You are on page 1of 13

2076

Dynamics of a hydropower generator subjected to


unbalanced magnetic pull
Y Calleecharan* and J-O Aidanpää
Department of Engineering Sciences and Mathematics, Division of Solid Mechanics, Luleå, Tekniska Universitet,
Luleå, Sweden

The manuscript was received on 6 September 2010 and was accepted after revision for publication on 23 February 2011.

DOI: 10.1177/0954406211403844

Abstract: Eccentricity leading to unbalanced magnetic pull (UMP) in electrical machines is a


significant concern in industry. The UMP is known to be composed of two components: a radial
component and a tangential one. Models that are used in industry tend to include the radial com-
ponent alone. In this article, a Jeffcott rotor model together with a new UMP model that incorpo-
rates both radial and tangential UMP constituents is studied for an industrial hydropower
generator. Characterizing the UMP as springs permits the model to inherit UMP stiffness contri-
bution. Interesting dynamics are observed with the new model for a wide range of external forcing
frequencies. It is shown firstly that the new UMP model is sensitive to forcing frequency in the rotor
movements. Secondly, it is found that this sensitivity to forcing frequency increases with decreasing
rotor system stiffness. Moreover, quasi-periodic motion in the rotor displacements is observed and
it is noted that the rotor does not need to be forced by frequencies above its critical speed for this less
desirable motion to occur. Thus, it becomes interesting to be able to account for the UMP stiffness
contribution in order to curb machine malfunction which might result from these UMP forces.

Keywords: eccentricity, whirling, UMP, radial, tangential, rotor, bifurcation, quasi-periodic motion

1 INTRODUCTION In literature, analysis has been mainly limited to


static eccentricity as it is more amenable to theoret-
Rotor–stator eccentricity in electrical machines is an ical conceptualizations. Static eccentricity may be
issue that have caught attention for a long time [1, 2] caused for example in turbine generators due to mis-
and is an important item in condition monitoring [3]. alignment of the rotor shaft with respect to the sup-
This undesirable condition arises due to either porting bearings. Some authors [5, 6] have considered
mechanical or electrical faults or due to a combina- mixed eccentricity conditions whereas others [7, 8]
tion of both types of faults. Rotor–stator eccentricity have given attention to each of static eccentricity
can be classified into static eccentricity, dynamic phenomenon and of dynamic eccentricity phenome-
eccentricity, and mixed eccentricity. Mixed eccentric- non separately in electrical machines.
ity is hard to analyse due to a multitude of factors that Mechanical vibrations due to rotor–stator eccen-
can lead to this type of eccentricity and in practice, tricity can lead to various types of operating malfunc-
dynamic eccentricity implicitly implies the existence tions such as bearing failures which can precede
of both types of eccentricities [4]. rotor–stator rub. It is known that the unbalanced
magnetic pull (UMP) has the adverse effect to
*Correspondence author: Department of Engineering Sciences and increase the rotor–stator eccentricity [9]. Now, the
Mathematics, Division of Solid Mechanics, Luleå, Tekniska UMP forces the rotor centre to move further away
Universitet, 971 87 Luleå, Sweden. from the stator bore centre, thereby augmenting the
email: yogcal@ltu.se possibility of operating malfunctions as stated earlier.

Proc. IMechE Vol. 225 Part C: J. Mechanical Engineering Science


Dynamics of a hydropower generator 2077

Earlier analysis considered only a radial UMP [10] related forces as described in an earlier paragraph,
whereas more recent works [11, 12] depict the UMP the rotor is seen to be affected by unwanted frequen-
as consisting of two components: a radial one and a cies also. In this article, the consequences of these
tangential component. The radial UMP constituent external forces shall be explored over a certain fre-
arises due to an unbalance in magnetic flux on oppo- quency range. These outside influences will affect
site sides of the rotor brought about by the rotor– the whirling of the rotor inside the stator and will
stator eccentricity. The tangential UMP part on the affect the magnitude of the UMP components
other hand originates from the presence of damper likewise.
windings or amortisseur windings that normally Now, the rotor is a major component in a generator
equip synchronous machines. More information system and it can have different aspect ratios, i.e. dif-
about the importance of damper windings can be ferent length-to-diameter ratios. The stiffness prop-
found, for example, in reference [13]. Unlike static erty of the rotor paves a way for simple mechanical
eccentricity which acts in a single direction, approx- dynamic analysis when the UMP constituents can
imately in the direction of the shortest air-gap length also be cast into stiffness formulations. An industrial
[14] that is fixed in space, dynamic eccentricity pro- hydropower synchronous generator will be consid-
duces a rotating UMP [4] with the shortest air-gap ered here [23]. A qualitative knowledge of the varia-
length not fixed in space. tion in rotor displacements inside the stator is of
In hydropower generators where the clearance significance and relevance in industry as sudden
between the rotor and the stator is of low order of changes in the displacements of the rotor centre
magnitude as compared to the diameter of the rotor, with respect to the centre of the stator bore can
rotor–stator eccentricity leading to vibration problems induce high amplitudes of vibration in other parts
becomes a crucial aspect in both design and operation of the generator which may in turn entail machine
in order to prevent outage. The length of an air-gap is a malfunction and premature wear.
design compromise between mechanical and electri- Present models in the hydropower industry use the
cal parameters, and it should not be too short, other- maximum radial UMP value only – more distinctly the
wise the UMP may be of very large magnitude. maximum radial stiffness that corresponds to the max-
Although several papers [15–19], for instance, have imum radial UMP in the computations of the dynamic
studied the UMP in electrical machines from the response of the rotor. We shall see that this approxi-
electrical point of view, fewer papers in the literature mation is insufficient if reliable information about
[20–22], for example, have attempted to model the the dynamics of the rotor–stator system is needed.
UMP for mechanical dynamic analysis. It is worth- Simulations are included in this article which cater
while to note that limitations in numerical computa- for different cases of interest of the UMP components.
tions in the earlier years did restrict rigorous These UMP components determine how the rotor is
investigations to be made. This article comes in as a moving inside the stator. We are interested here in the
first step to fill this void. A parametric model for the displacements of the rotor centre with respect to the
UMP was derived in reference [14] but it is unclear on stator centre in the radial direction under the effect of
how to use the model in mechanical analysis. the previously mentioned external forcing frequencies.
The main objective of this article remains the iden- Three important cases of interest come into play: the
tification and characterization of the dynamics of the first case, mentioned briefly at the beginning of this
UMP constituents on the radial displacement of the paragraph, assumes existence of the radial UMP only
rotor centre inside the stator bore, and tools such as and furthermore simplifies the analysis to having a
bifurcation diagrams and rotor centre orbit plots are single value of the radial UMP stiffness, namely the
used. In a hydroelectric plant, there are several maximum radial UMP stiffness, at all whirling fre-
machineries, often close together, that operate in syn- quencies or whirling angular velocities of the rotor.
ergy. This gives rise to a multitude of operating and The second case considers a new UMP model, which
vibrating frequencies that are picked up by the is the main theme of this article, and treats quantita-
machineries. In effect, the vibrating signature of a tive behaviours of both UMP parts over a rotor distinc-
particular machinery component will often be pol- tive whirling frequency range. Finally, the third case
luted by other external frequencies that belong to considers the same new model as in the second case
nearby operating machines. Also, depending on the but with the tangential UMP component discarded.
rate of water flow around the blades at the bottom These different models are investigated in this article
of the turbine, a rotor may pick up vibrations from and have as purpose to clarify the importance of
the blades’ movements. These external disturbances not neglecting the actual behaviour of the UMP com-
constitute forcing frequencies that act on the rotor. ponents over the whirling frequency range of a rotor.
Therefore, besides being affected by eccentricity The new model portrayed in this article is applied to

