2 PhDThesis Design Improvement of Crossflow Hydro Turbine

University of Calgary
PRISM: University of Calgary's Digital Repository
Graduate Studies The Vault: Electronic Theses and Dissertations
2016
Design Improvement of Crossflow Hydro Turbine
Adhikari, Ram
Adhikari, R. (2016). Design Improvement of Crossflow Hydro Turbine (Unpublished doctoral

thesis). University of Calgary, Calgary, AB. doi:10.11575/PRISM/25581
http://hdl.handle.net/11023/3335
doctoral thesis
University of Calgary graduate students retain copyright ownership and moral rights for their
thesis. You may use this material in any way that is permitted by the Copyright Act or through
licensing that has been assigned to the document. For uses that are not allowable under
copyright legislation or licensing, you are required to seek permission.
Downloaded from PRISM: https://prism.ucalgary.ca
UNIVERSITY OF CALGARY
Design Improvement of Crossflow Hydro Turbine
by
Ram Chandra Adhikari
A THESIS
SUBMITTED TO THE FACULTY OF GRADUATE STUDIES
IN PARTIAL FULFILMENT OF THE REQUIREMENTS FOR THE
DEGREE OF DOCTOR OF PHILOSOPHY
GRADUATE PROGRAM IN MECHANICAL AND MANUFACTURING ENGINEERING
CALGARY, ALBERTA
SEPTEMBER, 2016

c Ram Chandra Adhikari 2016
Abstract
Efficiency is a critical consideration in the design of turbines. A computational study has
been conducted, using steady and unsteady three-dimensional Reynolds-Averaged Navier-
Stokes computations, to improve the maximum efficiency of crossflow hydro turbines that are
typically used for small-scale, low-head remote power systems. Although these turbines are
simple to design and manufacture at low cost, a longstanding technical problem is their lower
maximum efficiency than their more advanced counterparts, such as Pelton and Francis.
Despite an inherently simple design, the flow field inside the crossflow turbine is highly
complex. Since internal flow physics govern the turbine performance, major emphasis is
put on understanding the important performance limiting flow mechanisms as an aid for
improving the efficiency. The study follows a direct design method, which involves defining
the turbine configurations, computing the flow field, and modifying the design by synthesizing
the information obtained from the computed flow fields to improve the design. Toward
this goal, computations were performed on two small-scale turbines whose performance was
experimentally determined, one with 69% efficiency and the other with 88% efficiency. The
findings of the simulations were synthesized to identify the key design problem and define
an optimum turbine configuration. Then detailed computations were performed on varying
turbine configurations for establishing the qualitative and quantitative links between the
geometric parameters and the key flow features and understanding their roles on turbine
performance. A critical part of the analysis involves matching two main components, the
nozzle and the impeller, in order to achieve higher efficiency. For the nozzle design, a simple
two-dimensional analytical model was developed with the intention of converting all the
turbine head into kinetic energy at the impeller inlet and computing approximately the nozzle
geometrical parameters as a guide for computationally intensive simulations. Systematic
computations were performed on a realistic design problem, which is the matching of the
nozzle and the impeller designs in the reference turbines mentioned above. The results
showed that crossflow turbines can achieve efficiencies more than 90%, computed efficiency
of which is a significant improvement in the existing design. An example turbine design with
91% efficiency has been achieved for the original 69% efficient turbine.
ii
Acknowledgements
I consider myself very fortunate to have had Professor David Wood as my supervisor.
First and foremost, I would like to thank him for bringing me at the University of Calgary
to pursue MSc and PhD studies. His indepth knowledge, guidance and encouragement were
invaluable in my growth as a researcher. He has been an indispensable source of knowledge
and support to me both within the research and personal life. I will never forget his support
in teaching me to become a problem solver of practical problems. I am most thankful to him
for always being available to answer my questions and willingness to help on what I need
and want. I will never forget those great times during our many trips to Nepal with the aim
of solving technical problems in small-scale remote power systems. You are a great source
of inspiration in my personal life as well.
I would like to thank the members of my PhD thesis committee, Professor Robert Mart-
inuzzi, Professor Craig Johansen, Professor Chris Morton and Professor Jennifer He for both
their time as thesis committee members and providing useful feedbacks. Special thanks goes
to Professor Robert Martinuzzi for providing useful feedbacks during the early phase of my
research. I would also like to thank Professor Eric Bibeau for agreeing to serve as an ex-
ternal examiner and providing useful comments and suggestions. I wish to thank Professor
Ed Nowicki for sharing his research expertise and ideas on remote power systems during our
trip to Nepal. I also wish to thank Cassandra Arnold for sharing great fun times during our
trip to Nepal. You are a great source of inspiration in my personal life as well.
I also would like to acknowledge my fellow labmates of the fluid dynamics research group
for their support: Eric, Dr. Hammam, Philip, Dr. Rif, Adrian, Anna, Dr. Arman, Iman,
Mohammed, Tahsina Loba, Ehsan, Matthew...I would also like to thank Dr. Jerson for
sharing ideas in my research. I would like to thank Dough Philips of WestGrid, Compute
Canada at UofC for helping me on using high-performance computing resources of WestGrid.
Finally I thank my family for their constant support and encouragement. I owe a great
debt to my brothers who were an immense source of support throughout my student life. I
will never forget your support and love. I would also like to thank my father for supporting
my education and constantly encouraging me to achieve higher. I want to thank my wife
Sushma for her unwavering love and support throughout the years of my PhD study. Without
her understanding, I could not have made it to this point. I owe a great debt to my son
Abhik, who was just innocent to allow his father to leave home for research. Thanks for all
the joyful moments you shared with me. I remember you would love to follow me when I had
to leave for school. I would also like to thank all the members of our Nepalese community
in Calgary for having incredibly great fun times during the past few years.
iii
Dedicated to the loving memory of my mother Keshar Kumari Adhikari
iv
Contents
Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ii
Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iii
List of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ix
List of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . x
List of Symbols . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xxiv
List of Abbreviations, Symbols, and Nomenclature xxiv
1 Introduction 1
1.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Types of Turbines and Design Requirements . . . . . . . . . . . . . . . . . . 4
2 Thesis Objectives 7
2.1 Motivation and Background . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.2 Problem Statement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.3 Research Questions and Goals . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.4 Flow Field Characteristics in Crossflow Turbines . . . . . . . . . . . . . . . . 14
2.5 Literature Review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.5.1 Experimental Studies . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.5.2 Numerical Studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
2.6 Part-load Operation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
2.7 Organization of the Thesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
v
3 Computational Method 28
3.1 Governing Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
3.2 Multiphase Flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
3.3 Computational Details . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
3.3.1 Computational Mesh . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
3.3.2 Grid Convergence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
3.3.3 Turbulence Modeling . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
3.3.4 Boundary Conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
3.4 Cavitation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
3.5 CFD Model Validation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
3.6 Methods of Computational Analysis . . . . . . . . . . . . . . . . . . . . . . . 48
3.7 Chapter Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
4 Validation and Performance Characterization 49

4.1 Turbine Efficiency and Performance Metrics . . . . . . . . . . . . . . . . . . 50
4.2 Results of Validation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
4.2.1 Computational Details and Results . . . . . . . . . . . . . . . . . . . 55
4.2.2 Unsteady RANS Computations . . . . . . . . . . . . . . . . . . . . . 62
4.3 Key Flow Features and Performance Characteristics . . . . . . . . . . . . . . 66
4.4 Power Production Mechanism - Stage Performance . . . . . . . . . . . . . . 85
4.5 Implications to Nozzle Design . . . . . . . . . . . . . . . . . . . . . . . . . . 94
4.6 Implications to Impeller Design . . . . . . . . . . . . . . . . . . . . . . . . . 95
4.7 Cavitation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
4.7.1 Cavitation Inception . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
4.7.2 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
4.7.3 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
4.8 Chapter Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
vi
5 The Key Components for Design: Nozzle and Impeller 107
5.1 Nozzle and Impeller Design Parameters . . . . . . . . . . . . . . . . . . . . . 109
5.2 Analytical Model for Nozzle Design . . . . . . . . . . . . . . . . . . . . . . . 113
5.3 Impeller Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
5.4 Inlet-flow Control Mechanism . . . . . . . . . . . . . . . . . . . . . . . . . . 121
5.5 Chapter Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
6 Design Improvement 124

6.1 Key Design Concepts for Improvement . . . . . . . . . . . . . . . . . . . . . 125
6.2 Sensitivity Analysis: 0.53 kW Turbine - What Matters Most? . . . . . . . . 126
6.2.1 Influence of Nozzle Design on the Efficiency of Impeller . . . . . . . . 127
6.2.2 Influence of Number of Blades (Nb ) . . . . . . . . . . . . . . . . . . . 137
6.2.3 Influence of Outer Blade Angle (β1b ) . . . . . . . . . . . . . . . . . . 145
6.2.4 Influence of Inner Blade Angle (β2b ) . . . . . . . . . . . . . . . . . . . 151
6.2.5 Influence of Impeller Diameter Ratio (D2 /D1 ) . . . . . . . . . . . . . 153
6.2.6 Performance at Part-Load - Slider Control Mechanism . . . . . . . . 158
6.2.7 Synthesis of Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167
6.3 Improved Design: 0.53 kW Turbine . . . . . . . . . . . . . . . . . . . . . . . 168
6.4 Improved Design: 7 kW Turbine . . . . . . . . . . . . . . . . . . . . . . . . . 171
6.4.1 Influence of Nozzle Design on the Efficiency of Impeller . . . . . . . . 172
6.4.2 Matching of Nozzle and Impeller Designs . . . . . . . . . . . . . . . . 181
6.5 Second-order Effects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 191
6.6 Synthesis of Results on Design Improvement . . . . . . . . . . . . . . . . . . 192
6.7 Chapter Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 192
7 Conclusions 194
7.1 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 194
7.2 Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 199
vii
A Analytical Equations for Efficiency 201
B Generalized Equation for Nozzle Design 204
C Error and Uncertainty 206
Bibliography 209
viii
List of Tables
2.1 Summary of empirical values for the design parameters used in the experimen-
tal studies by the previous researchers. The symbol [*] represents the optimal
value at the maximum efficiency point. For the symbols, refer Figure 2.1. . . 18
4.1 Design parameters of 7 kW turbine . . . . . . . . . . . . . . . . . . . . . . . 57

4.2 Operating conditions of 7 kW turbine . . . . . . . . . . . . . . . . . . . . . . 58
4.3 Design parameters of 0.53 kW turbine . . . . . . . . . . . . . . . . . . . . . 59
4.4 Operating conditions of 0.53 kW turbine . . . . . . . . . . . . . . . . . . . . 59
4.5 Summary of mesh convergence test for the 7 kW turbine at the maximum
efficiency point: 450 RPM (steady RANS calculations). . . . . . . . . . . . . 61
4.6 Summary of mesh convergence test for the 0.53 kW turbine at the maximum
efficiency point: 199.1 RPM (steady RANS calculations). . . . . . . . . . . . 61
6.1 Impeller design parameters at different outer blade angle β1b . . . . . . . . . 146
6.2 Impeller design parameters at different β2b . . . . . . . . . . . . . . . . . . . 152
6.3 Impeller design parameters at different D2 /D1 . . . . . . . . . . . . . . . . . 155
6.4 Operating parameters for the slider control . . . . . . . . . . . . . . . . . . . 159
6.5 New nozzle design parameters . . . . . . . . . . . . . . . . . . . . . . . . . . 173
6.6 Design parameters of the new impeller . . . . . . . . . . . . . . . . . . . . . 182
ix
List of Figures
1.1 Classification of various turbines based on head H and flow rate Q illustrating
their application range (Benzon et al., 2016). . . . . . . . . . . . . . . . . . . 5
1.2 Comparison of typical efficiencies of different turbines (Sinagra et al., 2014).
Note that the efficiency is the hydrodynamic efficiency, i.e. the efficiency of
energy extraction by the blades, not the efficiency of conversion into electrical
energy. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.1 Schematic illustration of the basic design features of a crossflow turbine. . . . 9

2.2 Schematic illustration of the guide vane in the crossflow turbine. . . . . . . . 23
3.1 A typical contour of water and air volume fractions in a crossflow turbine
computed using homogeneous multiphase free-surface model. The red and
blue regions show respectively water and air in the computational domain,
illustrating a discernible free-surface interface between them. The picture is
taken from one of the snapshots of the author’s unsteady RANS simulations. 31
3.2 An illustration of the conditions at an ideal free-surface of water and air in
the computational domain. Adapted from White and Corfield (2006). . . . . 32
3.3 An illustration of the typical computational domain for the crossflow turbine. 35
4.1 Steady flow through the control volume of the impeller. . . . . . . . . . . . . 52
x
4.2 A schematic illustration of the conditions of power extraction and loss across
the blades at the second stage of the impeller. Note that the inner blade angle
is 90◦ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
4.3 Nozzle dimensions of 7 kW turbine. Redrawn from Dakers and Martin (1982). 56
4.4 Impeller dimensions of 7 kW turbine. Data taken from Dakers and Martin
(1982). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
4.5 Schematic of the dimensions of the 0.53 kW turbine. Data taken from Desai
(1994). All the components have constant width W = 101.6 mm normal to
the page. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
4.6 An illustration of a typical computational mesh of the rotating domain: the
impeller. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
4.7 Zoomed view of the mesh resolution around the blades of the impeller. . . . 60
4.8 An illustration of a typical computational mesh of the stationary domain: the
nozzle and the casing. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
4.9 Comparison of the steady and unsteady RANS solutions with the experimental
results for the turbine power output of the 7 kW turbine. Experimental data
taken from Dakers and Martin (1982). . . . . . . . . . . . . . . . . . . . . . 63
4.10 Comparison of the steady RANS solutions with the experimental results for
the turbine power output of the 0.53 kW turbine. The uncertainty in the
experimental data is ± 2.4% [Q = 46 lps, H = 1.337 m and Nb = 30]. Exper-
imental data taken from Desai (1994). . . . . . . . . . . . . . . . . . . . . . 63
4.11 Comparison of the steady RANS solutions with the experimental results for
the efficiency of the 0.53 kW turbine. The uncertainty in the experimental
data is ± 2.4% [Q = 42.56 lps, H = 1.17 m and Nb = 25]. Experimental data
taken from Desai (1994). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
xi
4.12 Schematic illustration of the conventions used in the analysis. The azimuthal
position ψ, nozzle throat, left nozzle lip, right nozzle lip, first stage, and second
stage will be used. ψ will be used to refer to a specific position or a region in
the nozzle entry arc and the first and second stages. Configuration is that of
the 0.53 kW turbine. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
4.13 Schematic illustration of different flow regimes in a crossflow turbines. . . . . 70
4.14 Contour plot of the magnitude of mean water velocity in the 7 kW turbine
illustrating the main flow path in crossflow turbines. . . . . . . . . . . . . . . 70
4.15 Azimuthal variations of the computed absolute, tangential and radial veloci-
ties at the impeller inlet of the 7 kW turbine. Variations across the inlet are
due to the position of the blades. For a time averaged flow - which is assumed
in the simple nozzle model, these variations would average out. [Q = 105 lps
and H = 10 m]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
4.16 Comparison of the inlet flow angle β1 with the outer blade angle at the inlet
of the impeller of the 7 kW turbine at maximum efficiency [Q = 105 lps, H
= 10 m and N = 450 RPM]. . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
4.17 Close view of water velocity vectors illustrating the major flow separation on
the pressure sides of the blades at the first stage of the 7 kW turbine at the
maximum efficiency point [H = 10 m, Q = 105 lps and N = 450 RPM]. Note
that there is no separation in the second stage. . . . . . . . . . . . . . . . . . 73
4.18 Azimuthal variations of the computed absolute, tangential and radial veloci-
ties at the impeller inlet of the 0.53 kW turbine [Q = 46 lps and H = 1.33
m]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
4.19 Azimuthal variation of the inlet flow angle β1 at the impeller inlet of the 0.53
kW turbine at the maximum efficiency point: 88% [Q = 46 lps, H = 1.33 m
and N = 199.1 RPM]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
xii
4.20 Contours of the magnitude of the mean water velocity in the 0.53 kW turbine
at the maximum efficiency point [H = 1.33 m, Q = 46 lps and N = 199.1
RPM]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
4.21 Water velocity vectors in the first stage of the 0.53 kW turbine at the max-
imum efficiency point, illustrating the suction side separation on the blades
[H = 1.33 m, Q = 46 lps and N = 199.1 RPM]. . . . . . . . . . . . . . . . . 77
4.22 Computed exit flow angle from the first stage of the 7 kW turbine at maximum
efficiency [H = 10 m, Q = 105 lps and N = 450 RPM]. The local minima are
due to the wakes of each blade. . . . . . . . . . . . . . . . . . . . . . . . . . 78
4.23 Exit flow angle at the first stage of the 0.53 kW turbine at maximum efficiency
[H = 1.33 m, Q = 46 lps and N = 199.1 RPM]. The local minima are due to
the presence of wakes of each blade. . . . . . . . . . . . . . . . . . . . . . . . 78
4.24 Close view of the water velocity vectors illustrating the main flow pattern in
the 0.53 kW turbine at maximum efficiency. Note that the flow is almost fully
attached in both stages. Blade wakes from the first stage can also be seen.
The flow at the second stage is also aligned with the blades and is slightly
accelerated in the blade passages [H =1.33 m, Q = 46 lps and N = 199.1 RPM]. 79
4.25 Inlet flow angle β2i at the second stage of the 7 kW turbine at maximum
efficiency [H = 10 m, Q = 105 lps and N = 450 RPM]. Note that the local
minima due to the wakes of the first stage blades are decayed by the entry to
the second stage. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
4.26 Inlet flow angle β2i at the second stage of the 0.53 kW turbine at maximum
efficiency [88%]. Note that the effects of wakes have decayed at the inlet of
the second stage. Note that the local minima due to the wakes of the first
stage blades are decayed by the entry to the second stage. . . . . . . . . . . 81
4.27 Streamlines superimposed on the velocity contour at the second stage of the
0.53 kW turbine illustrating flow alignment to the blades. . . . . . . . . . . . 84
xiii
4.28 Streamlines superimposed on the velocity contour at the second stage of the
0.53 kW turbine illustrating flow alignment to the blades. The impeller speed
199 RPM corresponds to the maximum efficiency point. . . . . . . . . . . . . 84
4.29 Azimuthal variation of tangential and radial velocities at the exit of second
stage of the 7 kW turbine at maximum efficiency [H = 10 m, Q = 105 lps,
N = 450 RPM]. Note that the tangential velocity is small, implying that the
residual angular momentum exiting the second stage is minimum. . . . . . . 86
4.30 Azimuthal variation of power production at different impeller speeds for the
7 kW turbine [H = 10 m, Q = 105 lps]. Impeller speeds 400 - 450 RPM
correspond to the maximum efficiency. Note that the negative values represent
the power lost by the blades into the flow. The data values represent blade
positions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
4.31 Percentage contribution to the total power production by the first and second
stages of the 7 kW turbine. . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
4.32 Azimuthal variation of power production at different impeller speeds for the
0.53 kW turbine [H = 1.337 m, Q = 46 lps]. N = 199.1 RPM corresponds to
the maximum efficiency. Note that the negative values represent the power
lost by the blades into the flow. The symbols represent blade positions. . . . 91
4.33 Percentage contribution to the total power production by the first and second
stages at different impeller speeds of the 0.53 kW turbine. . . . . . . . . . . 92
4.34 Schematic illustration of pressure distribution on a cavitating blade in the
impeller. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
xiv
4.35 Contours of pressure distributions on the impeller blades at the second stage
[Q = 105 lps and H= 10 m] at the impeller speed 450 RPM, showing the
evidence of cavitation inception on the suction sides near the inner edge of
the blade. It is noted that the scale of the contour is only for the cavitation
region, and the pressure outside this region (red) is significantly higher than
the vapor pressure. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
[Q = 105 lps and H= 10 m] at the impeller speed 550 RPM, showing the
evidence of cavitation inception on the suction sides near the inner edge of
the blade. It is noted that the scale of the contour is only for the cavitation
region, and the pressure outside this region (red) is significantly higher than
the vapor pressure. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
at Q = 105 lps, H = 8 m and N = 450 RPM, showing the evidence of
cavitation inception on the suction sides near the inner edge of the blades.
It is noted that two blades have cavitation at H = 8 m, whereas only one
blade has cavitation at H = 10 m. It is noted that the scale of the contour
is only for the cavitation region, and the pressure outside this region (red) is
significantly higher than the vapor pressure. . . . . . . . . . . . . . . . . . . 102
4.38 CFD predicted cavitation region for the 7 kW turbine. . . . . . . . . . . . . 103
5.1 Schematic illustration of the turbine geometry with tangential entry nozzle.
Every component in the diagram has equal width W normal to the page. . . 110
5.2 Schematic illustration of the flow angles. . . . . . . . . . . . . . . . . . . . . 111
5.3 Schematic illustration of the geometry of the tangential entry nozzle used in
the analysis. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
5.4 Schematic illustration of the generalized nozzle geometry used in the analysis. 117
5.5 Slider control mechanism for controlling the inlet flow . . . . . . . . . . . . . 122
xv
6.1 Schematic illustration of the conventions used in the analysis. The azimuthal
position ψ, nozzle throat, left nozzle lip, right nozzle lip, first stage, and second
stage will be used. ψ will be used to refer to a specific position or a region in
the nozzle entry arc and the first and second stages. It is noted that for the
0.53 kW turbine, the left nozzle lip is at the azimuthal position 120◦ and the
right nozzle lip at 210◦ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
6.2 Schematic of the new nozzle with the original nozzle orientation. . . . . . . . 130
6.3 Schematic of the tangential entry nozzle. . . . . . . . . . . . . . . . . . . . . 130
6.4 Efficiency vs speed for different nozzle designs for the 0.53 kW turbine [H =
1.337 m and Q = 46 lps]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
6.5 Comparison of velocities at the impeller inlet for the new nozzle and existing
nozzle for the 0.53 kW turbine [H = 1.337 m and Q = 46 lps]. . . . . . . . . 132
6.6 Comparison of inlet flow angle β1 for the new nozzle and the existing nozzle
at maximum efficiency of the 0.53 kW turbine [H = 1.337 m, Q = 46 lps and
N = 199.1 RPM]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
6.7 Contour plot of the magnitude of mean water velocity for the original nozzle
at maximum efficiency of the 0.53 kW turbine [H = 1.337 m, Q = 46 lps and
N = 199.1 RPM]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
6.8 Contour plot of the magnitude of mean water velocity for the new nozzle at
the maximum efficiency point of the 0.53 kW turbine [H = 1.337 m, Q = 46
lps and N = 199.1 RPM]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
6.9 Contour plot of the magnitude of mean water velocity for the tangential nozzle
at the maximum efficiency point of the 0.53 kW turbine [H = 1.337 m, Q =
46 lps and N = 199.1 RPM]. . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
xvi
6.10 Comparison of the inlet flow angle β1 for the tangential nozzle and existing
nozzle at the maximum efficiency point of the 0.53 kW turbine. The azimuthal
position ψ for the tangential nozzle has been increased by 30◦ for the purpose
of comparison [H = 1.337 m, Q = 46 lps and N = 199.1 RPM]. . . . . . . . 135
6.11 Comparison of power outputs in the impeller for different nozzle designs at
the maximum efficiency point [199.1 RPM]. Note that the changes are due
to the change in the inlet flow angle β1 . The azimuthal position ψ for the
tangential nozzle has been increased by 30◦ for the purpose of comparison. . 135
6.12 Comparison of percentage power production in the impeller stages for the
different nozzles at the maximum efficiency point of the 0.53 kW turbine
[199.1 RPM]. Note that the changes are due to the change in the inlet flow
angle β1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136
6.13 Influence of blade number on the efficiency. . . . . . . . . . . . . . . . . . . . 138
6.14 Influence of blade number on the power extraction in the impeller at the
maximum efficiency point (199 RPM). The data are computed at each blade
position and normalized with the torque for 30 blades. . . . . . . . . . . . . 140
6.15 Influence of number of blades on percentage power production in the impeller
stages. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140
6.16 Contours of the magnitude of the mean water velocity in the impeller with
20 blades at the maximum efficiency point. Note the suction side separation
near the right nozzle lip, where the power extraction is lower compared to 30
or 35 blades. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
6.17 Water velocity vectors at the second stage of the impeller with 20 blades at
the maximum efficiency point illustrating the regions of main flow separations
in the impeller of the 0.53 kW turbine. . . . . . . . . . . . . . . . . . . . . . 141
xvii
6.18 Streamlines superimposed on the water velocity contours at the second stage
of the impeller with 20 blades at the maximum efficiency point. Note that
the inlet flow angle at the entry to the second stage in the lower portion of
the flow is less than or almost equal to the inner blade angle in the impeller
of the 0.53 kW turbine. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142
6.19 Contours of the magnitude of the mean water velocity in the impeller with 30
blades at the maximum efficiency point. . . . . . . . . . . . . . . . . . . . . 142
6.22 Influence of outer blade angle β1b on the efficiency of the turbine [H = 1.337
m, Q = 46 lps]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147
6.23 Influence of outer blade angle β1b on the power production in the first and
second stages of the impeller at the maximum efficiency points [H = 1.337 m,
Q = 46 lps]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147
6.24 Influence of outer blade angle β1b on the percentage power production at the
first and second stages at the maximum efficiency points [H = 1.337 m, Q =
46 lps]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 148
6.25 Water velocity contours in the impeller at β1b = 43◦ . Note the suction side
separation near the right nozzle lip, responsible for decrease in power extraction.149
6.26 Water velocity vectors in the impeller at β1b = 43◦ , illustrating the flow sep-
aration on the suction sides of the blades near the right nozzle lip and the
second stage. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150
6.27 Water velocity contours in the impeller for β1b = 39◦ at maximum efficiency. 150
6.28 Water velocity contours in the impeller for β1b = 35◦ at maximum efficiency. 151
xviii
6.29 Azimuthal variation of power production in the impeller at the maximum
efficiency point for β2b = 90 and 85◦ . . . . . . . . . . . . . . . . . . . . . . . 152
6.30 Close view of water velocity streamlines superimposed on the velocity contour
map in the second stage for β2b = 85◦ at maximum efficiency, illustrating that
a considerable portion of the flow at the entry of the second stage is misaligned
with the blades [inlet flow angle β2i greater than β2b = 85◦ ] that results into
significant performance loss in the second stage. . . . . . . . . . . . . . . . . 153
6.31 Velocity vectors in the impeller at the maximum efficiency point for β2b = 85◦ ,
illustrating that a considerable portion of the flow at the entry of the second
stage is misaligned with the blades [inlet flow angle β2i greater than β2b =
85◦ ] that results into significant performance loss in the second stage. . . . . 154
6.32 Influence of diameter ratio D2 /D1 on the impeller performance at the maxi-
mum efficiency point. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155
6.33 Influence of diameter ratio D2 /D1 on the percentage of total power production
at the first and second stages at the maximum efficiency point. . . . . . . . . 156
6.34 Close view of the slider at the inlet of the impeller [Q = 20 lps, H = 1.337
m]. Note that only the inlet section of the slider is modelled. . . . . . . . . . 160
6.35 Part-flow efficiencies at different impeller speeds of crossflow turbine with
slider control mechanism for the inlet flow. . . . . . . . . . . . . . . . . . . . 160
6.36 Part-flow efficiencies of crossflow turbine with slider and without slider control
mechanism. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161
6.37 Stage performances at various part-flow operating conditions. . . . . . . . . . 162
6.38 Velocity at the impeller inlet [Q = 20 lps, H = 1.337 m]. . . . . . . . . . . . 163
6.41 Inlet flow angle at the impeller inlet [Q = 20 lps, H = 1.337 m and N = 199.1].164
xix
6.44 Contour maps of the magnitude of the mean water velocity at the maximum
efficiency point for Q = 20 lps with slider. . . . . . . . . . . . . . . . . . . . 166
efficiency point for Q = 30 lps with slider. . . . . . . . . . . . . . . . . . . . 166
efficiency point of Q = 40 lps with slider. . . . . . . . . . . . . . . . . . . . . 167
6.47 Comparison of the efficiency of the new nozzle and the existing design. The
improved design has new nozzle and the impeller with 35 blades. . . . . . . . 169
6.48 Comparison of azimuthal variation of power production in the impeller of the
new design and the existing designs. The improved design has the new nozzle
and the impeller with 35 blades. . . . . . . . . . . . . . . . . . . . . . . . . . 169
6.49 Streamlines superimposed on the contour map of the magnitude of the mean
water velocity at the maximum efficiency point. Note that the flow is perfectly
aligned with the blade in both stages. The improved design has the new nozzle
and the impeller with 35 blades. . . . . . . . . . . . . . . . . . . . . . . . . . 170
6.50 Water velocity vectors in the impeller at the maximum efficiency point, illus-
trating the main flow in the impeller. . . . . . . . . . . . . . . . . . . . . . . 170
6.51 Schematic illustration of the new nozzle design for the 7 kW turbine [Q = 105
lps and H = 10 m]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 173
6.52 Azimuthal variations of the computed absolute, tangential and radial veloc-
ities at the impeller inlet with the new nozzle [Q = 105 lps and H = 10
m]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 174
6.53 Comparison of the inlet flow angle β1 and the blade angle at the maximum
efficiency points for the existing impeller with the new nozzle and the existing
nozzle. [Q = 105 lps and H = 10 m]. . . . . . . . . . . . . . . . . . . . . . . 175
xx
6.54 Streamlines superimposed on the contour map of the magnitude of the mean
water velocity at the maximum efficiency point. . . . . . . . . . . . . . . . . 176
6.55 Water velocity vectors at the maximum efficiency point, illustrating the flow
separation and flow angle at the entry of the second stage for the existing
impeller (20 blades) with the new nozzle. . . . . . . . . . . . . . . . . . . . . 177
6.56 Comparison of the computed flow angle and the inner blade angle at the exit
of the first stage of the existing impeller with new nozzle at the maximum
efficiency point [N = 500 RPM]. . . . . . . . . . . . . . . . . . . . . . . . . . 177
6.57 Comparison of the computed flow angle and the inner blade angle at the inlet
of the second stage of the existing impeller with new nozzle at the maximum
efficiency point. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 178
6.58 Comparison of power production in the impeller at the maximum efficiency
points between the existing and the new nozzle designs. . . . . . . . . . . . . 179
6.59 Comparison of percentage power production in the impeller with the original
and new nozzles at the maximum efficiency points. . . . . . . . . . . . . . . 179
6.60 Performance characteristics of the improved design [H = 10 m, Q = 105 lps]. 183
6.61 Comparison of the computed inlet flow angle β1 and the outer blade angle at
the inlet of the impeller at the maximum efficiency point: N = 500 RPM. . . 184
6.62 Comparison of the computed relative flow angle and the blade angle at the
exit of the first stage at the maximum efficiency point: N = 500 RPM. Note
that the azimuthal range: 180 - 220◦ corresponds to the region of maximum
power extraction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 185
6.63 Comparison of the computed relative flow angle and the blade angle at the
inlet of the second stage at the maximum efficiency point: N = 500 RPM.
Note that the azimuthal range: ψ = 240 - 260◦ corresponds to the region of
maximum power extraction. . . . . . . . . . . . . . . . . . . . . . . . . . . . 185
xxi
6.64 Comparison of torque production in the impeller between the improved and
the existing impeller with the new nozzle at the maximum efficiency points.
Note that the original impeller has Nb = 20 and β1b = 30◦ , whereas the new
impeller has Nb = 35 and β1b = 39◦ . For the purpose of comparison, the data
are computed at each blade position and normalized with the torque for 35
blades. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 186
6.65 Azimuthal variation of power production in the impeller of the improved de-
sign at different speeds. The impeller speed 500 RPM corresponds to the
maximum efficiency point (91%). . . . . . . . . . . . . . . . . . . . . . . . . 187
6.66 Contours of the magnitude of mean water velocity at N = 500 RPM corre-
sponding to the maximum efficiency (91%). . . . . . . . . . . . . . . . . . . . 188
6.67 Streamlines superimposed into the water velocity contour illustrating the main
flow path in the impeller. Note the increase in the inlet flow angle at the second
stage with the increase in impeller speed. . . . . . . . . . . . . . . . . . . . . 188
A.1 Comparison between the analytical model [Equations (A.2) and (A.3)] and
the experimental results for the efficiency η vs speed ratio U1 /u1 at α1 = 16◦
[Q = 105 lps and H = 10 m]. The experimental data are plotted for the 7
kW turbine (Dakers and Martin, 1982). . . . . . . . . . . . . . . . . . . . . . 203
C.1 Results of grid convergence test of the 7 kW turbine at maximum efficiency

