Computers & Fluids 88 (2013) 11–18

Contents lists available at ScienceDirect

Computers & Fluids

j o u r n a l h o m e p a g e : w w w . e l s e v i e r . c o m / l o c a t e / c o m p fl u i d

Numerical simulation and rotor dynamic stability analysis on a large

hydraulic turbine
Ru-Zhi Gong, Hong-Jie Wang ⇑, Quan-Zhong Liu, Ling-Feng Shu, Feng-Chen Li
School of Energy Science and Engineering, Harbin Institute of Technology, Harbin 150001, China

a r t i c l e i n f o a b s t r a c t

Article history: The rotor dynamic stability, which is a vital problem in large hydraulic turbine, is usually not taken into
Received 18 November 2011 consideration when a hydraulic turbine is designed. But the rotor dynamic character will influence the
Received in revised form 23 February 2013 performance (efficiency, noise, vibration, etc.), life-span and even the safety of the turbine. The method
Accepted 30 August 2013
for analyzing the rotor stability of hydraulic turbine has been studied in this paper. Stability of the rotor
Available online 7 September 2013
system of a real turbine was studied with the theory of rotor dynamics in this paper. The way how each
factor affects the dynamic character of labyrinth system was also investigated. It can be concluded that
Keywords:
the research methods can be widely used in the dynamic analysis of the rotor system with gap flow.
Fluid machinery
Rotor dynamics
Ó 2013 Elsevier Ltd. All rights reserved.
Stability analysis
Hydraulic turbine

1. Introduction the equation obtained, dynamic performance of the turbine rotor

It has been discussed a lot about the application of computa-
tional fluid dynamics (CFD) method on the flow simulation of
hydraulic turbines in the published literatures so far. Iwatsubo
[1] evaluated the instability forces of labyrinth seals in turbines
in 1980. Diewald and Nordmann [2] analyzed the fluid-mechanical
interaction in fluid machinery. Nilsson and Davidson [3] compared
the computational and the experimental results in a Francis tur-
bine and proved that CFD method is reasonable to simulate the
flow field in fluid machinery. And there are other works that stud-
ied the application of CFD method in the detection of behaviors of
hydroturbines [4,5]. There are also works studying the stability of
rotor systems. Alford [6] is the first person who found self-exited
whirl in the turbo machinery in 1965. Hashimoto et al. [7,8] exam-
ined the effects of wear on steady-state and dynamic character of
the theoretical and experimental methods under operating condi-
tions including turbulence. Ren et al. [9] analyzed dynamic stabil-
ity of a rotor system with labyrinth seals. However, there was few
works related to rotor dynamic stability of hydraulic turbines.
In the present paper, the flow field of the whole passage in the
turbine is at first simulated and forces on the journal are then ob-
tained through the simulation of the flow in the labyrinth seals
clearance. The Muszyska model for rotor system analysis is
adopted to describe the flow induced force in the labyrinth seal.
The parameters in rotor dynamics equation for the rotor system
of the turbine is fitted with the least square fitting method. With
system is analyzed and NewMark direct integral method is used.
To detect how the structure of the labyrinth seals affect the
dynamics of the turbine rotor system, the length, thickness and
pressure drop of the labyrinth clearance are changed for the
discussions.
2. Case description
For the present study, a real hydraulic turbine for BaiHeTan
hydropower station in China is adopted. The detailed geometrical
parameters of the turbine are provided by Harbin Institute of Large
Electrical Machinery, China. The 3D turbine for flow simulation is
shown in Fig. 1. The turbine has 15 runner blades, 23 stay vanes
including a special blade and 24 guide vanes with a runner diam-
eter of Dref = 8700 mm (runner radius Rref = 4350 m). The designed
head of the turbine is 120 m, the power is 1000 MW. And the rota-
tional speed at designed condition is 74.5 rpm. Table 1 shows the
geometrical parameters of the turbine. It is unusual to have laby-
rinth seal in hydraulic turbine; however, labyrinth seal is adopted
in the selected turbine. The structure of the turbine is shown in
Fig. 2 and the labyrinth seal is marked with red1 circle in the figure.
3. Numerical simulation considerations
The flow path in the turbine is from the spiral case inlet to the
draft tube outlet. The water was considered to be incompressible. A

⇑ Corresponding author. Tel.: +86 13836034807. 1

For interpretation of color in Fig. 2, the reader is referred to the web version of
E-mail address: Wanghongjie@hit.edu.cn (H.-J. Wang). this article.

