Upload
others
View
2
Download
0
Embed Size (px)
Citation preview
Numerical and Experimental Investigation on
the Flow in Rotor-Stator Cavities
Von der Fakultät für Ingenieurwissenschaften, Abteilung Maschinenbau der
Universität Duisburg-Essen
zur Erlangung des akademischen Grades
DOKTOR-INGENIEUR
genehmigte Dissertation
von
Bo Hu
aus
Lanzhou, China
Referent: Prof. Dr.-Ing. F.-K. Benra
Korreferent: Prof. Marcello Manna
Tag der mündlichen Prüfung: 11. 07. 2018
II
Acknowledgement
I would like to express my gratitude to all, who helped me during my study in Germany.
My sincere gratitude goes first and foremost to my supervisor Prof. Dr. –Ing. F.–K. Benra for his
inspirational discussions, constant guidance and encouragement throughout this work. He offers
me the opportunity to start this interesting topic, and gives me great help by providing me with
important materials, advice and inspiration of new ideas during my research. His suggestion as well
as his own technical insights have improved my research work substantially. Furthermore, he has
offered me a lot of opportunities to take part in some good international conferences where I was
able to exchange new research ideas, results and innovations together with other participants
working in my research field.
I am deeply indebted to all my colleagues in the Chair of Turbomachinery at University of
Duisburg-Essen. Special thanks should go to Dr. –Ing. H. J. Dohmen and Detlev Weniger, who
have given me valuable suggestions and patient guidance during the design and construction of my
test rig.
High tribute is also paid to CSC (China Scholarship Council), which provide the funds to cover my
cost of living in Germany and Chair of Turbomachinery at University of Duisburg-Essen, which
provides the funds to build-up the test rig, laboratory site and experimental facilities.
Last but not the least, my gratitude also extends to my family who has been assisting, supporting
and caring for me throughout my life.
Duisburg, Germany
July, 2018
Bo Hu
III
Contents
Abstract ______________________________________________________________________ 1
1. Introduction _________________________________________________________________ 3
1.1 Significance of This Thesis __________________________________________________ 3
1.2 Important Variables and Limitations of Previous Studies ___________________________ 4
1.2.1 Core Swirl Ratio _______________________________________________________ 4
1.2.2 Axial Thrust __________________________________________________________ 4
1.2.3 Frictional Torque ______________________________________________________ 5
2. State of the Art ______________________________________________________________ 9
2.1 Basic Equations ___________________________________________________________ 9
2.2 Thickness of Boundary Layers ______________________________________________ 15
2.3 Core Swirl Ratio _________________________________________________________ 20
2.3.1 Enclosed Rotor-Stator Cavity ____________________________________________ 20
2.3.2 Impact of Through-Flow ________________________________________________ 21
2.3.2.1 With Centripetal Through-Flow (𝑄 < 0 m3/s) __________________________ 23
2.3.2.2 With Centrifugal Through-Flow (𝑄 > 0 m3/s) __________________________ 24
2.3.3 Impact of Surface Roughness in an Enclosed Rotor-Stator Cavity _______________ 26
2.3.4 Impact of Pre-Swirl ____________________________________________________ 26
2.3.4.1 Centripetal Pre-Swirl Through-Flow ___________________________________ 28
2.3.4.2 Centrifugal Pre-Swirl Through-Flow ___________________________________ 29
2.4 Axial Thrust _____________________________________________________________ 31
2.5 Moment Coefficient _______________________________________________________ 33
2.5.1 The Free Disk ________________________________________________________ 33
2.5.2 Enclosed Rotor-Stator Cavity ____________________________________________ 34
2.5.3 Rotor-Stator Cavity with Through-Flow ___________________________________ 36
2.5.4 Impact of Surface Roughness on the Moment Coefficient ______________________ 38
2.6 Flow Separation near the Entrance ___________________________________________ 41
2.7 Influence of the Sealing Gap Height __________________________________________ 44
2.7.1 Flow Structure inside the Sealing Gap _____________________________________ 44
2.7.2 Leakage Volumetric Flow Rate through the Sealing Gap ______________________ 48
2.8 Side Chamber Flow in a Centrifugal Pump _____________________________________ 49
3. Experimental Set-Up _________________________________________________________ 51
3.1 Mechanical Set-Up _______________________________________________________ 51
3.2 Uncertainty Analysis ______________________________________________________ 56
3.3 Experimental Validation ___________________________________________________ 57
4. Numerical Simulation Set-Up __________________________________________________ 60
4.1 Turbulence Model ________________________________________________________ 60
IV
4.2 Grid Generation __________________________________________________________ 61
4.3 Simulation Set-Up ________________________________________________________ 63
4.4 Validation for Numerical Simulation _________________________________________ 65
5. Results and Discussion _____________________________________________________ 67
5.1 Rotor-Stator Cavity with Centripetal Through-Flow _____________________________ 67
5.1.1 Simulation Results of Velocity Distributions ________________________________ 67
5.1.2 Core Swirl Ratio ______________________________________________________ 70
5.1.2.1 Impact of Through-Flow with a Smooth Disk ____________________________ 70
5.1.2.2 Impact of Surface Roughness of the Disk _______________________________ 71
5.1.2.3 Impact of Centripetal Pre-Swirl Through-Flow with a Smooth Disk __________ 73
5.1.3 Pressure Distribution ___________________________________________________ 74
5.1.3.1 Impact of Through-Flow with a Smooth Disk ____________________________ 74
5.1.3.2 Impact of Surface Roughness of the Disk _______________________________ 75
5.1.3.3 Impact of Pre-Swirl on the Pressure Distribution with a Smooth Disk _________ 77
5.1.4 Thrust Coefficient _____________________________________________________ 78
5.1.4.1 Impact of Through-Flow with a Smooth Disk ____________________________ 78
5.1.4.2 Impact of Surface Roughness of the Disk _______________________________ 79
5.1.4.3 Impact of Pre-Swirl on the Thrust Coefficient with a Smooth Disk ___________ 81
5.1.5 3D Diagram with Centripetal Through-Flow ________________________________ 83
5.1.6 Moment Coefficient ___________________________________________________ 84
5.1.6.1 Impact of Through-Flow with a Smooth Disk ____________________________ 84
5.1.6.2 Impact of Surface Roughness of the Disk _______________________________ 86
5.1.6.3 Impact of Pre-Swirl on the Moment Coefficient with a Smooth Disk __________ 95
5.2 Rotor-Stator Cavity with Centrifugal Through-Flow _____________________________ 96
5.2.1 Simulation Results of Velocity Distributions ________________________________ 96
5.2.2 Core Swirl Ratio ______________________________________________________ 98
5.2.2.1 Impact of Through-Flow with a Smooth Disk ____________________________ 98
5.2.2.2 Impact of Surface Roughness of the Disk _______________________________ 99
5.2.2.3 Impact of Centrifugal Pre-Swirl Through-Flow with a Smooth Disk _________ 100
5.2.3 Pressure Distribution __________________________________________________ 102
5.2.3.1 Impact of Through-Flow with a Smooth Disk ___________________________ 102
5.2.3.2 Impact of Surface Roughness of the Disk ______________________________ 103
5.2.3.3 Impact of Pre-Swirl on the Pressure Distribution with a Smooth Disk ________ 104
5.2.4 Thrust Coefficient ____________________________________________________ 104
5.2.4.1 Impact of Centrifugal Through-Flow with a Smooth Disk _________________ 104
5.2.4.2 Impact of Surface Roughness of the Disk ______________________________ 105
5.2.4.3 Impact of Pre-Swirl on the Thrust Coefficient with a Smooth Disk __________ 107
5.2.5 3D Diagram with Centrifugal Through-Flow _______________________________ 108
5.2.6 Moment Coefficient __________________________________________________ 108
5.2.6.1 Impact of Through-Flow with a Smooth Disk ___________________________ 108
5.2.6.2 Impact of Surface Roughness of the disk _______________________________ 110
5.2.6.3 Impact of Pre-Swirl on the Moment Coefficient with a Smooth Disk _________ 113
V
6. Applications of the Results in Radial Pumps _____________________________________ 115
6.1 Flow in the Rear Chamber of a Submersible Multi-Stage Slurry Pump (SMSP) _______ 115
6.1.1 Sand Discharge Groove Design _________________________________________ 115
6.1.2 Geometrical Parameters _______________________________________________ 116
6.1.3 Numerical Simulation _________________________________________________ 117
6.1.4 Results and Discussion ________________________________________________ 118
6.2 The Axial Thrust in a Deep Well Pump ______________________________________ 123
6.2.1 Main Geometric Parameters ____________________________________________ 123
6.2.2 Simulation Set-Up ____________________________________________________ 124
6.2.3 Results for Axial Thrust Coefficient ______________________________________ 124
7. Summary of the Results ____________________________________________________ 126
8. Outlook ________________________________________________________________ 128
References __________________________________________________________________ 129
VI
Nomenclature
Latin Symbols
A Area m2
a Hub radius m
𝑎𝑅 Velocity factors for the rotor in case of separated boundary layers −
𝑎𝑆 Velocity factors for the stator in case of separated boundary layers −
𝐵+ Constant in the expression for the boundary layer thickness −
b Outer radius of the disk m
𝑏2 Outlet width of impeller mm
𝐶𝑎𝑚 Coefficient of the inlet angular momentum −
𝐶𝐷 Through-flow coefficient −
𝐶𝐹 Axial thrust coefficient −
𝐶𝐹𝑓 𝐶𝐹 on the front surface −
𝐶𝐹𝑓−𝑎𝑐 𝐶𝐹𝑓 when disk rotates anti-clockwise −
𝐶𝐹𝑓−𝑐 𝐶𝐹𝑓 when disk rotates clockwise −
𝐶𝐹𝑏 𝐶𝐹 on the back surface −
𝐶𝐹𝑠 𝐶𝐹 when the shaft is rotating without the disk −
𝐶𝑓 Skin friction coefficient −
𝐶𝑀 Moment coefficient −
𝐶𝑀𝑎 𝐶𝑀 for the source region −
𝐶𝑀𝑏 𝐶𝑀 for the core region −
𝐶𝑀𝑐 𝐶𝑀 due to the cylindrical shroud −
𝐶𝑀𝑐𝑦𝑙 𝐶𝑀 on the cylinder surface of the disk −
𝐶𝑀𝑠 𝐶𝑀 of the shaft −
𝐶𝑀𝑠−𝑎𝑐 𝐶𝑀𝑠 when disk rotates anti-clockwise −
𝐶𝑀𝑠−𝑐 𝐶𝑀𝑠 when disk rotates clockwise −
𝐶𝑀1 𝐶𝑀 for regime I −
𝐶𝑀2 𝐶𝑀 for regime II −
𝐶𝑀3 𝐶𝑀 for regime III −
𝐶𝑀4 𝐶𝑀 for regime IV −
𝐶𝑝 Pressure coefficient −
𝐶𝑞 Through-flow rate coefficient −
𝐶𝑞𝑟 Local flow rate coefficient −
𝐶𝑅 Factor in the flow model formulation −
𝐶𝑆 Factor in the flow model formulation −
𝐶+ Constant in the equation for the boundary layer thickness on the stator −
𝐷1 Diameter of the impeller eye mm
VII
𝐷2 Outer diameter of the impeller mm
𝑑𝑔𝑎𝑝 Diameter of the sealing gap m
𝑑ℎ Diameter of the horizontal pipe m
𝐸𝑘 Ekman number −
e𝑇 Relative error of the transducer −
𝑐 Constant −
𝑐𝑠 Suitable dimensionless radial gap −
𝑐∗ Constant −
e𝐷 Relative error due to the data acquisition device −
Fa Axial thrust N
Faf Force on the front surface of the disk N
Fab Force on the back surface of the disk N
Fap Force on the impeller passage N
Fas Impulse fore at the impeller eye N
𝐹𝑎−𝑠 Axial thrust when the shaft is rotating without the disk N
𝑓 Body force N
𝑓𝑟 Body force in radial direction N
𝑓𝜑 Body force in tangential direction N
𝑓𝑧 Body force in axial direction N
𝑓∗ Correction function −
G Non-dimensional axial gap −
H Pressure head of a pump m
K Core swirl ratio at 휁 = 0.5 −
𝐾0 K for enclosed rotor-stator cavity −
𝐾𝑏 K at 𝑟 = 𝑏 −
𝐾𝑒 K at the entrance −
𝐾ℎ K at the radius of pre-swirl nozzle −
𝐾𝑝 K with pre-swirl −
𝐾ℎ,𝑒𝑓𝑓 Effective pre-swirl ratio at the radius of pre-swirl nozzle −
𝑘𝑠 Equivalent surface roughness 𝜇m
𝑘𝑠𝑙 Transition point between the hydraulic smooth disks and the disks in the transition zone 𝜇m
L Angular momentum in the centripetal through-flow −
l Streamwise coordinates m
𝑀 Frictional torque Nm
𝑀𝑐𝑦𝑙 Frictional resistance on the cylinder surface of the disk Nm
𝑀𝑟 Measured range −
𝑀𝑠 Frictional torque when the shaft is rotating without the disk Nm
m Constant −
VIII
�̇� Mass flow rate kg/s
𝑁𝐷 Uncertainty of the data acquisition system −
𝑁𝑇 Uncertainty of the transducer −
∆𝑁 Uncertainty of the measured results −
n Speed of rotation rpm
𝑛1 Constant −
𝑛𝑇 Number of transducers −
𝑛𝑀 Measuring times to obtain one result −
𝑛𝑠 Specific speed of pump −
p Pressure Pa
𝑝𝑏 Pressure at 𝑟 = 𝑏 Pa
𝑝𝑟 Pressure drop ratio −
𝑝𝑥 Pressure at 𝑡ℎ = 𝑥 Pa
𝑝0 Pressure at 𝑡ℎ = 0 Pa
𝑝∗ Non-dimensional pressure −
Q Volumetric through-flow rate m3/s
𝑄𝑒 Flow rate at the point of best efficiency L/s
Re Global circumferential Reynolds number −
Rep Perimeter Reynolds number for the sealing gap −
Reφ Local circumferential Reynolds number −
r Radial coordinate m
𝑟𝑎 Radius of the hub m
𝑟ℎ Radius of the horizontal pipe m
∆𝑟 Radial distance from the disk to the wall m
∆𝑟𝑠𝑒𝑎𝑙 Sealing gap height m
s Axial gap of the front chamber m
sb Axial gap of the back chamber m
𝑇𝑎 Taylor number −
t Thickness of the disk m
𝑡ℎ Time h
U Velocity of the free stream m/s
u Velocity along the flat plate m/s
𝑢∗ Friction velocity m/s
𝑉𝑟 Non-dimensional radial velocity −
𝑉𝑧 Non-dimensional axial velocity −
𝑉𝜑 Non-dimensional tangential velocity −
𝑉𝜑ℎ 𝑉𝜑 at 𝑥 = 𝑥ℎ −
v Velocity m/s
IX
v𝑎𝑥𝑔𝑎𝑝 Sealing gap mean axial velocity m/s
v𝑔𝑎𝑝 Sealing gap circumferential velocity m/s
v𝑟 Radial velocity m/s
v𝑧 Axial velocity m/s
v𝜑 Tangential velocity m/s
v𝑚𝑒𝑎𝑛 Mean velocity m/s
x Non-dimensional radial coordinate −
𝑥ℎ Non-dimensional radius of the horizontal pipe −
𝑥𝑐 Radial location where source region ends −
𝑥∗ 𝑥∗ = 𝐾𝑝0.5 ∙ 𝑥𝑎 −
∆𝑥 Non-dimensional radial gap width −
𝑦 Spacing of the first layer node m
𝑦+ Non-dimensional wall distance −
𝑧 Axial coordinate m
𝑧𝑙 Normal coordinates m
Greek Symbols
𝛼 Index −
𝛽 Pre-swirl angle Deg
𝛽2 Outlet blade angle Deg
𝛾 Heat capacity ratio −
𝛶𝑅 Proportionality factor for the boundary layer thickness on the rotor −
𝛶𝑠 Proportionality factor for the boundary layer thickness on the stator −
𝛿 Thickness of the boundary layer m
𝜍 Distance from the buffer layer to the viscous sublayer m
𝛿𝑅 Thickness of the disk layer m
𝛿𝑠 Thickness of the wall layer m
휀 Diameter of spheres 𝜇m
휁 Non-dimensional axial coordinate m
휂 Pump efficiency −
휃 Angle in cylindrical coordinates Deg
휃1 Wrapping angle of blade Deg
𝜆𝑅 Friction factor for the rotor wall −
𝜆𝑆 Friction factor for the stator wall −
𝜆𝑇 Turbulent flow parameter −
𝜇 Dynamic viscosity of water N ∙ s m2⁄
𝑣 Kinematic viscosity of water m2 s⁄
𝜌 Density of water kg m3⁄
X
𝜏 Shear stress N m2⁄
𝜏𝑤 Wall shear stress N m2⁄
𝜑𝐺 Non-dimensional through-flow rate −
𝛹 Flow rate coefficient −
𝛺 Angular velocity of the disk rad s⁄
𝛺𝑓 Angular velocity of the fluid at 휁 = 0.5 rad s⁄
Abbreviations
DC Direct current
DNS Direct numerical simulation
DWP Deep well pump
FS Full scale
LDA Laser Doppler Anemometer
LDV Laser Doppler Velocimetry
LES Large eddy simulation
RANS Reynolds-averaged Navier-Stokes equations
RSM Reynolds Stress Models
rpm Revolution per minute
SMSP Submersible multi-stage slurry pump
SST Shear stress transport
TF Through-flow
SR-4 Simmons Ruge-4
XI
List of figures
Fig Name Page
1 Cross section of a centrifugal pump 3
2 Concerns during the design of a radial pump or turbine 4
3 Sources of the axial thrust for a radial pump 5
4 Typical velocity profiles for the four flow regimes 5
5 Distinguishing lines for flow regimes without through-flow (2D Daily&Nece diagram [28]) 6
6 Contents of this thesis 7
7 Main geometry of the test rig 7
8 Flow structure in an idealized rotor-stator cavity (left) and the velocity profiles for a wide
axial gap (replotted from Will [88]) 16
9 Radial velocity profiles in dependence on Eq. 36 and Eq. 37 (replotted from Will [88]) 18
10 Comparison of results from Eq. 51 and Eq. 52 24
11 Velocity profiles for Batchelor type flow, Couette type flow and Stewartson type flow 24
12 Results for K from different equations for centrifugal through-flow 25
13 Experimental results of K along the radius of disk for 𝐺 = 0.031, 𝑅𝑒 = 3.1 × 106 and
𝐶𝐷 = 0 by Kurokawa et al. [54] 26
14 Velocity triangles for the pre-swirl through-flow 27
15 Flow structure in the case of 𝐾 > 1 29
16 Sketch of the rotor-stator cavity (redrawn from Karabay et al. [49]) 30
17 Variation of 𝐾ℎ,𝑒𝑓𝑓 versus 𝐾ℎ by Karabay [49] 31
18 Flow structure around a free disk (According to Schlichting and Gersten [73]) 33
19 Torques on an annular volume element 33
20 Flow structure inside a rotor-stator cavity with (a) no through-flow, (b) with centrifugal
through-flow and 𝜆𝑇 < 0.219 and (c) with through−flow, 𝜆𝑇 > 0.219 37
21 Moment coefficient for a rotating disk 38
22 Schematic drawing of the test rig (redrawn from Daily and Nece [28]) 39
23 Rough disk torque data (replotted from Daily and Nece [28]) 40
24 Sketch of the test rig (replotted from Kurokawa et al. [54]) 41
25 Graphical representation of the velocity profile and the reverse flow which show the flow
separation [3] 41
26 Approximate separation line for centripetal through-flow 42
27 Solution procedure for the approximate separation line 43
28 Comparison of the results for the separation line (𝛺 = 0, 𝐶𝐷 = −5050, 𝐺 = 0.045) 43
29 One seventh segment of the back cavity (Will [88]) 44
30 Comparison between experimental and numerical results for the shroud side chamber (left)
and the hub side chamber (right) in case of 0.48 mm sealing gap height (Will [88]) 45
31 Comparison between experimental and numerical results for the hub side chamber in case of
0.24 mm sealing gap height (Will [88]) 45
XII
32 Flow in the sealing gap (∆𝑟 = 0.8 mm) for different leakage flow rates (replotted from Will
[88]) 48
33 Test rig design 52
34 Test rig set-up for each steps to measure 𝐹𝑎 53
35 Inlet swirlers in the horizontal pipe (for centrifugal through-flow) 53
36 Drawing of the centrifugal inlet swirler (Left: front view; Right: side view) 54
37 Geometry of the radial guide vanes and the flow in the rotor-stator cavities 54 to 55
38 Experimental results of 𝐶𝑀𝑠 and 𝐶𝐹𝑠 versus Re 57
39 Comparison of results when the shaft rotates in different direction 58
40 Comparison of 𝐶𝑝 along the radius for 𝑅𝑒 = 1.36 × 106 59
41 Domain for numerical simulation (𝐺 = 0.072) 64
42 Additional fluid domain for centrifugal pre-swirl through-flow 64
43 Mesh independence analysis 65
44 Comparison of radial pressure distribution for 𝑅𝑒 = 4.15 × 106 and 𝐺 = 0.036 66
45 Comparisons of radial pressure distribution for 𝑅𝑒 = 1.36 × 106 and 𝐺 = 0.0495 without
pre-swirl 66
46 Research procedure 67
47 Velocity profiles for 𝑅𝑒 = 1.9 × 106 and 𝐺 = 0.072 68
48 Velocity profiles for 𝑅𝑒 = 1.9 × 106 and 𝐺 = 0.018 69
49 K (𝐶𝑞𝑟) curves for centripetal through-flow 71
50 K (x) curves for 𝐺 = 0.031, 𝑅𝑒 =3.1× 106 and 𝐶𝐷 = 0 by Kurokawa et al. [54] 72
51 Impact of 𝑘𝑠 on K when G=0.072 73
52 Inlet boundary conditions with centripetal through-flow 73
53 𝐾𝑝 (𝑥) curves for 𝐺 = 0.072, 𝐶𝐷 = −5050 and 𝑅𝑒 = 1.9 × 106 74
54 Influence of 𝐶𝐷 on 𝐶𝑝 in dependence of Re and G (𝑘𝑠 = 0.4 𝜇m) 75
55 𝐶𝑝 (x) curves along the radius of the disks 76 to 77
56 𝐶𝑝 (x) curves for 𝐺 = 0.072, 𝐶𝐷 = −5050 and 𝑅𝑒 = 1.9 × 106 with centripetal through-
flow 77
57 𝐶𝐹 (𝐶𝐷) curves for 𝑅𝑒 = 1.32 × 106 78
58 𝐶𝐹 (𝑅𝑒) curves in dependence of 𝐶𝐷 and G 79
59 𝐶𝐹 (𝑅𝑒) curves in dependence of 𝐶𝐷, G and 𝑘𝑠 80 to 81
60 Comparison of the results of 𝐶𝐹 for 𝐺 = 0.05 and 𝐶𝑎𝑚 𝐶𝑞⁄ = −0.619 [52] 82
61 𝐶𝐹 (𝐶𝑎𝑚) curves for 𝐶𝐷 = −5050 for various 𝑅𝑒 and G 82 to 83
62 3D diagram distinguishing regime III and regime IV with centripetal through-flow for 0.3 ×106 ≤ 𝑅𝑒 ≤ 3.3 × 106 84
63 Comparison of the results of 𝐶𝑀 for 𝐺 = 0.018 and 𝐺 = 0.072 at 𝐶𝐷 = 0 85
64 𝐶𝑀 (Re) curves with centripetal through-flow 85 to 86
65 Results of 𝐶𝑀3/𝐶𝑀4 at the distinguishing lines for centripetal through-flow 86
66 Comparison of 𝐶𝑀 with different values of 𝑘𝑠 in an enclosed rotor-stator cavity 87
XIII
67 Comparison of 𝐶𝑀 from different equations 88
68 𝐶𝑀 (𝑅𝑒) curves at 𝐺 = 0.012 and 𝐺 = 0.027 for different values of 𝑘𝑠 and |𝐶𝐷| 89 to 90
69 𝐶𝑀 (Re) curves at 𝐺 = 0.047 and 𝐺 = 0.065 for different values of 𝑘𝑠 and |𝐶𝐷| 91 to 93
70 𝐶𝑀 (Re) curves at various G for different values of 𝑘𝑠 and 𝐶𝐷 93 to 94
71 𝐶𝑀 (𝐶𝑎𝑚) curves for |𝐶𝐷| = 5050 at different 𝑅𝑒 95
72 Velocity profiles for 𝑅𝑒 = 1.9 × 106 and 𝐺 = 0.072 96
73 Velocity profiles for 𝑅𝑒 = 1.9 × 106 and 𝐺 = 0.018 97
74 𝐾 (𝐶𝑞𝑟) curves 98
75 Large differences of K attributed to the geometry near the outlet 99
76 𝐾 (𝐶𝑞𝑟) curves for various 𝑘𝑠 at 𝐺 = 0.072 100
77 𝐾𝑝 (𝐶𝑞𝑟) curves for 𝑅𝑒 = 1.9 × 106, 𝐶𝐷 = 5050 and 𝐺 = 0.072 101
78 Distribution of 𝐶𝑝 along the radius 102
79 Distribution of 𝐶𝑝 along the radius for the rough disks 103
80 𝐶𝑝 (x) curves for various 𝐶𝑎𝑚 at 𝐶𝐷 = 5050 and G=0.072 104
81 𝐶𝐹 (𝑅𝑒) curves (𝑘𝑠 = 0.4 𝜇m) 105
82 𝐶𝐹 (𝑅𝑒) curves in dependence of 𝐶𝐷, 𝐺 and 𝑘𝑠 106
83 Mean 𝐶𝐹 (𝑅𝑒) curves (𝑘𝑠 = 0.4 𝜇m) 107
84 3D diagram distinguishing regime III and regime IV with centrifugal through-flow for 0.3 ×106 ≤ 𝑅𝑒 ≤ 3.3 × 106 108
85 Curves for 𝐶𝑀 in dependence of 𝑅𝑒 for different values of 𝐶𝐷 and 𝐺 with centrifugal
through-flow 109
86 Results of 𝐶𝑚3/𝐶𝑚4 at the distinguishing lines for centrifugal through-flow 110
87 Curves for 𝐶𝑀 in dependence of 𝑅𝑒 for different values of 𝐶𝐷 and 𝐺 with centrifugal
through-flow 111 to 113
88 Experimental results of 𝐶𝑀 versus 𝑅𝑒 for 𝑘𝑠 = 0.4 𝜇m and 𝐶𝐷 = 5050 114
89 Leakage flow and flow pattern inside the rear chamber of a SMSP 115
90 Sand discharge groove 116
91 Leakage flow and flow pattern inside the rear chamber of a SMSP 118
92 Distribution of the radial velocity 118
93 Movement of particles on the meridian plane 119
94 Distribution of non-dimensional radial velocity 120
95 Distribution of non-dimensional tangential velocity 121
96 Distribution of solid volume fraction 122
97 Pump performance from experiments 122
98 Pressure drop during the abrasion test 123
99 Geometry of the impeller (left) and the guide vane (right) 124
100 Axial thrust in a centrifugal single stage well pump [84]: (a) Pressure distribution and (b)
Comparison of 𝐹𝑎𝑏 − 𝐹𝑎𝑓 125
XIV
List of Tables
Tab Name Page
1 Current research states on flow in rotor-stator cavities 6
2 Values of 𝐵+ and m from the literature 17
3 Values of 𝑐 and 𝑐∗ by Kurokawa and Toyokura [53] 17
4 Core swirl ratio 𝐾0 in the literature for an enclosed rotor-stator cavity (reorganized from Will
[88]) 21
5 Parameters of the experiments 44
6 Estimation of 𝑅𝑒𝑔𝑎𝑝 and 𝑇𝑎 in Will [88] 47
7 Parameters of the experiments 52
8 Parameters of the experiments conducted in the test rig 53
9 Pre-swirl angles of the radial guide vanes 56
10 Surface roughness of the disks 56
11 Uncertainty analysis for the measurements 57
12 Selections of turbulence model in the literature 60
13 Moment coefficients from different turbulence models (K. N. Volkov [51]) 61
14 Grid number and maximum values of 𝑦+ 63
15 Qualities of the meshes 63
16 Geometrical parameters of the impeller 116
17 Parameters of the 6 rotor-stator cavities 117
18 Uncertainties of results 122
19 Main geometric parameters of the pump 123
XV
List of Equations
𝜕𝑝
𝜕𝑡+
𝜕(𝜌 ∙ v𝑖)
𝜕𝑥𝑖
= 0 Eq. 1
𝜕(𝜌 ∙ v𝑖)
𝜕𝑡+
𝜕(𝜌 ∙ v𝑖 ∙ v𝑖)
𝜕𝑥𝑗
= 𝑓𝑖 −𝜕𝑝
𝜕𝑥𝑖
+𝜕𝜏𝑖𝑗
𝜕𝑥𝑗
Eq. 2
𝜏𝑖𝑗 = 𝜇 ∙ (𝜕v𝑖
𝜕𝑥𝑗
+𝜕v𝑗
𝜕𝑥𝑖
−2
3∙ 𝛿𝑖𝑗 ∙
𝜕v𝑘
𝜕v𝑘
) Eq. 3
𝜌 ∙ (𝜕v𝑟
𝜕𝑡+ v𝑟 ∙
𝜕v𝑟
𝜕𝑟+
v𝜑
𝑟∙
𝜕v𝑟
𝜕𝜑−
v𝜑2
𝑟+ v𝑧 ∙
𝜕v𝑟
𝜕𝑧)
= 𝑓𝑟 −𝜕𝑝
𝜕𝑟+
1
𝑟∙
𝜕(𝑟 ∙ 𝜏𝑟𝑟)
𝜕𝑟+
1
𝑟∙
𝜕(𝜏𝑟𝜑)
𝜕𝜑−
𝜏𝜑𝜑
𝑟+
𝜕(𝜏𝑟𝑧)
𝜕𝑧
Eq. 4
𝜌 ∙ (𝜕v𝜑
𝜕𝑡+ v𝑟 ∙
𝜕v𝜑𝜑
𝜕𝑟+
v𝜑
𝑟∙
𝜕v𝜑
𝜕𝜑+
v𝜑 ∙ v𝑟
𝑟+ v𝑧 ∙
𝜕v𝜑
𝜕𝑧)
= 𝑓𝜑 −1
𝑟∙
𝜕𝑝
𝜕𝜑+
1
𝑟2∙
𝜕(𝑟2 ∙ 𝜏𝜑𝑟)
𝜕𝑟+
1
𝑟∙
𝜕(𝜏𝜑𝜑)
𝜕𝜑+
𝜕(𝜏𝜑𝑧)
𝜕𝑧
Eq. 5
𝜌 ∙ (𝜕v𝑧
𝜕𝑡+ v𝑟 ∙
𝜕v𝑧
𝜕𝑟+
v𝜑
𝑟∙
𝜕v𝑧
𝜕𝜑+ v𝑧 ∙
𝜕v𝑧
𝜕𝑧) = 𝑓𝑧 −
𝜕𝑝
𝜕𝑧+
𝜕(𝜏𝑧𝑟)
𝜕𝑟+
𝜏𝑧𝑟
𝑟+
1
𝑟∙
𝜕(𝜏𝑧𝜑)
𝜕𝜑+
𝜕(𝜏𝑧𝑧)
𝜕𝑧
Eq. 6
1
𝑟∙
𝜕(𝑟 ∙ v𝑟)
𝜕𝑟+
1
𝑟∙
𝜕v𝜑
𝜕𝜑+
𝜕v𝑧
𝜕𝑧= 0
Eq. 7
𝜕(𝜌 ∙ v𝑖)
𝜕𝑡+
𝜕(𝜌 ∙ v𝑖 ∙ v𝑖)
𝜕𝑥𝑗
= 𝑓𝑖 −𝜕𝑝
𝜕𝑥𝑖
Eq. 8
𝜌 ∙ (v𝑟 ∙𝜕v𝑟
𝜕𝑟−
v𝜑2
𝑟+ v𝑧 ∙
𝜕v𝑧
𝜕𝑧) = −
𝜕𝑝
𝜕𝑟+
𝜕(𝜏𝑟𝑧)
𝜕𝑧
Eq. 9
𝜌 ∙ (v𝑟 ∙𝜕v𝜑
𝜕𝑟+
v𝑟 ∙ v𝜑
𝑟+ v𝑧 ∙
𝜕v𝜑
𝜕𝑧) =
𝜕(𝜏𝜑𝑧)
𝜕𝑧
Eq. 10
𝜌 ∙ (v𝑟 ∙𝜕v𝑧
𝜕𝑟+ v𝑧 ∙
𝜕v𝑧
𝜕𝑧) = −
𝜕𝑝
𝜕𝑧+
𝜏𝑟𝑧
𝑟+
𝜕(𝜏𝑟𝑧)
𝜕𝑟
Eq. 11
𝜕v𝑟
𝜕𝑟+
v𝑟
𝑟+
𝜕v𝑧
𝜕𝑧= 0
Eq. 12
𝜌 ∙ (v𝑟 ∙𝜕v𝜑
𝜕𝑟+
v𝑟 ∙ v𝜑
𝑟) = 0
Eq. 13
v𝑟
𝑟∙
𝜕
𝜕𝑟(v𝜑 ∙ 𝑟) = 0
Eq. 14
𝜕𝑝
𝜕𝑟= 𝜌 ∙
v𝜑2
𝑟
Eq. 15
𝜕𝑝
𝜕𝑟= 𝜌(∙
v𝜑2
𝑟− v𝑟 ∙
𝜕v𝑟
𝜕𝑟)
Eq. 16
v𝑟 ∙𝜕v𝑟
𝜕𝑟−
v𝜑2
𝜕𝑟+ v𝑧 ∙
𝜕v𝑟
𝜕𝑧=
1
𝑟∙ [
𝜕
𝜕𝑟∙ (𝑟 ∙ v𝑟
2) +𝜕
𝜕𝑧∙ (𝑟 ∙ v𝑟 ∙ v𝑧) − v𝜑
2] Eq. 17
v𝑟 ∙𝜕v𝜑
𝜕𝑟−
v𝑟 ∙ v𝜑
𝑟+ v𝑧 ∙
𝜕v𝜑
𝜕𝑧=
1
𝑟2∙ [
𝜕
𝜕𝑟∙ (𝑟 ∙ v𝑟 ∙ v𝜑) +
𝜕
𝜕𝑧∙ (𝑟2 ∙ v𝑧 ∙ v𝜑)]
Eq. 18
XVI
𝜕
𝜕𝑟(𝑟 ∙ v𝑟) + 𝑟 ∙
𝜕
𝜕𝑧∙ (v𝑟 ∙ v𝑧) − v𝜑
2 = −𝑟
𝜌∙
𝜕𝑝
𝜕𝑟+
𝑟
𝜌∙
𝜕(𝜏𝑟𝑧)
𝜕𝑧
Eq. 19
𝜕
𝜕𝑟(𝑟2 ∙ v𝑟 ∙ v𝜑) +
𝜕
𝜕𝑧∙ (𝑟2 ∙ v𝑧 ∙ v𝜑) = −
𝑟2
𝜌∙
𝜕(𝜏𝜑𝑧)
𝜕𝑟
Eq. 20
( ∫𝜕𝑋(𝑟, 𝑧)
𝜕𝑟
𝑧2
𝑧1
𝑑𝑧) =𝜕
𝜕𝑟( ∫ 𝑋(𝑟, 𝑧)
𝑧2
𝑧1
𝑑𝑧) +𝜕𝑧1
𝜕𝑟∙ 𝑋(𝑟, 𝑧1 ) −
𝜕𝑧2
𝜕𝑟∙ 𝑋(𝑟, 𝑧2)
Eq. 21
𝜕
𝜕𝑟(𝑟 ∙ ∫ v𝑟
2
𝑧2
𝑧1
𝑑𝑧) + 𝑟 ∙𝜕𝑧1
𝜕𝑟∙ v𝑟1
2 − 𝑟 ∙𝜕𝑧2
𝜕𝑟∙ v𝑟2
2 + 𝑟 ∙ v𝑟2 ∙ v𝑧2 − 𝑟 ∙ v𝑟1 ∙ v𝑧1 − ∫ v𝜑2𝑑𝑧
𝑧2
𝑧1
= −𝑟
𝜌∙ ∫
𝜕𝑝
𝜕𝑟𝑑𝑧 +
𝑟
𝜌
𝑧2
𝑧1
∙ ∫𝜕(𝜏𝑟𝑧)
𝜕𝑧𝑑𝑧
𝑧2
𝑧1
Eq. 22
𝜕
𝑟2∙
𝜕
𝜕𝑟(𝑟2 ∙ ∫ v𝑟 ∙ v𝜑
𝑧2
𝑧1
𝑑𝑧) +𝜕𝑧1
𝜕𝑟∙ v𝑟1 ∙ v𝜑1 −
𝜕𝑧2
𝜕𝑟∙ v𝑟2 ∙ v𝜑2 + 𝑟 ∙ v𝑧2 ∙ v𝜑2 − v𝑧1 ∙ v𝜑1
=1
𝜌∙ ∫
𝜕(𝜏𝜑𝑧)
𝜕𝑧𝑑𝑧
𝑧2
𝑧1
Eq. 23
1
𝑟∙
𝜕
𝜕𝑟(𝑟 ∙ ∫ v𝑟
2
𝑠
0
𝑑𝑧) −1
𝑟∙ ∫ v𝜑
2
𝑠
0
𝑑𝑧 = −1
𝜌∙ ∫
𝜕𝑝
𝜕𝑟𝑑𝑧
𝑠
0
+1
𝜌∙ (𝜏𝑟𝑧𝑆 − 𝜏𝑟𝑧𝑅)
Eq. 24
1
𝑟2∙
𝜕
𝜕𝑟(𝑟2 ∙ ∫ v𝑟v𝜑
𝑠
0
𝑑𝑧) =1
𝜌∙ (𝜏𝜑𝑧𝑆 − 𝜏𝜑𝑧𝑅)
Eq. 25
𝑑𝐾
𝑑𝑟=
2 ∙ 𝜋 ∙ 𝑏
�̇� ∙ 𝛺∙ (𝜏𝜑𝑧𝑆 − 𝜏𝜑𝑧𝑅) −
2 ∙ 𝐾
𝑅
Eq. 26
𝜏 = 𝜆 ∙𝜌
8∙ v𝑚𝑒𝑎𝑛
2
Where v𝑚𝑒𝑎𝑛𝑅 = 𝑟2 ∙ 𝛺2 ∙ (1 − 𝐾)2;
v𝑚𝑒𝑎𝑛𝑆 = 𝑟2 ∙ 𝛺2 ∙ 𝐾2. Eq. 27
𝑑𝐾
𝑑𝑟=
�̅�2
4 ∙ 𝜑𝐺
∙ (𝜆𝑠 ∙ 𝐾2 − 𝜆𝑅 ∙ (1 − 𝐾)2) −2 ∙ 𝐾
𝑅
Where 𝜑𝐺 =𝑄
𝜋∙𝛺∙𝑏3. Eq. 28
𝜑𝐺 → 0, 𝐾 = 0.5 Eq. 29
𝜑𝐺 → ∞, 𝐾 ∙ 𝑥2 = 0.5 Eq. 30
∫ v𝑟v𝜑𝑧=𝑠
𝑧=0𝑑𝑧 = ∫ v𝑟v𝜑
𝑧=𝛿𝑅
𝑧=0𝑑𝑧+∫ v𝑟v𝜑
𝑧=𝑠−𝛿𝑠
𝑧=𝛿𝑅𝑑𝑧+∫ v𝑟v𝜑
𝑧=𝑠
𝑧=𝑠−𝛿𝑠𝑑𝑧
Eq. 31
𝛿𝑅 = 𝐵+ ∙𝑟
𝑅𝑒𝜑
15
∙ (1 − 𝐾)𝑚
Eq. 32
XVII
𝛿𝑠 =𝑓 ∙ 𝑟
(𝑟2 ∙ 𝜔
𝜈)
15
Where 𝑓 =1
𝑐∙𝐾∙ [𝑐∗ ∙ 𝑏 ∙ (1 − 𝐾)3 −
120
49∙
𝑄
2∙𝜋∙𝜔∙𝑟3 ∙ (𝑟2∙𝜔
𝜈)
1
5]. Eq. 33
𝑄 = ∫ v𝑟
𝑧=𝑠
𝑧=0
𝑑𝑧
Eq. 34
∫ v𝑟𝑆
𝑧=𝛿𝑠
𝑧=0
𝑑𝑧𝑠 + ∫ v𝑟𝑅
𝑧=𝛿𝑅
𝑧=0
d𝑧𝑅 =𝑄
2 ∙ 𝜋 ∙ 𝑟
Eq. 35
v𝑟𝑅 = 𝑎𝑅 ∙ (1 − 𝐾) ∙ 𝑟 ∙ 𝛺 ∙ (1 −𝑧𝑅
𝛿𝑅
)𝑛1 ∙ (𝑧𝑅
𝛿𝑅
)1𝑚
Eq. 36
v𝑟𝑆 = −𝑎𝑆 ∙ 𝐾 ∙ 𝑟 ∙ 𝛺 ∙ (1 −𝑧𝑠
𝛿𝑠
)𝑛1 ∙ (𝑧𝑠
𝛿𝑠
)1𝑚
Eq. 37
𝑎𝑅 = 1.18 ∙ (𝑅𝑒𝜑
105+ 2)−0.49
Eq. 38
𝑎𝑆 = 1.03 ∙ (𝑅𝑒𝜑
105+ 2)−0.387
Eq. 39
𝛿𝑠 = 0.304 ∙𝑐∗
𝑐∙
(1 − 𝐾)125
𝐾∙
𝑟
𝑅𝑒𝜑
15
−𝑄
0.408 ∙ 𝑐 ∙ 2 ∙ 𝜋 ∙ 𝑟2 ∙ 𝛺 ∙ 𝐾
Eq. 40
𝛿𝑅 = 𝛶𝑅 ∙ 𝑟35 ∙ (
𝜈
𝛺)
15
Eq. 41
𝛿𝑠 = 𝛶𝑠 ∙ 𝑟35 ∙ (
𝜈
𝛺)
15
Eq. 42
𝜆𝑅 =0.18
𝐶𝑅
∙ 𝑅𝑒𝜑−
15 ∙ (
1
1 − 𝐾)
14
Eq. 43
𝜆𝑆 =0.18
𝐶𝑆
∙ 𝑅𝑒𝜑−
15 ∙ (
1
𝐾)
14
Eq. 44
𝐶𝑅 = 0.315 Eq. 45
𝐶𝑆 = 𝐶𝑅 ∙ (1 − 𝐾0
𝐾0
)74
Eq. 46
𝐾0 =1
1 + √1 + 5 ∙ 𝐺
Eq. 47
𝐾 = 0.25 ∙ [−1 + √5 − 4 ∙ 𝜑𝐺
√𝑅𝑒𝜑
𝑥2]
2
Eq. 48
(1 − 𝐾)85 ∙ (1 − 0.51 ∙ 𝐾) − 0.638 ∙ 𝐾
45 = 0.25 ∙ [−1 + 4 ∙ 𝜑𝐺√
𝑅𝑒𝜑
𝑥2]
2
Eq. 49
XVIII
𝑑𝐾
𝑑𝑅=
𝑅2
4 ∙ 𝜑𝐺
∙ (𝑓∗ ∙ 𝜆𝑠 ∙ 𝐾2 − 𝜆𝑅 ∙ (1 − 𝐾)2) −2 ∙ 𝐾
𝑅
Where 𝑓∗ = 1 + (𝑠
𝑏+𝑙1−𝑎+ 5 ∙ 𝑅4 ∙ |1 −
𝐾
0.58|
6
5).
