Anda di halaman 1dari 15

971

Vibration analysis of rotors for the identication of


shaft misalignment
Part 1: theoretical analysis

P N Saavedra* and D E Ram rez


Universidad de Concepcion, Concepcion, Chile

Abstract: Shaft misalignment in machinery causes preload forces to be generated in couplings, which
are then transmitted to the different machine components, thus reducing their lifetime. Shaft
misalignment is a major cause of vibration in machines. Based on tests on coupling stiffness, a new
coupling nite element stiffness matrix has been deduced. This has been used in the nite element
analysis of a two-rotor system connected by a exible coupling, to calculate the mechanical vibration
resulting from mixed angular and parallel shaft misalignments with residual unbalance. The
calculated vibration spectra were determined for two exible couplings: a three-pin Renold coupling
and a three-jaw Lovejoy coupling. The results and analysis indicate that the vibration generated by
shaft misalignment is caused by the variation in coupling stiffness on rotation, and that the forcing
frequencies generated are harmonics of the shafts speed and directly depend on the frequency of the
variation in coupling stiffness. In addition, the amplitudes of the measured vibratory components
were found to rely directly upon the frequency response function that is related to the coupling and
measurement points.

Keywords: rotor vibration analysis, coupling misalignment, condition monitoring

NOTATION F0 hypothetical radial force applied at point


B of the coupling
A rst node of the coupling element, as a FFT fast Fourier transform
superscript indicating that it is relative to FRFij frequency response function between
node A nodes i and j in the specied degrees of
B second node of the coupling element, as a freedom
superscript indicating that it is relative to Gd gyroscopic matrix of the disc element
node B Ka axial coupling stiffness coefcient
C central point of the coupling Kfx x-direction exural coupling stiffness
Cb damping matrix of the bearing element coefcient
dm hypothetical xed displacement imposed Kfy y-direction exural coupling stiffness
on the coupling due to misalignment coefcient
dc nodal displacements vector for the Ki any of the stiffness coefcients of the
coupling element coupling
di displacement vector of node i of the nite Kt measured transverse stiffness coefcient of
element system the coupling
d0B misalignment vector measured on node B Ktx x-direction transverse coupling stiffness
of the coupling coefcient
Fp hypothetical preload force generated on Kty y-direction transverse coupling stiffness
the coupling as a result of misalignment coefcient
K0 hypothetical calculated radial coupling
stiffness
Kb stiffness matrix of the bearing element
The MS was received on 11 September 2003 and was accepted after Kc stiffness matrix of the coupling element
revision for publication on 20 May 2004.
* Corresponding author: Universidad de Concepcion, Casilla 160-C, Kcl lower half of the coupling stiffness
Concepcion, Correo 3, Chile. E-mail: psaavedr@udec.cl matrix
C18203 # IMechE 2004 Downloaded from pic.sagepub.com at University ofProc. InstnonMech.
Otago Library Engrs
December Vol.
22, 2014 218 Part C: J. Mechanical Engineering Science
972 P N SAAVEDRA AND D E RAMIREZ

KcL coupling stiffness matrix with the upper dy parallel misalignment in the y direction
half elements equal to zero measured with respect to the centre of the
Kcu upper half of the coupling stiffness matrix coupling
KcU coupling stiffness matrix with the lower dyB parallel misalignment in the y direction as
half elements equal to zero measured at node B of the coupling
Ks stiffness matrix of the shaft element dz axial misalignment in the z direction
L length of the coupling measured between Dy0 hypothetical radial displacement at point
nodes A and B B of the coupling due to a radial force F0
MX bending moment on the coupling about yix rotation of node i about the x axis
the x axis yiy rotation of node i about the y axis
MY bending moment on the coupling about j hypothetical misalignment angle
the y axis measured at the centre of the coupling
Mc mass matrix of the coupling element jx angular misalignment about the x axis
Mcl lower half of the coupling mass matrix jy angular misalignment about the y axis
Mcu upper half of the coupling mass matrix O rotational speed of the rotor system
Md mass matrix of the disc element
Ms mass matrix of the shaft element
n natural number 1, 2, 3, . . .
N number of identical geometric elements 1 INTRODUCTION
that shape a exible coupling
PX transverse force on the coupling in the x In all industries around the world, maintenance
direction strategies are changing from periodic inspection strate-
PY transverse force on the coupling in the y gies to condition-based predictive maintenance strate-
direction gies. Condition monitoring can improve the availability
PZ transverse force on the coupling in the z of machinery by being able to pinpoint and diagnose
direction problems at an early stage while the machinery is
Pb forcing vector on the bearing element working. This process avoids off-line inspections with
Pc forcing vector for the coupling element consequent loss of production and reduces unplanned
P cmisal loads generated by the misalignment on down time. Vibration analysis is certainly one of the
the coupling most powerful methods in current use in machinery
Pd forcing vector on the disc element diagnosis.
Ps forcing vector on the shaft element However, there is a lack of research on the ultimate
R point on the circumference of the coupling objective of diagnostic techniques, which is the determi-
that acts as a reference for the tachometer nation of reliable vibratory indicators to differentiate
and in the measurement of a between problems that exhibit similar frequency spectra
T transmitted torque on the coupling and waveform vibrations. Examples are shaft misalign-
ui displacement of node i on the x axis ment and cracked shafts. The importance of discrimi-
vi displacement of node i on the y axis nating between these problems can be seen in that, in the
wi displacement of node i on the z axis rst case, the machinery must be stopped as soon as
X1 Y1 Z1 reference system 1, where the z axis is possible to avoid a catastrophic failure and, in the
coincident with the rotational axis of second case, the machinery can usually wait for the next
rotor 1 scheduled maintenance period. Thus, to develop a
X2 Y2 Z2 reference system 2, where the z axis is reliable diagnosis, a thorough understanding of the
coincident with the rotational axis of physics of the problem and its effect on machine
rotor 2 vibration is necessary. The work presented in this paper
describes theoretical and experimental work on the
a rotation angle of the coupling measured vibration behaviour of coupled rotors with exible
between R and the horizontal plane ZX couplings.
aN parameter of Newmarks numerical Shaft misalignment is one of the most common causes
integration of vibration in rotating machinery. Shaft misalignment
dN parameter of Newmarks numerical is dened as the deviation of the shaft position relative
integration to a collinear axis of rotation measured at the points of
dx parallel misalignment in the x direction power transmission between two machines. There are
measured with respect to the centre of the two basic types of shaft misalignment: parallel and
coupling angular [1]. A combination of parallel and angular
dxB parallel misalignment in the x direction as misalignment in the vertical and horizontal directions is
measured at node B of the coupling the most common found in industry. Shaft misalignment
Proc. Instn Mech. Engrs Vol. 218 Part C: J. Mechanical Engineering
Downloaded Science
from pic.sagepub.com at University of Otago Library on December 22, 2014 C18203 # IMechE 2004
VIBRATION ANALYSIS OF ROTORS FOR THE IDENTIFICATION OF SHAFT MISALIGNMENT. PART 1 973