0045-7930/$ - see front matter Ó 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.compfluid.2013.08.014
12 R.-Z. Gong et al. / Computers & Fluids 88 (2013) 11–18

Fig. 1. Model of the hydraulic turbine.

Fig. 3. Grid validation result.

Table 1 governing equation, RNG k–e turbulence model provides two addi-
Parameters of the hydraulic turbine studied presently. tional equations:
Parameters Value  
@ðqkÞ @ðqui kÞ @ @k
Diameter of the runner, D1 (mm) 8700 þ ¼ ak ðl þ lt Þ þ qðpr  eÞ ð3Þ
Amount of blades, Z 15
@t @xi @xj @xj
Distribution radius of the guide vanes, D0 (mm) 1.145D1 = 9960
Amount of guide vanes, Z0 24
 
@ðqeÞ @ðqui eÞ @ @e q 
Height of guide vanes, B0 (mm) 1588 þ ¼ ae ðl þ lt Þ þ C 1e epr  C 2e e2 ð4Þ
Amount of stay vanes, Zs 23 @t @xi @xj @xj k

where,
2
k
lt ¼ qC u ð5Þ
e
ak, ae , Ce1 and Ce2 are constants and can be determined by experi-
ments and statistical theory. These constants are herein chosen as
ak = ae = 1.39, Ce1 = 1.42 and Ce2 = 1.68.
The mesh for simulation was generated by ICEM grid generator.
And grid independence is validated, as shown in Fig. 3. The whole
path was discretized with an unstructured hybrid mesh of tetrahe-
dron cells. Normally, the fluid flow simulation needs finer grid than
structure analysis. So the mesh size was determined according to
the recommendation by Ma and Zhou [10] to meet the require-
Fig. 2. Structure of the labyrinth seal.
ment for fluid flow simulation. It is acceptable when there is no
evident change with the increasing number of grid nodes. And
time-dependent Reynolds average Navier–Stokes (RANS) model the final mesh has 4.05  106 nodes.
was chosen for the governing equations of the flow and numerical The interaction between the rotor and stator is simulated with
simulation was performed in order to obtain the flow field in the interface. The kind of interface between the rotor and stator is slid-
turbine flow path. The continuity equation and the momentum ing surface. And the hydraulic conversation across the interface is
equations for the flow in RANS-model style are as follows: just about 0.0002%. It can be completely neglected. Multi Reference
Frame (MRF) model is used for the part of runner. The pressure
@ui drop between the inlet and the outlet of the seal is obtained from
¼0 ð1Þ the results of the flow field of the turbine, which is about
@xi
55,000 Pa.
    Then the flow field in the seal is simulated with the rotational
@ui @ui 1 @ @ui @uj 2 @uk speed of the shaft and pressure drop obtained as boundary condi-
þ uj ¼ pdij þ l þ  dij þ ðqu0i u0j Þ
@t @xj q @xj @xj @xi 3 @xk tions. In the simulation, different rotating speed of the shaft and
ð2Þ different shaft offset are taken into consideration. The pressure dis-
tributions in the labyrinth seal are acquired. Fig. 4 shows the pres-
The flow in hydraulic turbine and the labyrinth seal is very complex. sure distribution profile along the circumference under the same
It is impossible to find accurate or approximate analytical solutions. rotational speed but different shaft offset. And the eccentricity
Herein, the above equations are solved by the commercial CFD code direction is X coordinate, which means circumference angle is zero
FLUENT (Version 12.1.2). (or 2p).
The streamlines in the turbine have large curvature. Fortu- When the offset of the shaft is different, the pressure distribu-
nately, the re-normalization group (RNG) k–e turbulence model tion is different from each other. Because of the eccentricity, the
in FLUENT simulates rotational and swirl flow in the average flow pressure is higher on the side of the shaft deviation, and there is
better through improved calculation of the turbulent viscosity. The some phase difference, always before the eccentricity direction.
influence of turbulence in the hydraulic turbine is thus considered The pressure inhomogeneity is increasing with the increase of
with RNG k–e turbulence model in the present study. To close the the offset (xe) of the shaft.
 R.-Z. Gong et al. / Computers & Fluids 88 (2013) 11–18 13