Proc. IMechE Vol. 225 Part C: J. Mechanical Engineering Science


2078 Y Calleecharan and J-O Aidanpää

a simple Jeffcott rotor. Simplifications accompanying (a)


this model in order to simplify analysis are firstly that inner stator bore m
k, c
the bearings’ stiffnesses have been kept constant; sec-
ondly, a combined damping value has been used for
the rotor shaft and that of the bearings; and thirdly,
mass eccentricity effect on the rotor has been omitted.
It has been deemed sensible to exclude such parame-
ters as they would clutter the analysis making the
effects of the UMP to be less easily distinguishable.
Apart that, assumptions governing the proposed new y
(b)
UMP model are that saturation effects of the iron parts
are neglected and the UMP has a linear behaviour with FtUMP
eccentricity. FrUMP

2 MODEL r
θ
2.1 Simulating the rotor–stator unit
Cro x
Cst
The model used is a symmetrical Jeffcott rotor as ωro
a
shown in Fig. 1(a) and the main parameters of the
rotor–stator system are given in Fig. 1(b). The main
difference in this model as compared to the rigidly
supported Jeffcott rotor model [24] is the inclusion
of the bearings’ stiffnesses and damping. The mass
Fig. 1 (a) The symmetrical Jeffcott rotor supported on
of the rotor is m and this rotor rests on two isotropic bearings. (b) Rotor inside the stator with the
bearings each of stiffness kb via a supporting shaft main parameters in the Cartesian coordinates
having stiffness kro. The combined damping of the system and with the directions of the forces
shaft and that of the bearings is assumed to be c, FrUMP and FtUMP defined. The size of the stator
and this will be accounted in the equations of is grossly exaggerated. The rotor and stator cen-
motions by a damping ratio  as will be seen in tres are marked, in order, by points Cro and Cst.
Gravity is ignored in the model. The angular dis-
section 2.2. The effect of gravity is neglected in the
placement is given by  and the rotor–stator
model. For the Jeffcott rotor model shown in Fig. 1(a) eccentricity is represented by the distance r
together accounting for the contributions of the UMP
components FrUMP and FtUMP as depicted in Fig. 1(b),
equations of motion can be set up in the Cartesian
coordinates system when an external force having Table 1 Numerical data of the rotor of Porjus genera-
amplitude F with forcing frequency !fo exists. This tor Unit 8. All data pertain to the rotor unless
forcing frequency, !fo, is not to be confused with mentioned otherwise
the mechanical angular velocity of the rotor, !ro, the Parameter Value
latter being fixed (Table 1) for the system in consid- Mass, m [kg] 30 000
eration. The nature of the forcing frequency, !fo, was Rotor–stator eccentricity, r [m] 1.7  103
Mean air-gap length, am [m] 17  103
keyed out in section 1 and !fo encompasses the whirl- Mechanical angular velocity, !ro [rad/s] 22.44
ing frequencies in the rotor whirling range, the fre- Young’s modulus of elasticity, E [Pa] 200  109
quency values of which are introduced in section 3.1. Area moment of inertia, I [m4] 0.0635
Length of shaft, L [m] 3.6
As a rotor can experience both positive and negative Bearing stiffness (per unit), kb [N/m] 500  106
whirl, the forcing frequency will take on both posi-
tive and negative whirling frequency values in the
whirling range. This whirling range that identifies
the span of forcing frequencies to be applied on the
system is a limitation of the usefulness of the UMP
models put forward in this article since the UMP forces in the system. Returning back to Fig. 1, if we
curves only hold for this whirling frequency range assume that any external forcing frequencies may be
(section 3.1 and Fig. 2). It is to be underlined that combined crudely into a single-frequency compo-
angular whirling of the rotor is a consequence of nent !fo and with a unique amplitude F, equations

Proc. IMechE Vol. 225 Part C: J. Mechanical Engineering Science


Dynamics of a hydropower generator 2079

UMP
of motion using the Cartesian x- and y- directions stiffnesses Kr,t are amenable for estimation using
become
UMP
UMP Fr,t
" # ! " # ! " # ! Kr,t ¼ ð3Þ
m 0 x€ c 0 x_ k 0 x am  10%
þ þ
0 m y€ 0 c y_ 0 k y on the assumption of the linear behaviour of the UMP
! ! curves at low eccentricities. Equation (3) thus pro-
UMP
Fx F cos !fo t vides a formula to estimate the UMP components at
 ¼ ð1Þ
FyUMP F sin !fo t rotor–stator eccentricities other than at 10 per cent.
The actual non-linear behaviour of the UMP with
where k is the system stiffness and is given by the eccentricity together with saturation effects will
combined stiffness values of the shaft and that of limit the usefulness of this linear assumption to
the isotropic bearings, and the Cartesian UMP com- low-eccentricity values and one way to ensure that
ponents are given by this linearity holds is to select external forcing fre-
quencies with appropriate amplitudes.
FxUMP ¼ FrUMP cos   FtUMP sin  Now, equation (2) can be re-written in terms of
stiffnesses as
FyUMP ¼ FrUMP sin  þ FtUMP cos  ð2Þ
FxUMP ¼ KrUMP x  KtUMP y
Now, with a low rotor–stator eccentricity ratio FyUMP ¼ KrUMP y þ KtUMP x ð4Þ
(ratio of the displacement, r, of the rotor centre
from the centre of the stator bore to the mean air- where the Cartesian coordinates x and y are the loca-
gap length, am, in the radial direction), the UMP con- tions of the centre of the eccentric rotor with respect
stituents can be approximated by linear spring to that of the stator. Altogether, equation (1) can be
models with stiffnesses KrUMP and KtUMP in the conveniently transformed into
respective directions assuming linear behaviour of " # ! " # ! " # !
m 0 x€ c 0 x_ k 0 x
the UMP curves with eccentricity ratio. We therefore þ þ
note that FrUMP ¼ KrUMP r and FtUMP ¼ KtUMP r. The 0 m y€ 0 c y_ 0 k y
simulation points on the UMP curves in Fig. 2 were " UMP # ! !
Kr KtUMP x F cos !fo t
computed using 10 per cent dynamic rotor–stator  ¼ ð5Þ
eccentricity. It is worthy to note that the UMP KtUMP KrUMP y F sin !fo t