Eq. 50
𝐾 = 2 ∙ (−5.9 ∙ 𝐶𝑞𝑟 + 0.63)5
7 − 1 , 𝐶𝑞𝑟 =𝑄∙𝑅𝑒𝜑
0.2
2∙𝜋∙𝛺∙𝑟3 Eq. 51
𝐾 = [−8.85 ∙ 𝐶𝑞𝑟 + 0.5
𝑒(−1.45𝐶𝑞𝑟)]
54
Eq. 52
𝐾 =𝐾0
12.74𝑄
𝛺 ∙ 𝑏3 ∙ 𝑅𝑒𝜑0.2 ∙ (
𝑏𝑟
)
135
+ 1
Eq. 53
𝐾 = 0.032 + 0.32 × 𝑒−𝐶𝑞𝑟
0.028 Eq. 54
𝑘𝑠 =𝜋∙
8 , 휀 = 0.978 ∙ 𝑅𝑧
Eq. 55
𝑘𝑠𝑙 =100 ∙ 𝜈
(1 − 𝐾) ∙ 𝑟 ∙ 𝛺
Eq. 56
tan(𝛽) =𝑉𝜑
𝑉𝑟
Eq. 57
tan(𝛽) =𝑉𝜑
𝑉𝑧
Eq. 58
49
720𝑎∗ ∙ 𝑏 ∙ (1 − 𝐾𝑏)3 +
5
6𝐶𝑞 ∙ 𝐾𝑏 =
0.0225 ∙ 𝐺
𝑏14
∙𝐾𝑏
74
(1 − 𝐾𝑏)12
∙ [(𝑎∗ ∙1 − 𝐾𝑏
𝐾𝑏
+ 1)]38
Eq. 59
[5
6∙
𝐶𝑞
𝑅135
−49
240∙ 𝑎∗ ∙ 𝑏 ∙ (1 − 𝐾)2] 𝑅
𝑑𝐾
𝑑𝑅= 0.0225 ∙ {
[(𝑎∗)2 + 1]38
𝑏14
∙ (1 − 𝐾)54 −
(𝑎2 + 1)38
𝑓14
∙ 𝐾74}
−5
3∙
𝐶𝑞
𝑅135
∙ 𝐾 −1127
3600∙ 𝑎∗ ∙ 𝑏 ∙ (1 − 𝐾)3
Eq. 60
49
720𝑎∗ ∙ 𝑏 ∙ (1 − 𝐾𝑒)3 +
5
6𝐶𝑞 ∙ 𝐾𝑒 = 0.0225 ∙ 𝐺 ∙ (𝑎2 + 1)
38 ∙ (
𝐾𝑒7
𝑓𝑒
)
14
− 𝐶𝑎𝑚
Eq. 61
𝐾𝑝 =𝑉𝜑
𝛺 ∙ 𝑟= 𝐾ℎ ∙ 𝑥ℎ
2 ∙ 𝑥−2 Eq. 62
𝜆𝑇 ≥ 0.437 ∙ [1 − (𝐾ℎ ∙ 𝑥ℎ2)1.175]1.656 Eq. 63
𝐾𝑝 =𝑉𝜃,∞
𝛺 ∙ 𝑟= 𝐾ℎ,𝑒𝑓𝑓 ∙ 𝑥ℎ
2 ∙ 𝑥−2 Eq. 64
𝐾ℎ,𝑒𝑓𝑓
𝐾ℎ
= 1.053 − 0.062 ∙ 𝐾ℎ Eq. 65
𝐾ℎ,𝑒𝑓𝑓
𝐾ℎ
= 1 − 0.056 ∙ 𝐾ℎ Eq. 66
𝐹𝑎 = 𝐹𝑎𝑏 − 𝐹𝑎𝑓 Eq. 67
𝐹𝑎𝑓 = 𝜋 ∙ 𝑝𝑏 ∙ 𝑏2 − 𝐶𝐹𝑓 ∙ 𝜌 ∙ 𝛺2 ∙ 𝑏4 Eq. 68
𝐹𝑎𝑏 = 𝜋 ∙ 𝑝𝑏 ∙ (𝑏2 − 𝑎2)−𝐶𝐹𝑏 ∙ 𝜌 ∙ 𝛺2 ∙ (𝑏4 − 𝑎4) Eq. 69
XIX
𝜕𝑝
𝜕𝑟= 𝜌 ∙ 𝛺2 ∙ 𝐾2 ∙ 𝑟
Eq. 70
𝜕𝑝
𝜕𝑟= 𝜌 ∙ (
𝑣𝜑2
𝑟− 𝑣𝑟
𝜕𝑣𝑟
𝜕𝑟) = 𝜌 ∙ 𝐾2 ∙ 𝛺2 ∙ 𝑟 +
𝜌∙𝑄2
4∙𝜋2∙𝑠2∙𝑟3 Eq. 71
𝐶𝐹 = 9.96 ∙ 𝐶𝑎𝑚 + 0.039; 𝐶𝑎𝑚 = (𝐿
2∙𝜋∙𝑏5∙𝛺2) ∙ 𝑅𝑒1
5
Where L is the angular momentum which centripetal through-flow brings into the flow field. Eq. 72
𝐶𝑎𝑚 = −[1 − 𝜙 ∙ cot(𝛽)] ∙ 𝐶𝑞; 𝐶𝑞 = (𝑄
2∙𝜋∙𝑏3∙𝛺) ∙ 𝑅𝑒
1
5
Where 𝛽 =arctan(𝑉𝜑
𝑉𝑟).
Eq. 73
𝐶𝑎𝑚 = −(1 + 5 ∙ 𝐾𝑏) ∙𝐶𝑞
6; 𝐶𝑞 = (
𝑄
2∙𝜋∙𝑏3∙𝛺) ∙ 𝑅𝑒
1
5
Where 𝛽 =arctan(𝑉𝜑
𝑉𝑧).
Eq. 74
𝐶𝑀 = 3.68 ∙ 𝑅𝑒−12
Eq. 75
𝐶𝑀 = 0.146 ∙ 𝑅𝑒−15
Eq. 76
𝐶𝑀 = 1.935 ∙ 𝑅𝑒−12
Eq. 77
𝐶𝑀 = 0.982 ∙ (log10𝑅𝑒)−2.58 Eq. 78
𝐶𝑀 = 0.131 ∙ 𝑅𝑒−0.186 Eq. 79
𝐶𝑀 =𝜋
𝐺 ∙ 𝑅𝑒
Eq. 80
𝐶𝑀 = 0.0308 ∙ 𝐺−14 ∙ 𝑅𝑒−
14
Eq. 81
𝐶𝑀1 =𝜋
𝐺 ∙ 𝑅𝑒
Eq. 82
𝐶𝑀2 = 1.85 ∙ 𝐺1
10 ∙ 𝑅𝑒−12
Eq. 83
𝐶𝑀3 = 0.04 ∙ 𝐺−16 ∙ 𝑅𝑒−
14
Eq. 84
𝐶𝑀4 = 0.0501 ∙ 𝐺1
10 ∙ 𝑅𝑒−15
Eq. 85
𝐶𝑀 = 𝐶𝑀0 ∙ (1 + 13.9 ∙ 𝐾0 ∙ 𝜆𝑇 ∙ 𝐺−18)
Eq. 86
𝐶𝑀 = 𝐶𝑀𝑎 + 𝑐𝑀𝑏 Eq. 87 (a)
𝐶𝑀𝑎 = 0.146 ∙ 𝑅𝑒(−15
) ∙ 𝑥𝑐
235
Eq. 87 (b)
𝐶𝑀𝑏 = 0.0796 ∙ 𝑅𝑒(−1
5) ∙ {(1 − 𝑥𝑐
23
5 ) + 14.7 ∙ 𝜆𝑇 ∙ (1 − 𝑥𝑐2) + 90.4 ∙ 𝜆𝑇
2 ∙ [1 − 𝑥𝑐(−
3
5)]}
Eq. 87 (c)
𝑥𝑐 = 1.79 ∙ 𝜆𝑇(
513
)
Eq. 87 (d)
XX
𝐶𝑀 = 0.666 ∙ 𝐶𝐷 ∙ 𝑅𝑒−1 Eq. 88
𝐶𝑀𝑐 = 0.36 ∙ 𝛾(−14
) ∙ 𝐾74 ∙ (1 − 𝐾)
320 ∙ 𝑅𝑒(−
15
)
Eq. 89
𝛾 = [81 ∙ (1 + 𝛼2)
38
49 ∙ (23 + 37 ∙ 𝐾) ∙ 𝛼]
45
𝛼 = [2300 ∙ (1 + 8𝐾)
49 ∙ (1789 − 409 ∙ 𝐾)]
12
𝐾 = 0.087 ∙ 𝐾0 ∙ 𝑒[5.2∙(0.486−𝜆𝑇)−1]
𝐾0 = 0.49 − 0.57 ∙𝑠
𝑏
𝐶𝑀 = 0.52 ∙ 𝐶𝐷0.37 ∙ 𝑅𝑒−0.57 + 0.0028 Eq. 90
𝐶𝑀 = 0.108 ∙ (𝑘𝑠
𝑏)0.272
Eq. 91
𝐶𝑀−
12 = −5.37 ∙ log10 (
𝑘𝑠
𝑏) − 3.4 ∙ 𝐺
14
Eq. 92
𝑅𝑒 ∙ 𝐶𝑀−
12 ≈ 16000 ∙ (
𝑘𝑠
𝑏)
−1
10
Eq. 93
v𝜕v
𝜕𝑙= −
1
𝜌
𝑑𝑝
𝑑𝑙+ 𝜈
𝜕2v
𝜕𝑧𝑙2
Eq. 94
𝑇𝑎 =v𝑔𝑎𝑝 ∙ ∆𝑟𝑠𝑒𝑎𝑙
𝜈∙ √
∆𝑟𝑠𝑒𝑎𝑙
𝑟𝑔𝑎𝑝
=𝑅𝑒𝑝
2∙ √
∆𝑟𝑠𝑒𝑎𝑙
𝑟𝑔𝑎𝑝
Eq. 95
𝑅𝑒𝑝 =2 ∙ ∆𝑟𝑠𝑒𝑎𝑙 ∙ v𝑔𝑎𝑝
𝜈
Eq. 96
v𝑔𝑎𝑝 =2𝜋 ∙ 𝑟𝑔𝑎𝑝 ∙ 𝑛
60
Eq. 97
𝑅𝑒𝑔𝑎𝑝 = √(2 ∙ ∆𝑟𝑠𝑒𝑎𝑙 ∙ v𝑎𝑥𝑔𝑎𝑝
𝜈)2 + (
v𝑔𝑎𝑝 ∙ ∆𝑟𝑠𝑒𝑎𝑙
𝜈)2
Eq. 98
𝑄 = 𝛹 ∙ 𝐴 ∙ √2𝑔 ∙ ∆𝑝 Eq. 99
∆𝑁 = √𝑁𝑇2 + 𝑁𝐷
2; 𝑁𝑇 =√𝑛𝑇∙𝑛𝑀∙(𝑒𝑇∙𝑀𝑟)2
1.96∙√1000; 𝑁𝐷 =
√𝑛𝑇∙𝑛𝑀∙(𝑒𝐷∙𝑀𝑟)2
1.96∙√1000
Eq. 100
𝑦+ =𝑢∗ ∙ 𝑦
𝜈
Eq. 101
𝑢∗ = √𝜏𝑤
𝜌
Eq. 102
𝐶𝑓 =𝜏𝑤
12
∙ 𝜌 ∙ 𝑈2
Eq. 103
1
𝐶𝑓
12
= 1.7 + 4.15 ∙ log10(𝑅𝑒𝜑 ∙ 𝐶𝑓) Eq. 104
𝐶𝑓 = [2 ∙ log10((𝑅𝑒𝜑) − 0.65]−2.3 ; 𝑅𝑒𝜑 < 109 Eq. 105
𝐶𝑓 = (2.87 + 1.58 ∙ 𝑙𝑜𝑔10(𝑙/𝑘𝑠))−2.5 Eq. 106
XXI
𝑦 =𝑦+ ∙ 𝜈 ∙ √𝜌
√12
∙ 𝜌 ∙ 𝑈2 ∙ (2.87 + 1.58 ∙ log10(𝑙/𝑘𝑠))−2.5
Eq. 107
∆= |𝐶1.75 − 𝐶𝑥
𝐶1.75
| Eq. 108
𝐾𝑐~𝑐+1̅̅ ̅̅ ̅̅ ̅̅ ̅ = √
𝑝(𝑟𝑐) − 𝑝(𝑟𝑐+1) −𝜌 ∙ 𝑄2
8 ∙ 𝜋2 ∙ 𝑠2 (1
𝑟𝑐+12 −
1𝑟𝑐
2)
12
∙ 𝜌 ∙ 𝛺2 ∙ (𝑟𝑐2 − 𝑟𝑐+1
2)
Eq. 109
𝐾 = 0.97 ∙ [−8.5 ∙ 𝐶𝑞𝑟 + 0.5
𝑒(−1.45𝐶𝑞𝑟)]
54
Where −0.5 ≤ 𝐶𝑞𝑟 ≤ 0.03. Eq. 110
𝐾 = 0.97 ∙ 𝑒(600∙𝑘𝑠∙𝑟
𝑏2 )∙ [
−8.5 ∙ 𝐶𝑞𝑟 + 0.5
𝑒(−1.45𝐶𝑞𝑟)]
54
Where 0.018 ≤ 𝐺 ≤ 0.072;
𝑅𝑒 ≤ 3.17 × 106;
−5050 ≤ 𝐶𝐷 ≤ 0 and 𝑘𝑠 ≤ 58.9 𝜇m. Eq. 111
𝑝(𝑟) = 𝑝𝑏 + ∫ 𝜌 ∙ 𝐾2 ∙ 𝛺2 ∙ 𝑟𝑑𝑟𝑟
𝑏
+𝜌 ∙ 𝑄2
8 ∙ 𝜋2 ∙ 𝑠2(
1
𝑏2−
1
𝑟2)
Where ∫ 𝜌 ∙ 𝐾2 ∙ 𝛺2 ∙ 𝑟𝑑𝑟𝑟
𝑏≈
𝜌
2∙ 𝛺2 ∙ ∑ 𝐾𝑟𝑖+1
2 ∙ (𝑟𝑖2 − 𝑟𝑖+1
2)𝑐−1
0 ;
𝑟 = 𝑏 − 0.001 ∙ 𝑐 (m);
𝑟𝑖 − 𝑟𝑖+1 = −0.001 (m). Eq. 112
𝐶𝐹 = [6.6 ∙ 10−3 ∙ 𝑙𝑛 (𝑅𝑒) − 0.113] ∙ 𝑒(−1.2∙10−4∙𝐶𝐷) ∙ [0.122 ∙ 𝑙𝑛(𝐺) − 0.67] Eq. 113
𝐶𝐹 = [6.6 ∙ 10−3 ∙ ln(𝑅𝑒) − 0.113] ∙ 𝑒(−1.2∙10−4∙𝐶𝐷) ∙ [0.122 ∙ ln(𝐺) − 0.67] ∙ 𝑒(880∙𝑘𝑠𝑏
)
Where 0.018< 𝐺 ≤0.072;
𝑅𝑒 ≤ 3.17 × 106;
−5050 ≤ 𝐶𝐷 ≤ 0 and 𝑘𝑠 ≤ 58.9 𝜇m. Eq. 114
𝐶𝐹 =∫ 2𝜋 ∙ (𝑝𝑏 − 𝑝1) ∙ 𝑟𝑑𝑟
𝑟=𝑏
𝑟=𝑟𝑖1
𝜌 ∙ 𝜋 ∙ 𝛺2 ∙ 𝑏4
Eq. 115
𝐶𝑀𝑐𝑦𝑙 =2 ∙ |𝑀𝑐𝑦𝑙|
𝜌 ∙ 𝛺2 ∙ 𝑏5=
0.084 ∙ 𝜋 ∙ 𝑡
𝑏 ∙ (𝑙𝑔𝛺 ∙ 𝑏2
𝜐)1.5152
Eq. 116
𝐶𝑀3 = 0.011 ∙ 𝐺−16 ∙ 𝑅𝑒−
14 ∙ [𝑒(0.8∙10−4∙|𝐶𝐷|)]
Eq. 117
𝐶𝑀4 = 0.014 ∙ 𝐺1
10 ∙ 𝑅𝑒−15 ∙ [𝑒(0.46∙10−4∙|𝐶𝐷|)]
Eq. 118
𝐶𝑀3 = 0.32 ∙ 𝐺−16 ∙ 𝑅𝑒−
14 ∙ [𝑒(0.82∙10−4∙|𝐶𝐷|)] ∙ (
𝑘𝑠
𝑏)
0.272
Eq. 119
𝐶𝑀4 = 0.41 ∙ 𝐺1
10 ∙ 𝑅𝑒−15 ∙ [𝑒(0.46∙10−4∙|𝐶𝐷|)] ∙ (
𝑘𝑠
𝑏)
0.272
Eq. 120
𝐾 = 0.85 ∙ [0.032 + 0.32 × 𝑒(−𝐶𝑞𝑟
0.028)] Eq. 121
XXII
𝐾 = 0.85 ∙ 𝑒(600∙𝑘𝑠∙𝑟
𝑏2 )∙ [0.032 + 0.32 × 𝑒(
−𝐶𝑞𝑟
0.028)] Eq. 122
𝐾ℎ =𝑄
𝜋 ∙ 𝛺 ∙ (𝑑ℎ
2)3
∙ tan(90° − 𝛽)
Eq. 123
𝐾𝑝 =𝑄 ∙ 𝑥𝑎
2 ∙ 𝑡𝑎𝑛(90° − 𝛽)
𝜋 ∙ 𝑥2 ∙ 𝛺 ∙ (𝑑ℎ
2)3
Eq. 124
𝐶𝐹 = [6.6 ∙ 10−3 ∙ ln(𝑅𝑒) − 0.113] ∙ 𝑒(−1.6∙10−4∙𝐶𝐷) ∙ [0.122 ∙ ln(𝐺) − 0.67] Eq. 125
𝐶𝐹 = [6.6 ∙ 10−3 ∙ ln(𝑅𝑒) − 0.113] ∙ 𝑒(−1.6∙10−4∙𝐶𝐷) ∙ [0.122 ∙ ln(𝐺) − 0.67] ∙ 𝑒(880∙𝑘𝑠𝑏
)
Eq. 126
𝐶𝑀3 = 0.011 ∙ 𝐺−16 ∙ 𝑅𝑒−
14 ∙ [𝑒(10−4∙𝐶𝐷)]
Eq. 127
𝐶𝑀4 = 0.014 ∙ 𝐺1
10 ∙ 𝑅𝑒−15 ∙ [𝑒(0.6∙10−4∙𝐶𝐷)]
Eq. 128
𝐶𝑀3 = 0.32 ∙ 𝐺−16 ∙ 𝑅𝑒−
14 ∙ [𝑒(10−4∙𝐶𝐷)] ∙ (
𝑘𝑠
𝑏)
0.272
Eq. 129
𝐶𝑀4 = 0.41 ∙ 𝐺1
10 ∙ 𝑅𝑒−15 ∙ [𝑒(0.6∙10−4∙𝐶𝐷)] ∙ (
𝑘𝑠
𝑏)
0.272
Eq. 130
𝑝𝑟 =𝑃𝑥
𝑃0
Eq. 131
XXIII
Definitions of the Significant Non-dimensional Parameters
𝐶𝑎𝑚 = (𝐿
2 ∙ 𝜋 ∙ 𝑏5 ∙ 𝛺2) ∙ 𝑅𝑒
1
5
Centripetal through-flow (𝐶𝑞 < 0, 𝑄 < 0 m3/s):
𝐶𝑎𝑚 = −[1 − 𝜙 ∙ cot(𝛽)] ∙ (
𝑄
2 ∙ 𝜋 ∙ 𝑏3 ∙ 𝛺) ∙ 𝑅𝑒
1
5
Centrifugal through-flow (𝐶𝑞 > 0, 𝑄 > 0 m3/s):
𝐶𝑎𝑚 = −(1 + 5 ∙ 𝐾𝑏) ∙ (
𝑄
2 ∙ 𝜋 ∙ 𝑏3 ∙ 𝛺) ∙ 𝑅𝑒
15/6
𝐶𝐷 =�̇�
𝜇 ∙ 𝑏
𝐶𝐹 = ∫2 ∙ 𝜋 ∙ (𝑝
𝑏− 𝑝) ∙ 𝑟d𝑟
𝜌 ∙ 𝛺2 ∙ 𝑏4
𝑏
𝑎
𝐶𝑀 =2 ∙ |𝑀|
𝜌 ∙ 𝛺2 ∙ 𝑏5
𝐶𝑝 = 𝑝∗(𝑥 = 1) − 𝑝∗(𝑥); 𝑝∗ =𝑝
𝜌 ∙ 𝛺2 ∙ 𝑏2
𝐶𝑞 =𝑄
2 ∙ 𝜋 ∙ 𝛺 ∙ 𝑏3 ∙ 𝑅𝑒1
5
𝐶𝑞𝑟 =𝑄 ∙ 𝑅𝑒𝜑
0.2
2 ∙ 𝜋 ∙ 𝛺 ∙ 𝑟3
𝐸𝑘 =1
𝐺2 ∙ 𝑅𝑒
𝐺 =𝑠
𝑏
𝐾 =𝛺𝑓
𝛺
𝑛𝑠 =𝑛√𝑄
𝑒
𝐻3
4
𝑅𝑒 =𝛺 ∙ 𝑏2
𝑣
𝑅𝑒𝜑 =𝛺 ∙ 𝑟2
v
𝑉𝜑 =v𝜑
𝛺 ∙ 𝑏
𝑉𝑟 =v𝑟
𝛺 ∙ 𝑏
𝑉𝑧 =v𝑧
𝛺 ∙ 𝑏
𝑥 =𝑟
𝑏
∆𝑥 =∆𝑟
𝑏
𝜆𝑇 =𝐶𝐷
𝑅𝑒0.8
휁 =𝑧
𝑠
𝜑𝐺
=𝑄
𝜋 ∙ 𝛺 ∙ 𝑏3
1
Abstract
The leakage flow (centripetal or centrifugal through-flow) can be found in the side cavities between
the rotor and the stationary wall in nearly all kinds of radial pumps and turbines. The cavity flow
has a strong impact on the disk friction loss, leakage loss, which in this way influence the efficiency
of radial pumps and turbines. Many more effects are related to the side cavity flow, such as the
resulting axial force on the impeller and rotordynamic. To better understand the effects the flow in
rotor-stator cavities is investigated by means of analytical, numerical and experimental approaches
in this thesis.
In chapter 1 and chapter 2, the research status and the progress for the core swirl ratio, the axial
thrust coefficient and the moment coefficient are introduced.
In chapter 3, the design of the test rig is described. The uncertainties of the experimental parameters
are estimated. The experimental results from the test rig are also compared with those from
literature to show that the results from the test rig are reliable.
In chapter 4, the numerical simulation set-up is illustrated. To minimize the error, the selection of
a turbulence model and the generation of the mesh are accomplished. The simulation results are in
good agreement with those from the literature, indicating that the numerical simulation set-up is
reasonable.
In chapter 5, the experimental results are presented for the core swirl ratio, the axial thrust
coefficient and the moment coefficient. The former correlations for the core swirl ratio are modified
based on the pressure measurements of the author and are extended by introducing the impact of
surface roughness. The values of core swirl ratio deduced from the pressure measurements are in
good agreement with the simulation results. Correlations for the axial thrust coefficient are
determined which cover the impact of global Reynolds number, axial gap width, through-flow
coefficient, surface roughness for both centripetal and centrifugal through-flow. The experimental
results for the moment coefficient are also compared with those from the correlations in the
literature according to the flow regimes, where a large gap occurs. The gap is explained by the
difference of surface roughness. The former correlations therefore are modified by introducing the
surface roughness based on the torque measurements with rough disks. Some experimental results
are also provided to understand how the pre-swirl impacts the above mentioned parameters.
In chapter 6, two examples are presented on the applications of the results in this thesis. The first
example is to accomplish the geometry optimization of the rear chamber of a submersible multi-
stage slurry pump based on the flow pattern. The service life of the pump is dramatically improved
2
by around 30%. The second example is to predict the axial thrust for a deep-well pump. The axial
thrust from the correlation is also in good agreement with the experimental results. The applications
indicate that the results in this thesis should be reasonable.
All the results will provide a database for the calculation of the axial thrust and the frictional loss
in order to better design radial pumps and turbines.
3
1. Introduction
1.1 Significance of This Thesis
Nowadays, some radial turbomachines can achieve the efficiency around 80% with well-designed
impellers. To further improve the service life and the efficiency, the flow in rotor-stator cavities
becomes one of the major concerns. The flow in such cavities (between the rotating impeller and
the stationary wall) can be either radially inward (in single-stage pumps and turbines) or radially
outward (between the two adjacent stages in multi-stage pumps). In Fig. 1, the cross section of a
radial pump is sketched. The effect of leakage flow whose volumetric flow rate is noted as Q
through the sealing gap (centripetal through-flow) can be better understood by investigating the
cavity flow.
Fig. 1: Cross section of a centrifugal pump
The concerns in the process of radial pumps or turbines design mainly include two parts: impeller
design and rotor-stator cavity design, depicted in Fig. 2. The design of a rotor-stator cavity, which
is closely related to many practical problems such as axial thrust, leakage flow, etc., is quite
important. In the light of the results from the literature, the leakage flow can excessively impact
the axial thrust, which significantly reduces the service life of a turbomachine. In addition, the
prediction of the frictional loss when the leakage flow occurs also attracts extensive attention to
minimize the energy consumption. To meet the demands of industry, the sources of the axial thrust
and the frictional loss of the disk are investigated. The issue becomes more complicated concerning
the surface roughness and the pre-swirl. To close the knowledge gaps, this thesis tries to quantify
how the cavity flow impacts the axial thrust and the frictional loss of the disk.
Impeller (rotor)
Q
Centripetal through-flow
(leakage flow)
Front cavity
Back cavity
4
Fig. 2: Concerns during the design of a radial pump or turbine
1.2 Important Variables and Limitations of Previous Studies
1.2.1 Core Swirl Ratio
The core swirl ratio K (referring to the ratio of the angular velocity of the fluid 𝛺𝑓 to that of the
disk 𝛺 at a position half of the axial gap width) is utilized to show the dominant tangential motion
of the fluid. The distribution of the pressure along the rotor can be approximately determined based
on the core swirl ratio K. From the pressure distribution, the axial thrust acting on the rotor can be
predicted. The amounts of K are very sensitive to the through-flow, the surface roughness of the
rotor and the angular momentum, which are still not sufficiently investigated.
1.2.2 Axial Thrust
The forces determining the axial thrust (𝐹𝑎) of the impeller are presented in Fig. 3. As commonly
understood, the direction of the axial thrust is towards the suction side. The axial thrust 𝐹𝑎 mainly
includes four parts: the force on the front surface 𝐹𝑎𝑓 , the force on the back surface 𝐹𝑎𝑏 , the
impulse force at the impeller eye 𝐹𝑎𝑠 and the force on the impeller passage 𝐹𝑎𝑝. The parameters
𝐹𝑎𝑓 and 𝐹𝑎𝑏 can be calculated by the radial pressure distributions. The values of Re (global
Reynolds number), G (non-dimensional axial gap width), 𝐶𝐷 (through-flow coefficient) and 𝑘𝑠
(equivalent surface roughness) impact the thrust coefficient 𝐶𝐹 in a certain manner, which is up
to now still not precisely predictable.
Radial turbomachine design
Impeller design
Pressure head
Efficiency
Rotor-stator cavity design
Axial thrust
Frictional losses
Leakage losses
Kn
ow
led
ge
gap
5
Fig. 3: Sources of the axial thrust for a radial pump
1.2.3 Frictional Torque
To improve the efficiency of a turbomachine, the geometry of the side chambers should be carefully
designed. Daily and Nece [28] find that the moment coefficient 𝐶𝑀 (on a single surface) can be
predicted according to the flow regimes by classifying the tangential velocity profiles. They
examine the flow regimes and distinguish between four flow regimes on the basis of the measured
tangential velocity profiles. The typical profiles of tangential velocity and radial velocity for all
four flow regimes are shown schematically in Fig. 4. In radial pumps or turbines, the turbulent flow
regimes (regime III and regime IV) are more likely to occur.
Laminar flow (𝑅𝑒 ≤ 1.5 × 105) Turbulent flow (𝑅𝑒 > 1.5 × 105)
Regime I
Small axial gap
Regime II
Large axial gap
Regime III
Small axial gap
Regime IV
Large axial gap
Tan
gen
tial
vel
oci
ty
Rad
ial
vel
oci
ty
Fig. 4: Typical velocity profiles for the four flow regimes
𝐹𝑎𝑓 𝐹𝑎𝑏
𝐹𝑎𝑠 𝐹𝑎𝑝
𝐹𝑎
Wall Disk
Wall Disk
6
The distinguishing lines among the flow regimes for enclosed rotor-stator cavities are depicted in
Fig. 5 (Daily and Nece [28]). In most of the former studies, the frictional losses are predicted with
the correlations in Daily and Nece [28] according to the flow regimes in the 2D Daily&Nece
diagram for enclosed rotor-stator cavities. Kurokawa et al. [53][55] illustrate the impact of
through-flow on 𝐶𝑀 without investigating the impact of through-flow on the distinguishing lines.
Known the impact, more precise correlations for 𝐶𝑀 are also demanded for a rotor-stator cavity
with both through-flow and rough disks.
G
Fig. 5: Distinguishing lines for flow regimes without through-flow (2D Daily&Nece diagram [28])
1.3 Research Goals and Proposed Approach
In Table 1, a selection of the current research states of the flow in a rotor-stator cavity are listed. To
provide more confidence for calculating the values of K, CF and CM, the contents of this thesis
are depicted in Fig. 6.
Parameters: 𝐶𝐷 Re G 𝑘𝑠 Pre-swirl
K S [30][52][68] S [27][28][29] N [27][28][29] N [51][54] N [50][52][53]
𝐶𝐹 S [52][53] S [52][53] N [52][53] N [54] N [52][53][79]
𝐶𝑀 N [12][27][29][53][66] S [27][53][55] S [27][53][55] N [32][54] N [49][50][53]
Table 1: Current research states on flow in rotor-stator cavities (S: Sufficient, N: Not sufficient)
0
0.02
0.04
0.06
0.08
0.1
1.E+03 1.E+04 1.E+05 1.E+06 1.E+07
Regime II Regime IV
Regime III Regime I
Re 103 104 105 106 107
7
Fig. 6: Contents of this thesis
The main geometry of the rotor-stator cavity applied in this thesis is presented in Fig. 7. The front
chamber and the back chamber are separated by a disk in the middle. The front chamber is the test
region where the through-flow is imposed. The axial gap width of the front chamber s is changeable
while that of the back chamber 𝑠𝑏 is a fixed value (𝑠𝑏 = 8 mm).
Fig. 7: Main geometry of the test rig
The goals of this thesis are to:
1. Conduct the numerical simulation of the steady flow in a rotor-stator cavity with centripetal
or centrifugal through-flow;
2. Quantify the impacts of through-flow on K, 𝐶𝐹 and 𝐶𝑀 in a rotor-stator cavity with
through-flow for a smooth disk (𝑘𝑠 = 0.4 𝜇m);
3. Organize the 3D diagram distinguishing regime III and regime IV with a third axis through-
Co
nte
nts
Numerical simulation
Velocity distribution
3D Daily&Nece diagram
Pressure distribution
Axial thrust coefficient
Experiments
Pressure measurements
Core swirl ratio
Axial thrust coefficient
Axial thrust measurements
Axial thrust coefficient
Frictional torque measurements
Moment coefficient
Application of the results
Sand exclusion in a SMSP
Axial thrust in a DWP
New
dia
gra
ms
an
d c
orr
ela
tio
ns
Test region
Horizontal pipe
8
flow 𝐶𝐷;
4. Quantify the effect of surface roughness on K, 𝐶𝐹 and 𝐶𝑀 in a rotor-stator cavity with
centripetal or centrifugal through-flow;
5. Provide more results on how the pre-swirl impacts 𝐶𝐹 and 𝐶𝑀 in a rotor-stator cavity
with centripetal or centrifugal through-flow.
9
2. State of the Art
2.1 Basic Equations
The fundamental equations for any analytical analyses are the conservation laws for mass,
momentum and energy. In Will et al. [85][88], Eq. 1 is determined as the continuity equation for the
compressible, unsteady flow.
𝜕𝑝
𝜕𝑡+
𝜕(𝜌 ∙ v𝑖)
𝜕𝑥𝑖= 0
Eq. 1
The momentum equations predict the acceleration of a fluid particle to the surface and body forces
of the flow. As commonly understood, the change rate of the momentum per unit volume plus the
outflowing minus the inflowing momentum over the surface equals the sum of the forces acting on
a control volume element. The universal momentum equation for the compressible unsteady flow
is written in Eq. 2. The left hand side shows the inertia terms while the right hand side includes the
impact of external forces, pressure and friction [88].
𝜕(𝜌 ∙ v𝑖)
𝜕𝑡+
𝜕(𝜌 ∙ v𝑖 ∙ v𝑖)
𝜕𝑥𝑗= 𝑓𝑖 −
𝜕𝑝
𝜕𝑥𝑖+
𝜕𝜏𝑖𝑗
𝜕𝑥𝑗
Eq. 2
For compressible Newtonian fluids, the stress tensor is written based on the hypothesis of Stokes.
It is basically a generalization of Newton’s one dimensional shear stress approach (Schlichting
[81]):
𝜏𝑖𝑗 = 𝜇 ∙ (𝜕v𝑖
𝜕𝑥𝑗+
𝜕v𝑗
𝜕𝑥𝑖−
2
3∙ 𝛿𝑖𝑗 ∙
𝜕v𝑘
𝜕v𝑘)
Eq. 3
Due to the geometry of rotor-stator cavities, the correlations are better expressed with cylindrical
coordinates. For isothermal and incompressible flow with constant density, the energy equation can
be omitted. Without making any closure assumptions for the shear stresses, the full equations of
momentum with cylindrical coordinates are determined in Eq. 4~Eq. 6 by Will [88].
10
Radial momentum:
𝜌 ∙ (𝜕v𝑟
𝜕𝑡+ v𝑟 ∙
𝜕v𝑟
𝜕𝑟+
v𝜑
𝑟∙
𝜕v𝑟
𝜕𝜑−
v𝜑2
𝑟+ v𝑧 ∙
𝜕v𝑟
𝜕𝑧)
= 𝑓𝑟 −𝜕𝑝
𝜕𝑟+
1
𝑟∙
𝜕(𝑟 ∙ 𝜏𝑟𝑟)
𝜕𝑟+
1
𝑟∙
𝜕(𝜏𝑟𝜑)
𝜕𝜑−
𝜏𝜑𝜑
𝑟+
𝜕(𝜏𝑟𝑧)
𝜕𝑧
Eq. 4
Tangential momentum:
𝜌 ∙ (𝜕v𝜑
𝜕𝑡+ v𝑟 ∙
𝜕v𝜑𝜑
𝜕𝑟+
v𝜑
𝑟∙
𝜕v𝜑
𝜕𝜑+
v𝜑 ∙ v𝑟
𝑟+ v𝑧 ∙
𝜕v𝜑
𝜕𝑧)
= 𝑓𝜑 −1
𝑟∙
𝜕𝑝
𝜕𝜑+
1
𝑟2∙
𝜕(𝑟2 ∙ 𝜏𝜑𝑟)
𝜕𝑟+
1
𝑟∙
𝜕(𝜏𝜑𝜑)
𝜕𝜑+
𝜕(𝜏𝜑𝑧)
𝜕𝑧
Eq. 5
Axial momentum:
𝜌 ∙ (𝜕v𝑧
𝜕𝑡+ v𝑟 ∙
𝜕v𝑧
𝜕𝑟+
v𝜑
𝑟∙
𝜕v𝑧
𝜕𝜑+ v𝑧 ∙
𝜕v𝑧
𝜕𝑧)
= 𝑓𝑧 −𝜕𝑝
𝜕𝑧+
𝜕(𝜏𝑧𝑟)
𝜕𝑟+
𝜏𝑧𝑟
𝑟+
1
𝑟∙
𝜕(𝜏𝑧𝜑)
𝜕𝜑+
𝜕(𝜏𝑧𝑧)
𝜕𝑧
Eq. 6
The continuity equation for a rotor-stator cavity is given in Eq. 7 with cylindrical coordinates.
1
𝑟∙
𝜕(𝑟 ∙ v𝑟)
𝜕𝑟+
1
𝑟∙
𝜕v𝜑
𝜕𝜑+
𝜕v𝑧
𝜕𝑧= 0
Eq. 7
Several assumptions are made in order to simplify Eq. 4~Eq.7 by Will [88].
(1) Steady flow (𝜕
𝜕𝑡= 0);
(2) Axisymmetric flow (𝜕
𝜕𝜑= 0);
(3) No body force;
(4) No normal stresses due to viscosity;
(5) No shear stress due to gradients of the tangential velocity.
With above simplification, Eq. 4~Eq. 6 are valid for general flows even when they do not have
Newtonian viscosity. Neglecting the viscous terms 𝜕𝜏𝑖𝑗
𝜕𝑥𝑗 in Eq. 2, the equation is still non-linear since
it has the convective terms (Will [88]). Now Eq. 2 can be written as:
𝜕(𝜌 ∙ v𝑖)
𝜕𝑡+
𝜕(𝜌 ∙ v𝑖 ∙ v𝑖)
𝜕𝑥𝑗= 𝑓𝑖 −
𝜕𝑝
𝜕𝑥𝑖
Eq. 8
11
Likewise, the simplified momentum equations (Eq. 9~Eq. 11) and the continuity equation (Eq. 12) are
conducted by Will [88] for steady, axisymmetric flow in rotor-stator cavities.
Radial momentum:
𝜌 ∙ (v𝑟 ∙𝜕v𝑟
𝜕𝑟−
v𝜑2
𝑟+ v𝑧 ∙
𝜕v𝑧
𝜕𝑧) = −
𝜕𝑝
𝜕𝑟+
𝜕(𝜏𝑟𝑧)
𝜕𝑧
Eq. 9
Tangential momentum:
𝜌 ∙ (v𝑟 ∙𝜕v𝜑
𝜕𝑟+
v𝑟 ∙ v𝜑
𝑟+ v𝑧 ∙
𝜕v𝜑
𝜕𝑧) =
𝜕(𝜏𝜑𝑧)
𝜕𝑧
Eq. 10
Axial momentum:
𝜌 ∙ (v𝑟 ∙𝜕v𝑧
𝜕𝑟+ v𝑧 ∙
𝜕v𝑧
𝜕𝑧) = −
𝜕𝑝
𝜕𝑧+
𝜏𝑟𝑧
𝑟+
𝜕(𝜏𝑟𝑧)
𝜕𝑟
Eq. 11
Continuity:
𝜕v𝑟
𝜕𝑟+
v𝑟
𝑟+
𝜕v𝑧
𝜕𝑧= 0
Eq. 12
Neglecting the axial velocity, Will [88] conducts Eq. 13 as the tangential momentum equation for
the inviscid core region.
𝜌 ∙ (v𝑟 ∙𝜕v𝜑
𝜕𝑟+
v𝑟 ∙ v𝜑
𝑟) = 0
Eq. 13
In Will [88], Eq. 13 is also written in the following form:
v𝑟
𝑟∙
𝜕
𝜕𝑟(v𝜑 ∙ 𝑟) = 0
Eq. 14
Assuming that v𝑟 equals to the radial velocity in the central core, the radial pressure distribution
can be evaluated using the radial balance between the centrifugal force and the pressure forces,
written in Eq. 15:
𝜕𝑝
𝜕𝑟= 𝜌 ∙
v𝜑2
𝑟
Eq. 15
For a rotor-stator cavity with through-flow, the radial velocity conducts a radial transport of angular
momentum. The radial velocity v𝑟 should be therefore introduced in Eq. 15. The solution as a
12
potential swirl is assumed. Hence, the determination of the pressure distribution requires an
additional term taking into account, namely the radial convection (Will [88]), written in Eq. 16.