          
46000 Fx kxx kxy x cxx cxy x_ mxx €x
 ¼ þ þ
Fy kyx kyy y cyx cyy y_ myy y€
44000 xe=0.5mm and ye=0
ð6Þ
42000
xe=1.0mm and ye=0
xe=1.5mm and ye=0 where Fx and Fy stand for sealing forces in the X and Y directions,
40000 xe=2.0mm and ye=0 respectively; x and y stand for displacements in the horizontal
Presssure (Pa)

xe=2.5mm and ye=0 and vertical directions, respectively; kxx, kyy, kxy and kyx stand for
38000
seal stiffness, respectively; cxx, cyy, cxy and cyx stand for seal damp-
36000 ing; mxx and myy stand for inertia coefficient. xand _ y_ stand for the
first derivatives of (x, y), €x and y€ stand for the second derivatives
34000 of (x, y).
Dynamic factors of the system are all constant. However, when
32000
runner rotates as large-disturbance motion, self-excited vibration
eccentricity direction
30000 will be likely to produce. In this situation, sealing force will be
nonlinear.
28000
0 π /2 π 3π/2 2π Muszynska [12] thought that action force of fluid rotated as an
average angular velocity sx, where, s stands for average flow
Circumference angle (rad)
velocity, which is the key parameter of fluids dynamic behavior.
Fig. 4. Pressure distribution under different journal offset. xe and ye stand for the X And Muszynska prompted the frequently-used model for fluid in-
and Y coordinate (as shown in Fig. 1 for the coordinate system) eccentricities, duced force. Muszynska model reflects the nonlinear character of
respectively. fluid excitation force. Its sealing force equation is as follows:
  " #    
Fx K  mf s2 x2 sxD x D 2smf x x_
 ¼ þ
Fy sxD K  mf s2 x2 y 2smf x D y_
After the simulation of flow field in the turbine and the flow   
field in the seal, forces on the shaft are read from the FLUENT sol- mf 0 €x
þ ð7Þ
ver. Conditions under varies shaft eccentricity were simulated. And 0 mf y €
the computed rotational speed is from 35 rpm to 150 rpm. Table 2
where x and y are rotor orthogonal lateral deflection coordinates in
shows part of the data.
the selected coordinate system X and Y; Fx and Fy stand for fluid
When the rotating runner has small disturbance in the center of
force components acting on the rotor; x is the angular velocity of
equilibrium position, the sealing force is linear. It has lateral sym-
rotation; s is the fluid circumferential average velocity ratio; K, D
metry, and can be written as [11]:
and mf respectively express equivalent stiffness of sealing force,
equivalent damping and equivalent mass, respectively. K, D and s
are proved to be nonlinear functions of perturbation motion dis-
placement x and y, of which the expression is:
Table 2
The force on the shaft under different journal offset. n n 1
K ¼ K 0 ð1  e2 Þ ; D ¼ D0 ð1  e2 Þ ; n¼ 3 ð8Þ
Eccentricity in Eccentricity in Rotational Forces in the Forces in 2
the X direction, the Y direction, speed, nr X direction, the Y
x (mm) y (mm) (rpm) Fx (N) direction, Fy s ¼ s0 ð1  eÞb ; 0 < b < 1 ð9Þ
(N)
0.5 0 65 3560.78 1614.48 where e stands for the eccentricity
0.5 0 70 3322.77 1843.34 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
0.5 0 75 2970.27 1961.57 e¼ x2 þ y 2
0.5 0 80 2561.02 1950.76
0.5 0 85 2178.46 1854.52 and other parameters in above equations are:
1.5 0 65 10657.64 4827.40
1.5 0 70 9834.24 5305.75 K 0 ¼ l3 l0 ; D0 ¼ l1 l3 T; mf ¼ l2 l3 T ð10Þ
1.5 0 75 8787.94 5546.93
1.5 0 80 7651.43 5585.21 in which:
1.5 0 85 6555.08 5405.71   
0.5 1.0 65 6739.60 5476.12 2r2 2r2 E B 1
l0 ¼ Eð1  m0 Þ; l1 ¼ þ þE
0.5 1.0 70 7015.70 4871.09 1 þ n þ 2r 1 þ n þ 2r r 2 6
0.5 1.0 75 6951.12 4107.12
0.5 1.0 80 6603.79 3311.73  
0.5 1.0 85 6072.83 2561.52 1 r p R DP
l2 ¼ þE ; l3 ¼ ;
1.5 2.0 65 15016.41 8369.51 1 þ n þ 2r 6 k
1.5 2.0 70 "
15304.11 7470.69
 2 # 2 0
1þm