200

150
UMP
Fr
100
Force F, kN

50

0
F tUMP
–50

–100
–2 –1.5 –1 –0.5 0 0.5 1 1.5 2 2.5 3
Whirl ratio, w whr

Fig. 2 Least-squares fitted curves of FrUMP and FtUMP constituents as a function of the whirl ratio
!whr. Upper curve shows the radial UMP component FrUMP and bottom curve represents the
tangential UMP component FtUMP . The original data sets of 20 simulation points for each
curve are overlaid, in order, as transparent triangles and squares for FrUMP and FtUMP .
Synchronous whirling is identified at the point !whr ¼ 1 and this corresponds to a rotor
whirling angular velocity of !wh ¼ 22.44 rad/s (Table 1)

Proc. IMechE Vol. 225 Part C: J. Mechanical Engineering Science


2080 Y Calleecharan and J-O Aidanpää

The UMP stiffness matrix term in equation (5) 2.2 Non-dimensionalization of the
has been proposed earlier with the signs of the equations of motion
counter-diagonal elements reversed [25]. This UMP
stiffness matrix model was found to be incorrect Equation (7) is put into a non-dimensional form by
and hence was subsequently revised during 2009 choosing
(U. Lundin, 2009, personal communication). The t ¼ ½t  t ,
!fo ¼ ½!fo  !fo , r ¼ ½r r,
new model as given in equation (5) is believed to be    ð8Þ
UMP UMP UMP
right as regard to the direction of the tangential FtUMP Kr,t ¼ ½Kr,t Kr,t and F ¼ F F
component [12]. Now, equation (5) can be recast in
using the following scales
Polar coordinates form with the position of the rotor
centre being described by r for radial displacement 1 UMP

½t  ¼ , ½!fo  ¼ !n , ½r ¼ am , ½Kr,t ¼k
and  for angular displacement (positive  being in the !n
counterclockwise direction and _ is the angular whirl- and ½F  ¼ k am ð9Þ
ing velocity of the rotor), both with respect to the
together with the following relationships
centre of the stator Cst (Fig. 1(b)). Introducing x ¼ r
cos  and y ¼ r sin  in equation (5), we obtain k c
!2n ¼ and 2  !n ¼ ð10Þ
m m

mðr€ cos   2 r_ _ sin   r € sin   r _2 cos Þ giving


2
þ cðr_ cos   r _ sin Þ þ k r cos   KrUMP r cos  r 00 þ 2  r 0 þ ð1  0 r  KrUMP r ¼ F cosð!fo t  Þ

þ KtUMP r sin  ¼ F cos !fo t ð6aÞ r 00 þ 2 ðr 0 þ  r 0  KtUMP r ¼ F sinð!fo t  Þ
ð11Þ

mðr€ sin  þ 2 r_ _ cos  þ r € cos   r _2 sin Þ as the non-dimensional form of the equations of
motion where 0 denotes differentiation with respect
þ cðr_ sin  þ r _ cos Þ þ k r sin  to the dimensionless time t . From equations (8) and
 KrUMP r sin   KtUMP r cos  ¼ F sin !fo t ð6bÞ (9), it is seen that firstly the external forcing fre-
quency, !fo, is scaled by the natural frequency, !n,
Equation (6) can be simplified using trigonometry. To of the system in consideration. This natural fre-
this end, equation (6a) can be multiplied by cos  quency, !n, will vary with the stiffness, k, of the
and equation (6b) by sin  respectively, and added system. Secondly, the radial rotor–stator eccentricity,
together to form equation (7a) on simplification. r, is scaled by the mean air-gap length, am, which
Also, equation (6a) can be multiplied by  sin  and ensues that a normalized displacement, r, of the
equation (6b) by cos  respectively, and summed and rotor centre Cro of unity magnitude in the radial
reduced to equation (7b). The reduced and rear- direction leads to rotor–stator bore contact. Thirdly,
ranged equations turn into the force normalization by the factor kam represents
the static force which is required to move the rotor
to the stator bore surface when no UMP exists. Next,
m r€ þ c r_ þ ðk  m _2 Þr  KrUMP r equation (10) shows how the damping ratio (which
ð7aÞ
¼ F cosð!fo t  Þ takes into account the damping of both the shaft and
that of the bearings as stated in section 2.1) comes
into being in the equations of motion. The mass m in
m r € þ ð2 m r_ þ c rÞ_  KtUMP r ¼ F sinð!fo t  Þ equation (10) represents the mass of the rotor and its
ð7bÞ value is given in Table 1.

We note that all terms in equation (7) constitute


force terms and that the effect of the UMP compo-
3 COMPUTER SIMULATIONS
nents at different rotor-eccentricity positions are
immediately visible by the last term on the left- 3.1 Modelling and computation of the
hand side of each of equations (7a) and (7b). UMP components
Furthermore, it is important to make remark that
the forcing terms on the right-hand sides of equation The UMP components were numerically estimated at
(7) are only external forces as described in section 1 full load condition with 10 per cent dynamic rotor–
and that any force due to the mass unbalance of the stator eccentricity at the University of Uppsala [26]
rotor is absent in the model. using the parameters of the Porjus generator