𝜕𝑝
𝜕𝑟= 𝜌(∙
v𝜑2
𝑟− v𝑟 ∙
𝜕v𝑟
𝜕𝑟)
Eq. 16
To predict the flow in both the front and the back cavities (see Fig. 1), a common method is found
on the basis of the integral boundary layer theory. The equations therefore have to be integrated in
the axial direction. In this approach, the integral relations are only fulfilled for integrated values
across the boundary layer thicknesses. Senoo and Hayami [75] assume the thickness of the disk
boundary layer to be twice that of the wall boundary layer. The solution is reported to be little
affected by this assumption.
The integral boundary layer method requires the specification of velocity profiles. The profiles are
assumed instead of being found as a part of the solution. Accordingly, the quality of the solution
depends on the assumed profiles. Nevertheless, the integral quantities such as the frictional torque
or the axial thrust can be well predicted by the method because the main velocity component is in
the circumferential direction. The following identities are valid for the simplified radial and
tangential momentum equations, given in Eq. 17 and Eq. 18 (Owen and Rogers [66]):
v𝑟 ∙𝜕v𝑟
𝜕𝑟−
v𝜑2
𝜕𝑟+ v𝑧 ∙
𝜕v𝑟
𝜕𝑧=
1
𝑟∙ [
𝜕
𝜕𝑟∙ (𝑟 ∙ v𝑟
2) +𝜕
𝜕𝑧∙ (𝑟 ∙ v𝑟 ∙ v𝑧) − v𝜑
2] Eq. 17
v𝑟 ∙𝜕v𝜑
𝜕𝑟−
v𝑟 ∙ v𝜑
𝑟+ v𝑧 ∙
𝜕v𝜑
𝜕𝑧=
1
𝑟2∙ [
𝜕
𝜕𝑟∙ (𝑟 ∙ v𝑟 ∙ v𝜑) +
𝜕
𝜕𝑧∙ (𝑟2 ∙ v𝑧 ∙ v𝜑)]
Eq. 18
The radial and the tangential momentum now can be written as:
𝜕
𝜕𝑟(𝑟 ∙ v𝑟) + 𝑟 ∙
𝜕
𝜕𝑧∙ (v𝑟 ∙ v𝑧) − v𝜑
2 = −𝑟
𝜌∙
𝜕𝑝
𝜕𝑟+
𝑟
𝜌∙
𝜕(𝜏𝑟𝑧)
𝜕𝑧
Eq. 19
𝜕
𝜕𝑟(𝑟2 ∙ v𝑟 ∙ v𝜑) +
𝜕
𝜕𝑧∙ (𝑟2 ∙ v𝑧 ∙ v𝜑) = −
𝑟2
𝜌∙
𝜕(𝜏𝜑𝑧)
𝜕𝑟
Eq. 20
These equations are now integrated in the axial direction for a control volume from the axial
coordinates 𝑧1 to 𝑧2. Considering a possible variation of the limits 𝑧1 and 𝑧2 with radius, the
application of the Leibniz rule for a general variable X yields Eq. 21 (Will [88]).
13
( ∫
𝜕𝑋(𝑟, 𝑧)
𝜕𝑟
𝑧2
𝑧1
𝑑𝑧) =𝜕
𝜕𝑟( ∫ 𝑋(𝑟, 𝑧)
𝑧2
𝑧1
𝑑𝑧) +𝜕𝑧1
𝜕𝑟∙ 𝑋(𝑟, 𝑧1 ) −
𝜕𝑧2
𝜕𝑟∙ 𝑋(𝑟, 𝑧2)
Eq. 21
Combining Eq. 21 with Eq. 19 and Eq. 20, the conducted integrated forms of the radial and the
tangential momentum equations are given in Eq. 22 and Eq. 23.
𝜕
𝜕𝑟(𝑟 ∙ ∫ v𝑟
2
𝑧2
𝑧1
𝑑𝑧) + 𝑟 ∙𝜕𝑧1
𝜕𝑟∙ v𝑟1
2 − 𝑟 ∙𝜕𝑧2
𝜕𝑟∙ v𝑟2
2 + 𝑟 ∙ v𝑟2 ∙ v𝑧2 − 𝑟 ∙ v𝑟1 ∙ v𝑧1
− ∫ v𝜑2𝑑𝑧 = −
𝑟
𝜌∙
𝑧2
𝑧1
∫𝜕𝑝
𝜕𝑟𝑑𝑧 +
𝑟
𝜌
𝑧2
𝑧1
∙ ∫𝜕(𝜏𝑟𝑧)
𝜕𝑧𝑑𝑧
𝑧2
𝑧1
Eq. 22
𝜕
𝑟2∙
𝜕
𝜕𝑟(𝑟2 ∙ ∫ v𝑟 ∙ v𝜑
𝑧2
𝑧1
𝑑𝑧) +𝜕𝑧1
𝜕𝑟∙ v𝑟1 ∙ v𝜑1 −
𝜕𝑧2
𝜕𝑟∙ v𝑟2 ∙ v𝜑2 + 𝑟 ∙ v𝑧2 ∙ v𝜑2 − v𝑧1 ∙ v𝜑1
=1
𝜌∙ ∫
𝜕(𝜏𝜑𝑧)
𝜕𝑧𝑑𝑧
𝑧2
𝑧1
Eq. 23
In the following equations, a plane rotor-stator cavity is assumed. In Zilling [90], it is mentioned
that the conical walls can be treated as the parallel walls for the inclination angles smaller than 12
degrees. For a plane rotor-stator system with 𝑧1 = 0 and 𝑧2 = 𝑠, the following two correlations
are conducted since the velocity components v𝑧 and v𝑟 are zero at the walls (no-slip wall
condition):
1
𝑟∙
𝜕
𝜕𝑟(𝑟 ∙ ∫ v𝑟
2
𝑠
0
𝑑𝑧) −1
𝑟∙ ∫ v𝜑
2
𝑠
0
𝑑𝑧 = −1
𝜌∙ ∫
𝜕𝑝
𝜕𝑟𝑑𝑧
𝑠
0
+1
𝜌∙ (𝜏𝑟𝑧𝑆 − 𝜏𝑟𝑧𝑅)
Eq. 24
1
𝑟2∙
𝜕
𝜕𝑟(𝑟2 ∙ ∫ v𝑟v𝜑
𝑠
0
𝑑𝑧) =1
𝜌∙ (𝜏𝜑𝑧𝑆 − 𝜏𝜑𝑧𝑅)
Eq. 25
Eq. 24 and Eq. 25 are the integral momentum relations expressed in the cylindrical coordinates for a
steady, incompressible flow field. They are crucial for most analytical flow models in the literature
(Kurokawa et al. [52][53][54], Baibikov and Karakhan’yan [7], Baibikov [8]). A further
solution requires the specification of velocity profiles and suitable expressions for the wall shear
stresses.
14
To predict the radial gradient of the core rotation, Will [88] correlates Eq. 26 by expanding Eq. 25
with 2𝜋 and by introducing K. The wall shear stress is predicted with Eq. 27 in Will [88].
𝑑𝐾
𝑑𝑟=
2 ∙ 𝜋 ∙ 𝑏
�̇� ∙ 𝛺∙ (𝜏𝜑𝑧𝑆 − 𝜏𝜑𝑧𝑅) −
2 ∙ 𝐾
𝑅
Eq. 26
𝜏 = 𝜆 ∙𝜌
8∙ v𝑚𝑒𝑎𝑛
2 Eq. 27
Where v𝑚𝑒𝑎𝑛𝑅 = 𝑟2 ∙ 𝛺2 ∙ (1 − 𝐾)2;
v𝑚𝑒𝑎𝑛𝑆 = 𝑟2 ∙ 𝛺2 ∙ 𝐾2.
Thus, Eq. 26 becomes:
𝑑𝐾
𝑑𝑟=
�̅�2
4 ∙ 𝜑𝐺∙ (𝜆𝑠 ∙ 𝐾2 − 𝜆𝑅 ∙ (1 − 𝐾)2) −
2 ∙ 𝐾
𝑅
Eq. 28
Where 𝜑𝐺 =𝑄
𝜋∙𝛺∙𝑏3.
For 𝜆𝑅 = 𝜆𝑆, the limits of Eq. 28 can be obtained by considering a vanishing small through-flow or
leakage (𝜑𝐺 → 0) and the opposite case of an infinite high leakage (𝜑𝐺 → ∞). The values of K are
given in Eq. 29 and Eq. 30 in Will [88].
Solid body rotation (forced vortex):
𝜑𝐺 → 0, 𝐾 = 0.5 Eq. 29
Potential swirl (free vortex):
𝜑𝐺 → ∞, 𝐾 ∙ 𝑥2 = 0.5 Eq. 30
According to Lomakin [62], a simplified flow model for the radial distribution of K is deduced
from Eq. 25 assuming a constant value along the axial gap width. The parameter v𝜑 is therefore
taken outside the integral sign. The idea is that the circumferential velocity is almost constant in
the core region. A significant change appears only within the thin boundary layers. The axial
distance with a significant change of v𝜑, compared to the region with an almost constant value,
can thus be neglected. This assumption restricts the model validity in case of stronger variations in
the velocity profiles, for example, in case of flow regime III. Numerical simulations as well as a
plenty of experimental investigations, however, show that this is an admissible assumption in most
cases. Velocity profiles have to be assumed in order to accomplish the integration across the axial
15
gap. In principle, variable limits are used to account for the geometrical changes of the cavity in
the radial direction. For example, Will [88] fragments the integral into:
∫ v𝑟v𝜑𝑧=𝑠
𝑧=0𝑑𝑧 = ∫ v𝑟v𝜑
𝑧=𝛿𝑅
𝑧=0𝑑𝑧+∫ v𝑟v𝜑
𝑧=𝑠−𝛿𝑠
𝑧=𝛿𝑅𝑑𝑧+∫ v𝑟v𝜑
𝑧=𝑠
𝑧=𝑠−𝛿𝑠𝑑𝑧
Eq. 31
Will [88] solves Eq. 31 by assuming that the value of v𝑟 in the central core is zero. From the results
of numerical simulation by Will [88], the largest value of 𝑉𝑟 is 0.08 at 𝑅𝑒 = 0.38 × 106 (𝑛 =
300 rpm), 𝐺 = 0.018 ( 𝑠 = 0.002 mm ) and 𝐶𝐷 = 5050 ( 𝑄 = 2 m3/h ) at 𝑥 = 0.4 . This
value is much smaller than 𝑉𝜑. The outer radius is found to be the dominant region with respect to
the frictional torque and the axial thrust. Hence, the simplification will not result in large errors. If
the radial velocity in the core is considered zero, the second term on the right hand side of Eq. 31
becomes zero.
2.2 Thickness of Boundary Layers
Almost all the flow models that emerge from the integral boundary layer theory require information
about the radial evolution of the boundary layer thickness along the radius of the rotating disk and
the stationary wall. The boundary layer thickness essentially impacts the rotation of the core and
the moment coefficient. In principle, smaller boundary layer thicknesses result in higher frictional
resistances for larger velocity gradients. Several different correlations for the boundary layer
thickness are determined in the literature. Some of them are mentioned in this thesis. Unfortunately,
no experimental data are available to validate these theoretical expressions.
Kurokawa and Sakuma [55] mention that a centrifugal through-flow favors the merging of the
boundary layers. More on, transition from the laminar flow to the turbulent flow is additionally
influenced by the through-flow. The reason is that the externally applied leakage is in a turbulent
state in general. Consequently, in this case, the whole flow in the cavity is likely to become entirely
turbulent. For a wide axial gap, an intermediate velocity establishes between the walls, which is
referred to as “core region”. Due to the dominant tangential movement, a radial pressure gradient
establishes with the maximum pressure value at the outer radius of the disk. In the core, the
centrifugal force and the pressure force can balance each other without any additional forces. This
state is usually known as radial equilibrium. At the stationary wall, the tangential velocity is
reduced to zero and therefore a radial inflow develops in the stator boundary layer caused by the
dominant pressure forces. Near the rotor surface, the dominant centrifugal forces result in a radial
outflow. Since the main fluid motion is in the circumferential direction, the radial velocity
16
component is noted as secondary flow. In Fig. 8, when the axial gap is large enough to allow a
formation of two boundary layers on the rotor and the stator (Will [88]), the most frequently
encountered flow pattern is depicted. The boundary layers can either be merged or not; the flow
can be laminar or turbulent. In case of flow regimes II or IV (separated disk boundary layer and
wall boundary layer), which is supposed to apply in the majority of practical applications (Gülich
[34], Hamkins [41]). The basic flow structure is indicated by Senoo and Hayami [75]: A boundary
layer on the rotor (IV) and the stator (I), a core region (III) and an intermediate layer (II) between
the stator boundary layer and the central core with a small radial outflow. Due to mass conservation,
an axial convection of fluid from the stator to the rotor takes place at small radius. Although the
axial velocity is very small, the axially convected angular momentum is not (Hamkins [41]). To
investigate the cavity flow, the thickness of both the disk boundary layer and the wall boundary
layer is one of the major concerns.
Fig. 8: Flow structure in an idealized rotor-stator cavity (left) and the velocity profiles for a wide
axial gap (replotted from Will [88])
In Daily et al. [27], the velocity profiles in the case of a centrifugal leakage flow are measured.
Based on the experimental results, an empirical correlation for the thickness of the disk boundary
layer is determined in Eq. 32.
𝛿𝑅 = 𝐵+ ∙𝑟
𝑅𝑒𝜑
15
∙ (1 − 𝐾)𝑚
Eq. 32
휁
𝑥
Roto
r
Sta
tor
𝑉𝑟
𝑉𝜑
I II III IV
I II III IV
17
In Eq. 32, 𝐵+ and m are two constants to fit the experimental results. This correlation is used with
slight modifications in all the flow models proposed by Kurokawa et al. [54][55]. Some of the
parameter combinations for the constants 𝐵+ and m are given in Table 2.
𝐵+ m
Daily et al. [27] 0.4 2
Kurokawa and Toyokura [54] 0.526 2
Kurokawa and Sakuma [55] 0.54 2.5
Table 2: Values of 𝑩+ and m from the literature
The thickness of the wall boundary layer can be predicted with Eq. 33 by Kurokawa et al. [52].
They also give the values of 𝑐 and 𝑐∗ based on the experimental results in Table 3.
𝛿𝑠 =𝑓 ∙ 𝑟
(𝑟2 ∙ 𝜔
𝜈 )15
Eq. 33
Where 𝑓 =1
𝑐∙𝐾∙ [𝑐∗ ∙ 𝑏 ∙ (1 − 𝐾)3 −
120
49∙
𝑄
2∙𝜋∙𝜔∙𝑟3 ∙ (𝑟2∙𝜔
𝜈)
1
5].
Kurokawa and Toyokura [53] 𝑐 = 0.374 𝑐∗ = 0.220
Table 3: Values of 𝒄 and 𝒄∗ by Kurokawa and Toyokura [53]
The volume flow rate passing through the cavity is defined as follows:
𝑄 = ∫ v𝑟
𝑧=𝑠
𝑧=0
𝑑𝑧
Eq. 34
The distribution of the stator boundary layer thickness can also be predicted from the continuity
equation, assuming a zero radial velocity in the core (Will [88]):
∫ v𝑟𝑆
𝑧=𝛿𝑠
𝑧=0
𝑑𝑧𝑠 + ∫ v𝑟𝑅
𝑧=𝛿𝑅
𝑧=0
d𝑧𝑅 =𝑄
2 ∙ 𝜋 ∙ 𝑟
Eq. 35
The radial velocity in the disk boundary layer and the wall boundary layer are noted as v𝑟𝑅 and
v𝑟𝑆, respectively. According to the assumption of the velocity distributions by Eq. 36 and Eq. 37 in
Kurokawa et al. [54][55], the profiles of radial velocity are shown in Fig. 9. For the successive
considerations, 𝑛1=2 is used since the profile appears to be closer to the actual flow physics in
Kurokawa et al. [54], while 𝑛1 equals to 1. The value of m is determined as 7 [54][55].
18
v𝑟𝑅 = 𝑎𝑅 ∙ (1 − 𝐾) ∙ 𝑟 ∙ 𝛺 ∙ (1 −
𝑧𝑅
𝛿𝑅)𝑛1 ∙ (
𝑧𝑅
𝛿𝑅)
1𝑚
Eq. 36
v𝑟𝑆 = −𝑎𝑆 ∙ 𝐾 ∙ 𝑟 ∙ 𝛺 ∙ (1 −
𝑧𝑠
𝛿𝑠)𝑛1 ∙ (
𝑧𝑠
𝛿𝑠)
1𝑚
Eq. 37
𝑉𝑟𝑅 𝑉𝑟𝑆
Fig. 9: Radial velocity profiles in dependence on Eq. 36 and Eq. 37 (replotted from Will [88])
The velocity factors 𝑎𝑅 and 𝑎𝑠 are determined from the empirical correlations given in
Kurokawa and Sakuma [55] based on the flow angle measurements for flow regime IV:
𝑎𝑅 = 1.18 ∙ (𝑅𝑒𝜑
105+ 2)−0.49
Eq. 38
𝑎𝑆 = 1.03 ∙ (𝑅𝑒𝜑
105+ 2)−0.387
Eq. 39
Will [88] determines Eq. 40 to predict the impact of through-flow on the thickness of the wall
boundary layer. Two constants of 0.304 and 0.408 are used in the equation. The amounts of c and
𝑐∗ are in Table 3. For centripetal through-flow, the volume flow rate is negative and the thickness
of the stator boundary layer therefore increases since the flow passes the cavity mostly in the
vicinity of the stator.
𝛿𝑠 = 0.304 ∙𝑐∗
𝑐∙
(1 − 𝐾)125
𝐾∙
𝑟
𝑅𝑒𝜑
15
−𝑄
0.408 ∙ 𝑐 ∙ 2 ∙ 𝜋 ∙ 𝑟2 ∙ 𝛺 ∙ 𝐾
Eq. 40
0
0.02
0.04
0.06
0.08
0.1
0.12
0 0.2 0.4 0.6 0.8 1
-0.12
-0.1
-0.08
-0.06
-0.04
-0.02
0
0 0.2 0.4 0.6 0.8 1
𝑧𝑅 𝛿𝑅⁄
𝑛1 = 1 𝑛1 = 2
1 − 𝑧𝑠 𝛿𝑠⁄
19
A common approach (e.g. Schultz-Grunow [74], Daily and Nece [27][28][29], Zilling [90],
Möhring [59], Senoo and Hayami [75], Kurokawa et al. [52] 55], Lauer [58]) is the
implementation of the Blasius law of the velocity distributions for pipe flow. Further evaluations
require information about the boundary layer thickness. Commonly, the dependencies for the free
disk are used. In contrast, in the laminar case, the boundary layer thickness of both the disk
boundary layer and the wall boundary layer under turbulent conditions are found to increase with
the 3/5 power of the radial coordinate (Dorfman [32][33]):
𝛿𝑅 = 𝛶𝑅 ∙ 𝑟35 ∙ (
𝜈
𝛺)
15
Eq. 41
𝛿𝑠 = 𝛶𝑠 ∙ 𝑟35 ∙ (
𝜈
𝛺)
15
Eq. 42
The proportionality factors (𝛶𝑅 and 𝛶𝑠 ) are now different from the free disk value (𝛶𝑅 = 𝛶𝑠 =
0.526). The friction factors for the rotor (𝜆𝑅) and the stator (𝜆𝑠) are in dependency upon the local
Reynolds number (Will [88]):
𝜆𝑅 =
0.18
𝐶𝑅∙ 𝑅𝑒𝜑
−15 ∙ (
1
1 − 𝐾)
14
Eq. 43
𝜆𝑆 =
0.18
𝐶𝑆∙ 𝑅𝑒𝜑
−15 ∙ (
1
𝐾)
14
Eq. 44
In Möhring [63], the constant 𝐶𝑅 is determined by comparing the impeller torque with the results
of Schultz-Grunow [74] and adjustment with his own results to:
𝐶𝑅 = 0.315 Eq. 45
The momentum of the secondary flow can be estimated by assuming equilibrium between the
torque on the walls and the momentum difference of the secondary flow in case of zero leakage
(Will [88]). An implementation is reasonable by increasing the frictional resistance of the casing
𝐶𝑆 (in Eq. 44) in Möhring [63]:
𝐶𝑆 = 𝐶𝑅 ∙ (
1 − 𝐾0
𝐾0)
74
Eq. 46
20
2.3 Core Swirl Ratio
2.3.1 Enclosed Rotor-Stator Cavity
In the case of solid body rotation (forced vortex) in an enclosed rotor-stator cavity, the core swirl
ratio only depends on the geometric magnitudes. Will [88] implements Eq. 47 to predict the impact
of G on 𝐾0.
𝐾0 =
1
1 + √1 + 5 ∙ 𝐺
Eq. 47
A lot of researches are accomplished to determine the values of 𝐾0. Some of them are listed below.
Radtke and Ziemann [78] show that the core swirl ratio decreases because of the higher ratio of
decelerating to accelerating surfaces with increasing radial coordinate.
Itoh et al. [40] measure the distributions of both the radial and the tangential velocities in an
enclosed rotor-stator cavity (𝐺 = 0.08) with a stationary shroud at 𝑅𝑒 = 106 using a hot-wire
anemometry. Cheah et al. [19] present the LDA and the hot-film velocimetry measurements in a
similar configuration for a wider cavity (𝐺 = 0.127 ). The local Reynolds number 𝑅𝑒𝜑 varies
from 0.3 × 106 to 1.6 × 106. In contrast to the results of Itoh et al. [40], they argue that the
turning at the outer radial shroud is mainly responsible for the entirely turbulent flow in the stator
boundary layer by causing upstream disturbances which destabilize the flow. Although the local
Reynolds number is confirmed to be the major parameter for the local velocity profiles, the authors
suggest that both the global Reynolds number and the axial gap width have to be taken into account
as well. At the lowest Reynolds number investigated, the flow along the rotor is almost completely
laminar while the wall boundary layer remains turbulent. As a consequence, the fluid which flows
in the axial direction from the stator to the rotor must relaminarize.
Watanabe et al. [83] study the influence of the fine spiral grooves on the frictional resistance of
an enclosed rotating disk in case of both the merged (flow regime III) and the separated (flow
regime IV) boundary layers. In practical cases, the bounding walls of rotor-stator cavities are
frequently inclined. The inclination angle influences the effective area in the formulation of the
torques on the rotating and the stationary wall.
Some of the typical values of 𝐾0 found in the literature are summarized in Table 4 for enclosed
rotor-stator cavities. The parameter 𝐺 is close to 0 for the disks with nearly infinite radius.
21
Authors Year 𝐾0 G Flow state From
Schultz-Grunow [74] 1935 0.512 ≈ 0
Turbulent
Theoretical
0.357 0.3 Experimental
Daily and Nece [28] 1960
0.46 0.0637
Experimental
0.454 Theoretical
0.444 0.102
Experimental
0.432 Theoretical
0.412 0.217
Experimental
0.388 Theoretical
0.46 0.051
Laminar
Experimental
0.504 Theoretical
0.44 0.102
Experimental
0.46 Theoretical
0.36 0.217
Experimental
0.386 Theoretical
Lance and Rogers [56] 1962 0.3
≈ 0
Simulation Cooper and Reshotko [20] 1975
0.3135
0.5
Turbulent Zilling [90] 1973 0.5 Theoretical
Möhring [63] 1979 0.5
Kurokawa [54] 1978 0.43 0.078 Laminar Experimental
Dijkstra and Van Heijst [27] 1983 0.313 ≈ 0 Turbulent
Simulation
Radtke and Ziemann [78] 1983 0.41 0.125 Experimental
Owen [63] 1989 0.382
0.069 Laminar Theoretical
0.426 Turbulent
Theoretical
Itoh et al. [40] 1992 0.42
0.08
Experimental 0.31 Laminar
Cheah et al. [19] 1994 0.4
0.127
Turbulent 0.35
Andersson and Lygren [5] 2006 0.4 0.1
Simulation 0.47 0.0632
Table 4: Core swirl ratio 𝑲𝟎 in the literature for an enclosed rotor-stator cavity (reorganized
from Will [88])
2.3.2 Impact of Through-Flow
Owen [63] develops a flow model to calculate the core swirl ratio and the frictional torque by
solving the Ekman equations for flow regime IV. In his configuration, the flow direction is usually
radially outward (e.g. Rabs et al. [70], Da Soghe et al. [26]). The Ekman equations are simplified,
linearized forms of the equations of motion. The functional relation for the core swirl ratio in case
of the laminar flow is finally obtained by using the continuity equation disregarding the outer
cylindrical wall:
𝐾 = 0.25 ∙ [−1 + √5 − 4 ∙ 𝜑𝐺
√𝑅𝑒𝜑
𝑥2]
2
Eq. 48
22
For an enclosed rotor-stator cavity, Eq. 48 implies 𝐾 = 𝐾0 = 0.382 , which is in very good
agreement with the measurements by Daily et al. [29]. The core swirl ratio is completely
suppressed for a critical value of the non-dimensional through-flow rate (𝜑𝐺 ∙ √𝑅𝑒𝜑
𝑥2 = 1 ) which
is approximately 16% bigger than the entrainment rate for the laminar flow over the free disk. It
appears when a leakage flow rate is sufficient to completely suppress the core rotation. For larger
leakage flow rates, a flow structure according to Stewartson [77] occurs. For the turbulent flow,
Owen [63] uses the integral method according to Van Kármán [47] to solve the non-linear
equations of motion for the disk boundary layer. This solution is combined with a simplified (linear)
approach on the stator side. Balancing the flow rates finally yields Eq. 49 for moderate through-flow
rates (𝜋 ∙ 𝜑𝐺 ∙ 𝑅𝑒𝜑2 ∙ 𝑥2.6 < 0.1 ). In the case of zero leakage, Eq. 49 implies 𝐾 = 𝐾0 = 0.426 ,
which is again in good agreement with the experiments by Daily et al. [29].
(1 − 𝐾)85 ∙ (1 − 0.51 ∙ 𝐾) − 0.638 ∙ 𝐾
45 = 0.25 ∙ [−1 + 4 ∙ 𝜑𝐺√
𝑅𝑒𝜑
𝑥2]
2
Eq. 49
Tests with the flow model in its original form indicate that in the case of lower entrance rotation
and large axial gap width, the computed curves show a slightly strong increase in the core rotation
compared to the available measurements performed by Lauer [58]. The decelerating effect of a
radial shroud onto the core rotation, especially at the outer radius, has yet not been accounted for
in the present flow model. The basic formulation of the model is derived both from the general
Navier-Stokes equations as well as from the principle of conservation of the angular momentum
for a small cylindrical volume element, given in Eq. 50 (see next page). In the latter case, it is obvious
that the influence of an outer casing is not taken into consideration. The overall decelerating
influence of an outer (stationary) shroud on the core rotation can be observed for the enclosed
rotating disk where usually both the measured and the computed values are markedly below the
theoretical maximum of 0.5. If a centripetal through-flow is present, the shear stress created by the
interaction of the entering leakage with the shroud gives rise to an additional “dynamic stress”
whose influence decreases at small radius. To include both effects in the flow model, a correction
function 𝑓∗ for the friction factor of the stator is introduced.
𝑑𝐾
𝑑𝑅=
𝑅2
4 ∙ 𝜑𝐺∙ (𝑓∗ ∙ 𝜆𝑠 ∙ 𝐾2 − 𝜆𝑅 ∙ (1 − 𝐾)2) −
2 ∙ 𝐾
𝑅
Eq. 50
Where 𝑓∗ = 1 + (𝑠
𝑏+𝑙1−𝑎+ 5 ∙ 𝑅4 ∙ |1 −
𝐾
0.58|
6
5).
23
Zilling [90] and Lauer [58] account for a torque resulting from a turbulent shear stress acting on
the lateral area of the control volume. Möhring [63] argues that this effect can be neglected because
the velocity gradient within the boundary layers in the axial direction (which determines for
example the wall frictional torque) is much greater than the velocity gradient in the radial direction
responsible for the turbulent shear stress.
2.3.2.1 With Centripetal Through-Flow (𝑸 < 𝟎 𝒎𝟑/𝒔)
In centrifugal pumps, the wear-ring leakage is driven by the pressure difference through the sealing
gap. The through-flow can also affect the boundaries between the different flow regimes by Daily
and Nece [24] because of its influence on the boundary layer thickness. For a rotor-stator cavity
with through-flow, the values of K are greatly influenced by the amount of through-flow. In this
thesis, the amounts of Q, 𝐶𝐷 and 𝐶𝑞𝑟 are negative for centripetal through-flow.
Altmann [1][2] measures the radial pressure distribution and determines the corresponding core
rotation distribution from the assumption of radial equilibrium between the pressure and the
centrifugal forces in the core region. In the case of a centripetal through-flow, a strong
magnification of the core rotation towards the center is observed. In the opposite flow direction,
the contrary situation prevails. The values of K decrease with increasing radial coordinate.
Using a two-component LDA system, Poncet et al. (2005) determine Eq. 51 to evaluate the core
swirl ratio K with centripetal through-flow when −0.2 ≤ 𝐶𝑞𝑟 ≤ 0.035 . Debuchy et al. [30]
determine Eq. 52 to calculate the amounts of K for a wider range −0.5 ≤ 𝐶𝑞𝑟 ≤ 0.035. The results
are compared in Fig. 10.
𝐾 = 2 ∙ (−5.9 ∙ 𝐶𝑞𝑟 + 0.63)5
7 − 1 , 𝐶𝑞𝑟 =𝑄∙𝑅𝑒𝜑
0.2
2∙𝜋∙𝛺∙𝑟3 Eq. 51
𝐾 = [
−8.85 ∙ 𝐶𝑞𝑟 + 0.5
𝑒(−1.45𝐶𝑞𝑟)]
54
Eq. 52
24
Fig. 10: Comparison of results from Eq. 51 and Eq. 52
2.3.2.2 With Centrifugal Through-Flow (𝑸 > 𝟎 𝒎𝟑/𝒔)
In a rotor-stator cavity with centrifugal through-flow, the flow can be classified into three flow
types: Batchelor type flow (regime IV), Couette type flow (regime III) and Stewartson type flow.
The main profiles of the non-dimensional tangential velocity 𝑉𝜑 and the non-dimensional radial
velocity 𝑉𝑟 for Batchelor type flow (Batchelor [10]), Couette type flow (Batchelor [10]) and
Stewartson type flow (Poncet et al. [68]) are shown in Fig. 11. Based on the experimental results
from Poncet et al. [68], the converting point of the rotation dominant flow (Batchelor or Couette
type flow) and the through-flow dominant flow (Stewartson type flow) is at 𝐶𝑞𝑟 ≈ 0.035.
Rotation dominant 𝐶𝑞𝑟 ≤ 0.035 Through-flow dominant 𝐶𝑞𝑟 > 0.035
Batchelor type (Regime IV) Couette type (Regime III) Stewartson type
Fig. 11: Velocity profiles for Batchelor type flow, Couette type flow and Stewartson type flow
Daily et al. [29] correlate Eq. 53 for the radial evolution of the core rotation in case of centrifugal
through-flow with zero entrance rotation. The equation is in good agreement with the
measurements of tangential velocity. 𝐾0 can be assumed to be 0.5 since Daily et al. [29] do not
find a unique dependence on the axial gap width G.
0
1
2
3
-0.5 -0.4 -0.3 -0.2 -0.1 0
𝐶𝑞𝑟
Eq. 51
Eq. 52
Disk Wall
x
𝑉𝜑
Disk
x
Wall
𝑉𝜑
𝑉𝑟
휁
x 𝑉𝑟
휁
휁
Disk Wall
𝑉𝑟
𝑉𝜑
K
25
𝐾 =𝐾0
12.74𝑄
𝛺 ∙ 𝑏3 ∙ 𝑅𝑒𝜑0.2 ∙ (
𝑏𝑟)
135
+ 1
Eq. 53
Kurokawa and Sakuma [55] mention that a radially outward through-flow favors the merging of
the boundary layers. Transition from the laminar to the turbulent flow is additionally influenced by
the through-flow because in general, the externally applied leakage is in a turbulent state.
Consequently, in this case, the whole flow in the cavity likely becomes entirely turbulent.
Bayley and Owen [13] investigate the influence of a centrifugal through-flow on the flow structure
between a rotating and a stationary disk (𝐺 = 0.008, 𝐺 = 0.03). The system is open towards the
atmosphere at the outer radius. The boundary layer equations are solved numerically using a finite
difference method and the flow is treated as turbulent for 𝑅𝑒𝜑 > 3 × 105. The externally applied
flow is without swirl. Consequently, the measured torque of the disk increases with increasing flow
rate because the core swirl ratio decreases. This agrees with the principal relations found by
Truckenbrodt [82] in case of a free disk subjected to an impinging external flow. From the results
of a two-component LDA system, Poncet et al. (2005) derive Eq. 54 for Stewartson type flow. The
results from the Eq. 51, Eq. 52 and Eq. 54 are depicted in Fig. 12.
Stewartson type flow (𝐶𝑞𝑟 > 0.035):
𝐾 = 0.032 + 0.32 × 𝑒−𝐶𝑞𝑟
0.028 Eq. 54
K K (Magnification of abscissa in the frame in the
left figure)
(a) Comparison of the results (b) Near the converting point
Eq. 51 Eq. 52 Eq. 54
Fig. 12: Results for K from different equations for centrifugal through-flow
0
0.1
0.2
0.3
0.4
0.5
0 0.1 0.20
0.1
0.2
0.3
0.4
0.5
0 0.02 0.04 0.06 0.08 0.1
𝐶𝑞𝑟 𝐶𝑞𝑟
26
2.3.3 Impact of Surface Roughness in an Enclosed Rotor-Stator Cavity
The measured surface roughness 𝑅𝑧 is the average value of five peaks and five valleys. It is a
challenging work to introduce the measured 𝑅𝑧 into Eq. 51, Eq. 52 and Eq. 54. Based on the results
of Adams et al. [81], a directly use of the measured surface roughness is not adequate. A conversion
of the measured surface roughness 𝑅𝑧 to the equivalent sand-grain roughness 𝑘𝑠 can be done
using Eq. 55.
𝑘𝑠 =𝜋∙
8 , 휀 = 0.978 ∙ 𝑅𝑧
Eq. 55
Schlichting [22] determines Eq. 56 to predict the transition point between a hydraulic smooth disk
and a disk in the transition zone, noted as 𝑘𝑠𝑙. In this thesis, 𝑘𝑠𝑙 ≈ 6 𝜇m is applied.
𝑘𝑠𝑙 =
100 ∙ 𝜈
(1 − 𝐾) ∙ 𝑟 ∙ 𝛺
Eq. 56
In Kurokawa et al. [54], the impact of surface roughness on the values of K is investigated. The
results show that K becomes larger towards the outer radius with rougher disks when 𝑘𝑠𝑙 < 𝑘𝑠 ≤
56 𝜇m (the disks are in the transition zone). Based on their velocity measurements, the variations
of K towards the outer radius of the shaft are depicted in Fig. 13.
Fig. 13: Experimental results of K along the radius of disk for 𝑮 = 𝟎. 𝟎𝟑𝟏, 𝑹𝒆 = 𝟑. 𝟏 × 𝟏𝟎𝟔 and
𝑪𝑫 = 𝟎 by Kurokawa et al. [54]
2.3.4 Impact of Pre-Swirl
In a turbomachine, the entering leakage flow normally contains a significant amount of angular
momentum (from the impeller outflow). Kurokawa and Toyokura [53], for example, show that
the angular momentum flux (instead of the volume flow rate) has the most significant effect on the
𝑘𝑠
=0
.6 𝜇
m
0.4
0.5
0.6
0.7
0.8
0.9
1
0.4 0.5 0.6
K
x
27
axial thrust for centripetal through-flow.
One major issue in this thesis is the specification of proper boundary conditions, especially with
respect to the swirl component of the leakage flow. Guide vanes or swirlers, oriented either in the
radial direction (centripetal through-flow) or in the axial direction (centrifugal through-flow) are
used in the experiments in this thesis to deflect the fluid in the circumferential direction. Since
insufficient information about the respective flow magnitudes are given in the literature, the
circumferential velocity component must be estimated from the geometrical outlet angle of the
guide vanes (assuming that the flow angle equals to the geometrical vane outlet angle). The
meridional component is calculated from the continuity equation disregarding the obstruction by
the guide vanes. However, the flow angle does not correspond to the geometrical angle of the
profile and it further varies in the axial direction. The velocity triangles when the fluid leaves the
inlet swirlers are plotted in Fig. 14, for centripetal (Fig. 14 (a)) and centrifugal (Fig. 14 (b)) through-
flow.
(a) Centripetal (24 passages) (b) Centrifugal (four swirlers)
Fig. 14: Velocity triangles for the pre-swirl through-flow
𝛽
View along
“A” direction
(Ignoring the
thickness of
the swirler)
v𝜑
𝛽
“A” direction
Swirler
Horizontal pipe
v𝑧
28
According to Fig. 14, the tangential velocities are calculated from the following relationships:
Centripetal through-flow:
tan(𝛽) =
𝑉𝜑
𝑉𝑟
Eq. 57
Centrifugal through-flow:
tan(𝛽) =
𝑉𝜑
𝑉𝑧
Eq. 58
Kurokawa et al. [52] derive Eq. 59 to predict the values of 𝐾𝑏 (K at 𝑟 = 𝑏). 𝐶𝑞 is the through-
flow rate coefficient.
49
720𝑎∗ ∙ 𝑏 ∙ (1 − 𝐾𝑏)3 +
5
6𝐶𝑞 ∙ 𝐾𝑏 =
0.0225 ∙ 𝐺
𝑏14
∙𝐾𝑏
74
(1 − 𝐾𝑏)12
∙ [(𝑎∗ ∙1 − 𝐾𝑏
𝐾𝑏+ 1)]
38
Eq. 59
In Kurokawa et al. [52], the distribution of K along the radius is determined with Eq. 60 (for
definition of f see Eq. 33).
[5
6∙
𝐶𝑞
𝑅135
−49
240∙ 𝑎∗ ∙ 𝑏 ∙ (1 − 𝐾)2] 𝑅
𝑑𝐾
𝑑𝑅= 0.0225 ∙ {
[(𝑎∗)2 + 1]38
𝑏14
∙ (1 − 𝐾)54 −
(𝑎2 + 1)38
𝑓14
∙ 𝐾74}
−5
3∙
𝐶𝑞
𝑅135
∙ 𝐾 −1127
3600∙ 𝑎∗ ∙ 𝑏 ∙ (1 − 𝐾)3
Eq. 60
2.3.4.1 Centripetal Pre-Swirl Through-Flow
In a rotor-stator cavity with centripetal pre-swirl through-flow, the core rotates faster than the disk
at lower radius which results in a creation of an attachment line where the fluid and the rotor rotate
at the same speed. Outgoing from this point, the dominant pressure gradient drives the fluid radially
inwards even in the rotor boundary layer for decreasing radius.
Based on the measured velocity profiles, Radtke and Ziemann [71] give a qualitative streamline
pattern of the flow structure when the core rotates faster than the disk (𝐾 > 1). Two counter rotating
vortices are formed by radial outflow in the core region, shown in Fig. 15 (a). Owen and Rogers
[66] also discuss this issue. Their theoretical streamline pattern dose not show any sign of positive
radial outflow. The fluid is directed radially inwards along the complete gap, illustrated in Fig. 15
(b). This is also confirmed by the LDA measurements by Poncet et al. [72].
29
(a) (b)
Fig. 15: Flow structure in the case of 𝑲 > 𝟏
Kurokawa et al. [52] correlate Eq. 61 to associate the value of 𝐾𝑒 (value of K at the entrance) with
𝐶𝑎𝑚 and 𝐶𝑞. 𝐶𝑎𝑚 is the coefficient of the inlet angular momentum in the through-flow.
49
720𝑎∗ ∙ 𝑏 ∙ (1 − 𝐾𝑒)3 +
5
6𝐶𝑞 ∙ 𝐾𝑒 = 0.0225 ∙ 𝐺 ∙ (𝑎2 + 1)
38 ∙ (
𝐾𝑒7
𝑓𝑒)
14
− 𝐶𝑎𝑚 Eq. 61
2.3.4.2 Centrifugal Pre-Swirl Through-Flow
In a rotor-stator cavity with centrifugal pre-swirl through-flow, the flow structure depends
principally on the turbulent flow parameters 𝜆𝑇 and 𝐾𝑒 (Owen and Rogers [66]). Karabay et
al. [49] show that, at sufficiently large amounts of 𝜆𝑇, free vortex flow occurs in the core outside
the boundary layer on the rotating surface. Under these conditions, 𝑉𝜑 is in proportion to 𝑟−1,
where 𝑉𝜑 is the tangential component of the velocity. 𝐾𝑝 follows Eq. 62 for an ideal system with
no losses. The test rig is depicted in Fig. 16. The pre-swirl nozzles instead of inlet swirlers at the
inlet are designed to generate the pre-swirl through-flow. The parameter 𝑥ℎ is the non-
dimensional radial coordinate at the radius of pre-swirl nozzle.
𝐾𝑝 =
𝑉𝜑
𝛺 ∙ 𝑟= 𝐾ℎ ∙ 𝑥ℎ
2 ∙ 𝑥−2 Eq. 62
𝐾 > 1
Pre-swirl
vanes
Q Q
Cas
ing
Cas
ing
Dis
k
Shaft
𝐾 > 1 Q
Q
Dis
k
Cas
ing
30
Fig. 16: Sketch of the rotor-stator cavity (redrawn from Karabay et al. [49])
In the cover-plate system where 𝐾ℎ > 1, Karabay et al. [49] show that free vortex flow will take
place throughout the cavity when:
𝜆𝑇 ≥ 0.437 ∙ [1 − (𝐾ℎ ∙ 𝑥ℎ2)1.175]1.656 Eq. 63
In practical cover-plate systems, there are losses between the outlet of the pre-swirl nozzles and the
inlet to the rotor-stator cavity. If 𝐾ℎ > 1, viscous effects cause a loss of angular momentum of the
fluid, reducing the effective swirl ratio. Karabay et al. [49] determine Eq. 64 to correlate 𝐾𝑝 with
𝐾ℎ,𝑒𝑓𝑓 (effective pre-swirl ratio at 𝑥 = 𝑥ℎ).