1.5 2.0 75 15154.49 6388.75 Rv

m0
1.5 2.0 80 14422.67 5130.57 k ¼ n0 Ra 1 þ ð11Þ
1.5 2.0 85 13417.26 3953.76
Ra
2.0 1.5 65 16667.47 3447.26
2.0 1.5 70 16699.72 2527.09 and
2.0 1.5 75 16257.63 1588.99
2
2.0 1.5 80 15213.64 654.07 Rv
Ra
 m0 1þn R xd 2v d
2.0 1.5 85 13918.76 146.83 B ¼ 2 
2 ; E¼ ; Rv ¼ ; Ra ¼ ð12Þ
2.5 2.0 65 Rv 2ð1 þ n þ 2rÞ t t
19775.14 4621.18
Ra
þ1
2.5 2.0 70 19465.18 3589.95
2.5 2.0 75 18724.29 2609.30 Least Square Method is used to fit the rotor-dynamics equations
2.5 2.0 80 17688.73 1642.85
with the data obtained in Table 2, resulting in: n = 9.6379,
2.5 2.0 85 16233.27 694.34
m0 = 0.1383, n = 0.7378, n0 = 0.0290, s = 0.2525 and b = 0.4143.
14 R.-Z. Gong et al. / Computers & Fluids 88 (2013) 11–18

where mD stands for mass of the rotor; DR stands for damping of the
rotor; KR stands for stiffness of the rotor and x stands for rotational
speed of the rotor.
For a rotor system, if dividing the shaft into N  1 sections, then
it contains N nodes. For the finite element model, there are 4
generalized displacement coordinates. Sum the motion equation
of each node, the differential equation for the rotor system contain-
ing 4N freedoms is obtained, which reads:

€ þ ½DfZ_ þ ½KfZg ¼ fFðZ; Z;

½MfZg € Z;
€ XÞg þ X2 fPðtÞg ð14Þ

where [M] stands for mass matrix of the rotor system; [D] stands for
damping matrix of the rotor system (in order to simplify the calcu-
lation, Rayleigh damping matrix is applied); [K] stands for stiffness
matrix of the system; fFðZ; Z;_ Z;
€ XÞg stands for flow induced force;
X2{P(t)} stands for unbalance motivational force.
The finite element model of hydraulic turbine rotor system is
shown in Fig. 5. The shaft is divided into 13 sections (14 nodes)
and the flow induced force described by Muszyska model shown
Fig. 5. Finite element model of hydraulic turbine rotor system. above is added to the part of rotor (node 14). Parameters of each
shaft section are listed in Table 3. The stiffness matrix and mass
matrix of the shaft are computed with a program using finite ele-
The following equation describes the dynamic character of the ment method. The bearings are important factors which influence
rotor system: the character of the system. The main bearings in the system are
         the upper guide bearing and turbine guide bearing. The bearing
KR x DR x_ mD €x
þ þ support stiffness is 1  108 as designed. Then NewMark direct inte-
KR y DR y_ mD y €
    " # gral method is used to analyze the character of the rotor system.
Fx 0 mex2 cosðxtÞ
¼ þ þ ð13Þ Fig. 6 is the response of the turbine rotor system. At low rota-
Fy mg mex2 sinðxtÞ tional speeds, the only lateral vibration component in the spectrum
is a small synchronous vibration component (1), due to some
unbalance in the rotor. The rotor unbalance acts as a centrifugal
rotating force, exciting rotor 1 responses. At low rotational
Table 3
Parameters of each shaft element.
speeds, these synchronous vibrations are stable. At a higher
rotational speed, which is known as the threshold of instability
No. of Section Inside Outside Section for the considered rotor/bearing system, a new, relatively
section length, l diameter, d diameter, D mass, m (kg)
high-amplitude, sub-synchronous vibration component appears
(mm) (mm) (mm)
in the spectrum (at about 265 rpm). At that speed, the previous ro-
1 1000 1350 1650 5513.5
tor regime becomes unstable, and the overall lateral vibration
2 1180 1350 1650 6505.9
3 1300 4000 4500 33846.7 amplitudes begin to increase.
4 1390 4000 4500 36189.9
5 1100 2650 3100 17436.4
6 1000 2650 3100 15851.3 4. Discussions
7 1000 2650 3100 15851.3
8 1500 3745 4000 18148.3 The flow field of the turbine is simulated and the stability of the
9 2030 3745 4000 24560.8
turbine rotor system is then calculated following the above proce-
10 1000 3745 4000 12098.9
11 1570 3745 4000 18995.3
dures. To analyze the factors influencing the stability character of
12 2100 3750 4000 24925.6 the rotor system, the structural parameters and the pressure drop
13 2100 3750 4000 24925.6 between the gap inlet and the gap outlet are analyzed to detect the
influencing mechanism.

