the frequency analysis is normally employed,
which gives a much earlier indication of the development of a fault and, secondly, the source of the
fault.

7.2 Basic Measurement chain

7.3 Time Domain Analysis

Detection uses the most basic form of vibration measurement, where the overall vibration level is measured
in a broad frequency range 10 Hz to 1 kHz, or 10 Hz to 10 kHz. This kind of measurements seems to be reach
in content but provides in fact little information. Although, time domain measurements represents a quick
and low cost method of vibration monitoring, it is less sensitive to incipient defects and has a limited
diagnostic capability. Also, individual contributions ,e.g. unbalance, to overall machine vibration are
difficult to identify. The spikkiness of the vibration signal may be indicated by the Crest Factor or even
better by Kurtosis factor.

7.3.1 RMS and Delta RMS

The Root Mean Square (RMS) value of a vibration signal is a time analysis feature that evaluate the power
content in the vibration signal. The equation used to calculate the RMS value of a data series xi over length
N is:

(7.1)

Delta RMS is simply the difference between the current RMS value and the previous.

The main disadvantage of RMS measurements is that these do not provide any information on which
component is failing. Also, the RMS level may not show appreciable changes in the early stages of folling
contact damages.
 7.3.2 The Crest Factor. I

Crest Factor is defined as the ratio of the peak level of the input signal to the RMS level, Figure 7.3 The
purpose of the Crest Factor calculation is to offer a quick idea of how much impactting is accurring in a
waveform. In a perfecrt sine signal with an amplitude of “1”, the RMS value equal to 0.707, and the crest
factor is equal to 1.41. A sine wave ccontains no impacting and therefore Crest Factorwith a value greater
than 1.41 indicates that there is some degree of impacting. For normall operation, Crest Factor may reach
values between 2 and 6. A value above 6 is usually associated with machinery problems.

Fig. 7.3 Peak Level, RMS Level and Crest Factor

In machine where there is little vibration other than from the bearings, spikiness of the vibration indicated by
the Crest Factor (the ratio Peak/RMS) may imply incipient defects whereas the high energy level provided
by the RMS level may indicate severe defects. Indeed an incipient fault provides short bursts of high energy
which increase the Peak level of the vibratory signal, but have little influence on the overall RMS level. As
the fault progresses, more Peaks will be generated until finally the Crest Factor will reduce but RMS vibration
level wil increase.

7.3.3 The Kurtosis Factor.

The Kurtosis factor represents a better way to point out the spikiness of the vibratory signal, and it is defined
as the fourth centred moment of the probability density function:

(7.3)

where x(t) is the instanrtaneous amplitudeat time t, is the mean value of x(t), and is the sample
standard deviation of x(t).

A rolling bearing in good condition provides waveform vibration with Kurtosis Factor K<3.5.

Rolling bearing faults like spalling on ball/rollers and craks on raceways create impulses in time domain
waveforms that can be picked up by large values of Kurtosis factor, K>3.5.

Trend analysis involves plotting the vibration as a function of time and using this to predict when the
machine must be taken out for repair.
 7.4 Frequency spectrum analysis.

7.4.1 Common frequency spectrum

A fault developing in a bearing will provide increasing vibration at certain characteristic bearing
frequencies. A frequency domain analysis allows an earlier stage detection.

The definition of the Fast Fourier Transform implies that any signal can be approximated as a sum of a set of
sine waves. That works well with periodic events, but some missinterpretations might appear when
non.periodic events as shocks and ramdom noise. Although shocks and random noise are quite diffenent
events they are able to provide the same continuous form spectrum, Figue 7.4. The Crest Factor appears to
be quite useful to obtain a quick idea of what is basically occurring in yhe time waveform.

Fig. 7.4 – FFT spectra for a impact and a random wave.

7.4.2 High frequency resonance technique (Envelope spectrum).

The vibration signal provided by the early stage of developing bearing fault is usually masked by machine
noise and consequently difficult to detect the fault’s presence by spectrum analysis alone.