𝐾𝑝 =
𝑉𝜃,∞
𝛺 ∙ 𝑟= 𝐾ℎ,𝑒𝑓𝑓 ∙ 𝑥ℎ
2 ∙ 𝑥−2 Eq. 64
Karabay et al. [49] find that the effects of 𝑅𝑒𝜑 and 𝜆𝑇 on 𝐾ℎ,𝑒𝑓𝑓 are negligible. They proposed
Eq. 65 for the computed values of 𝐾ℎ,𝑒𝑓𝑓. Based on the experimental results, they modify Eq. 65 into
Eq. 66, compared in Fig. 17.
𝐾ℎ,𝑒𝑓𝑓
𝐾ℎ= 1.053 − 0.062 ∙ 𝐾ℎ
Eq. 65
𝐾ℎ,𝑒𝑓𝑓
𝐾ℎ= 1 − 0.056 ∙ 𝐾ℎ
Eq. 66
Blade-cooling passage
Pre-swirl nozzles
Q
Dis
k
Cas
ing
Cas
ing
Shaft
Q
31
𝐾ℎ,𝑒𝑓𝑓
Fig. 17: Variation of 𝑲𝒉,𝒆𝒇𝒇 versus 𝑲𝒉 by Karabay [49]: Solid line: Eq. 65; Dashed line: Eq.
66; Point: Exp
2.4 Axial Thrust
The difference of the forces on both sides of the disk is the main source for the axial thrust (𝐹𝑎),
calculated with Eq. 67 (direction see Fig. 3). 𝐹𝑎𝑓 (calculated with Eq. 68) and 𝐶𝐹𝑓 respectively
represent the force and the thrust coefficient on the front surface of the disk (in the front chamber,
shown in Fig. 7), while 𝐹𝑎𝑏 (calculated with Eq. 69) and 𝐶𝐹𝑏 are those on the back surface of the
disk (in the back chamber). The parameters a and 𝑝𝑏 represent the radius of the hub (see Fig. 7)
and the pressure at 𝑥 = 1 , respectively. The back chamber (𝐺 = 0.072 ), shown in Fig. 7, is
supposed to be an enclosed cavity. The values of 𝐶𝐹𝑏 are obtained when 𝐶𝐷 = 0 and the axial
gaps of both cavities have the same width for different Re (under that condition 𝐶𝐹𝑓 = 𝐶𝐹𝑏). After
obtaining those amounts, the values of 𝐶𝐹𝑓 with different amounts of 𝐶𝐷 can be calculated.
𝐹𝑎 = 𝐹𝑎𝑏 − 𝐹𝑎𝑓 Eq. 67
𝐹𝑎𝑓 = 𝜋 ∙ 𝑝𝑏 ∙ 𝑏2 − 𝐶𝐹𝑓 ∙ 𝜌 ∙ 𝛺2 ∙ 𝑏4 Eq. 68
𝐹𝑎𝑏 = 𝜋 ∙ 𝑝𝑏 ∙ (𝑏2 − 𝑎2)−𝐶𝐹𝑏 ∙ 𝜌 ∙ 𝛺2 ∙ (𝑏4 − 𝑎4) Eq. 69
To get an estimate of the expected forces, the simplified radial equilibrium, presented in Eq. 70, is
used in Will [88].
𝜕𝑝
𝜕𝑟= 𝜌 ∙ 𝛺2 ∙ 𝐾2 ∙ 𝑟
Eq. 70
0
0.5
1
1.5
2
2.5
3
3.5
4
0 1 2 3 4 5 6
𝐾ℎ
32
Will et al. [85][86][87] determine Eq. 71 to evaluate the pressure distribution along the radius of the
disk for the incompressible, steady flow. It is obtained directly from the radial momentum equation
when the turbulent shear stress is neglected. In a rotor-stator cavity, the cross sectional area changes
along the radius. Consequently, the pressure must also change since the velocity changes in the
radial direction according to the continuity equation with a known pressure distribution along the
radius, the axial thrust at the surface of the rotor can be estimated by integrating the following
equation:
𝜕𝑝
𝜕𝑟= 𝜌 ∙ (
𝑣𝜑2
𝑟− 𝑣𝑟
𝜕𝑣𝑟
𝜕𝑟) = 𝜌 ∙ 𝐾2 ∙ 𝛺2 ∙ 𝑟 +
𝜌∙𝑄2
4∙𝜋2∙𝑠2∙𝑟3 Eq. 71
Stepanoff [79] investigates the resulting axial force in centrifugal pumps. For theoretical analysis,
it is assumed that the fluid rotates as a solid body with a fixed value for the core rotation factor
(𝐾 = 0.5). The assumption of a constant core rotation is nowadays proved as inefficient in a rotor-
stator cavity with through-flow.
Kurokawa et al. [52] derive Eq. 72 to estimate the thrust coefficient 𝐶𝐹 for the case of 𝐺 = 0.05.
𝐶𝐹 = 9.96 ∙ 𝐶𝑎𝑚 + 0.039; 𝐶𝑎𝑚 = (𝐿
2∙𝜋∙𝑏5∙𝛺2) ∙ 𝑅𝑒1
5 Eq. 72
Where L is the angular momentum which centripetal through-flow brings into the flow field.
Kurokawa et al. [53] determine Eq. 73 and Eq. 74 to calculate 𝐶𝑎𝑚 in a radial pump. The parameter
𝛽 is shown in Fig. 14. According to their results, the values of 𝜙 range from 0.1 to 0.15.
Centripetal through-flow (𝐶𝑞 < 0, 𝑄 < 0 m3/s):
𝐶𝑎𝑚 = −[1 − 𝜙 ∙ cot(𝛽)] ∙ 𝐶𝑞; 𝐶𝑞 = (𝑄
2∙𝜋∙𝑏3∙𝛺) ∙ 𝑅𝑒
1
5 Eq. 73
Where 𝛽 =arctan(𝑉𝜑
𝑉𝑟).
Centrifugal through-flow (𝐶𝑞 > 0, 𝑄 > 0 m3/s):
𝐶𝑎𝑚 = −(1 + 5 ∙ 𝐾𝑏) ∙ 𝐶𝑞/6; 𝐶𝑞 = (𝑄
2∙𝜋∙𝑏3∙𝛺) ∙ 𝑅𝑒
1
5 Eq. 74
Where 𝛽 =arctan(𝑉𝜑
𝑉𝑧).
33
2.5 Moment Coefficient
2.5.1 The Free Disk
To study the flow in a rotor-stator cavity, a starting point is the flow near the free disk. The general
flow structure of the free disk problem is shown schematically in Fig. 18.
Fig. 18: Flow structure around a free disk (According to Schlichting and Gersten [73])
The frictional torque and the axial thrust are essentially influenced by the flow conditions at the
outer radius. Zilling [90] and Lauer [58] account for a torque resulting from a turbulent shear stress
acting on the lateral area of the control volume, shown in Fig. 19.
Fig. 19: Torques on an annular volume element
𝑉𝑍
𝑉𝑟
𝑉𝜑
𝑧
𝑥
𝑦
Disk
𝑧 r x
y
𝑟 + 𝑑𝑟
𝛺
𝛺
Disk Disk Fluid
34
The flow created by a rotating disk is three dimensional from inception and is first theoretically
investigated by Von Kármán [47] in 1921. Two analytical expressions are developed for 𝐶𝑀 for
a free disk under both the laminar and the turbulent flow conditions (for a single surface of the
disk):
Laminar flow: 𝐶𝑀 = 3.68 ∙ 𝑅𝑒−12 Eq. 75
Turbulent flow: 𝐶𝑀 = 0.146 ∙ 𝑅𝑒−15 Eq. 76
Cochran [22] give a more exact solution of the ordinary differential equations derived by Von
Kármán [47]. He solves the equations using a power series expansion in the near wall vicinity
which is combined with an asymptotic approach in the outer region. The results of Von Kármán
[47] and Cochran [19] are principally valid for the case of a disk with infinite radius. Eq. 77 is
derived from the integration of the shear stress across the radius for 𝐶𝑀.
𝐶𝑀 = 1.935 ∙ 𝑅𝑒−
12 Eq. 77
Dorfman [32][33] uses analytical methods to calculate the moment coefficient of a free disk using
a logarithmic boundary layer velocity profile with Eq. 78 (for a single surface of the disk).
𝐶𝑀 = 0.982 ∙ (log10𝑅𝑒)−2.58 Eq. 78
Experiments on a single disk are performed by Bayley and Owen [13], with a radius 𝑏 =
0.381 m, rotating at speeds of up to 4500 rpm . The frictional torque is measured using an
optical torquemeter built into the drive shaft. Based on the experimental results, they correlate Eq.
79 for the moment coefficient (for a single surface of the disk).
𝐶𝑀 = 0.0655 ∙ 𝑅𝑒−0.186 Eq. 79
2.5.2 Enclosed Rotor-Stator Cavity
Apart from the fundamental free disk case, a disk rotating in a casing more closely resembles the
configurations commonly encountered in actual turbomachinery applications.
Schultz-Grunow [74] is one of the first to analyze the flow caused by a rotating disk in a stationary
housing. For small gap sizes and circumferential Reynolds numbers, the tangential velocity varies
35
linearly with the axial coordinate. Neglecting both the radial and the axial velocities in the gap, a
simple relationship for the moment coefficient is found analytically (laminar flow, for a single
surface of the disk):
𝐶𝑀 =𝜋
𝐺 ∙ 𝑅𝑒
Eq. 80
Soo [78] theoretically investigates the laminar flow of a rotor-stator system for small through-flow
and gap widths in order to resolve some inconsistencies from the older reports. The basic solution
is similar to the one obtained by Schultz-Grunow [74] (for a single surface of the disk), written in
Eq. 81.
𝐶𝑀 = 0.0308 ∙ 𝐺−
14 ∙ 𝑅𝑒−
14 Eq. 81
Daily and Nece [28] examine the flow of an enclosed rotating disk both analytically and
experimentally. They distinguish the four flow regimes (see Fig. 5) by correlating different
empirical equations of the moment coefficients, written in Eq. 82~Eq. 85 (for a single surface of the
disk). The turbulent flow regimes III and IV are the most likely to occur in practice. The small gap
cases (regime I and regime III) essentially represent a viscous Couette-type of flow while the other
two regimes correspond to the Batchelor-flow solution characterized by an inviscid core region.
For merged boundary layers (regime I and III), the frictional resistance decreases with increasing
axial gap width as a consequence of the reduced velocity gradient. In contrast, the frictional
resistance increases with increasing axial gap width in case of separated boundary layers (regime
II and IV) due to the braking influence of the outer cylindrical shroud.
Regime I 𝐶𝑀1 =
𝜋
𝐺 ∙ 𝑅𝑒
Eq. 82
Regime II 𝐶𝑀2 = 1.85 ∙ 𝐺1
10 ∙ 𝑅𝑒−12 Eq. 83
Regime III 𝐶𝑀3 = 0.04 ∙ 𝐺−16 ∙ 𝑅𝑒−
14 Eq. 84
Regime IV 𝐶𝑀4 = 0.0501 ∙ 𝐺1
10 ∙ 𝑅𝑒−15 Eq. 85
36
2.5.3 Rotor-Stator Cavity with Through-Flow
According to Daily et al. [29], the moment coefficient increases with increasing through-flow rate
(for centripetal through-flow). The correlation given in Eq. 86 allows the calculation of the moment
coefficient. The parameter 𝐶𝑀0 is the moment coefficient without through-flow, calculated by
Daily and Nece [28] (from Eq. 82 to Eq. 85 according to the flow regime).
𝐶𝑀 = 𝐶𝑀0 ∙ (1 + 13.9 ∙ 𝐾0 ∙ 𝜆𝑇 ∙ 𝐺−18) Eq. 86
Bayley and Owen [13] investigate the effects of a superimposed radial outflow on both windage
and pressure distribution in a rotor-stator system. They use a disk of radius 𝑏 = 0.381 𝑚, rotating
up to speeds of 4500 rpm, with a stationary shroud. The tests cover the range of rotational
Reynolds numbers 4 × 105 ≤ 𝑅𝑒 ≤ 4 × 106. Three values of axial gap width G are used: 𝐺 =
0.06, 0.12 and 0.18. The windage torque on the disk is measured using a shaft mounted optical
torque meter. They observe that, as G decreases, the sensitivity of the torque to variations in G also
decreases.
Dibelius et al. [31] carry out tests on a shrouded rotor-stator system with a disk (𝑏 = 0.4 m). It
has the capacity for both centripetal and centrifugal through-flow ( |𝐶𝐷| ≤ 1.4 × 105 ).
Measurements are obtained for the static pressure, the velocity distribution and the disk frictional
torque. The range of Re tested is 2 × 106 ≤ 𝑅𝑒 ≤ 3 × 107. The torque measurements show that
the moment coefficient for centrifugal through-flow exceeds that for centripetal through-flow.
Owen [65] distinguishes between the source region close to the axis of rotation and the core region
radially outward of the source region. He distinguishes the Stewartson type flow from the Batchelor
type based on the amounts of 𝜆𝑇. He also correlates an analytical approach to predict the moment
coefficient in a shrouded rotor-stator cavity for Batchelor type flow. The correlations for 𝜆𝑇 <
0.219 are given in Eq. 87 (a~d). It can be seen that Eq. 87 (b) is the same as that from von Kármán
[48] for the free disk solution. The flow structures are depicted in Fig. 20 depending on the values
of 𝜆𝑇. The flow structure for an enclosed rotor-stator cavity is sketched in Fig. 20 (a). When the
centrifugal through-flow is imposed and 𝜆𝑇 < 0.219, the centripetal radial velocity near the wall
can still be found, depicted in Fig. 20 (b). In Fig. 20 (c), the case with through-flow for 𝜆𝑇 >
0.219 is depicted. All the radial velocity in the cavity seems to be centrifugal. The parameter 𝑥𝑐
is defined as the radial location at which the source region ends and the core rotation begins. In the
case where 𝜆𝑇 > 0.219, the source region is assumed to apply for the whole rotor-stator cavity
(𝑥𝑐 > 1).
37
𝐶𝑀 = 𝐶𝑀𝑎 + 𝑐𝑀𝑏 Eq. 87 (a)
𝐶𝑀𝑎 = 0.146 ∙ 𝑅𝑒(−15
) ∙ 𝑥𝑐
235 Eq. 87 (b)
𝐶𝑀𝑏 = 0.0796 ∙ 𝑅𝑒(−
1
5)
∙ {(1 − 𝑥𝑐
23
5 ) + 14.7 ∙ 𝜆𝑇 ∙ (1 − 𝑥𝑐2) + 90.4 ∙ 𝜆𝑇
2 ∙ [1 − 𝑥𝑐(−
3
5)]}
Eq. 87 (c)
𝑥𝑐 = 1.79 ∙ 𝜆𝑇(
513
) Eq. 87 (d)
Fig. 20: Flow structure inside a rotor-stator cavity with (a) no through-flow, (b) with centrifugal
through-flow and 𝝀𝑻 < 𝟎. 𝟐𝟏𝟗 and (c) with through−flow, 𝝀𝑻 > 𝟎. 𝟐𝟏𝟗
It can be seen that 𝐶𝑀𝑎 is the same as the free disk solution from von Kármán [48]. In the case
where 𝜆𝑇 > 0.219, the source region is assumed to apply for the whole rotor-stator cavity (𝑥𝑐 >
1). However, the free disk solution is inapplicable therefore, in this case, the moment coefficient
𝐶𝑀 is calculated with Eq. 88.
𝐶𝑀 = 0.666 ∙ 𝐶𝐷 ∙ 𝑅𝑒−1 Eq. 88
Gartner [37][38] uses an analytical approach to extend Eq. 87 (a) for the frictional torque, in a
shrouded rotor-stator system with radial outflow, to include the effects of the cylindrical shrouding.
In the case where 𝜆𝑇 < 0.219 , a third term 𝐶𝑀𝑐 is added into Eq. 87 (a), which represents the
moment coefficient due to the cylindrical shroud. The expression for 𝐶𝑀𝑐 is defined in Eq. 89. This
correlation shows a small improvement on Eq. 87 in Owen [65], when compared with the measured
torque. Coren et al. [23] produce a correlation giving the moment coefficient in terms of Reand
𝐶𝐷, given in Eq. 90.
s
b
𝛺
Dis
k D
isk
Dis
k
Wal
l Wal
l
Wal
l
(a) (b) (c)
𝑥𝑐
𝑥𝑐
𝑥𝑐
38
𝐶𝑀𝑐 = 0.36 ∙ 𝛾(−
14
)∙ 𝐾
74 ∙ (1 − 𝐾)
320 ∙ 𝑅𝑒(−
15
)
𝛾 = [81 ∙ (1 + 𝛼2)
38
49 ∙ (23 + 37 ∙ 𝐾) ∙ 𝛼]
45
𝛼 = [
2300 ∙ (1 + 8𝐾)
49 ∙ (1789 − 409 ∙ 𝐾)]
12
𝐾 = 0.087 ∙ 𝐾0 ∙ 𝑒[5.2∙(0.486−𝜆𝑇)−1]
Eq. 89 𝐾0 = 0.49 − 0.57 ∙
𝑠
𝑏
𝐶𝑀 = 0.52 ∙ 𝐶𝐷0.37 ∙ 𝑅𝑒−0.57 + 0.0028 Eq. 90
2.5.4 Impact of Surface Roughness on the Moment Coefficient
Dorfman [32] calculates the frictional torque of a rough disk and develops Eq. 91 for the moment
coefficient for an enclosed rotor-stator cavity. The results are plotted in Fig. 21. With the increase
of 𝑏
𝑘𝑠, the values of 𝐶𝑀 from his analytical correlation drop and are in very good agreement with
those from Eq. 91.
𝐶𝑀 = 0.108 ∙ (
𝑘𝑠
𝑏)0.272
Eq. 91
Fig. 21: Moment coefficient for a rotating disk (circles: from an analytical correlation; dashed
line: Eq. 91; triangle: test)
Daily and Nece [28] study the effects of disk roughness in enclosed rotor-stator systems. The test
fluids are water and three solvent-refined, paraffin-based commercial lubricating oils. The test rig
is illustrated in Fig. 22. During the experiments, there is no through-flow, despite a recirculation
system equipped with a heat exchanger is provided for cooling the test fluid. Torques are measured
by means of four SR-4 (Simmons Ruge-4) bonded strain gages connected to a battery-powered
0
4
8
12
16
20
24
28
32
36
40
1 10 100
10
00
𝐶𝑀
𝑘𝑠
𝑏/100𝑘𝑠
39
bridge circuit and mounted to the inside wall of a recess in the shaft. Voltage differences from the
circuit are taken off through slip rings mounted on the shaft. A DC galvanometer serves as the
indicating device. As the shaft recess containing the strain gages is situated on the fluid side of the
shaft seal, the only torque measured is due to the fluid friction on the disk surface. No deductions
are necessary for bearing and sealing friction.
Fig. 22: Schematic drawing of the test rig (redrawn from Daily and Nece [28])
The commercial grid papers are bonded to both the rotor and the stator. The experiments are
conducted for relative roughnesses of 𝑘𝑠
𝑏= 10−3, 5 × 10−4 and 3.2 × 10−4, for G=0.023, 0.061
and 0.112, and for 4 × 103 < 𝑅𝑒 < 6 × 106. It is found that roughness had no significant effect
on the moment coefficient in the laminar regimes I and II; for 𝑅𝑒 > 2 × 105, in the turbulent
regime III and IV, 𝐶𝑀 increases as the roughness increases. From results obtained at the largest
values of Re, Daily and Nece [28] determine the correlation:
𝐶𝑀
−12 = −5.37 ∙ log10 (
𝑘𝑠
𝑏) − 3.4 ∙ 𝐺
14
Eq. 92
They also conclude that roughness effects start to become significant for:
𝑅𝑒 ∙ 𝐶𝑀
−12 ≈ 16000 ∙ (
𝑘𝑠
𝑏)
−1
10
Eq. 93
Shaft
Disk
Wall
Strain gage
To motor
Bearing
Slip rings Seal
40
Within the Reynolds number regions of significant roughness effect, roughness changes produce
much larger changes in 𝐶𝑀 than changes in axial spacing do (Daily and Nece[28]), plotted in Fig.
23.
𝐶𝑀
Fig. 23: Rough disk torque data (replotted from Daily and Nece [28])
Kurokawa et al. [53] study the moment coefficient in a rotor-stator cavity with different through-
flow coefficients, noted as 𝐶𝐷. The sketch of their test rig is given in Fig. 24. Kurokawa et al. [54]
also investigate the roughness effects on the flow in an enclosed rotor-stator cavity. Their analytical
results are in a satisfactory agreement with the measured data in both the hydraulic smooth and the
complete rough regions. In their experiments, a disk of 164 mm radius is rotated in a water-filled
casing with the radial and the axial gaps of 2 mm (𝐺 = 0.012 ) and 12.8 mm (𝐺 = 0.078 )
respectively.
0.001
0.01
0.1
10000 100000 1000000 10000000
𝑅𝑒
𝑘𝑠
𝑏=
1
3200, 𝐺 = 0.0609
104 105 106 107
Smooth disk and wall: 𝐺 = 0.0255
Smooth disk and wall: 𝐺 = 0.0609
𝑘𝑠
𝑏=
1
1000, 𝐺 = 0.0609
𝑘𝑠
𝑏=
1
2000, 𝐺 = 0.0609
41
Fig. 24: Sketch of the test rig (replotted from Kurokawa et al. [54])
2.6 Flow Separation near the Entrance
When the velocity at the inlet is large enough, the flow is through-flow dominant. Since there is an
area change when the flow enters the cavity from the guide vane, flow separation may occur when
the boundary layer travels far enough against an adverse pressure gradient that the speed of the
boundary layer relative to the object falls almost to zero (Anderson, John D. [3]). The fluid flow
becomes detached from the surface of the object, and instead takes the forms of eddies and vortices,
depicted in Fig. 25. The flow reversal is primarily caused by an adverse pressure gradient imposed
on the boundary layer by the outer potential flow. The direction 𝑙 and 𝑧𝑙 are also shown in Fig.
25. The approximately streamwise momentum equation inside the boundary layer is stated in Eq.
94.
Fig. 25: Graphical representation of the velocity profile and the reverse flow which show the
flow separation [3]
𝑄
𝑧𝑙
𝑙
𝑄
I III IV
∆𝑟
𝑏
𝑟
I: Wall boundary layer
III: Core
IV: Disk boundary layer
𝑉𝑟
𝑉𝜑
𝛺
Disk
Wall
Surface streamline Wall
v
42
v
𝜕v
𝜕𝑙= −
1
𝜌
𝑑𝑝
𝑑𝑙+ 𝜈
𝜕2v
𝜕𝑧𝑙2
Eq. 94
In the case with centripetal through-flow, flow separation may occur when there is a jet flow
through each guide vane passage. The approximate separation line is depicted with green dotted
line in Fig. 26.
Fig. 26: Approximate separation line for centripetal through-flow
The solution procedure for the approximate separation line is shown in Fig. 27. When the jet flow
is very strong through the guide vane channel and the disk rotates slowly, the tangential velocity
component can be neglected compared with the radial velocity. The cavity flow is therefore
through-flow dominant. The pressure and the velocity at each radius should be calculated to
understand the cavity flow. The position of the separation line is very important for the calculation
of above parameters. The case when 𝛺 = 0 , 𝐶𝐷 = −5050 (centripetal through-flow, 𝑄 =
−2 m3/h ) and 𝐺 = 0.045 (𝑠 = 0.005 m ) is taken as an example, shown in Fig. 28. The
approximate separation line (dashed line) from the results of a numerical simulation is in relative
good agreement with those from the solution procedure in Fig. 27. Known the position of the
separation line, the cross section 𝐴𝑛 at each radius can be calculated. Then, the average values of
p and v𝑛 can be estimated. The shape of the separation lines from the solution procedure and the
numerical simulation are in relative good agreement with each other, depicted in Fig. 28.
Separation line:
Green dotted line
𝑝𝑛, v𝑛, 𝐴𝑛
𝑝𝑛+1, v𝑛+1, 𝐴n+1
𝑙
𝑧𝑙
Disk
Wall Wall
Control
volume
1, 2, … … , 𝑎
1,2
,……
,𝑑
43
Starting parameter:
𝑎 = 1, 𝑑 = 1, n=1
𝑙𝑑, 𝑛
𝑧𝑙𝑎
𝐴𝑛+1
𝐴𝑛v𝑛 = 𝐴𝑛+1 ∙ v𝑛+1, 𝑉𝑛𝜕v
𝜕𝑙= −
1
𝜌
𝑑𝑝
𝑑𝑙+ 𝜈
𝜕2v
𝜕𝑍2
𝑝𝑛+1 = 𝑝𝑛 +𝜌v𝑛
2
2−
𝜌 ∙ 𝐴𝑛2v𝑛
2
2 ∙ 𝐴𝑛+12 −
𝜌 ∙ 𝐴𝑛2v𝑛
2
2 ∙ 𝐴𝑛+12 (
𝐴𝑛+1
𝐴𝑛− 1)2
Get velocity profile along 𝑧𝑙: ν𝜕2v
𝜕𝑍2 = 𝑉𝑛𝜕v
𝜕𝑙+
1
𝜌
𝑑𝑝
𝑑𝑙
Get Average velocity: v̇𝑛+1
0.999 ∙ v̇𝑛+1 < v𝑛+1 < 1.001 ∙ v̇𝑛+1
Save 𝑙𝑑 , 𝑧𝑙𝑎, 𝑝𝑛+1, , v𝑛+1, 𝑛
Fig. 27: Solution procedure for the approximate separation line
𝑉𝑟 from numerical simulation 𝑉𝑟 from Matlab (Program see Fig. 27)
Blue: Radially inward velocity; Red: Radially outward velocity.
Fig. 28: Comparison of the results for the separation line (𝜴 = 𝟎, 𝑪𝑫 = −𝟓𝟎𝟓𝟎, 𝑮 = 𝟎. 𝟎𝟒𝟓)
𝑎 = 𝑎 + 1 𝑑 = 𝑑 + 1
𝑛 = 𝑛 + 1
Numerical
simulation From
solution
procedure
𝑥 = 0.83
Disk Wall 0.005 − 𝑧 (mm)
𝑥 = 1.01 Inlet
Wal
l
Wall
Dis
k
Inlet
𝑥 = 0.83
𝑥 = 1.01
𝑙 (m
)
Dis
k
Wall
Wal
l
44
2.7 Influence of the Sealing Gap Height
2.7.1 Flow Structure inside the Sealing Gap
The sealing gaps in centrifugal pumps are usually very narrow in order to keep the leakage flow
rate as low as possible. During the operation, the gap height will change due to wear or variations
in the tangential direction because of an eccentricity of the rotor.
Will [88] investigates the effect of the sealing gap height in the back cavity of a radial pump. The
domain for numerical simulation is depicted in Fig. 29. Two sealing gap heights (given in Table 5)
are tested.
Fig. 29: One seventh segment of the back cavity (Will [88])
Config. 1 Config. 2
Gap height shroud side chamber 0.5 mm 0.5 mm
Gap height hub side chamber 0.48 mm 0.24 mm
Table 5: Parameters of the experiments
Will [88] compares the measured pressure with those from numerical simulation, depicted in Fig.
30. The results for the front cavity are satisfactory while they have bigger deviations in the back
cavity.
Q
Front view Cross section
Sealing gap
Q
Disk/impeller
Casing
Balance hole
Hub
45
𝑥 𝑥
Points: Pressure measurements; Dashed lines: Numerical simulation.
Fig. 30: Comparison between experimental and numerical results for the shroud side chamber
(left) and the hub side chamber (right) in case of 0.48 mm sealing gap height (Will [88])
When the sealing gap height decreases to 0.24 mm, smaller deviations between the numerical and
the experimental results are apparent across the sealing gap in the back cavity (hub side), depicted
in Fig. 31 (compared with those in the right figure of Fig. 30). The sealing gap controls the amount
of leakage that passes through the cavity and vice versa determines the pressure difference across
the gap. Mechanical energy conveyed to the fluid in the impeller is dissipated into heat by the
choking effect of the sealing which directly influences the efficiency of the pump. The values of
the 𝑝∗ drop more tremendously across the sealing gap from the results of both numerical
simulation and pressure measurements, compared with those in Fig. 30 for wider sealing gap at the
hub side.
𝑥
Points: Pressure measurements; Dasheded lines: Numerical simulation.
Fig. 31: Comparison between experimental and numerical results for the hub side chamber in
case of 0.24 mm sealing gap height (Will [88])
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0.5 0.7 0.9 1.1 1.3 1.5
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0.5 0.7 0.9 1.1 1.3 1.5
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0.5 0.7 0.9 1.1 1.3 1.5
𝑝∗
𝑝∗
𝑝∗
46
One major parameter for the pressure drop is the gap height which is also pointed out by Iino et al.
[44] for multistage radial pumps. In order to reduce the leakage losses, the gap height is generally
very small and therefore difficult to determine under operating conditions. As expected, the
pressure drop considerably increases and the agreement with the experimental data improves.
Unfortunately, some discrepancies are still apparent behind the sealing gap which might be
attributed to an insufficient resolution of the flow in the gap itself. Currently, the flow is only
resolved by 8 cells in the radial direction in the shroud side cavity and by 13 cells in the hub side
cavity. For example, Tamm [80] uses 6 cells and notes that doubling the number does not change
the results.
According to Gülich [38], the flow in the gap is unstable because both the circumferential velocity
and the centrifugal forces decrease from the rotating shaft to the stationary casing. A similar
situation is presented in the radial gap between the shroud of the rotating disk and the radial casing
in the numerical study of the enclosed rotating disk. The stability criterion for the formation of
Taylor vortices in the gap resulting from this instability is the Taylor number 𝑇𝑎, given in Eq. 95.
The perimeter Reynolds number for the sealing gap 𝑅𝑒𝑝 is calculated with Eq. 96. The
circumferential velocity in the sealing gap v𝑔𝑎𝑝 is predicted with Eq. 97.
𝑇𝑎 =v𝑔𝑎𝑝 ∙ ∆𝑟𝑠𝑒𝑎𝑙
𝜈∙ √
∆𝑟𝑠𝑒𝑎𝑙
𝑟𝑔𝑎𝑝=
𝑅𝑒𝑝
2∙ √
∆𝑟𝑠𝑒𝑎𝑙
𝑟𝑔𝑎𝑝
Eq. 95
𝑅𝑒𝑝 =
2 ∙ ∆𝑟𝑠𝑒𝑎𝑙 ∙ v𝑔𝑎𝑝
𝜈
Eq. 96
v𝑔𝑎𝑝 =
2𝜋 ∙ 𝑟𝑔𝑎𝑝 ∙ 𝑛
60 Eq. 97
For 𝑇𝑎 ≥ 41.3, counter-rotating Taylor vortices are expected in the sealing between the rotating
shaft and the stationary casing which strongly influence the pressure drop and energy dissipation
in the gap. To determine whether the flow is turbulent or not, both v𝑔𝑎𝑝 and v𝑎𝑥𝑔𝑎𝑝 (mean axial
velocity in the sealing gap) have to be taken into account considering the gap Reynolds number
𝑅𝑒𝑔𝑎𝑝, given in Eq. 98. The characteristic linear dimension for a fluid moving between two plane
parallel surfaces is 2 ∙ ∆𝑟𝑠𝑒𝑎𝑙.
𝑅𝑒𝑔𝑎𝑝 = √(2 ∙ ∆𝑟𝑠𝑒𝑎𝑙 ∙ v𝑎𝑥𝑔𝑎𝑝
𝜈)2 + (
v𝑔𝑎𝑝 ∙ ∆𝑟𝑠𝑒𝑎𝑙
𝜈)2
Eq. 98
47
Transition to the turbulent flow takes place at 𝑅𝑒𝑔𝑎𝑝 = 2000 while the flow will always be
turbulent for 𝑅𝑒𝑔𝑎𝑝 = 4000 (Gülich [38]). For the turbulent flow in the gap, the friction factor
for the sealing additionally depends strongly on the wall roughness due to the very small gap height
(the relative roughness is comparatively large).
An estimation of the Taylor number and the gap Reynolds number 𝑅𝑒𝑔𝑎𝑝 of both cavities indicates
that the flow is turbulent. Taylor vortices are to be expected (see Table 6) (Will [88]).
Shroud disk cavity
Hub disk cavity
𝑅𝑒𝑔𝑎𝑝 12841 4450
𝑇𝑎 457 147
Table 6: Estimation of 𝑹𝒆𝒈𝒂𝒑 and 𝑻𝒂 in Will [88]
Will [88] states that the flow in the sealing gap is unstable and favors the formation of Taylor
vortices. An important parameter influencing the flow in the small sealing gap is the through-flow.
To further investigate this detail, a parametric study is accomplished. The through-flow rate is first
set to zero and afterwards successively increased by varying the axial velocity component at the
cavity inlet. These simulations are performed with a sealing gap height ∆𝑟 = 0.8 mm to ensure
that the resulting Taylor number of 𝑇𝑎 = 1003 is far above the critical number of 𝑇𝑎 = 41.3. The
results in Fig. 32 clearly prove the expected dependency of the flow structure on the externally
applied through-flow rate. For a very small through-flow rate, several vortex pairs are clearly
visible in the gap and the flow is dominated by the instabilities rather than the through-flow. If the
through-flow rate increases, the vortex pairs gradually diminish. This trend continues with
increasing through-flow until the streamlines are perfectly aligned in the axial direction and the
flow structure is controlled by the through-flow. The critical through-flow rate in this thesis
amounts approximately 16% of the nominal leakage flow. For higher values, no more vortices
occur (Will [88]).
48
𝑄 = 0.0064 m3/h, v𝑎𝑥𝑔𝑎𝑝 = 0.003 m s⁄
𝑄 = 0.0139 m3/h, v𝑎𝑥𝑔𝑎𝑝 = 0.0065 m s⁄
𝑄 = 0.0642 m3/h, v𝑎𝑥𝑔𝑎𝑝 = 0.03 m s⁄
Fig. 32: Flow in the sealing gap (∆𝒓 = 𝟎. 𝟖 𝐦𝐦) for different leakage flow rates (replotted from
Will [88])
2.7.2 Leakage Volumetric Flow Rate through the Sealing Gap
The through-flow coefficient 𝐶𝐷 , calculated with the leakage volumetric flow rate, is very
important during the calculations of K, 𝐶𝐹 and 𝐶𝑀. Baskharone et al. [9] estimate the leakage
flow between pump stages based on finite-element analyses. Zhao et al. [91] investigate the sealing
gap height on the pump performance based on a CFD method. Wu and Squires [89] correlate Eq.
Q
Q
Q
Q
Q
Q
Rotor (Impeller)
Casing
Casing
Rotor (Impeller)
Rotor (Impeller)
Casing
∆𝑟
∆𝑥 =∆𝑟
𝑏
49
99 to predict the volumetric leakage flow rate through the sealing gap of a multi-stage centrifugal
pump. They mention that for the radial pump, the pressure drops can be estimated with ∆𝑝 = 0.6𝐻
(H is the pump pressure head in m) and 𝛹 = 0.5~0.6.
𝑄 = 𝛹 ∙ 𝐴 ∙ √2𝑔 ∙ ∆𝑝 Eq. 99
2.8 Side Chamber Flow in a Centrifugal Pump
The design process of radial pumps requires reliable information about the fluid dynamics in every
single component of the machine. Aside from the flows which are responsible for the energy
transmission in the main parts of a machine, such as the blade passages in the impeller or the volute
casing, the flow in regions resulting from construction conversion is of great importance. These
parts, such as the side cavities between the rotating impeller and the stationary casing, do not serve
the purpose of increasing the energy level of the fluid. They lead to a reduction of the machine
efficiency and can affect the operational reliability. In the side chambers, the frictional torque of
the side walls has a great impact on the required power of the machine. Especially for low specific
speed impellers, this is an inherent problem. Gülich [38] mentions that for low specific speed
machines, disk friction losses can amount up to 50% of the useful power and the leakage losses can
make up to 12% of the nominal flow rate (Bahm [6]). Furthermore, in general, the leakage flow,
which passes the sealing gaps, leads to a reduction of the hydraulic efficiency.
The flow in the side chambers directly influences the mechanical design of a radial pump because
the resulting axial force, whose knowledge is necessary for the design of the bearing, is essentially
determined by the static pressure distribution on the outer side walls of the impeller. The pressure
distribution depends on the rotation of the fluid in the side chamber which is strongly influenced
by the incoming swirling leakage flow. The leakage again is a function of the operating point of
the machine. Besides, the central role of the leakage flow with respect to the axial thrust has already
been noticed in Bahm [6].
The flow structure in the volute depends on the operating point and an axisymmetric pressure
distribution is usually only presented in design flow conditions (Pavesi [67]). According to Bahm
[6], the volute acts like a diffuser at part load. Consequently, the flow in the volute is retarded and
the pressure increases in tangential direction. For overflow conditions, the volute acts like a nozzle
accelerating the fluid in the tangential direction. This leads to a pressure decrease in the
circumferential direction. Further, the authors observe an inward directed flow close to the wall
whose velocity and direction depend on the circumferential position. This imbalance results in a
50
radial load which deflects the impeller into an eccentric position. As a result, the line of influence
of the axial force does not coincide with the axis of rotation leading to an additional torque. In
centrifugal pumps, part load recirculation is a well-known problem.
Gülich [38] notices that especially low energy backflow fluid entering the side chambers
significantly dampens the rotation of fluid and thus affects the axial force. A reduced fluid rotation
leads to a smaller pressure drop which results in a larger pressure force. In dependence on the axial
displacement of the impeller, the recirculation region can occur close to either of the side chambers
at part load conditions. Under certain circumstances, this can even cause a reversal of the resulting
axial force in multi-stage pumps.
The pressure distribution at the impeller outlet is non-stationary (Gülich [38]) due to the complex
interaction of the rotating impeller and the stationary casing or the guide vanes. This causes a
fluctuating behavior of the flow magnitudes that even extends into the side cavities.
Before investigating a centrifugal pump, the flow in a rotor-stator cavity should be investigated
carefully. The assumption of the steady, axisymmetric flow conditions is fundamental to nearly all
publications concerning rotor-stator cavities. A lot of correlations are, however, still not satisfactory
even for the rotor-stator cavity model. Therefore, the following study tries to provide more results
and determine some correlations.
51
3. Experimental Set-Up
3.1 Mechanical Set-Up
The design of both radial pumps and turbines demands more accurate values of 𝐶𝐹 and 𝐶𝑀 .
Above two parameters are still not precisely predictable based on the results and the correlations
from the literature. Measurements are still required to provide more results aiming to determine
some correlations. In order to investigate these relationships experimentally, a test rig is designed
at the University of Duisburg-Essen. The cross section of the test rig is depicted in Fig. 33 (a). The
through-flow (volumetric through-flow rate 𝑄) is supplied with water by a pump system. A tank
with 3 m3 pure water at 20°C is used to feed the water in the pump and also to recollect the water
out of the test rig. The view along the “A” direction is sketched in Fig. 33 (b). In Fig. 33 (c), the
shaft sealing with the radius of 10 mm at the back cavity is depicted. A picture of the test rig is
shown in Fig. 33 (d). The shaft is driven by an electrical motor. A frequency converter is utilized
to adjust the speed of rotation (0 to 2500 rpm) with the absolute uncertainty of 7.5 rpm. The speed
of rotation is increased softly and gradually. In this thesis, only the axial gap of front chamber is
changeable (by installing six sleeves (I) with different length). There are 24 channels in the radial
guide vane (II) instead of entirely open at the periphery, shown in Fig. 33 (b). With centripetal
through-flow, one of the guide vanes with different blade angle is installed to produce pre-swirl.
For centrifugal through-flow, there are four inlet swirlers (XI) in the horizontal pipe. A radial
directed guide vane (II) with wide passages is installed at the centrifugal outlet. Other parameters
of the experiments are given in Table 7. The transducers in the test rig include 12 pressure tubes, a
torque meter and three tension compression transducers. A thrust plate (VIII) is fixed by a ball
bearing and a nut from each side to convey the axial thrust to the tension compression transducers.
A linear bearing (VI) is utilized to minimize the frictional resistance during the axial thrust
measurements. In the process of the axial thrust measurements, the axial thrust transducers are
calibrated when the axial gap width of the front chamber is changed.
52
(V)
(I)
(VI) (II)
(VII) (III)
(VIII) (IV)
(XI) (IX)
(X)
(I). Sleeves (to change the axial gap), (II). Guide vane (24 channels), (III). Front chamber, (IV). Disk, (V). Back cover,
(VI). Linear bearing, (VII). Tension compression sensor, (VIII). Thrust plate, (IX). Nut, (X). Shaft, (XI) Swirler
(a) Cross section of the test rig
(b) View along “A” direction (c) Shaft sealing (d) Test rig
Fig. 33: Test rig design
b (mm) n (rpm) |𝑄| (m3/s) s (mm) sb (mm) a (mm) t (mm)
110 0 to 2500 0 to 5.56 × 10−4 2 to 8 8 23 10
Table 7: Parameters of the experiments
The measurements of the axial thrust coefficient 𝐶𝐹 include two steps. The test rig set-up for each
step is drawn in Fig. 34. The first step, plotted in Fig. 34 (a), is to measure the results imposed by
the drive end of the motor when the shaft without the disk is rotating at different speeds of rotation
in the air. For the second step, the measurements of 𝐹𝑎 are performed at different Re, G, 𝐶𝐷 and
𝐶𝑎𝑚. All the measured results are modified by subtracting above values obtained in the first step
according to the speed of rotation. The shaft head in Fig. 34 (b) is considered as a part of the front
surface (the area of front surface is 𝛺 ∙ 𝑏2). The values of 𝐶𝐹𝑏 are obtained when 𝐶𝐷 = 0 and
both the cavities have the same axial gap width for different Re (under that condition 𝐶𝐹𝑓 = 𝐶𝐹𝑏)
according to the area difference between the front surface and the back surface.