[Q = 105 lps, H = 10 m and N = 450 RPM]. . . . . . . . . . . . . . . . . . 207
xxii
C.2 Results of grid convergence test of the 0.53 kW turbine at maximum efficiency
[Q = 46 lps, H = 1.337 m and N = 199.1 RPM]. . . . . . . . . . . . . . . . 208
xxiii
List of Abbreviations, Symbols, and
Nomenclature
Symbols
D rate of diffusion of turbulence kinetic energy

D1 impeller outer diameter
D2 impeller inner diameter
h0 nozzle throat
H operating head
k turbulent kinetic energy
K kinetic energy
L characteristic length
N impeller speed (RPM)
U mean-velocity or impeller peripheral velocity
U0 Tangential velocity at the nozzle throat
ur radial velocity of the flow
uθ tangential velocity of the flow
t time
u Cartesian x−wise velocity component
xxiv
v Cartesian y−wise velocity component
w Cartesian z−wise velocity component
Sij strain tensor
pv water vapor pressure
P rate of production of turbulence kinetic energy
Q flow rate
Re Reynolds number
R1 impeller outer radius
R2 impeller inner radius
u0 absolute or computed total inlet velocity
ut theoretical total inlet velocity
U impeller peripheral velocity
W relative velocity or impeller and nozzle width
Greek Symbols
ε rate of dissipation of turbulence kinetic energy

α angle of attack
αa volume fraction of air
αw volume fraction of water
β reduced frequency
β1b outer blade angle
β2b inner blade angle
β1 inlet flow angle
β2 outlet flow angle
β2i inlet flow angle at the second stage
xxv
ρ density
µ dynamic viscosity
µa dynamic viscosity of air
µa dynamic viscosity of air
µw dynamic viscosity of water
µef f effective viscosity
ν kinematic viscosity
ω angular velocity
η efficiency
γ unit gravity force or gravity force per unit volume
σi cavitation inception number
ψ Azimuthal angle
θs nozzle entry arc angle
Other Symbols
∇ gradient operator
Abbreviations
2D Two Dimensional
3D Three Dimensional
CF L Courant-Friedrichs-Lewy Number
CV Control Volume
DES Detached Eddy Simulation
DN S Direct Numerical Simulation
xxvi
LES Large Eddy Simulation
RAN S Reynolds Averaged Navier Stokes
RM S Root-Mean-Square
U RAN S Unsteady Reynolds Averaged Navier Stokes
xxvii
Chapter 1
Introduction
The next-generation turbine designs for small-scale hydropower systems seek higher efficiency
and lower design and manufacturing costs. This thesis studies the fluid dynamic mechanisms
in crossflow turbines for improving their maximum efficiency, motivated by their inherent
simplicity in design and manufacturing at low cost and the potential for performance im-
provement using recent advances in computational fluid dynamic simulations. Crossflow
turbines are typically used for small-scale, low-head remote power systems. Current cross-
flow turbines are reported to achieve maximum efficiency upto 85%, but due to lack of proper
design theory, most turbines in practice only achieve about 80% maximum efficiency. The
computational study presented in this thesis shows that crossflow turbines can achieve more
than 90% efficiency and maintain high efficiency at part-flow operations if the inlet flow is
regulated using a circular slider type control device at the inlet periphery of the impeller.
1.1 Background
Energy is at the heart of modern society, playing critical roles from human civilization
to supporting modern technological developments, such as manufacturing, transportation,
food, medicine and communication. Amongst different forms of energy, electricity is the
most advanced. Electricity is possibly the most remarkable discovery in the history of hu-
1
man civilization, which has enabled tremendous transformations in the pace of technological
developments and human civilization. Today almost every advanced technological product
or device, which constitutes an essential part of modern society, requires electricity to per-
form its intended function. Therefore, access to electricity has become an essential part
of modern-life as well as an enabler to accelerate technological advancements and human
civilization.
At the present, however, not everyone in the world is equally fortunate to enjoy the
benefits of electricity. The International Energy Agency (IEA) estimates that about 1.2
billion people, particularly those living in the developing nations, lack access to electricity,
and about 80% of people live in rural areas (IEA, 2013). The main obstacle is that expanding
electricity networks to those places is expensive as well as technically challenging, further
exacerbated by limited technical and politico-economic capabilities of those nations. Lack
of energy has hindered education, health, food, transportation and communication systems.
Tackling this wicked problem requires partly politico-economic solutions, but technological
advancements in clean and affordable energy generation and supply greatly accelerate the
efforts. For this, innovative designs of renewable and climate-friendly energy technologies and
decentralized rural electrification methods to deploy a large number of systems are thought
to be the key elements in meeting the United Nations’ sustainable development goals 1 .
In developed countries, such as Europe and Americas, nearly all people have access to
electricity. Power is generated mostly from fossil fuels. Amidst global competition in con-
suming natural resources 2 , global issues, such as climate change, environmental degradation,
and declining fossil fuel resources, have become critical. Therefore clean energy systems are
needed to meet these demands and tackle the global issues of environmental pollution and
climate change caused by fossil fuel consumption. As a result, the demand for renewable
and cost-effective sustainable energy systems has grown rapidly worldwide. Amongst differ-
1
United Nations has set a number of sustainable development goals to address the issues of energy access
in the developing world (UNDP, 2015).
2
In efforts to support increasing energy demand due to population growth and for improving the quality
of modern life
2
ent sources of renewable and environmental-friendly technologies, hydropower is the oldest
and the cheapest and reliable energy technology, and thus it is expected to continue as an
important part of the future sustainable energy systems.
Sustainable development 3 requires harnessing renewable energy resources cost effectively
to meet the demand of clean energy in both developing and developed nations. Amongst
different renewable energy technologies, small-scale hydropower systems, ranging from a
few kWs to 200 kWs, offer economical and reliable means to generate electricity in both
developing and developed nations, where small streams of water are locally available in many
areas. The idea behind the small-scale hydropower is technically simple. These systems do
not require a dam to impound the water, and thus have little impact on the ecology. Water is
diverted from a small river to a canal. A surge tank or a small reservoir is built at the end of
the canal. From the reservoir, the penstock pipe carries the water to the power house, where
it turns a turbine that drives an electric generator, producing electricity. The typical small-
hydro systems for remote power work under the head of a few meters to 100 m with flow rates
upto 200 litres per second. Small-hydro systems that generate power in the range of 10 kW
- 200 kW are most popular. Depending on the capacity of the systems, the electricity can
be used for lighting, supporting water, health, education and communication systems, and
running small industries. Small hydropower systems, which have been used for a long time,
particularly in the rural villages of the developing nations, can provide the best solutions
to the sustainable development goals by providing electricity to support the local economy,
education, health, food and communications systems. To ensure their cost competitiveness
for deploying a large number of systems in the future, they require further technological
advancements, primarily the cost reductions in design and manufacturing and improvements
in efficiency of turbines. Within this broader context of sustainable development, this thesis
seeks to contribute toward the design of high-efficiency crossflow turbines, which are most
popularly used in small-scale hydropower systems.
3
A development framework that works well for many future generations without harming the environment,
human health and economy (UNDP, 2015).
3
1.2 Types of Turbines and Design Requirements
The turbines used in hydropower can be broadly classified as impulse and reaction turbines
based on the degree of reaction (Dixon and Hall, 2013). Degree of reaction is the relative
amount of pressure drop in the rotor and the nozzle, and is defined as the ratio of static
pressure drop in the rotor to the static pressure drop in the stator or nozzle plus the rotor.
Impulse turbines, such as Pelton, work on the principle of impulse action of high jet velocity
impinging on the impeller blades. All the available energy of the flow is converted into kinetic
energy by the nozzle at atmospheric pressure before the flow passes through the turbine
blades. Angular momentum is extracted by the blades mainly due to the dynamic pressure
difference between the two surfaces of the blade, while the static pressure difference across
the blade surface is atmospheric. Thus the impulse turbines have zero reaction. In contrast
to this, the reaction turbines, such as Francis, Kaplan and Propeller, work on the principle of
reaction forces developed across the blade surfaces. Angular momentum is extracted mainly
due to the pressure drop across both the stationary guide vanes and the impeller blades. The
pressure drop occurs both in the nozzle and the rotor (Shepherd, 1956). Thus the reaction of
these turbines is non-zero. An alternative and widely used classification, from the viewpoint
of selection of turbines and their design, can be made using the concept of specific speed Nsp
given as (Shepherd, 1956):
√
N Q
Nsp = (1.1)
(H)3/4
Equation (1.1) shows that the specific speed classifies turbines based on the operating vari-
ables: the flow rate Q, the head H, and the operating speed N . A high head turbine, such
as Pelton, has a lower specific speed. The low-to-medium head turbine, such as Francis,
has a medium specific speed. Similarly, a low head turbine, such as Propeller or axial flow
turbine, has a very high specific speed and can handle very high flow rates. Crossflow turbine
falls under low-to-medium head turbine with high specific speed. Figure 1.1 shows a typical
4
Figure 1.1: Classification of various turbines based on head H and flow rate Q illustrating
their application range (Benzon et al., 2016).
classification of various turbines based on head H and flow rate Q. The typical efficiencies
of those turbines are shown in Figure 1.2.
In the selection of turbines for small-scale hydropower systems, the maximum efficiency
4
, the simplicity in design and manufacturing, and the cost are tightly coupled. As shown in
Figure 1.1, the crossflow turbine is the most suitable turbine for low-to-medium head small
systems. The simplicity in design and manufacturing at low cost is the most prominent
feature amongst its counterparts. However, as shown in Figure 1.2, its maximum efficiency
tends to be in the range 70 - 86%, which is lower than that of more commonly used advanced
turbines, such as Pelton, Francis, and Kaplan, which have typical maximum efficiencies above
90% (Dixon and Hall, 2013; Sinagra et al., 2014; Elbatran et al., 2015). It is also known
that compared to other turbines, crossflow turbines have relatively flat efficiency curves over
a wide range of operating conditions as shown in Figure 1.2. The most efficient impulse
turbine, the Pelton wheel, normally operates above 50 m, whereas the reaction turbines like
4
efficiency is defined as the ratio of shaft power to total input power at the turbine inlet. Losses in the
penstock pipe are not taken into account.
5
Figure 1.2: Comparison of typical efficiencies of different turbines (Sinagra et al., 2014). Note
that the efficiency is the hydrodynamic efficiency, i.e. the efficiency of energy extraction by
the blades, not the efficiency of conversion into electrical energy.
Francis and Kaplan can operate at low heads, but usually require high flow rates to operate
at high efficiency. Use of these turbines in low-head systems requires the design of scaled-
versions, which tend to have lower efficiency than the full-scale versions. In addition, they are
highly complex to design and manufacture locally and are also relatively expensive. Similarly,
Turgo turbines, which require heads of about 50 m, is not as efficient as Pelton and Francis
turbines, and has efficiency below 90% (Benzon et al., 2016). Moreover, their blade design
and manufacturing is also complex (Benzon et al., 2016). Therefore, improving the maximum
efficiency would make crossflow turbines an ideal choice for small-scale applications. This is
the fundamental motivation underlying the work of this thesis.
6
Chapter 2
Thesis Objectives
2.1 Motivation and Background
The turbine is the most important component of a hydropower system. The turbine affects
not only its own performance, but also the performance of the entire system. Crossflow
turbines are widely adopted in small-scale, low-to-medium head hydropower systems. The
well-known and highly desirable advantages of crossflow turbine over its counterparts, such as
Pelton and Francis, are the simplicity in design and manufacturing, low cost, and relatively
flat efficiency characteristics. Moreover, the turbine is mechanically rugged and reliable.
Unlike complex design shapes, such as the blades of Pelton and Francis turbines, the most
prominent design feature is the use of circular-section blades, which can be easily designed
and rolled from thin steel sheet using simple machines at low cost. This is crucial because
it is the blade that presents many challenges to the design and manufacturing of high-
efficiency turbines at low cost. The circular section blades are arranged radially around
the axis of rotation and the blades are fixed to two circular discs at the two ends. Despite
these advantages (design simplicity, local manufacturability, affordability, reliability, etc), a
well-known major issue is that crossflow turbines suffer from a lower maximum efficiency (70
- 86%) compared to more advanced turbines, such as Pelton and Francis, which can easily
7
achieve efficiencies more than 90% (Dixon and Hall, 2013; Sinagra et al., 2014; Elbatran
et al., 2015). The flow mechanisms in those turbines have been extensively investigated and
well-understood, and the achievable performance gains through advanced design studies can
be assumed only marginal. In contrast to this, less research has been conducted on crossflow
turbines to understand the fundamental flow mechanisms underlying the power extraction,
and the literature review indicates that lower maximum efficiency is due to the lack of
fundamental research and not an inherent limitation of the design. The fluid dynamics of the
flow in a crossflow turbine is rather complex, and has not been well-studied in the past from
the viewpoint of improving the design. This thesis aims to improve the efficiency of crossflow
turbines by improving our understanding of the underlying flow physics, identifying the
dominant performance limiting flow mechanisms, and developing a rigorous design strategy
to improve efficiency.
Crossflow turbines consist of two main components as illustrated in Figure 2.1: a nozzle
to control the flow entering the impeller blades and an impeller to extract the power from
the flow. The impeller is open to the atmosphere. The impeller blades extract the power
from the flow by creating a change in angular momentum of the flow passing through the
impeller. The energy exchange is based on the kinetic energy of the water that enters and
leaves the impeller at atmospheric pressure. The high velocity water enters the impeller with
a significant circumferential velocity, traverses the central air-space almost diametrically, and
then exits the impeller with reference to the axis of rotation. The flow first passes through
the first stage blades, then crosses the central region of the impeller, and passes through
the second stage rotating blades before exiting the impeller at atmospheric pressure. Due to
this unique flow characteristic and its operation at atmospheric pressure, crossflow turbines
are considered as an impulse radial turbine (Shepherd, 1956). Many crossflow turbines have
a guide vane in the nozzle, which helps to maintain high velocity at a suitable flow angle
during part-flow operations. The smaller models are usually designed without guide vanes.
8
Figure 2.1: Schematic illustration of the basic design features of a crossflow turbine.
The key design features are illustrated in Figure 2.1 with the exception of the flow control
device which is usually a guide vane.
In general, there are two methods for the design of modern turbines, direct design and
inverse design (Yang, 1991; Logan Jr, 2003). A standard turbine design process follows the
direct design method, which involves two main steps (Lakshminarayana, 1996; Logan Jr,
2003). The first is the preliminary design, which consists of establishing the turbine con-
figuration and the calculation of design parameters (e.g. radius of blades, inner to outer
diameter ratio, outer and inner blade angles, number of blades etc.) based on experience. In
conventional designs, the calculation of design parameters is mostly based on previous ex-
perimental studies, which provide an empirical understanding of factors that are important
9
to turbine design. The author’s experience, however, with small-scale hydro turbine manu-
facture in Nepal is that no account is taken of the actual flow behaviour in the design. The
second is the detailed design phase, which involves detailed investigations of the specific flow
physics (e.g. flow separation on blades) or guiding the design improvement via high-fidelity
computations of Navier-Stokes equations, such as Reynolds-Averaged Navier-Stokes (RANS)
simulations or Large Eddy Simulations (LES). In this design phase, the entire flow field is
computed and the flow field information is synthesized to examine the loss mechanisms as
an aid for improving the performance. Numerical simulations provide deeper understand-
ing of the flow mechanisms that govern performance. This two-step design process is then
iteratively repeated until an acceptable design is obtained. The main obstacle here is the
requirement of a large number of simulations and evaluations of the flow field data, which
demand for high-performance computers, long computational time, and efforts on flow field
analysis. In the inverse design method, the objective function (e.g. pressure distributions on
the blades) is prescribed and an optimum blade geometry is then computed as part of the
design solution (Yang, 1991). Some limited inverse design techniques using simplified flow
models have been developed for the blade design of radial inflow turbines (Yang, 1991), but
the inverse design using high-fidelity simulations are currently computationally expensive to
be useful for practical design purpose. In this thesis, the direct design approach is followed
using RANS simulations to approximately compute the flow field in the turbine, which helps
in understanding the underlying fluid dynamic design problem as an aid for improving the
efficiency.
2.2 Problem Statement
A longstanding problem in the design of crossflow turbines is the improvement in efficiency. A