Enveloping, known also as high frequency resonance technique, is used to monitor the high frequency
response of the mechanical system to periodic impacts such as bearing or gear faults.

Let us consider a fault on the outer raceway. An impulse is produced each time a loaded rolling element
strikes the fault and consequently the outer ring will be excited by a shock type load. The shock load has
an extremely short duration compared to the interval between pulses, and the energy of the defect pulse
will be distributed at very low level over a wide range of frequencies.

This characteristic, low energy level with a wide distribution, is that makes bearing defects so difficult to
detect by conventional spectrum analysis when they are in the presence of higher amplitude vibrations
from other transmission parts or equipments. Fortunately, this shock load may excite resonances of the
outer ring at a much higher frequencies than those of the vibrations generated by other components.
These bursts of high amplitude decays rapidly but they appear again and again due to the fault strikes by
the following roling elements. The resulting time signal will contain a high frequency component amplitude-
modulated at the rolling element pass frequency on the outer ring. But, there is a but, this high frequency
component is very small and very difficult to detect in raw frequency spectrum. The envelope techinque is
a method to enhance the needed component and point out the fault signature.
The high frequency resonance technique (envelope technique) focuses on processing stucture resonance
energy with an envelope detector. The bearing ring resonance is extract by band pass-filtering the data
around the ring resonance frequency. The band-pass filtered signal is then processed by an envelope
detector , which consists of a half-wavwe rectifier and a peak-hold smoothing section. A simple envelope
detector processing flow diagram is shown in Figure 7.3.

Fig. 7.3 Basic flow diagram for a envelope envelope detection process.

The center frequency of the band-pass filter should be selected to coincide with the structure resonance
frequency of the component, in our case the outer ring. The bandpass of the filter should be at least
double the highest characteristic defect frequency, that will ensure that the filter will pass the carrier
frequency and at least one pair of modulation sidebands. In practice, the bandwidth should be sufficiently
large to accommodate the first two pairs of modulation sidebands around the carrier frequency.

The rectifier in the envelope detector turns the alternate filtered signal into a unipolar waveform. After that
the rectified signal is introduced into a low pass filter to envelope, or demodulate, the signal, whose
frequency corresponds to the repetition rate of the defect. In this way the low frequency information is
extruded from carrying resonance frequency band. After these operation the signal is transformed to
frequency domain.

Exemple 1. In Figure 7.4 the Envelope Spectra of the signals btained from the support bearings of a 15 kW
electic motor are exemplified. One bearing was in good condition whereas the other had an fault on the
outer raceway. Both spectra revealthe bearing pass frequency of 89.3 Hz. The calculated value according
Equation ( ) gives the frequency of 89.7 Hz with the sliding factor o.96 (4% slide). The first peak is visible at
24.5 Hz that correspond to the rotational speed. Also the 24,5 Hz sidebands can be found. The highest peak
at 125 Hz in the envelope spectrum of the good conditio bearing is not a bearing pass
Figure 7.3 - Envelope spectra of the signal provided by a bearing in good condition and a similar bearing but with
an incipient fault on outer raceway.

Example 2 Figure 7.4 presents example of spectra obtained from a taper roller bearing with a 432
diameter bore rotating at 394 rot/min, that means a frequency . The bearing shaft was
dirven with a drive shaftspeed of 936 rot/min (2.375 reduction) giving a gear mesh frequency of 374.4
Hz.

The rotation shaft frequency and its harmonics are clearly revealed in the spectrum.

The 374.4 Hz component is related to gear mesh frequency, with sidebands at rotational speed 6.56 Hz.

Another clearly evident frequency is that of 62.38 Hz that corrresponds with twice the rotational
frequency of the roller, accompanyed by a number of harmonics, i.e. 186.5 Hz (x3), 497 Hz (x8), 748
Hz(x12), 873 Hz (x14) and 936 Hz (x15). This presence of this frequency and its harmonics suggest a
tapered roller, or rollers, deteriration , that was confirmed upon examination of the bearing. The
frequency spectrum also shows discrete peaks spaced at cage speed, 2.93 Hz which again is
consistent with roller deterioration.