Proc. IMechE Vol. 225 Part C: J. Mechanical Engineering Science


Dynamics of a hydropower generator 2081

Unit 8 [23]. Some important numerical data pertain- The second case is the full new model (equation
ing to this generator are given in Table 1. The UMP (5)) which takes into account both UMP components
components FrUMP and FtUMP were simulated at 20 and is given by equation (11) which is repeated here
points in the whirling ratio range from 2 to 3 [27], for convenience.
i.e. from 44.88 to 67.32 rad/s where the whirling 2
ratio, !whr, is the ratio of the rotor angular whirling r 00 þ 2  r 0 þ ð1  0 r  KrUMP r ¼ F cosð!fo t  Þ

velocity, !wh, to its mechanical angular velocity, !ro. r 00 þ 2 ðr 0 þ  r 0  KtUMP r ¼ F sinð!fo t  Þ
This makes an angular whirling velocity of value ð14Þ
!wh ¼ 22.44 rad/s (Table 1) to correspond to unity
whirling ratio (!whr ¼ 1). The equations of motion in the new model of equa-
The simulated UMP components were fitted [28] tion (14) compute both UMP parts for each whirling
using the least-squares criterion with rational func- ratio according to equation (12). Finally, the third
tions as a function of the whirling ratio !whr. The esti- case is a subset of the second case and equation
mated rational function models, which are smooth in (11) or equation (14) reduces to
the prescribed whirling ratio range, are given, in 2
r 00 þ 2  r 0 þ ð1  0 r  KrUMP r ¼ F cosð!fo t  Þ
order, by equations (12a) and (12b), and the coeffi- 
cients A’s and B’s are provided in the Appendix. r 00 þ 2 ðr 0 þ  r 0 ¼ F sinð!fo t  Þ
Figure 2 shows the fitted curves for the radial and ð15Þ
tangential UMP together with the simulation points.
The commonality among the three cases is that the
A0r þ A1r  !whr þ A2r  !2whr corresponding UMP stiffness(es) is(are) established
FrUMP ð!whr Þ ¼  from Fig. 2. The computations are done using
B0r þ B1r  !whr þ B2r  !2whr
non-dimensional variables as expressed in equations
ð12aÞ
(13) to (15).
A0t þ A1t  !whr A single-step fourth-order Runge–Kutta procedure
FtUMP ð!whr Þ ¼ 
B0t þ B1t  !whr þ B2t  !2whr has been chosen as the time integration algorithm.
ð12bÞ This popular integration routine has a global error
term of O(h4) [29]. A C90 [30] code was written to this
end and the simulations were carried out using
3.2 Time integration Microsoft Visual Cþþ 2010 Express [31] on a computer
equipped with an IntelÕ -based T7700 Core 2 Duo
Equation (11) can be integrated with a time integra- Mobile Processor. All the computations were made
tion procedure to find the normalized radial displace- using long double precision which is 15 decimal signif-
ment r. The degree of variation in the amplitude of r icant digits in Microsoft Visual Cþþ 2010 Express.
is a direct indication of the amount of movement of
the rotor centre with respect to the stator bore centre. 3.3 Results
Varying the normalized forcing frequency !fo term in
equation (11) in the time integration procedure can The simulation results that will follow in sections
point out forcing frequencies, !fo, at which large var- 3.3.1 and 3.3.2 comprise mainly of two parts. In
iation in amplitudes of the rotor movement may lead both simulation parts, it was found informative to
to vibration problems. As pointed out in section 2.1, vary the system stiffness k in equations (13) to (15)
the external forcing frequencies comprise the whirl- and observe the resulting change in the normalized
UMP
ing frequencies for which models of Fr,t exist, i.e. displacement r of the rotor centre Cro in the radial
the span of !fo is equal to the range of !wh which is direction with the forcing frequency !fo. The informa-
from 44.88 to 67.32 rad/s (section 3.1). tion gained from such an analysis is beneficial in the
Three separate cases worthy of interest are investi- study of different rotor designs since rotors can, for
gated in this article. They are all based on equation example, be stiff or slender in relation to other parts of
(11). The three cases were introduced in section 1. In a hydropower machine. It is important to note that
the first case, the maximum radial UMP stiffness the generator under consideration [23] is regarded as
value is used at all whirling angular velocities and a stiff machine. Therefore in the simulations, it was
equation (11) reduces to deemed relevant and reasonable to form a subrange
of stiffness values from the nominal stiffness value of
2 [23]. More specifically, stiffness ratios were intro-
r 00 þ 2  r 0 þ ð1  0 r  KrUMP
max
r ¼ F cosð!fo t  Þ
0
 0 duced with the nominal stiffness value of [23] taken
00
r  þ 2 ðr þ  r  ¼ F sinð!fo t  Þ ð13Þ as unity and stiffness ratios above unity were

Proc. IMechE Vol. 225 Part C: J. Mechanical Engineering Science


2082 Y Calleecharan and J-O Aidanpää

excluded as such stiffer machines are less likely to so as to avoid trespassing higher eccentricity ratios,
exist. r, where the proposed UMP model in equation (5) will
Section 3.3.1 shows the outcome of varying the not genuinely hold (section 2.1).
system stiffness on the maximum normalized dis-
placement r max and section 3.3.2 presents bifurcation 3.3.1 Variation of the maximum normalized
diagrams for the parameter r and here likewise con- displacement with system stiffness
sidering variation of the system stiffness k or equiva-
lently the system stiffness ratio k. Rotor centre orbits Varying the overall stiffness ratio of the rotor system
estimation ends section 3.3.2 for some stiffness ratio can provide useful information in that a knowledge of
values. The overall stiffness ratio, k, is the ratio of the variation in maximum r with forcing frequency !fo
actual stiffness value used in the simulations to the will indicate whether the rotor can be wobbling
nominal system stiffness value. The nominal stiffness dangerously inside the stator which can cause
ratio can be readily computed using the parameters rotor–stator rub in severe cases. Figure 3 lays out
given in Table 1, with the shaft stiffness calculated as the variation of the maximum normalized radial dis-
a bending stiffness. Of importance is to note that the placement of r, r max , with respect to the forcing fre-
system damping ratio, , was set to a value of 0.1 in quency !fo. The normalized stiffness ratio values
both the simulations of sections 3.3.1 and 3.3.2. This chosen in these simulations were k ¼ 0:2, 0.5, 0.7,
value has been chosen in accordance with the typical and unity, respectively. Three cases of interest pertain
 value of 5 to 10 per cent that is normally used in the to Fig. 3. These cases were mentioned at the end of
hydropower industry. In addition, the amplitude of section 1. Their respective equations of motion were
the forcing frequency !fo has been sensibly chosen laid out in section 3.2 and some elaborations follow

(a) (b)
0.7 0.18
Max. normalised radial

Max. normalised radial

0.16
displ. rmax

0.5
displ. rmax

0.14

0.3
0.12

0.1 0.1
–60 –40 –20 0 20 40 60 80 –60 –40 –20 0 20 40 60 80
Forcing frequency wfo, rad/s Forcing frequency wfo, rad/s