generates additional loads on bearings, seals, shafts and that the 1 6 component did not change as the degree
couplings, so decreasing their operating lives, and, when of misalignment increased.
it is excessive, shaft misalignment can induce early Xu and Marangoni presented a theoretical model of a
failure, such as oil leakage at the bearing seals, loosening motor with a exible coupling rotor for describing the
foundation and coupling bolts, and shaft breakages mechanical vibration resulting from angular misalign-
occurring at or close to the inboard bearing [1]. Because ment [4, 5]. To take the effect of misalignment into
of the above, ideally, it is desired to maintain perfect account, they assumed that the exible coupling
alignment of the driving and driven shafts. However, behaved exactly as a universal joint, and they concluded
this is not possible in practical applications because, that misalignment can be characterized by a
even if perfect alignment is achieved initially, it cannot 2 6 component. However, the effects of misalignment
be maintained over an extended period of time as shafts were not always apparent, since the forcing frequency
can change their position when rotating equipment (assumed to be twice the shafts speed) of the misalign-
starts. The most common causes of this movement are ment was not close enough to one of the systems
temperature changes arising in the machinery casing due natural frequencies to excite the system appreciably. In
to friction in the bearings, or from thermal changes that this case, the misalignment can be hidden and so does
occur in the process liquids and gases, or from reaction not show up in the vibration spectrum.
forces and couples in attached piping. Sekhar and Prabhu [6] proposed a theoretical model
Flexible couplings must be used to compensate for the of a rotor-bearing system using higher-order nite
negative effects generated by unavoidable misalignment. elements. They assumed that the imbalance force had
Flexible couplings join two anges of the rotating a 1 6 and 2 6 components and, using this assumption,
equipment, while permitting some degree of misalign- they observed an increase in these frequencies with
ment or end movement (or both). There are three basic increasing misalignment. The effect of parallel misalign-
types of exible coupling: mechanical, elastomeric and ment on the lateral and torsional responses of two
metallic. The mechanical element type generally obtains Jeffcott rotors connected by a rigid mechanical coupling
exibility from loose-tting parts, or from the rolling or was analysed by Al-Hussain and Redmond [2]. They
sliding of mating surfaces, e.g. a gear coupling. The concluded that, in the steady state regime, only the
elastomeric element type obtains exibility from stretch- 1 6 component was present in the vibratory response.
ing or compressing a resilient material, e.g. a jaw This study did not provide any evidence for the presence
coupling. The metallic element type obtains exibility of second- or higher-order harmonic responses, which is
from the exing of thin metallic discs or diaphragms, a characteristic commonly observed in misaligned
e.g. diaphragm couplings. rotating shaft systems.
Although the use of exible couplings helps to Simon [7] and Bachschmid et al. [8] modelled the
overcome the negative effects of misalignment, they effect of misalignment by introducing an equivalent
are not a cure for severe alignment problems. It is still system of external forces and couples. When rigid
absolutely necessary to count on precise diagnostic tools coupling halves were forced together by bolting, the
for early detection of this problem, and for evaluation if rotation of the shaft produced a rotating pair of forces
corrective actions are needed. So far, only a few arising from the radial offset or parallel misalignment,
researchers have paid attention to shaft misalignment, and a rotational pair of bending moments arising from
due to the complexity in modelling the phenomenon. the angular misalignment. The rotational frequencies of
According to Al-Hussain and Redmond [2], there is these pairs were equal to the shafts speed.
much discussion in industry regarding the interpretation In several practical or eld studies, shaft misalignment
of vibration signals caused by misalignment in has been reported to show up in the frequency domain
machines, but there is not enough academic research of the rst harmonic of the shafts rotational speed (1 6,
that explains this phenomenon in a simple way. 2 6, 3 6, . . . ). However, to the present authors
Dewell and Mitchell [3] determined that the expected knowledge, no theoretical model or formulations are
vibration frequencies for a misaligned metallic-disc available that describe the mechanical vibration indi-
exible coupling containing four bolts were multiples cated above.
of four times the shafts speed (called the This work is focused on establishing a clear correla-
4 6 component), with side bands at the shafts speed tion between misalignment and the vibratory symptoms
(called the 1 6 component), when the forcing frequency that it produces to aid machinery diagnostic engineers to
was equal to the shafts speed (due to unbalance), and distinguish between shaft misalignment and other
that the expected vibration frequencies were multiples of problems that exhibit similar vibration symptoms (e.g.
twice the shafts speed (called the 2 6 component) when cracked shafts, mechanical looseness and case distor-
the forcing frequency was twice the shafts speed (due to tion) using vibration analysis. To accomplish this
an oval shaft, or from Hookes joint effect). Further, objective, this paper is divided into two parts. In Part
they concluded from experiments that the 2 6 and 1, a new theoretical model capable of describing the
4 6 components exhibited the largest changes, and mechanical vibration resulting from shaft misalignment
C18203 # IMechE 2004 Downloaded from pic.sagepub.com at University ofProc. InstnonMech.
Otago Library Engrs
December Vol.
22, 2014 218 Part C: J. Mechanical Engineering Science
974 P N SAAVEDRA AND D E RAMIREZ