Bearings defects normally provide a signal which is amplitude modulated. As was already presented,
by demodulating the signal and analysing the encvelope early fault detection can be revealed. Figure
7.5 shows the envelope spectrum, where discrete peaks appear at 62.5 Hz and its harmonics.
Fig. 7.4 Spectra obtained from the housing of a tapered roller bearing
Fig. 7.4 Envelope spectrum obtained from the housing of a tapered roller bearing

Example 3. Figure 7.5 presents a sample of accelerationtime signal collected from the housing of a 23036
spherical roller bearing , located at the main shaft of an impact crusher. The rotation speed of the bearing
shaft was 540 rot/min. The modulation of the time signal at the cage rotational speed is clearly revealed.

Fig. 7.5- Acceleration time signal measured radially on the housing of 23036 spherical roller bearing

The frequency spectrum of acceleration is presented in Figure 7.6 and reveals a number of harmonics
at the outer raceway roller pass frequency, 101 Hz. A dominant peak exists at 404 Hz ( 4xfw-e) at shaft
rotational frequency, 9 Hz. The envelope spectrum, Figure 7.7, presents a peak at 4.0 Hz that
correspond to cage rotational frequency, accompanied by a large number of harmonics

Fig. 7.6-Spectrum of the acceleration time signal presented in Figure 7.5

Fig. 7.7- Envelope spectrum of the acceleration spectrum presented in Figure 7.6
Examples

1. Vibration spectra of a belt driven exhaust fan

Let us examine a simple practical example to illustrate the possibilities of vibration analysis: a belt-driven fan
unit had failed due to excessive vibration. Since the most severe vibration level was measured on the drive
motor, the motor seemed the logical candidate for examination. Vibration analysis showed, however, that
the extremely severe vibration (15.2 mm/s) at the motor was occurring primarily at a frequency which was
conducted to the motor via the belt drive. When the belt drive wheel on the fan was balanced, vibration
decreased to acceptable levels of 2.3 mm/s on the fan and 3.2 mm/s on the drive motor. This case
presents a typical method of operation: a simple measurement of overall vibration level allows the
machine condition to be rated as ‘good’, ‘satisfactory’, ‘unsatisfactory’ and ‘unacceptable’. In the case
of excessive vibration, the root cause - drive belt wheel unbalance - was made clear by checking the
frequency peaks in the FFT vibration spectrum.

Fig. 7.8 Vibration spectra of a belt driven exhaust fan

2. Machine condition trending

A rational approach to successful and effective condition monitoring is that of trending the
development of characteristic overall value measurements of machine condition over time. The trend
readings are plotted as shown here and compared with appropriate warning and alarm thresholds.
When thresholds are exceeded (and not before then), detailed vibration diagnosis is performed in
 order to locate the exact source of trouble and to determine the corresponding maintenance remedy.
Let us examine, then, the vibration monitoring and diagnosis techniques that hold particular relevance
for electric motors.

Fig. 7.8 Machine condition trending

3. Level1 /Level 2 Condition Monitoring Strategy

Machine condition monitoring calls for measurement of suitable vibration characteristic overall values,
which allow the general vibration condition of the machine to be estimated. The trend development of
these characteristic overall values points out condition deterioration, i.e. damage progression. This type of
overall vibration measurement is characterized as ‘Level 1’ as shown here. It allows monitoring of many
aggregates without imposing high demands in terms of equipment and manpower. Characteristic overall
value (Level 1) measurements such as these, however, are insufficient for precise localization of defects, as
this requires closer analysis of the machine spectrum. Most types of damage can be detected by their
characteristic frequencies or typical pattern of frequencies. ‘Level 2’ vibration.
Fig. 7.9 - Level1 /Level 2 Condition Monitoring Strategy

4. Vibration Severity according to ISO Standards

DIN ISO 10816-3 plays a very important role for maintenance technicians in the evaluation of machine
vibrations. Part 3 of this standard, which is the section that is of relevance to Condition Monitoring, has
been revised. Groups 3 and 4 of Part 3, which dealt with pumps, have been removed. Instead, the
standard was expanded to include Part 7 – namely, DIN ISO 10816-7. This new part deals entirely with
vibrations in centrifugal pumps. The new DIN ISO 10816-7 has been in effect since August 2009

Fig. 7.10 - Vibration Severity according to ISO Standards

5. Motor components vulnerable to damage

This illustration gives an overview of the electric motor components most vulnerable to damage.
Some types of damage exhibit typical vibration spectra patterns, and each of these phenomena
shall now be explained in detail.