(c) 0.16 (d)


0.13
Max. normalised radial

Max. normalised radial

0.14
displ. rmax

displ. rmax

0.12

0.12
0.11

0.1 0.1
–60 –40 –20 0 20 40 60 80 –60 –40 –20 0 20 40 60 80
Forcing frequency wfo, rad/s Forcing frequency wfo, rad/s

Fig. 3 Time integration of equations (13)–(15) using the fourth-order Runge–Kutta routine for
different stiffness ratio values k = 0.2, 0.5, 0.7, and 1, respectively. Variations in steady
state values of the maximum r values, r max , are shown over the 300-point forcing frequency
range. The mechanical angular velocity of the rotor is !ro ¼ 22.44 rad/s. The red curves
represent the first case with only consideration of the maximum KrUMP value from Fig. 2.
The black curves denote the second case in which both FrUMP and FtUMP fitted graphs of Fig. 2
in the new UMP model are accounted for. Last, the blue curves refer to the third case in
which the contribution of the tangential UMP component FtUMP is taken away from the
second case. Details for these three different cases are given in section 3.3.1. Overlapping of
the blue and black curves exists in the diagrams for some frequency regions

Proc. IMechE Vol. 225 Part C: J. Mechanical Engineering Science


Dynamics of a hydropower generator 2083

here for convenience. The first case described model includes both the radial and tangential UMP
by equation (13) and represented by red curves in constituents. This particular case indeed gives a more
Fig. 3 considers the maximum radial UMP stiffness accurate representation of the effects of the UMP and
KrUMP
max
. This implies that the same maximum value of it can be seen from Fig. 3 with reference to the black
UMP UMP FrUMP
Kr , i.e. Krmax ¼ am 0:1 ¼ 87:485 MN/m (Fig. 2) is
max
curves that interesting dynamics in terms of multiple
used at all angular whirling velocities of the rotor. discontinuities arise in the plots. Hence, the bifurca-
The second case which is described by equation (14) tion diagrams estimated in this section aim at further
is depicted by black curves in Fig. 3 and this case refers investigating the dynamics of the new model with
to the new model with consideration of both UMP both FrUMP and FtUMP incorporated.
components FrUMP and FtUMP (Fig. 2). Lastly, the The same initial conditions as mentioned in sec-
third case described by equation (15) and represented tion 3.3.1 were used here. Steady-state conditions
by blue curves in Fig. 3 denotes the new model but were taken after 800 corresponding periods of !fo.
with the tangential UMP component FtUMP left out. In The number of data points was 5  106. Following
other words, the third case views the radial UMP only this, 100 Poincaré sections were taken at appropriate
and this according to FrUMP as shown in Fig. 2. time intervals. The strobe frequency applied was
With initial conditions set to 0.01 for the states r, r 0 , the external forcing frequency !fo. In locating the
 and 0 in the respective equations of motion for the Poincaré sections, the step size was set to
different cases, i.e. equations (13) to (15), a non- ð1=3000 th of the rotor mechanical angular velocity
dimensional time step of t ¼ 0:004 was found suffi- period.
cient and the maximum r, r max , was computed after Figure 4 shows the bifurcation diagrams from the
steady state conditions were reached in the time inte- simulations with k ¼ 0:2, 0.5, 0.7, and unity, respec-
gration process. The last one-tenth of the total of tively, with 2000 frequency points in the prescribed
2.5  106 data samples used in the simulation was forcing frequency range. Horizontal gaps appear in
utilized in estimating r max . every plot of Fig. 4 in the vicinity of !fo ¼ 0 where
Two points becomes essential to comment here. the integration algorithm was stopped so as to pre-
Firstly, in each time step at a given forcing frequency vent too long a computation time. Figure 5 displays
!fo, the UMP stiffnesses Kr,t UMP
are computed form the zoomed-in portions of the plots in Fig. 4. These
prescribed force curves Fr,t UMP
from Fig. 2 except for enlarged regions lie between the forcing frequencies
the first case where the same KrUMP is used at all whirl- of 26 rad/s and 34 rad/s and they aid to emphasize
max
ing frequencies for all forcing frequencies applied. visually the interesting dynamics happening in the
These stiffnesses after normalization with the nomi- system. More specifically, multiple solutions arise in
nal stiffness, k, (equation (9)) are used in equations some frequency ranges. The combination of single
(13) to (15) to estimate the UMP components as a solutions and multiple solutions form together the
result of a change in the rotor–stator eccentricity or bifurcation diagrams that are observed in Fig. 4.
more appropriately, the normalized radial displace- So as to gain a further understanding of the dynam-
ment r. This procedure demonstrates the linear rela- ics involved, rotor centre orbit plots are presented in
tionship Fr,t UMP UMP
¼ Kr,t r which has been assumed at Fig. 6 for two stiffness ratio values, namely k ¼ 0:2
the onset of the derivation of the equations of motion and k ¼ 0:5. These plots again apply to the second
UMP
in section 2.1. Secondly, the amplitude of !fo in equa- case where consideration is given to Fr,t according
tions (13) to (15) has been suitably chosen in the sim- to Fig. 2. Orbit plots permit visualization of the rotor
ulations to cater for the worst case scenario (Fig. 3(a)) at motion inside the stator and furnish a simple way to
low stiffness ratio values in preventing r to exceed unity discern periodic motions from non-periodic ones.
since larger displacements are observed at low k. The orbit plots were constructed using the last
200 000 data points of the 2.5  106 total data samples
3.3.2 Bifurcation diagrams for the radial used in the simulation at the forcing frequency of
displacement of the rotor centre !fo ¼ 11.03 rad/s. This frequency value coincides
roughly with the midpoint value of the forcing
Bifurcation diagrams help to identify abrupt changes frequency range which is from 44.88 rad/s to 67.32
in topologies of the phase portrait of a system due to rad/s (section 3.1).
stability changes. These abrupt changes occur as one
or more parameters pass(es) through some critical 4 DISCUSSIONS AND CONCLUSIONS
value(s) [32]. In this section, bifurcation diagrams
for the parameter r are estimated for different stiff- A first observation from the plots contained in Figs 3
ness ratio values k as in section 3.3.1. However, only and 4 is that the forces belonging to the UMP cer-
the second case is considered, i.e. where the UMP tainly bring about modifications in the dynamics of