is developed. To consider this effect, the formulation of have to repeat at the same values. Therefore, the basic
a nite element model for a exible coupling is frequency of vibration and the computed stiffness K0 a
developed. The model is used to study the steady produced by a misaligned jaw coupling should be three
vibration behaviour of two coupled test rotors with times the shafts speed. This is in agreement with a
misaligned shafts. In Part 2, experimental tests are similar analysis performed by Dewell and Mitchell [3]
performed to conrm data from the model, and to for the moments produced by a exing disc pack on a
obtain useful vibration analysis rules for misalignment misaligned four-bolt metallic-disc coupling. He stated
diagnosis, which can serve as an aid to the practical that the moment had a periodicity of four times per
engineer. It will be seen that good agreement between turn. Because the stiffness coefcient has a complex
theory and experiment conrms the theoretical model periodic waveform, it is useful to translate K0 a into the
developed. frequency domain by means of the classical Fourier
transform technique. Figure 1b shows a typical order
spectrum obtained. The spectrum in Fig. 1b contains a
2 QUALITATIVE ANALYSIS OF THE PROBLEM 36 main component (order, 3, or three times per turn)
and some secondary components at orders of 3n (where
The following qualitative analysis, using a simplied n 2, 3, 4, . . .).
approach, aims for an easier understanding of the basic To quantify the previous qualitative analysis, several
hypothesis on which this work is based. This is that commercial couplings were experimentally tested in a
vibration generated by misalignment is caused by similar way to the hypothetical case shown in Fig. 1b,
periodic variations in the coupling stiffness while the within a 1 mm maximum radial offset. The results for a
shaft is rotating; i.e. the origin of the vibration is of a three-jaw Lovejoy coupling (sketched in Fig. 2a) are
parametric nature. On this point, only the dependence of shown in Figs 2b and c. As expected, the stiffness
the coupling stiffness on the rotation angle will be coefcient spectrum shows one major 36 component,
considered. Further, its inuence on the dynamic and some secondary components at the harmonics of
behaviour of the coupled rotors is discussed. 36, but it also shows minor components at all orders
arising from coupling asymmetry generated by imper-
fections in manufacturing or assembly. For these test
2.1 Coupling stiffness analysis couplings, the imperfections are small, and the symme-
try (three times per revolution) has a strong inuence on
Consider an ideal jaw exible coupling composed of two the waveform and value of the spectral components of
three-jaws hubs and one six-legged intermediate elasto- the measured static stiffness coefcient Kt a. However,
meric element (a spider). As manufacturers do not other couplings that were tested showed that these
publish information on coupling stiffness and, as such, ndings cannot be generalized a priori for any given
information is vital for the vibration analysis of exible coupling. Figures 2e and f show the waveform
misaligned rotors, a given coupling must be evaluated and frequency spectrum results of tests performed on a
on a test stand. The static transverse stiffness can be commercial three-pin Renold coupling (sketched in
measured as shown schematically in Fig. 1b. The Fig. 2d). The spectrum shows a 36 main component
angular position of the coupling is dened by the angle but also that the 16 component has a similar
a, which is measured from a specic point R on the amplitude. In this case, the inuence of geometric
circumference of the coupling; this point R acts as the symmetry is not as dominant as in the earlier case. It
reference point for the tachometer signal, as shown in is important for the eld engineer to know that, in
Fig. 1a. misaligned shafts connected by a exible coupling made
To measure the stiffness, one hub of the coupling is up of N elements, the frequencies of the main spectral
xed A, and then a constant radial force F0 is applied components of the variation in coupling stiffness and, as
to the other hub (B), as shown in Fig. 1b. As a a consequence, in the vibration generated in the rotor,
consequence, a certain radial displacement is obtained are not always pure harmonics of the fundamental
on B, denoted by Dy0 . Then, a stiffness coefcient frequency (N6 components), as has been previously
K0 F0 =Dy0 , associated with a 0 , can be computed. indicated by other workers.
As a varies, the relative position of the applied force
with respect to the coupling structure distribution
changes, and so Dy0 is not constant but depends on a, 2.2 Inuence on the dynamic response of rotors
as shown in Fig. 2 for two commercial couplings. By
studying Fig. 1b, it can be seen that, if the coupling When shaft misalignment occurs, a xed amount of
construction and assembly are perfectly symmetric, then relative deformation between the coupling faces is
the same geometrical distribution will be repeated for imposed. This permanent deformation will introduce
each successive third turn of the shaft; i.e. each 1208 preload forces on both rotors, and subsequently on the
a 0 or 1208 or 2408) and, in addition, Dy0 would bearings and other elements of the rotors. If a exible
Proc. Instn Mech. Engrs Vol. 218 Part C: J. Mechanical Engineering
Downloaded Science
from pic.sagepub.com at University of Otago Library on December 22, 2014 C18203 # IMechE 2004
VIBRATION ANALYSIS OF ROTORS FOR THE IDENTIFICATION OF SHAFT MISALIGNMENT. PART 1 975