number of challenges exist in the design and analysis of crossflow turbines because crossflow
turbines manifest a unique flow feature, in which a high velocity flow from the rectangular
10
nozzle passes through the rotating blades twice before exiting the impeller. To the best
knowledge of the author, there is no satisfactory method available for the design of the
nozzle, and its importance to the overall efficiency is not understood. It is noted that power
is extracted in two stages, and the role and significance of the second stage energy extraction
is not well-understood. More precisely, whether the second stage merely acts to extract
whatever energy is missed in the first stage or there exists a certain flow condition between
the stages, which optimizes power extraction from both stages. Knowledge of relative flow
angles at the inlet of the impeller, exit of the first stage, and the inlet of the second stage is
important because the flow coming out of the first stage can negatively affect the performance
of the second stage. This has not been studied previously. It is thus necessary to characterize
the fundamental flow patterns in the turbine that influence the performance behaviour. This
is the key design problem for any attempt to improve the maximum efficiency. This represents
a complex design problem because the flow field is complicated due to three-dimensionality
and turbulence, impeller rotation and multiphase flow (two or three-phase) with free-surface
effects (water and air). Even if it is feasible to compute this flow field with high accuracy, the
fine details of the flow field are not helpful in tackling this problem. It is therefore necessary
to identify the major aspects of the flow and turbine geometry that affect efficiency and,
hopefully, ignore those that do not. Examples of ignored features are the surface condition
of the nozzle and blades, the impeller shaft, and any leakage flow from the nozzle that does
not enter the impeller. Approximate computations of average flow properties are much more
useful to identify what is a dominant flow and what is not for the efficiency. The challenge is
to develop a design which accurately captures the relevant dynamics of the flow, and more
importantly the interactions between the two. Consideration of fluid dynamic effects is vital
at the design stage to determine what combinations of design parameters would improve the
maximum efficiency.
Although improving the maximum efficiency is the primary objective, the multitude of
possible parameters that influence the efficiency and the complexity of the computational
11
fluid dynamic simulations means that typical numerical optimization is not possible. In-
stead, this thesis will use previous studies and a new analysis of nozzle flow to identify
the important parameters prior to undertaking a design which has the goal of exceeding
90% efficiency. More precisely, the mathematical formulation and solution of the design
optimization problem are either highly complex or computationally prohibitive as it incurs
repeated analyses of a high-fidelity model in different turbine configurations. The mathe-
matical formulation is complex because the flow field in turbines is highly turbulent and
three-dimensional with rotation. Efficiency, which is multi-dimensional in character, de-
1
pends on so many design parameters that it is not possible to formulate a well-posed
mathematical optimization problem. Furthermore, even if it were possible to formulate such
a mathematical problem approximately, the solution of this mathematical formulation re-
quires computations of Navier-Stokes equations, which govern the fluid flow in the turbine.
Computing the Navier-Stokes equations, even for the lower-fidelity models such as Reynolds-
Averaged Navier-Stokes equations, is too expensive to be feasible for such an optimization
which may require the evaluation of the objective function many thousands of times. This
leads us to an approach of design improvement, in which relevant flow physics and vari-
ous design concepts are studied by using Reynolds-Averaged Navier-Stokes computations
2
. Computational analysis of experimentally studied turbines available in the literature can
effectively lead to optimum design. The fundamental idea behind the design improvement
for the efficiency is the establishment of fundamental principles such that all the head should
be turned into kinetic energy at the impeller inlet and to develop a better understanding of
the dominant flow features in the impeller, which simplifies this mathematical complexity.
The basic principle of converting the entire head into kinetic energy is used routinely for
Pelton turbine design, which achieves high efficiency, but this principle has not yet been
1
a well-posed mathematical problem is the one which has a unique solution(s) and the function is con-
tinuous for the specified initial and boundary conditions. The optimization yields a unique global solution,
whereas design improvement most likely yields only the local solutions, but guides towards the global optimal
solution.
2
see Chapter 3 for the computational method
12
applied to crossflow turbines. Previous studies that are reviewed in Section 2.5 have mainly
focused on experimental studies of turbine efficiency as a function of different geometrical
parameters, however, they do not provide fundamental descriptions of the underlying flow
physics, the links between the geometrical parameters and the dominant flow features, and
their influence on turbine performance. Therefore, a design study that takes into account
the links between the geometrical parameters and the flow field characteristics is needed. In
this thesis, RANS simulations are used to tackle this design problem.
2.3 Research Questions and Goals
The overall goal of this thesis is to improve the efficiency of crossflow turbine through the im-
provement of nozzle and impeller designs and their matching to achieve maximum efficiency.
To improve the efficiency, this thesis seeks to answer the following research questions:
1. Nozzle Design: How do different nozzle design parameters affect the conversion of
head to kinetic energy, the distribution of angular momentum flux and the flow angle
at the impeller inlet? What are the key fluid dynamic criteria that set the nozzle
performance?
2. Impeller Design: How do different inlet flow conditions affect the impeller perfor-
mance? How do different impeller design parameters and matching the nozzle and
impeller designs affect the internal flow and efficiency? A substantial portion of the
total power is produced at the first stage, however, the performance of the second stage
to extract the remaining power is also important. The design problem is how to design
an impeller to match a good nozzle to extract power most efficiently from both stages.
This occurs if all the head is approximately converted into kinetic energy, the flow is
directed at a suitable angle to the impeller, and the flow exiting the first stage aligns
to the second stage blades. It is noted that the blade inlet angle is set by the inlet flow
angle so the only remaining parameter is the blade outlet angle as the study follows
13
the crossflow turbine design tradition and uses circular section blades. Thus the key to
improving the efficiency is to extract as much power as possible from the first stage and
direct the flow to the second stage in such a way that the flow aligns to the rotating
blades. What are the flow mechanisms that limit the performance of the impeller?
2.4 Flow Field Characteristics in Crossflow Turbines
Crossflow turbines are a two-stage radial turbine that operate at atmospheric condition.
Crossflow turbines manifest a unique flow feature, in which the flow from the nozzle passes
twice radially through the impeller blades before exiting the impeller as indicated in Figure
2.1. Thus power is extracted at two stages. This implies that the flows at two stages are cou-
pled. The flow field is three-dimensional with turbulence 3 and multiphase flow, consisting of
either two-phase (water and air) or three-phase (water, water-vapor, and air if cavitation is
present) with free-surface effects on rotating walls between water and air. Free-surface effects
arise if the inertia force is significantly higher than the surface tension force, which is char-
acterized by the Weber number = ρU 2 L/γ (White and Corfield, 2006). Therefore, crossflow
turbines represent one of the most complicated hydroturbines with respect to computational
simulations. These complex flows can only be investigated approximately through numerical
simulations. The computational methods for simulating the flow in crossflow turbines are
described in the next chapter.
In turbomachinery flows, the effects of solid boundary walls or the geometry on velocity
and flow deflection must be examined carefully in order to improve the performance (Greitzer
et al., 2007). Substantial pressure gradients exist and rotation is a major factor influencing
the flow behaviour (Baskharone, 2006; Greitzer et al., 2007). Losses can result from physical
phenomena such as three-dimensional boundary layers, interaction between rotating blades
and end-wall boundary layers, blade-tip clearance vortices, leakage in the nozzle, blade trail-
3
turbulence is a phenomenon resulting from irregular and significant variation of velocity field in time
and space. An important characteristics of turbulence is its ability to transport and mix fluid much more
effectively than laminar flow.
14
ing edge vortices and wakes and mixing. This thesis will not investigate these losses as they
have only minor impacts in the overall performance. Since the flow field is directly linked to
flow passage geometry (Greitzer et al., 2007), complex geometrical features, such as nozzle
geometry, curvature and shape of the flow passage, blade spacings, blade inlet and exit an-
gles etc complicate the flow structure in the turbine, such as separation on the blades, and
govern its performance. In this thesis, only the dominant flow structures are investigated
rather than on the fine details of the above mentioned flow phenomena without significant
loss of information. The dominant features will be identified from previous experimental
studies and the new analysis of the nozzle presented in Section 5.2. The features that will
be ignored include viscous losses, blade tip-clearance leakage etc.
2.5 Literature Review
Crossflow turbines were conceived by Michell (1904), and since then many researchers have
attempted to improve their maximum efficiency. Improving the maximum efficiency has
been an important objective of previous investigations. In this section, a brief review of the
relevant literature, which provides motivation for this thesis, is presented.
The previous studies can be divided in two groups: experimental and numerical. Mac-
more and Merryfield (1949) developed analytical equations for the turbine efficiency, but
their usefulness in turbine design is highly limited. Examples are provided in Appendix A,
and are thus omitted from this review. The bulk of the literature deals with the experimental
measurements of turbine performance by varying the design parameters. At least some of
these studies provide an empirical understanding of the design parameters that are important
to the turbine efficiency. No information is available from experimental measurements about
the internal flow physics and its impact on efficiency. In other words, the only experimental
information available is measurements of turbine efficiency as a function of impeller speed
for different combinations of flow rate and head. Similarly, the numerical studies are mainly
15
focused on performance prediction of specific turbines using RANS computations, rather
than on the investigation of internal flow physics that are crucial for improving the maxi-
mum efficiency. The main conclusions presented in the existing literature are summarized
as follows.
2.5.1 Experimental Studies
The geometrical parameters for turbine design are well known, but their effects on turbine
performance are not yet clearly understood, and thus are the subjects of many experimental
investigations in the past. Most previous studies (Macmore and Merryfield, 1949; Durali,
1976; Nakase et al., 1982; Dakers and Martin, 1982; Khosrowpanah, 1984; Khosrowpanah
et al., 1988; Fiuzat and Akerkar, 1989; Desai, 1994; Totapally and Aziz, 1994) on efficiency
improvement have entirely relied on experimental measurements of efficiency, which empha-
size the influence of geometric parameters on the turbine efficiency while de-emphasizing the
importance of underlying flow physics in understanding the design problem. The results of
these studies are summarized in Table 2.1. Most of these studies have reported the max-
imum efficiency of less than 82%, which is significantly lower than the typical maximum
efficiency of more than 90% for the Pelton, Francis and Kaplan turbines. There are only
three studies that have reported the maximum efficiency in the range 88 - 90%. Based on
a limited experimental study, Fiuzat and Akerkar (1989) found a maximum efficiency of
89%. Similarly, Desai (1994) was able to develop a small-scale turbine of capacity 0.53 kW
with the maximum efficiency of 88% based on an extensive experimental testing of various
designs. In continuation of his work, Totapally and Aziz (1994) achieved a remarkable maxi-
mum efficiency of 90% just by increasing the number of blades from 30 to 35. These turbines
represent the most efficient crossflow turbines reported in the literature so far; no larger-
scale turbines of 88 - 90% efficiency have been reported in the literature. Fortunately, the
references Desai (1994) and Totapally and Aziz (1994) also give sufficient information about
the turbine geometry, operating conditions and test results for turbine efficiency to allow
16
computational simulations and validation, and so this test case is one of the two used in this
thesis. More importantly, the underlying flow physics in these turbines has not been studied
which is crucial to guide the design of more efficient turbines. The experimental works of
Desai (1994) and Totapally and Aziz (1994) thus serve as an impetus for the computational
study in this thesis for improving an understanding of the dominant flow features and the
design problem, which is crucial in improving the turbine efficiency.
During 1916 - 1918, Donat Banki conducted extensive experimental investigations on
the crossflow turbine and developed a basic theory for its design and operation (Fiuzat
and Akerkar, 1991). Macmore and Merryfield (1949) extended Banki’s work. Using a two-
dimensional inviscid flow analysis, they formulated a simple analytical equation for the max-
imum efficiency. The key conclusion of their study to achieve higher efficiency was to design
a turbine with a small angle of attack α at the inlet of the impeller. They defined α as the
angle between the absolute velocity and the tangent at the periphery of the impeller inlet as
shown in Figure 2.1. Following Macmore and Merryfield (1949), many researchers (Durali,
1976; Nakase et al., 1982; Dakers and Martin, 1982; Khosrowpanah, 1984; Khosrowpanah
et al., 1988; Fiuzat and Akerkar, 1989; Desai, 1994; Totapally and Aziz, 1994) measured the
effects of different geometrical parameters on turbine efficiency. A brief summary of previous
studies is presented here.
The influence of the angle of attack α on the efficiency was studied by Nakase et al.
(1982), Khosrowpanah (1984), Fiuzat and Akerkar (1989), Desai (1994), Totapally and Aziz
(1994) and others. Khosrowpanah (1984), Nakase et al. (1982) and others found that α
= 16◦ gave the maximum efficiency. They found the maximum efficiency of 80 and 82%
respectively. Fiuzat and Akerkar (1989) found the maximum efficiency of 89% at α = 24◦ ,
and Desai (1994) reported the maximum efficiency of 88% at α = 22◦ . Similarly, Totapally
and Aziz (1994) found the maximum efficiency of 90% at α = 22◦ . However, systematic
design criteria for obtaining the desired values of α are still lacking because these studies
have assumed the inclination of the lower nozzle wall at the impeller inlet as α as shown in
17
Authors α β1 β2 D2 /D1 Nb θs η
[deg] [deg] [deg] [-] [-] [deg] [%]
Michell (1904) 16 38 90 0.68 20 90 92 (Theory)
Macmore and Merryfield (1949) 16 30 90 0.66 20 - 68
Varga (1959) 16 39 - 0.66 30 - 77
Durali (1976) 16 30 90 0.68 24 - 76
Johnson and White (1982) 16 39 - 0.68 18 60 80
Nakase et al. (1982) 15 39 - 0.68 26 90 82
Dakers and Martin (1982) 22 30 90 0.67 20 69 69
Durgin and Fay (1984) 16 39 - 0.68 20 63 66
Khosrowpanah (1984) 16 39 90 0.68 15 58,78,90* 80
Hothersall (1985) 16 - - 0.66 21 - 75
Ott and Chappell (1989) 16 - - 0.68 20 - 79
Fiuzat and Akerkar (1989) 20-24* 39 90 0.68 20 90 89
Desai (1994) 22*-32 39 90 0.60,0.68*,0.75 30 90 88
Totapally and Aziz (1994) 22*-24 39 55*-90 0.68 35 90 90
Table 2.1: Summary of empirical values for the design parameters used in the experimental
studies by the previous researchers. The symbol [*] represents the optimal value at the
maximum efficiency point. For the symbols, refer Figure 2.1.
Figure 2.1. The above condition for maximum efficiency is valid only if the entire impeller
inlet section receives a uniform flow at this value of α. However, these studies did not
measure the actual angle of attack to check whether the assumed values of α were obtained
at the inlet of the impeller or not.
The influence of nozzle rear wall shape was studied by Nakase et al. (1982). He found
that both the circular and logarithmic spiral shapes of the rear wall gave the same maximum
efficiency. However, it is not known how the shape of the nozzle rear wall affects the inlet
flow conditions and the impeller performance.
The influence of nozzle entry arc angle θs on the efficiency was studied by Nakase et al.
(1982), Khosrowpanah (1984), Fiuzat and Akerkar (1989) and others. They found that θs =
90◦ gave the maximum efficiency. Fiuzat and Akerkar (1989) and Totapally and Aziz (1994)
found that vertically oriented nozzles with θs = 90◦ would give the maximum efficiency of
about 90%.
The influence of the outer blade angle β1b and the inner blade angle β2b have been studied
by several researchers (Desai, 1994; Totapally and Aziz, 1994). These angles determine the
conditions for flow separation on the blades, the impeller performance, and the relative
magnitudes of power production in the first and second stages. The majority of previous
18
studies found the maximum efficiency at β1 = 39◦ , whereas β2 was kept at 90◦ . Desai (1994)
found that β2 = 90◦ gave the maximum efficiency, whereas Totapally and Aziz (1994) found
that β2 = 55◦ gave slightly greater maximum efficiency than β2 = 90◦ .
Studies on the influence of inner diameter to outer diameter ratio D2 /D1 on the efficiency
have determined that D2 /D1 = 0.68 is the optimum value (Macmore and Merryfield, 1949;
Durali, 1976; Nakase et al., 1982; Dakers and Martin, 1982; Khosrowpanah, 1984; Khos-
rowpanah et al., 1988; Fiuzat and Akerkar, 1989; Desai, 1994; Totapally and Aziz, 1994).
Smaller diameter ratio means longer blades, whereas larger diameter ratio means shorter
blades. The diameter ratio impacts the size of the inner air-space of the impeller, where
the water streams from different blade passages from the first stage combine before passing
through the second stage. As such, it influences the contributions of power extraction from
the first and second stages. The influence of this parameter on the internal flow structure
and the impeller performance is not well-known.
The influence of the number of blades Nb on the maximum efficiency was studied in
detail by Nakase et al. (1982), Khosrowpanah (1984), Desai (1994), and Totapally and Aziz
(1994). Khosrowpanah (1984) conducted experiments with Nb = 10, 15 and 20, and found
that impellers with 15 blades were the most efficient, giving about 80% maximum efficiency.
Desai (1994) and Totapally and Aziz (1994) investigated the influence of the number of blades
and found respectively that 30 and 35 are the optimum number of blades, and obtained the
maximum efficiencies of 88% and 90% respectively. However, Fiuzat and Akerkar (1989)
found a maximum efficiency of 89% with 20 blades on a similar size of the impeller used
by Desai (1994) and Totapally and Aziz (1994). The effect of Nb on flow separation on the
blades and efficiency has not been studied systematically. The most important parameters
that influence the selection of Nb are likely to be the operating head, the flow rate, the outer
diameter of the impeller, inner to outer diameter ratio and blade angles. For the design study
of a particular turbine, the researchers varied Nb and observed the changes in efficiency. For
example, Nakase et al. (1982) found that Nb = 26 is the optimum Nb for an impeller with
19
D1 = 315 mm, whereas Desai (1994) and Totapally and Aziz (1994) found Nb = 30 and 35
as the optimum Nb for a similar size of impeller diameter (304 mm). The operating heads
in these cases were 1.54 m and 1.337 m respectively. The maximum efficiencies were 82, 88
and 90% respectively. The influence of Nb on the impeller weight and its consequences on
efficiency has not been studied so far, but the weight of the impeller increases, which may
increase vibration and frictional losses.
The influence of the nozzle design and the operating conditions (head and flow rate) was
studied by Dakers and Martin (1982). For a 7 kW turbine, impeller was designed with the
outer blade angle β1b = 30◦ , the inner blade angle β2b = 90◦ , the inner to outer diameter
ratio D2 /D1 = 0.68 and Nb = 20 for H = 10 m and Q = 105 lps. By changing the nozzle
configurations, e.g. rear wall shape and orientation, and keeping the impeller design same,
they achieved a maximum efficiency of 69%. Extensive testing was carried out at different
H, Q, and N . Their work also gives sufficient information of the turbine geometry, the
operating conditions, and test results to allow a detailed computational simulation, and thus
this test case is one of the two major ones used in this thesis. This turbine will be described
in more detail in Chapter 4.
There are a few studies that attempt to determine the contribution of each stage to
the total power output. Fiuzat and Akerkar (1991) designed a special turbine with a flow
diverter inside the turbine to measure the relative contributions of the two stages to the
total power output. A significant portion of the power was extracted at the second stage.
They concluded that the efficiency can be improved if efforts are concentrated on improving
the performance of the second stage. They found that the contribution of the second stage
is at least 45% for the nozzle entry arc angle θs = 90◦ , and at least 41% for θs = 120◦ . They
argued that the best way to improve turbine efficiency would be to increase the cross flow
in the interior region (refer Figure 2.1), which will increase the performance of both stages.
The above experimental studies did not study the influence of geometric parameters
on internal flow characteristics and their influence on turbine efficiency. Durgin and Fay
20
(1984) performed an experimental study to investigate the overall internal flow pattern using
stroboscopic and photographic techniques. They studied an open-ended, cantilevered runner
to observe the flow and also placed external flow deflector inside the impeller to alter the flow
deflection and measure the contributions of each stage. They used a nozzle of varying entry
arc to study the effects of entry arc angle on the flow and efficiency. They measured the
amount of flow crossing the interior region and the flow trapped in the blade passages after
the first stage, hereafter called trapped flow, by diverting the flow from the interior region,
i.e. not allowing to enter the second stage. However, they could perform the experiments
only at part-flow conditions, but not at design flow rate due to high flow interaction. The
most significant results they obtained was the low-percentage of trapped flow inside the blade
passages. The trapped flow was found to be thrown tangentially, which was found to increase
as the speed increased or as the amount of crossflow decreased. Also as the entry arc angle
was increased from 30 to 80◦ , the amount of trapped flow increased. They obtained the
maximum efficiency of 61% and found only 17% contribution from the second stage to the
total power production. The first stage contributed about 83% to the total output power.
However, their findings were not generalized, and are of limited value in designing high
efficiency turbines in the range of 88 - 90% or above. However, the split between the stages
is interesting, particularly because this is the only such result experimentally determined
after (Fiuzat and Akerkar, 1991). They also observed that a significant amount of trapped
flow did not contribute to power extraction at the second stage. In a crossflow turbine, the
flow is assumed to follow a path through the central region of the impeller in the form of
an ideal, “well-developed single jet” (Macmore and Merryfield, 1949), however, Durgin and
Fay (1984) did not observe this flow pattern in their experiment. They concluded that the
trapped flow caused significant incidence losses at the second stage. Moreover, they observed
that the amount of trapped flow also varied depending upon the impeller speed. They also
4
found that θs impacted the amount of crossflow increased, the efficiency improved. They
4
the amount of flow through the central air-space region in the impeller.
21
modified the existing design to take into account the effects of entrapped flow, which slightly
improved the maximum efficiency to 66%. However, this value is significantly lower than
the experimentally reported maximum efficiency of 88 - 90% by Desai (1994) and Totapally
and Aziz (1994). Therefore, their observations are less likely to be useful in the design
improvement.
2.5.2 Numerical Studies
Although the effects of key geometric parameters on turbine efficiency have been experi-
mentally studied, the current knowledge cannot be readily applied to design high efficiency
turbines. As it is challenging as well as expensive to measure and visualize the flow fields
in the impeller, the alternative method to study the effects of geometric parameters on the
flow field and efficiency is computational simulations. Recent computational studies include
Choi et al. (2008), De Andrade et al. (2011), Sammartano et al. (2013), Sinagra et al. (2014),
Acharya et al. (2015), and Sammartano et al. (2016), however, these studies contain very
little description of the flow physics, and do not provide any description of the design prob-
lem and the methods to address it. These studies have used steady RANS computations
employing k − and shear stress transport (SST) k −ω turbulence models with homogeneous,
two-phase free-surface models for water and air. These studies performed on low-efficiency
turbines are primarily aimed at predicting the turbine performance. Moreover, the flow pat-
terns reported are only illustrative of the general flow field, which is common to all crossflow
turbines, rather than a detailed treatment of their specific characteristics to identify the
reasons for losses, and their influence on the impeller performance. These studies lack an
important design aspect, the consideration of the need for conversion of head into kinetic
energy before the flow passes to the impeller. There is no information on the role of inlet
flow at the impeller inlet on the impeller performance.
Choi et al. (2008) showed that guide vane angle in the nozzle plays a significant role in
improving the turbine performance during part-flow operations, possibly due to improving
22
Guide vane
Flow First Stage
free-stream,
air-space region
Second stage
Figure 2.2: Schematic illustration of the guide vane in the crossflow turbine.
the nozzle flow and maintaining a suitable angle of attack. A schematic illustration of the
guide vane is shown in Figure 2.2. To cope with the part-load fluctuation and reduction in
flow rate, guide vane is needed, but their role on maximum efficiency has not been studied
previously. They also showed that an air-layer in the blade passages, via suction in air vents,
improved the turbine performance, primarily due to the removal of recirculation regions
in the blade passages. However, the experimental studies of Desai (1994) on small-scale
models showed that turbines without guide vanes or such air-layers in the passage achieved
88% efficiency, which is significantly higher than the efficiency Choi et al. (2008) achieved
(80%). Moreover, guide vanes can improve efficiency at part-flow conditions, but not at the
design-point condition. The design-point has the maximum Q at constant H. Similarly re-
circulation regions can be removed by improving flow guidance within the blade passages, not
by providing suction air. Thus the outcomes of these studies are less useful for understanding
the performance limiting flow mechanisms aimed at developing more efficient turbines (e.g.
more than 90% efficiency).
23
De Andrade et al. (2011) studied the internal flow using 3-D steady RANS computations
with a water-air homogeneous flow model with free-surface effects and the k − turbulence
model. They found that about 68% of the power is produced at the first stage and the
remaining at the second stage in a turbine of about 70% maximum efficiency. They char-
acterized the relative flow angles at the inlet and exit of the first stage blades and the inlet
of the second stage. They found a significant difference and variation of the computed flow
angles at the above locations from that of the designed blade angles. However, the authors
did not attempt any design improvement based on the information of the flow field. More
importantly, the reasons for the low efficiency and the ways to improve the efficiency have
not been explored.
Sammartano et al. (2013) presented a two-step design methodology for the optimal design
of crossflow turbines. The first step provides a simple analytical nozzle design equation. The
second step gives a set of empirical values for the impeller design, which can be refined using
steady and unsteady RANS computations. By simulating a 5.2 kW turbine with their opti-
mal design methodology using steady and unsteady RANS computations with the SST k − ω
turbulence model and two-phase homogeneous model with free-surface effects, they reported
5
a maximum hydraulic efficiency of 86%. However, the hydraulic efficiency is greater than
the overall efficiency as used in this thesis, and defined by Equation (4.1), so the maximum
efficiency obtained by the authors must be less than the overall maximum efficiency. As
discussed above, maximum efficiency in the range of 88 - 90% has already been achieved by
Fiuzat and Akerkar (1989), Desai (1994) and Totapally and Aziz (1994), and thus this study
may not provide desirable design attributes for designing higher efficiency turbines. More
importantly, their CFD simulations lack a fundamental analysis of the flow characteristics,
characterization of loss mechanisms, and their influence on the impeller performance.
Acharya et al. (2015) carried out steady RANS computations on a turbine with 63%
maximum efficiency. They used the SST k − ω turbulence model with the homogeneous mul-
5
they defined the maximum efficiency as the ratio of shaft power and the total input power to the impeller
(= power at the impeller inlet - power at the impeller outlet).
24
tiphase model for modeling water and air with free-surface effects. They modified the radius
of the nozzle rear wall of circular shape, adjusted the guide vane opening, and modified the
number of blades from the reference turbine. They showed an improvement in the maximum
efficiency from 63 to 76%. However, their study lacks any explanation of the fundamental
differences in the flow field, and their quantitative effects on the impeller performance. The
study does not provide a detailed analysis on the nozzle performance, inlet flow conditions,
and power extraction in the two stages, which are the basic informations necessary to assess
the turbine performance. Therefore, the results are of limited use for design improvement.
In a recent computational study, Adhikari et al. (2016) have shown that cavitation can
occur in crossflow turbines. Using the steady RANS computations with SST k−ω turbulence
model and homogeneous multiphase model for water and air with free-surface effects, they
studied cavitation inception on a 7 kW turbine. For this particular turbine design, cavitation
started at the inner edges of the second stage blades at and above the impeller speeds
corresponding to the maximum efficiency. This study indicates that cavitation may be an
important consideration in the design of crossflow turbines. No further information about
cavitation in crossflow turbines is available in the literature.
While several parametric studies have been carried out and suggested in the literature,
a rigorous quantitative causal link between the nozzle and impeller geometry and the flow
mechanisms leading to performance losses is missing. More importantly, the performance
of the nozzle and the matching with the impeller have not been investigated in any detail.
This important design aspect needs to be investigated in detail in any attempt on design
improvement. In the numerical simulations, the effects of flow alignment with the blades in
the first and second stages have not been assessed for its role on limiting the second stage
performance. In summary, there is a lack of quantitative traceability of any flow process
or mechanism responsible for performance losses. The quantitative relationships between
the overall flow parameters, i.e. velocity and flow angles, which are the basic information
necessary to assess the impeller performance, are missing in the current literature.
25
2.6 Part-load Operation
In the operation of hydro turbines, part-load means lower Q at constant H because the
latter is practically always constant. It will be shown in Section 6.2.6 that it is not possible
to achieve high efficiency by electronically controlling the generator of a crossflow turbine
at part-load. For controlling part-load operations, a guide vane is generally installed in the
nozzle passage as shown in Figure 2.2. Even by using the guide vanes, better part-load
efficiency can be achieved than any other hydro turbines. However, the problem with using
a guide vane in the nozzle passage is that it may induce extra losses due to blockage of
the flow passage area and viscous dissipation on the walls and may even form trailing edge
wake at the impeller inlet although there are no studies on these issues. An alternative to
guide vane is a slider type flow regulator at the inlet periphery of the impeller as described
in Sinagra et al. (2014). Sinagra et al. (2014) concluded that at constant head, a constant
velocity at the impeller inlet was obtained at all reduced flow rates, which is the key in
keeping the efficiency high at part-load operations. This will be described in Chapter 5 in
more details and investigated in Chapter 6.
2.7 Organization of the Thesis
The remainder of the thesis is set out as follows:

Chapter 3 presents a description of the computational models employed in the flow com-
putations. Chapter 4 describes the validation and performance characterization. The main
focus is on developing an improved understanding of the flow physics in crossflow turbines
and identify key methods to improve the design. Chapter 5 presents an integrated nozzle-
impeller design framework for the computational simulations. A simple analytical model for
nozzle design is presented. For operating the turbines efficiently during part-load conditions,
a slider type control device is described. Chapter 6 expands the analysis of the turbines
under the new design framework developed in Chapter 5, synthesizes the results of Chap-
26
ter 4 to develop a new design strategy for design improvement, and presents the results of
design improvement. It describes the requirements for designing a high-efficiency turbine.
Computational evaluations of different designs are presented, with the aim of providing es-
sential information on how to improve the design. Finally, Chapter 7 summarizes the major
findings and recommendations for designing high-efficiency crossflow turbines of more than
90% efficiency and outlines future works.
27
Chapter 3
Computational Method
In order to compute the three-dimensional flow field in a crossflow turbine, one must adopt
some computational models. This chapter describes the computational models for computing
and analyzing the flow in the turbine in terms of computational mesh, boundary conditions,
multiphase and turbulence models, and validation and uncertainty. The flow computations
were performed using the commercial computational fluid dynamics code ANSYS CFX (AN-
SYS, 2016). The capability of the computational model using steady and unsteady RANS
computations to capture the flow in crossflow turbines was evaluated by following a series
of steps: 1) assessed the predictive capability of steady and unsteady RANS models with
turbulence and multiphase models for the turbine power output; error and uncertainty in
the computed results were quantified by comparing against the experimental results, 2) as-
sessed the sensitivity of the changes in turbine power output in response to the change in
grid resolution, 3) assessed the validity and consistency of the results by comparing them
against experimental results for a range of turbine designs and operating conditions, and
4) assessed the effects of unsteadiness on power output due to nozzle-impeller interaction
by comparing unsteady RANS computations with the steady RANS computations. A more
detailed discussion on URANS simulations is given in Section 4.2.2.
28
3.1 Governing Equations
An incompressible isothermal flow through a turbomachine is fully described by the conti-

nuity and momentum equations, which are called the Navier-Stokes equations. The Navier-
Stokes equations for an incompressible isothermal flow are written as:
∂ui
=0 (3.1)
∂xi
∂ui ∂ui 1 ∂p ∂ 2 ui
+ uj =− +υ (3.2)
∂t ∂xj ρ ∂xi ∂xj ∂xj
where ui is the velocity, p is the pressure, ρ is the density of fluid and υ is the kinematic
viscosity of the fluid. It is noted that in an isothermal flow, there is no change in fluid tem-
perature due to heat addition or removal from the fluid as a result of flow processes occurring
in the turbine, e.g. due to the power extraction in the impeller or viscous dissipation. There
are no general solutions for these non-linear equations, and one has to solve these equations
numerically for the vast majority of flows with the specifications of appropriate boundary and
initial conditions. Since solving the Navier-Stokes equations is computationally expensive
for high Reynolds number flows in complex geometries, such as in a crossflow turbine, RANS
equations are generally solved to determine the mean velocity field. Computing the flows in
crossflow turbines is challenging due to the complexity of the flow physics involved. The flow
involves either a two-phase flow consisting of water and air or a three-phase cavitating flow
consisting of water, water-vapor and air with free-surface effects on rotating walls (Adhikari
et al., 2016). The mean values of the flow variables are usually the quantities of interest,
especially in predicting the performance of turbomachines. RANS equations are obtained by
time-averaging the Navier-Stokes equations for the mean values of the flow variables over a
29
sufficiently long period compared to the frequencies of turbulent fluctuations, and are written
as:
∂Ui
=0 (3.3)
∂xi

∂U ∂Ui 1 ∂P ∂ ∂Uj ∂Ui 0 0
+ Uj =− + υ( + ) − ui uj ) (3.4)
∂t ∂xj ρ ∂xi ∂xj ∂xi ∂xj
where U = time-averaged mean velocity, ui 0 = fluctuating velocity due to turbulence and

0 0
−ρui uj = Reynolds shear stress. An unsteady turbulent flow can often be considered as
consisting of two parts: a slowly varying mean flow plus a rapidly fluctuating component
related to turbulence as shown in Equation (3.4). In Equation (3.4), the Reynolds shear
0 0
stress (−ρui uj ) is modelled with a turbulent viscosity model, which will be discussed in a
later section.
3.2 Multiphase Flow
Crossflow turbine differs from other turbomachines (except Pelton and Turgo) by having
free-surface multiphase flow. Thus the simulation of the flow is complex and poses many
modeling challenges because of the existence of free-surface effects between water and air on
the rotating impeller blades, and the interaction of the two-phase flow with the turbulent
boundary layer. There are many multiphase models, but none is a generalized and robust
model (Brennen, 2005). Accurate simulation of turbulent free-surface flows around impeller
blades has a central role in the optimal design of such machines.
In a multiphase flow, the fluids are not mixed on a microscopic scale, but they are mixed
on a macroscopic scale with a distinct resolvable interface between the fluids (Brennen,
2005) (e.g. Figure 3.1). The free-surface multiphase flow is constituted by water and air
separated by a thin interface, in which the fluid or the water ends at an open or free-surface
30
Figure 3.1: A typical contour of water and air volume fractions in a crossflow turbine com-
puted using homogeneous multiphase free-surface model. The red and blue regions show re-
spectively water and air in the computational domain, illustrating a discernible free-surface
interface between them. The picture is taken from one of the snapshots of the author’s
unsteady RANS simulations.
exposed to an atmosphere of ambient air. The crossflow turbine operates with ambient air
pressure on the free-surface (White and Corfield, 2006). In a more complex situation, the
atmosphere exerts not only pressure but also shear, heat flux and mass flux at the water
surface (water-vapor interface) (White and Corfield, 2006). This situation arises in the case of
cavitation in the turbine. The two required conditions for the former free-surface, illustrated
schematically in Figure 3.2, are that the fluid particles at the surface must remain attached
(kinematic condition) and the water and the atmospheric air pressure must balance except for
surface-tension effects (White and Corfield, 2006). For modeling the free-surface in a RANS
simulation, interface capturing method is generally used, in which the governing equations
are solved for both the air and the water phases with the help of boundary conditions at the
free-surface.
31
Patm
Y
Air
P
water
X
Figure 3.2: An illustration of the conditions at an ideal free-surface of water and air in the
computational domain. Adapted from White and Corfield (2006).
Among the two types of multiphase models, namely the Eulerian-Eulerian and a La-
grangian particle tracking multiphase models (Brennen, 2005; ANSYS, 2016), only the
Eulerian-Eulerian model is relevant to this problem. For modeling a two-phase flow us-
ing Eulerian-Eulerian, two models are available, the homogeneous (or phase separated flow)
and the inhomogeneous (disperse flow) models (Brennen, 2005; ANSYS, 2016). In an incom-
pressible, homogeneous multiphase flow, both water and air have the same velocity at the
interface or zero relative velocity, as well as other common flow fields such as pressure and
turbulence (Brennen, 2005). In inhomogeneous flow, where the phases are dispersed, have
separate velocity, pressure and turbulence fields. In other words, the RANS equations must
be solved for each phase at the interface. For a given transport process, the homogeneous
model assumes that the transported quantities for that process are the same for both water
and air at the interface. As an illustration of homogeneous free-surface flow, water and air
are separated by a distinct resolvable interface as depicted in Figure 3.1.
Although the accuracy of free-surface homogeneous model has not been assessed in the
present work or in previous CFD studies on crossflow turbines, it is reasonable to assume
here that this model would give satisfactory results for the turbine performance. It will be
shown later that the water does all the energy transfer, i.e. the water volume fractions at the
impeller inlet and outlet are both one, and it only decreases significantly after the energy has
32
been extracted. The homogeneous multiphase model is a reasonable assumption in crossflow
turbines because the most dominant parameter for phase separation is the density difference
between the phases (Brennen, 2005). This is the simplest multiphase flow model and is also
computationally efficient. For simulating the flow in crossflow turbines, De Andrade et al.
(2011) and Sammartano et al. (2013) used the homogeneous free-surface model, whereas
Choi et al. (2008) used the inhomogeneous model. These studies have reported a satisfactory
agreement with the experimental results. It remains, however, that the choice of multiphase
models needs to be assessed for more accurate flow computations as it can affect the accuracy
of CFD predictions and thus merits further examination. This aspect is not examined in
this work as satisfactory results have been obtained with the homogeneous free-surface flow
model, which will be discussed in the next chapter.
Time-averaged continuity equation for an incompressible homogeneous flow is written as:
∂
(αw Uw + αa Ua ) = 0 (3.5)
∂xi
where αw = volume fraction of water, αa = volume fraction of air, αa + αw = 1 and the indices
represent: w= water and a= air. It is noted that for homogeneous multiphase flow, Uw =
Ua , thus Equation (3.5) reduces to Equation (3.3).
Similarly the momentum equation is written as:
0
∂ρU ∂ρUi ∂P ∂ ∂Uj ∂Ui
+ Uj =− + µef f ( + ) (3.6)
∂t ∂xj ∂xi ∂xj ∂xi ∂xj
where µef f = effective viscosity accounting for turbulence given by µef f = µ + µt , µ =

0 0 2ρk
αa µa +αw µw , µt = turbulent viscosity, P = modified pressure defined as P = P + 3
, k
= turbulent kinetic energy, U = (αw ρw Uw + αa ρa Ua )/ρ and ρ = αw ρw + αa ρa . The symbols
αw , Uw , ρw , αa , ρa and Ua are respectively the volume fraction of water, mean velocity of
water, water density, volume fraction of air, air density and mean velocity of air.
33
3.3 Computational Details
3.3.1 Computational Mesh
The used computational model involves solving RANS/URANS equations via a cell-centered
discretization on unstructured tetrahedral mesh using a finite volume scheme. In this scheme,
discretization can be directly done in the physical coordinate system and combined with a
cell-centered discretization. This scheme integrates the mass and momentum fluxes over
each cell to calculate the time rate of change of flow properties at the centre of the cell.
The algorithm used in the simulations is based on high resolution, second order accurate
and backward discretization scheme with double precision. For modeling water and air in
the computational domain, the homogeneous free-surface multiphase model described in the
previous section is used. To compute the flow properties in stationary and rotating frames of
reference, domain decomposition is used; the computational domain is divided into two sub-
domains: the stationary domain consisting of nozzle and casing and the rotating domain
consisting of the impeller as illustrated in Figure 3.3. General grid interface is used for
connecting these two sub-domains (ANSYS, 2016).
Computational domain was first defined in the geometry. In the 7 kW turbine, shaft was
not included in the geometry as there was no information avaiable on its diameter. In the
case of 0.53 kW turbine, shaft was included, so the angular momentum exiting the first stage
may not be equal to the angular momentum entering the second stage because there might
be some losses on the shaft. This means that shaft may have some losses, which in turn
affect the impeller performance. All the angular momentum exiting the first stage would not
be equal to the angular momentum entering the second stage.
CFD requires generation of suitable mesh in the computational domain. In turbomachin-
ery design analysis, structured hexahedral mesh, which are mostly aligned to the main flow
path, is preferred to get more accurate solutions at less computational time (Tucker, 2011;
Ali and Tucker, 2014). However, generating a structured hexahedral mesh is challenging in
34
Figure 3.3: An illustration of the typical computational domain for the crossflow turbine.
complex geometries, such as in a crossflow turbine. Recently unstructured hybrid meshes

have been used in turbomachinery simulations, in which hexahedral or prismatic cells are
used in the boundary layers (near wall regions) and tetrahedral and prismatic cells are used
away from the wall (Blazek, 2015; Tucker, 2011). In this work, both structured (in the near
wall region) and unstructured meshes (away from the wall), called the hybrid mesh, are used
in all the turbine configurations considered in this thesis. Previous computational studies on
crossflow turbines mentioned in the previous chapter have used this type of mesh. To gen-
erate the computational mesh, relevant geometries of the computational domain were first
defined in the computer-aided design (CAD) environment, SolidWorks. The CAD model
was cleaned (e.g. checked if there were any geometrical errors) in ANSYS ICEM meshing
software before it was used for meshing. After the mesh was generated in ICEM, a more
detailed analysis of the mesh quality was performed. ICEM allows quality check, smoothing
and refinement based on different criteria (e.g. orthogonality, aspect ratio, skewness etc.).
The mesh quality was carefully examined and improved through quality criteria (e.g. mini-
mum orthogonality = 0.3, maximum aspect ratio = 10 and maximum skewness = 0.25 etc.),
35
smoothing and refinement features available in ICEM as recommended in ICEM meshing
guidelines (ANSYS, 2016). These are essential to ensure good quality mesh not to hinder
numerical convergence. In order to improve the mesh quality, at least ten prismatic lay-
ers were generated on all walls and blades in an attempt to capture the flow features of
the turbulent boundary layer. For turbomachinery simulations using SST k − ω tubulence
model (Denton and Dawes, 1998; Montomoli et al., 2011; Tucker, 2011), the dimensionless
wall distance y + < 1 (y + = uτ y/ν) is required in order to accurately capture the important
features of the turbulent boundary layers. In this analysis, especially in the impeller blades,
not all the nodes nearest to the wall were in the region y + < 1. Some of the regions were
found in the range y + < 100. To get this mesh resolution, few simulations were performed by
gradually refining the prismatic layers and computing the y + in the solution. The automatic
wall function available in SST k − ω model allows a smooth switch between the wall function
(in the wall region where y + > 1) and the low Reynolds number grids (in the wall region
where y + < 1) (ANSYS, 2016). According to the analysis reported by Kalitzin et al. (2005)
and Weide (2004), use of wall function allows use of relatively coarser grids (e.g. y + = 100)
without a significant loss in accuracy. The y + sensitivity has been discussed in more detail
by Nichols (2008).
The impeller inlet and blade passages are regarded as highly important regions of the
flow field, where the energy extraction takes place and high velocity gradients exist. In addi-
tion, the narrow tip-gap, the region between the blade-tip and the casing wall, significantly
increased the complexity of the grid topology. Even more complex regions are the sharp
leading and trailing edges of the blades. In order to minimize the discretization error, a high
resolution mesh is required near the tip regions. It was found in this study that unstructured
grids are easy to create for complex geometries, but the quality of these grids deteriorates
with complex geometries and the large number of mesh elements are required. As a result,
there is a large computational overhead due to increased number of mesh elements compared
to an equivalent hexahedral mesh.
36
3.3.2 Grid Convergence
In order to determine the minimum grid size or mesh resolution required to resolve the
boundary layer and the mean flow features, a grid independence study was conducted. This
allows minimization of errors and uncertainties in the predicted results, for example the
impeller power output or the efficiency. As the geometry is highly complex with curved walls
and narrow tip regions, very fine grids were required. Therefore grid convergence analysis
has been carried out considering power output (in this study by monitoring torque) as a
parameter of significant interest. Dubbioso et al. (2014) considered the forces and moments
(thrust) in the detached-eddy simulation of propeller in their grid convergence analysis.
After attempting various grid refinements in order to obtain the converged results for the
turbine power output, it was determined that about 5.4 million mesh elements were required
for the 7 kW turbine described in Chapter 4, to produce grid-independent results for the
power output. For this, steady RANS simulations were performed by gradually increasing
the number of mesh elements until a small difference of about 122 watts between the last
three solutions was obtained. This numerical uncertainty due to grid size corresponds to
less than 0.1% of the total power output. In each simulation, steady state torque values,
and the difference between the maximum and the minimum of the fluctuation were recorded
to estimate the uncertainty in converged solution. This is the numerical model generated
numerical uncertainty. The above values indicate how much the solution for the power output
is likely to change with the use of this grid and RANS model. The grid sensitivity was tested
at the impeller speed of 250 RPM and 450 RPM. In general, the grid convergence is examined
by halving or doubling the grid spacing and Richardson’s extrapolation (Wilcox, 2006), but
in this study grid refinement was done as discussed above to obtain grid independent results.
The results of grid independence study are presented in Chapter 4.
37
3.3.3 Turbulence Modeling
Current approaches to flow modeling are classified into four types: 1) RANS/URANS sim-
ulations, 2) Large eddy simulations (LES), 3) hybrid RANS-LES and 4) Direct numerical
simulation (DNS). In RANS, time-averaged flow properties, the pressure and the velocity, are
computed. RANS simulations with appropriate turbulence models have been widely used for
turbomachinery design analyses due to their low computational cost and satisfactory predic-
tive capability for average device performance (e.g. Montomoli et al. (2011); Tucker (2013)).
In LES, Navier-Stokes equations are solved for the large scale eddies and the small scale ed-
dies of the flow are modelled. This demands greater computational resources than the RANS
methods, so it is still computationally costly to be practical for turbomachinery design anal-
ysis (Larsson and Wang, 2014). In hybrid RANS-LES approach, the inner boundary layer
is modelled with RANS using the Spalart-Allmaras (SA) turbulence model and outside the
boundary layer, LES is used (Tucker, 2011). A more promising hybrid RANS-LES approach
is the detached eddy simulation (DES) (Tucker, 2011). As the computational cost is higher
than the RANS for multi-parametric design evaluations, this approach is not used in this
study. In DNS, numerical solutions of the Navier-Stokes equations are obtained, which have
the capability to simulate every fine details of the flow field. As the computational cost is
huge, this method is in general impractical for turbomachinery design analysis, at least at
the present time. In this research, RANS/URANS simulations are used as these are compu-
tationally practicable and are also suitable for most of the turbomachinery design analysis
(Denton and Dawes, 1998; Denton, 2010; Montomoli et al., 2011; Tucker, 2011; Bourgeois
et al., 2011; Tucker, 2013; Cornelius et al., 2014). In the design of such machines, average
flow properties are important rather than a detailed flow field properties. The large com-
putational resources associated with LES-DES or LES computations render the direct CFD
approach unsuitable in the design phase or the parametric exploration of the turbine.
In this study, SST k − ω turbulence model of Menter (1994) is used. Although k − ε
turbulence model is a general purpose standard model in most practical problems, the stan-
38
dard k −ε model shows poor performance in turbomachinery flows, which involve separation,
curvature and rotation (Tucker, 2011). The SST k − ω turbulence model is a two equation
model for the turbulence kinetic energy k and the specific dissipation rate ω that consists
of blending of k − ω model and k − models based on proximity to the walls. This model
improves the k − ω model by using the blending function in the near-wall region, which
activates the standard k − w model and transforms to the k − ε model away from the wall.
This idea originates from Menter’s observation that the basic k − ω model over predicts
the level of shear stress in adverse pressure gradient boundary layers (Menter, 1994). The
k − ε model is found to perform poorly in adverse pressure gradients but better in free-shear
layers (Menter, 1994). Therefore, turbulent viscosity is modified using a limiter function
to account for the transport of turbulent shear stress, which shows better agreement with
experimental results in adverse pressure gradients, streamline curvature and the onset of
separation (Menter, 1994). These features make the SST k − ω model more accurate and
reliable in turbomachinery flows than the standard k − ω model or k − model. The pre-
dictive capability of this model for the onset and separation at adverse pressure gradients
has been found better than other two-equation turbulence models, such as k − or k − ω
(Menter, 1994).
Transport equations are used to compute the turbulent kinetic energy k and the turbulent
viscosity νt . The turbulent viscosity turbulence models use the Bousinessq approximation
for the Reynolds stresses, which is written as:
0 0 2 ∂Ui ∂Uj
−ρui uj = ρkδij − µt ( + ) (3.7)
3 ∂xj ∂xi
0 0
where k = 21 uk uk . The transport equation for the turbulent kinetic energy k for the k − ε
model is written as:

∂k ∂Uj k ∂ νt ∂k
+ = ν+ + Pk − (3.8)
∂t ∂xj ∂xj σk ∂xj
39
0 0
where the production of turbulent kinetic energy, k, is given by: Pk = ui uj ∂Ui /∂xj . Similarly,
the transport equation for the turbulence dissipation rate ε is:

∂ε ∂Uj ε ∂ υt ∂ε ε
+ = υ+ + (Pk Cε1 − Cε2 ε) (3.9)
∂t ∂xj ∂xj σε ∂xj k
where Cε1 = 1.44, Cε2 = 1.92, σε = 1.3 , Cµ = 0.09 and σk = 1 are the model constants
(Jones and Launder, 1972). The turbulent viscosity υt is given by:
k2
υt = Cµ (3.10)
ε
The transport equation for the turbulent kinetic energy k for the k − ω model is written
as:

∂k ∂Uj k ∂ νt ∂k 0
+ = ν+ + Pk − β kω (3.11)
∂t ∂xj ∂xj σk ∂xj
Similarly, the transport equation for the specific dissipation rate, ω ∝ ε/k, is:

∂ω ∂Uj ω ∂ υt ∂ω ω
+ = υ+ + α Pk − βω 2 (3.12)
∂t ∂xj ∂xj σω ∂xj k
0
where υt = k/ω, ε = Cµ k/ω and the model constants are: β = 0.09, σk = 2.0, σω = 2.0,
α = 5/9 and β = 0.075. Far from the wall, SST k − ω is designed to transform into the k − ε
model because k − ε performs better in free shear layers. This solves the shortcoming of the
k − ω model that it is strongly sensitive to the free shear layers. Menter suggested the use
of the following boundary condition:
60ν
ω= (3.13)
βy 2
40
where y is the distance from the wall to the center of the first cell above the wall. The SST
k − ω model uses the blending functions, so the transport equation for k − ω becomes:

∂ω ∂Uj ω ∂ υt ∂ω ω 1 ∂k ∂ω
+ = υ+ + α Pk − βω 2 + 2(1 − F 1)σω2 (3.14)
∂t ∂xj ∂xj σω ∂xj k ω ∂xj ∂xj
The transformation between models is accomplished by the blending function, F1 , applied

to the coefficients of the ω equation. α and β change from their k − ω values to their k − ε
values. The blending function is given by:
F1 = tanh(arg14 ) (3.15)
" √ #
k 500υ 4ρk
arg1 = min max( 0 , 2 ), (3.16)
β ωy y ω CDkw σω2 y 2
where CDkω is written as:
1 ∂k ∂ω
CDkω = max(2ρ , 1.0 × 10−10 ) (3.17)
σw2 ω ∂xj ∂xj
The turbulent viscosity νt , which includes the shear stress limiter, is expressed as:
a1 k
νt = (3.18)
max(a1 w, SF2 )
F2 = tanh(arg22 ) (3.19)
and √
2 k 500ν
arg2 = max( 0 , 2 ) (3.20)
β ωy y ω
∂Uj
2Sij Sij , Sij = 21 ( ∂U
p
where S = ∂xj
i
+ ∂xi
) and F2 is the second blending function.
41
3.3.4 Boundary Conditions
Computation of the flow in the turbine requires the specifications of inlet, outlet and wall
boundary conditions in the computational domain. These are described below.
Inlet boundary condition
The inlet boundary condition corresponds to the experimentally tested operating conditions.
In the simulation of crossflow turbines, two types of inlet boundary conditions have been used
in the literature, either the total pressure or the velocity (Choi et al., 2008; De Andrade et al.,
2011; Sammartano et al., 2013; Sinagra et al., 2014; Acharya et al., 2015). These studies
have shown good agreements with the experimental results for both boundary conditions.
In this study, total pressure corresponding to the operating head was specified at the inlet.
Uniform pressure was assumed as no measurements are available for the inlet flow. Water
and air volume fractions were chosen to be 1 and 0 respectively. As a test case, another
set of simulations using uniform mass flow rate was also performed and similar results were
found for the turbine power output. Turbulence intensity of 5% was assumed at the inlet.
The inlet was put far away from the impeller so that the flow at this location is unaffected
by the downstream non-uniformities, if any, in the flow created by the impeller motion. The
specific requirement depends upon the length scale of the downstream component; in this
case, the chord length of the impeller blades. The objective is to minimize the uncertainty
on how far upstream will the influence of the non-uniformity extend? (Greitzer et al., 2007).
Outlet boundary condition
As the flow exits the impeller at an atmospheric condition, the outlet boundary condition
must be specified appropriately. Following the previous works of Choi et al. (2008), De An-
drade et al. (2011) and Acharya et al. (2015), bulk mass flow rate was specified as an outflow
boundary condition for the case with inlet pressure boundary condition. As a test case, for
the mass flow rate type inlet boundary condition, the opening type of outlet boundary con-
dition was used with the specification of a uniform average pressure of 1 atm. The opening
42
type of boundary allows the flow of air in and out of the computational domain, whereas
water can only leave the outlet (air volume fraction = 1 and water volume fraction = 0).
Air vents were specified opening type of boundary condition with 1 atm pressure (air vol-
ume fraction = 1 and water volume fraction = 0). The outlet is chosen at a sufficient far
downstream location of the impeller, where the non-uniformities in the flow have decayed so
that the flow can be considered uniform.
Wall boundary conditions
The wall boundary conditions are applied on the walls of the nozzle, the casing and the
blades. All the walls in the stationary domain are treated as no-slip, isothermal stationary
walls, whereas the walls of the rotating blades are treated as no-slip, isothermal, rotating
walls.
Interface
For the steady and unsteady RANS computations, there are two types of interface models
for computing the flow in the turbine. The frozen rotor model is available for the steady
RANS, and transient stator-rotor model is available for the unsteady RANS. For the steady
RANS computations, the interface connecting the stationary and rotating domains is chosen
as frozen rotor, which is approximated by a mixing-plane approach (ANSYS, 2016). In this
model, appropriate transformation equations in the rotating frame of reference are applied
and the solution has no dependence on the relative positioning of the impeller and the casing.
This steady state approximation is called mixing plane approach, which is achieved by aver-
aging the conserved variables (mass and momentum) at the interface. This approximation
also significantly reduces or eliminates any potential interaction effects between impeller and
nozzle. In other words, this model cannot take into account the losses incurred in the real
(transient) situation as the flow is mixed between stationary and rotating components. In
addition, it cannot model the transient effects of the flow over time caused by the blades
43
rotating past a circumferentially varying inlet flow. This model consists of averaging the
circumferential non-uniformities of the upstream flow at many spanwise locations, incurring
a local, instantaneous loss at the interface. This approximation also significantly reduces or
eliminates any potential interaction effects between rotor and stator. The unsteadiness or
transient effects can be considered as a cyclic flow due to impeller rotation (Greitzer et al.,
2007; Tucker, 2011). For URANS simulations, the interface is chosen as transient rotor sta-
tor, which has the capability to capture the effects of transient flow between the impeller and
the nozzle and casing. The results of steady simulations were used to initialize the unsteady
calculations. The details of the mixing plane approach are described in ANSYS (2016).
3.4 Cavitation
One of the important flow phenomena influencing the internal flow and performance loss
in hydraulic turbines is cavitation. Therefore, one of the major requirements in the design
and operation of hydraulic turbines is to avoid cavitation. To the author’s knowledge, the
only study on cavitation in crossflow turbines reported in the literature is by Adhikari et al.
(2016). The main findings of this study will be presented in Chapter 4. Since high-performing
turbines are very important for the overall sustainability of small-scale systems, the issue
of cavitation must be addressed in future turbine designs. Experimental observations and
measurements of cavitation in turbines involve complexity as well as high cost. Therefore,
computational studies employing the solutions of the Navier-Stokes equations of various
fidelity 1 , because of their ability to determine pressure everywhere in the flow, are required
to gain fundamental understanding of the underlying flow structures and the possibility of
cavitation. It is extremely important to characterize cavitation, if present, in order to avoid
or minimize it and improve the overall turbine performance. The ability to avoid or minimize
cavitation leads to better turbine designs and longer lifetimes of installed turbines.
1
such as the RANS and URANS used here
44
Cavitation occurs when the pressure falls sufficiently low in some regions of the flow
field, and as a result vapor bubbles are formed. Cavitation is common mostly in reaction
types of hydraulic turbines (Escaler et al., 2006); however, there are no previous studies
on cavitation phenomena in crossflow turbines. Crossflow turbines are commonly used for
low-head operations, therefore cavitation analysis is potentially important. Cavitation is
characterized in terms of cavitation number σ, which is defined as (Brennen, 2005):
p∞ − pv
σ= 1 (3.21)
ρ U2
2 w ∞
where U∞ , p∞ and T∞ are respectively the reference velocity, pressure and temperature in
the upstream flow, ρw is the water density and pv (T∞ ) is the saturated vapor pressure. The
inception of cavitation is defined by σi at which it is equal to the minimum value of the
coefficient of pressure Cpmin . The condition for the inception of cavitation is written as
(Brennen, 2005):
σi = −Cpmin (3.22)
Further reduction of σ from σi increases the cavitation. In this work, only the cavitation
inception is examined, based on the grounds that mapping a cavitation-free operating range
is important for the design and operation of the turbine, rather than description of its full
development mechanisms.
3.5 CFD Model Validation
With CFD solutions, it is always crucial to proceed with caution as noted by H.L. Mencken
that “there is always an easy solution to every human problem-neat, plausible and wrong”
(Wilcox, 2006). This also applies when using CFD results in the analysis of turbines. There-
fore, CFD solutions must indicate the error bands and uncertainties incurred in the results
45
in order to have confidence in the predicted results 2 . Validation is related to quantifying
the accuracy of the computational results by comparing with the experimental data. Un-
certainties are associated with specifying the physical parameters in the model or due to
lack of understanding of the physical model, whereas errors are associated with translation
of mathematical formulation into numerical algorithm (round-off, convergence) (Iaccarino,
2012). However quantifying CFD uncertainties and errors is a difficult problem. It is there-
fore desirable that uncertainties and errors be minimized (Roache, 1997; Stern et al., 1999).
Thus validation of CFD results (degree of agreement with experimental results) is needed.
In doing so, it is important to know how accurately the true flow physics is described by a
validated solution, which is called uncertainty.
Iteration convergence and grid independence are the two requirements for any CFD
solutions (Wilcox, 2006) to be useful for engineering problems. Iteration convergence ensures
that a converged solution has been obtained (Wilcox, 2006). Similarly, grid independence
ensures that magnitude of discretization errors has reduced to an acceptable level (Wilcox,
2006). For grid independence study, grid refinement is performed unless numerical results
converge to an acceptable accuracy (Durbin and Reif, 2001). The sufficiency of a grid is
tested by ensuring that statistics (e.g. velocity and pressure, torque or power output in the
present study) are relatively insensitive to that resolution. The objective is to demonstrate
that the model will not lead to a deterioration in accuracy of results for different sets of
geometrical configuration of the turbine. As the important considerations to be taken into
account are the turbulence models and grids, grid independence study is the first step in
assessing the capability of a turbulence model. Numerical solutions may be sensitive to near
wall grid resolution (y + ) when using automatic wall functions in the turbulence models, such
as in k − ω and SST k − ω (Wilcox, 2006). The y + sensitivity is discussed in more details
by Nichols (2008).
2
“it is almost impossible to find a precise definition of the terms verification and validation”
46
The first criteria for a converged solution is to check the values of the residual in the
discretized mass and momentum equations. Residual is defined as the magnitude of the
difference between the left-hand-side of the discretized mass and momentum equations and
the right-hand-side in the solution. In ANSYS CFX, it is recommended that a reasonably
converged solution requires a maximum residual level not higher than 5 × 10−4 or the root
mean square (RMS) residual as 0.5 × 10−4 ; RMS of 10−4 may be sufficient for engineering
applications (ANSYS, 2016). In this study, considering the computational time and available
resources, calculations were considered converged when all RMS residuals reached 10−5 , or
when both residuals and monitored points for velocity and torque displayed periodicity
with slight variations in the values. There were 5 monitor points spaced at key points
throughout the computational domain (e.g. both nozzle throat and outlet and blade leading
and trailing edges, as well as in the impeller-nozzle clearance. In cases where the latter
convergence criterion was used, the RMS residual for mass conservation was at the desired
10−5 level while the momentum residuals were between 10−4 and 10−5 . In general, for
complex geometry with unstructured mesh, more computational time is required to get
lower RMS residual values. Moreover, the computational time is significantly increased
in multiphase simulations (ANSYS, 2016). It is important to check whether the monitored
parameter, for example velocity or torque, has achieved the steady state value with minimum
fluctuation during the simulation.
It is important to identify the sources of errors in turbomachinery simulations and at-
tempt to minimize them. According to Denton (2010), the errors in turbomachinery sim-
ulations are roughly classified as originating from 1) numerical errors because of domain
discretization and 2) errors due to turbulence and multiphase modeling, where the true
physics is too complex to capture with a numerical model and unknown geometrical shapes,
such as leading and trailing edge shapes.
47
3.6 Methods of Computational Analysis
The most important fluid dynamic quantities to assess the turbine performance are the inlet
velocity, the relative flow angles at the inlet and exit of the blades and the angular momen-
tum. These parameters describe the fundamental flow physics and how the impeller extracts
power from the fluid. These parameters are the input parameters for computing the dimen-
sions of the nozzle and the impeller, and also set the criteria to match the two components.
These quantities can be analyzed locally and globally. The flows in turbomachines are rather
complex, and so the issues arising from the dominant flow structures are helpful to have a
unifying principle rather than the fine details of the flow field. To analyze the performance
of each stage by splitting the power into stage contributions, it is assumed that the angular
momentum leaving the first stage is equal to the angular momentum entering the second
stage. In the case of 7 kW turbine, the shaft was not modelled as there was no information
available in the literature, and thus its effect was not captured. However, in the case of 0.53
kW turbine, the shaft was modelled as the size of the shaft is available in the literature.
3.7 Chapter Summary
This chapter presented the description of the computational method for the simulation of
flow in crossflow turbines. RANS formulation with SST k −ω turbulence model and homoge-
neous two-phase (water and air) flow with free-surface (water and air) effects was described.
Methods of generating computational mesh, specifying the boundary conditions, and the
methods of assessing errors and uncertainties in the computed results were described. Fi-
nally, a brief outline of the methods of computational analysis of the flow in the turbine was
presented.
48
Chapter 4
Validation and Performance

Characterization
This study requires a detailed assessment of all the potential sources of errors and uncer-
tainties that might occur in the computed results. This is important for determining the
confidence in the computed results. A standard procedure of validation is to compare the pa-
rameters of interest with the experimental measurements. A validated computational model
can then be used for characterizing the flow field for other operating conditions or for changes
in the turbine geometry. In this chapter, results from steady and unsteady RANS computa-
tions are compared with the experimental data to assess the capability of the computational
model to capture the turbine’s performance along the speed-line and the changes in turbine
performance due to variations in operating conditions and design parameters. The predictive
capability is evaluated based on the maximum relative error between the computed and the
experimental results for the turbine power output or the efficiency. No detailed information
of the flow, such as measurements of velocity profiles at the impeller inlet or in the nozzle,
was available for comparison. The uncertainty in the computed results is evaluated based
on the difference between the maxima and minima in the turbine power output during the
simulation in a converged solution. Grid convergence studies and comparisons between the
49
steady and unsteady RANS results are made to quantify the error bands and minimize nu-
merical uncertainty in the predicted results. This establishes the uncertainty and error band
in the computed results.
In crossflow turbines, the fundamental flow processes or mechanisms responsible for lim-
iting the turbine performance are still not well-understood. There are no measurements of
the internal flow and its influence on turbine performance. Therefore, in this study, per-
formance characterization has been carried out with particular emphasis on understanding
the dominant flow features that characterize behaviour. For example, flow separation on the
blades is a major cause of performance loss in the impeller. The results of the computed flow
field are then synthesized to suggest a design strategy for improving the turbine efficiency.
Characterization of the mean flow field and turbine performance allows one to determine
differences which arise due to the changes in turbine design parameters, and identify key
features that govern turbine performance. For this purpose, comparisons are made between
the mean flow fields of two turbine designs: one 7 kW turbine with a maximum efficiency
of 69% and the other 0.53 kW turbine with a maximum efficiency of 88%, which provide
insights on the reasons for inefficiency as well as guidance for improving the efficiency. The
aim of this chapter is therefore to characterize the fundamental flow features and power
extraction mechanisms of these turbines and establish a basis for improved design.
4.1 Turbine Efficiency and Performance Metrics
The performance of a turbine is measured in terms of maximum efficiency over a range of

operating conditions. The turbine efficiency (η) is defined as:
shaft power Tω
η= = (4.1)
available power ρgHQ
where T is the shaft torque, ω is the angular velocity of the impeller, ρ is the density of
water, g is the acceleration due to gravity, H is the operating head, and Q is the flow
50
rate. Assuming a steady flow at the impeller boundary surface as illustrated in Figure 4.1,
Reynolds transport theorem for the total change in angular momentum gives:
ˆ ˆ
T =− ρrur uθ ni dA (4.2)
where T = total torque exerted on the blades by the fluid, ρ is the density of water, r = radius,
ur = radial velocity of water, uθ = tangential velocity of water, and A = impeller boundary
surface area. It is noted that the effect of air is assumed negligible. The implications
of this formulation in turbine performance analysis can be elucidated by employing the
control volume analysis as follows. Equation (4.2) shows that the turbine shaft torque T is
´´
given by the angular momentum change ρrur uθ ni dA produced at the impeller boundary
surface, which is the quantity of interest in the performance analysis. Therefore the key
objective of any attempt to improve the turbine performance must be to increase the change
in angular momentum at the impeller boundary surface. This is achieved by maximizing
the angular momentum flux of the flow at the impeller inlet and minimizing the angular
momentum leaving the impeller. From the design viewpoint, this implies the significance of
the individual performance of the nozzle and the impeller and matching the flows in these
components.
The fluid forces acting on the control volume or the control surfaces of the impeller are
composed of pressure and viscous forces. To obtain maximum torque T for a given speed
in Equation (4.2), the flow must have a maximum angular momentum flux at the impeller
inlet, whereas at the exit of the impeller, the angular momentum of the flow should be
negligible. Thus the design challenge is to understand the design issues arising from the
flow conditions at the impeller inlet and the flow patterns developed within the impeller.
This further implies that matching the nozzle and the impeller geometries is crucial. The
objective is to understand how the turbine performance will change when the flow field is
perturbed by changes in geometrical parameters (i.e. nozzle and impeller geometries) for a
given operating condition (i.e. flow rate, head and impeller speed).
51
Figure 4.1: Steady flow through the control volume of the impeller.
Incoming flow
Incoming flow
0
β2i 90
0
β2i 90
Ω Ω
(a) power extraction across (b) power loss across

the blades the blades
Figure 4.2: A schematic illustration of the conditions of power extraction and loss across the
blades at the second stage of the impeller. Note that the inner blade angle is 90◦ .
52
Although the flow within an impeller is inherently unsteady, the steady flow model can
still be made of by considering the steady flow in and out of the impeller boundary surface
as illustrated in Figure 4.1 to compute the change in angular momentum at the impeller
boundary surface, which gives the torque T . There are two noteworthy points for analyzing
the flow field in the impeller and its performance.
1
1. If the relative tangential velocity wθ decreases across a blade row (where positive wθ
is defined in the same direction as the impeller rotation), then work is removed from
the flow. Then the impeller blade extracts power from the flow. This is illustrated in
Figure 4.2.
2. If the relative tangential velocity wθ increases across the blades (where positive wθ is
defined in the same direction as the impeller rotation), then work is added to the flow.
Then the impeller blade loses power to the fluid. This study has found this phenomenon
to happen generally at the second stage of the impeller, but at high impeller speeds,
this occurs at the first stage as well. This aspect will be discussed in more details in a
later section.
The flow at the inlet of the impeller is predominantly in the tangential direction (in the
same direction as that of the impeller motion) so as to increase the angular momentum flux.
The flow at the exit of the impeller should be in the radial direction, so as to reduce the
angular momentum. It is important to note that the inlet and exit of the impeller have the
same geometry; this imposes a design constraint on blade design. To gain qualitative and
quantitative insights on the regions where the impeller is more efficient and where it is less
efficient in extracting the energy from the fluid, ur and uθ or angular momentum change
at the impeller boundary provide useful information. For example, Yang (1991) used ruθ ,
the angular momentum per unit mass, to optimize radial turbine blades using inverse design
methods.
1
wθ = uθ - ω r, where uθ is the absolute tangential velocity at the inlet or exit and ω is the angular
velocity of the impeller.
53
The simulation of the entire flow field enables computation of the velocity, pressure, and
flow angles at the inlet and outlet of the impeller. Such quantitative information is important
for investigating the fluid dynamic mechanisms of power extraction in the impeller. For
example, angular momentum change per blade, which gives the power extraction per blade,
can be computed and characterized. The local flow field characteristics associated with the
power extraction, such as flow separation in blade passages, will provide guidance toward
performance improvement. The relative flow angle β is defined as:
ur
β = tan−1 ( ) (4.3)
uθ − ωr
where ur = radial velocity, uθ = tangential velocity, ω = angular velocity of the impeller,

and r = R1 or R2 . Common turbomachinery design principles, e.g. Dixon and Hall (2013),
suggest that this angle should be equal to the blade inlet angle. It will be shown that this
is approximately true at maximum efficiency but the situation is complicated by the ability
of the second stage to extract energy that is “missed” at the first stage.
4.2 Results of Validation
For the purpose of validation and performance characterization, two small-scale crossflow
turbines reported in the literature were considered. The first turbine is a 7 kW low-efficiency
turbine with a maximum efficiency of 69%, which was experimentally studied by Dakers and
Martin (1982). However, the dominant flow physics and the reasons for its poor performance
are not known in any detail. The characteristics Reynolds number considering the chord
length of the blade and relative velocity at the impeller inlet corresponding to the impeller
speed at maximum efficiency is about 2.8×106 . The second turbine is a 0.53 kW high-
efficiency turbine with a maximum efficiency of 88%, which was experimentally studied
by Desai (1994). A further improvement in maximum efficiency of the same turbine was
obtained by Totapally and Aziz (1994), who achieved a maximum efficiency of 90% by
54
increasing the number of blades from 30 to 35. Thus far, this is the most efficient turbine
design reported in the literature. Furthermore, the internal flow features of this turbine are
not known in any detail. The characteristics Reynolds number considering the chord length
of the blade and relative velocity at the impeller inlet corresponding to the impeller speed at
maximum efficiency is about 0.8×106 . A better understanding of the key flow features and
the reasons for its high performance provides the essential features of the design requirements
for high efficiency turbines and avenues for further improvement. These are the primary
motivations for investigating this particular turbine. Moreover, a detailed description of
the geometry of the turbine, the operating conditions, and the experimental results of the
turbine performance are available for both turbines, which is essential for the purpose of
computational validation and performance characterization. A large number of operating
conditions and geometrical configurations are typically needed for assessing and validating
the computational model to build confidence in the analysis. These two turbines provide
these details. The operating conditions and the geometrical details for the assessment of
the computational model and investigation of the internal flow and turbine performance are
presented in the next section.
Large computational resources associated with unsteady RANS (URANS) computations
render the direct application of computational simulations unsuitable in the design phase
or the parametric exploration. Therefore, URANS computations are only used to verify the
results of steady RANS, and steady RANS provide an affordable and acceptable accuracy
to assess the performance of different turbine configurations.
4.2.1 Computational Details and Results
The 7 kW turbine is depicted in Figures 4.3 and 4.4 and the corresponding design parameters
are presented in Table 4.1. As a point of clarification, the shaft has not been included in
the impeller of the 7 kW turbine as there was no information about the size of the shaft. It
is therefore assumed in the analysis that the angular momentum exiting the first stage will
55
600
Flange
Inlet flow
120
26
68
179
170
159
0
69
30.5 47
231
All dimensions in mm
Figure 4.3: Nozzle dimensions of 7 kW turbine. Redrawn from Dakers and Martin (1982).
be equal to the angular momentum entering the second stage. The operating parameters
are given in Table 4.2. For this turbine, both steady and unsteady RANS computations
were carried out at various operating conditions along the experimentally determined speed-
lines. Similarly, the design configuration of the 0.53 kW turbine is depicted in Figure 4.5,
and the corresponding design parameters are presented in Table 4.3. It is noted that the
turbine has constant width as well as constant blade cross-section. For the 0.53 kW turbine,
only steady RANS computations were carried out at various operating conditions along the
experimentally determined speed-lines. The operating parameters are given in Table 4.4.
First of all, grid convergence tests were carried out on both turbines. For the 7 kW tur-
bine, a total number of 5.4 million mesh elements were required to obtain mesh independent
results for the power output (Table 4.5). A solution was considered grid independent for
less than 0.1% difference in the power output between three different consecutive mesh sizes.
Similarly, for the 0.53 kW turbine, about 14.5 million elements were required to obtain mesh
independent results. The results of mesh independence study for the 0.53 kW turbine is
summarized in Figures 4.6. Mesh independent test was done considering the impeller speed
56
flow
β1b
A
Rb
C
blade centre blade
β2b B
R1
flow
R2
D2=211.72 mm O = Impeller centre

D1 = 316 mm
Figure 4.4: Impeller dimensions of 7 kW turbine. Data taken from Dakers and Martin
(1982).
Table 4.1: Design parameters of 7 kW turbine
Design parameter Value

Outer diameter (D1 ), [mm] 316
Inner diameter (D2 ), [mm] 211.72
Angle of attack (α), [deg] 16
Outer blade angle (β1b ), [deg] 30
Inner blade angle (β2b ), [deg] 90
Blade radius (Rb ), [mm] 52.14
Number of blades (Nb ) 20
Impeller and nozzle width (W ), [mm] 150
Nozzle thickness or throat (h0 ), [mm] 65
Nozzle entry arc (θs ), [deg] 69
57
Table 4.2: Operating conditions of 7 kW turbine
Operating conditions Value

Flow rate (Q), [lps] 105
Head (H), [m] 10
Head (H), [m] 8
Head (H), [m] 5
Head (H), [m] 3
Impeller speed (H), [RPM] 200 - 600
151 mm
flow
Y
nozzle throat:
220
h0 =89.44 mm
β1b=390
Rb = 52 mm
0
30 β2b =90 0
103.6 mm
0
θs =90
Z
152.4 mm
Figure 4.5: Schematic of the dimensions of the 0.53 kW turbine. Data taken from Desai
(1994). All the components have constant width W = 101.6 mm normal to the page.
58
Table 4.3: Design parameters of 0.53 kW turbine

Outer diameter (D1 ), [mm] 304.8
Inner diameter (D2 ), [mm] 207.2
Angle of attack (α), [deg] 22
Outer blade angle (β1b ), [deg] 39
Inner blade angle (β2b ), [deg] 90
Blade radius (Rb ), [mm] 52
Blade thickness (t), [mm] 3.2
Number of blades (Nb ) 30
Impeller and nozzle width (W ), [mm] 101.6
Nozzle thickness or throat (h0 ), [mm] 89
Nozzle entry arc angle (θs ), [deg] 90
Table 4.4: Operating conditions of 0.53 kW turbine
Operating conditions Value

