Model Based Fault Identification of Unbalance and

Misalignment Simultaneously Present in a Rotor System

A. K. Jalan and A. R. Mohanty

Department of Mechanical Engineering,
Indian Institute of Technology, Kharagpur 721 302, West Bengal, India
[email protected]

Abstract

Vibration in rotating machinery is mostly caused by unbalance, misalignment, mechanical looseness,

shaft crack, and other malfunctions. The diagnostics of rotor faults have been gaining importance in recent
times. Many researchers have dealt with single fault identification at any given instance, but normally
more than one fault can occur in a rotor system, simultaneously. In the present study a model based
technique is described for the on-line identification of rotor having more than one fault. The rotor has been
modeled using finite element method with unbalance and misalignment together as the considered faults.
Using the residual generation technique, residual vibrations are generated from experimental results
for unbalance and misalignment present simultaneously in a rotor system, and then the residual forces
due to presence of these faults are calculated. These residual forces are compared with the equivalent
virtual forces due to presence of these faults. The fault condition and location of faults are successfully
identified by this model based technique.

Keywords: Fault diagnosis, Condition monitoring, Finite element method, Residual generation

Nomenclature

C Damping matrix
e Eccentricity
F X1, F X2 . . . Reaction forces
K Stiffness matrix
Ka , Kb Axial & bending spring rate per degree per disk pack
M Unbalance mass
M Mass Matrix
MX1, MX2 . . . Reaction moments
q Generalized co-ordinates
Tq Torque

X1,
Y 1 Misalignment in X, Y direction at node 1

X2,
Y 2 Misalignment in X, Y direction at node 2
Z3 Center of articulation

Corresponding author.

ADVANCES IN VIBRATION ENGINEERING, 12(1) 2013 KRISHTEL eMAGING SOLUTIONS PRIVATE LIMITED
24 A. K. JALAN AND A. R. MOHANTY / ADVANCES IN VIBRATION ENGINEERING, 12(1) 2013

1 sin1 (
Y 1/Z3)

2 sin1 (
Y 2/Z3)
Misalignment angle
1 sin1 (
X1/Z3)
2 sin1 (
X2/Z3)

1 Introduction
Rotating machines are important assets in most industries. One of the major areas of interest in the
modern day condition monitoring of rotating machinery is that of vibration. Misalignment and unbalance
are the two major sources of rotating machinery vibration. The vibration due to such sources affects
critical parts of the system such as bearings, gears, motor, seals, couplings etc.
When the shafts of the driving and driven machines are eccentric because of improper machine
assembly, it is known as shaft misalignment. In practical applications perfect alignment of the driving
and driven shafts cannot be achieved. Even if perfect alignment is achieved initially, it would not be
possible to maintain it during operation of the machine due to factors like, thermal distortion of housing
supports of the bearings, differential thermal growth of machine parts, piping forces due to variation in
pressure and temperature, movement of foundation etc. Therefore, a misalignment condition is virtually
always present in machine-trains. Usually flexible couplings are used to accommodate the existing
misalignment between shafts and to transmit rotary power without torsional slip. Model of flexible
coupling as given by Kramer [1] is utilized in this study. Misalignment in rotating machinery causes
reaction forces and moments to be generated in the coupling which leads to vibration in the system. The
effect of these forces and moments on the system has been described by[24] to reveal their influence on
machine components such as bearings, shaft etc.
When the center of mass of a rotating disc is not coincident with the center of rotation, it is known as
unbalance. Like misalignment, unbalance in a rotor system is unavoidable and it can not be completely
eliminated. In practice, balancing is done to balance the rotor system but due to reasons like porosity in
casting, non uniform density of material, manufacturing tolerances, and gain or loss of material during
operation, rotors can never be perfectly balanced. A fault model for unbalance has been developed by
Platz and Markert[5] by considering equivalent loads at particular point due to unbalance.
The generalized equation of motion for a complete rotor-bearing system is derived using the finite
element method given by Nelson and McVaugh[6] . In the present study a model based fault diagnosis
technique is developed for identifying the faults of a rotor system. This model based fault diagnosis
technique is based on the residual generation technique, which is elaborately described by Isermann[7] .
Jalan and Mohanty[8] used this model based technique to identify the unbalance and misalignment
acting separately in a rotor system. In this study, model based residual generation technique is applied
to identify unbalance and misalignment simultaneously present in a rotor system.