Fig. 6. Dynamic character of the rotor system.

 R.-Z. Gong et al. / Computers & Fluids 88 (2013) 11–18 15

Fig. 7. The response of the system under different labyrinth length. Labyrinth width W = 0.6 mm and pressure drop P = 55,000 Pa.

Fig. 8. The relationship between seal length and threshold speed.

4.1. Length change of the labyrinth seal
The length of the seal in the model is changed to study the
influence on the stability of the rotor system. Fig. 7 shows the
responses of the system under different labyrinth length. The rotor
system has a lower instability threshold when the labyrinth seal is
longer.
Instability threshold under different labyrinth lengths are
plotted in Fig. 8 with the labyrinth length as the X coordinate
and instability as the Y coordinate. It is shown that the instability
threshold decreases when the labyrinth seal length increases.
4.2. Width change of the labyrinth seal
The width is an important parameter in the labyrinth seal and it
will influence the flow induced forces that act on the shaft. The
width of the labyrinth seal clearance in the model is changed to
study the influence on the stability of the hydraulic turbine rotor
16 R.-Z. Gong et al. / Computers & Fluids 88 (2013) 11–18

Fig. 9. The response of the rotor system under different labyrinth width. Labyrinth length L = 0.36 m and pressure drop P = 55,000 Pa.

Fig. 10. The relationship between seal width and threshold speed.

system. Fig. 9 shows the responses of the system under different
labyrinth seal width.
Instability threshold under different labyrinth seal widths are
plotted in Fig. 10 with the labyrinth seal width as the X coordinate
and instability as the Y coordinate. It can be seen that the rotor sys-
tem has a lower instability threshold when the labyrinth seal
width is larger. In other words, the stability of the rotor system will
become worse when the labyrinth seal width increases.
4.3. Change of the pressure-drop between the inlet and the outlet of
the labyrinth seal
When the rotor dynamic equations of the turbine rotor system
are created, the pressure drop in the clearance is used. With the
precondition that we assume the pressure drop will affect the rotor
dynamic behavior of the system, the pressure drop in the clearance
is changed to study the influence on the stability of the turbine ro-
 R.-Z. Gong et al. / Computers & Fluids 88 (2013) 11–18 17

Fig. 11. The response of the rotor system under different pressure drops. Labyrinth length L = 0.36 m and labyrinth width W = 6 mm.