Flow rate [lps] 46
Head (H), [m] 1.337
Flow rate [lps] 42
Head (H), [m] 1.17
Impeller speed (N ), [RPM] 160 - 224
that corresponds to the maximum efficiency (199.1 RPM). To compute numerical uncertainty
(e.g. the difference between the maxima and minima of the torque in the solution) in the
calculations, torque was monitored during the simulations. The steady RANS computations
were performed on the finest grid at different operating points. The typical mesh of the
impeller and the casing are shown in Figures 4.6 - 4.8. The results of RANS computations
for these turbines are summarized in Figures 4.9 and 4.10 respectively.
The unsteady RANS computations were carried out only for the 7 kW turbine, and are
described in the next section. The computed results of different test cases for the 7 kW
turbine are summarized in Figure 4.9. A comparison of the steady and unsteady RANS
computations against experimental results shows that RANS computations provide a good
agreement with the experiment. A maximum relative error of 6.11%, between the CFD
and experimental data, was found for the 7 kW turbine and 3.8% relative error for the
0.53 kW turbine at the design - point operating condition (Q = 46 lps and H = 1.337 m).
59
Figure 4.6: An illustration of a typical computational mesh of the rotating domain: the
impeller.
Figure 4.7: Zoomed view of the mesh resolution around the blades of the impeller.
60
Table 4.5: Summary of mesh convergence test for the 7 kW turbine at the maximum efficiency
point: 450 RPM (steady RANS calculations).
Mesh No of elements Power output Numerical

(million) (kW) error (%)
Mesh 1 0.96 6.83 13.87
Mesh 2 1.32 6.87 3.49
Mesh 3 2.12 7.091 3.11
Mesh 4 2.96 7.122 0.43
Mesh 5 3.47 7.157 0.49
Mesh 6 3.94 7.169 0.16
Mesh 7 4.53 7.181 0.16
Mesh 8 5.24 7.192 0.15
Mesh 9 5.40 7.199 0.10
Table 4.6: Summary of mesh convergence test for the 0.53 kW turbine at the maximum
efficiency point: 199.1 RPM (steady RANS calculations).
Mesh No of elements Power output Numerical

(million) (Watts) error (%)
Mesh 1 2.17 498 -
Mesh 2 2.76 493 1.01
Mesh 3 5.35 511 3.65
Mesh 4 7.92 519 1.56
Mesh 5 9.23 521 0.38
Mesh 6 12.23 525 0.76
Mesh 7 13.14 528 0.56
Mesh 8 14.51 529 0.18
At off-design operating conditions (Q = 42 lps and H = 1.17 m), the relative error was
about 5.3%. The error bars represent the numerical uncertainty involved in the computed
results. As shown in Figures 4.9, 4.10 and 4.11, satisfactory agreement with the experimental
measurements was obtained for both turbines. Thus, the two-phase computational model has
satisfactory predictive capability for the turbine power output over a wide range of operating
conditions (flow rate, head and speed), as well as varying geometrical parameters. Therefore,
steady and unsteady RANS computations are adequate to capture the overall features of the
flow, evaluate different turbine designs and improve the efficiency. Previous researchers
(Choi et al., 2008; De Andrade et al., 2011; Sammartano et al., 2013; Sinagra et al., 2014;
61
Figure 4.8: An illustration of a typical computational mesh of the stationary domain: the
nozzle and the casing.
Acharya et al., 2015) have also reported a similar predictive capability for crossflow turbine
performance using the steady RANS computations with two-phase homogeneous free-surface
model.
4.2.2 Unsteady RANS Computations
Fluid flows within turbines are inherently non-uniform and unsteady (Greitzer et al., 2007).
It is noted that the flow is circumferentially varying at the impeller inlet because the nozzle
entry arc angle is less than 360◦ . This is a geometry generated non-uniformity, primarily due
to contraction of the nozzle rear-wall in close proximity to the impeller. Thus the influence
of impeller motion on the upstream flow may be important. Two important design issues
are: 1) effect of nozzle wall shape on flow non-uniformity at the impeller inlet and the effect
62
7.5
6.5
Power output(kW) 6
5.5
4.5
Steady SST k−ω
4 Steady k−ω
3.5 URANS SST k−ω
Experiment
3 Steady SST k−ω
2.5 Experiment
2
200 250 300 350 400 450 500 550 600
Impeller speed (RPM)
Figure 4.9: Comparison of the steady and unsteady RANS solutions with the experimental
results for the turbine power output of the 7 kW turbine. Experimental data taken from
Dakers and Martin (1982).
0.54
Experiment
0.535 Steady SST k−ω
0.53
Power Output [kW]
0.525
0.52
0.515
0.51
0.505
0.5
150 160 170 180 190 200 210 220 230
Impeller Speed [RPM]
Figure 4.10: Comparison of the steady RANS solutions with the experimental results for the
turbine power output of the 0.53 kW turbine. The uncertainty in the experimental data is ±
2.4% [Q = 46 lps, H = 1.337 m and Nb = 30]. Experimental data taken from Desai (1994).
63
86
Experiment
85
CFD
84
83
Efficiency [%]
82
81
80
79
78
77
76
75
130 140 150 160 170 180 190 200 210
Impeller Speed [RPM]
Figure 4.11: Comparison of the steady RANS solutions with the experimental results for the
efficiency of the 0.53 kW turbine. The uncertainty in the experimental data is ± 2.4% [Q =
42.56 lps, H = 1.17 m and Nb = 25]. Experimental data taken from Desai (1994).
of this non-uniformity on the impeller performance and 2) influence of impeller motion on

the unsteadiness and its impact on the impeller performance. The resulting non-uniformity
and unsteadiness modify the velocity, pressure and angular momentum flux distributions at
the impeller inlet. The focus here is on the accuracy of URANS vs RANS simulations that
is of significance in improving the performance. This means determining whether dealing
with the azimuthal nonuniformity improves the accuracy of the simulations. The effects
of unsteadiness on the impeller performance was quantitatively assessed by comparing the
URANS solutions with the steady RANS solutions and the experimental data. A variety
of parameters can be used to quantify the effects of unsteady flow in fluid systems, but
according to Greitzer et al. (2007), the magnitude of unsteadiness, due to the azimuthally
non-uniform geometry, is measured by using reduced frequency β, given as:
β = ωL/U (4.4)
64
where ω = angular velocity of the impeller, L = characteristic length of the impeller (blade
spacing), and U = characteristic through-flow velocity or peripheral velocity of the impeller.
Reduced frequency β represents the “change in local flow quantities during the passage
of the fluid particle” (Greitzer et al., 2007). In Equation (4.4), if β 1, unsteady effects
are small; if β 1, unsteady effects dominate; and if β ≈ 1, unsteady effects are important.
Equation (4.4) shows that unsteady effects become important once β rises above a transi-
tional value of the order of unity, i.e., the time-scale of the oscillation or unsteadiness is
comparable to that of flow propagation. According to Greitzer et al. (2007), the appropriate
length scale L for calculating β is the blade spacing or may be entry arc length. For the
7 kW turbine, the characteristic value for β was found to be about 0.25, so that β < 1.
Therefore, unsteady effects should be negligible. In this work, the results of steady and un-
steady RANS computations are compared to the experimental results for the 7 kW turbine
to examine the influence of any potential unsteady effects on the turbine performance. Blade
passing frequency relative to the stationary nozzle is computed by N ( rev/min
60
)× Nb . For
example, the 0.53 kW impeller with 30 blades rotating at 200 RPM has the blade passing
200
frequency of 100 Hz (= 60
× 30).
In crossflow turbines, there is a relative motion between the rotating impeller and the
stationary nozzle, which has a limited extent at the impeller inlet. As a consequence, the
flow at the impeller inlet can be considered as a cyclic flow. Steady RANS solutions deal
with unsteadiness but URANS is needed if there is additional unsteadiness associated with
the azimuthal non-uniformity of the nozzle and walls and the impeller. Since steady RANS
simulation does not capture such unsteady effects, URANS simulations were conducted in
order to examine whether unsteady effects are critical (by comparing the results of steady
and URANS for the power output). The effects on power output and the velocity at the
impeller inlet are important.
To perform URANS simulations, solutions of steady RANS simulations were used as the
initial conditions, which considerably reduced the computational time. To determine the
65
appropriate time-step for the URANS simulations, which is important in determining the
numerical convergence and the computational time in simulating a multiphase flow (ANSYS,
2016), the frequency of the largest eddy was taken as a reference. This frequency can be
approximately estimated from the impeller angular velocity ω as 0.2/ω (ANSYS, 2016). This
time-step value did not produce desired convergence criteria of 5 × 10−5 RMS residual, so
smaller time step of 0.002 sec was used. For the URANS simulations, the time required
for a typical simulation on 96 core parallel computing was about 72 hours for the total
simulation time of 4 sec (e.g. at 450 RPM, there are 7.5 impeller revolutions per sec or 30
revolutions in total) and time-step of 0.002 sec (2000 total time-steps) to achieve numerical
convergence of RMS of 5 × 10−5 . For a typical simulation at 450 RPM of the 7 kW turbine,
the maximum Courant-Friedrichs-Lewy (CFL) number was 3.92. For faster convergence and
numerical stability in the solution, a CFL number less or equal to 1 is recommended (ANSYS,
2016). For this, smaller time-steps and finer grids should be used, but this will substantially
increase the computational cost. In order to reduce the computational time, the simulation
was conducted with the CFL number of 3.92, which was obtained by gradually reducing the
time step as discussed above.
4.3 Key Flow Features and Performance Characteris-
tics
Flow features in this work refers to the flow structures or processes, which have dominant
roles on determining the turbine performance. For example, conversion of head into kinetic
energy in the nozzle and major flow separation on blades of the impeller. Fundamental
understanding of such performance related flow phenomena is important because the flow
conditions developed at the impeller inlet and consequently, within the impeller will affect
the overall turbine performance.
66
It is emphasized here that the analyses are aimed at presenting an overall qualitative
and quantitative picture of the flow mechanisms of power extraction, rather than obtaining
a detailed time-accurate quantitative data in a situation where the basic fluid-dynamical
processes are not yet understood. To make the analysis tractable and of practical value,
the emphasis is on the global features of the flow, for example the velocities and the flow
angles at the inlet and interior regions of the impeller, since it is these global parameters
that are significant for the turbine performance. Flow separation is another important flow
phenomena that limits the impeller performance. The key aim is to minimize the occurrence
of flow separation on the blades. Thus, the detailed features of the boundary layer separation,
including the precise location and development and vortex formation are assumed to be of
secondary importance. Average velocity, the roles of flow angles on flow separation, and
angular momentum extraction by the blades are the basic information necessary to assess
the internal flow and the performance of the turbine.
In this analysis, important flow features in the 7 kW (69% efficiency) and the 0.53 kW
(88% efficiency) turbines are investigated, compared and characterized to gain an under-
standing of what flow structures in the turbine result in lower efficiency and what flow
structures result in higher efficiency. Since the overall essential features of the flow in these
turbines should be similar, it is these features that this study seeks to clarify as well. Only
the results of steady RANS computations are used in this analysis. Average velocity, angu-
lar momentum extraction by the blades, the roles of flow angles and flow separation etc on
the turbine efficiency are investigated. Computations were performed at different operating
conditions as used in the experimental measurements. The analysis is done on the radial-
tangential (r − θ) plane because only the flow on this plane takes part in power extraction.
The conventions used in the analysis are shown in Figure 4.12.
The flow field in a crossflow turbine can be divided into four distinct regimes as follows:
1) the high kinetic energy flow in the nozzle and at the impeller inlet, with some interaction
between the inlet flow on the blades and the impeller motion, 2) the power extraction regime
67
Y
left nozzle lip flow
nozzle throat
first stage
Ψ
θs
Z
right nozzle lip
second stage
Figure 4.12: Schematic illustration of the conventions used in the analysis. The azimuthal
position ψ, nozzle throat, left nozzle lip, right nozzle lip, first stage, and second stage will
be used. ψ will be used to refer to a specific position or a region in the nozzle entry arc and
the first and second stages. Configuration is that of the 0.53 kW turbine.
at the first stage, where the interactions of the flow with the blade surfaces produce a
change in angular momentum to extract the fluid power (converging blade passages radially
inward), 3) the free-stream flow regime between the first and the second stages, i.e. the
interior air-space region of the impeller, where the exiting flow streams from the first stage
blade passages to a single flow stream at atmospheric condition before passing through the
second stage blades, and 4) the second stage, where the interactions of the flow with the
blade surfaces produce a change in angular momentum to extract the fluid power (diverging
blade passages radially outward). These distinct flow regimes are illustrated in Figure 4.13,
where the azimuthal angle has its origin at the left nozzle lip in Figure 4.12. This origin
for the angle is used on all subsequent Figures showing impeller inlet velocities, flow angles,
power production etc. As the impeller is open to atmosphere, there is a similarity between
68
the assumption of total conversion of head into kinetic energy and the matching of the
jet kinetic energy to the head in Pelton turbines. However, the flow through the blades
in crossflow turbines is parallel to the blade surfaces, which is similar to the runner of
the Francis turbine, rather than impinging jets as in the buckets of Pelton turbine. These
designs have very similar considerations. These distinct flow regimes illustrated in Figure 4.14
implies that the flow is highly complex and coupled between the nozzle and the two stages.
For example, the flow conditions at the impeller inlet govern the flow regimes within the
impeller and, consequently its performance. This implies that design improvement requires
an understanding of the flow in these regimes to understand the mechanisms to minimize
losses due to major flow separations on the blades at both stages. It is noteworthy that the
flow at the first stage encounters converging blade passages, whereas the flow at the second
stage encounters diverging blade passages.
The flow condition at the nozzle outlet is important for the impeller performance. How-
ever, this important design aspect has not been investigated in detail while studying the
performance of the impeller by the previous researchers. The flow in the nozzle results into
a predominantly tangential flow at the impeller inlet, which has a strong influence on the
performance of the downstream impeller. Therefore, at the inlet of the impeller, the absolute
velocity and the relative flow angle β1 or the angular momentum flux, are important to assess
the impeller performance. For an ideal nozzle, the absolute velocity u0 at the impeller inlet
√
is equal to the theoretical jet velocity, i.e. 2gH. As the impeller is open to atmosphere,
there is a similarity between this assumption and the matching of the jet kinetic energy to
the head in Pelton turbines (Dixon and Hall, 2013). To assess the nozzle performance or
the quality of the flow produced at the impeller inlet, azimuthal variations of the absolute,
tangential, and radial velocities and inlet flow angle are plotted in Figures 4.15 - 4.19 for
both turbines.
In an ideal turbine operation at maximum efficiency, the inlet flow angle β1 is equal to
the outer blade angle β1b . In the case of 7 kW turbine (69% efficiency), it can be seen from
69
Nozzle:
high velocity flow regime
First Stage:
converging blade passages
Control
Flow volume
boundaries
free-stream,
air-space region
Second stage:
diverging blade passages
Figure 4.13: Schematic illustration of different flow regimes in a crossflow turbines.
Figure 4.14: Contour plot of the magnitude of mean water velocity in the 7 kW turbine
illustrating the main flow path in crossflow turbines.
70
14
12
10
Velocity [m/s] 8 Computed total inlet velocity

6 Radial velocity
Tangential velocity
4 Total inlet velocity
−2
−4
−6
60 70 80 90 100 110 120 130
Nozzle entry arc azimuthal position [deg]
Figure 4.15: Azimuthal variations of the computed absolute, tangential and radial velocities
at the impeller inlet of the 7 kW turbine. Variations across the inlet are due to the position
of the blades. For a time averaged flow - which is assumed in the simple nozzle model, these
variations would average out. [Q = 105 lps and H = 10 m].
Figure 4.15 that there is a significant difference between the computed u0 (≈ 11 m/s) and
√
the theoretical jet velocity ( 2gH = 14 m/s) at the impeller inlet. This implies that the
head H is not fully converted into kinetic energy, which is an important criterion for the
best performance of impulse turbines. As a result, the impeller is receiving significantly
lower angular momentum flux than it is actually possible to obtain. Similarly, at maximum
efficiency, it can be observed from Figure 4.16 that β1 has decreased almost linearly from
75 to 50◦ over the impeller inlet, which is significantly greater than the outer blade angle
(30◦ ). This means that the flow is not parallel to the blade surface. As a result, considerable
boundary layer separation has occurred on all the blades at the first stage as shown in
Figure 4.17. This is because separation on the blades is strongly related to the inlet flow
conditions. This reveals that flow separation is an important loss mechanism in the impeller
of the 7 kW turbine. The flow separates from the blade surface if the wall shear stress
τw becomes zero or negative (τw = µ∂u/∂y). If the pressure coefficient rapidly increases
71
behind the leading edge of the blade, an adverse pressure gradient is generated. As soon
as the pressure gradient becomes too large, the kinetic energy of the fluid will not suffice
to keep the fluid following the blade surface. As a result, it causes an abrupt change in
blade loading. Flow separation is generally accompanied by recirculating flow, and affects
the impeller performance in two ways. First, the reversed flow provides a low-speed region
with high turbulence, which promotes mixing and viscous losses. In addition, a large amount
of vorticity is carried by the separated flow, which will eventually cause performance loss.
Secondly, the blockage created by the flow separation region reduces the effective flow passage
area, and consequently increases the flow velocity. As a consequence, it reduces the power
extraction capability the impeller. Since separation is strongly related to upstream flow
conditions, the results of the 7 kW turbine indicate that the nozzle performance is very
poor, primarily on aligning the flow to the impeller blades, the penalty of which on impeller
performance is expected to be significant. It is noted that if separation occurs in the first
stage, it reduces the quality of the inlet flow at the second stage and thus results into the
performance loss of the second stage. It can be concluded that the partial conversion of head
H into kinetic energy in the nozzle and the flow separation on the blades at the first stage
due to misalignment of the inlet flow with the blades are the primary reasons for the poor
performance of the 7 kW turbine. The influence of the nozzle and impeller designs and their
matching on the impeller performance will be examined in more details in Chapter 6.
In the 0.53 kW turbine (88% efficiency), the computed absolute velocity u0 at the impeller
√
inlet is approximately equal to the theoretical jet velocity, i.e. 2gH = 5.12 m/s, as shown
in Figure 4.18. This implies that H is approximately converted into kinetic energy. However,
similar to the case of 7 kW turbine, there are considerable azimuthal variations of tangential
(uθ ) and radial (ur ) velocities, particularly near the nozzle throat in the azimuthal range:
120◦ ≤ ψ ≤ 160◦ . As a result, β1 varies azimuthally from 75 to 35◦ over the impeller inlet as
shown in Figure 4.19. In addition, it can be seen that β1 is considerably greater than β1b =
39◦ in 120◦ ≤ ψ ≤ 160◦ (ur is high), but is approximately equal to β1b over a larger extent
72
80
Computed angle of attack α
70 Computed flow angle
Designed outer blade angle
60
Flow angle [deg] 50
40
30
20
10
0
60 70 80 90 100 110 120 130
Figure 4.16: Comparison of the inlet flow angle β1 with the outer blade angle at the inlet of
the impeller of the 7 kW turbine at maximum efficiency [Q = 105 lps, H = 10 m and N =
450 RPM].
Figure 4.17: Close view of water velocity vectors illustrating the major flow separation on the
pressure sides of the blades at the first stage of the 7 kW turbine at the maximum efficiency
point [H = 10 m, Q = 105 lps and N = 450 RPM]. Note that there is no separation in the
second stage.
73
6
5
4
Absolute velocity
3
Radial Velocity
Velocity [m/s]
2 Tangential Velocity
1 Theoretical jet velocity
0
−1
−2
−3
−4
−5
120 130 140 150 160 170 180 190 200 210
Nozzle azimuthal position [deg]
at the impeller inlet of the 0.53 kW turbine [Q = 46 lps and H = 1.33 m].
of the impeller inlet. In 120◦ ≤ ψ ≤ 160◦ , β1 has decreased almost linearly from 75 to about
39◦ . It is noteworthy here that in the actual design of this turbine by Desai (1994), it’s likely
that β1 was assumed to be azimuthally uniform. However, this is not explicitly mentioned by
the author. Clearly, the difference between β1 and β1b and the azimuthal non-uniformity of
the inlet flow (i.e. angular momentum flux) do not have a catastrophic lowering of efficiency
in the 0.53 kW turbine. The main reasons for this are 1) H is approximately converted into
kinetic energy, 2) power is extracted in two stages; the inlet flow with high β1 is missed at
the first stage but can still be captured in the second stage, 3) less flow separation due to a
large number of blades, and 4) close matching between β1 with β1b over a large extent of the
impeller inlet. The influence of β1 on the power production in the impeller will be examined
in the next section. The influence of the inlet flow conditions on the flow field characteristics
in the impeller is examined as follows.
The flow patterns in the first stage can be explained as follows. The high velocity flow
passes radially inward through the blades, and as a result, torque is generated due to the
difference in dynamic pressure between the pressure and the suction surfaces of the blade.
The flow is parallel to the blade surface, which is similar to Francis and Turgo turbines. It
74
90
Computed flow angle
80 Designed blade angle
70
Inlet flow angle [deg]

60
50
40
30
20
10
0
120 130 140 150 160 170 180 190 200 210
Figure 4.19: Azimuthal variation of the inlet flow angle β1 at the impeller inlet of the 0.53
kW turbine at the maximum efficiency point: 88% [Q = 46 lps, H = 1.33 m and N = 199.1
RPM].
Figure 4.20: Contours of the magnitude of the mean water velocity in the 0.53 kW turbine
at the maximum efficiency point [H = 1.33 m, Q = 46 lps and N = 199.1 RPM].
75
is noted that a unique design feature of the blade passages in the impeller is the difference
in the shape that the flow encounters at the first and second stages. At the first stage, the
passages are converging, whereas at the second stage, the blade passages are diverging, which
is inherently likely to cause flow separation. This is because the inner to outer diameter ratio
D2 /D1 or the outlet to inlet area of the stage is 0.68. The inlet at the first stage becomes the
outlet at the second stage, and the outlet of the first stage becomes the inlet of the second
stage. Therefore, the flow is similar to a nozzle flow at the first stage, whereas it is similar
to a diffuser flow at the second stage as depicted in Figure 4.13. As a consequence of this
geometrical feature, the flow exits the first stage as a converging flow radially inward and
passes almost diametrically in the free-stream or interior air-space region before it passes
through the second stage. This requires that the flows at the two stages must be matched to
extract maximum power from both stages. It is interesting to note that although the second
stage blade passages are diverging, there is no flow separation in the second stage. This is
due the fact that the flow is converging toward the second stage, and is slightly accelerated
at the entry to the second stage as can be observed from Figures 4.20 and 4.21.
To illustrate the mean flow field in the first stage of the 7 kW turbine, the velocity
vectors were plotted in Figure 4.17. The Figure shows that there is a major flow separation
in the first stage as discussed earlier, which is primarily attributed to a significant difference
between β1 and β1b as well as due to a wider blade spacing or insufficient number of blades.
Similarly, in contrast to the 7 kW turbine, there is negligible flow separation in the first
stage blades of the 0.53 kW turbine as shown in Figures 4.20 and 4.21 despite the fact that
β1 is significantly greater in 120◦ ≤ ψ ≤ 160◦ . This is partly due to the higher mass and
momentum fluxes through 120◦ ≤ ψ ≤ 160◦ and mostly the use of larger number of blades
Nb . It is interesting to note that there is negligible suction side separation near the trailing
edges of the blades near the right nozzle lip: 190◦ ≤ ψ ≤ 210◦ . This is due to the fact
that β1 is slightly less than β1b = 39◦ . The 0.53 kW turbine has relatively narrower blade
spacing (Nb = 30) compared to the 7 kW turbine (Nb = 20) for a comparable size of the
76
Figure 4.21: Water velocity vectors in the first stage of the 0.53 kW turbine at the maximum
efficiency point, illustrating the suction side separation on the blades [H = 1.33 m, Q = 46
lps and N = 199.1 RPM].
impeller outer diameter (304 mm and 316 mm respectively). It can be concluded that the
greater efficiency of the 0.53 kW turbine most likely relies on better nozzle performance,
smaller difference between β1 and β1b over a large extent of the inlet, use of a larger Nb and
an almost fully attached flow. The consequence of this on power extraction behaviour of
the first stage will be examined in the next section. Thus, the significant amount of flow
separation at the first stage of the 7 kW turbine indicates that the nozzle and the impeller
are poorly designed and are not properly matched to align the inlet flow with the blades.
After exiting the first stage, the flow stream passes through the interior air-space region
or free-stream region toward the second stage. In this region, it is seen that streams from
different blade passages have converged to a single flow stream at atmospheric air before
passing through the second stage. The flow is converging by about the factor of inner to
outer diameter ratio (0.67 for the 7 kW turbine and 0.68 for the 0.53 kW turbine). The
direction toward the second stage is mostly governed by the inner blade angle β2b and the
77
180
Computed exit flow angle
140
Flow angle [deg] 120
100
80
60
40
20
0
70 80 90 100 110 120 130 140
Azimuthal position [deg]
Figure 4.22: Computed exit flow angle from the first stage of the 7 kW turbine at maximum
efficiency [H = 10 m, Q = 105 lps and N = 450 RPM]. The local minima are due to the
wakes of each blade.
140 Computed exit flow angle

