Professional Documents
Culture Documents
Presented at the International Gas Turbine and Aeroengine Congress & Exhibition
Birmingham, UK June 10-13, 1996
This paper has been accepted for publication in the Transactions of the ASME
Discussion of it will be accepted at ASME Headquarters until September 30, 1996
Downloaded From: https://proceedings.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/conferences/asmep/82210/ on 02/05/2017 Terms of Use: http://www.asme.org/abo
analysis for a compressor by Horlodc and (1981).
Subscripts A radially averaged (2-0), inviscid,
0- near upstream of the actuator incompressible flow was assumed. The
disc in the radius scale connecting conditions at the disc were provided
analysis by the equations of continuity, momentum, and
0+ near downstream of an empirical correlation for turbine efficiency.
the actuator disc in the The flow redistribution was shown to increase the
radius scale analysis magnitude of the lateral forces above what would
1-9 Casing instrument be predicted on the basis of efficiency variation
stations at x/c=-3.56, - alone. However, the assumption of perfect flow
2.38, -2.08, -021, 0.06, guidance by the blades ignored the effects of the
0.029, 0.53, 1.42 and 2.84 tip gap flow , and, therefore, the predictions were
respectively unsatisfactory.
design value This paper presents a new radius-scale
indicates a force due to analysis which uses the results from the blade
action of nonuniform scale analysis as its coupling conditions. Thus,
pressures this model does not resort to empiricism in
vsd indicates a force due to accounting the bcaled effects (e.g. losses)
tangential variation of associated with the tip clearance flow. A
work extraction comparison of the model predictions with the
experimental data is also given.
Superscripts
non-axisymmetric 2. Analytical Model
perturbation As in the blade scale analysis, the stage
azimuthal is collapsed into an actuator disc (Figure 1). x is
mean, or axisymmetric the through-flow direction, y is the azimuthal
value direction, and z is the radial direction. The axial
A complex amplitude locations of stationsreferred to in the analysis
are as follows. The "actuator disc" consists of a
1. Introduction turbine stage collapsed into a plane at x=0. The
The problem of self-excited vibration in axial stations far upstream and far downstream of
rotor systems, or rotordynamic instability, has the actuator disc, on the radius scale, are
plagued development of various turbomachines. referred to as 03 and 03, respectively. The
The rotordynamic instability caused by stations near upstream and near downstream of
asymmetric tip clearance in turbines has been the actuator disc, on the radius scale, are
the focus of this investigation. The experimental referred to as 0- and 0+, respectively (Figure 4 in
results were presented in Martinez-Sanchez et the companion paper). Although the actuator
al. (1993) and an analytical model for the blade disc is assumed to have zero axial thickness on
scale effects of a finite rotor tip clearance is the radius scale, it actually consists of a full span
presented in a companion paper. stator row and a partial span rotor row as shown
In a turbine with an eccentric rotor, the in the figure.
efficiency is higher in the region with the smaller To model a turbine whirling at some
tip clearance, resulting in a higher local torque. amplitude, e, much smaller than the blade span,
Upon azimuthal integration, the torque Hb , a linear perturbation approximation is used.
asymmetry leads to a force perpendicular to the Thus, the flow variables are assumed to contain a
rotor eccentricity which feeds energy into the small perturbation about the mean, which in this
forward whirling mode (Figure 1 in the companion case is the centered turbine case. The solution
paper). The implicit assumption was that the flow procedure is based on previous efforts by Leung
remains perfectly uniform upstream of the (1991) and Yoo (1993).
eccentric turbine, as'well as downstream of it. In The assumptions of this model are
reality, the presence of an eccentric rotor would equivalent to those found in the blade scale
induce an azimuthal flow redistribution, causing a model, with the obvious exception of the
force variation even without the efficiency relaxation of the axisymmetric flow assumption to
variation. facilitate the analysis of a whirling turbine. Other
Oiu (1985) developed an actuator disc newly incorporated effects include the following:
analysis to examine the azimuthal distribution First, a non-axisymmetric, mass storage effect of
effect on the lateral forces in a turbine with a the axial gap between the stator and the rotor
whirling rotor. This was an adaptation of an was incorporated. Second, the non-
c y )= VO ;V2=0 (3)
0.11 + +
cx (0 ,y,t) = Re[Exo + Koe
(4) + (E v)e + vp o (10)
at
0 0 O 0 a + +
A53
O 0 0 A 0 YR ciA = t cy (0-. y))
64 (15)
U U H
c
x4
+
c 4
1 po in the last equation in Eq. (19) is time relate ear , &I , 42,, and /32 to gap variations,
dependent and is given by Eq. (8).
DR
(23)
and across the (quasi-steady, ideal) stator, Notice that in the limit d/R ) O.
= Pn. i.e., 1/2, /47 and
exlies0
E tana = For finite stator-
y2r11 2 y2s
LOAE + +E o = a- +
x2sz1 y2s y2s rotor axial gap, d, this formulation introduces
P differences between stator exit and rotor inlet
2 1 velocity components.
exO( 1 4- tan c t 2 )e xl
2.5 Calculation of Rotordynamic
Or Coefficients
The excitation force coefficients can be
obtained from the perturbation quantities.