Direction of 𝐹𝑎
“A” direction
Shaft connected
to an electric
motor Shaft sealing
𝑟𝑠𝑒𝑎𝑙 = 10 mm
Shaft
Shaft sealing
53
(a) Step 1 (b) Step 2
Fig. 34: Test rig set-up for each steps to measure 𝑭𝒂
To compute the torque precisely, the friction due to the shaft needs to be subtracted. Therefore,
the measurements without the disk and at different rotating speed are accomplished. The values of
𝐶𝑀 at the back surface are obtained when 𝐶𝐷 = 0 and the axial gaps of the both cavities are of
the same size for different Reynolds numbers Re.
The positions for the pressure measurement are presented in Table 8.
Type of measurement Location Number of positions X
min max
p Front chamber 12 0.454 0.955
Table 8: Parameters of the experiments conducted in the test rig
The inlet swirler system is designed in the horizontal pipe, depicted in Fig. 35. In the test rig, four
swirlers, fixed with screws, is angled every 15° from −90° to 90°. The component drawing of
the centrifugal inlet swirler is illustrated in Fig. 36.
Top view
Fig. 35: Inlet swirlers in the horizontal pipe (for centrifugal through-flow)
𝛺
Dis
k
Wat
er a
t 2
0
°C
Air
Shaft Shaft 𝛺
Wal
l
Wal
l
Wal
l
Wal
l
Screw to adjust
pre-swirl angle
54
Fig. 36: Drawing of the centrifugal inlet swirler (Left: front view; Right: side view)
The geometry of the radial guide vanes (for centripetal pre-swirl through-flow) and the flow in the
rotor-stator cavities are depicted in Fig. 37. Every single part in green color refers to the flow in
each guide vane passage. The guide vanes can be categorized into two groups: with small passages
and with large passages. The through-flow is depicted by the arrows. A jet flow, which becomes
stronger for the guide vanes with smaller passages, exists near each cavity inlet in the guide vane.
Guide vane Flow
No.1
(sm
all
pas
sages
)
Blade
Q
Screw
Unit: mm
Flow in one passage 24 Passages
55
No
.2 (
smal
l pas
sag
es)
No.3
(la
rge
pas
sages
)
No.
4(l
arge
pas
sages
)
No
. 5
(la
rge
pas
sages
)
Fig. 37: Geometry of the radial guide vanes and the flow in the rotor-stator cavities
Q
Q
Q
24 Passages
24 Passages
24 Passages
24 Passages
Flow in one passage
Flow in one passage
Flow in one passage
Flow in one passage
Q
56
The pre-swirl angles (angle between the vane and the tangential direction) of the radial guide vanes
in Fig. 37 are given in Table 9.
Guide vane No.1 No.2 No.3 No.4 No.5
Passage Small Small large large large
Pre-swirl angle 0° 60° 0° 52° 26°
Table 9: Pre-swirl angles of the radial guide vanes
Six disks are manufactured with different surface roughness for comparisons. In Table 10, the
values of measured 𝑅𝑧 and the amounts of 𝑘𝑠 are given. The values of 𝑘𝑠 for the disks ranges
from 0.4 𝜇m to 81.3 𝜇m, which covers the parameter range of a majority of turbomachines.
The values of 𝑅𝑧 on all other wet surfaces of the test rig are below 1.6 μm.
Disk No. 0 1 2 3 4 5
𝑅𝑧 (𝜇m) 1.0 9 29.6 69 153 211
𝑘𝑠 (𝜇m) 0.4 3.5 11.4 26.5 58.9 81.3
Table 10: Surface roughness of the disks
3.2 Uncertainty Analysis
In Table 11, the relative error (e𝑇) and the measuring range ((𝑀𝑟)) of the pressure, the torque and
the axial thrust transducers are listed. All the experimental results are an ensemble average of 1000
samples. The uncertainties of the measured results ∆𝑁 , which is estimated with the root sum
squared method with Eq.100, are the differences between the real values and the measured values.
The parameter 𝑁𝑇 is the uncertainty due to the transducers while the parameter 𝑁𝐷 is that due to
the data acquisition system. The input voltage signals are the following ranges: 0 to 10 V for the
pressure transducers and the torque transducer, −10 V to 10 V for the axial thrust transducers. The
absolute accuracy of the data acquisition system is 4.28 mV. The random noise and the zero order
uncertainty are neglected because they are considered to be very small. The distributions of the
results are considered as normal distributions and the normal distribution coefficient is selected as
1.96 (95% confidence level). The parameter 𝑛𝑇 represents the number of transducers used to
obtain one result together. To evaluate 𝐹𝑎 and M on the front surface, the total results, the results
on the back surface (obtained when 𝐶𝐷 = 0 and 𝑠 = 𝑠𝑏) and the results when the shaft is rotating
without a disk in the air are measured with the transducers. Hence, the measuring times to obtain
one result 𝑛𝑀 is 3 for the axial thrust, the frictional torque and Re (by measuring n). The
uncertainties of the measurements are also given in Table. 11.
57
∆𝑁 = √𝑁𝑇
2 + 𝑁𝐷2; 𝑁𝑇 =
√𝑛𝑇∙𝑛𝑀∙(𝑒𝑇∙𝑀𝑟)2
1.96∙√1000; 𝑁𝐷 =
√𝑛𝑇∙𝑛𝑀∙(𝑒𝐷∙𝑀𝑟)2
1.96∙√1000
Eq. 100
p (bar) 𝐹𝑎 (N) M (Nm) Re 𝐶𝐷
e𝑇 1% (FS) 0.5% (FS) 0.1% (FS) − −
𝑀𝑟 0 to 2.5 bar −100 to 100 N 0 to 10 Nm − −
∆𝑁 4.04× 10−4 2.43× 10−2 3.00× 10−4 9.01× 104 4.1
𝑛𝑇 1 3 1 1 1
𝑛𝑀 1 3 3 3 1
Table 11: Uncertainty analysis for the measurements (FS: full scale)
3.3 Experimental Validation
In the experiments, the values of M and 𝐹𝑎 when the shaft without the disk is rotating at different
speeds of rotation are subtracted. The parameters 𝑀𝑠 and 𝐹𝑎−𝑠 respectively are the measured
torque and the measured axial thrust without the disk. The values of 𝐶𝑀𝑠 and 𝐶𝐹𝑠 respectively
are plotted in Fig. 38 (a) and Fig. 38 (b) at different Re.
𝐶𝑀
𝑠=
2∙|
𝑀𝑠|
𝜌∙𝛺
2∙𝑏
5
𝐶𝐹
𝑠=
2∙|
𝐹 𝑎−
𝑠|
𝜌𝜋
𝛺2
𝑏4
(a) 𝐶𝑀𝑠 versus Re (b) 𝐶𝐹𝑠 versus Re
Fig. 38: Experimental results of 𝑪𝑴𝒔 and 𝑪𝑭𝒔 versus Re
In some experiments of 𝐶𝐹 and 𝐶𝑀 concerning pre-swirl, the shaft has to counter-rotate to get a
negative pre-swirl angle. It is therefore important to check whether the axial thrust is sensitive to
the direction of rotation. The ratio of 𝐶𝑀𝑠−𝑐/𝐶𝑀𝑠−𝑎𝑐 versus Re are shown in Fig. 39 (a). The
subscript “c” refers to a clockwise rotation of the disk, while the subscript “ac” represents an anti-
clockwise rotation of the disk. The values of both 𝐶𝑀𝑠−𝑐/𝐶𝑀𝑠−𝑎𝑐 and 𝐶𝐹𝑠−𝑐/𝐶𝐹𝑠−𝑎𝑐 range from
0.96 to 1.04, shown in Fig. 39. Hence, the experimental results for 𝐶𝑀 and 𝐶𝐹 remain almost
unaffected by the direction of rotation.
0
0.005
0.01
0.015
0.02
0.025
0.03
0 1 2 3 4
0
0.005
0.01
0.015
0.02
0.025
0.03
0 1 2 3 4
Re
(106)
Re
(106)
58
𝐶𝑀
𝑠−
𝑐
𝐶𝑀
𝑠−
𝑎𝑐
(a) 𝐶𝑀𝑠−𝑐/𝐶𝑀𝑠−𝑎𝑐
𝐶𝐹
𝑠−
𝑐
𝐶𝐹
𝑠−
𝑎𝑐
𝐶𝐹𝑠−𝑐/𝐶𝐹𝑠−𝑎𝑐
Fig. 39: Comparison of results when the shaft rotates in different direction
In Fig. 40, the measured pressure coefficients at 𝑅𝑒 = 1.36 × 106, 𝐶𝐷 = 0 and 𝐺 = 0.0454 are
compared with those in Kurokawa et al. [52] (𝐺 = 0.046). They are in good accordance with each
other both using smooth and plane disks. The results from the pressure transducers therefore are
considered to be reliable.
0.95
0.96
0.97
0.98
0.99
1
1.01
1.02
1.03
1.04
1.05
0 1 2 3 4
0.95
0.96
0.97
0.98
0.99
1
1.01
1.02
1.03
1.04
1.05
0 1 2 3 4
Re
(106)
Re
(106)
59
𝑥
Fig. 40: Comparison of 𝑪𝒑 along the radius for 𝑹𝒆 = 𝟏. 𝟑𝟔 × 𝟏𝟎𝟔
0.4
0.5
0.6
0.7
0.8
0.9
1
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08
Measurements G=0.0454
𝐶𝑝
Measurements in Kurokawa et al. [52]
G=0.046
60
4. Numerical Simulation Set-Up
In this thesis, the velocity profiles are predicted using the ANSYS CFX 14.0 code. The distributions
of K from numerical simulations (varying G, Re, 𝐶𝐷 and 𝐶𝑎𝑚) are compared with those calculated
based on the pressure measurements in Chapter 5. This can improve the reliability of the correlation
of K derived based on the pressure measurements. The profiles of tangential velocity from
numerical simulation are also analyzed to estimate the impact of 𝐶𝐷 on the distinguishing lines
between flow regime III and flow regime IV for various G and Re.
4.1 Turbulence Model
During the past 60 years, numerous numerical studies are devoted to the flow in a rotor-stator cavity.
The turbulence models turn out to be a key feature. Besides, these flows are very well suited to
study the effects of rotation on turbulence. Part of the selected turbulence models in the literature
are listed in Table 12.
Authors Through-flow Shroud Models Findings
Morse [64] Yes Yes Low Re 𝑘 − 휀 Under-predicted 𝐶𝑀
Chew and Vaughan [20] No Yes Mixing length
formulation Succeed to predict 𝐶𝑀
Iacovides and Theofanopoulus [42] No Yes 𝑘 − 휀 and
algebraic stress
models
No single model is satisfactory in all
cases
Iacovides and Toumpanakis [43] No Yes 𝑘 − 휀 and 𝑘 − 𝜔
Both fail to predict the size of the
laminar region in the rotor boundary
layer
Elena and Schiestel [34] Yes Yes RSM Succeed to predict the cavity flow
Launder et al. [60] No Yes Second moment
closure approach Embedded pressure reflection impacts
Randriamampianina et al. [72] No Yes RSM More accurate predictions
Haddadi and Poncet [39] Yes Yes
Low Re second-
order full stress
closure
Different flow regimes coexist
Owen [65] No Yes RANS Asymmetrical and unsteady flow
Wu and Squires [89] No No LES
Promoting sweeps and ejections
contribute to the shear stress
production
Anderson and Lygren [4][5] No Yes DNS Isotropic eddy viscosity models
produce reliable results
Anderson and Lygren [5] No No LES Idealized model of an unshrouded
rotor-stator system
Séverac et al. [86] No Yes LES Can not predict the Large scale
rotating structures
Craft et al. [21] No Yes RANS No large scale structures in the core
region
Poncet et al. [69] Yes Yes RSM Fail to predict the influence of G
Itoh et al. [45] No Yes Realizable 𝑘 − 휀 Over prediction of the core rotation
Poncet et al. [69] Yes Yes RSM Fail to predict the influence of G
Will [88] No Yes SST 𝑘 − 𝜔 Closest to the pressure measurements
Barabas et al. [11] No Yes SST 𝑘 − 𝜔 Closest to the pressure measurements
Table 12: Selections of turbulence model in the literature
61
Will [88] compares the simulation results of both the tangential velocity and the radial velocity
from the realizable 𝑘 − 휀 turbulence model and the SST 𝑘 − 𝜔 turbulence model. He finds that
the velocity distributions obtained with the SST 𝑘 − 𝜔 turbulence model are closest to the
velocity measurements in rotor-stator cavities from the literature.
Barabas et al. [11] and Hu et al. [16] show that the simulation results of pressure from the SST
𝑘 − 𝜔 turbulence model in combination with the scalable wall functions is in good agreement with
those from pressure measurements in rotor-stator cavities with air and water.
The study by Volkov [51] focuses on the moment coefficients of the disk in an enclosed rotor-stator
cavity both numerically and experimentally. The moment coefficient from the SST 𝑘 − 𝜔
turbulence model shows good agreement with that from torque measurements, compared in Table
13. Hence, in this thesis, the SST 𝑘 − 𝜔 turbulence model is selected.
Models M (Nm) 𝐶𝑀 ∙ 103
𝑘 − 휀 2.3026 3.5115
𝑘 − 휀 2.4138 3.7125
SST 𝑘 − 𝜔 2.2350 3.4376
RSM 2.3729 3.6496
Table 13: Moment coefficients from different turbulence models (K. N. Volkov [51])
4.2 Grid Generation
One of the most time-consuming procedures in the CFD simulation is the generation of meshes for
the computational domains. The convergence of the numerical simulation and the accuracy of the
results from numerical simulation are strongly dependent on the quality of the meshes. In this thesis,
structured meshes are generated for all the simulation models with ICEM 14.0 and Turbogrid 14.0.
The wall laws of the SST 𝑘 − 𝜔 turbulence model have restrictions on the values of 𝑦+ (the non-
dimensional wall distance). The values of 𝑦+, used to describe how fine a mesh is for a particular
flow pattern, are important for determining the proper size of the elements near the walls. In the
fluid dynamics, the law of the wall shows that according to different flow, the boundary layer can
be divided into three parts from the wall to the free-stream flow region, namely the viscous sublayer,
the buffer layer and the log-law region. If the distance from the buffer layer to the viscous sublayer
is 𝜍, then the log-law region will extend to about 100𝜍 away from the wall. In both the viscous
sublayer and the buffer layer, the viscous force plays a dominant role, while the inertial force can
be ignored. The viscous force is linearly related to the velocity gradient. The first node should be
placed in the log-law region. The non-dimensional wall distance 𝑦+ can be predicted with the
friction velocity 𝑢∗, written in Eq.101.
62
𝑦+ =
𝑢∗ ∙ 𝑦
𝜈
Eq. 101
Where y is the distance from the first node to the wall.
The values of 𝑦+ are controlled below 10 in the grid generation processes for all the simulation
models. To estimate the values of y, the friction velocity 𝑢∗ should be estimated with Eq. 102. The
parameter 𝜏𝑤 is the wall shear stress, defined as 𝜏𝑤 = 𝜇 (𝜕𝑢
𝜕𝑦)
𝑦=0. The velocity gradient, however,
is not available. Hence, a new parameter, namely the skin friction coefficient 𝐶𝑓, is introduced to
predict the wall shear stress, written in Eq. 103.
𝑢∗ = √𝜏𝑤
𝜌
Eq. 102
𝐶𝑓 =𝜏𝑤
12 ∙ 𝜌 ∙ 𝑈2
Eq. 103
Where U is the velocity of the free stream.
Von Kármán [48] determines Eq. 104 to predict the values of 𝐶𝑓.
1
𝐶𝑓
12
= 1.7 + 4.15 ∙ log10(𝑅𝑒𝜑 ∙ 𝐶𝑓)
Eq. 104
Schlichting and Gersten [73] deduce Eq. 105 to predict the values of 𝐶𝑓 . The parameter 𝐶𝑓
increases with increasing roughness. As a consequence, the friction velocity 𝑢∗ increases.
𝐶𝑓 = [2 ∙ log10((𝑅𝑒𝜑) − 0.65]−2.3 ; 𝑅𝑒𝜑 < 109 Eq. 105
Above two equations, however, fail to consider the roughness effect, which results in large errors
during the predictions of 𝐶𝑓 and 𝑦+ . Schlichting and Gersten [73] propose Eq. 106 for 𝐶𝑓
including the roughness effect. With the increase of roughness, 𝐶𝑓 increases. As a consequence,
the friction velocity 𝑢∗ will also increase. The parameter 𝑙 is the streamwise coordinate.
𝐶𝑓 = (2.87 + 1.58 ∙ 𝑙𝑜𝑔10(𝑙/𝑘𝑠))−2.5 Eq. 106
63
Combining Eq. 101, Eq. 102, Eq. 103 and Eq. 106, the values of y during the generation of the meshes
can be estimated with Eq. 107 for rough walls.
𝑦 =𝑦+ ∙ 𝜈 ∙ √𝜌
√12
∙ 𝜌 ∙ 𝑈2 ∙ (2.87 + 1.58 ∙ log10(𝑙/𝑘𝑠))−2.5
Eq. 107
The simulation models are generated with structured meshes with different element numbers. When
𝑅𝑧 = 211 𝜇𝑚 ( 𝑘𝑠 = 81.3 𝜇𝑚 ), 𝐶𝐷 = −5050 and 𝑅𝑒 = 1.9 × 106 , the variation of the
maximum value of 𝑦+ versus the total element numbers from the numerical simulations are given
in Table 14. The qualities of the meshes are given in Table 15.
Total element numbers (106) 0.18 0.4 0.47 0.796 0.97 1.4 1.75
Maximal 𝑦+ 89 50 41 26 21 13.4 11.2
Table 14: Grid number and maximum values of 𝒚+
Criterion Determinant 3 × 3 × 3 Aspect ratio Angle
Values ≥ 0.89 ≤ 41 ≥ 27°
Table 15: Qualities of the meshes
4.3 Simulation Set-Up
To predict the cavity flow, numerical simulations are carried out using the ANSYS CFX 14.0 code.
There are 24 channels in the guide vane with small or large passages (see Fig. 37). The domain for
numerical simulation for 𝐺 = 0.072 is depicted with yellow color in Fig. 41. Considering the
axial symmetry of the problem, a segment of 15 degrees (with one radial guide vane passage) of
the whole domain is modeled and a rotational periodic boundary condition is applied. The
discretization scheme is set as second order upwind. The simulation type is set as steady state. The
convergence criteria are set as 10−5 in maximum type. The no-slip wall condition in relative
frame is set for all the walls.
For non-pre-swirl through-flow, the boundary conditions at the inlet and at the outlet respectively
are pressure inlet and mass flow outlet. The values of pressure at the inlet are set according to the
pressure sensor at the pump outlet.
For centripetal pre-swirl through-flow, the boundary condition at the inlet is velocity inlet (with
both the radial and the tangential velocity components) and at the outlet is pressure outlet (1 bar
reference pressure).
64
Centripetal through-flow
Centrifugal through-flow
Fig. 41: Domain for numerical simulation (𝑮 = 𝟎. 𝟎𝟕𝟐)
For centrifugal pre-swirl through-flow, the geometry of the swirlers is considered when generating
the simulation models and the meshes. Additional fluid domains with different pre-swirl angles are
connected to the inlet (at the horizontal pipe), depicted in Fig. 42. The length of the domain is five
times the inner diameter of the horizontal pipe.
𝛽 = 15° 𝛽 = 45°
Fig. 42: Additional fluid domain for centrifugal pre-swirl through-flow
Mesh sensitivity analyses are accomplished for all the simulation models. The pressure and the
tangential velocity at 𝑥 = 0.955 and 휁 = 0.5 for 𝐺 = 0.072 , 𝐶𝐷 = −5050 and 𝑅𝑒 = 1.9 ×
106 are compared in Fig. 43. The parameter Δ is a non-dimensional parameter which shows the
relative change rate of the results for different total element numbers, defined in Eq. 108. The
parameter 𝐶𝑥 represents the value of a certain variable when the total element number is x. The
parameter 𝐶1.75 is the value of the variable when the total element number is 1.75 million. When
the total element numbers exceed 1.4 million, the results are independent of the total element
numbers. Hence, the total element numbers should exceed 1.4 million to improve the accuracy of
Disk
Domain
Casing
Shaft
Disk
Domain
Casing
Shaft
Q
Q
Q
𝛽
65
the numerical simulations.
∆= |
𝐶1.75 − 𝐶𝑥
𝐶1.75|
Eq. 108
∆
Fig. 43: Mesh independence analysis
4.4 Validation for Numerical Simulation
In Fig. 44, the results from numerical simulations for 𝐶𝑝 along the radius are compared with those
from pressure measurements by Poncet et al. [68] at 𝑅𝑒 = 4.15 × 106 and 𝐺 = 0.036 in a
rotor-stator cavity with centrifugal through-flow. The reference pressures are taken at 𝑥 = 0.92
instead of 𝑥 = 1 to compare the results with those from the measurements by Poncet at al. [68].
The simulation results of the pressure are in very good agreement with the measurements in Poncet
at al. [68].
0
0.02
0.04
0.06
0.08
0.1
0.12
0 0.5 1 1.5 2
Total grid number (106)
Pressure
Tangential velocity
66
X
Colors: Black: 𝐶𝐷 = 0, Red: 𝐶𝐷 = 2579, Green: 𝐶𝐷 = 5159; Solid lines: RSM from Poncet et al. [68]; Dashed
lines: Simulation results (SST 𝑘 − 𝜔); Points: Measurements by Poncet et al. [68].
Fig. 44: Comparison of radial pressure distribution for 𝑹𝒆 = 𝟒. 𝟏𝟓 × 𝟏𝟎𝟔 and 𝑮 = 𝟎. 𝟎𝟑𝟔
In Fig. 45, the results from numerical simulation for 𝐶𝑝 along the radius are compared with those
from the pressure measurements by Kurokawa et al. [52]. The values of 𝐶𝑞 are negative for
centripetal through-flow. The results of the simulations and the measurements are in a very good
accordance.
X
Dashed lines: Simulation results (SST 𝑘 − 𝜔); Points: Measurements by Kurokawa et al. [52].
Fig. 45: Comparisons of radial pressure distribution for 𝑹𝒆 = 𝟏. 𝟑𝟔 × 𝟏𝟎𝟔 and 𝑮 = 0.0495
without pre-swirl
The comparison of the simulation results with the experimental results from the literature shows
that the simulation set-up should be reasonable.
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0 0.02 0.04 0.06 0.08 0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0 0.05 0.1 0.15 0.2 0.25 0.3
𝐶𝑝
𝐶𝑝
𝐶𝑞 = 0
𝐶𝑞 = −0.00273
𝐶𝑞 = −0.00829
𝑄
𝛺
𝑄 𝛺
67
5. Results and Discussion
According to the results from the literature, the most influential parameters in a rotor-stator cavity
are 𝐶𝐷, G, Re, 𝑘𝑠 and 𝐶𝑎𝑚. This section is divided into two parts on the basis of through-flow
direction. In each part, first the through-flow without pre-swirl is considered in a rotor-stator cavity
varying the values of G, Re and 𝐶𝐷 using a disk with a smooth surface (𝑘𝑠 = 0.4 𝜇m). Afterwards,
the influence of the disk surface roughness is taken into consideration. In the third step, the impact
of centripetal or centrifugal pre-swirl through-flow on K, 𝐶𝑝 , 𝐶𝐹 and 𝐶𝑀 is investigated.
According to the results from both the numerical simulations and the measurements, the flow in a
rotor-stator cavity is analyzed with both centripetal (Hu et al. [16][17]) and centrifugal (Hu et al.
[18]) through-flow. The research procedure of this chapter is depicted in Fig. 46.
Fig. 46: Research procedure
5.1 Rotor-Stator Cavity with Centripetal Through-Flow
5.1.1 Simulation Results of Velocity Distributions
The inlet boundary condition greatly affects the velocity profiles. The non-dimensional velocities
in the front chamber therefore are discussed in this part. There is a small jet flow through each
channel in the radial guide vane (see Fig. 37). All presented velocities are made non-dimensional
with 𝛺 ∙ 𝑏 . The values of 𝑉𝑧 (non-dimensional axial velocity) are positive when they have a
direction from the disk to the wall. The velocity profiles at three radial positions for 𝑅𝑒 = 1.9 ×
106 and 𝐺 = 0.072 (small passages in the guide vane) are shown in Fig. 47. The non-
dimensional radial velocities 𝑉𝑟 are not exactly zero in the central cores, shown in Fig. 47 (a to c).
The distribution of 𝑉𝜑 (non-dimensional tangential velocity) shows that there are central cores at
all the investigated radial positions with nearly zero velocity gradient, shown in Fig. 47 (d to f). At
𝑥 = 0.955 and 𝑥 = 0.79 , the values of 𝑉𝜑 are smaller at 휁 = 0.5 when |𝐶𝐷| increases from
Par
amet
ers:
K ,
𝐶𝑝
, 𝐶
𝐹,
𝐶𝑀
Introduce through-flow (Smooth disk)
Introduce disk roughness
Introduce pre-swirl (Smooth disk)
Introduce through-flow (Smooth disk)
Introduce disk roughness
Introduce pre-swirl (Smooth disk)
5.1: Centripetal
5.2: Centrifugal
Step 1
Step 1
Step 2
Step 2
Step 3
Step 3
68
1262 to 3787 and 5050, depicted in Fig. 47 (d, e). The profiles contrast with the predictable pattern
according to the measurements from Poncet et al. [68] and Debuchy et al. [30]. Presumably, the
increase of |𝐶𝐷| should lead to in an increase of K for centripetal through-flow. As a consequence,
the tangential velocity should increase rather than decrease. Probably this disagreement can be
attributed to the jet flow through the guide vane channel at the inlet, which is stronger for large
|𝐶𝐷|. At 𝑥 = 0.955, the values of |𝑉𝑧| become smaller for larger |𝐶𝐷| in general, shown in Fig.
47 (g). The direction of 𝑉𝑧 is from the disk towards the wall for |𝐶𝐷| = 0 and |𝐶𝐷| = 1262 at
𝑥 = 0.955, while it is from the wall towards the disk at |𝐶𝐷| = 3787 and |𝐶𝐷| = 5050, but with
a very small amount. From the axial velocity profiles, the axial circulations of the fluid exist in the
front chamber.
𝑥 = 0.955 𝑥 = 0.79 𝑥 = 0.57
𝑉𝑟 (a) 𝑉𝑟 (b) 𝑉𝑟 (c)
𝑉𝜑 (d) 𝑉𝜑 (e) 𝑉𝜑 (f)
𝑉𝑧 (g) 𝑉𝑧 (h) 𝑉𝑧 (i)
Colors: Black: |𝐶𝐷| = 0; Blue: |𝐶𝐷| = 1262; Green: |𝐶𝐷| = 3787; Red: |𝐶𝐷| = 5050.
Fig. 47: Velocity profiles for 𝑹𝒆 = 𝟏. 𝟗 × 𝟏𝟎𝟔 and 𝑮 = 𝟎. 𝟎𝟕𝟐
-0.2
-0.15
-0.1
-0.05
0
0.05
0.1
0.15
0 0.5 1-0.2
-0.15
-0.1
-0.05
0
0.05
0.1
0.15
0 0.5 1-0.2
-0.15
-0.1
-0.05
0
0.05
0.1
0.15
0 0.5 1
0
0.2
0.4
0.6
0.8
1
0 0.5 1
0
0.2
0.4
0.6
0.8
1
0 0.5 10
0.2
0.4
0.6
0.8
1
0 0.5 1
-0.03
-0.01
0.01
0.03
0 0.5 1
-0.03
-0.01
0.01
0.03
0 0.5 1
-0.03
-0.01
0.01
0.03
0 0.5 1
휁 휁 휁
휁 휁 휁
Disk Wall
휁 휁 휁
69
The velocity profiles at three radial coordinates for 𝑅𝑒 = 1.9 × 106 and 𝐺 = 0.018 (small gap)
are depicted in Fig. 48. The values of 𝑉𝑟 vary along 휁, shown in Fig. 48 (a to c). Near the wall,
they mainly increase with increasing |𝐶𝐷| and become negative all along 휁 for high through-flow
rate. The tangential velocity 𝑉𝜑 decreases constantly along 휁, shown in Fig. 48 (d to f), which is
the characteristic of flow regime III. At 𝑥 = 0.955 and 𝑥 = 0.79 , the values of tangential
velocity are much smaller for large |𝐶𝐷|, shown in Fig. 48 (d). The reason is that the impact of the
jet flow at the inlet becomes stronger for smaller G. The profiles of 𝑉𝑧 are quite different at 𝑥 =
0.955 in Fig. 48 (g), compared with those in Fig. 47 (g). Most of the values of |𝑉𝑧| are smaller
than those from Fig. 48 (h, i), which indicates the weaker axial circulations of the fluid for smaller
G.
𝑥 = 0.955 𝑥 = 0.79 𝑥 = 0.57
𝑉𝑟 (a) 𝑉𝑟 (b) 𝑉𝑟 (c)
𝑉𝜑 (d) 𝑉𝜑 (e) 𝑉𝜑 (f)
𝑉𝑧 (g) 𝑉𝑧 (h) 𝑉𝑧 (i)
Colors: Black: |𝐶𝐷| = 0; Blue: |𝐶𝐷| = 1262; Green: |𝐶𝐷| = 3787; Red: |𝐶𝐷| = 5050.
Fig. 48: Velocity profiles for 𝑹𝒆 = 𝟏. 𝟗 × 𝟏𝟎𝟔 and 𝑮 = 𝟎. 𝟎𝟏𝟖
-0.2
-0.15
-0.1
-0.05
0
0.05
0.1
0.15
0 0.5 1-0.2
-0.15
-0.1
-0.05
0
0.05
0.1
0.15
0 0.5 1
-0.2
-0.15
-0.1
-0.05
0
0.05
0.1
0.15
0 0.5 1
0
0.2
0.4
0.6
0.8
1
0 0.5 1
0
0.2
0.4
0.6
0.8
1
0 0.5 10
0.2
0.4
0.6
0.8
1
0 0.5 1
-0.05
-0.03
-0.01
0.01
0.03
0.05
0 0.5 1-0.05
-0.03
-0.01
0.01
0.03
0.05
0 0.5 1
-0.05
-0.03
-0.01
0.01
0.03
0.05
0 0.5 1
휁 휁 휁
Disk Wall
휁 휁 휁
휁 휁 휁
70
5.1.2 Core Swirl Ratio
5.1.2.1 Impact of Through-Flow with a Smooth Disk
To predict the axial thrust, the distribution of K along the radius should be estimated. In a majority
of the rotor-stator cavities in radial pumps or turbines, measuring the velocities is quite expensive
and complicated. By derivation of equations, Will et al. [85][86][87][88] deduce Eq. 71 to associate
K with the measured pressure. The average values of K between each two adjacent pressure tubes
(12 tubes from 𝑟 = 0.05 m (𝑥 = 0.455 ) to 𝑟 = 0.105 m (𝑥 = 0.954 )), noted as 𝐾𝑐~c+1̅̅ ̅̅ ̅̅ ̅̅ ̅ , are
calculated according to the pressure measurements with Eq. 109. Since the radial distances between
the adjacent pressure tubes are small, the application of the average values of K results a small error
only. The values of 𝐶𝑞𝑟 for the experimental results are calculated with 𝑟 =𝑟𝑐+𝑟𝑐+1
2.
𝐾𝑐~𝑐+1̅̅ ̅̅ ̅̅ ̅̅ ̅ = √
𝑝(𝑟𝑐) − 𝑝(𝑟𝑐+1) −𝜌 ∙ 𝑄2
8 ∙ 𝜋2 ∙ 𝑠2 (1
𝑟𝑐+12 −
1𝑟𝑐
2)
12 ∙ 𝜌 ∙ 𝛺2 ∙ (𝑟𝑐
2 − 𝑟𝑐+12)
Eq. 109
Poncet et al. [68] and Debuchy et al. [30] neglect the effect of G on K according to the LDA
measurements. In most of the radial pumps or turbines, the wall is not parallel to the disk, resulting
in a variable G. A simplified correlation with good accuracy can help to easily account for varying
G. Based on the pressure measurements of the author, Eq. 110 is correlated to predict the values of
K when G ranges from 0.018 to 0.072 (Batchelor or Couette type flow) in a rotor-stator cavity with
centripetal through-flow. The results from Eq. 110 are in good accordance with the simulation results
and those from Eq. 51 and Eq. 52, depicted in Fig. 49.
𝐾 = 0.97 ∙ [
−8.5∙𝐶𝑞𝑟+0.5
𝑒(−1.45𝐶𝑞𝑟)]
5
4 Eq. 110
Where −0.5 ≤ 𝐶𝑞𝑟 ≤ 0.03.
71
Hollow points: Simulation; Solid points: By pressure measurements (Eq. 109); Dashed lines: Equations.
Colors (points): Black: 𝐺 = 0.018; Green: 𝐺 = 0.036; Yellow: 𝐺 = 0.054; Red: 𝐺 = 0.072.
Colors (Dashed lines): Red: Eq. 51; Green: Eq. 52; Purple: Eq. 110.
Fig. 49: K (𝑪𝒒𝒓) curves for centripetal through-flow
5.1.2.2 Impact of Surface Roughness of the Disk
The distributions of K along a rough disk are estimated with Eq. 109 in a rotor-stator cavity with
centripetal through-flow. Based on the pressure measurements, Eq. 110 is extended by introducing
𝑘𝑠, written in Eq. 111.
𝐾 = 0.97 ∙ 𝑒
(600∙𝑘𝑠∙𝑟
𝑏2 )∙ [
−8.5 ∙ 𝐶𝑞𝑟 + 0.5
𝑒(−1.45𝐶𝑞𝑟)]
54
Eq. 111
Where 0.018 ≤ 𝐺 ≤ 0.072;
𝑅𝑒 ≤ 3.17 × 106;
−5050 ≤ 𝐶𝐷 ≤ 0 and 𝑘𝑠 ≤ 58.9 𝜇m.
0
0.5
1
1.5
2
2.5
3
-0.5 -0.4 -0.3 -0.2 -0.1 0
𝐶𝑞𝑟
𝐾
72
Most of the experimental results of K by Kurokawa et al. [54] are in good agreement with those
from Eq. 111, depicted in Fig. 50. Relatively large errors only occur when 𝑘𝑠 = 56 𝜇m. Near the
outer radius of the disk (where the values of |𝐶𝑞𝑟| are of small values), K increases faster for larger
𝑘𝑠 , which is in consistent with the trend of the measurements by Kurokawa et al. [54] in an
enclosed rotor-stator cavity.
Solid points: By pressure measurements (Eq. 109); Dashed lines: Eq. 111.
Colors : Blue: 𝑘𝑠 = 0.6 𝜇m; Red: 𝑘𝑠 = 22 𝜇m; Green: 𝑘𝑠 = 24 𝜇m; Red: 𝑘𝑠 = 56 𝜇m.
Fig. 50: K (x) curves for 𝑮 = 𝟎. 𝟎𝟑𝟏, 𝑹𝒆 =3.1× 𝟏𝟎𝟔 and 𝑪𝑫 = 𝟎 by Kurokawa et al. [54]
The impact of the surface roughness is analyzed according to the results of 𝐾/𝐾𝑘𝑠=0.4 𝜇m. Since
the results of K are almost not affected by changing G, only the results when 𝐺 = 0.072 are
compared, shown in Fig. 51. The values of K increase with increasing 𝑘𝑠 and |𝐶𝑞𝑟|. The results
indicate that the impact of surface roughness can be ignored for the two hydraulic smooth disks
(𝐾𝑘𝑠=3.5 𝜇m ≈ 𝐾𝑘𝑠=0.4 𝜇m with given 𝐶𝑞𝑟). For a disk in the transition zone (11.4 𝜇m ≤ 𝑘𝑠 ≤
58.9 𝜇m), the results of K can be predicted using Eq. 111 with good accuracy.
0.4
0.5
0.6
0.7
0.8
0.9
1
0.4 0.45 0.5 0.55 0.6
𝐾
𝑥
73
Hollow points: Simulation; Solid points: By pressure measurements (Eq. 109); Dashed lines: Eq. 111.
Colors : Black: 𝑘𝑠 = 3.5 𝜇m; Green: 𝑘𝑠 = 11.4 𝜇m; Yellow: 𝑘𝑠 = 26.5 𝜇m; Red: 𝑘𝑠 = 58.9 𝜇m.
Fig. 51: Impact of 𝒌𝒔 on K when G=0.072
5.1.2.3 Impact of Centripetal Pre-Swirl Through-Flow with a Smooth Disk
The inlet boundary conditions for the small and the large guide vane passages are depicted in Fig.
52. The simulation results of 𝐾𝑝 are compared with those by pressure measurements with Eq. 109
for 𝐺 = 0.072 , 𝑅𝑒 = 1.9 × 106 , 𝐶𝐷 = −5050 , depicted in Fig. 53. Based on the results, the
boundary conditions at the centripetal inlet (see Fig. 37) play an important role in the distributions
of 𝐾𝑝. The shaft rotates in the opposite direction to get negative values of pre-swirl angle 𝛽 (flow
counter-rotates to the shaft when entering the cavity). For small guide vane passages where the jet
flow occurs, the values of 𝐾𝑝 are much smaller compared with those with large guide vane
passages. The jet flow is weakening the rotation of the fluid. Currently, it is quite hard to quantify
the impact of 𝐶𝑎𝑚 on 𝐾𝑝, which deserves further investigations.
Small guide vane passages Large guide vane passages
Fig. 52: Inlet boundary conditions with centripetal through-flow
1
1.05
1.1
1.15
1.2
1.25
-0.5 -0.4 -0.3 -0.2 -0.1 0
𝐶𝑞𝑟
𝐾/𝐾
𝑘𝑠=
0.4
𝜇
m
Disk Disk + −
𝛺 − +
𝛺
74
x Small guide vane passages x Large guide vane passages
Solid points: By pressure measurements (Eq. 109); Solid lines: Numerical simulation.
Colors: Gray: 𝐶𝑎𝑚 = −0.0078 (𝛽 = 60°); Red: 𝐶𝑎𝑚 = −0.0075 (𝛽 = 52°); Blue: 𝐶𝑎𝑚 =
−0.0069 (𝛽 = 26°); Green: 𝐶𝑎𝑚 = 0 (without pre-swirl); Purple: 𝐶𝑎𝑚 = 0.0083 (𝛽 = −26°);
Orange: 𝐶𝑎𝑚 = 0.0089 (𝛽 = −52°); Dark blue: 𝐶𝑎𝑚 = 0.0093 (𝛽 = −60°).
Fig. 53: 𝑲𝒑 (𝒙) curves for 𝑮 = 𝟎. 𝟎𝟕𝟐, 𝑪𝑫 = −𝟓𝟎𝟓𝟎 and 𝑹𝒆 = 𝟏. 𝟗 × 𝟏𝟎𝟔
5.1.3 Pressure Distribution
5.1.3.1 Impact of Through-Flow with a Smooth Disk
The pressure measurements are performed to analyze the impact of centripetal through-flow on 𝐶𝐹
with a smooth disk (𝑘𝑠 = 0.4 𝜇m). Based on Eq. 71, the pressure along the radius can be calculated
with Eq. 112 based on the values of K from Eq. 110. K is a variable along the radius of the disk. A
simplification is made as follows: K is a fixed value every 1 mm in the radial direction. The
parameter 𝑝𝑏 represents the pressure at 𝑥 = 1. Due to the construction of the geometry, however,
there is no pressure tube at 𝑥 = 1 . The values of 𝑝𝑏 are calculated combining the measured
pressure at 𝑥 = 0.955 (𝑟 = 0.105 m) with Eq. 112.
𝑝(𝑟) = 𝑝𝑏 + ∫ 𝜌 ∙ 𝐾2 ∙ 𝛺2 ∙ 𝑟𝑑𝑟
𝑟
𝑏
+𝜌 ∙ 𝑄2
8 ∙ 𝜋2 ∙ 𝑠2(
1
𝑏2−
1
𝑟2)
Eq. 112
Where ∫ 𝜌 ∙ 𝐾2 ∙ 𝛺2 ∙ 𝑟𝑑𝑟𝑟
𝑏≈
𝜌
2∙ 𝛺2 ∙ ∑ 𝐾𝑟𝑖+1
2 ∙ (𝑟𝑖2 − 𝑟𝑖+1
2)𝑐−1
0 ;
𝑟 = 𝑏 − 0.001 ∙ 𝑐 (m);
𝑟𝑖 − 𝑟𝑖+1 = −0.001 (m).
Since the pressure drops towards the shaft, the values of the pressure coefficient 𝐶𝑝 are positive
values. In Fig. 54, 𝐶𝑝 is plotted versus x for three values of Re and G (with small passages in the
guide vane). The pressure measurements show that the values of 𝐶𝑝 increase with increasing |𝐶𝐷|.
At a given |𝐶𝐷|, 𝐶𝑝 decrease with the increase of Re and G. When 𝑅𝑒 = 1.9 × 106 and 𝑅𝑒 =
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
-0.2 0 0.2 0.4 0.6
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
-0.2 0 0.2 0.4 0.6 0.8 1𝐾𝑝
𝑥ℎ 𝑥ℎ
𝐾𝑝
75
2.79 × 106 , the uncertainties of the 𝐶𝑝 are 2.7 × 10−4 and 1.3 × 10−4 , respectively. This is
very small compared with the measured results. Hence, they are not plotted in Fig. 54 (d~i).
G=0.018 G=0.036 G=0.072
Re=
0.76
×1
06
x (a) x (b) x (c)
Re=
1.9×
10
6
x (d) x (e) x (f)
Re=
2.79
×1
06
x (g) x (h) x (i)
Solid points: Pressure measurements; Solid lines: Numerical simulation.
Colors: Black: |𝐶𝐷| = 0; Blue: |𝐶𝐷| = 1262; Green: |𝐶𝐷| = 3787; Red: |𝐶𝐷| = 5050.