Proc. IMechE Vol. 225 Part C: J. Mechanical Engineering Science


2084 Y Calleecharan and J-O Aidanpää

(a) (b) 0.16


0.4
0.14

Normalised radial
Normalised radial

displacement r
displacement r

0.3

0.12
0.2

0.1 0.1

0 0.08
–60 –40 –20 0 20 40 60 80 –60 –40 –20 0 20 40 60 80
Forcing frequency wfo, rad/s Forcing frequency wfo, rad/s

(d)
(c)
0.14 0.12

Normalised radial
Normalised radial

displacement r
0.13
displacement r

0.12 0.11

0.11

0.1 0.1

0.09
–60 –40 –20 0 20 40 60 80 –60 –40 –20 0 20 40 60 80
Forcing frequency wfo, rad/s Forcing frequency wfo, rad/s

Fig. 4 Time integration of equation (14) using the fourth-order Runge–Kutta routine for different
stiffness ratio values k ¼ 0.2, 0.5, 0.7, and 1, respectively, represented as bifurcation dia-
grams. These plots apply to the second case where both UMP components are taken into
account (section 3.3.1). Steady-state values of the rotor centre displacement, r, at Poincaré
sections are shown over the 2000-point forcing frequency range. Horizontal gaps in the
plots in the normalized radial displacement, r, values around !fo ¼ 0 are where the integra-
tion process becomes excessively long. The mechanical angular velocity of the rotor is
!ro ¼ 22.44 rad/s

the simple Jeffcott rotor model of Fig. 1(a). The signs in the tangential UMP component at !whr ¼ 1
system stability is indeed affected by the UMP. thus plays a vital role in the stability of the system.
Different explored cases for the type of UMP contri- Figure 3(a) also shows overlapping of the blue curves
bution as described in section 3.3.1 result in different and that of the black curves. This overlapping is to be
system response behaviours as can be seen in Fig. 3. expected as the second case is a superset of the third
Now, stability of any mechanical or electrical system case, the latter being the absence of the tangential
about its operating point is a fundamental require- UMP from the second case.
ment. Yet, for the simple system in Fig. 1 it is noted Another look at Figs 3 and 4 clearly reveals that the
that small deviations in the rotor angular whirling system does in fact need not be forced by frequencies
velocity (in the forward or in the reverse whirl above the critical speed of the rotor for the UMP com-
motion) can produce undesirable dynamics. The ponents to cause strange and unexpected dynamic
effects due to the UMP are more pronounced for behaviour. Changes in the stiffness of a mechanical
the second case (represented by the black curves in system are not uncommon, and in both Figs 3 and 4
Fig. 3) in which both radial and tangential UMP com- we note that a reduced stiffness ratio has as conse-
ponents exist. When both UMP parts are included, quence a large variation in amplitudes in rotor centre
lots of discontinuities take place around the forcing movements. Here again, the second case with both
frequency that corresponds to the mechanical angu- UMP constituents in the model shows marked differ-
lar velocity of the rotor (i.e. at synchronous whirl ences from the other cases. Discontinuities occur
!whr ¼ 1 in Fig. 2). The presence of no discontinuities over longer frequency ranges (compare the black
in the red curves unlike to those displayed by the blue curves for the second case with the blue curves for
and black curves (Fig. 3) substantiates the importance the third case in Fig. 3). As mentioned in the previous
of not omitting the tangential UMP. The change in paragraph, there is overlapping of the response

Proc. IMechE Vol. 225 Part C: J. Mechanical Engineering Science


Dynamics of a hydropower generator 2085

(a) (b)
0.14
0.24
Normalised radial

Normalised radial
0.22
displacement r

displacement r
0.13
0.2
0.18 0.12

0.16
0.11
0.14

26 27 28 29 30 31 32 33 34 26 27 28 29 30 31 32 33 34
Forcing frequency wfo, rad/s Forcing frequency wfo, rad/s

(c) 0.13 (d)


0.12
Normalised radial

Normalised radial
0.115
displacement r

displacement r
0.12

0.11
0.11
0.105

0.1 0.1
26 27 28 29 30 31 32 33 34 26 27 28 29 30 31 32 33 34
Forcing frequency wfo, rad/s Forcing frequency wfo, rad/s

Fig. 5 Zoomed-in windows of the bifurcation plots in Fig. 4 for the frequency range 26 rad/s to
34 rad/s for different stiffness ratio values k ¼ 0.2, 0.5, 0.7, and 1, respectively. These plots
apply to the second case in which both UMP components are taken into account (section
3.3.1). The amplitudes on the y-axis represent the normalized rotor centre displacement, r,
in the radial direction. The mechanical angular velocity of the rotor is !ro ¼ 22.44 rad/s

(a) (b)
0.3 0.15
Normalised rotor displacement

Normalised rotor displacement

0.2 0.1

0.1 0.05
in y direction

in y direction

0 0

–0.1 –0.05

–0.2 –0.1

–0.3 –0.15
–0.3 –0.2 –0.1 0 0.1 0.2 0.3 –0.15 –0.1 –0.05 0 0.05 0.1 0.15
Normalised rotor displacement Normalised rotor displacement
in x direction in x direction

Fig. 6 Rotor centre orbit plots for different stiffness ratio values of k ¼ 0.2 and 0.5, respectively.
These plots apply to the second case in which both UMP components are taken into account
(section 3.3.1). The steady state values of r were taken as the last 200 000 data points in the
integration process (section 3.3.2) and the forcing frequency at which these diagrams were
simulated was !fo ¼ 11.03 rad/s. The normalized Cartesian components are calculated as
x ¼ r cos 0 and y ¼ r sin 0