112 = 2 . nR - Defining the tangential force exerted on the
cs0 1+ tall - cx1 (24)
co turbine per azimuthal length, fy , as the sum of
x )
forces by the underturned flow and the bladed
Using Eqs. (21), (22) and (24), the flow quantities flow,
can be solved for in terms of el H and one of the
unknowns. Since 42
, = pHe2 , connects 1y = 42r (52r c'y+3) +0
directly with the x-z analysis of a centered rotor,
(28)
q2 , is chosen as this remaining parameter. The
results are: Then the azimuthal mean and the perturbation of
fy are , respectively
+ f Olaf2 ! +
=
H
(25) 1Y 42r( ry2r c3) + 0- 472r(52.
d 2 (29)
x 1 i()W + ( 7? ) fi
1; = 42r(F y2 r 51-3)
fr2 xl
A- (26)
cy' 22. c;+3 (30)
/x43 exo
it(13) RD 0.208
(a + ia (33) 1.508
k "Mal lifo
0.580
-i)
Eq. (33) summarizes the results of the theory, I71 1.508
and its linear nature makes it straightforward in
application. Parametric variations and specific
MN 0.000
Figure 10. AU of the coefficients decrease as decreases almost linearly. Consequently, the
TIN is increased; however, the direct force azimuthal pressure gradient dpi dy is reduced,
coefficients are more sensitive to FIR than the and layd is decreased. Thus, the damping
cross force coefficients, and, as will be shown,
both trends are weaker than those seen in the effect is seen to be due to the kinematics of the
data. rotor whirling motion.
Figure 11 shows the calculated effects The slope of cross excitation force
of unsteady whirring motion on the total excitation coefficient is an indication of direct damping,
force coefficients. The f2RIU range of -1.0 to which along with cross stiffness, determines the
stability of the rotor system. Figures 11 and 12
1.0 spans the frequency ratios from backward
make clear that the damping effect is provided
synchronous whirl to the forward synchronous primarily by the pressure force. The work
whirl. The restoring nature of the direct force and extraction forces actually show negative
damping for MN less than +0.3, but the cross
force due to workextraction variation remains a
2 This "mass storage" effect was originally destabilizing force for forward whirl.
described by Millsaps (1992) in relation to the
lateral forces which arise due to whirling motion
of labyrinth seals.
Table 3: Comparison of cross force predictions from existing correlations and the data from Configuration 4
[Ref. 131 at IP =0.58.
mechanism was assumed to our 180 ' out of
This section presents a comparison of phase with the clearance distribution. Thus, zero
the experimental data and the analytical direct force was predicted.
predictions. First, the experimental data are
compared against predictions from the pre- Author aX (roxkl) aIcs (meanue ) aXmrAdaseanwed
existing models. Then, the newly predicted and
measured magnitudes and the composition of
forces are compared. h is to be noted that no Afford 0.0 - 0.6 - 3.4
data exist for dynamic (whirling) conditions. In Traupel 0.0 - 0.6 - 3.4
view of the significant predicted damping, this Farokhi 0.0 - 0.6 - 3.4
should be a priority for further testing. Dunham 8 0.0 - 0.6 - 3.4
The experimental data are compared Came _
against predictions from a few widely used
correlations and models for efficiency loss due to
Table 4: Comparison of direct force predictions
clearance effects, which were available before
the development of the current model. The from existing correlations and the data from
implied assumption is that the work loss and the Configuration 4 (Song, 1995) at = 0.58.
efficiency loss are equal. This is not accurate
because, as shown by the new model, the Next, the same experimental data are
pressure drop (a factor in the efficiency compared to the predictions from the new model.
definition)is also strongly affected by clearance Table 5 shows a comparison of the predictions
changes. The correlations of Alford (1965), and the dynamometer measurements. a tow
Traupel (1958), Farokhi (1988), and Dunham 8 represents the sum of the pressure and blade
Came 3 (1970) were selected. The equations for force contributions.
the correlations are listed in Song [13]. Table 3
presents a comparison of the cross excitation
Direction ,(model) total (measured)
force coefficient. While the prediction of Dunham
and Came's correlation is better than others, it X - 2.7 - 3.4
misses the pressure contribution, and because +3.2 +3.5
of this, the experimental value is still twice the
predicted value. Thus, Table 3 shows the poor Table 5: Comparison of the predicted and the
state of existing predictive capability. measured total excitation force coefficients from
Table 4 shows the comparison of direct
forces. The azimuthal redistribution of flow due Configuration 4 (Song, 1995) at = 0.58.
to a rotor offset was previously neither
recognized nor modeled, and the work defect While the model still underpredicts excitation
coefficients, the predictions of both cross and
direct excitation force coefficients are within 20%
3 This correlation is an adaptation of the of the measured values
correlation originally developed by Ainley au The new model 's prediction of the total
Mathieson (1951) As noted by Demon (1993) force in each direction consists of those due to
this correlation has been widely used by the non-axisymmetric pressure and work extraction.
engine manufacturers.