Fig. 1 Measurement of coupling stiffness

coupling is used and the coupling inertia effects and component with the same frequency content of K0 a,
displacements other than the magnitude of misalign- which will dynamically excite the rotors, causing
ment are momentarily rejected, the preload induced by vibration. Figure 3 shows a graphical overview of this
misalignment can be considered as an actionreaction scenario, showing both hypothetical cases of constant
force system acting on both ends of the rotors in the coupling stiffness and periodic coupling stiffness.
opposite direction with an equivalent force given by The universal joint effect indicated as a major
Fp K0 dm , where dm is the xed displacement imposed vibration source resulting from shaft misalignment by
by a misalignment. This implies that, if the coupling other workers [35] is not considered in this work for the
stiffness is constant, then the only effect of misalignment following reasons:
on the rotors should be this static preload condition
which would cause no dynamic effects, as shown in 1. There is no justication to generally attribute this
Fig. 3a. However, as stated earlier in the tested coupling effect, inherent to universal joint kinematics, to other
tests, the coupling stiffness is not constant but depends types of coupling.
on the angular position a, with a Ot, where O is the 2. The universal joint effect only considers angular
rotational speed. This involves that the preload force Fp misalignment, and not parallel or combined mis-
also depends on a, and that it has an alternating periodic alignment.

C18203 # IMechE 2004 Downloaded from pic.sagepub.com at University ofProc. InstnonMech.


Otago Library Engrs
December Vol.
22, 2014 218 Part C: J. Mechanical Engineering Science
976 P N SAAVEDRA AND D E RAMIREZ

3. If the rotor shaft and the motor shaft are angularly 3 CONFIGURATION OF A ROTORBEARING
misaligned by an angle j, then a bending moment SYSTEM
T sinj that acts on both the driving and driven
shaft is generated, where T denotes the transmitted To investigate the effect of misalignment on the
torque. Even if the effect of a universal joint was behaviour of rotor vibration, a simple but general
applicable to shaft misalignment, then this effect dynamic model system, as illustrated in Fig. 4, will be
would be negligible compared with the forces analysed. This consists of two rotors connected by a
generated by the preload caused by misalignment exible coupling. Each bearingrotor system consists of
and uctuations in stiffness. a set of interconnecting components consisting of rigid

Fig. 2 (continued over)

Proc. Instn Mech. Engrs Vol. 218 Part C: J. Mechanical Engineering


Downloaded Science
from pic.sagepub.com at University of Otago Library on December 22, 2014 C18203 # IMechE 2004
VIBRATION ANALYSIS OF ROTORS FOR THE IDENTIFICATION OF SHAFT MISALIGNMENT. PART 1 977

Fig. 2 Results of experimental tests on two commercial couplings

discs, shaft segments with distributed mass and elasti- 3.1 Equations of motion of the elements
city, and two antifriction-type bearings. Thus, any
arbitrary rotor bearing system of this type can be Extending the work of Nelson and McVaugh [9] to
represented using a combination of the four types of consider nite elements of ve displacements per node,
nite element: shaft elements, disc elements, bearing and disregarding the inuence of the transverse mass
elements and coupling elements. moment of inertia and the gyroscopic effect on the shaft
The components are discretized by the nite element element, the following equations of motion for the shaft,
scheme, as shown in Fig. 4. The numbers indicated are disc and bearing elements are obtained. The equation of
the numbers of the nodes utilized. Five degrees of motion for the coupling element is developed in the next
freedom are considered for each node. In this case, section. The mass matrix for the shaft element, Ms , the
the displacement vectors are denoted by d i mass matrix for the disc element, Md , the stiffness
f ui vi wi yix yiy gT , where ui , vi and wi are the matrix for the shaft element, Ks , the stiffness matrix for
node displacements in the x-, y- and z-axis directions the bearing element, Kb , the damping matrix for the
respectively, and yix and yiy are the rotations about the x bearing element, Cb , and the gyroscopic matrix for
and y axes respectively. the disc element, Gd , have been detailed in reference [9].
The forcing terms Ps , Pd and P b include the external and

Fig. 3 Inuence of misalignment and coupling stiffness on the dynamic response

C18203 # IMechE 2004 Downloaded from pic.sagepub.com at University ofProc. InstnonMech.