Inner blade angle
120
Exit flow angle [deg]
100
80
60
40
20
130 140 150 160 170 180 190 200 210 220
Figure 4.23: Exit flow angle at the first stage of the 0.53 kW turbine at maximum efficiency
[H = 1.33 m, Q = 46 lps and N = 199.1 RPM]. The local minima are due to the presence
of wakes of each blade.
78
Figure 4.24: Close view of the water velocity vectors illustrating the main flow pattern in
the 0.53 kW turbine at maximum efficiency. Note that the flow is almost fully attached in
both stages. Blade wakes from the first stage can also be seen. The flow at the second stage
is also aligned with the blades and is slightly accelerated in the blade passages [H =1.33 m,
Q = 46 lps and N = 199.1 RPM].
impeller speed. In the 7 kW turbine, it is observed that β2 is considerably greater than

the β2b (90◦ ) by more than 15◦ and it is gradually increasing toward the right nozzle lip as
shown in Figure 4.22. This means that the flow is highly deflected because β1 is decreasing
toward the right nozzle lip. The difference between β2 and β2b is called the deviation angle.
Despite this significant deviation angle, there is not much separation around the trailing edge
because the flow is converging in the blade passages as well as in the interior free-stream
region. Likewise, in the 0.53 kW turbine, β2 is greater than β2b (90◦ ) and the deviation
angle is about 10◦ as shown in Figure 4.23. It is observed that β2 is uniform in 130◦ ≤ ψ
≤ 190◦ , but there is a significant increase near the right nozzle lip in 190◦ ≤ ψ ≤ 210◦ . As
will be discussed in the next section, this is the region at the first stage where a significant
amount of power is extracted from the flow. Thus, a highly deflected flow with a significant
streamline curvature can be observed in the interior region, which can be attributed to the
consequence of small β1 and significant power extraction from the flow.
79
As shown in Figure 4.24, wakes can be seen just downstream of the blade trailing edges.
The wake is a region of non-zero vorticity and low velocity downstream of the blade. In the
0.53 kW turbine, the size of the wake increased toward the right nozzle lip. This is because
the difference between β2 and β2b is increasing. Similarly, the size of wakes is increasing
toward the right nozzle lip in the case of 7 kW turbine because the difference between β1
and the inner blade angle is increasing, and the effects are more pronounced in this turbine
due to separation in the first stage. This implies that there are relatively more viscous losses
in the wakes of the 7 kW turbine compared to the 0.53 kW turbine. The reason is that the
0.53 kW turbine has more blades and less separation in the first stage. However, a detailed
quantitative analysis was not performed as the design goal is to minimize them. These results
imply that the size of wakes can be reduced by minimizing the difference between β2 and β2b .
The velocity distribution in the wakes is complex. The spikes, i.e. the sharp rise and fall
of values or maxima and minima of values, in the plots of β2 are due to the wakes, but far
downstream of the blade trailing-edges, the wakes must decay and the streamlines become
approximately parallel. Long spikes means that the wakes are significant. The wake decay is
monotonic, i.e. there will always be a deficit on entry to the second stage; the issue is what
is the minimum size of this that affects second stage performance?. In the case of 0.53 kW
turbine, a short wake, whose velocity defect decays shortly downstream, can be seen. This
is closely associated with fully attached flow or negligible flow separations on the blades. In
contrast to this, the wake length is significant in the case of 7 kW turbine, and thus requires
design modifications to address flow separation. This can be addressed by better matching
of nozzle and impeller design, increasing the number of blades and matching β1 and β2 with
β1b and β2b respectively.
In the free-stream region, streams from different blade passages mix at ambient air pres-
sure because the turbine is open to the atmosphere. In this region, the flow may also pass
around the shaft and lose some energy, but as shown in the case of 0.53 kW turbine, the
mean flow path is above the shaft center. As such, there is only negligible loss. As the
80
180
Computed inlet flow angle
140
Flow angle [deg] 120
100
80
60
40
20
0
150 160 170 180 190 200
Figure 4.25: Inlet flow angle β2i at the second stage of the 7 kW turbine at maximum
efficiency [H = 10 m, Q = 105 lps and N = 450 RPM]. Note that the local minima due to
the wakes of the first stage blades are decayed by the entry to the second stage.
140 Computed inlet flow angle

Designed blade angle
120
Exit flow angle [deg]
100
80
60
40
20
230 240 250 260 270 280 290

Figure 4.26: Inlet flow angle β2i at the second stage of the 0.53 kW turbine at maximum
efficiency [88%]. Note that the effects of wakes have decayed at the inlet of the second stage.
Note that the local minima due to the wakes of the first stage blades are decayed by the
entry to the second stage.
81
individual velocities and relative angles between them are different, which is evident from
the variation in β2 in both turbines, the flows cannot converge to a flow of uniform velocity
across its cross-section. More precisely, velocity gradients occur across the cross-section of
the converged flow. This means that the flow at the inlet of the second stage has significant
azimuthal variations of both the velocity and the flow angle as depicted in Figures 4.25 and
4.26. In both turbines, a very similar trend in the azimuthal variation of the flow angle at
the entry of the second stage can be seen.
In the 0.53 kW turbine, considerable streamline curvature of the flow can be seen at the
upper region of the free-stream flow due to significantly greater exit flow angle β2 than β2b
(90◦ ), whereas the flow has almost straight streamlines around the center of the impeller,
with about β2 ≈ 90◦ . This lower portion of the flow has greater mass and momentum fluxes
than at the upper region. The inlet flow angle β2i at the second stage is the flow angle that
determines second stage performance. If β2i does not match β2b , there will be losses due to
flow separation or it can even create negative torque (if β2i is greater than β2b ). Thus, β2i
at the entry to the second stage is crucial because the flow at the second stage is confined
to a narrow region, and if β2i is greater than β2b it would result into negative power. This
has been schematically illustrated in Figure 4.2 in Section 4.1. It is noted that β2i is mostly
governed by β2b and the impeller speed. Figure 4.26 reveals that β2i is greater than β2b after
ψ = 260◦ , where the blades produce negative power. Similarly, Figure 4.25 shows that after
ψ = 185◦ , the blades produce negative power. These aspects will be discussed in the next
section.
As the flow is converging at the entry to the second stage, the flow has accelerated slightly
in the second stage as dictated by continuity. As a result, the flow is not separated even if
the second stage blades are diverging as can be observed from Figure 4.24. As shown in the
water velocity contour maps in Figure 4.20 and β2i in Figure 4.26, it is observed that the
flow passing through 120◦ ≤ ψ ≤ 160◦ runs straight through the impeller centre and passes
through the second stage blades in 90◦ ≤ ψ ≤ 130◦ , where it produces the maximum power.
82
It is also observed that in 90◦ ≤ ψ ≤ 130◦ , β2i is less than 90◦ . A more interesting feature
about the development of the flow in the free-stream region can be observed when β2 and
β2i are compared. Clearly, β2i is directly related to β2 , which means the relation between
them determines the power production in the second stage as will be discussed in the next
section.
The development of the flow patterns at different impeller speeds at the second stage of
0.53 kW turbine is shown in Figures 4.27 and 4.28. At the upper inlet region of the second
stage: 230◦ ≤ ψ ≤ 245◦ , the streamlines are highly curved, and so β2i is relatively small,
i.e. about 25◦ , whereas the β2b is 90◦ . As a result, separation has occurred on the suction
side of a blade. Close examination of the velocity contours at different speeds reveals that
the size of suction side separation has decreased with the increase in impeller speed. This
is due to the fact that more flow has been deflected upward. The flow is aligned with the
blades at around 199 RPM, which is the maximum efficiency point. The contour plots reveal
that there is a strong recirculating flow within the blade passage at the upper region of the
second stage, which may be called the “inter-blade vortex”, and is due to the trapped flow
in the blade passage carried from the first stage. Experiments have also confirmed that some
flow is trapped within the blade passage from the first stage and is dragged downstream
by the blades (Durgin and Fay, 1984). This trapped flow, which looks similar to the fluid
motion in a rotating channel, is convected downstream once it meets the incoming flow and
a favourable pressure gradient is developed, and so cannot last long.
The velocity contours show that the flow at the entry to the blade passages of the second
stage is accelerated, and is particularly on the suction sides. Note that this is due to the
converging flow coming out of the first stage. This type of accelerating flow, where pressure
decreases in the flow direction, is called a favourable pressure gradient 2 , and is highly
desired for improved blade loading or extracting more power. Acceleration on the suction
side reduces the susceptibility to flow separation. Therefore, there is no flow separation in
2
strong favourable pressure gradient prevents major separation
83
(a) 160 RPM (b) 186 RPM
Figure 4.27: Streamlines superimposed on the velocity contour at the second stage of the
0.53 kW turbine illustrating flow alignment to the blades.
(a) 199 RPM (b) 224 RPM
Figure 4.28: Streamlines superimposed on the velocity contour at the second stage of the
0.53 kW turbine illustrating flow alignment to the blades. The impeller speed 199 RPM
corresponds to the maximum efficiency point.
the second stage in both turbines. As β2i has significant azimuthal variation at the inlet of
the second stage, power production in this stage is also varying azimuthally. However, as
β1b is usually selected to match β1 at the impeller inlet, the only option to change β2i is by
modifying β2b . It is reiterated that at the first stage, most of the power has been extracted
where β1 approximately matches to the outer blade angle. Streamline plots at the second
stage of the 0.53 kW turbine, shown in Figures 4.27 - 4.28, reveal that streamlines are mostly
84
aligned to the blades at the second stage, despite highly curved streamlines at around one
blade position at the upper region, where the flow separates due to small β2i . This flow
separation cannot be avoided as it has occurred at around the water-air free-surface (water
is only partially covering the blades), and mainly involves the trapped flow. It is also seen
that the flow has greater β2i than β2b (90◦ ) after ψ = 260◦ . This means that the blades
passing after ψ = 260◦ are losing power to the flow. This means that β2i has a significant
role in the blade performance. Thus matching the flow angles between the first and second
stages to avoid flow separation and the production of negative power lies at the heart of
improving the power produced by the second stage. Since β1b is fixed by first stage efficiency
considerations, only the first stage outlet/second stage inlet angle (inner blade angle) can be
altered. In the case of 7 kW turbine, the plot of radial and tangential velocities depicted in
Figure 4.29 show that the tangential velocity is almost negligible, implying that the residual
angular momentum of the flow leaving the impeller is minimum. As the flow is confined to
a narrow region at the second stage, one way to producing more second stage power is to
increase the number of blades. Also, the increase in the number of blades should, according
to Totapally and Aziz (1994), increase the power production from the first stage until the
optimal number reached, but how it affects the second stage is not known.
4.4 Power Production Mechanism - Stage Performance
It is reasonable to assume that power production in crossflow turbines takes place by two
flow mechanisms: 1) conversion of head into kinetic energy and maximization of angular
momentum flux at the impeller inlet, and 2) extraction of angular momentum due to the
interaction of the flow with the rotating blades in two stages. As the impeller is open to
atmosphere, there is a similarity between this assumption and the matching of the jet kinetic
energy to the head in Pelton turbines. However, the flow through the blades in crossflow
turbines is parallel to the blade surfaces, which is similar to the runner of the Francis turbine,
85
8
Velocity [m/s]
4
Radial velocity
2 Tangential velocity
−2
150 155 160 165 170 175 180
Figure 4.29: Azimuthal variation of tangential and radial velocities at the exit of second
stage of the 7 kW turbine at maximum efficiency [H = 10 m, Q = 105 lps, N = 450 RPM].
Note that the tangential velocity is small, implying that the residual angular momentum
exiting the second stage is minimum.
rather than impinging jets as in the buckets of Pelton turbine. The significance of these flow
mechanisms in turbine performance is investigated here.
The flow structure in 1) is essential because the flow into the impeller is atmospheric,
which is similar to the jet of a Pelton turbine, and any remaining pressure cannot be converted
into power, and thus the flow with maximum angular momentum flux should be directed to
the impeller blades at a suitable flow angle, so that there will be negligible flow separation
on the blades. In principle, an inlet flow of rectangular cross-section with high velocity and
uniform mass and angular momentum fluxes is required so that maximum power can be
extracted. However, the first stage does not convert all available energy into power. This
implies that there may be an optimum inlet flow angle β1 that results in a flow condition
at the first stage, which contributes significantly in power extraction. Since the outer blade
angle β1b is fixed by the impeller inlet flow, only the inner blade angle β2b can influence the
second stage performance.
Maximum torque occurs when there is a maximum conversion of inlet angular momentum
into torque and by having negligible angular momentum leaving the outlet. Maximum power
86
is produced when that torque is generated at an angular velocity that allows the conversion
of all H into kinetic energy at the impeller inlet. After passing through the first stage blades,
the streams from different blade passages at the first stage are combined in the interior air-
space region, where it is developed approximately as a single flow stream, and passes again
through the second stage blades. In principle, each blade should receive uniform mass and
angular momentum fluxes in order to extract maximum power from the flow. The situation
for crossflow turbines is complicated because power is extracted in two stages and the flow
has to pass through the second stage. If all the power is extracted from the first stage, the
second stage may well be unable to merely pass the flow without losing power. In addition,
the flow in the interior region of the impeller can not be diverted along the axis of the shaft
to bypass the second stage. Thus, there should be some optimal flow passage that would
allow maximum power extraction from both stages, which is strongly related to the flow
angles at the inlet of the impeller, exit of the first stage, and the inlet of the second stage.
The performance of the first and second stages were examined at various operating con-
ditions for both turbines. As the flows in the first and second stages are coupled, it follows
that an optimum flow passage (from inlet to the exit of the impeller) must be established to
extract maximum power. By generalising the results for the inefficient 7 kW turbine, it is
reasonable to assume that such an optimal flow passage results into minimum flow separation
on the blades. As discussed in Section 4.3, a significant amount of separation occurred in the
first stage of the 7 kW turbine. This resulted in a highly non-uniform azimuthal flow entering
the second stage. As a result, the second stage performance deteriorates as the flow enters
at improper flow angles. This is crucial because the impeller blades at the second stage may
extract or lose power depending upon the magnitude and direction of the relative velocity
wθ as illustrated earlier in Figure 4.2. Flow matching, which means matching the flow angles
at the impeller inlet, exit of the first stage, and the inlet of the second stage, determines
how much power is extracted from the first stage and how much from the second stage.
Potential performance benefit results from a reduction in flow separation in the first stage
87
2
250 RPM
300 RPM
1.5 350 RPM
400 RPM power loss
450 RPM
Power [kW] 1 500 RPM
550 RPM
0.5
0
power loss
first second
−0.5 stage stage
−1
60 80 100 120 140 160 180 200
Figure 4.30: Azimuthal variation of power production at different impeller speeds for the 7
kW turbine [H = 10 m, Q = 105 lps]. Impeller speeds 400 - 450 RPM correspond to the
maximum efficiency. Note that the negative values represent the power lost by the blades
into the flow. The data values represent blade positions.
and reduction of angular momentum increase (loss) in the second stage (see Figure 4.2). A
number of challenges exist in the design and performance assessment of such configurations.
One is that definition of the requirements for flow matching becomes more difficult because
the concepts of flow matching between the first and second stages, via β1b and β2b , involves a
number of other impeller geometrical parameters. Nevertheless, the key parameters to tailor
the flow would be outer and inner blade angles, inner to outer diameter ratio and number
of blades. This will be investigated in Chapter 6. The results now presented provide the
background to that investigation.
The power extraction behaviour in the impeller of the 7 kW turbine is shown in Figure
4.30. The Figure shows the power produced per blade, which means that as the blades
rotate past the entry arc, the power production will vary azimuthally as shown in the Figure.
The power production was computed by using the control volume method for each blade.
Similarly, the percentage contribution from each stage is shown in Figure 4.31.
88
80
250 RPM
300 RPM
Total power production [%]

70
350 RPM
60 400 RPM
450 RPM
50 500 RPM
550 RPM
40
30
20
10
0
1 2
Impeller stage
Figure 4.31: Percentage contribution to the total power production by the first and second
stages of the 7 kW turbine.
It is observed from Figure 4.30 that there is a significant azimuthal variation in power
production in the first stage of the 7 kW turbine. This is due to the fact that β1 is azimuthally
non-uniform as well as is greater than β1b near the nozzle throat: 60◦ ≤ ψ ≤ 80◦ (see Figure
4.16). As a result, the power production in this section is low compared to the remaining
portion of the first stage, where there is less difference between the inlet flow angle β1 and
the outer blade angle β1b . As a result, the size of flow separation on the blades has also
decreased. Similarly, there is a significant azimuthal variation in the power production in
the second stage. This is also due to the azimuthal variation in the flow angle β2i at the
entry of the second stage (see Figure 4.25). Despite the fact that flow is highly separated
in the first stage and there is a significant variation in β2i at the entry to the second stage,
there is no flow separation in the second stage (see Figure 4.17) because the flow exiting the
first stage is converging and is confined to a narrow region. The flow angle can be related
to the performance of individual blade as they rotate around the impeller center.
It is interesting to note that as the speed increases above the maximum efficiency (450
RPM) [see Figure 4.30], β1 is increased (see Equation 4.3). As a consequence, the blades in
the first stage produce negative power as shown in Figure 4.30, particularly in 60◦ ≤ ψ ≤
89
80◦ , where β1 is very high. After the impeller speed N = 450 RPM, β1 is too high or the
relative velocity wθ becomes negative that the blades lose power to the flow. This is the
main reason that the impeller is inefficient above 450 RPM; power production increases due
to the decrease in β1 upto the value of outer blade angle β1b . At N = 450 RPM, the power
extraction is almost linearly increasing over the entire azimuthal range. At lower speeds N
= 200 - 300 RPM, the power extraction increased only slightly toward the right nozzle lip
for 80◦ ≤ ψ ≤ 130◦ . At very high impeller speeds, say above N = 450 RPM, the power
production has sharply increased in the azimuthal range: 80◦ ≤ ψ ≤ 130◦ , and is increased
as N increases. The power extraction increased toward the right nozzle lip, defined in Figure
4.12, because the difference between β1 and β1b is small at all impeller speeds and the exit
flow angle β2 is significantly greater than β2b . Therefore this is the most efficient azimuthal
region in the first stage. This results into efficient power extraction from the first stage, but
β2 significantly greater than β2b results into small inlet angle β2i at the entry to the second
stage. This causes suction side separation in the second stage, but its effect will be only
marginal on power production.
In the second stage, there is a significant variation in the inlet flow angle β2i compared
to the inlet flow angle β1 at the first stage. As a result, there is a significant azimuthal
variation in the power production. At N = 250 RPM, it is observed that in 150◦ ≤ ψ ≤
170◦ , where β2i is significantly greater than β2b , the blades lose power to the flow. After
ψ = 170◦ , the power production increases almost linearly. As N increases from 250 RPM,
the power production in ψ = 150 - 170◦ increased sharply. At maximum efficiency, power
production is significant in ψ = 150 - 185◦ . After N = 450 RPM, the blades lose power after
ψ = 190◦ , and is very small. This means that after N = 450 RPM, the blades are losing
power to the flow at both stages, but the loss in the first stage is greater than in the second
stage. This is the main reason why the impeller has a poor performance and its maximum
efficiency is limited at N = 450 RPM.
90
400
100 RPM
160 RPM
300 173 RPM
186 RPM power loss
194 RPM
200 199 RPM
Power [W]
211 RPM
224 RPM
100
329 RPM
first second
−100 stage stage
−200
120 140 160 180 200 220 240 260 280
Azimuthal blade position [deg]
Figure 4.32: Azimuthal variation of power production at different impeller speeds for the
0.53 kW turbine [H = 1.337 m, Q = 46 lps]. N = 199.1 RPM corresponds to the maximum
efficiency. Note that the negative values represent the power lost by the blades into the flow.
The symbols represent blade positions.
Figure 4.31 shows that the percentage power production from the first stage is greater
than the second stage at lower impeller speed. For example, at N = 250 RPM, the first
stage contributes about 72% and the remaining 28% comes from the second stage. At the
maximum efficiency point N = 450 RPM, the first stage contributes about 63% and the
remaining 37% comes from the second stage. At N = 550 RPM, the power production in
the first stage is about 52%. These results reveal that β1 determines the percentage power
contribution of the first stage, which in turn influences the second stage. It can be concluded
that β1 greater than β1b results in less power extraction, and limits the performance of the
first stage. Therefore, the selection of optimum β1 is crucial.
The power production in the impeller of the 0.53 kW turbine is shown in Figure 4.32.
Similarly, the percentage power contribution of each stage is presented in Figure 4.33. It is
observed that the power production is almost uniform over a large extent of the first stage
for N = 100 - 224 RPM. The power production gradually increased only in ψ = 200 - 215◦ ,
where β1 is slightly less than the outer blade angle, otherwise it is uniform for ψ = 120 -
91
160 RPM
70