Proc. IMechE Vol. 225 Part C: J. Mechanical Engineering Science


2086 Y Calleecharan and J-O Aidanpää

curves (blue and black curves in Fig. 3) principally on Pre-existing eccentricity coupled with a wide vari-
the continuous parts of the curves and this overlap- ation in the displacements of the rotor centre origi-
ping increases with increasing stiffness ratio, thereby nating from the effect of UMP can be a central cause
demonstrating the adverse effect of the UMP at low for rotor–stator rub. Besides, the plots in Figs 3 and 4
stiffness ratios on the system response. The current do exhibit very sudden or abrupt changes in the rotor
model that is typically used in industry is shown by radial displacement. Apart from the wear and tear,
the red curves in Fig. 3 and these are very smooth and these brusque movements induce the damper wind-
well behaved. It is observed in Fig. 3 how dropping the ings to work hard to damp these sudden and in some
contribution of the tangential UMP from the analysis cases excessive amplitudes, and in doing so we may
can lead to a false representation of the system expect that the damper windings can heat up consid-
response as mentioned in the earlier paragraph. An erably which would shorten the lifespan of such
oddity is that this model from the first case has a windings. Damper windings in hydropower genera-
strong effect on the natural frequency of the system tors have been studied since a long time [13, 35]. It is
at low stiffness ratios. In particular, we note in Fig. seen in Fig. 3 and from equation (5) that the damper
3(a) that the resonance of the system is decreased windings may not be considered as functioning as
from an expected value of around 78 rad/s to a dampers in the mechanical sense. More erratic move-
value of around 57 rad/s and this is due to the maxi- ments of the rotor are in fact observed in Fig. 3 by the
F UMP
mum radial UMP stiffness, namely KrUMP max
¼ amr0:1
max
black curves which belong to the second case (section
(Fig. 2) which is used at all angular whirling velocities 3.3.1). It is important to make remark here that a
of the rotor to calculate the product KrUMP
max
r in equa- linear relationship between the UMP components
tion (13). and the rotor–stator eccentricity r was assumed
Turning to Fig. 6, we see that quasi-periodic motion (section 2.1). This linear relationship is only valid at
of the rotor is distinctly exposed. The diagrams low eccentricity ratios r till around 25 per cent eccen-
therein apply to the second case namely with both tricity and yet normalized amplitudes of around
UMP components accounted for. At the forcing fre- r max ¼ 0:4 are observed in Fig. 3(a) for the second
quency value of !fo ¼ 11.03 rad/s at which the dia- UMP case. This is not to be regarded as a flagrant
grams of Fig. 6 were created, the trajectories never violation of this linear assumption on which the
close and they fill the whole torus. Quasi-periodicity equations of motion, i.e. equation (11), rest upon.
in a mechanical system is undesirable since the paths Nonetheless, this hypothesis of linear relationship
of the trajectories never repeat and therefore the would definitely be invalid at too high a eccentricity
motion of the system may become structurally unsta-
ratio.
ble [33]. Quasi-periodic motion is one route leading The model that has been proposed in Fig. 1 has lent
to chaos [32], and the plots in Figs 4 and 5 show inter-
itself to a simple analysis of a rotor–stator system sub-
ruptions in the quasi-periodic motion. These regions
jected to UMP. This model, given analytically in equa-
are known as frequency-locked states where the
tion (5), is attractive in that it is formulated as a matrix
incommensurate frequencies become rational in
of order two involving only stiffness UMP parameters.
relation to each other. For the duration of the phase
Three different cases of UMP influence were put for-
or frequency-locked states, the system becomes peri-
ward and treated in this article. With solely mechan-
odic though this is hard to be observed in practice and
ical dynamics introduced into the modelling process,
be identified in a power spectrum. These frequency-
it has been shown that insights can be readily arrived
locked states occur as a result of non-linearity in the
at. More specifically, it has been seen that the UMP
system (equation (11)) when some coupling exists
affects the magnitudes of the displacement of the
between a fixed frequency and some other variable
rotor centre Cro in the radial direction with respect
frequency in the system [34]. For the system in con-
to the centre of the stator bore Cst. Assumptions and
sideration, the fixed frequency is the mechanical
simplifications were made on the mechanical aspects
angular velocity of the rotor, !ro, whereas the variable
in order for the analysis to become tangible and easily
frequency is the external forcing frequency on the
grasped. While an enhanced analysis that would
rotor, !fo, and frequency locking will only appear for
cover electrical parameters as well would provide a
some frequencies at which there is strong coupling
between these two afore-mentioned frequencies. In more in-detailed knowledge about the overall system
addition, the plots in Fig. 4(a) and (b) point out that behaviour and characteristics, and information on
with a low stiffness ratio, the system is more likely to the interplay between mechanical and electrical
be pulled into quasi-periodic regimes at more fre- parameters as regard to both the rotor and the
quencies than for a system having higher stiffness stator of a hydropower generator, it is nevertheless
ratio values (Figs. 4(c) and (d)). hoped that the new model containing both UMP