Fig. 54: Influence of 𝑪𝑫 on 𝑪𝒑 in dependence of Re and G (𝒌𝒔 = 𝟎. 𝟒 𝝁𝐦)
5.1.3.2 Impact of Surface Roughness of the Disk
When |𝐶𝐷|, Re and G are given, the amounts of K increase with increasing 𝑘𝑠 at a given radial
position. This is in accordance with the velocity measurements by Kurokawa et al. [54]. Hence,
the pressure drops faster towards the shaft with rougher disks. The values of 𝐶𝑝 when 𝐺 = 0.072
from Eq. 112 (K is calculated with Eq. 111) are in good accordance with those by pressure
measurements with large passages in the guide vane, presented in Fig. 55. All the results indicate
that Eq. 112 can predict the effect of 𝑘𝑠 on K with good accuracy when 𝑘𝑠 ≤ 58.9 𝜇m.
0.4
0.6
0.8
1
0 0.05 0.1 0.150.4
0.6
0.8
1
0 0.05 0.1 0.150.4
0.6
0.8
1
0 0.05 0.1 0.15
0.4
0.6
0.8
1
0 0.05 0.1 0.150.4
0.6
0.8
1
0 0.05 0.1 0.15
0.4
0.6
0.8
1
0 0.05 0.1 0.15
0.4
0.6
0.8
1
0 0.05 0.1 0.15
0.4
0.6
0.8
1
0 0.05 0.1 0.150.4
0.6
0.8
1
0 0.05 0.1 0.15
𝐶𝑝
|𝐶𝐷|
𝐶𝑝 𝐶𝑝 𝐶𝑝
𝐶𝑝 𝐶𝑝
𝐶𝑝 𝐶𝑝 𝐶𝑝
|𝐶𝐷| |𝐶𝐷|
|𝐶𝐷| |𝐶𝐷| |𝐶𝐷|
|𝐶𝐷| |𝐶𝐷| |𝐶𝐷|
76
|𝐶𝐷| =0 |𝐶𝐷| =2525 |𝐶𝐷| =5050 G
=0
.072,
𝑘
𝑠=
0.4
𝜇m
x x x
x x x
G=
0.0
72,
𝑘
𝑠=
3.5
𝜇m
G=
0.0
72,
𝑘
𝑠=
11.4
𝜇m
x x x
G=
0.0
72,
𝑘
𝑠=
26.5
𝜇m
x x x
0.5
0.6
0.7
0.8
0.9
1
0 0.05 0.10.5
0.6
0.7
0.8
0.9
1
0 0.1 0.20.5
0.6
0.7
0.8
0.9
1
0 0.1 0.2 0.3
0.5
0.6
0.7
0.8
0.9
1
0 0.05 0.10.5
0.6
0.7
0.8
0.9
1
0 0.1 0.2
0.5
0.6
0.7
0.8
0.9
1
0 0.1 0.2 0.3
0.5
0.6
0.7
0.8
0.9
1
0 0.05 0.10.5
0.6
0.7
0.8
0.9
1
0 0.1 0.2
0.5
0.6
0.7
0.8
0.9
1
0 0.1 0.2 0.3
0.5
0.6
0.7
0.8
0.9
1
0 0.05 0.1
0.5
0.6
0.7
0.8
0.9
1
0 0.1 0.20.5
0.6
0.7
0.8
0.9
1
0 0.1 0.2 0.3
𝐶𝑝
Re
𝐶𝑝
𝐶𝑝 𝐶𝑝
𝐶𝑝 𝐶𝑝 𝐶𝑝
𝐶𝑝 𝐶𝑝 𝐶𝑝
𝐶𝑝
𝐶𝑝
Re
Re
Re
Re Re
Re Re Re
Re Re Re
77
G=
0.0
72,
𝑘
𝑠=
58
.9 𝜇
m x x x
Solid points: Pressure measurements; Dashed lines: Numerical simulation.
Colors: Red: 𝑅𝑒 = 0.76 × 106; Purple: 𝑅𝑒 = 1.14 × 106; Green: 1.90 × 106; Yellow: 𝑅𝑒 = 3.17 × 106.
Fig. 55: 𝑪𝒑 (x) curves along the radius of the disks
5.1.3.3 Impact of Pre-Swirl on the Pressure Distribution with a Smooth Disk
In Fig. 56, the impacts of pre-swirl on the pressure distribution for 𝐺 = 0.072, 𝐶𝐷 = −5050 and
𝑅𝑒 = 1.9 × 106 with small or large guide vane passages are presented. With increasing 𝛽
(−𝐶𝑎𝑚), the centripetal through-flow brings larger tangential velocity component into the cavity
(the values of 𝐾 become larger). Hence, greater values of 𝛽 result in a larger pressure drop
towards the shaft. The results of 𝐶𝑝 from numerical simulation are in relative good agreement
with those from pressure measurements.
x Small guide vane passage x Large guide vane passage
Solid points: Pressure measurements; Solid lines: Numerical simulation.
Colors: Gray: 𝐶𝑎𝑚 = −0.0078 (𝛽 = 60°); Red: 𝐶𝑎𝑚 = −0.0075 (𝛽 = 52°); Blue: 𝐶𝑎𝑚 = −0.0069 (𝛽 =
26°); Green: 𝐶𝑎𝑚 = 0 (without pre-swirl); Purple: 𝐶𝑎𝑚 = 0.0083 (𝛽 = −26°); Orange: 𝐶𝑎𝑚 =
0.0089 (𝛽 = −52°); Dark blue: 𝐶𝑎𝑚 = 0.0093 (𝛽 = −60°).
Fig. 56: 𝑪𝒑 (x) curves for 𝑮 = 𝟎. 𝟎𝟕𝟐, 𝑪𝑫 = −𝟓𝟎𝟓𝟎 and 𝑹𝒆 = 𝟏. 𝟗 × 𝟏𝟎𝟔 with centripetal
through-flow
0.5
0.6
0.7
0.8
0.9
1
0 0.05 0.1
0.5
0.6
0.7
0.8
0.9
1
0 0.1 0.20.5
0.6
0.7
0.8
0.9
1
0 0.1 0.2 0.3
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0 0.1 0.2 0.3
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0 0.12 0.24 0.36
𝐶𝑝 𝐶𝑝
𝐶𝑝
Disk
𝐶𝑝
Disk
𝐶𝑝
Re
Re Re
78
5.1.4 Thrust Coefficient
5.1.4.1 Impact of Through-Flow with a Smooth Disk
According to the axial thrust measurements, a correlation for the thrust coefficient 𝐶𝐹 is
determined for a smooth disk (𝑘𝑠 = 0.4 𝜇m), given in Eq. 113. The through-flow will result in a
significant decrease of the pressure towards the shaft, compared with the case for an enclosed rotor-
stator cavity. Kurokawa et al. [53] mention that 𝐶𝐹 is not affected so much by the through-flow
rate without inlet angular momentum. However, according to their experimental results, shown in
Fig. 57, the values of 𝐶𝐹 decrease by around 30% with the increase of 𝐶𝐷. The conclusion is
therefore considered not precise. Although the results from Eq. 113 are in line with the trend of the
experimental results in Kurokawa et al. [53], there is a large gap between them at a given set of
𝐶𝐷 and 𝐺. Possibly the differences can be attributed to the uncertainty of the surface roughness of
the disks.
𝐶𝐹 = [6.6 ∙ 10−3 ∙ 𝑙𝑛 (𝑅𝑒) − 0.113] ∙ 𝑒(−1.2∙10−4∙𝐶𝐷) ∙ [0.122 ∙ 𝑙𝑛(𝐺) − 0.67] Eq. 113
𝐶𝐹
Solid points: Pressure measurements in Kurokawa et al. [53]; Solid lines: Eq. 113; Dashed lines: Theoretical analyses in
Kurokawa et al. [53].
Colors: Blue: 𝐺 = 0.024; Yellow: 𝐺 = 0.048; Red: 𝐺 = 0.097.
Fig. 57: 𝑪𝑭 (𝑪𝑫) curves for 𝑹𝒆 = 𝟏. 𝟑𝟐 × 𝟏𝟎𝟔
The values of 𝐶𝐹 increase with increasing |𝐶𝐷|, while decrease with increasing Re, plotted in Fig.
58. For increasing |𝐶𝐷| , the values of 𝐶𝐹 increase, which can be attributed to the drop of the
pressure. The values of 𝐶𝐹 are smaller for higher values of G and Re.
0
0.01
0.02
0.03
0.04
0.05
-3000 -2000 -1000 0
𝐶𝐷
79
𝐶𝐹 𝐺 = 0.018 𝐶𝐹 𝐺 = 0.036
𝐶𝐹 𝐺 = 0.054 𝐶𝐹 𝐺 = 0.072
Solid points: Axial thrust measurements; Dashed lines: Eq. 113.
Colors: Black: |𝐶𝐷| = 0; Blue: |𝐶𝐷| = 1262; Yellow: |𝐶𝐷| = 2525; Green: |𝐶𝐷| = 3787; Red: |𝐶𝐷| = 5050.
Fig. 58: 𝑪𝑭 (𝑹𝒆) curves in dependence of 𝑪𝑫 and G
5.1.4.2 Impact of Surface Roughness of the Disk
Based on the experimental results, an empirical correlation for the thrust coefficient is determined
to predict the thrust coefficient for rough disks, given in Eq. 114.
𝐶𝐹 = [6.6 ∙ 10−3 ∙ ln(𝑅𝑒) − 0.113] ∙ 𝑒(−1.2∙10−4∙𝐶𝐷) ∙ [0.122 ∙ ln(𝐺) − 0.67] ∙ 𝑒(880∙
𝑘𝑠𝑏
) Eq. 114
Where 0.018< 𝐺 ≤0.072;
𝑅𝑒 ≤ 3.17 × 106;
−5050 ≤ 𝐶𝐷 ≤ 0 and 𝑘𝑠 ≤ 58.9 𝜇m.
The values of 𝐶𝐹 can also be directly calculated according to the pressure distribution with Eq. 115.
𝐶𝐹 =
∫ 2𝜋 ∙ (𝑝𝑏 − 𝑝1) ∙ 𝑟𝑑𝑟𝑟=𝑏
𝑟=𝑟𝑖1
𝜌 ∙ 𝜋 ∙ 𝛺2 ∙ 𝑏4
Eq. 115
0.01
0.03
0.05
0.07
0.3 0.8 1.3 1.8 2.3 2.8 3.30.01
0.03
0.05
0.07
0.3 0.8 1.3 1.8 2.3 2.8 3.3
0.01
0.03
0.05
0.07
0.3 0.8 1.3 1.8 2.3 2.8 3.3
0.01
0.03
0.05
0.07
0.3 0.8 1.3 1.8 2.3 2.8 3.3
Re (106)
Re (106) Re (106)
Re (106)
|𝐶𝐷| |𝐶𝐷|
|𝐶𝐷| |𝐶𝐷|
80
The results of 𝐶𝐹 from Eq. 114 are in good agreement with those from Eq. 115 and thrust
measurements for 𝑘𝑠 ≤ 58.9 𝜇m, depicted in Fig. 59. The impact of 𝑘𝑠 on 𝐶𝐹 is strong. With
the increase of |𝐶𝐷|, the values of 𝐶𝐹 increase, which can be attributed to the drop of the pressure.
The values of 𝐶𝐹 increase with increasing 𝑘𝑠 at a given Re. The results also indicate that Eq. 111
is capable to describe the impact of 𝑘𝑠 on K in a rotor-stator cavity with centripetal through-flow
when 𝑘𝑠 ≤ 58.9 𝜇m.
𝐺 = 0.018 𝐺 = 0.036 𝐺 = 0.072
𝑘𝑠
=0
.4
𝜇m
𝐶𝐹 𝐶𝐹 𝐶𝐹
𝐶𝐹 𝐶𝐹 𝐶𝐹
𝑘𝑠
=3
.5
𝜇m
𝐶𝐹 𝐶𝐹 𝐶𝐹
𝑘𝑠
=1
1.4
𝜇
m
0
0.02
0.04
0.06
0.08
0.1
0.3 1.3 2.3 3.3
0
0.02
0.04
0.06
0.08
0.1
0.3 1.3 2.3 3.3
百万
0
0.02
0.04
0.06
0.08
0.1
0.3 1.3 2.3 3.3
0
0.02
0.04
0.06
0.08
0.1
0.3 1.3 2.3 3.3
0
0.02
0.04
0.06
0.08
0.1
0.3 1.3 2.3 3.3
百万
0
0.02
0.04
0.06
0.08
0.1
0.3 1.3 2.3 3.3
0
0.02
0.04
0.06
0.08
0.1
0.3 1.3 2.3 3.3
0
0.02
0.04
0.06
0.08
0.1
0.3 1.3 2.3 3.30
0.02
0.04
0.06
0.08
0.1
0.3 1.3 2.3 3.3
|𝐶𝐷|
Re
(106)
Re
(106)
Re
(106)
Re
(106)
Re
(106)
|𝐶𝐷| |𝐶𝐷|
|𝐶𝐷| |𝐶𝐷| |𝐶𝐷|
|𝐶𝐷|
Re
(106)
Re
(106)
Re
(106)
Re
(106)
|𝐶𝐷| |𝐶𝐷|
81
𝐶𝐹 𝐶𝐹 𝐶𝐹
𝑘𝑠
=2
6.5
𝜇m
𝐶𝐹 𝐶𝐹 𝐶𝐹
𝑘𝑠
=5
8.9
𝜇
m
Solid points: Axial thrust measurements; Hollow points: Eq. 115; Dashed lines: Eq. 114.
Colors: Black: |𝐶𝐷| = 0; Blue: |𝐶𝐷| = 1262; Yellow: |𝐶𝐷| = 2525; Green: |𝐶𝐷| = 3787; Red: |𝐶𝐷| = 5050.
Fig. 59: 𝑪𝑭 (𝑹𝒆) curves in dependence of 𝑪𝑫, G and 𝒌𝒔
5.1.4.3 Impact of Pre-Swirl on the Thrust Coefficient with a Smooth Disk
Kurokawa et al. [52] state that when the through-flow contains angular momentum, 𝐶𝐹 increases
with the increase of |𝐶𝑎𝑚| . 𝐶𝐹 shows little change so long as 𝐶𝑎𝑚 is a constant, even if 𝐶𝑞
varies. In Fig. 60, the impact of the pre-swirl on 𝐶𝐹 with different hub radius ratio (a/b) is depicted
[52]. It seems that the hub radius also plays an important role in calculating the axial thrust, which
deserves further investigation in the future work.
0
0.02
0.04
0.06
0.08
0.1
0.3 1.3 2.3 3.30
0.02
0.04
0.06
0.08
0.1
0.3 1.3 2.3 3.3
0
0.02
0.04
0.06
0.08
0.1
0.3 1.3 2.3 3.3
0
0.02
0.04
0.06
0.08
0.1
0.3 1.3 2.3 3.3
0
0.02
0.04
0.06
0.08
0.1
0.3 1.3 2.3 3.30
0.02
0.04
0.06
0.08
0.1
0.3 1.3 2.3 3.3
Re
(106)
Re
(106)
Re
(106)
Re
(106)
|𝐶𝐷| |𝐶𝐷| |𝐶𝐷|
Re
(106)
Re
(106)
|𝐶𝐷| |𝐶𝐷| |𝐶𝐷|
82
𝐶𝐹
Fig. 60: Comparison of the results of 𝑪𝑭 for 𝑮 = 𝟎. 𝟎𝟓 and 𝑪𝒂𝒎 𝑪𝒒⁄ = −𝟎. 𝟔𝟏𝟗 [52]
The experimental results of 𝐶𝐹 by axial thrust measurements are depicted versus 𝐶𝑎𝑚 in Fig. 61
for 𝐶𝐷 = −5050 (𝑘𝑠 = 0.4 𝜇m) at various Re and G. Due to the drop of 𝐾𝑝, 𝐶𝐹 decreases with
increasing 𝐶𝑎𝑚 at a fixed Re. For wider axial gap, the impact of 𝐶𝑎𝑚 on 𝐶𝐹 becomes lesser.
According to the experimental results, it is recognized that the axial thrust is greatly influenced by
the angular momentum at the guide vane entrance. The impact of 𝐶𝑎𝑚 on 𝐶𝐹 therefore should be
carefully considered, which is worthy of further investigation.
𝐶𝐹 G = 0.018 𝐶𝐹 G = 0.036
0
0.02
0.04
0.06
0.08
0.1
0.12
-0.006 -0.005 -0.004 -0.003 -0.002 -0.001 0
0.02
0.025
0.03
0.035
0.04
0.045
0.05
0.055
0.06
0.065
0.07
-0.002 -0.001 0 0.001 0.0020.02
0.025
0.03
0.035
0.04
0.045
0.05
0.055
0.06
0.065
0.07
-0.002 -0.001 0 0.001 0.002
𝐶𝑎𝑚 𝐶𝑎𝑚
𝛽=
0°
Co-rotating Counter-rotating
𝛽=
0°
𝐶𝑎𝑚
83
𝐶𝐹 G = 0.054 𝐶𝐹 G = 0.072
Colors of points: Red: 𝑅𝑒 = 0.76 × 106; Purple: 𝑅𝑒 = 1.14 × 106; Blue: 𝑅𝑒 = 1.9 × 106; Green: 𝑅𝑒 = 3.17 × 106.
Fig. 61: 𝑪𝑭 (𝑪𝒂𝒎) curves for 𝑪𝑫 = −𝟓𝟎𝟓𝟎 for various 𝑹𝒆 and G
5.1.5 3D Diagram with Centripetal Through-Flow
The typical tangential velocity profiles for regime III (merged disk boundary layer and wall
boundary layer) and regime IV (separated disk boundary layer and wall boundary layer) are given
in Fig. 4. According to the tangential velocity profiles by numerical simulations, the 2D
Daily&Nece diagram (see Fig. 5) is extended into 3D by distinguishing the tangential velocity
profiles at 𝑥 = 0.955, 𝑥 = 0.79 and 𝑥 = 0.57, depicted in Fig. 62. The scope of this thesis is
the following parameter ranges: |𝐶𝐷| ≤ 5050 , 0.38 × 106 ≤ 𝑅𝑒 ≤ 3.17 × 107 and 0.018 ≤
𝐺 ≤ 0.072. They are categorized into two regimes, namely regime III (below the distinguishing
lines) and regime IV (above the distinguishing lines). Near the distinguishing surface, there is a
mixing zone where regime III and regime IV coexist in the front chamber (the flow regimes at the
above three radial positions differ). In this thesis, it is not plotted in Fig. 62 for its small size.
Currently, five distinguishing lines are found for different |𝐶𝐷| , shown in Fig. 62 (b). The
distinguishing line at 𝐶𝐷 = 0 is almost equal to that from Daily and Nece [28]. The distinguishing
lines become steeper for larger |𝐶𝐷|. The approximate distinguishing surface is drawn through the
lines, shown in Fig. 62 (a). Below and above the surface respectively are regime III and regime IV.
0.02
0.025
0.03
0.035
0.04
0.045
0.05
0.055
0.06
0.065
0.07
-0.002 -0.001 0 0.001 0.0020.02
0.025
0.03
0.035
0.04
0.045
0.05
0.055
0.06
0.065
0.07
-0.002 -0.001 0 0.001 0.002
𝛽=
0°
𝐶𝑎𝑚
𝛽=
0°
𝐶𝑎𝑚
84
G G
Fig. 62: 3D diagram distinguishing regime III and regime IV with centripetal through-flow for
𝟎. 𝟑 × 𝟏𝟎𝟔 ≤ 𝑹𝒆 ≤ 𝟑. 𝟑 × 𝟏𝟎𝟔
By Kurokawa et al. [54], the surface roughness of the disk has almost no influence on the thickness
of the disk layer when 𝑘𝑠 ≤ 95 𝜇m . Hence, the impact of 𝑘𝑠 on the distinguishing lines is
neglected in this thesis.
5.1.6 Moment Coefficient
5.1.6.1 Impact of Through-Flow with a Smooth Disk
According to the experimental results from Han et al. [40], the moment coefficient on the cylinder
surface of the disk 𝐶𝑀𝑐𝑦𝑙 can be estimated with Eq. 116 for smooth disks. For 𝐶𝐷 = 0, the values
of 𝐶𝑀 for 𝐺 = 0.018 (regime III) and 𝐺 = 0.072 (regime IV) are compared in Fig. 63 (a) and
Fig. 63 (b), respectively. The differences between the experimental results and those from the
correlations by Daily and Nece [28] for both regime III and regime IV, are colossal. Hence, Eq. 117
and Eq. 118 are determined to satisfy the own experimental results (𝑘𝑠 = 0.4 μm for the disk).
𝐶𝑀𝑐𝑦𝑙 =
2 ∙ |𝑀𝑐𝑦𝑙|
𝜌 ∙ 𝛺2 ∙ 𝑏5=
0.084 ∙ 𝜋 ∙ 𝑡
𝑏 ∙ (𝑙𝑔𝛺 ∙ 𝑏2
𝜐 )1.5152
Eq. 116
Regime III: 𝐶𝑀3 = 0.011 ∙ 𝐺−16 ∙ 𝑅𝑒−
14 ∙ [𝑒(0.8∙10−4∙|𝐶𝐷|)] Eq. 117
Regime IV: 𝐶𝑀4 = 0.014 ∙ 𝐺1
10 ∙ 𝑅𝑒−15 ∙ [𝑒(0.46∙10−4∙|𝐶𝐷|)] Eq. 118
0.01
0.02
0.03
0.04
0.05
0.06
0.07
0.08
0.1 1.0 10.0
百万
𝐺 = 0.072
𝐺 = 0.054
𝐺 = 0.036
𝐺 = 0.018
Re (106)
85
𝐶𝑀 𝐶𝑀
(a) 𝐺 = 0.018, regime III (b) 𝐺 = 0.072, regime IV
Exp Eq. 117 Eq. 84 ( Daily and Nece) Eq. 118 Eq. 85 (Daily and Nece)
Fig. 63: Comparison of the results of 𝑪𝑴 for 𝑮 = 𝟎. 𝟎𝟏𝟖 and 𝑮 = 𝟎. 𝟎𝟕𝟐 at 𝑪𝑫 = 𝟎
To introduce the influence of 𝐶𝐷 on the moment coefficient, the results of 𝐶𝑀 from both the
torque measurements and above two correlations are plotted versus Re in Fig. 64. With increasing
Re, the flow regime may change from regime III to regime IV (see the distinguishing lines in Fig.
62). For 𝐺 = 0.018 and 𝐺 = 0.036, most of the flow is within regime III and the results are close
to those from Eq. 117, shown in Fig. 64 (a) and Fig. 64 (b). With increasing G, the flow regime shifts
from regime III to regime IV a given Re and |𝐶𝐷|. The regime changes can also be found according
to the experimental results of 𝐶𝑀. For example, in Fig. 64 (b) for G=0.036 and 𝐶𝐷 = 1262, the
results of 𝐶𝑀 from Eq. 118 rather than from Eq. 117 are approaching those from torque measurements
when 𝑅𝑒 ≥ 0.9 × 106. For 𝐺 = 0.054 and 𝐺 = 0.072, most of the flow regime is regime IV
and the results are close to those from Eq. 118, shown in Fig. 64 (c) and in Fig. 64 (d). The results
of 𝐶𝑀 from the equations are in good agreement with those by torque measurements. The values
of 𝐶𝑀 increase with increasing |𝐶𝐷|, while they decrease with the increase of Re. For large values
of Re, the impact of 𝐶𝐷 on 𝐶𝑀 becomes lesser.
𝐶𝑀 (a) 𝐺 = 0.018 𝐶𝑀 (b) 𝐺 = 0.036
0
0.001
0.002
0.003
0.004
0.3 1.3 2.3 3.3
0
0.001
0.002
0.003
0.004
0.3 1.3 2.3 3.3
百万
0.0004
0.0009
0.0014
0.3 1.3 2.3 3.3
百万
0.0004
0.0009
0.0014
0.3 1.3 2.3 3.3
Re (106)
|𝐶𝐷|
Re (106)
|𝐶𝐷|
Re (106) Re (106)
86
𝐶𝑀 (c) 𝐺 = 0.054 𝐶𝑀 (d) 𝐺 = 0.072
Solid points: Measurements; Dashed lines: Eq. 117; Solid lines: Eq. 118.
Colors: Black: |𝐶𝐷| = 0; Blue: |𝐶𝐷| = 1262; Yellow: |𝐶𝐷| = 2525; Green: |𝐶𝐷| = 3787; Red: |𝐶𝐷| = 5050.
Fig. 64: 𝑪𝑴 (Re) curves with centripetal through-flow
The results from Eq. 117 and Eq. 118 at the lines distinguishing regime III and regime IV should be
equal. The ratios of 𝐶𝑀3/𝐶𝑀4 at the distinguishing lines (see Fig. 62) are presented in Fig. 65. The
differences, attributed to the existence of the mixing zone, cover an amount less than ±4%. The
results indicate that the lines distinguishing regime III and regime IV are reasonable.
𝐶𝑀3/𝐶𝑀4
Colors: Black: |𝐶𝐷| = 0; Blue: |𝐶𝐷| = 1262; Yellow: |𝐶𝐷| = 2525; Green: |𝐶𝐷| = 3787; Red: |𝐶𝐷| = 5050.
Fig. 65: Results of 𝑪𝑴𝟑/𝑪𝑴𝟒 at the distinguishing lines for centripetal through-flow
5.1.6.2 Impact of Surface Roughness of the Disk
Combining the torque measurements with the results from Daily and Nece [28] and Dorfman
[32][33], two correlations can be determined to predict the influence of surface roughness on the
moment coefficient (for a single surface), given in Eq. 119 and Eq. 120.
0.0004
0.0009
0.0014
0.3 1.3 2.3 3.3
0.0004
0.0009
0.0014
0.3 1.3 2.3 3.3
0.96
0.97
0.98
0.99
1
1.01
1.02
1.03
1.04
0.1 1 10
Re (106) Re (106)
Re (106)
|𝐶𝐷| |𝐶𝐷|
87
Regime III: 𝐶𝑀3 = 0.32 ∙ 𝐺−
16 ∙ 𝑅𝑒−
14 ∙ [𝑒(0.82∙10−4∙|𝐶𝐷|)] ∙ (
𝑘𝑠
𝑏)
0.272
Eq. 119
Regime IV: 𝐶𝑀4 = 0.41 ∙ 𝐺
110 ∙ 𝑅𝑒−
15 ∙ [𝑒(0.46∙10−4∙|𝐶𝐷|)] ∙ (
𝑘𝑠
𝑏)
0.272
Eq. 120
The first step is to introduce the influence of surface roughness 𝑘𝑠 in an enclosed rotor-stator
cavity. The results from Eq. 119 and Eq. 120 (with the geometry from Kurokawa et al. [54]) are
compared with the analytical solutions from Kurokawa et al. [54] for a single surface, depicted in
Fig. 66. For 𝐺 = 0.115 and 𝑅𝑒 = 106, the flow regime is regime IV in Fig. 62. The results from
Eq. 120 show the same trend as those of analytical solutions by Kurokawa et al. [54][54]. For 𝐺 =
0.0255 and 𝑅𝑒 = 106 , the flow regime is regime III and the results from Eq. 119 are also in
acceptable agreement with the analytical solutions. More important, Eq. 119 and Eq. 120 cover a wider
parameter range of 𝑘𝑠.
Colors: Green: Analytical solutions by Kurokawa et al. [54]; Blue: Eq. 119 or Eq. 120 according to G.
Lines: Solid: G=0.115; Dashed: G=0.0255.
Fig. 66: Comparison of 𝑪𝑴 with different values of 𝒌𝒔 in an enclosed rotor-stator cavity
The second step is to compare the results from Eq. 119 and Eq. 120 with those from Eq. 87 by Owen
[65]. Since the surface roughness is not available in the work of Owen [68], only the results for
disk 3 (𝑘𝑠 = 26.5 𝜇m, which is considered reasonable) are compared in Fig. 67. In Fig. 62, the
flow regime is regime IV when 𝑅𝑒 ≥ 0.7 × 106 for |𝐶𝐷| = 2525, 𝐺 = 0.047 and when 𝑅𝑒 ≥
2.3 × 106 for |𝐶𝐷| = 5050 , 𝐺 = 0.047 . At 𝐺 = 0.012 , the flow regime is regime III for all
through-flow coefficients. The experimental results are in good agreement with those from Eq. 119
and Eq. 120 according to the flow regime, shown in Fig. 67. The trend of the results by Eq. 119 and
Eq. 120 is similar with those from Eq. 87. The difference can be attributed to the uncertainty of surface
roughness.
0
0.0005
0.001
0.0015
0.002
0.0025
0.003
0.0035
0.004
0 2000 4000 6000 8000 10000
𝑏
𝑘𝑠
𝐶𝑀
88
𝐶𝑀 |𝐶𝐷|=2525, G=0.047 𝐶𝑀 |𝐶𝐷|=5050, G=0.047
𝐶𝑀 |𝐶𝐷| =2525, G=0.012 𝐶𝑀 |𝐶𝐷| =5050, G=0.012
Triangles: Exp; Dot dash line: Eq. 87; Dashed line: Eq. 119; Solid line: Eq. 120.
Fig. 67: Comparison of 𝑪𝑴 from different equations
The third step is to compare the experimental results of 𝐶𝑀 for a rotor-stator cavity with centripetal
through-flow and rough disks with those from Eq. 119 and Eq. 120. For 𝐺 = 0.012 and 𝐺 = 0.027,
most of the flow regimes are regime III with increasing |𝐶𝐷| (see Fig. 62). The results of 𝐶𝑀 by
torque measurements for various 𝑘𝑠 are in very good consistent with those from Eq. 119, depicted
in Fig. 68. The values of 𝐶𝑀 increase as 𝑘𝑠 increases. The values of 𝐶𝑀 also increase with the
increase of |𝐶𝐷|. The values of 𝐶𝑀 drop faster with the increase of Re for smaller axial gap. When
G increases from 0.012 to 0.027 (right column of diagrams), the values of 𝐶𝑀 decrease, which is
the characteristic for regime III.
0.001
0.0015
0.002
0.0025
0.003
0.0035
0.004
0.0045
0.005
0.3 2.3 4.3 6.30.001
0.0015
0.002
0.0025
0.003
0.0035
0.004
0.0045
0.005
0.3 2.3 4.3 6.3
0.001
0.0015
0.002
0.0025
0.003
0.0035
0.004
0.0045
0.005
0.3 2.3 4.3 6.3
0.001
0.0015
0.002
0.0025
0.003
0.0035
0.004
0.0045
0.005
0.3 2.3 4.3 6.3
From regime III
to regime IV
From regime III
to regime IV
Regime III
Re
(106) Re
(106)
Re
(106)
Re
(106)
Regime III
89
𝐶𝑀 𝐺 = 0.012, 𝑘𝑠 = 0.4 𝜇m 𝐶𝑀 𝐺 = 0.027, 𝑘𝑠 = 0.4 𝜇m
𝐶𝑀 𝐺 = 0.012, 𝑘𝑠 = 3.5 𝜇m 𝐶𝑀 𝐺 = 0.027, 𝑘𝑠 = 3.5 𝜇m
𝐶𝑀 𝐺 = 0.012, 𝑘𝑠 = 11.4 𝜇m 𝐶𝑀 𝐺 = 0.027, 𝑘𝑠 = 11.4 𝜇m
0.0003
0.0008
0.0013
0.0018
0.30 1.30 2.30 3.30
百万
0.0003
0.0008
0.0013
0.0018
0.30 1.30 2.30 3.30
百万
0.0005
0.001
0.0015
0.002
0.0025
0.003
0.30 1.30 2.30 3.30
百万
0.0005
0.001
0.0015
0.002
0.0025
0.003
0.30 1.30 2.30 3.30
百万
0.001
0.0015
0.002
0.0025
0.003
0.0035
0.004
0.30 1.30 2.30 3.30
百万
0.001
0.0015
0.002
0.0025
0.003
0.0035
0.004
0.30 1.30 2.30 3.30
百万
Re
(106)
Disk 0
|𝐶𝐷|
Disk 0
Re
(106)
Disk 1
Re
(106)
Disk 2
Re
(106)
Disk 2
Re
(106)
Re
(106)
Disk 1
|𝐶𝐷| |𝐶𝐷|
|𝐶𝐷| |𝐶𝐷|
|𝐶𝐷|
90
𝐶𝑀 𝐺 = 0.012, 𝑘𝑠 = 26.5 𝜇m 𝐶𝑀 𝐺 = 0.027, 𝑘𝑠 = 26.5 𝜇m
𝐶𝑀 𝐺 = 0.012, 𝑘𝑠 = 58.9 𝜇m 𝐶𝑀 𝐺 = 0.027, 𝑘𝑠 = 58.9 𝜇m
𝐶𝑀 𝐺 = 0.012, 𝑘𝑠 = 81.3 𝜇m 𝐶𝑀 𝐺 = 0.027, 𝑘𝑠 = 81.3 𝜇m
Solid points: Measurements; Dashed lines: Eq. 119; Solid lines: Eq. 120.
Colors: Black: |𝐶𝐷| = 0; Blue: |𝐶𝐷| = 1262; Yellow: |𝐶𝐷| = 2525; Green: |𝐶𝐷| = 3787; Red: |𝐶𝐷| = 5050.
Fig. 68: 𝑪𝑴 (𝑹𝒆) curves at 𝑮 = 𝟎. 𝟎𝟏𝟐 and 𝑮 = 𝟎. 𝟎𝟐𝟕 for different values of 𝒌𝒔 and |𝑪𝑫|
0.001
0.0015
0.002
0.0025
0.003
0.0035
0.004
0.0045
0.005
0.30 1.30 2.30 3.30
百万
0.001
0.0015
0.002
0.0025
0.003
0.0035
0.004
0.0045
0.005
0.30 1.30 2.30 3.30
百万
0.001
0.0015
0.002
0.0025
0.003
0.0035
0.004
0.0045
0.005
0.0055
0.006
0.30 1.30 2.30 3.30
百万
0.001
0.0015
0.002
0.0025
0.003
0.0035
0.004
0.0045
0.005
0.0055
0.006
0.30 1.30 2.30 3.30
百万
0.002
0.0025
0.003
0.0035
0.004
0.0045
0.005
0.0055
0.006
0.0065
0.007
0.30 1.30 2.30 3.30
百万
0.002
0.0025
0.003
0.0035
0.004
0.0045
0.005
0.0055
0.006
0.0065
0.007
0.30 1.30 2.30 3.30
百万
Re
(106)
Disk 3
|𝐶𝐷|
Disk 4
Disk 5
Re
(106)
Disk 5
Re
(106)
Disk 4
Re
(106)
|𝐶𝐷|
|𝐶𝐷|
Re
(106)
Re
(106)
|𝐶𝐷|
|𝐶𝐷|
|𝐶𝐷|
Disk 3
91
For 𝐺 = 0.047 and 𝐺 = 0.065, the results of 𝐶𝑀 by torque measurements are also in very good
agreement with those from Eq. 119 and Eq. 120 according to the flow regimes in Fig. 62, depicted in
Fig. 69. The impact of 𝑘𝑠 and 𝐶𝐷 on 𝐶𝑀 is similar to those from 𝐺 = 0.012 and 𝐺 = 0.027.
The flow regime may change from regime III to regime IV with the increase of Re (see Fig. 62).
For example, when |𝐶𝐷| = 5050 and 𝐺 = 0.047, the experimental results of 𝐶𝑀 are closer to
those from Eq. 119 until 𝑅𝑒 ≈ 2.8 × 106 (in the 3D Daily&Nece diagram, the transition point is at
𝑅𝑒 ≈ 2.3 × 106 in Fig. 62). When |𝐶𝐷| = 5050 and 𝐺 = 0.065, the values of 𝐶𝑀 are closer to
those from Eq. 120 when 𝑅𝑒 ≥ 0.8 × 106, while in Fig. 62, the transition point is at 𝑅𝑒 ≈ 1 × 106.
Considering the existence of a mixing zone, where the regime III and regime IV coexist, the
difference is acceptable. With centripetal through-flow, the moment coefficient for rough disks can
be predicted with sufficient accuracy with Eq. 119 and Eq. 120 according to the 3D diagram in Fig.
62.
𝐶𝑀 𝐺 = 0.047, 𝑘𝑠 = 0.4 𝜇m 𝐶𝑀 𝐺 = 0.065, 𝑘𝑠 = 0.4 𝜇m
𝐶𝑀 𝐺 = 0.047, 𝑘𝑠 = 3.5 𝜇m 𝐶𝑀 𝐺 = 0.065, 𝑘𝑠 = 3.5 𝜇m
0.0004
0.0009
0.0014
0.3 1.3 2.3 3.3
百万
0.0004
0.0009
0.0014
0.3 1.3 2.3 3.3
百万
0.0005
0.001
0.0015
0.002
0.0025
0.3 1.3 2.3 3.3
百万
0.0005
0.001
0.0015
0.002
0.0025
0.3 1.3 2.3 3.3
百万
Re
(106)
Disk 0
|𝐶𝐷|
Re
(106)
Disk 0
Re
(106)
Disk 1
Re
(106)
Disk 1
|𝐶𝐷| |𝐶𝐷|
|𝐶𝐷|
92
𝐶𝑀 𝐺 = 0.047, 𝑘𝑠 = 11.4 𝜇m 𝐶𝑀 𝐺 = 0.065, 𝑘𝑠 = 11.4 𝜇m
𝐶𝑀 𝐺 = 0.047, 𝑘𝑠 = 26.5 𝜇m 𝐶𝑀 𝐺 = 0.065, 𝑘𝑠 = 26.5 𝜇m
𝐶𝑀 𝐺 = 0.047, 𝑘𝑠 = 58.9 𝜇m 𝐶𝑀 𝐺 = 0.065, 𝑘𝑠 = 58.9 𝜇m
0.001
0.0015
0.002
0.0025
0.003
0.3 1.3 2.3 3.3
百万
0.001
0.0015
0.002
0.0025
0.003
0.3 1.3 2.3 3.3
百万
0.0015
0.002
0.0025
0.003
0.0035
0.004
0.3 1.3 2.3 3.3
百万
0.0015
0.002
0.0025
0.003
0.0035
0.004
0.3 1.3 2.3 3.3
百万
0.002
0.0025
0.003
0.0035
0.004
0.0045
0.005
0.3 1.3 2.3 3.3
百万
0.002
0.0025
0.003
0.0035
0.004
0.0045
0.005
0.3 1.3 2.3 3.3
百万
Re
(106)
Disk 4
Re
(106)
Disk 4
Re
(106)
Disk 2
Re
(106)
Disk 3
Re
(106)
Disk 3
Re
(106)
Disk 2
|𝐶𝐷|
|𝐶𝐷| |𝐶𝐷|
|𝐶𝐷|
|𝐶𝐷| |𝐶𝐷|
93
𝐶𝑀 𝐺 = 0.047, 𝑘𝑠 = 81.3 𝜇m 𝐶𝑀 𝐺 = 0.065, 𝑘𝑠 = 81.3 𝜇m
Solid points: Measurements; Dashed lines: Eq. 119; Solid lines: Eq. 120.
Colors: Black: |𝐶𝐷| = 0; Blue: |𝐶𝐷| = 1262; Yellow: |𝐶𝐷| = 2525; Green: |𝐶𝐷| = 3787; Red: |𝐶𝐷| = 5050.
Fig. 69: 𝑪𝑴 (Re) curves at 𝑮 = 𝟎. 𝟎𝟒𝟕 and 𝑮 = 𝟎. 𝟎𝟔𝟓 for different values of 𝒌𝒔 and |𝑪𝑫|
Some more experimental results are presented in Fig. 70. All the results show that the impact of
𝑘𝑠 can be predicted with Eq. 119 and Eq. 120 with good accuracy when 𝑘𝑠 ≤ 81.3 𝜇m.
𝐺 = 0.018 𝐺 = 0.036 𝐺 = 0.072
𝑘𝑠
=0
.4 𝜇
m
𝐶𝑀 𝐶𝑀 𝐶𝑀
𝑘𝑠
=3
.5 𝜇
m
𝐶𝑀 𝐶𝑀 𝐶𝑀
0.002
0.0025
0.003
0.0035
0.004
0.0045
0.005
0.3 1.3 2.3 3.3
百万
0.002
0.0025
0.003
0.0035
0.004
0.0045
0.005
0.3 1.3 2.3 3.3
百万
0.0004
0.0009
0.0014
0.3 1.3 2.3 3.3
百万
0.0004
0.0009
0.0014
0.3 1.3 2.3 3.30.0004
0.0009
0.0014
0.3 1.3 2.3 3.3
0.0005
0.001
0.0015
0.002
0.0025
0.3 1.3 2.3 3.3
百万
0.0005
0.001
0.0015
0.002
0.0025
0.3 1.3 2.3 3.3
0.0005
0.001
0.0015
0.002
0.0025
0.3 1.3 2.3 3.3
Re
(106)
Disk 5
Re
(106)
Re
(106)
|𝐶𝐷|
Re
(106)
Re
(106)
Re
(106)
Re
(106)
Disk 5
|𝐶𝐷|
|𝐶𝐷|
Re
(106)
|𝐶𝐷| |𝐶𝐷|
|𝐶𝐷| |𝐶𝐷| |𝐶𝐷|
94
𝑘𝑠
=2
6.5
𝜇m
𝐶𝑀 𝐶𝑀 𝐶𝑀
𝑘𝑠
=5
8.9
𝜇m
𝐶𝑀 𝐶𝑀 𝐶𝑀
𝐶𝑀 𝐶𝑀 𝐶𝑀
𝑘𝑠
=8
1.3
𝜇m
Solid points: Measurements; Dashed lines: Eq. 119; Solid lines: Eq. 120.
Colors: Black: |𝐶𝐷| = 0; Blue: |𝐶𝐷| = 1262; Yellow: |𝐶𝐷| = 2525; Green: |𝐶𝐷| = 3787; Red: |𝐶𝐷| = 5050.