Fig. 7.11 - Motor components vulnerable to damage

6. Rotor unbalance. Shaft misalignment

Unbalance is understood to be an eccentric distribution of rotor mass. When an unbal-anced rotor begins
to rotate, the resulting rotating centrifugal force produces additional forces on bearings and rotor vibration
at the exact frequency of rotation. This characterizes the spectrum of an unbalanced machine, i.e. the
rotation frequency appears as a ‘peak’ with elevated amplitude, and this can signifi-cantly degrade the
overall vibration condition of the machine. The necessary redistribution of rotor mass is achieved by
balancing the motor rotor either with a balancing machine following disassembly or on-site using a vibra-
tion-based balancing instrument. Reference #3 indicates acceptable residual unbalance for rigid rotors.

Shaft misalignment of directly coupled ma-chines results primarily in elevated vibration at twice the shaft
rotation frequency, sometimes with the peak at shaft rotation frequency elevated as well. If the radial
misalignment (i.e. shaft offset) dominates, then this peak is most pronounced for measurements taken in
radial direction (perpendicular to the shafts). If angular misalignment (coupling gap) is predominant, then
vibration elevation will be most noticeable in frequency spectra of axial measurements. Many
manufacturers and op-erators of electric machines have adopted the use of modern laser-optical shaft
alignment systems such as OPTALIGN® to correct exces-sive shaft misalignment. Recommended align-ment
tolerances are outlined in Note #4.

3 ISO 3945 Mechanical vibration of large rotating machines with speed range from 10 to 200 rev/s;
Measurement and evaluation of vibration severity in situ, 12/1985

4 OPTALIGN® PLUSOperating instructions and alignment handbook,PRÜFTECHNIK AG, Ismaning, Germany,

03/1997
Fig. 7.12 - Rotor unbalance. Shaft misalignment

7. Practical vibration diagnosis. Rotor unbalance

The vibration spectrum exhibits a typical un-balance pattern. The levels of vibration severi-ty
measured at several locations on the ma-chine point indicate that the source of excita-tion lies
near the coupling. Simple rotor bal-ancing of the brake disk reduced motor vibration to 3.5
mm/s and gearbox vibration to 3.1 mm/s.
Fig. 7.13 - Practical vibration diagnosis. Rotor unbalance

8. Shaft misalignment

The vibration spectrum shows a distinct peak at twice the shaft rotation frequency, which clearly indicates
shaft misalignment. Follow-ing shaft alignment, the peak has disap-peared, but the rotor unbalance
evident in the previous spectrum remains to be corrected.
Fig. 7.14 Shaft misalignment

9. Loose belt drive wheel

A press drive motor had developed severe vibration and was producing unusual noises that had become
more pronounced from one day to the next. In stark contrast to the usual vibration spectrum, the rotation
frequency was hardly visible at all, but multiples of the rotation frequency were quite obvious. These
symptoms remained unchanged when the drive belt was removed from the motor. The source was found
to be loose mounting of the belt drive wheel on the motor shaft. The problem was resolved by remachining
the motor shaft and reattaching the belt drive wheel.
Fig. 7.15 - Loose belt drive wheel

10 Bearing evaluation characteristic overall values

As a rule, bearing race damage cannot be detected by elevated levels of low-frequency vibration
parameters until damage is quite severe. The reason for this is that when the rolling elements pass over a
damaged area of the race, a shock pulse is created that can be detected only in the high-frequency
range at first. This is why special bearing characteristic overall values were developed for anti-friction
bearing monitoring; there is no internationally-accepted standard for these so far, and so a variety of
different characteristic overall values can be found in use today.