2 Model Based Fault Identification Method

For the improvement of reliability, safety and efficiency of technical processes, fault diagnosis is used
as an advanced supervision tool in present industries. The automated early detection and localization of
 MODEL BASED FAULT IDENTIFICATION OF UNBALANCE AND MISALIGNMENT 25

faults in machines are of paramount importance. The general procedure of model based fault diagnosis
can be roughly divided into generation of residuals (i.e., functions that are emphasized by fault vector),
and identification of faults.

2.1 Mathematical description

The vibrations of the healthy rotor system is represented by the vector x0 (t) at N degrees of freedom
for the operating load F (t) and is described by the linear equation of motion

M x0 (t) + C x0 (t) + Kx0 (t) = F (t) (1)

where, M, C and K are mass, damping, and stiffness matrices of any complex rotor system which
includes the effect of bearings, foundations, gyroscopic effects etc.
The presence of a fault in the system changes its dynamic behavior. The extent of the change depends
on the vector , which describes the fault parameters such as type, location, magnitude etc. of the fault,
for e.g., for misalignment: type of misalignment (angular or parallel), amount of misalignment. The
fault-induced change in the vibrational behavior of the system, and is represented by the equation

+ C x(t)
M x(t) + Kx(t) = F (t) +
F (, t) (2)

The residual vibrations induced represent the difference of previously measured normal vibrations x0 (t)
of healthy system from currently measured vibrations x(t) of faulty system. Hence, the residuals of
vibration may be written as

x(t) = x(t) x0 (t); = x(t)

x(t) x0 (t); = x(t)

x(t) x0 (t). (3)

Subtraction of the equations of motion for the healthy system (1) from that of faulty system (2), and
using (3), the equation of motion for residual vibration can be written as

+ C
x(t)
M
x(t) + K
x(t) =
F (, t) (4)

The system matrices remain unchanged and the rotor model remains linear, whereas the equivalent
loads induce the change in the dynamic behavior of the healthy linear rotor model. To identify the fault
parameters, the difference of the theoretical fault model loads and the measured equivalent loads will
be minimized by some statistical algorithm like least square fitting.
To calculate residual vibrations, there is a need of measured vibration data for both healthy and
faulty rotor system at same operating and measurement conditions but directly matching data is usually
not available. Hence some signal processing has to be carried out to achieve the same conditions like,
different rotor speeds are compensated by adjusting the time scale of the recorded normal vibrations to
the time scale of currently measured vibrations. Phase shifts are avoided by recording a trigger signal
during the measurement [5, 9] .

2.2 Scheme of fault diagnosis

To identify the fault in any system a model based scheme is established with the help of residual
generation. The flow chart of this scheme is shown in Fig. 1.
The measured vibration signal data for the healthy rotor system in the form of displacement can be
written as X0 (t). Now in order to find out the fault in the system, measured vibration signal data for
26 A. K. JALAN AND A. R. MOHANTY / ADVANCES IN VIBRATION ENGINEERING, 12(1) 2013

Fig. 1 Flow chart for fault diagnosis

the faulty system are stored in the form of displacement which can be written as X(t). The residual
vibration is calculated from the measured vibration displacement data for both the healthy system and
the faulty system at same operating conditions. The residual displacement is given by
X0 (t).
Next a transition matrix is estimated to make a relationship between theoretical vibration response
of the healthy system and measured vibration response of healthy system. This transition matrix [] is
used to compute the equivalent residual vibration for the theoretical model. Later, equivalent residual

velocities (
x(t))
and accelerations (
x(t)) for theoretical model are obtained by differentiating the
equivalent residual displacement signal.
To calculate the equivalent loads from the mathematical model of the rotor system there is a need of
measured residual vibrations at all degrees of freedom of the model. But practically it is not possible
to install the vibration sensors at all the locations. The number of measurement locations is much
smaller than the number of models degrees of freedom. To resolve this problem the residual vibrations
of the non-measurable degrees of freedom have to be estimated from the measured residual vibrations.
Therefore, the residual vibrations need to be reconstructed via shape expansion.
 MODEL BASED FAULT IDENTIFICATION OF UNBALANCE AND MISALIGNMENT 27

The equivalent forces characterizing the unknown faults are calculated by substituting the residual
displacements, residual velocities and residual accelerations of the full vibrational state into (4).