Fig. 70: 𝑪𝑴 (Re) curves at various G for different values of 𝒌𝒔 and 𝑪𝑫
0.001
0.0015
0.002
0.0025
0.003
0.0035
0.004
0.3 1.3 2.3 3.3
0.001
0.0015
0.002
0.0025
0.003
0.0035
0.004
0.3 1.3 2.3 3.3
0.001
0.0015
0.002
0.0025
0.003
0.0035
0.004
0.3 1.3 2.3 3.3
百万
0.0015
0.002
0.0025
0.003
0.0035
0.004
0.0045
0.005
0.0055
0.3 1.3 2.3 3.3
0.0015
0.002
0.0025
0.003
0.0035
0.004
0.0045
0.005
0.0055
0.3 1.3 2.3 3.3
0.0015
0.002
0.0025
0.003
0.0035
0.004
0.0045
0.005
0.0055
0.3 1.3 2.3 3.3
百万
0.0015
0.002
0.0025
0.003
0.0035
0.004
0.0045
0.005
0.0055
0.006
0.0065
0.3 1.3 2.3 3.3
0.0015
0.002
0.0025
0.003
0.0035
0.004
0.0045
0.005
0.0055
0.006
0.0065
0.3 1.3 2.3 3.3
0.0015
0.002
0.0025
0.003
0.0035
0.004
0.0045
0.005
0.0055
0.006
0.0065
0.3 1.3 2.3 3.3
百万
Re
(106)
Re
(106)
Re
(106)
Re
(106)
Re
(106)
Re
(106)
Re
(106)
|𝐶𝐷|
Re
(106)
Re
(106)
|𝐶𝐷|
|𝐶𝐷|
|𝐶𝐷|
|𝐶𝐷|
|𝐶𝐷|
|𝐶𝐷|
|𝐶𝐷|
|𝐶𝐷|
95
5.1.6.3 Impact of Pre-Swirl on the Moment Coefficient with a Smooth Disk
In Fig. 71, the 𝐶𝑀 (𝐶𝑎𝑚 ) curves for |𝐶𝐷| = 5050 varying the pre-swirl angle 𝛽 and Re are
plotted. The values of 𝐶𝑎𝑚 are calculated with Eq. 73 (𝛽 ranges from −60° to 60°, see Table 9).
As commonly understood, near the external diameter of the disk, the tangential velocity gradient
on the disk surface has enormous impact on 𝐶𝑀 (Kurokawa et al [53]). By torque measurements,
the pre-swirl leads to a decrease of 𝐶𝑀 when the fluid co-rotates with the disk at the entrance. For
the counter-rotating flow when entering the cavity (with a large tangential velocity gradient), 𝐶𝑀
increases significantly, especially for small Re. The values of 𝐶𝑀 are dominant by the angular
momentum at the inlet (𝑥 = 1.01) rather than G. It seems that the experimental results of 𝐶𝑀 vary
versus 𝐶𝑎𝑚 following an exponent law for the inlet boundary conditions.
𝐶𝑀 𝐺 = 0.018 𝐶𝑀 𝐺 = 0.036
𝐶𝑀 𝐺 = 0.054 𝐶𝑀 𝐺 = 0.072
Colors: Red: 𝑅𝑒 = 0.76 × 106; Purple: 𝑅𝑒 = 1.14 × 106; Blue: 𝑅𝑒 = 1.9 × 106; Green: 𝑅𝑒 = 3.17 × 106.
Fig. 71: 𝑪𝑴 (𝑪𝒂𝒎) curves for |𝑪𝑫| = 𝟓𝟎𝟓𝟎 at different 𝑹𝒆
0
0.002
0.004
0.006
0.008
0.01
0.012
-0.02 -0.01 0 0.01 0.02
0
0.002
0.004
0.006
0.008
0.01
0.012
-0.02 -0.01 0 0.01 0.02
0
0.002
0.004
0.006
0.008
0.01
0.012
-0.02 -0.01 0 0.01 0.020
0.002
0.004
0.006
0.008
0.01
0.012
-0.02 -0.01 0 0.01 0.02
𝐶𝑎𝑚
𝐶𝑎𝑚 𝐶𝑎𝑚
Counter-rotating Co-rotating
𝐶𝑎𝑚
96
5.2 Rotor-Stator Cavity with Centrifugal Through-Flow
5.2.1 Simulation Results of Velocity Distributions
The velocity profiles at three chosen radial positions for 𝑅𝑒 = 1.9 × 106 and 𝐺 = 0.072 (small
passages in the radial guide vane) are presented in Fig. 72. The non-dimensional radial velocities
are not exactly zero in the central cores (휁 = 0.5), shown in Fig. 72 (a to c). From the distribution
of tangential velocity, central cores are identified at all the investigated radial positions by a
constant value of 𝑉𝜑 along 휁, shown in Fig. 72 (d to f). The values of the tangential velocity
decline at 휁 = 0.5 when 𝐶𝐷 increases, depicted in Fig. 72 (d, e). The trends of the tangential
velocity are in good agreement with the measured tangential velocity in the literature (such as by
Poncet et al. [68] and by Debuchy et al. [30]). The values of |𝑉𝑧| become smaller towards the
shaft.
𝑥 = 0.955 𝑥 = 0.79 𝑥 = 0.57 𝑉𝑟 (a) 𝑉𝑟 (b) 𝑉𝑟 (c)
𝑉𝜑 (d) 𝑉𝜑 (e) 𝑉𝜑 (f)
𝑉𝑧 (g) 𝑉𝑧 (h) 𝑉𝑧 (i)
Colors: Black: 𝐶𝐷 = 0; Blue: 𝐶𝐷 = 1262; Green: 𝐶𝐷 = 3787; Red: 𝐶𝐷 = 5050.
Fig. 72: Velocity profiles for 𝑹𝒆 = 𝟏. 𝟗 × 𝟏𝟎𝟔 and 𝑮 = 𝟎. 𝟎𝟕𝟐
-0.1
-0.05
0
0.05
0.1
0 0.5 1
-0.1
-0.05
0
0.05
0.1
0 0.5 1
-0.1
-0.05
0
0.05
0.1
0 0.5 1
0
0.2
0.4
0.6
0.8
1
0 0.5 1
0
0.2
0.4
0.6
0.8
1
0 0.5 10
0.2
0.4
0.6
0.8
1
0 0.5 1
-0.03
-0.01
0.01
0.03
0 0.5 1
-0.03
-0.01
0.01
0.03
0 0.5 1
-0.03
-0.01
0.01
0.03
0 0.5 1
휁 휁 휁
Disk Wall
휁 휁 휁
휁 휁 휁
97
The velocity profiles at the three selected radial coordinates for 𝑅𝑒 = 1.9 × 106 and 𝐺 = 0.018
(small gap) are presented in Fig. 73. The values of 𝑉𝑟 vary along 휁, shown in Fig. 73 (a to c). The
values of 𝑉𝑟 increase with the increase of 𝐶𝐷. At 𝐶𝐷 = 3787 and 𝐶𝐷 = 5050, all the values of
𝑉𝑟 are positive (all the flow in the boundary layers is centrifugal). The flow type is therefore of the
Stewartson type. From the disk to the wall, 𝑉𝜑 decreases constantly, which is the characteristic of
the regime III, shown in Fig. 73 (d to f). The values of |𝑉𝑧| are very small, compared with those
in Fig. 72 (g~i). This indicates that the axial circulation of the fluid is weaker for a small axial gap
width.
𝑥 = 0.955 𝑥 = 0.79 𝑥 = 0.57
𝑉𝑟 (a) 𝑉𝑟 (b) 𝑉𝑟 (c)
𝑉𝜑 (d) 𝑉𝜑 (e) 𝑉𝜑 (f)
𝑉𝑧 (g) 𝑉𝑧 (h) 𝑉𝑧 (i)
Colors: Black: 𝐶𝐷 = 0; Blue: 𝐶𝐷 = 1262; Green: 𝐶𝐷 = 3787; Red: 𝐶𝐷 = 5050.
Fig. 73: Velocity profiles for 𝑹𝒆 = 𝟏. 𝟗 × 𝟏𝟎𝟔 and 𝑮 = 𝟎. 𝟎𝟏𝟖
-0.1
-0.05
0
0.05
0.1
0.15
0 0.5 1
-0.05
0
0.05
0.1
0 0.5 1
-0.1
-0.05
0
0.05
0.1
0 0.5 1
0
0.2
0.4
0.6
0.8
1
0 0.5 1
0
0.2
0.4
0.6
0.8
1
0 0.5 1
0
0.2
0.4
0.6
0.8
1
0 0.5 1
-0.01
-0.005
0
0.005
0.01
0 0.5 1
-0.01
-0.005
0
0.005
0.01
0 0.5 1
-0.01
-0.005
0
0.005
0.01
0 0.5 1
휁 휁 휁
Disk Wall
휁 휁 휁
휁 휁 휁
98
5.2.2 Core Swirl Ratio
5.2.2.1 Impact of Through-Flow with a Smooth Disk
In most of the radial pumps or turbines, the parameter G is a variable along the radius. A simplified
correlation is required with good accuracy over the whole range of G. By pressure measurements
and numerical simulations of core swirl ratio, Eq. 121 is correlated to predict the values of K when
G ranges from 0.018 to 0.072 for Stewartson type flow. The results are compared in Fig. 74. The
results from Eq. 110 and Eq. 121 are in good accordance with results of 𝐾 both from numerical
simulations and from the pressure measurements with Eq. 109.
Stewartson type flow (𝐶𝑞𝑟 ≥ 0.04):
𝐾 = 0.85 ∙ [0.032 + 0.32 × 𝑒(
−𝐶𝑞𝑟
0.028)] Eq. 121
K
Hollow points: Simulation; Solid points: By pressure measurements (Eq. 109); Dashed lines: Eq. 110 or Eq. 121.
Colors: Black: 𝐺 = 0.018; Green: 𝐺 = 0.036; Yellow: 𝐺 = 0.054; Red: 𝐺 = 0.072.
Fig. 74: 𝑲 (𝑪𝒒𝒓) curves
0
0.1
0.2
0.3
0.4
0.001 0.01 0.1
Batchelor/Couette type flow
Stewartson type flow
Co
nver
tin
g r
egio
n
𝐶𝑞𝑟
Eq. 121 Eq. 110
99
Some results of K, however, do not fit to the results from Eq. 110 and Eq. 121, especially at 𝑥 = 0.955
for wider axial gaps. A selection of the results is shown in Fig. 73. Near the outlet, an area change
from the front cavity to the channel in the guide vane for 𝐺 = 0.036, 0.054 and 0.072 occurs. The
measured pressure at 𝑥 = 0.955 is strongly influenced by the geometry at the outlet of the test rig.
Based on the simulation results (𝐺 = 0.072, 𝐶𝐷 = 5050), small vortexes near the outer radius of
the disk exist and therefore the measured values at 𝑥 = 0.955 are only partly used during the
calculation of K.
K
(a) Part of the results do not fit Eq. 110 and Eq. 122 (b) Surface streamlines near the outlet
Hollow points: Simulation; Solid points: By pressure measurements; Dashed lines: Eq. 110 or Eq. 121.
Colors: Black: 𝐺 = 0.018; Green: 𝐺 = 0.036; Yellow: 𝐺 = 0.054; Red: 𝐺 = 0.072.
Fig. 75: Large differences of K attributed to the geometry near the outlet
5.2.2.2 Impact of Surface Roughness of the Disk
By pressure measurements, Eq. 122 is determined to describe the impact of surface roughness on K
for Stewartson type flow.
Stewartson type flow (𝐶𝑞𝑟 ≥ 0.04):
𝐾 = 0.85 ∙ 𝑒
(600∙𝑘𝑠∙𝑟
𝑏2 )∙ [0.032 + 0.32 × 𝑒(
−𝐶𝑞𝑟
0.028)] Eq. 122
For Batchelor or Couette type flow, the impact of 𝑘𝑠 on K is the same as the case for centripetal
through-flow. The amounts of K by pressure measurements are compared with those from Eq. 122
for 𝐺 = 0.072, presented in Fig. 72. When 𝐶𝑞𝑟 ≤ 0.03, the experimental results of K are close to
0
0.1
0.2
0.3
0.4
0.5
0.001 0.01 0.1
𝐶𝑞𝑟
𝑥 = 0.955
Disk Wall
Guide vane
Eq. 110 Eq. 121 C
on
ver
tin
g r
egio
n
100
those from Eq. 111 for Batchelor or Couette type flow. For 𝐶𝑞𝑟 ≥ 0.04 , the results are in good
agreement with those from Eq. 122 for Stewartson type flow. For rougher surfaces, the K (𝐶𝑞𝑟) curves
become steeper. A faster increase of K can be spotted with the decrease of 𝐶𝑞𝑟 (with the increase
of radial coordinates) for rougher surfaces, which is in accordance with the trend in Kurokawa et
al. [54] for an enclosed rotor-stator cavity. In the transition zone, the values of K can be predicted
with both Eq. 111 and Eq. 122.
K
Solid points: By pressure measurements; Solid lines: Eq. 111; Dashed lines: Eq. 122.
Colors: Blue: 𝑘𝑠 = 0.4 𝜇m; Orange: 𝑘𝑠 = 11.4 𝜇m; Green: 𝑘𝑠 = 26.5 𝜇m; Red: 𝑘𝑠 = 58.9 𝜇m.
Fig. 76: 𝑲 (𝑪𝒒𝒓) curves for various 𝒌𝒔 at 𝑮 = 𝟎. 𝟎𝟕𝟐
5.2.2.3 Impact of Centrifugal Pre-Swirl Through-Flow with a Smooth Disk
Four inlet guide vanes exist in the horizontal pipe to generate the pre-swirl. When the flow passes
the guide vane, it will obtain a tangential velocity component. When introducing the through-flow,
Karabay et al. [49][50] mention that the values of 𝐾ℎ can be calculated with Eq. 123 at 𝑟 =𝑑ℎ
2
(with inlet pre-swirl nozzles, see Fig. 16, instead of swirlers).
𝐾ℎ =
𝑄
𝜋 ∙ 𝛺 ∙ (𝑑ℎ2 )3
∙ 𝑡𝑎𝑛(90° − 𝛽)
Eq. 123
0
0.1
0.2
0.3
0.4
0.5
0.6
0.01 0.02 0.03 0.04 0.05 0.06 0.07
Stewartson type flow
Bat
chel
or/
Couet
te
type
flo
w
Co
nver
tin
g r
egio
n
𝐶𝑞𝑟
101
Based on Eq. 62 from Karabay et al. [49][50], the values of K along the radius can be calculated
with Eq. 124.
𝐾𝑝 =
𝑄 ∙ 𝑥𝑎2 ∙ 𝑡𝑎𝑛(90° − 𝛽)
𝜋 ∙ 𝑥2 ∙ 𝛺 ∙ (𝑑ℎ2 )3
Eq. 124
The results of 𝐾𝑝 from numerical simulations for 𝐶𝐷 = 5050, 𝑅𝑒 = 1.9 × 106 and 𝐺 = 0.072
are depicted in Fig. 77. The simulation models include the geometry of the inlet swirler (see Fig.
35). The fluid domain near the swirler can be seen in Fig. 42. When compared with those from Eq.
124 by Karabay et al. [49][50], large differences can be observed. This indicates that Eq. 124 can not
be used to predict the value of 𝐾𝑝 with the boundary conditions used in this thesis. The values of
𝛽 are positive when the fluid co-rotates with the disk when entering the cavity. The results from
four pre-swirl angles (𝛽 = ±15° and 𝛽 = ±45°) are selected for comparison. Overall, for the
co-rotating flow, the amounts of 𝐾𝑝 will firstly decrease, then increase towards the outer radius of
the disk. For the counter-rotating flow, the values of K increase gradually by the disk from 𝑥 = 𝑥ℎ
to 𝑥 = 1. How to quantify the values of angular momentum introduced by the centrifugal through-
flow deserves further investigation.
Solid points: By pressure measurements; Solid lines: Numerical simulations; Dashed lines: Eq. 124.
Colors: Red: 𝛽 = 45°; Blue: 𝛽 = 15°; Green: no pre-swirl; Purple: 𝛽 = −15°; Orange: 𝛽 = −45°.
Fig. 77: 𝑲𝒑 (𝑪𝒒𝒓) curves for 𝑹𝒆 = 𝟏. 𝟗 × 𝟏𝟎𝟔, 𝑪𝑫 = 𝟓𝟎𝟓𝟎 and 𝑮 = 𝟎. 𝟎𝟕𝟐
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
-0.15 -0.1 -0.05 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35
𝑉𝑧
𝛽
𝑉𝜑 𝑄
𝑥
𝐾𝑝 𝑥 = 𝑥ℎ
Swirler at the inlet
102
5.2.3 Pressure Distribution
5.2.3.1 Impact of Through-Flow with a Smooth Disk
Like the case for centripetal through-flow, a reference pressure is taken at the non-dimensional
radial coordinate 𝑥 = 1 . The pressure values at 𝑥 = 1 are calculated by Eq. 112. The pressure
coefficient 𝐶𝑝 are positive values because the pressure drops towards the shaft. In Fig. 78, the
values of 𝐶𝑝 are plotted versus 𝐶𝐷. The experimental results show that 𝐶𝑝 decreases with the
increasing 𝐶𝐷, Re and G in general. The experimental results are in good agreement with those
from Eq. 112 based on the equations for K (Eq. 121 or Eq. 110 by distinguishing 𝐶𝑞𝑟 every 1 mm).
When 𝑅𝑒 = 2.79 × 106, the uncertainty of 𝐶𝑝 is 1.3 × 10−4, which is very small compared with
the measured results. Hence, it is neglected in Fig. 78 (d~f).
x x x
(a) 𝑅𝑒 = 0.76 × 106, 𝐺 = 0.072 (b) 𝑅𝑒 = 0.76 × 106, 𝐺 = 0.036 (c) 𝑅𝑒 = 0.76 × 106, 𝐺 = 0.018
x x x
(d) 𝑅𝑒 = 2.79 × 106, 𝐺 = 0.072 (e) 𝑅𝑒 =2.79× 106, 𝐺 = 0.036 (f) 𝑅𝑒 =2.79× 106, 𝐺 = 0.018
Solid points: By pressure measurements; Solid lines: Eq. 112.
Colors: Black: 𝐶𝐷 = 0; Blue: 𝐶𝐷 = 1262; Green: 𝐶𝐷 = 3787; Red: 𝐶𝐷 = 5050.
Fig. 78: Distribution of 𝑪𝒑 along the radius
0.4
0.6
0.8
1
0 0.01 0.02 0.03 0.04 0.05
0.4
0.6
0.8
1
0 0.01 0.02 0.03 0.04 0.050.4
0.6
0.8
1
0 0.01 0.02 0.03 0.04 0.05
0.4
0.6
0.8
1
0 0.01 0.02 0.03 0.04 0.05
0.4
0.6
0.8
1
0 0.01 0.02 0.03 0.04 0.05
0.4
0.6
0.8
1
0 0.01 0.02 0.03 0.04 0.05
𝐶𝑝 𝐶𝑝
𝐶𝑝
𝐶𝑝
𝐶𝑝
𝐶𝑝
𝐶𝐷
103
5.2.3.2 Impact of Surface Roughness of the Disk
The effect of surface roughness on the pressure distribution is shown in Fig. 79. The results of K
based on the pressure measurements are compared with those from Eq. 112 based on the equations
for K (Eq. 122 or Eq. 111 by distinguishing 𝐶𝑞𝑟 every 1 mm). The pressure drops faster towards the
shaft with the increase of 𝑘𝑠. With the increase of 𝐶𝐷, the values of 𝐶𝑝 decrease, which is caused
by the decrease of K (see Eq. 122 and Eq. 111).
𝐶𝐷 = 0 𝐶𝐷 = 5050
x x
𝑅𝑒
=0
.76
×1
06
x x
𝑅𝑒
=1
.9×
10
6
Solid points: By pressure measurements; Solid lines: Eq. 112.
Colors: Black: 𝑘𝑠 = 0.4 𝜇m; Blue: 𝑘𝑠 = 11.4 𝜇m; Green: 𝑘𝑠 = 26.5 𝜇m; Red: 𝑘𝑠 = 58.9 𝜇m.
Fig. 79: Distribution of 𝑪𝒑 along the radius for the rough disks
0.4
0.6
0.8
1
0 0.05 0.10.4
0.6
0.8
1
0 0.05 0.1
0.4
0.6
0.8
1
0 0.05 0.10.4
0.6
0.8
1
0 0.05 0.1
𝐶𝑝
𝐶𝑝
𝐶𝑝
𝐶𝑝
𝑘𝑠
104
5.2.3.3 Impact of Pre-Swirl on the Pressure Distribution with a Smooth Disk
To show the impact of centrifugal pre-swirl on the pressure distribution, the distributions of 𝐶𝑝
along x for 𝑅𝑒 = 0.76 × 106 and 𝑅𝑒 = 1.9 × 106 at G=0.072 with different 𝐶𝑎𝑚 at 𝐶𝐷 =
5050 are depicted in Fig. 80. With the increase of 𝛽, the pressure drops faster towards the shaft.
The values of 𝐶𝑝 are smaller for larger Re from both the simulation results and the pressure
measurements.
x 𝑅𝑒 = 0.76 × 106 x 𝑅𝑒 = 1.9 × 106
Solid Lines: Simulation results; Points: Pressure measurements;
Colors: Red: 𝛽 = 45°; Blue: 𝛽 = 15°; Green: no pre-swirl; Purple: 𝛽 = −15°; Orange: 𝛽 = −45°
Fig. 80: 𝑪𝒑 (x) curves for various 𝑪𝒂𝒎 at 𝑪𝑫 = 𝟓𝟎𝟓𝟎 and G=0.072
5.2.4 Thrust Coefficient
5.2.4.1 Impact of Centrifugal Through-Flow with a Smooth Disk
Based on the measurements, Hu et al. [18] correlate Eq. 113 for the thrust coefficient in a rotor-
stator cavity with centripetal through-flow ( 𝑘𝑠 = 0.4 𝜇m ). It is organized based on the
experimental results for centripetal through-flow. When compared with the experimental results, it
is similar to centrifugal through-flow, written in Eq. 125. In this thesis, 𝐶𝐷 is positive for centrifugal
through-flow. The comparisons of the results of 𝐶𝐹 for different G and 𝐶𝐷 are shown in Fig. 81.
The experimental results of 𝐶𝐹 are in good agreement with those based on the calculated pressure
along the disk with Eq. 115 and Eq. 125. The values of 𝐶𝐹 decrease with increasing 𝐶𝐷. In a rotor-
stator cavity with centripetal through-flow (𝐶𝐷 is negative) studied by Hu et al. [18], however, the
values of 𝐶𝐹 increase with increasing |𝐶𝐷|. The values of 𝐶𝐹 are smaller for a wider axial gap in
general.
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0 0.023 0.046 0.069 0.092
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0 0.011 0.022 0.033 0.044
𝐶𝑝 𝐶𝑝
105
𝐶𝐹 = [6.6 ∙ 10−3 ∙ ln(𝑅𝑒) − 0.113] ∙ 𝑒(−1.6∙10−4∙𝐶𝐷) ∙ [0.122 ∙ ln(𝐺) − 0.67] Eq. 125
𝐶𝐹 G=0.018 𝐶𝐹 G=0.036
𝐶𝐹 G=0.054 𝐶𝐹 G=0.072
Hollow points: Eq. 115; Solid points: Measurements; Dashed lines: Eq. 125.
Colors: Black: |𝐶𝐷| = 0; Blue: |𝐶𝐷| = 1262; Yellow: |𝐶𝐷| = 2525; Green: |𝐶𝐷| = 3787; Red: |𝐶𝐷| = 5050.
Fig. 81: 𝑪𝑭 (𝑹𝒆) curves (𝒌𝒔 = 𝟎. 𝟒 𝝁𝐦)
5.2.4.2 Impact of Surface Roughness of the Disk
For the rough disks, the impact of surface roughness on 𝐶𝐹 can be predicted with Eq. 126. A rougher
surface results in higher values of 𝐶𝐹 . This can be explained by the increase of 𝐶𝑝 . The
experimental results are also close to those based on the integral of pressure along the radius.
During the calculation of the pressure, the values of 𝐶𝑞𝑟 are calculated every 1 mm. Then the
values of K can be predicted based on the flow type. The results from Eq. 115 and Eq. 126 are in good
agreement with the experimental results, depicted in Fig. 82.
0
0.01
0.02
0.03
0.04
0.3 1.3 2.3 3.3
0
0.01
0.02
0.03
0.04
0.3 1.3 2.3 3.3
0
0.01
0.02
0.03
0.04
0.3 1.3 2.3 3.3
0
0.01
0.02
0.03
0.04
0.3 1.3 2.3 3.3
𝑅𝑒 (106) 𝑅𝑒 (106)
𝑅𝑒 (106) 𝑅𝑒 (106)
𝐶𝐷 𝐶𝐷
𝐶𝐷 𝐶𝐷
106
𝐶𝐹 = [6.6 ∙ 10−3 ∙ ln(𝑅𝑒) − 0.113] ∙ 𝑒(−1.6∙10−4∙𝐶𝐷) ∙ [0.122 ∙ ln(𝐺) − 0.67] ∙ 𝑒(880∙
𝑘𝑠𝑏
) Eq. 126
𝐺 = 0.018 𝐺 = 0.036 𝐺 = 0.072
𝑘𝑠
=3
.5 𝜇
m
𝐶𝐹 𝐶𝐹 𝐶𝐹
Hollow points: Eq. 115; Solid points: Measurements; Dashed lines: Eq. 126.
Colors: Black: |𝐶𝐷| = 0; Blue: |𝐶𝐷| = 1262; Yellow: |𝐶𝐷| = 2525; Green: |𝐶𝐷| = 3787; Red: |𝐶𝐷| = 5050.
Fig. 82: 𝑪𝑭 (𝑹𝒆) curves in dependence of 𝑪𝑫, 𝑮 and 𝒌𝒔
0
0.01
0.02
0.03
0.04
0.05
0.06
0.3 1.3 2.3 3.30
0.01
0.02
0.03
0.04
0.05
0.06
0.3 1.3 2.3 3.3
0
0.01
0.02
0.03
0.04
0.05
0.06
0.3 1.3 2.3 3.3
𝑘𝑠
=2
6.5
𝜇m
𝐶𝐹 𝐶𝐹 𝐶𝐹
𝑘𝑠
=5
8.9
𝜇m
𝐶𝐹 𝐶𝐹 𝐶𝐹
0
0.01
0.02
0.03
0.04
0.05
0.06
0.3 1.3 2.3 3.3
0
0.01
0.02
0.03
0.04
0.05
0.06
0.3 1.3 2.3 3.3
0
0.01
0.02
0.03
0.04
0.05
0.06
0.3 1.3 2.3 3.3
0
0.01
0.02
0.03
0.04
0.05
0.06
0.3 1.3 2.3 3.30
0.01
0.02
0.03
0.04
0.05
0.06
0.3 1.3 2.3 3.30
0.01
0.02
0.03
0.04
0.05
0.06
0.3 1.3 2.3 3.3
𝑅𝑒 (106)
𝑅𝑒 (106)
𝑅𝑒 (106)
𝑅𝑒 (106)
𝑅𝑒 (106)
𝑅𝑒 (106)
𝑅𝑒 (106)
𝑅𝑒 (106)
𝑅𝑒 (106)
𝐶𝐷
𝐶𝐷
𝐶𝐷 𝐶𝐷 𝐶𝐷
𝐶𝐷 𝐶𝐷
𝐶𝐷 𝐶𝐷
107
5.2.4.3 Impact of Pre-Swirl on the Thrust Coefficient with a Smooth Disk
The pre-swirl has a large impact on 𝐶𝐹 by changing the pressure distribution. The experimental
results of 𝐶𝐹 are compared for 𝐶𝐷 = 5050 and 𝑘𝑠 = 0.4 𝜇m in Fig. 83. With the increase of
𝛽, 𝐶𝐹 increases in general. The impact becomes lesser for wider axial gap width. The impact of
𝐶𝑎𝑚 on 𝐶𝐹 should be carefully considered, especially for small Re and small axial gaps. With the
increase of Re, the amounts of 𝐶𝐹 decline.
𝐶𝐹 𝐺 = 0.018 𝐶𝐹 𝐺 = 0.036
𝐶𝐹 𝐺 = 0.054 𝐶𝐹 𝐺 = 0.072
Points: Experiments.
Colors: Red: 𝛽 = 45°; Green: 𝛽 = 15°; Black: no pre-swirl; Yellow: 𝛽 = −15°; Blue: 𝛽 = −45°.
Fig. 83: Experimental results of 𝑪𝑭 versus 𝑪𝒂𝒎 for 𝑪𝑫 = 𝟓𝟎𝟓𝟎 at 𝐝𝐢𝐟𝐟𝐞𝐫𝐞𝐧𝐭 𝑹𝒆 and G
0
0.005
0.01
0.015
0.02
0.025
0.3 1.3 2.3 3.30
0.005
0.01
0.015
0.02
0.025
0.3 1.3 2.3 3.3
0
0.005
0.01
0.015
0.02
0.025
0.3 1.3 2.3 3.3
0
0.005
0.01
0.015
0.02
0.025
0.3 1.3 2.3 3.3
𝑅𝑒 (106)
𝑅𝑒 (106)
𝑅𝑒 (106)
𝑅𝑒 (106)
108
5.2.5 3D Diagram with Centrifugal Through-Flow
In this part, the 2D Daily&Nece diagram is extended with centrifugal through-flow by classifying
the tangential velocity profiles at x=0.945, x=0.79 and x=0.57 based on the results of numerical
simulations. Currently, five distinguishing lines can be found, depicted in Fig. 84 (a) (Hu et al.
[18]). Below and above the distinguishing lines are regime III and regime IV, respectively. The
distinguishing surface is drawn through the five distinguishing lines, shown in Fig. 84 (b). Near the
distinguishing surface, there is a mixing zone, where regime III and regime IV coexist in the cavity.
The distinguishing surface for centripetal through-flow (Hu et al. [16][17]) is also plotted to make
it a complete diagram.
G G
(a) Distinguishing line (b) Distinguishing surface
Fig. 84: 3D diagram distinguishing regime III and regime IV with centrifugal through-flow for
𝟎. 𝟑 × 𝟏𝟎𝟔 ≤ 𝑹𝒆 ≤ 𝟑. 𝟑 × 𝟏𝟎𝟔
5.2.6 Moment Coefficient
5.2.6.1 Impact of Through-Flow with a Smooth Disk
Comparing the torque measurements with the results from Daily and Nece [28] and Dorfman
[32][33], two correlations can be determined to predict the moment coefficient for centrifugal
through-flow (for a single surface of the disk, 𝑘𝑠 = 0.4 𝜇m), given in Eq. 127 and Eq. 128.
𝐶𝑀3 = 0.011 ∙ 𝐺−
16 ∙ 𝑅𝑒−
14 ∙ [𝑒(10−4∙𝐶𝐷)] Eq. 127
𝐶𝑀4 = 0.014 ∙ 𝐺
110 ∙ 𝑅𝑒−
15 ∙ [𝑒(0.6∙10−4∙𝐶𝐷)] Eq. 128
0
0.01
0.02
0.03
0.04
0.05
0.06
0.07
0.08
0.1 1 10 𝐶𝐷
Centripetal
through-flow
Centrifugal
through-flow
𝐶𝐷 = 0
𝐶𝐷 = 2525
𝐶𝐷 = 5050
𝐶𝐷 = 3787
𝐶𝐷 = 1262
𝑅𝑒 (106)
𝐺 = 0.072
𝐺 = 0.054
𝐺 = 0.036
𝐺 = 0.018
𝑅𝑒 (106)
109
The experimental results of 𝐶𝑀 are compared with those from Eq. 127 and Eq. 128, depicted in Fig.
85. For 𝐺 = 0.018 and 𝐺 = 0.036, most of the flow regimes are regime III, shown in Fig. 85 (a)
and Fig. 85 (b). When G increases to 0.036, the flow regimes change from regime III to regime IV
with the increase of Re for 𝐶𝐷 = 0, 1262 and 2525, which is also indicated by the experimental
results of 𝐶𝑀 . For 𝐺 = 0.054 and 𝐺 = 0.072 , most of the flow regimes are regime IV. The
results of 𝐶𝑀 from experiments are in very good agreement with those from Eq. 128 in general. The
regime III may occur at small Re and large 𝐶𝐷. The amounts of 𝐶𝑀 increase with the increase of
𝐶𝐷. The values of 𝐶𝑀 drop faster with the increase of Re for smaller values of G. Compared with
measured 𝐶𝑀 for centripetal through-flow (Hu et al. [18]), the centrifugal through-flow will result
in larger values of 𝐶𝑀 at the same values of |𝐶𝐷|, which is in accordance with the conclusion of
Dibelius et al. [31].
𝐶𝑀 G=0.018 𝐶𝑀 G=0.036
𝐶𝑀 G=0.054 𝐶𝑀 G=0.072
Solid points: Measurements; Dashed lines: Eq. 127; Solid lines: Eq. 128.
Colors: Black: |𝐶𝐷| = 0; Blue: |𝐶𝐷| = 1262; Yellow: |𝐶𝐷| = 2525; Green: |𝐶𝐷| = 3787; Red: |𝐶𝐷| = 5050.
Fig. 85: Curves for 𝑪𝑴 in dependence of 𝑹𝒆 for different values of 𝑪𝑫 and 𝑮 with centrifugal
through-flow
0
0.0005
0.001
0.0015
0.002
0.30 1.30 2.30 3.30
百万
0
0.0005
0.001
0.0015
0.002
0.30 1.30 2.30 3.30
百万
0
0.0005
0.001
0.0015
0.002
0.3 1.3 2.3 3.30
0.0005
0.001
0.0015
0.002
0.30 1.30 2.30 3.30
百万
𝑅𝑒 (106)
𝐶𝐷
𝑅𝑒 (106)
𝑅𝑒 (106)
𝑅𝑒 (106)
𝐶𝐷
𝐶𝐷 𝐶𝐷
110
On the distinguishing lines (see Fig. 84), the results from Eq. 127 should be equal to those from Eq.
128. The results of 𝐶𝑀3/𝐶𝑀4 for a non-dimensioned gap width G at the distinguishing lines are
presented in Fig. 86. The differences (may be attributed to the existence of the mixing zone) are
very small in general and cover an amount less than 5%.
𝐶𝑀3/𝐶𝑀4
𝐶𝐷 = 1262 𝐶𝐷 = 2525 𝐶𝐷 = 3787 𝐶𝐷 = 5050
Fig. 86: Results of 𝑪𝒎𝟑/𝑪𝒎𝟒 at the distinguishing lines for centrifugal through-flow
5.2.6.2 Impact of Surface Roughness of the disk
Comparing the torque measurements with the results from Daily and Nece [28] and Dorfman
[32][33], two correlations can be determined to predict the influence of surface roughness on the
moment coefficient (for a single surface), given in Eq. 129 and Eq. 130.
𝐶𝑀3 = 0.32 ∙ 𝐺−
16 ∙ 𝑅𝑒−
14 ∙ [𝑒(10−4∙𝐶𝐷)] ∙ (
𝑘𝑠
𝑏)
0.272
Eq. 129
𝐶𝑀4 = 0.41 ∙ 𝐺
110 ∙ 𝑅𝑒−
15 ∙ [𝑒(0.6∙10−4∙𝐶𝐷)] ∙ (
𝑘𝑠
𝑏)
0.272
Eq. 130
The results from Eq. 129 and Eq. 130 are compared with those from experiments in Fig. 87. For small
axial gaps (𝐺 = 0.018 and 𝐺 = 0.036), most of the experimental results of 𝐶𝑀 are in very good
agreement with those from Eq. 129. Regime IV may only occur at small 𝐶𝐷 and large Re for 𝐺 =
0.036 . For large axial gaps (𝐺 = 0.054 and 𝐺 = 0.072 ), the experimental results of 𝐶𝑀 are
close to those from Eq. 130. The experimental results of 𝐶𝑀 do not indicate the change of flow
regimes (see Fig. 84) with the increase of surface roughness, which is in accordance with the case
for centripetal through-flow. Kurokawa et al. [54] state that the impact of surface roughness of the
rotor on the thickness of the disk layer is weak in an enclosed rotor-stator cavity based on the
velocity measurements. Hence, the distinguishing lines are considered valid for the rotors when
𝑘𝑠 ≤ 81.3 𝜇m.
0.95
0.97
0.99
1.01
1.03
1.05
0.1 1 10
𝑅𝑒 (106)
111
𝐶𝑀 G=0.018 𝐶𝑀 G=0.036 𝑘
𝑠=
3.5
𝜇m
𝐶𝑀 G=0.054 𝐶𝑀 G=0.072
𝑘𝑠
=3
.5 𝜇
m
𝑘𝑠
=1
1.4
𝜇m
𝐶𝑀 G=0.018 𝐶𝑀 G=0.036
𝑘𝑠
=1
1.4
𝜇m
𝐶𝑀 G=0.054 𝐶𝑀 G=0.072
0.0003
0.0008
0.0013
0.0018
0.0023
0.0028
0.30 1.30 2.30 3.30
百万
0.0003
0.0008
0.0013
0.0018
0.0023
0.0028
0.30 1.30 2.30 3.30
百万
0.0003
0.0008
0.0013
0.0018
0.0023
0.0028
0.3 1.3 2.3 3.30.0003
0.0008
0.0013
0.0018
0.0023
0.0028
0.30 1.30 2.30 3.30
百万
0.0003
0.0008
0.0013
0.0018
0.0023
0.0028
0.0033
0.0038
0.30 1.30 2.30 3.30
百万
0.0003
0.0008
0.0013
0.0018
0.0023
0.0028
0.0033
0.0038
0.30 1.30 2.30 3.30
百万
0.0003
0.0008
0.0013
0.0018
0.0023
0.0028
0.0033
0.0038
0.3 1.3 2.3 3.3
0.0003
0.0008
0.0013
0.0018
0.0023
0.0028
0.0033
0.0038
0.30 1.30 2.30 3.30
百万
Re
(106)
Re
(106)
Re
(106)
Re
(106)
Re
(106)
Re
(106)
𝐶𝐷
Re
(106)
Re
(106)
𝐶𝐷
𝐶𝐷
𝐶𝐷 𝐶𝐷
𝐶𝐷
𝐶𝐷 𝐶𝐷
112
𝐶𝑀 G=0.018 𝐶𝑀 G=0.036 𝑘
𝑠=
26
.5 𝜇
m
𝐶𝑀 G=0.054 𝐶𝑀 G=0.072
𝑘𝑠
=2
6.5
𝜇m
𝑘𝑠
=5
8.9
𝜇m
𝐶𝑀 G=0.018 𝐶𝑀 G=0.036
𝑘𝑠
=5
8.9
𝜇m
𝐶𝑀 G=0.054 𝐶𝑀 G=0.072
0.0003
0.0008
0.0013
0.0018
0.0023
0.0028
0.0033
0.0038
0.0043
0.30 1.30 2.30 3.30
百万
0.0003
0.0008
0.0013
0.0018
0.0023
0.0028
0.0033
0.0038
0.0043
0.30 1.30 2.30 3.30
百万
0.0003
0.0008
0.0013
0.0018
0.0023
0.0028
0.0033
0.0038
0.0043
0.3 1.3 2.3 3.30.0003
0.0008
0.0013
0.0018
0.0023
0.0028
0.0033
0.0038
0.0043
0.30 1.30 2.30 3.30
百万
0.00030.00080.00130.00180.00230.00280.00330.00380.00430.00480.00530.0058
0.30 1.30 2.30 3.30
百万
0.0003
0.0008
0.0013
0.0018
0.0023
0.0028
0.0033
0.0038
0.0043
0.0048
0.0053
0.0058
0.30 1.30 2.30 3.30
百万
0.00030.00080.00130.00180.00230.00280.00330.00380.00430.00480.00530.0058
0.3 1.3 2.3 3.3
0.00030.00080.00130.00180.00230.00280.00330.00380.00430.00480.00530.0058
0.30 1.30 2.30 3.30
百万
Re
(106)
Re
(106)
Re
(106)
Re
(106)
Re
(106) Re
(106)
Re
(106)
Re
(106)
𝐶𝐷 𝐶𝐷
𝐶𝐷 𝐶𝐷
𝐶𝐷 𝐶𝐷
𝐶𝐷 𝐶𝐷
113
𝑘𝑠
=8
1.3
𝜇m
𝐶𝑀 G=0.018 𝐶𝑀 G=0.036
𝑘𝑠
=8
1.3
𝜇m
𝐶𝑀 G=0.054 𝐶𝑀 G=0.072
Solid points: Measurements; Dashed lines: Eq. 129; Solid lines: Eq. 130.
Colors: Black: |𝐶𝐷| = 0; Blue: |𝐶𝐷| = 1262; Yellow: |𝐶𝐷| = 2525; Green: |𝐶𝐷| = 3787; Red: |𝐶𝐷| = 5050.
Fig. 87: Curves for 𝑪𝑴 in dependence of 𝑹𝒆 for different values of 𝑪𝑫 and 𝑮 with centrifugal
through-flow
5.2.6.3 Impact of Pre-Swirl on the Moment Coefficient with a Smooth Disk
The results of 𝐶𝑀 for centrifugal pre-swirl through-flow are depicted in Fig. 88 for 𝐶𝐷 = 5050
and 𝑘𝑠 = 0.4 𝜇m. The pre-swirl reduces the tangential velocity gradient in the disk layer with the
increase of 𝛽. Then, the wall shear stress on the disk surface will decrease correspondingly. 𝐶𝑀
therefore decreases with increasing 𝛽 (inlet angular momentum) for a set of given values of G and
Re. These trends are in accordance with theoretical analyses from Kurokawa et al. [52]. Currently,
the angular momentum at the inlet is not quantified, which deserves further investigations.