This illustration lists the most well-known of these bearing parameters. In Germany, for example, the shock
pulse method has established itself over the past 25 years as an easy-to-use and reliable measurement
technique for monitoring anti-friction bearings. In contrast to all other bearing parameters, this method uses
two parameters for evaluation. The shock pulse maximum value dBm, which indicates the severity of shocks
in the rolling behavior of the bearing, is useful in detecting initial damage to bearing races. The ‘carpet
level’ of shock pulses, dBc, indicates the base noise level of the bearing, which increases primarily due to
lubrication problems, general wear of races, insufficient bearing clearance or residual stress due to
improper installation.

One typical characteristic of all anti-friction bearing parameters is the dependency of their levels upon
various influences such as rolling velocity, i.e. bearing size and rpm, signal damping, bearing load and
lubrication. This is why it is practically always necessary to take a comparative measurement in good
condition or to normalize readings relative to good condition.
Fig. 7.16 - Bearing evaluation characteristic overall values

11. Normalization of shock pulse measurement

This illustration shows the normalization procedure that PRÜFTECHNIK instruments use during shock pulse
measurement to compensate the influence of rolling velocity differences. The initial level, and in turn the
adjusted initial value dBia, are determined by taking a comparative measurement in good condition. This
serves as the reference for relative level measurement of maximum shock pulse value dBm and the shock
pulse carpet value dBc. This procedure allows measurements from different bearings to be compared using
the same level scale so that tolerances do not have to be set individually for every single measurement
location.
Fig. 7.17 - Normalization of shock pulse measurement

12. Practical bearing diagnosis- inner race damage

This shows an example of advanced damage to the inner race. The great increase in shock
pulse levels, especially that of peak value dBm from 18 to 48 dBsv, signified serious bearing
damage. Envelope spectrum analysis indicat-ed a pattern typical of inner race damage, which
was then confirmed following bearing replacement: one of the two races of the inner ring
already exhibited a damaged sur-face area of about 15 x 15 mm / 5/8” x 5/8”.
Fig. 7.18 Practical bearing diagnosis- inner race damage
MEASURING SHOCK PULSE

When we look at the rotating component that gives us the most concern, it usually comes down to the
bearing. I think it’s fair to say that 70-80% of rotational problems are bearing related. Whether the causes
are due to under or over lubrication, contamination, installation faults, secondary forces or just plain
fatigue, we need to know the operating condition of bearings most frequently. So it’s very important to
determine the best technique for identifying your particular bearing problems. The other rotational
problems certainly need to be identified as well, so again, choose the most cost effective, efficient
technique to accomplish that.

That is, around 80% of equipment needs to be monitored without the need of spectral data and large
amounts of data collection. You could then utilize spectrum analyzing only on the equipment that needs
it. For those pieces of equipment that are so critical that periodic monitoring is not enough, then
continuous monitoring needs to be considered

The Shock Pulse Method (SPM) is the front line technique the Hallsta paper mill chose to quickly manage
input from its 800 rolls, with 4000 machines and 16,000 measurement points. With 8 inspectors, they need a
quick method to know whether bearings need to be greased or not, or that damage is present and needs
to be monitored more frequently.

What is shock pulse ?

What we loosely call ‘machine vibration’ is a very complex form of movement that has many different
causes and that can be described and measured in many different ways. Vibration exists in all machines
with moving parts, because some of the force, which makes the machine work, is directed against the
machine structure and tries to shift it from its position. Thus, vibration is normal up to a degree, and all
machines are constructed to withstand a certain amount of vibration without malfunctions. In order to use
vibration monitoring to diagnose machine condition, we have to:
i. Find a suitable way of measuring vibration, and
ii. Decide what normal vibration is and what excessive vibration is for any par-ticular machine.