F (t) = M
x(t)
+ C
x(t)
+ K
x(t) (5)

For identifying the fault parameter, the results from the fault models and measured equivalent loads
from (5) should be compared. If there is no noise and no errors due to shape expansion, the equivalent
loads from the measured data would match exactly that of the fault model. But practically, the measured
signals and their processing are always associated with some noise and inevitable errors. To resolve the
associated error least-squares algorithm is used in the time domain by achieving best curve fitting. The
objective function to be minimized for the measured equivalent forces
F (t) and the modeling forces

F (, t) can be expressed as

 2
 

F (, t)
F (t) dt = Minimum (6)

The algorithm iterates for the values of fault parameter for all suspected faults taken into account. The
vector gives the information about the magnitude of fault for e.g., unbalance: amount of unbalance
and its phase. The type of the faults is to be identified from the signals; location of fault is identified
from the plot of equivalent forces.

3 Modeling of Rotor System

The rotor-coupling-bearing system is considered to be a set of interconnecting components comprising
a rigid disc, shaft segments with distributed mass and elasticity, and bearings.

3.1 Equation of motion

The assembled equation of motion of rotor system may be written as

+ [C]{q}
[M]{q} + [K]{q} = {Q}, (7)

where, the mass matrix [M] includes the translational and rotary mass matrices of shaft elements and
rigid disc and the matrix [C] includes gyroscopic moments and bearing damping. The stiffness matrix
[K] includes stiffness of the shaft elements and bearing stiffness. The excitation matrix {Q} comprises
weight of rigid disc, unbalance excitation force, bearing force, forces due to fault in rotor system and
other external forces.

3.2 Fault models

For the model based fault identification method, each fault has to be represented by a mathematical
model describing the relation between the fault parameters and the equivalent force.

3.2.1 Disc unbalance

Presence of unbalance changes the dynamic behavior of the system. This change is taken into account by
equivalent loads acting on the healthy system model. Equivalent loads on the system generate a dynamic
28 A. K. JALAN AND A. R. MOHANTY / ADVANCES IN VIBRATION ENGINEERING, 12(1) 2013

behavior identical to that of the real faulty system.

The fault model for unbalance is developed
by Platz and Markert[5] . The static unbalance
is present in the system due to shift of mass
center of mounted disc, resulting in an equivalent
force.
The unbalance forces are effective in the lateral
directions at the disc, in x y plane, and are given
by: Fig. 2 Static disc unbalance
Fx = me cos t,
2
Fy = me sin t
2

(8)
where, m is mass of unbalance in the disc located at distance e as shown in Fig. 2. is angular speed
of rotor and t is time.
The excitation matrix {Q} of (7) can be written as

0
0 ..
..
.

.
0
0
me2 cos t
{Q} =
Qd = 2 sin t
(9)
0 me

.. 0
.
..
.
0
0
where, {Qd } is excitation matrix for disc static unbalance.
3.2.2 Coupling misalignment
Misalignment occurs when the components of the rotor system are not coaxial while rotating. Usually
couplings are used to accommodate the existing misalignment between shafts. The flexibility in the
coupling can be modeled by assuming the coupling to be comprised of two shaft elements connected
together by a frictionless joint. A Dynamic analysis of misaligned rotor system is given by Rao and
Sreenivas[11] .
Misalignment in rotating machinery causes reaction forces and moments to be generated in the
coupling. There are two basic types of shaft misalignment, parallel misalignment and angular misalign-
ment, as shown in Fig. 3. Gibbons[2] has given the reaction forces and moments developed due to parallel
misalignment. The reaction forces and moments due to angular misalignment are given by Sekhar and
Prabhu[3] and Prabhakar et al.[4] . The reaction forces and moments which the coupling exerts on the
machines shaft are as follows:
For parallel misalignment:
MX1 = T q sin 1 + Kb 1 , MX2 = T q sin 2 Kb 2 , MY 1 = T q sin 1 Kb 1 ,
MY 2 = T q sin 2 + Kb 2 , F X1 = (MY 1 MY 2)/Z3, F X2 = F X1,
F Y 1 = (MX1 + MX2)/Z3, F Y 2 = F Y 1 (10)
 MODEL BASED FAULT IDENTIFICATION OF UNBALANCE AND MISALIGNMENT 29