0.00030.00080.00130.00180.00230.00280.00330.00380.00430.00480.00530.00580.0063
0.30 1.30 2.30 3.30
百万
0.00030.00080.00130.00180.00230.00280.00330.00380.00430.00480.00530.00580.0063
0.30 1.30 2.30 3.30
百万
0.00030.00080.00130.00180.00230.00280.00330.00380.00430.00480.00530.00580.0063
0.3 1.3 2.3 3.30.00030.00080.00130.00180.00230.00280.00330.00380.00430.00480.00530.00580.0063
0.30 1.30 2.30 3.30
百万
Re
(106)
Re
(106)
Re
(106)
Re
(106)
𝐶𝐷 𝐶𝐷
𝐶𝐷 𝐶𝐷
114
𝐶𝑀 G=0.018 𝐶𝑀 G=0.036
𝐶𝑀 G=0.054 𝐶𝑀 G=0.072
Colors: Red: 𝛽 = 45°; Green: 𝛽 = 15°; Black: no pre-swirl; Yellow: 𝛽 = −15°; Blue: 𝛽 = −45°.
Fig. 88: Experimental results of 𝑪𝑴 𝐯𝐞𝐫𝐬𝐮𝐬 𝑹𝒆 for 𝒌𝒔 = 𝟎. 𝟒 𝝁𝐦 and 𝑪𝑫 = 𝟓𝟎𝟓𝟎
0.0005
0.001
0.0015
0.002
0.3 1.3 2.3 3.3
0.0005
0.001
0.0015
0.002
0.3 1.3 2.3 3.3
0.0005
0.001
0.0015
0.002
0.3 1.3 2.3 3.3
0.0005
0.001
0.0015
0.002
0.3 1.3 2.3 3.3
Re
(106)
𝛽
𝛽 𝛽
𝛽
Re
(106)
Re
(106)
Re
(106)
115
6. Applications of the Results in Radial Pumps
All the results presented in the previous chapters are obtained with the rotor-stator cavity model. It
is very important to apply the results to the design of the turbomachines. In this chapter, two
examples are given to show the applications of the results.
6.1 Flow in the Rear Chamber of a Submersible Multi-Stage Slurry Pump (SMSP)
6.1.1 Sand Discharge Groove Design
The complicated solid-liquid flow in a grooved rotor-stator cavity is strongly influenced by both
the axial and the radial gaps between the rotor and the wall. This part of the thesis presents the
results from numerical simulation on the influences of above two gaps on the solid-liquid flow
inside the grooved rear chamber, aiming to accomplish effective micro-sized sand exclusion. The
leakage flow in the rear chamber of a SMSP is selected as the research model, shown in Fig. 89 (a).
The flow pattern in the rear chamber of a SMSP is shown in Fig. 89 (b).
① : Shaft; ②: Shaft shoulder; ③:Disk; ④: Casing; ⑤: Wall layer; ⑥: Core; ⑦: Disk layer
Fig. 89: Leakage flow and flow pattern inside the rear chamber of a SMSP
Rear chamber
𝑧
Casing
∆r
Broken O-ring
Q
(a)
① ②
③
④
⑤ ⑥
⑦
(b)
0.8
b
b
Rotor
116
Since the micro-sized sand moves with the conveying fluid (Hu et al. [14][15]), the flow patterns
of solid-liquid flow with micro-sized sand may be similar to that of pure water. Due to mass
conservation, an axial convection of the fluid from the outer radius of the disk to that of the stator
takes place. A sand discharge groove, shown in Fig. 90 (a), is designed based on that axial
convection. The groove includes two parts: an annular groove at the outer radius of the stator and
a spiral groove on the inner surface of the stator. The flow in the groove is shown in Fig. 90 (b).
Since the sand moves from the outer radius of the disk to that of the stator, it may enter the annular
groove before it moves towards the disk. Due to the dominant tangential motion, the sand will be
discharged out of the rear chamber along the spiral groove. In this thesis, the ANSYS CFX 14.0
code is used to predict the solid-liquid flow in 6 grooved rear chambers. This part focuses on
selecting the optimum dimensions of both the axial and the radial gaps for accomplishing sand
exclusion.
(a) Design of the sand discharge groove (b) Flow in the sand discharge groove
Fig. 90: Sand discharge groove
6.1.2 Geometrical Parameters
The main geometries of the impeller are given in Table 16.
𝛽2 (o) 𝐷2 (mm) 𝐷1 (mm) 𝑏2 (mm) 휃1 (o) n (rpm) 𝑛𝑠
23 220 80 10 130 2900 690
Table 16: Geometrical parameters of the impeller
Four non-dimensional parameters, namely G, ∆x (non-dimensional radial gap width), Re and 𝐸𝑘
(Ekman number), are used to distinguish the different flow patterns. When 𝐺 << 1, 𝑅𝑒 >> 1
and 𝐸𝑘 << 1, the turbulent sheared flow is considered with dominant inertia effects. Sand is more
likely to move towards the shaft before entering the groove when the axial gap is too wide. At the
same time, the periphery flow will dramatically influence the cavity flow unless the radial gap
Flow in spiral groove
Annular groove
Spiral groove
Flow in annular groove
117
decreases to a certain value (Haudt et al. [35]). Hence, G and Δ𝑥 should be decreased. 6 cavities
are designed based on the above principle, noted from No.1 to No.6. The first three chambers are
designed to find the influence of the radial gap on the flow in the rear chambers. The radial gap of
the last three rear chambers will be designed based on that value, noted as 𝑐𝑠. The amounts of G
for No. 4 to No. 6 are minimized to increase the amount of sand which is transported into the groove.
According to the parameters of the 6 cavities, given in Table 17, the flow regimes in all the cavities
are flow regime IV. In this thesis, the diameter and the lead of the spiral groove are designed as 5
mm and 85 mm, respectively. The diameter of the annular groove is 5 mm.
Cavities Δx 𝐺 Re 𝐸𝑘
No.1 0.4×10-1
3.9×10-1
0.87×106 to 1.2×106
5.5×10-6 No.2 0.3×10-1
No.3 0.2×10-1
No.4
𝑐𝑠
3.1×10-1 8.6×10-6
No.5 2.3×10-1 1.5×10-5
No.6 1.6×10-1 3.4×10-5
Table 17: Parameters of the 6 rotor-stator cavities
6.1.3 Numerical Simulation
The analysis type is set as steady state. All the simulations are carried out using the Multiple Frames
of reference method. The value of inlet pressure is set according to the average pressure at the
entrance of the pump last stage from numerical simulation. The 3D numerical simulation is
conducted with the Ansys CFX 14.0 code. The simulation model includes five domains: inlet,
impeller, rotor-stator cavity, leakage gap and volute, shown in Fig. 91. The meshes of the rear
chamber with the spiral groove are generated using hybrid grids. All other domains are meshed
with structured grids. Since the volute has two outlets, the flows through the two outlet pipes are
considered as an equal value. The standard 𝑘 − 휀 model is used to predict the turbulent flow.
Schiller Naumann model is used to calculate the drag coefficient. Instead of resolving the boundary
layers, the scalable wall functions are used to model the viscous effects in the near wall regions.
The temperature of the water is set to 20℃. Based on the design requirements, the value of the
mass flow rate at the inlet, the pressure at the outlet and the solid volume fraction are set as 5.4
kg/s, 2.6 MPa and 10%, respectively. A mesh independent analysis is conducted to minimize the
error associated with a poor quality mesh [14][15]. The values of 𝑦+ range from 30.6 to 78, which
indicates that the scalable wall functions are valid.
118
Fig. 91: Domains and meshes of the simulation model
6.1.4 Results and Discussion
The contours of the non-dimensional radial velocity are shown in Fig. 92. The positive and the
negative values of 𝑉𝑟 indicate radial outward flow (centrifugal through-flow) and radial inward
flow (centripetal through-flow), respectively. Both the disk boundary layer (where 𝑉𝑟 ≥ 0.03) and
the wall boundary layer (where 𝑉𝑟 ≤ −0.03) can be predicted inside of the three cavities. The
values of 𝑉𝑟 in most of the cavities are close to zero, which indicates that there is a core in each
cavity. The simulation results show that when the non-dimensional radial gap ∆𝑥 decreases to
0.03, a small vortex near the shaft shoulder occurs. The vortex almost remains the same when the
non-dimensional radial gap ∆𝑥 decreases further to 0.02. It indicates that the periphery flow has
little influence on the cavity flow for ∆𝑥 ≤ 0.03. Considering the shaft deformation, ∆𝑥 = 0.02
is not selected as the optimal value. Hence, the value of 𝑐𝑠 for No.4, No.5 and No.6 is set as 0.03.
① : 𝑉𝑟 ≥ 0.03; ② : −0.03 < 𝑉𝑟 <0.03; ③ : 𝑉𝑟 ≤ −0.03
Fig. 92: Distribution of the radial velocity
The distribution of streamlines on the meridian plane is shown in Fig. 93. The movement of sand
is shown by the direction of the arrows. The flow near the spiral groove is turbulent. As shown in
Impeller
Volute with two outlets
Inlet
Cavity with sand discharge groove
Leakage gap
∆𝑥 ∆𝑥 ∆𝑥
119
Fig. 93 (a), most of the sand rotates in the chamber and moves towards the shaft before entering
the groove. Only a small portion of sand flows into the groove. From Fig. 93 (a to c),it seems that
the amount of sand entering the groove increases with decreasing axial gap width. The cavity flow,
which is obviously not influenced by the spiral groove and the shoulder, is similar to Batchelor
type flow. In Fig. 93 (d), some particles move out of the groove towards the shaft.
(a) No. 2 (b) No. 3
(c) No. 5 (d) No. 6
Fig. 93: Movement of particles on the meridian plane
The distribution of the non-dimensional radial velocity 𝑉𝑟 is shown in Fig. 94 (a to d). The axial
distribution of the non-dimensional radial velocity reveals that there is a central core in each cavity
where the value is close to zero. Near both the disk and the wall, the values of the non-dimensional
radial velocity drop obviously with increasing radial coordinate after reaching their peaks.
Batchelor type flows can be predicted in all of the rear chambers. According to Fig. 94 (d), the
negative values of the non-dimensional radial velocity indicate that the flow moves towards the
shaft. One reason may be that the influence of the spiral groove and the shoulder on the fluid is
strong at that axial gap width.
120
r/(r+Δr) No.2 r/(r+Δr) No. 4
(a) (b)
r/(r+Δr) No.5 r/(r+Δr) No. 6
(c) (d)
𝑧 = 0.2𝑠
𝑧 = 0.4𝑠
𝑧 = 0.6𝑠
𝑧 = 0.8𝑠
𝑧 = 0.9𝑠
𝑧 = 0.2𝑠: close to the disk; 𝑧 = 0.9𝑠: Close to the wall.
Fig. 94: Distribution of non-dimensional radial velocity
The simulation results of the non-dimensional tangential velocity are depicted in Fig. 95. The non-
dimensional tangential velocity increases with decreasing axial gap width. There is a balance
between the centrifugal force and the pressure gradient. Due to the no-slip wall condition, the non-
dimensional tangential velocities on the surface of the rotor and the stator are considered as 0.8
(the radius of the sealing is 0.8𝑏, see Fig. 89) and 0, respectively. With the increase of the non-
dimensional tangential velocity, the sand is likely to move away from the shaft. The simulation
results of the tangential velocity indicate that the axial gap should be minimized.
0.8
0.85
0.9
0.95
1
-0.1 0 0.1
0.8
0.85
0.9
0.95
1
-0.1 0 0.1
0.8
0.85
0.9
0.95
1
-0.1 0 0.10.8
0.85
0.9
0.95
1
-0.1 0 0.1
𝑉𝑟 𝑉𝑟
𝑉𝑟 𝑉𝑟
121
r/(r+Δr) No.2 r/(r+Δr) No. 4
(a) (b)
r/(r+Δr) No.5 r/(r+Δr) No. 6
(c) (d)
𝑧 = 0.2𝑠
𝑧 = 0.4𝑠
𝑧 = 0.6𝑠
𝑧 = 0.8𝑠
𝑧= 0.9𝑠
𝑧 = 0.2𝑠: close to the disk; 𝑧 = 0.9𝑠: Close to the wall.
Fig. 95: Distribution of non-dimensional tangential velocity
The distribution of the solid volume fraction is shown in Fig. 96. The highest solid volume fractions
are shown in the spiral groove. A relative low solid volume fraction is predicted for case No.5. The
value of particle concentration in case No.6 is higher than that in case No.5 in general. Although
the tangential velocity is higher in case No.6, the concentration of sand in the groove is less than
that of case No.4 and case No.5. These results can be explained by analyzing the radial velocity
distributions in Fig. 94 (d). Near the outer radius of the disk, the flow moves towards the shaft
before it enters the groove. It seems that the geometries of the spiral groove and the shoulder have
a negative effect on the sand exclusion in case No. 6.
0.8
0.85
0.9
0.95
1
0.3 0.4 0.5 0.6 0.7 0.8
0.8
0.85
0.9
0.95
1
0.3 0.4 0.5 0.6 0.7 0.8
0.8
0.85
0.9
0.95
1
0.3 0.4 0.5 0.6 0.7 0.8
0.8
0.85
0.9
0.95
1
0.3 0.4 0.5 0.6 0.7 0.8
𝑉𝜑 𝑉𝜑
𝑉𝜑 𝑉𝜑
122
①: 𝜑𝑠 > 2%; ②: 1% ≤ 𝜑𝑠 < 2%; ③: 0.5%≤𝜑𝑠 < 1%; ④: 0.2% ≤ 𝜑𝑠 < 0.5%; ⑤: 𝜑𝑠 < 0.2%
Fig. 96: Distribution of solid volume fraction
In order to further analyze the accuracy of the results from the numerical simulations, the SMSP
with cavity No.5 is manufactured and tested. The test rig precedes the requirements of international
B-grade precision (ISO9906-1999). The uncertainties of the results are given in Table. 18. The
density and the kinematic viscosity of the working fluid are 1160 kg/m3 and 1.06 m2/s, respectively
at 20℃. The test rig has a circulating system at the bottom, which can reduce the settlement of sand.
The size of the sand is selected with screens. During the test, the concentration of the sand is
measured every 48 hours. An appropriate amount of sand is added to the slurry when the sand
concentration decreases more than 20%. The measured pump performance is given in Fig. 97.
Pressure Flow rate n Electric Voltage Electric current Power η
0.1% 0.2% 0.2% 0.1% 0.1% 0.14% 0.26%
Table 18: Uncertainties of results
0 5 10 15 20 25 30 35 400
50
100
150
200
250
300
350
400
450
Po
wer(
kW
)E
ffic
ien
cy
(%)
H
H (
m)
Q (m3/h)
0
10
20
30
40
50
ا¦
30
40
50
60
70
P
Fig. 97: Pump performance from experiments
① ① ① ① ② ②
②
②
③ ③
③
③
④ ④
④
④ ⑤ ⑤
⑤
⑤
H
η
P
No.2 No.4 No.5 No.6
Flow (m3/h)
Pre
ssu
re h
ead
(m
)
123
An abrasion test is also accomplished to verify the correctness of sand exclusion. The outlet
pressure is monitored with a pressure sensor during the experiments. When the mechanical seal is
broken, a high pressure difference contributes to a large leakage flow from the rear chamber to the
environment. It results in the plunge of the pressure head, shown in Fig. 98. The parameter 𝑝𝑟 is
defined in Eq. 131. The mechanical seal is considered broken when the values of 𝑝𝑟 are below 0.9.
According to the experimental results, the service life of the SMSP is significantly improved by
around 30% with cavity No. 5.
𝑝𝑟 =
𝑃𝑥
𝑃0
Eq. 131
Fig. 98: Pressure drop during the abrasion test
6.2 The Axial Thrust in a Deep Well Pump
6.2.1 Main Geometric Parameters
The bearing systems of pumps are fragile, especially when the large leakage flow occurs. Shi et al.
[84] describe an investigation of the axial thrust of a deep well pump based on both the numerical
simulation and the axial thrust measurements. The main geometric parameters of the pump are
listed in Table 19.
Geometric parameters Values Geometric parameters Values
Blade number 7 Impeller diameter at the back cover 132 mm
Blade inlet angle 30° Impeller inlet diameter 𝐷1 62 mm
Blade outlet angle 𝛽2 35° Impeller outlet width 𝑏2 12 mm
Wrapping angle 휃1 115° Impeller hub diameter 38 mm
Impeller diameter at the front cover 148 mm Shaft diameter 28 mm
Table 19: Main geometric parameters of the pump
0.6
0.7
0.8
0.9
1
0 200 400 600 800 1000
1
2
Prototype
Cavity No.5
𝑡ℎ (hour)
𝑝𝑟
Broken mechanical seal
124
The 3D model of the impeller and the guide vane are depicted in Fig. 99 (a) and Fig. 99 (b),
respectively [84].
Fig. 99: Geometry of the impeller (left) and the guide vane (right)
6.2.2 Simulation Set-Up
The fluid domains are modeled in Creo 2.0. To predict the inner flow, the axial thrust and the
frictional torque, numerical simulations are carried out using the ANSYS CFX 14.0 code by Shi et
al. [84]. The CFX set-up by Shi et al. [84] is described in this part. The simulation type is set as
steady state. The standard 𝑘 − 휀 turbulence model in combination with the scalable wall functions
is selected. The chosen boundary conditions are total pressure at the inlet and mass flow at the
outlet, respectively. The convergence criteria for all the numerical simulations are set as 10−4 in
maximum type. To minimize the error associated with the quality of the mesh, a mesh sensitivity
analysis is accomplished by comparing the simulation results of the axial thrust.
6.2.3 Results for Axial Thrust Coefficient
The pressure acting on the impeller of a centrifugal pump is shown in Fig. 100 (a). Based on Eq.
124, the thrust coefficient at surface 1 and surface 2 can be calculated when the leakage flow is
estimated (centripetal through-flow). The maximum difference between the simulation results [84]
and the measurements [84] of the axial thrust is 5.9%. The values of 𝐹𝑎𝑏 − 𝐹𝑎𝑓 are plotted versus
𝐶𝐷 in Fig. 100 (b). The experimental results of 𝐹𝑎𝑏 − 𝐹𝑎𝑓 are obtained by subtracting the sum of
the impulse force at the impeller eye 𝐹𝑎𝑠 and the force on the impeller passage 𝐹𝑎𝑝 (from
numerical simulations [84]). The trend of the results from Eq. 124 is in better agreement with the
125
experimental results than that from the equation by Kurokawa et al. [53] when |𝐶𝐷| ranges from
1150 to 4630.
(a) (b)
Fig. 100: Axial thrust in a centrifugal single stage well pump [84]: (a) Pressure distribution and
(b) Comparison of 𝑭𝒂𝒃 − 𝑭𝒂𝒇
-50
0
50
100
150
200
1000 2000 3000 4000 5000
𝐹𝑎
𝑏−
𝐹 𝑎𝑓
Suction
pressure
|𝐶𝐷|
Pre
ssu
re
in
the
bac
k c
ham
ber
Pressu
re in th
e
front ch
amb
er
Surface 3
Su
rfac
e 2
Su
rface 1
𝐹𝑎
126
7. Summary of the Results
The key parameters studied within the scope of this thesis are listed as follows:
a. Through-flow coefficient 𝐶𝐷:
① Centripetal through-flow;
② Centrifugal through-flow;
b. Core swirl ratio K;
c. Non-dimensional axial gap width G;
d. Global Reynolds number Re;
e. Local circumferential Reynolds number 𝑅𝑒𝜑;
f. Equivalent surface roughness 𝑘𝑠;
g. Pre-swirl;
h. Axial thrust coefficient 𝐶𝐹;
i. Moment coefficient 𝐶𝑀.
According to the results from both the numerical simulations and the experiments, some achieved
results and determined correlations are as follows:
a. Two correlations to evaluate the core swirl ratio K with a smooth disk for both centripetal
and centrifugal through-flow;
b. Two correlations for K by introducing the impact of surface roughness of the disk for both
centripetal and centrifugal through-flow;
c. Some experimental results of K in a rotor-stator cavity with centripetal or centrifugal pre-
swirl through-flow with a smooth disk;
d. Two equations for the axial thrust coefficient 𝐶𝐹 for the first time in the field of
turbomachinery to introduce the impacts of 𝐶𝐷, Re, G and 𝑘𝑠 for centripetal or centrifugal
through-flow;
e. A 3D diagram distinguishing regime III and regime IV for both centripetal and centrifugal
through-flow;
f. Two sets of correlations to predict the influences of 𝐶𝐷, Re, G and 𝑘𝑠 on 𝐶𝑀 for regime
III and regime IV.
The applications of the results in the future are offered:
a. An extensive data base of 𝐶𝐹 and 𝐶𝑀, which is of a huge worth for the designers of radial
pumps and turbines;
b. New correlations for 𝐶𝐹 and 𝐶𝑀 with good accuracy, which can be easily implemeted
into the design tools;
127
c. Effective sand exclusion in the rear chamber of radial slurry pumps, which can improve
the service life of sealings;
d. Reduction of the axial thrust, which can improve the service life of bearing systems in
radial pumps or turbines;
e. Reduction of the frictional losses of the disks, which can reduce the energy consumption
of a radial turbomachine by modifying the chamber geometry.
128
8. Outlook
Nevertheless a large amount of work has been accomplished, there are still some limitations in this
work.
1. The distinguishing lines for regime III and regime IV are obtained by evaluating the
tangential flow component based on numerical simulations. This has to be put on an
experimental level by measuring the velocity components in both the tangential and the
radial directions with a two-component LDV system in the future.
2. To quantify the impact of pre-swirl on the cavity flow, more investigations should be
accomplished in the future with more variations of the pre-swirl angle.
3. Currently, it is not possible to obtain rougher surfaces than those investigated in this thesis.
More complicated, the surface roughness is a 3D parameter, while the measured 𝑅𝑧 is the
average of the five highest peaks and the five deepest valleys (2D parameter). How to
expand the 2D surface roughness into 3D more correctly also deserves further investigation.
4. The impacts of boundary conditions at the inlet on K, 𝐶𝐹 and 𝐶𝑀 deserve further
investigation.
5. To better understand the scientific meaning of the correlations, massive fundamental
researches are still required in the future work.
6. Experiments in real machines have to be conducted to verify the results obtained in this
thesis and to show the applicability of the deduced correlations.
129
References
1. Altmann, D., 1972, Beitrag zur Berechnung der turbulenten Strömung im Axialspalt zwischen
Laufrad und Gehäuse von Radialpumpen, Dissertation, TU Magdeburg.
2. Altmann, D., 1984, Bestimmung der Radreibungsverluste bei radialen Kreiselpumpen mit offenen
Rückenschaufeln, Technische Mechanik, Band 5, Nr. 4, Seite 53-61.
3. Anderson, John D., 2004, Introduction to Flight, Section 4.20, 5th edition.
4. Andersson, H. I., Lygren, M., 2001, Turbulent flow between a rotating and a stationary disk, Journal
of Fluid Mechanics, Vol. 426, pp. 297-326.
5. Andersson, H. I., Lygren, M., 2006, LES of open rotor-stator flow, International Journal of Heat
and Fluid Flow, Vol. 27, No. 4, pp. 551-557.
6. Bahm, F. U., 2000, Das Axialschubverhalten einstufiger Kreiselpumpen mit Spiralgehäuse,
Dissertation, Universität Hannover.
7. Baibikov, A. S., V.K. Karakhan yan, 1973, Method of Calculating Axial Force in a Centrifugal
Pump with Allowance for Leakage, Journal Thermal Engineering, Vol. 20, No. 9, pp. 85-88.
8. Baibikov, A. S., 1998, Method of Calculation of a Turbulent Flow in an Axial Gap with a Variable
Radius between a Rotating Disk and an Axisymmetric Casing, Journal of Engineering Physics and
Thermophysics, Vol. 71, No. 6, pp. 1072-1081.
9. Baskharone, E. A, Wyman, N. J., 1999, Primary/leakage flow interaction in a pump stage, ASME
Journal of Fluid Engineering, 121(3), pp. 133-138.
10. Batchelor, G. K.,1951, Note on a class of solutions of the Navier-Stokes equations representing
steady rotationally-symmetric flow, The Quarterly Journal of Mechanics and Applied Mathematics,
Vol. 4, No. 1, pp. 29-41.
11. B.Barabas, S. Clauss, S. Schuster, F.-K. Benra, H.J. Dohmen, 2015, Experimental and numerical
determination of pressure and velocity distribution inside a rotor-stator cavity at very high
circumferential Reynolds numbers, 11th European Conference on Turbomachinery, Madrid, Spain.
12. Bayley, F. J., And Conway, L., 1964, Fluid Friction and Leakage between a Stationary and a
Rotating Disc, Journal of Mechanical Engineering Science, Vol. 6, pp.164-172.
13. Bayley, F. J., Owen, J.M., 1969, Flow between a rotating and a stationary disc, Aeronautical
Quarterly, Vol. 20, pp. 333-354.
14. Bo Hu, Benra, F.-K., Weigang Lu, 2015, Investigation on the solid-liquid flow in the rear chamber
of a submersible multi-stage slurry pump. 11th European Conference on Turbomachinery Fluid
Dynamics and Thermodynamics, Madrid, Spain.
15. Bo Hu, Benra, F.-K., Dohmen, H. J. and Dieter Brillert, 2015, Investigation on solid-liquid flow in
the rear chamber of multi-stage slurry pumps with leakage flow, 16th International Symposium on
Transport Phenomena and Dynamics of Rotating Machinery, ISROMAC, Honululu, Hawaii.
16. Bo Hu, Brillert. D, Dohmen, H. J. and Benra, F.-K., 2017, Investigation on the flow in a rotor-stator
cavity with centripetal through-flow, Int. J. Turbomach. Propuls. Power 2017, 2, 18;
DOI:10.3390/ijtpp2040018.
17. Bo Hu, Brillert. D, Dohmen, H. J., Benra, F.-K., 2017, Investigation on the influence of surface
roughness on the moment coefficient in a rotor-stator cavity with centripetal through-flow, ASME
2017 Fluids Engineering Division Summer Meeting, Waikoloa, Hawaii.
18. Bo Hu, Brillert. D, Dohmen, H. J. and Benra, F.-K., 2017, Investigation on thrust and moment
coefficients of a centrifugal turbomachine, 17th International Symposium on Transport Phenomena
and Dynamics of Rotating Machinery, ISROMAC, Maul, Hawaii.
19. Cheah, S. C., Iacovides, H., Jackson, D. C., Ji, H., Launder, B. E., 1994, Experimental Investigation
130
of Enclosed Rotor-Stator Disk Flows, Experimental Thermal and Fluid Science, Vol. 9, No. 4, pp.
445-455.
20. Chew, J. W., Vaughan, C. M., 1988, Numerical Predictions for the Flow Induced by an Enclosed
Rotating Disc, Gas Turbine and Aeroengine Congress and Exposition, Paper No. 89-GT-127,
Amsterdam, Netherlands.
21. Cooper, P., Reshotko, E., 1975, Turbulent Flow Between a Rotating Disc and a Parallel Wall, AIAA
Journal, Vol. 13, No. 5, pp. 573-578.
22. Cochran, W. G., 1934, The flow due to a rotating disk, Proceedings of the Cambridge Philosophical
Society, Vol. 30, pp. 365-375.
23. Coren D., Childs P. R. N., Long C. A., 2009, Windage sources in smooth-walled rotating disc
systems, Proceedings of the Institution of Mechanical Engineers, Part C: Journal of Mechanical
Engineering Science, Vol. 223, No. 4, pp. 873-888.
24. Craft, T., Iacovides, H., Launder, B., Athanasios, Z., 2008, Some Swirling-flow Challenges for
Turbulent CFD, Flow, Turbulence and Combustion, Vol. 80, No. 4, pp. 419-434.
25. Czarny, O., Iacovides, H., Launder, B.E., 2002, Precessing vortex structures in turbulent flow within
rotor-stator disc cavities, Flow Turbulence and Combustion, Vol. 69, No. 1, pp. 51-61.
26. Da Soghe, R., Facchini, B., Innocenti, L., Micio, M., 2010, Analysis of Gas Turbine Rotating
Cavities by a One-Dimensional Model: Definition of New Disk Friction Coefficient Correlations
Set, Journal of Turbomachinery, Vol. 133, No. 2, pp. 1-8.
27. Daily, J. W., Ernst, W. D., Asbedian, V. V., 1964, Enclosed Rotating Discs with Superposed
Through-Flow, Massachusetts Institute of Technology, Report No. 64.
28. Daily, J. W., and Nece, R. E., 1960, Chamber dimension effects on induced flow and frictional
resistance of enclosed rotating disks, Journal of Basic Engineering, Vol. 82, pp. 217-232.
29. Daily, J. W., and Nece, R. E., 1964, Enclosed Rotating Discs with Superposed Throughflow, Dept.
of Civil Engineering, MIT, Report No. 64.
30. Debuchy, R., Abdel Nour, F., and Bois, G., 2008, On the flow behavior in rotor-stator system with
superimposed flow, International Journal of Rotating Machinery, 10.1155/2008/719510.
31. Dibelius, G., Radtke, F. and Ziemann, M., 1984, Experiments on Friction, Velocity and Pressure
Distribution of Rotating Discs, Heat and mass transfer in rotating machinery, Washington DC,
Hemisphere Publishing Corp., pp. 117-130.
32. Dorfman, L. A., 1958, Resistance of a rotating rough disc, Technical physics Zhurnal tekhnicheskoi
fiziki, pp. 353-367.
33. Dorfman, L. A., 1963, Hydrodynamic resistance and the heat loss of rotating solids, Oliver and
Boyd, Edinburgh.
34. Elena, L., Schiestel, R., 1995, Turbulence Modeling of Confined Flow in Rotating Disk Systems,
AIAA Journal, Vol. 33, No. 5, pp. 812-821.
35. Emmanuel D’Haudt, Stefania Della Gatta, Roger Debuchy, Gerard Bois, Francesco Martelli, 2006,
Assessment of experimental and numerical flow investigation in rotating-disc systems, The
Eleventh International Symposium on Transport Phenomena and Dynamics of Rotating Machinery,
Honolulu, Hawaii, USA.
36. Gartner, W., 1997, A Prediction Method for the Frictional Torque of a Rotating Disc in a Stationary,
ASME international gas turbine & aero engine conference paper 97-GT-204.
37. Gartner, W., 1998, A Momentum Integral Method to Predict the Frictional Torque of a Rotating
Disc with Protruding Bolts, ASME international gas turbine & aero engine conference, pp. 98-GT-
138.
38. Gülich, J. F., 2010, Handbuch Kreiselpumpen, Springer-Verlag, Berlin, Germany.
131
39. Haddadi, S., Poncet, S., 2008, Turbulence Modeling of Torsional Couette Flows, International
Journal of Rotating Machinery, Volume 2008, Article ID 635138, 27 pages.
40. Han Hongbiao, Gao Shanqun, Li Jishun, Zhang Yongzhen, 2015, Exploring fluid resistance of disk
rotor based on boundary layer theory, Mechanical Science and Technology for Aerospace
Engineering, Vol. 34, No.10, pp. 1621-1625.
41. Hamkins, C. P., 1999, The Surface Flow Angle in Rotating Flow: Application to the Centrifugal
Pump Impeller Side Gap, Dissertation, TU Kaiserslautern.
42. Iacovides, H., Theofanopoulos, I. P., 1991, Turbulence modeling of axisymmetric flow inside
rotating cavities, International Journal of Heat and Fluid Flow, Vol. 12, No. 1, pp. 2-11.
43. Iacovides, H., Toumpanakis, 1993, Turbulence modelling of flow in axisymmetric rotor-stator
systems, 5th International Symposium on Refined Flow Modelling and Turbulence Measurements,
Paris, Int. Assoc. Hydr. Res., pp. 835-842.
44. Iino, T., Sato, H., Miyashiro, H., 1980, Hydraulic Axial Thrust in Multistage Centrifugal Pumps,
ASME Journal Fluids Engineering, Vol. 102, No. 1, pp. 64-69.
45. Itoh, M., Yamada, Y., Imao, S. Gonda, M., 1992, Experiments on turbulent flow due to an enclosed
rotating disk, Experimental Thermal and Fluid Science, Vol. 5, No. 3, pp. 359-368.
46. Itoh, M., 1995, Experiments on the turbulent flow in the narrow clearance between a rotating and a
stationary disk, Turbulent Flows - Proceedings of the Forum, 1997 ASME/JSME Fluids
Engineering and Laser Anenometry Conference and Exhibition, Aug. 13-18 1995, edited by B.F.
Carroll, T. Kobayashi, and M. J. Morris, Vol. 208, pp. 27-32.
47. Kármán, Th. v., 1921, Über laminare und turbulente Reibung, Zeitschrift für angewandte
Mathematik und Mechanik, Band 1, Nr. 4, pp. 233-252.
48. Kármán, Th. v., 1934, Turbulence and Skin Friction, Journal of the Aeronautical Sciences, Vol. 1,
No 1, pp. 1-20.
49. Karabay, H., Chen, J. X., Pilbrow, R., Wilson, M. and Owen, J. M., 1999, Flow in a "cover-plate"
pre-swirl rotor-stator system, J. Turbomachinery, v. 121, pp 160- 166.
50. Karabay, H., Wilson, M. and Owen, J. M., 2001, Predictions of effect of swirl on flow and heat
transfer in a rotating cavity, Int. J. Heat and Fluid Flow, v. 22, n. 2, pp 143-155.
51. K. N. Volkov, 2006, Moment of resistance of a disk rotating in a closed axisymmetric cavity,
Journal of applied Mechanics and Technical Physics, Vol. 47, No.1, pp. 127-133.
52. Kurokawa, J., Toyokura, T., 1972, Study on axial thrust of radial flow turbomachinery, The second
international JSME Symposium Fluid Machinery and Fluid Mechanics, Tokyo, pp. 31-40.
53. Kurokawa J., Toyokura T., 1976, Axial Thrust, Disc Friction Torque and Leakage Loss of Radial
Flow Turbomachinery, International Conference on Pump and Turbine Design and Development,
Vol. 1, Paper 5-2, Glasgow England.
54. Kurokawa J., Toyokura T., 1978, Roughness Effects on the Flow along an Enclosed Rotating Disc,
Bulletin of the JSME, Vol. 21, pp. 1725-1732.
55. Kurokawa, J., Sakuma, M., 1988, Flow in a Narrow gap along an Enclosed Rotating Disk with
Through-Flow, JSME International Journal, Series II, Vol. 31, No. 2, pp. 243-251.
56. Lance, G. N., Rogers, M. H., 1962, The Axially Symmetric Flow of Viscous Fluid between Two
Infinite Rotating Discs, Proceedings of the Royal Society of London, Series A, Mathematical and
Physical Sciences, Vol. 266, pp. 109-121.
57. Langtry, R. B., 2006, A Correlation-Based Transition Model using Local Variables for Unstructured
Parallelized CFD Codes, Dissertation, University of Stuttgart.
58. Lauer, J., 1999, Einfluss der Eintrittsbedingung und der Geometrie auf die Strömung in den
Radseitenräumen von Kreiselpumpen, Dissertation, TU Darmstadt.
132
59. Launder, B., Poncet, S. and Serre, E., 2010, Laminar, Transitional, and Turbulent Flows in Rotor-
Stator cavities, Annual Review of Fluid Mechanics, Vol. 42, pp. 229-48.
60. Lazauskas, Leo Victor, 2005, Hydrodynamics of Advanced High-Speed Sealift Vessels, Master
Thesis, University of Adelaide, Australia
61. Littell, H. S., Eaton, J. K., 1994, Turbulence characteristics of the boundary layer on a rotating disk,
Journal of Fluid Mechanics, Vol. 266, pp. 175-207.
62. Lomakin, A. A., 1940, Axial Pressure in Centrifugal Pumps, Taking into Account the Influence of
the Gap Size in the Packing Rings, Sovietskoe Kotloturbostroenie, Vol. 12, pp. 431-437.
63. Möhring, U. K., 1976, Untersuchung des radialen Druckverlaufes und des übertragenen
Drehmomentes im Radseitenraum von Kreiselpumpen bei glatter, ebener Radseitenwand und bei
Anwendung von Rückenschaufeln, Dissertation, TU Braunschweig.
64. Morse, A. P., 1991, Assessment of Laminar-Turbulent Transition in Closed Disk Geometries,
Journal of Turbomachinery, Vol. 113, pp. 131-138.
65. Owen, J. M., 1987, An Approximate Solution for the Flow between a Rotating and a Stationary
Disc, Thermo-Fluid Mechanics Research Centre, University of Sussex, Report.
66. Owen, J.M., Rogers, R.H., 1989, Flow and Heat Transfer in Rotating-Disc Systems, Volume 1:
Rotor-Stator Systems, Research Study Press, Taunton, Somerset, UK.
67. Pavesi, G., 2006, Impeller Volute and Diffuser Interaction, presented during the AVT-143 RTO
AVT/VKI Lecture Series held at the von Karman Institute, Rhode St. Gense, Belgium.
68. Poncet, S., Chauve, M.P., Le Gal, P., 2005, Turbulent Rotating Disk Flow with Inward Through-
Flow, Journal Fluid Mechanics, Vol. 522, pp. 253-262.
69. Poncet, S., Schiestel, R., Chauve, M.P., 2005, Turbulence Modelling and Measurements in a Rotor-
Stator System with Throughflow, Engineering Turbulence Modelling and Experiments ETMM6,
Elsevier (New-York), pp. 761-770.
70. Rabs, M., Benra, F. K., Dohmen, H.J., Schneider, O., 2009, Investigation of Flow Instabilities Near
The Rim Cavity Of A 1.5 Stage Gas Turbine, Proceedings of ASME Turbo Expo 2009: Power for
Land, Sea and Air, GT2009-59965, Orlando, Florida, USA.
71. Radtke, F., Ziemann, M., 1983, Experimentelle und theoretische Untersuchung des
Reibungseinflusses an rotierenden Scheiben, FVV Forschungsberichte Nr. 331, Frankfurt.
72. Randriamampianina, A., Elena, L., Schiestel, R., Fontaine, A.P., 1997, Numerical Prediction of
Laminar, Transitional and Turbulent Flows in Shrouded Rotor-Stator systems, Physics of Fluids,
Vol. 9, No. 6, pp. 1696-1713.
73. Schlichting, H., Gersten, K., 2006, Grenzschicht-Theorie, Springer-Verlag, Berlin, Germany.
74. Schultz-Grunow, F., 1935, Der Reibungswiderstand rotierender Scheiben in Gehäusen, Zeitschrift
für angewandte Mathematik und Mechanik, Band 15, Seite 191-204.
75. Senoo, Y., Hayami, H., 1976, An Analysis on the Flow in a Casing Induced by a Rotating Disk
Using a Four-Layer Flow Model, Transactions of the ASME, Journal of Fluids Engineering, Vol.
99, pp. 192-198.
76. Séverac, É., Poncet, S., Serre, É., 2007, Large Eddy Simulation and Measurements of Turbulent
Enclosed Rotor-Stator Flow, Physics of Fluids, Vol. 19, No. 8.
77. Stewartson, K., 1953, On the Flow between Two Coaxial Rotating Disks, Proceedings of the
Cambridge Philosophical Society, Vol. 49, pp. 333-341.
78. Soo, S. L., 1958, Laminar flow over an enclosed rotating disk, Journal of Basic Engineering, Vol.
80, No. 2, pp. 287-296.
79. Stepanoff, A., 1932, Leakage Loss and Axial Thrust in Centrifugal Pumps, ASME Paper HYD-54-
5, pp. 65-103.
133
80. Tamm, A., 2002, Beitrag zur Bestimmung der Wirkungsgrade einer Kreiselpumpe durch
theoretische, numerische und experimentelle Untersuchungen, Dissertation, TU Darmstadt.
81. Thomas Adams and Christopher Grant, 2012, A Simple Algorithm to Relate Measured Surface
Roughness to Equivalent Sand-grain Roughness, International Journal of Mechanical Engineering
and Mechatronics, Vol. 1, pp. 66-71.
82. Truckenbrodt, E., 1954, Die turbulente Strömung an einer angeblasenen rotierenden Scheibe,
Zeitschrift für angewandte Mathematik und Mechanik, Band 34, Heft 4/5, Seite 150-162.
83. Watanabe, K., Budiarso, Ogata, S., Uemura, K., 2007, Drag Reduction of an Enclosed Rotating
Disk with Fine Spiral Grooves, Journal of Environment and Engineering, Vol. 2, No. 1, pp. 97-107.
84. W. Shi, H. Wang, L. Zhou, P. Zou, 2010, The Estimation and Experiment of Axial Force in Deep
Well Pump Basing on Numerical Simulation. I.J.Modern Education and Computer Science, Vol. 2,
pp. 53-61.
85. Will B. C., Benra F. -K. 2009, Investigation of the Fluid Flow in a Rotor-Stator Cavity with Inward
Through-Flow, Proceedings of FEDSM2009, ASME Fluids Engineering Conference,
FEDSM2009-78503, Vail, Colorado, USA
86. Will, B. C., Benra, F.-K., Dohmen, H. J., 2010, Numerical and Experimental Investigation of the
Flow in the Side Cavities of a Centrifugal Pump, The 12th International Symposium on Transport
Phenomena and Dynamics of Rotating Machinery, ISROMAC13-2010-0002, Honolulu, Hawaii,
USA.
87. Will, B. C., Benra, F.-K., Dohmen, H. J., 2011, Investigation of the Flow in the Side Chambers of
a Centrifugal Pump with Volute Casing, 10th International Symposium on Experimental
Computational Aerothermodynamics of Internal Flows, Brussels, Belgium.
88. Will, B. C., 2011, Theoretical, Numerical and Experimental Investigation of the Flow in Rotor-
Stator Cavities with Application to a Centrifugal Pump, Dissertation, University of Duisburg-Essen.
89. Wu, X., Squires, K. D., 2000, Prediction and investigation of the turbulent flow over a rotating disk,
Journal of Fluid Mechanics, Vol. 418, No. 1, pp. 231-264.
90. Zilling, H., 1973, Untersuchung des Axialschubs und der Strömungsvorgänge in den
Radseitenräumen einer einstufigen radialen Kreiselpumpe mit Leitrad, Dissertation, TU-Karlsruhe.
91. Zhao W. G., 2006, Research on wearing-rings of centrifugal pump based on CFD code, Wuhan:
Huazhong University of Science and Technology, No. 4, pp. 27-28.