10
Table 6: Comparison of the predicted and the measured rotordynamic coefficients (from the aerodynamic
data) from Configuration 4 (Song, 1995) at 0 = 0.58.
11
Thus, the main conclusions from this Dunham, J., and.- Came,. P.M.,
investigations include the following: 'Improvements to the Ainley-Mathieson
Method of Turbine Performance
1)An actuator disc based approach can Prediction', ASME Paper 70 GT 2, - -
12
Traupel, W., Therm ische Figure 1: A radius scale view of the actuator disc
lushomakckumen. Vol. 1, Springer-
Verlag, Berlin, 1958. (Yoo, 1993)
Variationfltfachints. Vol.
197c, 1983, pp. 173-178.
13
f; A
thU(e/M2
-6
Figure 3: Tangential cross-sectional flow area. Ay. 0 90 180 270 360
between the stator and the rotor. Figure 4: Predicted nondimensional tangential force
perturbation vs. azimuthal angle (maximum gap at 180).
6
. - , i1 .. 6
c ./u(a1H) .
4
.
. I -
. - - (11,)/oui(eni) 4 :
.
I
2 . i . -" "
2 . I Cirtotal)
Ir. ...', .. ... ! ...
.... .. ..7 ....;I
0 ...
, .. . . . . .. .
.- I ......... I ........................ 1 . c.11.12)....:
-- . . 0
i I
I I
.. ......... I ..
- .... t.--
2 I
.
-"..-. .--.
2
. .. 1 aXacaD:
I 1. I .
i -4
. I1 I1 I I :
6
-6
0 90 180 270 360
04 0.5 0.6 0.7 0.8 0.9
Figure 5: Predicted perturbations in the upstream SD
axial velocity (station 1 on the blade scale) and Figure 6: Predicted total and pressure excitation
the rotor region pressure (average of inlet and exit)
vs. azimthal angle (maximum gap at 180). force coefficients vs. design flow coefficient
6 6
4 4
2 2
. .
I
... .......... ..
... *: ...
. 1 I
0 1
a lto ) ; i: i sY(p)
X(p) -
-2 . I lux(r4
i
%ran:
4 4
, . .
I I
0.5 0.6 0.7 0.8 Os 0.5 2 23 3 33
0.4
13
Figure 7: Predicted total and pressure excitation Predicted total and pressure excitation
Figure 8:
force coefficients vs. mean flow coefficient force coefficients vs. mean work coefficient
14
2 .,......- .. .
.
,
,
2. . ...... _
! I
..... -- t
ir
,
-
0 0 ; aY(14
- .... t .. I 1 aX(p)
-2 : .... .........-- .
.....
. ag 1
- 4
1
I I anon!)
,
' .
4 1 I
0.05 0.1 0.15 0.2
0 0.01 0.02 0.03 0.04 0.05
dat
VET
Figure 9: Predicted total and pressure excitation Figure 10: Predicted total and pressure excitation
force coefficients vs. interblade distance. force coefficients vs. mean radial tip gap.
6 6
4 4
' aNIP)
czY(toor) ..... .......
...
2 2
aY(wd)
0
2 -2
: ago
4
ax()
;
4 -6
-OS 0 0.5 -03 0 0.5 1
MVO fllt/U
Figure 11: Predicted total excitation force Figure 12: Predicted pressure and tangential excitation
coefficients vs. whirling frequency. force coefficients vs. whirling frequency.
Tkeay 7bemy
Expeiwaau Experiment
P:
PU2
0.04 0.04
0.02 0.02
0
-OM
-0.04 -0.04
-0.06 .0.06
0 90 180 270 360 0 90 180 210 360
Figure 13: Measured and predicted tangential force Figure 14: Measured and predicted rotor inlet
perturbation vs. azimuthal angle for pressure perturbation vs. azimuthal angle for
Configuration 4 at if) = 0.58 & embso.oi (maximum gap at 1801. Configuration 4 at 4,) = 0.58 &
(maximum gap at 180).
15
6 .
4
1
2
1
0
-2
1.
.:
I
P I
4. 1
. .
4
0.4 0.7
OS 0.8 03
0.6 .
ti
Figure 15: Measured and predicted total force excitation coefficients vs. mean flow coefficient for
Configuration 4.
& Experiment
aX(total) neer/
ty(toad) them & Experiment
I
4
0 0.02 0.04 0.06 0.08 - 0.1
Wa
Figure 16: Measured and predicted total force excitation coefficients vs. interblade distance
(Configurations 1 & 2 at = 0.58).
aX(extd) Thttry & Experiment
chtr(mul) Theory at Exeeriman
4
i 1
2 i
. I
0
.
i
I I
. 1 .
2
.
.
. 1
.
I I
1
I '
4
0 0.01 O. 0.03 0.04 0.05
ifit
Figure 17: Measured and predicted total force excitation coefficients vs. mean radial tip gap
(Configurations 2 & 4 at ot) = 0.58).
16