Fig. 3 Coupling co-ordinate system (a) Parallel misalignment (b) Angular misalignment, ref. [2]

For angular misalignment:

MX1 = 0.0, MX2 = Kb , MY 1 = 0.0, MY 2 = T q sin ,
F X1 = (MY 1 MY 2)/Z3, F X2 = F X1,
F Y 1 = (MX1 + MX2)/(MX1 + MX2)/Z3, F Y 2 = F Y 1 (11)

4 Experimental Setup
The schematic diagram of experimental rotor-coupling-bearing system using a machinery fault simulator
is shown in Fig. 4. The shaft, 15.875 mm in diameter and 362 mm in length, is supported by two identical
rolling element bearings ER-10 K. A disc is mounted at the mid span of the shaft. The diameter of disc
of mass 0.782 kg is 152.4 mm and thickness 15.875 mm.
The balanced rigid disc is mounted on the shaft by radial screws and is located at the mid span of
the shaft. The rotor shaft is driven by a 0.375 kW three phase AC motor connected through a flexible
coupling. The power supply to the motor can be varied through delta VFD-S AC inverter which controls
the speed of the motor from 0 to 5000 rpm. The rotor rig was instrumented with two numbers of B & K
4321 tri-axial accelerometer along with B & K 2692 NEXUS signal conditioning amplifier. The rotor
rig had a built-in optical sensor for rotational speed measurements. All the signals were analyzed on a
OROS 8 channel OR-25 FFT analyzer.
Time response measurements were taken in vertical and horizontal direction from both the accelero-
meters. Two numbers of accelerometers were placed next to each other at the same location on the
structure in order to measure the angular rotational motions responses[10] . The general arrangement
for measurement of rotational motion responses is illustrated in Fig. 5 which also shows the required
responses X and .
The responses X and can be deduced from the translation responses XA and XB which were measured
from two different sensors placed at a distance l from each other. The responses X and are given as
X = 0.5(XA + XB )
= (XA XB )/ l (12)
30 A. K. JALAN AND A. R. MOHANTY / ADVANCES IN VIBRATION ENGINEERING, 12(1) 2013

Fig. 4 Line diagram of experimental rotor-coupl- Fig. 5 Measurement of rotational res-

ing-bearing system ponse, ref. [10]

These two tri-axial transducers are placed on left side bearing (bearing 1) block to measure the
acceleration amplitude in horizontal (X axis) and vertical (Y axis) directions.

5 Results
The analysis has been carried out using the
finite element method for flexural vibrations.
The rotor-coupling-bearing system is divided
into seventeen nodes as shown in Fig. 6.
At node-11 disc is mounted on the shaft which Fig. 6 FE model of rotor-coupling-bearing system
is supported by two bearings at node-5 and showing node locations
node-17 where coupling ends are represented
by node-1 and node-3. Each node having four degrees of freedom (two translation motions and two
rotational motions) was considered. Therefore, in total 68 degrees of freedom were considered in the
finite element model.

Fig. 7 Frequency response at bearing 1 (a) Along x-direction (b) Along y-direction of a rotor-bearing system
with 4.298 104 kg-m unbalance and 1 mm parallel misalignment for 30 Hz rotor speed
 MODEL BASED FAULT IDENTIFICATION OF UNBALANCE AND MISALIGNMENT 31

Fig. 8 Residual force at different locations along the rotor-bearing system with 4.298 104 kg-m unbalance
and 1mm parallel misalignment in (a) Horizontal; (b) Vertical direction for 30 Hz rotor speed