Otago Library Engrs
December Vol.
22, 2014 218 Part C: J. Mechanical Engineering Science
978 P N SAAVEDRA AND D E RAMIREZ

Fig. 4 Schematic drawing of two coupled rotors and location of the nodes

the interconnection forces. No coupling between the similar way to the calculations made by Gibbons [10].
axial, bending and torsion modes is assumed in deriving From the work of Schwerdlin [11], and from the tests
these matrices. In addition, torsional movement is not made by the present authors, it will be inferred that, in
considered in this analysis. The bearings are considered general, the stiffness coefcients Ki for a specic exible
to have linear behaviour and are modelled in a similar coupling type depend upon the rotation angle a, the
way to that shown in reference [1]. This approximation transmitted torque T, the rotation speed O, the
prevents any unnecessary complication of the analysis misalignment displacement and the misalignment angle:
and is a justied approximation for small amplitudes of
the journal motion. For the shaft element, Ki Ki a, T, O, dx , dy , dz , jx , jy 4

Ms d Ks d s P s In this work, it is considered that the stiffness


s
1
coefcients only depend on the rotation angle for the
For the disc element, following reasons:
1. All the theoretical simulations and experimental tests
Md d  OGd d_ P d
d d
2
are performed under constant transmitted torque and
For the bearing element, rotational speed.
2. The experimental results published in reference [12]
Cb d_ Kb d b Pb
b
3 indicate that the changes in the coefcient of stiffness
are insignicant over the variation range of the
misalignment displacement magnitude when the
rotor is running; i.e. the stiffness coefcients will
only depend on a and Ki Ki a within the dynamic
3.2 Coupling element range of movement of the coupling.
A coupling stiffness matrix is constructed to include the
coupling element in the nite element analysis of the
dynamic behaviour of misaligned shafts. For this
purpose, we consider the coupling element as a couple
3.2.1 Stiffness matrix of coupling element
of rigid blocks, joined together by an elastic element, as
shown in Fig. 5a. The elastic element is characterized by A stiffness matrix for the coupling, shown in Fig. 6,
ve springs: three linear springs along the z axis (axial), relating the displacement vector d c to the corresponding
x axis (transverse) and y axis (transverse) respectively, force vector P c is constructed as follows. The forces and
and two bending springs in the yz plane (x-axis rotation) displacements on the coupling end nodes are related
and in the zx plane (y-axis rotation) respectively. Each through the relationship
of these ve springs has its own associated stiffness
coefcient, namely Ka , Ktx , Kty , Kfx and Kfy . As manu- L A L
PBX  Ktx uA  Ktx yy Ktx uB  Ktx yBy
facturers do not publish these stiffness coefcients, they 2 2
L A L B
must be determined by experiment. PY  Kty v Kty yx Kty v Kty yx
B A B
Figure 5b shows two coupling hubs that are 2 2
misaligned in both the parallel and the angular planes. PBZ  Ka wA Ka wB
The magnitudes of the misalignment displacements,  
L A L2 L B
dx , dy and dz , and the magnitudes of the misalignment MX  Kty v Kty
B
 Kfx yA x Kty v
2 4 2
angles, jx and jy , can be computed using the values and  
directions of the displacements obtained from a L2
Kty Kfx yBx
graphical plot of the reverse indicator readings in a 4

Proc. Instn Mech. Engrs Vol. 218 Part C: J. Mechanical Engineering


Downloaded Science
from pic.sagepub.com at University of Otago Library on December 22, 2014 C18203 # IMechE 2004
VIBRATION ANALYSIS OF ROTORS FOR THE IDENTIFICATION OF SHAFT MISALIGNMENT. PART 1 979

Fig. 5 Characterization of a coupling

  1. Only small displacements are considered.


L L2 L B
MYB Ktx uA Ktx  Kfy yA
y  Ktx u 2. The x and y axes are considered to be principal
2 4 2
  stiffness axes to the bending movement, because their
L2
Ktx Kfy yBy 5 cross-stiffness coefcients are insignicant in relation
4 to their direct coefcients. Then, the relative bending
angles between the coupling anges with respect to
by considering the following assumptions: the x and y axes can be expressed by MX =Kfx and

Fig. 6 A coupling element

C18203 # IMechE 2004 Downloaded from pic.sagepub.com at University ofProc. InstnonMech.


Otago Library Engrs
December Vol.
22, 2014 218 Part C: J. Mechanical Engineering Science
980 P N SAAVEDRA AND D E RAMIREZ

MY =Kfy respectively, where M is the bending appropriate node displacement vectors must refer to the
moment at the coupling and Kf is the direct bending correct reference system that the nodes belong to. The
stiffness coefcient. nodes of the shaft, disc and bearing elements belong to
3. The coupling between the transverse, axial and the same reference system, but the coupling element has
bending displacements is considered to be insignif- one node belonging to one reference system, and
icant, then the relative shear displacement between another node belonging to the second system. Because
the coupling anges in the x and y directions can be of this, the motion equation of the coupling element
expressed by PX =Ktx and PY =Kty respectively, and [equation (8)] dened for one reference system, as shown
the axial displacement by PZ =Ka , where Ktx and Kty in Fig. 6, must be changed by coordinate transform in
are the direct shear stiffness coefcients and Ka is the such a way that the displacements for nodes A and B be
direct axial stiffness coefcient. referred to coordinate system X1 Y1 Z1 and X2 Y2 Z2
respectively.
The static equilibrium equations are
Equation (8) can be split into two subsystems:
X PX 0
PA B
Mcu d1 Kcu ad c1 P A
c
1 9
Y PY 0
PA B
c c
Ml d 2 Kl ad 2 P B2
c c
10
Z PZ 0
PA B