All vibration measurement starts with a time record, a registration of vibration over a length of time. A
transducer converts the movement into an electric signal, which an instrument quantifies, displays and
stores. The signal can then be evaluated in terms of ‘good’ or ‘bad’.

One way of looking at vibrations is to define the type of force, which causes it.

Most industrial machines are rotating, so the main force is rotational, operating on masses which are
imperfectly balanced. This accounts for approximately 99% of the total vibration energy. Rotational forces
are continuous and cyclic – the force does not stop (while the machine is running under power) and the
movement is repeated once per revolution of a part.

About 1% of machine vibration is due to shock. Shock forces are not continuous but can be repeated,
either at regular or irregular intervals.

The remaining small amount of vibration, about 0.1%, is attributed to frictional forces.
Even bearing damage can be detected through vibration analysis. A bearing produces a group of peaks
in the vibration spectrum, caused by the rolling elements passing, at different speeds, over the inner race
and the outer race, and by spinning around their axis. A further peak is caused by cage rotation. Given
the small mass of the bearing in relation to the large mass of the machine, these peaks normally have very
low amplitudes and many times are difficult to pick up with a spectrum before there is severe damage.

A shock pulse transducer contains a reference mass (m) and responds with a dampened oscillation when
hit by a shock wave. Attached to the reference mass is a piezoelectric crystal which produces a voltage
when compressed by the movement of the reference mass. This voltage is proportional to the amplitude
of the oscillation and thus to the energy of the shock wave. The principle is the same as used in
accelerometers for vibration measurement. There is, however, an important difference.

When a mass is excited at its resonance frequency, it will oscillate with much greater amplitude than at any
other frequency. For vibration measurement, one normally stops measuring far below the resonance
frequency of the transducer. On the other hand, shock pulse meters are mechanically and electrically
tuned to operate exclusively at their resonance frequency of 32 kHz (fm), where the resulting signal is
strongest. This gives us a very sensitive transducer for shocks only, but which will not react to ”normal”
machine vibration frequencies.

When a ball hits a damaged area in the raceway, it produces a shock wave. Shock waves are ”transients”
or short-lived waves starting with relatively high amplitude that quickly dampen out. In a time record
displayed by an oscilloscope, these transients are often clearly seen, superimposed on the continuous
wave produced by shaft rotation (see Figure SPM.1). When the distance between transients is constant
and corresponds to the ball pass frequency, this is clear evidence of bearing damage.

Fig. SPM.1 Transients generated by rolling element-fault impact superimposed on wave from shaft rotation.

In the spectrum, however, peak amplitude is determined by the energy contents of the vibration at any
given frequency. In relation to the energy at the shaft frequency, the energy of the shocks produced by
the damaged bearing can be negligible. Thus, the ball pass frequency line has low amplitude and is easily
lost among the ”noise”, as shown in Figure SPM.2.
Fig. BPM.2 - Rolling Element Pass Frequency masked by the spectrum noise.

In the area around the resonance frequency, we can record a time signal, which clearly shows the
transients produced by the damaged bearing. Each shock is a single event, but is also repeated at a
regular rate, the interval being the time between one ball passing the damage and the next. The signal is
treated by rectifying (which cuts off the negative amplitudes) and by enveloping (which produces well-
defined peaks). Figure SPM.3 illustrates this process.

Fig. SPM.3 – Illustration of the effect on signal after band pass, rectifying and enveloping.

The enveloping technique used by vibration analysis attempts, by manipulating the signal, to make shocks
visible and measurable in the frequency domain, simply because frequency analysis is the general
technique used to detect machine faults.

The main strength of the Shock Pulse Method is its specialization on shock detection.

The transducer and measuring instrument are designed to measure the magnitude of shocks directly in the
time domain. All generations of shock pulse meters give readouts of both the magnitude of the peaks
(maximum value dBm) and of the signal level between peaks (carpet value dBc). Together, these two
values can be directly translated into bearing condition information by utilizing the bearing bore diameter
and rpm.