Proc. IMechE Vol. 225 Part C: J. Mechanical Engineering Science


Dynamics of a hydropower generator 2087

constituents which has been presented herein would Electrical and Communications Engineering,
be appealing in a first mechanical dynamic analysis. Laboratory of Electromechanics, Helsinki
University of Technology, Finland).
13 Kimbark, E. W. Power System Stability Volume III:
FUNDING
Synchronous Machines, IEEE Press Series on Power
Engineering, Piscataway, NJ, 1995 (Wiley IEEE
This project was initially sponsored by Elforsk AB, a Press, New York).
Swedish electricity producers joint research and 14 Burakov, A. and Arkkio, A. Low-order parametric
development company, through the ELEKTRA elec- force model for eccentric-rotor electrical machine
tric power technology research programme. equipped with parallel stator windings and rotor
cage. IET Electr. Power Appl., 2007, 1(4), 532–542.
ß Authors 2011 15 Bahri, A., Ben Chaabane, S., and Seguier, G.
Synchronous machines with high number of poles:
REFERENCES reduction of the resultant radial magnetic attraction
by suitable connection of the stator windings.
In International Conference on Electrical Machines,
1 Walker M. Specification and design of dynamo- Lausanne, Switzerland. 1984, pp. 691–694.
electric machinery, 1915, Longmans’ Electrical 16 Belmans, R., Vandenput, A., and Geysen, W.
Engineering Series (Longmans, London). Calculation of the flux density and the unbalanced
2 Rosenberg, E. Magnetic pull in electrical machines. pull in two pole induction machines. Electr. Eng.
Trans. Am. Inst. Electr. Eng., 1918, 37(2), 1425–1469.
(Archiv fur Elektrotechnik), May 1987, 70(3),
3 Ozelgin, I. Analysis of magnetic flux density for
151–161.
airgap eccentricity and bearing faults. Int. J. Sys.
17 Smith, A. C. and Dorrell, D. G. Calculation and mea-
Appl., Eng. Dev., 2008, 4(2), 162–169.
surement of unbalanced magnetic pull in cage
4 Dorrell, D. G., Thomson, W. T., and Roach, S.
induction motors with eccentric rotors. part 1:
Analysis of airgap flux, current and vibration signals
Analytical model. IEE Proc.—Electr. Power Appl.,
as a function of the combination of static and
1996, 143(3), 193–201.
dynamic airgap eccentricity in 3-phase induction
18 Stoll, R. L. Simple computational model for
motors. IEEE Trans. Ind. Appl., 1997, 33(1), 24–34.
calculating the unbalanced magnetic pull on a
5 Nandi, S., Bharadwaj, R. M., and Toliyat, H. A.
two-pole turbogenerator rotor due to eccentricity.
Performance analysis of a three-phase induction
motor under mixed eccentricity condition. IEEE IEE Proc.—Electr. Power Appl., 1997, 144(4), 263–270.
Trans. Energy Conversion, 2002, 17(2), 392–399. 19 Stancheva, R. and Iatcheva, I. Dynamic behaviour
6 Faiz, J. and Ebrahimi, B. M. Mixed fault diagnosis in investigation of electromagnetic force densities. J.
three-phase squirrel-cage induction motor using Mater. Process. Technol., April 2005, 161(1-2), 258–262.
analysis of air-gap magnetic field. Prog. Electromag. 20 Rosenberg, L. T. Abnormal vibration problems in
Res., PIER, 2006, 64, 239–255. large turbine-driven generators and their solutions.
7 Dorrell, D. G. The sources and characteristics of IEEE Trans. Power Apparatus Syst., 1982,
unbalanced magnetic pull in cage induction PAS-101(10), 4131–4135.
motors with either static or dynamic rotor eccentric- 21 Guo, D., Chu, F., and Chen, D. The unbalanced mag-
ity, In Stockholm Power Tech, IEEE International netic pull and its effects on vibration in a three-phase
Symposium on Electric Power Engineering, generator with eccentric rotor. J. Sound Vibration,
Stockholm, Sweden, 18–22 June 1995, Volume on July 2002, 254(2), 297–312.
Electrical Machines and Drives pp. 229–234. 22 Werner, U. and Binder A. Rotor dynamic analysis of
8 Joksimovic, G. M. Dynamic simulation of cage asynchronous machines including the Finite-Element-
induction machine with air gap eccentricity. IEE Method for engineering low vibration motors. In
Proc.—Electric Power Appli., July 2005, 152(4), International Symposium on Power Electronics,
803–811. Electrical Drives, Automation and Motion, SPEEDAM
9 Burakov, A. and Arkkio, A. Comparison of the 2006, Taormina, 23–26 May 2006, pp. 88–96.
unbalanced magnetic pull mitigation by the parallel 23 Porjus Hydropower Centre, Porjus, Jokkmokk, Sweden.
paths in the stator and rotor windings. IEEE Trans. 24 Chen, J. W. and Gunter, E. J. Introduction to
Mag., 2007, 43(12), 4083–4088. dynamics of rotor-bearing systems, 2007 (Trafford
10 Ohishi, H., Sakabe, S., Tsumagari, K., and Publishing, Nictaria, BC).
Yamashita, K. Radial magnetic pull in salient pole 25 Lundström, L., Gustavsson, R., Aidanpää, J.-O.,
machines with eccentric rotors. IEEE Trans. Energy Dahlbäck, N., and Leijon, M. Influence on the sta-
Conversion, 1987, EC-2(3), 439–443. bility of generator rotors due to radial and tangential
11 Holopainen, T. P. Electromechanical interaction in magnetic pull force. IET Electric Power Applications,
rotordynamics of cage induction motors. PhD Thesis, 2007, 1(1), 1–8.
2004 (VTT Technical Research Centre of Finland, 26 Division for Electricity, The Ångström Laboratory,
Finland). Uppsala Universitet, Sweden.
12 Burakov, A. Modelling the unbalanced magnetic pull 27 Karlsson, M. Modelling and analysis of multiphysical
in eccentric-rotor electrical machines with parallel interactions in hydropower rotor systems. PhD
windings. PhD Thesis, 2007 (Department of Thesis, Luleå Tekniska Universitet, Sweden, 2008.

Proc. IMechE Vol. 225 Part C: J. Mechanical Engineering Science


2088 Y Calleecharan and J-O Aidanpää

28 DATAPLOT. National Institute of Standards and It is to be noted that the constant terms B0r and B0t
Technology, Technology Administration, U.S. occurring in the denominators of FrUMP and FtUMP ,
Department of Commerce. Version 5/02/06. respectively (equation (12)) were set to unity in the
29 Gerald, C. F. and Wheatley, P. O. Applied numerical
least-squares fit estimation.
analysis, fifth edition, 1994 (Addison Wesley,
Reading, MA).
30 ISO/IEC 9899:1990 Programming languages—C. Notation
31 Microsoft visual studio 2010, Microsoft Corporation.
a air-gap length (m)
32 Sprott, J. C. Chaos and time-series analysis,
5th edition, 2003 (Oxford University Press, c damping (N s/m)
Oxford, USA). C centre
33 Kim C.-M., Hwang D.-U. and Park, Y.-J. E Young’s modulus of elasticity (Pa)
Stabilization of quasiperiodic output in a q-switched F force (N)
Nd:YAG laser. In Proceedings of the 2nd FIR finite impulse response
International Conference on control of oscillations I area moment of inertia (m4)
and Chaos, St Petersburg, Russia, vol. 3, 2000, k stiffness (N/m)
pp. 431–434.
K unbalanced magnetic pull stiffness (N/m)
34 Carroll, T. L. and Pecora, L. M. Nonlinear dynamics
in circuits, 1996 (World Scientific Publishing L rotor length (m)
Company, New Jersey, USA). m rotor mass (kg)
35 Wagner, C. F. Damper windings for water-wheel r radial rotor–stator eccentricity (m)
generators. AIEE Trans., 1931, 50, 140–151. UMP unbalanced magnetic pull
 damping ratio
 angular displacement (rad)
APPENDIX ! angular velocity (rad/s)
Subscripts
Coefficients of the rational UMP b bearing
components fo forcing
m mean
Radial UMP component n natural frequency
r Polar radial direction
The coefficients of equation (12a) to five decimal ro rotor
places are as follows: st stator
A0r ¼ 50 592:183 59 t tangential direction
wh whirl
A1r ¼ 94 379:109 38
whr whirl ratio
A2r ¼ 49 365:472 66 x Cartesian x-direction
B0r ¼ 1 y Cartesian y-direction
B1r ¼ 1:976 09
B2r ¼ 1:013 78 ðA:1Þ

Tangential UMP component

The coefficients of equation (12b) to five decimal


places are given next:
A0t ¼ 18 651:650 39
A1t ¼ 18 670:914 06
B0t ¼ 1
B1t ¼ 1:937 78
B2t ¼ 0:971 65 ðA:2Þ

Proc. IMechE Vol. 225 Part C: J. Mechanical Engineering Science

You might also like