Fig. 12. The effect of the pressure drop under different labyrinth seal width and length.

tor system. Fig. 11 shows the responses of the system under
different pressure drops. It can be seen that the rotor system has
a lower instability threshold when the pressure drop is low.
Instability threshold under different pressure drops are plotted
in Fig. 12 with the labyrinth seal width as the X coordinate and
instability as the Y coordinate. It shows that the instability
threshold decreases with the increase of the pressure drop.
In the above three sections, the influences of the labyrinth seal
length, labyrinth clearance and pressure drop between the
labyrinth inlet and outlet are calculated, and the mechanisms are
found out. There are synchronous vibrations components 1, and
the system will go unstable when the speed of the shaft (runner)
reach to a certain value. The stability of the rotor system will be
influenced by the structure of the labyrinth seal and the pressure
drop in the seal.
The stability of turbine rotor system is an important problem in
the performance of the power station. When the hydraulic turbine
is designed, the rotor stability should be considered in order to
18 R.-Z. Gong et al. / Computers & Fluids 88 (2013) 11–18
improve the performance of the turbine. According to the above re-
search, surely, the actual situation should be considered. Overall,
the labyrinth seal length should be shorter, the labyrinth clearance
thickness should be thinner and the pressure drop should be con-
trolled to a relatively lower value in the range allowed.
5. Conclusions
In the present paper, flow field of the hydraulic turbine includ-
ing the labyrinth seal clearance is simulated. And the stability of
the rotor system is analyzed. Some factors which influence the sta-
bility of the rotor system are studied. Conclusions drawn from the
research are as follows.
It is realizable to analyze the stability of the hydraulic turbine
with the method of rotor dynamics.
The length of the labyrinth seal clearance, the radial clearance
of the labyrinth and the pressure drop between the inlet and outlet
of the clearance seriously affect the characters of the rotor system.
The threshold speed is lower with the increase of the length of
labyrinth seal clearance. And when the length of the seal is longer,
low frequency vibration appears in the rotor system. However, the
threshold speeds are higher with the increase of the width of lab-
yrinth seal clearance and the increase of the pressure drop. And
when the pressure drop in the clearance is higher, the rotor system
behaves low frequency vibration.
Acknowledgments
The work is supported under National Key Technology R&D
Program-research on the gap flow induced vibration and transient
stability of 1000 MW hydraulic turbine (Project No.
2007BAA05B01). And I appreciate the help of Harbin Institute of
Large Electrical Machinery.
References
[1] Iwatsubo T. Evaluation of instability forces of labyrinth seals in turbines or
compressors. In: The 1st workshop on rotor dynamic instability problems in
high performance turbo machinery, Taxas A&M University; 1980. p. 139–67.
[2] Diewald W, Nordmann R. Dynamic analysis of centrifugal pump rotors with
fluid mechanical interaction. J Vibr Acoust 1989;111:370–8.
[3] Nilsson H, Davidson L. Validations of CFD against detailed velocity and
pressure measurements in water turbine runner flow. Int J Numer Methods
Fluids 2003;41:863–79.
[4] Zhou LJ. Analysis of dynamic stresses in Kaplan turbine blades. Int J Computer-
Aided Eng Softw 2007;24(8):753–62.
[5] Flemming F, Foust J, Koutnik J, Fisher RK. Overload surge investigation using
CFD data. In: 24th IAHR symposium on hydraulic machinery and systems;
2008. p. 1–15.
[6] Alford JS. Protecting turbo-machinery from self-excited whirl. ASME J Eng
Power 1965;5:333–44.
[7] Hashimoto H, Wada S, Nojima K. Performance characteristics of worn journal
bearings in both laminar and turbulent regimes. Part I: steady state
characteristics. ASLE Trans 1986;29(4):565–71.
[8] Hashimoto H, Wada S, Nojima K. Performance characteristics of worn journal
bearings in both laminar and turbulent regimes. Part II: dynamic
characteristics. ASLE Trans 1986;29(4):572–7.
[9] Ren XM, Gu JL, Qin WY. Dynamic stability analysis of rotor system with
labyrinth seals. Chinese J Appl Mech 1996;13(2):77–83.
[10] Ma WS, Zhou LJ. The effect of grid on the result of CFD simulation of turbine. J
Hydroelectric Eng 2006;1:72–5.
[11] Black HF, Jenssen DN. Effects of high-pressure ring seals on pump rotor. ASME
paper, No.71-WA/FF-38; 1971.
[12] Muszunska A. A model testing of rotor/bearing systems. Int J Anal Exp Modal
Anal 1996;1(3):15–34.