MXA MXB  PBY L 0 The subscripts 1 and 2 refer to the X1 Y1 Z1 and X2 Y2 Z2


coordinate systems respectively, so that fd1c g is a 1061
MYA MYB PBX L 0 6 vector displacement of the coupling with reference to the
Combining the loaddisplacement relationships and the X1 Y1 Z1 system, and fd c2 g is a 1061 vector displacement
equilibrium forces, the following static equilibrium of the coupling with reference to the X2 Y2 Z2 system.
equation for the coupling element is obtained: The subscripts u and l refer to the upper and the lower
halves respectively of the stiffness and mass 5610
Kc ad c P c 7 matrices. The vectors d c1 and d c2 contain the displace-
ments of both nodes in reference to the same system
where the coupling stiffness matrix Kc is
2 3
Ktx
6 0 Kty 7
6 7
6 0 0 Ka 7
6 7
6 0 7
2
 Kty L2 0 Kty L4 Kfx
6 7
6 Ktx L 0 0 0
2
Ktx L4 Kfy 7
6 2 7
Kc 6  Ktx 0 0 0  Ktx L2 Ktx 7
6 7
6 0  K 0 K L
0 0 Kty 7
6 ty ty 2 7
6 0 0  Ka 0 0 0 0 Ka 7
6 7
6 0  Kty L2 0
2
Kty L4  Kfx 0 0 Kty L2 0
2
Kty L4 Kfx 7
4   5
2 2
Ktx L2 0 0 0 Ktx L4  Kfy  Ktx L2 0 0 0 Ktx L4 Kfy

The equation of motion for the coupling element using and, as was indicated previously, the displacements for
the previously dened stiffness matrix is node A must be referred to the X1 Y1 Z1 system, and the
displacements for node B must be referred to the
Mc d Kc ad c Pc
c
8 X2 Y2 Z2 system. For this purpose, the vectors d c1 and
d c2 are split in half; i.e. d c1 fd A
1 , d 1 g and d 2 fd 2 , d 2 g.
B c A B
where the mass matrix is conformed by two discrete B A
Then, d 1 and d 2 can be referred to the X2 Y2 Z2 and
masses at the coupling nodes with values equal to hubs
X1 Y1 Z1 systems respectively, using a simple coordinate
mass.
transformation.
For small misalignment angles, it can be seen from
Fig. 5b that
3.3 Equations of motion of the system
1 d 2 d 0B ,
dA A
d B1 d B2 d 0B
Figure 7 shows a schematic representation of two
d_1 d_2 ,
A A
d_1 d_2
B B

misaligned rotors with Z1 and Z2 denoting shaft
d1 d2 , d B1 d B2
A A
centre-lines connected by a exible coupling. For
practical reasons, one reference system for each rotor 11
is dened as X1 Y1 Z1 and X2 Y2 Z2 respectively. For the  T
assembling procedure of all the system elements, the where d 0B dxB dyB dz jx jy is a constant
Proc. Instn Mech. Engrs Vol. 218 Part C: J. Mechanical Engineering
Downloaded Science
from pic.sagepub.com at University of Otago Library on December 22, 2014 C18203 # IMechE 2004
VIBRATION ANALYSIS OF ROTORS FOR THE IDENTIFICATION OF SHAFT MISALIGNMENT. PART 1 981

Fig. 7 Reference system for each rotor

vector. Then, Assembling the equations of motion for the shaft,


  disc, bearing and coupling elements, a nite element
dA model suitable for simulating the dynamic response of
Kcu ad1c Kcu a 1
d B2 d 0B misaligned rotors is obtained. The model allows us to
  introduce all types of shaft misalignment and all types of
0
Ku ad Ku a
c c c
12 exible coupling. The equation of motion
d 0B
Md Cd_ Kd P 16
and
  obtained for the rotor system is solved by a numerical
Kcl ad c2 Kcl a 1  d 0B
dA integration method.
d B2 By analysing equation (14), it can be inferred that two
 