The dBi value is calculated once the RPM and shaft diameter have been input into the Tester. This
becomes the starting point for 0-60 scale.
What is the Shock Pulse Method

Many years ago SPM took the Shock Pulse technology and developed it into the Shock Pulse Method.
Through actual testing in bearing test labs, empirical data was developed by using the bore diameter and
rpm. With this info a dBi value is determined, which positions the normalized condition color alarm scale
onto the dynamic range of the shock pulse transducer. This enables users to utilize a standardized alarm
scale, regardless of the rpm or bearing bore diameter (see Figure SPM.4).

Fig. SPM.4 - Shock Pulse Meter Evaluation: Range/Scale.

The dBm is the maximum value, the measured value of the strongest pulses detected during the measuring
interval. While the bearing surfaces are undamaged, the difference between dBm and dBc (decibel level
Carpet) is small. A high dBm and a large difference between dBm and dBc are caused by surface
damage or foreign particles between rolling element and raceway.

Due to the sensitivity of the Shock Pulse Method, bearing lubrication condition is measurable through the
signal monitored as dBc. The dBc is measured in the time wave signal of the shock pulse transducer. The
filtered transducer signal reflects the pressure variation in the rolling interface of the bearing. When the oil
film in the bearing is thick, the shock pulse level is low, without distinctive peaks (green area, Figure SPM.5).
The level increases when the oil film is reduced, but there are still no distinctive peaks (yellow area, Figure
SPM. 5). Damage causes strong pulses at irregular intervals (red area, Figure SPM.5).
Fig. SPM.5 – Shock Pulse Method Readings and what they mean.

In 2002, SPM expanded the SPM Method by performing an FFT on the same 32 kHz signal utilized, which
resulted in a more indepth analysis capability. By identifying the different bearing frequencies (symptoms)
we can now see the matches of those frequencies within the SPM Spectrum. Likewise typical symptoms
such as imbalance or looseness can also be introduced for more accurate pattern recognition.

The x-axis of the SPM Spectrum is scaled in Hz. The y-axis is in SD (Shock Distribution unit). The amplitude in
the SPM spectrum should be used in conjunction with the SPM values. A new damage can cause high SD
readings and an older more severe damage can have lower SD values. Primarily the SPM Spectrum is used
for pattern recognition. It is known, but not quantified, that the delta (difference between high peaks and
average level) in a spectrum is related to the bearing status.

Figure 6 shows a typical Shock Pulse Bearing Condition chart. The x-axis represents the time frame. The Y-
axis is signal strength intensity divided up as a Green-Yellow-Red condition code. As explained earlier the
Alarm level is determined by the shaft diameter and RPM that is programmed into the instrument and/or
the PC software. That defines the baseline and from there the Green/Yellow/Red divisions are further
defined. On the chart we see the readings in the Green zone until about March 2002. Then they take off
into the Yellow then the Red zones. Plus we see the development of a large delta (dBm-dBc), which also
indicates bearing dam-age in progress.
Fig. SPM.6 – Typical SPM Chart showing the dBm and dBc of a failed roller at a Paper Mill.
References

1.
2. Lebold M., McClintic K., Cambell R., Byington C., Maynard K., Review of vibration analysis
methods for gearbox diagnostics and prognostics, Proceedings of the 54-th Meeting of the
Society for Machine Failure Prevention Technology, Virginia Beach, May 1-4, 2000, pp. 623-634.
3. Lacey S.J. An Overview of Bearing Vibration Analysis, Maintenance & Asset Management, vol
23, No 6, Nov/Dec 2008-
4. Spatenka P., Lindh T., Ahola J., Partanen, Embeded DSP based system for bearing condition
monitoring of electric motors Lappeenranta University of Technology, Research Report 17, 2003
5. Peter W.Tse, Wen-xian Yang, H. Y. Tam, Machine Fault diagnosisthrough an effective excat
wavelet analysis, Journal of Sound and Vibration, 277, (2004), pp. 1005-1024.
6. Gafitanu M., Cretu S., Dragan B., Vibroacoustic diagnosis of machine and equipments, Ed.
Tehnica, Bucharest, 1989