In order to identify the faults using the

model based scheme disc unbalance and coupling
misalignment were introduced simultaneously on
rotor-coupling-bearing system. Coupling misalign-
ment was introduced as 1 mm parallel misalign-
ment along horizontal direction and unbalance was
introduced as a mass of 6.14 gm was attached on the
disc at a distance of 70 mm from the center of the
disc. The time responses were measured at bearing
1 along horizontal and vertical direction for 30 Hz
rotor speed with faults (unbalance and misalign-
ment) acting simultaneously. The time response
acceleration data were numerically integrated to
get the response as displacement. The measured Fig. 9 3D plot of Residual force at different loca-
frequency responses at bearing 1 along horizontal tions along the rotor-bearing system with
and vertical direction with unbalance and misalign- 4.298104 kg-m unbalance and parallel
ment are shown in Fig. 7. misalignment (
X = 1.0 mm) in vertical
Figure 7 show that the amplitude of 1 direction for 30 Hz rotor speed
running speed component is higher compare to that
of 2 running speed component. Due to presence of misalignment the amplitude of 2 running speed
component also increases but due to the dominant nature of unbalance over misalignment amplitude of
1 running speed component is higher. Therefore these frequency responses are unable to identify the
fault. Now, to identify the faults the time response data were used in model based scheme. The experi-
mental residual forces were calculated and compared with numerical model forces. The numerical model
forces for 1mm parallel misalignment are determined at coupling ends node-1 and node-3. Whereas the
numerical forces for 4.298104 kg-m disc unbalance are determined at node-11. The residual forces at
a particular instance of time, for numerical and experimental studies, in horizontal and vertical direction
with unbalance and misalignment acting simultaneously on the system are shown in Fig. 8.
32 A. K. JALAN AND A. R. MOHANTY / ADVANCES IN VIBRATION ENGINEERING, 12(1) 2013

Figure 8 clearly shows that the residual forces are observed at node-1, node-3 and node-11. Node-1
and node-3 are representing the coupling ends (see Fig. 6) and at node-11 the unbalanced disc is mounted
on the shaft. Now to find the type of fault, residual force in a 3-D diagram (with node position, time and
residual force) is plotted. The 3-D diagram of residual force, with parallel misalignment and unbalance,
at 30 Hz rotor speed is shown in Fig. 9.
The presence of residual forces at node-1 and node-3 clearly indicate the fault as a coupling misalign-
ment whereas the residual force at node-11 indicates the presence of disc unbalance. It is also observed
that numerical fault model forces almost match with experimental residual forces (shown in Fig. 8)
which signify that this method predicts the quantity of faults successfully.

6 Conclusion
The rotor-coupling-bearing system was modeled using finite element method. Then by simple measure-
ment of the radial vibration and application of the proposed model based technique, obtained results were
clearly identifying the simultaneously presence of unbalance and coupling misalignment in the rotor
system. Therefore, this method has been demonstrated the model based fault detection for a simple rotor
system. This method may be useful for large systems like in motors, turbines, gearboxes and the like.
The present approach allows for on-line fault identification effectively.

References
[1] Kramer, E., Dynamics of rotors and foundations, Springer-Verlag, Berlin, 1993.
[2] Gibbons, C. B., Coupling misalignment forces, Proceedings of the Fifth Turbomachinery
Symposium, Gas Turbine Laboratories, Texas, pp. 1111116, 1976.
[3] Sekhar, A. S. and Prabhu, B. S., Effects of coupling misalignment on vibrations of rotating
machinery, J. Sound Vibration, Vol. 185(4), pp. 655671, 1995.
[4] Prabhakar, S., Sekhar, A. S. and Mohanty, A. R., Crack versus coupling misalignment in a transient
rotor system, J. Sound Vibration, Vol. 256(4), pp. 773786, 2002.
[5] Platz, R. and Markert, R., Fault models for on-line identification of malfunctions in rotor system,
Transaction of Fourth International Conference Acoustical and Vibratory Surveillance, France,
pp. 435446, 2001.
[6] Nelson, H. D. and McVaugh, J. M., The Dynamics of rotor-bearing system using finite elements,
American Society of Mechanical Engineers Journal of Engineering for Industry, Vol. 98(2),
pp. 593600, 1976.
[7] Isermann, R., Fault diagnosis of machines via parameter estimation and knowledge processing,
Automatica, Vol. 29(4), pp. 815835, 1993.
[8] Jalan, A. K. and Mohanty, A. R., Model based fault diagnosis of a rotor-bearing system
for misalignment and unbalance under steady-state condition, J. Sound Vibration, Vol. 327,
pp. 604622, 2009.
[9] Merkert, R., Platz, R. and Seidler, M., Model based fault identification in rotor systems by least
square fitting, International Journal of Rotating Machinery, Vol. 7(5), pp. 311321, 2001.
[10] Ewins, D. J., Modal testing: Theory and practice, John Wiley and Sons, New York, 1984.
[11] Rao, J. S. and Sreenivas, R., Dynamic analysis of misaligned rotor system, Adv. Vib. Eng., Journal
of Vibration Institute of India, Vol. 2(1), p. 1, 2003.