d 0B physical sources for the generated vibration in shaft
Kl ad Kl a
c c c
13 misalignment are present:
0
(a) a periodic variation in the coupling stiffness matrix
Introducing equations (11), (12), and (13) into equation and, therefore, in the global stiffness matrix of the
(8) yields the following equation, which can be system with frequencies equal to the harmonics of
assembled into the whole nite element model: the rotational speed, depending on the type of
exible coupling used, as shown in Fig. 2;
Mc d Kc ad c P c Pcmisal
c
14 (b) periodic dynamic loads generated by misalignment,
Pcmisal , which are proportional to the variation in
where stiffness, as is indicated by equation (15), with
    therefore, frequencies equal to the harmonics of the
d 0B 0
Pcmisal KcL a  KcU a 15 rotational speed.
0 d 0B
According to the experimental results shown in Fig. 2,
and KcU and KcL are the upper and lower halves it can be seen that the amplitude of the variation in
respectively of the stiffness matrix. stiffness is about two or three times lower than the mean
From equation (15), it can be seen that loads value. This implies that the second vibration source
generated by misalignment are directly proportional to dened above dominates the dynamic behaviour of
the magnitude of misalignment and the coupling misaligned shafts.
stiffness. An important practical conclusion can be As a consequence of the previous, to elucidate the full
drawn from this; if the magnitude of misalignment effects of a qualitative analysis of the problem, it is
increases, then the vibration amplitude will increase if possible to assume that the variation in the stiffness
the coupling stiffness increases or stays constant with matrix K is negligible and, therefore, it is possible to
increasing magnitude of misalignment. More misalign- dene frequency response functions (FRFs) that are
ment means more vibration, more bearing overload and independent of the angular position of the coupling, and
shorter bearing life; i.e. the measurement of vibration is equal to the values determined obtained when the rotors
a good indicator for detecting misalignment in machin- are not running. The FRFs calculated this way are used
ery that is running, and for evaluating the severity of as a reference for how the misalignment loads will be
misalignment. This statement is not always true if the translated into vibration on the bearings. Figure 8 shows
stiffness decreases with increasing magnitude of mis- a hypothetical case, where the vibration is measured at
alignment. two different rotation speeds. This assumes a special
C18203 # IMechE 2004 Downloaded from pic.sagepub.com at University ofProc. InstnonMech.
Otago Library Engrs
December Vol.
22, 2014 218 Part C: J. Mechanical Engineering Science
982 P N SAAVEDRA AND D E RAMIREZ

Fig. 8 The inuence of rotation speed and FRFs on measured vibrations

importance when the frequency of a spectral component of the vibratory spectrum are amplied, while the lower
of the excitation force generated by misalignment is orders are attenuated.
close to a natural frequency of the system, because a The similarities between the Kt spectrum and steady
resonance condition will occur. In this case, the state response spectrum are repeated in most of the
amplitude of this component can increase greatly if the simulation results. The inuence of the frequency
damping is low, and can become a major spectral response functions is also observed, particularly when
component, as shown in the experience discussed in Part a spectral component is close to a natural frequency of
II. the system, where it can be amplied and become one of
In any case, periodic elements on the global stiffness the main components of the spectrum.
matrix are fully considered in the following calculation The results from the numerical simulations are in
of the steady state response of the rotor system. agreement with previous qualitative analyses, showing
that vibration generated in misaligned shafts is caused
by the variation in coupling stiffness with rotation angle
3.4 Calculation of the steady state vibratory response a and that the frequencies of the steady vibratory
response are the same as the stiffness coefcients,
The equation of motion shown in equation (16) was showing similar spectral patterns. The inuence of the
solved numerically using Newmarks integration FRFs is also illustrated in Fig. 9. More simulation
method with aN 0:3025 and dN 0:6, to obtain the results are shown in Part 2 of this paper.
steady state vibratory response of a system using the jaw The nite element model allows for the inclusion of
and the three-pin couplings for various operation speeds unbalance forces on the system nodes. In this case, the
and different combined misalignments. Figure 9a shows effect of unbalancing on the steady state vibratory
the spectrum of the steady state response at node 6 in response of misaligned rotors can be studied. Analysis
the x direction for the three-pin coupling, and the of the effect of the angular position for residual
misalignment in the horizontal xz plane with dx unbalance in the vibratory spectrum was performed
 0:35 mm and jy 0:105 . It can be seen that the main using the same degree of misalignment as in the
component of the vibratory spectrum is of third order, previous analysis. Figure 10 shows the simulated
which is the same as in the Kt spectrum (Fig. 9b). Figure vibration spectra at node 3 in the x direction for a
9c shows the magnitude of FRF56 (mechanical mobi- balanced rotor and for an unbalanced rotor with
lity), where it can be seen a greater mechanical mobility unbalances positioned at three different angles. The
at higher frequencies, which explains that higher orders phase angles that dene the angular position of the

Proc. Instn Mech. Engrs Vol. 218 Part C: J. Mechanical Engineering


Downloaded Science
from pic.sagepub.com at University of Otago Library on December 22, 2014 C18203 # IMechE 2004
VIBRATION ANALYSIS OF ROTORS FOR THE IDENTIFICATION OF SHAFT MISALIGNMENT. PART 1 983

Fig. 9 Vibratory response, stiffness spectrum and FRF for the three-pin coupling

unbalance and the phase of the rst-order vibratory magnitude and phase, and they must be added as
components were calculated with respect to a hypo- vectors. Because of this, the 16 component due to
thetic tachometric pulse. The phase of the vibratory unbalance may increase, may decrease or may not
component at 16 is indicated in each spectrum. change at all the magnitude of the 16 misalignment
The 16 spectral vibratory components generated by component. Therefore, the unbalance can be determi-
simultaneous misalignment and unbalance have a nant on the value of the 16 spectral component. In the

C18203 # IMechE 2004 Downloaded from pic.sagepub.com at University ofProc. InstnonMech.


Otago Library Engrs
December Vol.
22, 2014 218 Part C: J. Mechanical Engineering Science
984 P N SAAVEDRA AND D E RAMIREZ

Fig. 10 Simulated vibration response at node 3 in the x direction for a three-pin coupling, O 30 Hz and
unbalance quality of G 6.3 located at different angular positions

eld, a 1808 difference in phase between the 16 spectral dependent on the frequencies of the variation in
component measured on bearings in the same direction coupling stiffness.
at both sides of the coupling is frequently taken as a The magnitude of misalignment is not easy to
practical criterion for diagnosing a shaft misalignment evaluate on machinery that is running. A systems
problem. According to our previous analysis, this misalignment forces transmitted from a coupling to a
assumption would be true only if the 1808 phase shaft is mainly static, i.e. unidirectional and constant,
difference between the action and reaction forces acting and is difcult to measure externally with commercial
on each rotor generated by the misalignment were not instruments, as has been indicated by Piotrowski [1].
modied by the FRFs related to the coupling force and However, it is possible to estimate the degree of
bearing vibration, and also by other sources of vibra- misalignment of the rotor shafts through the measure-
tion, such as unbalance. Because of this, it would be ment of the effect of these forces on the bearing
better to perform phase analysis using higher spectral vibrations. From equation (15), it can be seen that
components. This point will be illustrated with experi- loads generated by misalignment are directly propor-
ments in Part 2 of this paper. tional to the degree of misalignment and the coupling
stiffness. Because of this, vibration measurement is a
good qualitative indicator for detecting the degree of
misalignment in machinery while it is running.
The vibratory response of the bearings depends
4 CONCLUSIONS greatly on the frequency response functions between
the coupling misalignment forces and the bearings
A theoretical model suitable for determining the vibration. This makes qualitative analysis difcult, even
vibratory response of misaligned shaft rotors connected if the coupling stiffness spectrum is known, especially
by a exible coupling has been developed. Recognizing when some particular component is close to a natural
that spectral analysis is a basic technique used to frequency. An inverse method to determine dynamic
diagnose the mechanical condition of machinery, the misalignment forces from vibrations measured on the
expected vibration spectra for two exible couplings, a rotor bearings is now being investigated.
three-pin Renold and a three-jaw Lovejoy coupling, In Part 2 of this paper, experimental tests and a
were analysed. The calculated results and the qualitative discussion of the results are presented to evaluate the
analysis indicate that the vibration generated by shaft real effectiveness of the developed model. Total agree-
misalignment is caused by a variation in coupling ment between theory and experimental results observed
stiffness as the shaft rotates, and the forcing frequencies validates the theoretical model. The results obtained
generated by shaft misalignment are composed of the from our model simulations and from experimental tests
harmonics of the shaft speed, which are directly are synthesized to give the eld engineer useful

Proc. Instn Mech. Engrs Vol. 218 Part C: J. Mechanical Engineering


Downloaded Science
from pic.sagepub.com at University of Otago Library on December 22, 2014 C18203 # IMechE 2004
VIBRATION ANALYSIS OF ROTORS FOR THE IDENTIFICATION OF SHAFT MISALIGNMENT. PART 1 985

misalignment diagnostic tools for monitoring a 5 Xu, M. and Marangoni, R. D. Vibration analysis of a
machines health. motorexible couplingrotor system subject to misalign-
ment and unbalance. Part II: experimental validation. J.
Sound Vibr., 1994, 176(5), 681691.
6 Sekhar, A. S. and Prabhu, B. S. Effects of coupling
ACKNOWLEDGEMENTS misalignment on vibrations of rotating machinery. J.
Sound Vibr., 1995, 185(4), 655671.
The authors would like to acknowledge the Fondo 7 Simon, G. Prediction of vibration behaviour of large
Nacional de Ciencia y Tecnolog a of Chile, Project turbomachinery on elastic foundations due to unbalance
1030323, for support of this work. and coupling misalignment. Proc. Instn Mech. Engrs, Part
C: J. Mechanical Engineering Science, 1992, 206(C1), 29
39.
8 Bachschmid, N., Pennachi, P. and Vania, A. Identication
REFERENCES of multiple faults in rotor systems. J. Sound Vibr., 2002,
254(2), 327366.
1 Piotrowski, J. Shaft Alignment Handbook, 1986 (Marcel 9 Nelson, H. D. and McVaugh, J. M. The dynamics of rotor
Dekker, New York). bearing systems using nite elements. Trans. ASME, J.
2 Al-Hussain, K. M. and Redmond, I. Dynamic response of two Engng for Industry, 1976, 98(2), 593600.
rotors connected by a rigid mechanical coupling with parallel 10 Gibbons, C. B. Coupling misalignment forces. In Proceed-
misalignment. J. Sound Vibr., 2002, 249(3), 483498. ings of the Fifth Turbomachinery Symposium, Gas
3 Dewell, D. L. and Mitchell, L. D. Detection of a misaligned Turbine Laboratories, Texas A & M University, College
disk coupling using spectrum analysis. Trans. ASME, J. Station, Texas, 1976, pp. 111116.
Vibr., Acoust., Stress and Reliability Des., 1984, 106, 915. 11 Schwerdlin, H. Reaction forces in elastomeric couplings.
4 Xu, M. and Marangoni, R. D. Vibration analysis of a Mach. Des., 1979, 51(16), 7679.
motorexible couplingrotor system subject to misalign- 12 Ram rez, D. E. Analysis of the vibratory behaviour of
ment and unbalance. Part I: theoretical model and analysis. rotating machines. Masters thesis, Universidad de Con-
J. Sound Vibr., 1994, 176(5), 663679. cepcion, Chile, 2003.

C18203 # IMechE 2004 Downloaded from pic.sagepub.com at University ofProc. InstnonMech.


Otago Library Engrs
December Vol.
22, 2014 218 Part C: J. Mechanical Engineering Science

Anda mungkin juga menyukai