173 RPM
60 186 RPM
194 RPM
50 199 RPM
211 RPM
40 224 RPM
30
20
10
0
1 2
Impeller stage
Figure 4.33: Percentage contribution to the total power production by the first and second
stages at different impeller speeds of the 0.53 kW turbine.
200◦ . Compared to the 7 kW turbine, which has a significant azimuthal variation of power
production in the first stage, the power extraction in this turbine is more uniform over a
large extent of the first stage. This uniformity is due to the fact that β1 is greater than the
outer blade angle over a small section of the inlet: ψ = 120 - 150 (see Figure 4.19), and more
importantly, there is negligible separation despite β1 is greater than the outer blade angle in
this section. Moreover, the 0.53 kW turbine has more blades, which contributes to a more
uniform power production. It is noted that in the most efficient azimuthal region: ψ = 200 -
215◦ , β1 is slightly smaller than β1b and β2 is significantly greater than β2b . This flow results
in efficient power extraction in the first stage. The consequence of high β2 is flow separation
at the second stage because of significantly smaller β2i compared to β2b . A high β2 means
small inlet flow angle β2i at the second stage. This results in poor blade performance, but is
not critical because most of the power is extracted in the first stage. The only issue is that
it must pass through the second stage at least without losing power to the flow.
In the first stage, it is interesting to observe that at a very high N = 329 RPM, the power
loss is significant between ψ = 120 - 175◦ , and the power production is confined to a very
narrow region: ψ = 180 - 210◦ . This behaviour can be explained as follows. Increasing N
92
increases β1 , which means that the difference between β1 and β1b increases. This results in
increased separation and loss in power production. At very high N , such as 329 RPM, the
relative velocity wθ (= uθ -ωR1 ) at the impeller inlet becomes small (note that the tangential
velocity uθ is already small here), which results in power loss. However, at the right nozzle
lip, where β1 is small, wθ (= uθ - ωR1 ) is significantly higher and positive (note that uθ is
high). As a result, there is substantial amount of power production in this region.
In the second stage, power production takes place over a very narrow region: ψ = 220 -
270◦ compared to the first stage. As is evident from the Figure, there is a substantial increase
in power production with increasing N in ψ = 220 - 235◦ . There is a very narrow region ψ
= 235 - 245◦ , where substantial power is produced in ψ = 245 - 255◦ . After ψ = 255◦ , there
is loss of power for higher N , but the power production has sharply increased for lower N
(e.g. 100 RPM), i.e. no power loss at lower N . By looking at the plots of β2i at the entry
of the second stage and the plots of water velocity streamlines shown in Figures 4.27 and
4.28, it is evident that β2i is greater than β2b = 90◦ , which means that the relative velocity
wθ is opposing the impeller motion. As a result, the blades contribute negative power in
this region. At lower N , β2i is less than 90◦ and wθ is in the same direction as that of the
impeller rotation, and thus the flow contributes positive power. Thus, it is observed that
sharp changes in power production and loss take place within a narrow region in the second
stage, which may be useful from the viewpoint of design optimization.
It can be observed from Figure 4.33 that the percentage contribution to the total power
output from the first stage is greater than the second stage at all impeller speeds. In contrast
to the 7 kW turbine, the power production from the first stage is at least 65% at the lowest
impeller speed. At maximum efficiency, i.e. 199 RPM, the first stage contributes about
69% and the second stage contributes about 31%. At higher impeller speeds, the first stage
contributes less than 69%. Amongst the cases, the maximum contribution of the first stage
was found to be 70% at N = 173 RPM. It can be concluded that β1 greater than β1b results in
less power extraction from the first stage. In summary, even though there is a large difference
93
in the efficiency of these two turbines, there is only a small difference in the percentage power
contribution of the second stage at the maximum efficiency points. However, the first stage
performance of the 7 kW turbine is relatively poor compared to the 0.53 kW turbine, and
needs to be redesigned for improving its efficiency.
4.5 Implications to Nozzle Design
The nozzle design plays an important role in establishing the impeller performance at both
design and off-design conditions as it converts head to kinetic energy and directs the flow
to the impeller at a suitable angle. Therefore, depending on the design of nozzle and its
matching with the impeller, it can be the component limiting the turbine performance. Past
studies have focused mostly on impeller designs, and little is known about the influence
of nozzle design and its matching with the impeller to obtain higher performance. In the
previous sections, quantitative and qualitative analyses were carried out to illustrate the
main features connected with the inlet flow conditions and the impeller performance. In
contrast to what has been supposed in the nozzle design, the resulting inlet flow appeared to
be non-uniform azimuthally over a large extent of the impeller inlet, which does not appear
to have been recognized previously. More importantly, in the case of 7 kW turbine (69%
efficiency), the nozzle was unable to convert all the head into kinetic energy. Therefore, this
design aspect must be investigated, particularly in computing the nozzle design parameters
so that they meet conditions of an ideal nozzle. The calculations of the 0.53 kW turbine
showed that the nozzle is approximately optimal in terms converting H into kinetic energy.
The results also indicate that azimuthal variation of β1 was not very critical at least to
achieve 88% efficiency, provided that all the head is approximately converted into kinetic
energy and there is no separation on the blades. So these two design aspects are useful in
analyzing the impeller performance. It is noted that there is no power extraction in the
nozzle passage, other than negligible viscous dissipatio, and the nozzle contains converging
94
flow which is usually better behaved than expanding flow in a diffuser. It mainly involves
converting all the head into kinetic energy and directing the flow at a suitable angle to the
impeller, which is constant over the entry arc. Therefore, an inviscid, two-dimensional (2-D)
analysis can be used because the dominant effects are inertial in nature, viscous losses are
negligible, and the flow is mainly confined to tangential and radial directions.
4.6 Implications to Impeller Design
The implications of the flow patterns within the impeller observed in the computed results
to impeller design are summarized as the following.
1. The massive flow separations on the blades of the 7 kW turbine showed that flow
separation is the result of mismatch between the inlet flow condition and the impeller design.
This implies that the blades must be designed to match the inlet flow so that flow separation
and performance loss can be avoided. Moreover, the fully attached flow in the 0.53 kW
turbine implies that in order to minimize flow separation, optimal number of blades must
be selected. The calculations of 0.53 kW showed that azimuthal variation of β1 is not very
critical at least to achieve 88% efficiency, provided that all H is approximately converted
into kinetic energy and there is no separation on the blades.
2. The impeller design should be aimed at reducing the amount of counter torque at the
second stage. This can be done by deflecting the flow appropriately within the blade passage
in such a way that β2i is not greater than 90◦ at the entry of the second stage. This can be
attempted by modifying the β1b and β2b , number of blades, diameter ratio etc. Sensitivity
analysis provides guidance toward choosing optimal parameters to partly address this design
problem.
95
4.7 Cavitation
Cavitation is a common flow phenomena in most hydraulic turbines and has the potential
to cause vibration, blade surface damage and performance loss. Cavitation occurs when the
local fluid pressure falls below the vapor pressure of water due to the dynamic conditions.
Although cavitation in most hydraulic turbines, such Francis, Kaplan and Pelton, has been
extensively investigated by a number of researchers due to its strong impacts on turbine
vibration, blade surface erosion and performance loss, it has not been studied in any detail
in the case of crossflow turbines. In this work, the cavitation inception is examined based on
the grounds that mapping a useful cavitation-free operating range is more important than
the description of its full development. To investigate the basic mechanisms of cavitation
inception, the computed results of the steady RANS computations with two-phase fluid
model (water and air) are used. At the cavitation inception, the pressure field can be
computed with two-phase flow model, however, in the cavitation region, three-phase model
is required. Since the presence of cavitation bubbles is usually accompanied by turbulent
three-phase (water, water-vapor and air) interaction with complex dynamic behaviour, high
fidelity computational models as well as multi-phase model are needed to characterize the
flow field. Therefore, the present work is focused only on predicting the cavitation inception.
4.7.1 Cavitation Inception
This section provides a brief overview of the methods to determine the onset of cavitation. It
is reiterated here that the emphasis is on the conditions under which cavitation begins, rather
than its full development, on the grounds that mapping the boundary between cavitating
and non-cavitating flow is the most important design requirement if it can be guaranteed
that turbine performance in the non-cavitating regime includes the peak performance point.
Cavitation is a two-phase (water and water-vapor) interaction that involves vaporization
of water and condensation of water-vapor. In crossflow turbine, if it is present, cavitation
96
involves three-phase interaction (water, water-vapor and air). Vapor bubbles are formed
when the local pressure falls below the vapor pressure of water (3.17 kPa at 25 ◦ C) (White
and Corfield, 2006). Cavitation primarily occurs in separated, accelerated, and recirculating
flow regions near the rotating surfaces (Escaler et al., 2006), such as near the impeller
blades of the crossflow turbine. Local flow acceleration and the recirculating fluid elements
(in vortex regions) cause a decrease in local pressure. Cavitation is characterized using a
dimensionless parameter, called cavitation number (σ), which is defined as (Brennen, 2005):
p∞ − pv
σ= 1 (4.5)
ρ W2
2 w ∞
where p∞ (= patm + ρw gH) and pv are respectively the reference and saturated water-vapor
pressures in the upstream flow, i.e. inlet of the impeller, patm is the atmospheric pressure, ρw
is the density of water and W∞ is the reference relative velocity in the upstream flow (inlet
of the impeller). Equation (4.5) shows that cavitation is directly related to the ratio of drop
in local pressure head and the dynamic head. Note that the dynamic head is closely related
to the inertia of the local fluid motion, influenced by the impeller rotational speed N . For
a given geometry and the flow conditions, the inception of cavitation (σi ) occurs when σ is
equal to the minimum value of the coefficient of pressure Cpmin (Brennen, 2005). Here the
pressure coefficient Cp is defined by:
p − p∞
Cp = 1 (4.6)
ρ W2
2 w ∞
where p and W∞ are respectively the local pressure on the blade surface and the relative
velocity of water at the impeller inlet. So the condition for the inception of cavitation (σi )
is written as:
σi = −Cpmin (4.7)
97
Y
Cp
x/c
pressure side
suction side
Cpmin
cavitation
Figure 4.34: Schematic illustration of pressure distribution on a cavitating blade in the

impeller.
It is noted from Equations (4.6) and (4.7) that further reduction of σ from σi increases
cavitation. Cpmin is an important parameter in the design as it is related to hydrodynamic
loading of the blades and can be used as a criterion (Cpmin + σi > 0) to avoid cavitation.
Furthermore, Cpmin depends on the blade geometry (or impeller geometry) and the Reynolds
number. To avoid cavitation, Cp on the blades can be increased by modifying the geometry
of the blades and their number. The distribution of Cp on a cavitating blade in the impeller
is schematically illustrated in Figure 4.34. As Cp is reduced to σi , cavitation bubbles form on
the low-pressure surfaces, which generally start at the leading edge, of the impeller blades and
when they move into the higher-pressure regions, they collapse implosively. If the pressure in
the neighbourhood of the cavity rises above the vapor pressure, the cavity collapses and it is
often heard as a loud noise in the turbine. As a result, performance is significantly reduced.
In addition to performance deterioration, continuous collapse of cavities can rapidly erode
the blade surfaces, produce vibration and eventually destroy them.
Using the same computational model as the rest of this chapter, inception of cavitation
can be predicted based on the criteria of Equation (4.7).
98
4.7.2 Results
The pressure distributions on the impeller blades were examined around the maximum effi-
ciency points for different Q and H. The first series of computations were performed at Q
= 105 lps and H = 10 m at different impeller speeds (N = 400, 450, 500 and 550 RPM)
to examine the effects of impeller speed N on the onset of cavitation at constant Q and H.
Similarly, the second series of computations were performed at different Q (94, 73 and 56
lps), H (8, 5 and 3 m) and N (300 - 450 RPM) to examine the combined effects of reducing
Q and H on cavitation inception. Finally, three computations were performed at different H
(8, 10 and 12 m) and keeping Q and N constant at Q = 105 lps and N = 450 RPM respec-
tively to determine the effects of H on cavitation inception. By examining the computed
pressure distributions on the blades using contour plots, cavitation inceptions were observed
at several operating conditions.
The contours of absolute pressure distributions for different Q, H and N are shown in
Figures 4.35, 4.36 and 4.37. The examination of the absolute pressure distributions for Q =
105 lps and H = 10 m at different N (400, 450, 500 and 550 RPM) showed that cavitation
inception occurred at N = 450 RPM in the second stage. It is noteworthy that cavitation
begins in the second stage with little separation whereas there is massive separation in the
first stage. Figures 4.35 and 4.36 show that the size of cavitation region has increased with the
increase in impeller speed N . A general explanation for these differences is that inertia of the
local fluid motion is increased (or reduction in local fluid pressure) with the increase in blade
tip-speed as delineated by the Equations (4.5) and (4.6). Similarly pressure distributions at
lower Q and H were examined. A comparative analysis of these results showed that the size
of cavitation region increased at Q = 94 lps, H = 8 m and N = 450 RPM when compared
to the cavitation at Q = 105 lps, H = 10 m and N = 450 RPM or at lower Q and H. As
shown in Figures 4.35 and 4.36, cavitation inception has occurred at the blade tip on the
suction side of one blade at N = 450 RPM and cavitation has increased on the same blade
99
cavitation
Figure 4.35: Contours of pressure distributions on the impeller blades at the second stage [Q
= 105 lps and H= 10 m] at the impeller speed 450 RPM, showing the evidence of cavitation
inception on the suction sides near the inner edge of the blade. It is noted that the scale
of the contour is only for the cavitation region, and the pressure outside this region (red) is
significantly higher than the vapor pressure.
at N = 550 RPM, covering a larger potion of the blade section. At Q = 56 lps and H = 3
m, cavitation was not found at any impeller speed.
As discussed in the previous section, the inlet head H is important in determining the con-
dition for cavitation in the turbine because it affects the cavitation number σ (see Equations
(4.5) and (4.6)). To examine the effects of H on cavitation at Q = 105 lps and N = 450 RPM
(corresponding to the maximum efficiency point), three simulations were performed at H =
8, 10 and 12 m. The results showed that cavitation increased with the reduction in H from
10 m to 8 m, whereas cavitation was not found at H = 12 m. The corresponding pressure
distributions are shown in Figure 4.37. The results of various simulations are summarized
in Figure 4.38 as a performance map of the turbine, which would be useful in selecting the
100
cavitation
Figure 4.36: Contours of pressure distributions on the impeller blades at the second stage [Q
= 105 lps and H= 10 m] at the impeller speed 550 RPM, showing the evidence of cavitation
inception on the suction sides near the inner edge of the blade. It is noted that the scale
of the contour is only for the cavitation region, and the pressure outside this region (red) is
significantly higher than the vapor pressure.
suitable values of Q, H and N to avoid cavitation. The results show that cavitation occurs
after the maximum efficiency operating points, which is a highly desirable result. In all the
cases considered above, it is noteworthy that the cavitation inception regions are confined
to narrow tip regions, primarily on the suction side near the inner edge of the blades at the
second stage.
As the flow field contains water and air with free-surface effects, it might be possible that
cavitation bubbles are strongly influenced by the air content in the water stream (Batchelor,
2000). The air volume fraction may be significant in the water stream in the second row of
blades as the flow velocity is reduced at the first stage and the turbine operates at atmo-
spheric condition. A free-surface (water and air interface) can be seen in the second stage
101
cavitation
cavitation
Figure 4.37: Contours of pressure distributions on the impeller blades at the second stage
at Q = 105 lps, H = 8 m and N = 450 RPM, showing the evidence of cavitation inception
on the suction sides near the inner edge of the blades. It is noted that two blades have
cavitation at H = 8 m, whereas only one blade has cavitation at H = 10 m. It is noted that
the scale of the contour is only for the cavitation region, and the pressure outside this region
(red) is significantly higher than the vapor pressure.
as illustrated in the Figures presented in a previous section. It has been shown by previous
studies that the erosive power of such cavitation is relatively weak (Escaler et al., 2006;
Batchelor, 2000). Nevertheless, this needs further investigation employing a more accurate
three-phase and cavitation models and turbulence simulation techniques, such as detached
102
9
Experimentally determined performance map
8 CFD predicted cavitation region
69%
65% 65%
7 60% 60%
Power output [kW]

6 50% 50%
5 ← Cavitation region
η = 40%
40%
4
H = 10 m
3 Q = 105 lps
H=8m
2 Q = 94 lps
1 H = 5 m, Q = 73 lps
H = 3 m, Q = 56 lps
0
0 100 200 300 400 500 600 700 800 900
Impeller speed [RPM]
Figure 4.38: CFD predicted cavitation region for the 7 kW turbine.
eddy simulation or large eddy simulation to examine the dynamics of pressure field in the
multiphase flow.
4.7.3 Conclusions
Cavitation inception was examined on the 7 kW turbine at various operating conditions us-
ing three-dimensional Reynolds-Averaged Navier-Stokes computations. Homogeneous, free-
surface two-phase flow model was used. The pressure fields in the impeller were examined
in order to map the cavitation inception at different operating conditions by varying Q, H
and N . While cavitation was observed in the 7 kW turbine. The main conclusions drawn
from this study are summarized below.
The results of this study show that the flow in crossflow turbines is characterized by a
three-dimensional turbulent flow, either two-phase non-cavitating flow (water and air) or
three-phase cavitating flow (water, water-vapor and air) with free-surface effects. It was
found that cavitation occurs in the second stage of the turbine and was observed on the
suction side near the inner edge of the blades. In all the operating conditions considered in
103
this study, onset of cavitation was found to occur only in the second stage of the impeller. The
results showed that an increase in the impeller speed increased cavitation at constant flow
rate and head. Similarly, increase in head at constant flow rate and impeller speed reduced
cavitation. By reducing both the flow rate and head at a constant speed, cavitation increased
upto a specific operating condition and then decreased again. For this particular turbine
design, cavitation always occurred at shaft speeds greater than that giving the maximum
efficiency for each combination of Q and H. The implication is that the useful operating
range of crossflow turbines is up to and including the maximum efficiency point.
4.8 Chapter Summary
This chapter assessed the capability of RANS computations for predicting the crossflow tur-
bine performance, investigated the key flow features in two different small-scale crossflow
turbines, and characterized the performance behaviour. One was a 7 kW low-efficiency tur-
bine with 69% maximum efficiency, and the other was a 0.53 kW turbine with 88% maximum
efficiency. The dominant flow mechanisms, which characterize the power production in the
impeller, were studied systematically. The computational analysis performed on these two
reference turbines provide important insights into the crossflow turbine design problem. The
main findings are summarized as follows.
1. Good agreement was obtained between the measured and computed turbine power
outputs using steady and unsteady RANS computations for the simulation of two-
phase, free-surface flow in crossflow turbines. The steady RANS computation with
SST k − ω homogeneous, two-phase free-surface model offers reduced computational
cost compared to an unsteady RANS computation, detached-eddy simulation (DES)
and large-eddy simulation (LES). Satisfactory accuracy can be obtained from steady
RANS computations with the use of high quality meshes around the highly complex
blade shapes of the impeller. Thus, steady RANS computations are adequate for
104
conducting parametric evaluations of various turbine designs and different operating
conditions.
2. The results provided an explanation for the main reasons of inefficiency and efficiency
for both turbines. The results are consistent with the view that the flow or physical
mechanisms limiting the turbine performance have been identified. The most beneficial
effects on the impeller performance come mainly from the conversion of the head into
kinetic energy and a suitable inlet flow angle. For the impeller, the outer blade angle
and the number of blades were identified important in order to minimize major flow
separation in the impeller.
3. The difference between the inlet flow angle and the outer blade angle in the 7 kW
turbine causes significant flow separation on the blades at the first stage. This has
been determined to have a significant penalty on the performance of both stages. This
suggests that an optimal matching of the inlet flow conditions in the two stages is
crucial to efficiently extract power from the flow. This is the key design problem for
improving the efficiency of crossflow turbines. The higher performance of the 0.53
kW turbine is attributed to the conversion of head into kinetic energy, approximately
optimal inlet flow angle, and an optimal number of blades despite significant azimuthal
non-uniformity of the inlet flow. No separation was observed in the 0.53 kW turbine
at maximum efficiency.
4. The performance was characterized in terms of azimuthal distribution of the angular

momentum or torque production as the blades rotate. The performance of each stage
was computed in terms of percentage contribution to the total output power. It was
found that the flow in both the first and second stages can result in negative contri-
butions to the power, particularly when the relative velocity is negative or the flow
opposes the impeller motion.
105
5. Cavitation inception was observed around the inner edges at the second stage of the 7
kW turbine, but not in the 0.53 kW turbine. It was found that cavitation occurs only
in the second stage of the turbine on the suction side near the inner edge of the blades.
cavitation was found at and above the maximum efficiency points. Cavitation can be
avoided simply by operating the turbine up to the maximum efficiency point, and not
beyond it.
106
Chapter 5
The Key Components for Design:

Nozzle and Impeller
One of the most critical challenges in turbine design is the calculation of design parameters
and quantifying their influence and uncertainties in the performance during the preliminary
design phase (Opgenoord et al., 2016). Since conventional turbine design is mostly empir-
ical, the calculated design parameters for the design become very uncertain. This chapter
is aimed at providing a preliminary design guideline for crossflow turbine design, so that
design calculations would not lead to significant performance difference from the assumed
performance of the turbine. Based on the insights into the key flow mechanisms of the high
efficiency turbine (88%) gained in the previous chapter, this chapter expands the analysis
to include the influence of nozzle design on the impeller performance and matching their
designs. A key element of the design is the matching between the nozzle and the impeller.
The same basic principle used in the Pelton turbine is applicable to crossflow turbine. In the
Pelton turbine, the optimum jet velocity converts all the head H into kinetic energy and then
the impeller converts this into torque and then power. Similarly in crossflow turbines, the
nozzle should convert all the head H into kinetic energy, and the impeller must be designed
to convert as much as possible of the angular momentum into power. These design aspects,
107
which are the basic principles of designing efficient crossflow turbines, will be examined in
this chapter.
The specific goal is to develop a low-order model for the nozzle and the impeller designs to
guide a more advanced computational analysis and design for achieving higher performance.
This work is important because of the apparent lack of a design methodology for crossflow
turbines. This work combines an analytical model for the nozzle design with the existing
emperical model for impeller design. This is the main element of the design framework
developed in this work. In a typical design process, the overall design is assumed to be
matched, that is the designed nozzle and the impeller have a desired flow or performance.
In reality, the complexity of the design problem creates a situation, where the nozzle flow
differs significantly from the one desired for the best impeller performance. This has been
revealed from Chapter 4 in the case of both 7 kW and 0.53 kW turbines. This nozzle to
impeller flow variation from the assumed or desired flow is known as mismatching of the
nozzle and the impeller designs, and as a result, the overall performance of the turbine can
be significantly reduced.
The basic principle of converting the head H into kinetic energy can be easily stated
and tested. The next stage, i.e. designing the impeller to achieve best conversion into
power is more complex and can only be handled with CFD simulations. In the preliminary
design phase, the design challenge is at determining as close as possible the optimal values
of the design parameters. This is very important because if the initial calculations of the
design parameters are far from the optimal values, the detailed computational analysis using
higher-fidelity models would be too expensive. As a consequence, it’s likely that the optimal
design would be very difficult to be determined. Restricting the design to a few key design
parameters and understanding the associated flow field characteristics would be an effective
approach to tackle this type of multi-dimensional problem. Therefore, the insights gained
from the high-fidelity computations need to be integrated into the empirical models. In
the following, the information on optimal design parameters will be synthesized to develop
108
a simple design framework, which can be used to perform more advanced computational
analysis to further improve the design.
5.1 Nozzle and Impeller Design Parameters
The characteristic design parameters used in turbine designs are depicted in Figure 5.1
for a typical crossflow turbine configuration without a guide vane which would reduce the
efficiency. It is noted that only the circular profile blades are used in the impeller. Since the
flow field in the nozzle is much simpler than in the impeller, the analysis can be simpler.
However, matching the nozzle design with the impeller is not straightforward as it influences
the inlet flow conditions, which govern the impeller performance. To reiterate, the flow field
in the impeller is much more complex and can only be handled with CFD simulations.
For the simplest possible analysis, the inlet tangential to the impeller as shown in Figure
5.1, upto the nozzle throat h0 , is assumed to be a straight rectangular section and its design
is straightforward because continuity can be used to calculate the nozzle throat dimensions.
After the throat, the rear wall converges radially inward. Since the impeller has partial flow
admission, about 25% of the impeller, it is not fully submerged in the flow and only about
half the remainder receives the flow from the nozzle. The function of the nozzle is to convert
all the head into kinetic energy and direct the flow at a suitable angle to the impeller to
match with the blade angle.
Figure 5.1 shows the schematic illustration of the nozzle at the entry to a turbine. The
nozzle design parameters are the nozzle throat h0 , nozzle width W , rear-wall shape R(θ)
from the impeller centre, and nozzle entry arc angle θs . W is equal to the impeller width
and is constant. These parameters govern the impeller inlet flow conditions as well as the
flow field and performance of the downstream impeller. These parameters are summarized
as follows.
109
Y
h0 U0 β1b
flow
A
R(θ) Rb
C
blade centre blade
R2 θ
R1
θs β2b B
Z R1
R2
O = Impeller centre
X
Figure 5.1: Schematic illustration of the turbine geometry with tangential entry nozzle.
Every component in the diagram has equal width W normal to the page.
1. Nozzle throat area(W h0 ): This parameter is partly responsible for the conversion
of head into kinetic energy. This area must be calculated by using the continuity with
the maximum flow rate Q. W is likely to be set partly by the power requirements of
the turbine (H and Q), in which case h0 is then fixed. Otherwise a design algorithm
would be needed to vary W , determine h0 , estimate the efficiency, and repeat the
analysis until a desired efficiency is obtained. For the present, it is assumed that Q
is the maximum possible flow; consideration of part-flow operation is considered in
Section 5.4. It is also noted that designing for maximum efficiency at maximum Q
should prevent the turbine operating with cavitation.
2. Nozzle aspect ratio (W/h0 ): For the same entry arc area (R1 θs W ), this non-
dimensional parameter controls the nozzle thickness and the width W and entry arc
angle θs of the impeller.
110
flow
β1
A
Rb
C
blade centre blade
β2 B
R1
flow
R2
O = Impeller centre
Figure 5.2: Schematic illustration of the flow angles.
3. Nozzle rear wall shape (R(θ)): This parameter controls the radial velocity ur ,
tangential velocity uθ , and the inlet flow angle β1 .
4. Nozzle entry arc angle (θs ): This parameter controls the radial velocity ur and the
mass and angular momentum fluxes at the impeller inlet.
5. Nozzle throat to impeller inlet arc ratio (h0 /R1 θs ): This non-dimensional pa-
rameter influences the inlet velocity profile, the flow angle, and the mass and angular
momentum fluxes at the impeller inlet, as well as the ratio of tangential to radial
velocity (uθ /ur ) at the impeller inlet.
Despite the fact that the flow field within the impeller is highly complex, the impeller
geometry is relatively simple to design. The design parameters for the impeller are:
1. Angle of attack (α): This is the angle of the absolute velocity at the impeller inlet.
This angle is determined by the radial velocity ur and the tangential velocity uθ at the
impeller inlet.
111
2. Outer blade angle (β1b ): The angle of the blade inlet β1b for the first stage is equal to
the outlet blade angle of the second stage. The outer blade angle β1b is important as its
alignment/misalignment with the inlet flow angle β1 determines the flow separation on
the blades and the performance of the impeller. 2.1 shows that this angle is generally
selected as 30 - 39◦ . As discussed in Section 5.1, for the maximum efficiency, the
flow must exit radially from the impeller outlet to avoid efflux of angular momentum.
Therefore, β1b is important to adjust the flow angles at the inlet of the first stage and
at the exit of the second stage.
3. Inner blade angle (β2b ): The inner edge blade angle β2b is generally specified as 90◦
to have no angular momentum leaving the first stage. This parameter can be freely
adjusted to improve the second stage performance. β2b affects the exit flow angle β2 of
the first stage and the inlet flow angle β2i at the entry of the second stage. Based on a
limited experimental study, Totapally and Aziz (1994) found that β2b = 55◦ would give
a higher maximum efficiency than β2b = 90◦ . However, the consequence of lower values
of β2b than 90◦ on the flow field and impeller performance is not well-understood.
4. Inner to outer diameter ratio (D2 /D1 ): Changing the diameter ratio D2 /D1 has
two geometrical consequences. D2 /D1 defines the chord length and the radius of the
blades Rb . The effect of D2 /D1 on impeller performance has been experimentally
studied by many researchers and have reported that D2 /D1 = 0.64 - 0.68 will give
the best efficiency (see Table 2.1). However, the influence of D2 /D1 on the flow field
has not been studied so far. Increasing or decreasing the D2 /D1 ratio changes the
chord length and curvature of the blades which are assumed to always be circular arcs.
So D2 /D1 may result in poor flow guidance, separation on the blades and ultimately
decreased efficiency.
112
5. Blade radius (Rb ): The blade radius Rb has important fluid dynamic consequences,
such as flow guidance in the blade passages, boundary layer separation on the blades,
and flow deflection. Rb is geometrically related to D2 /D1 , β1b and β2b by
2
D2
D1 1− D1
Rb = D2
(5.1)
4 cosβ1b − D1
cosβ2b
In conventional design, β2b is specified as 90◦ . With β2b = 90◦ , Equation (5.1) reduces
to
D12 − D22
Rb = (5.2)
4D1 cosβ1b
6. Number of blades (Nb ): Increasing the number of blades gives better flow guidance
or reduces the tendency to boundary layer separation, but viscous losses will increase
in the boundary layers due to increased surface area and create create back pressure,
i.e blockage in the passage area. Therefore, the selection of optimal Nb is important.
The selection of Nb more likely depends on the working head, flow rate, the outer
diameter of the impeller, and the operating speed. However, no systematic studies on
the selection of Nb are available in the literature.
5.2 Analytical Model for Nozzle Design
Nozzle design is an important aspect because the inlet flow condition, particularly the an-
gular momentum flux and the flow angle at the impeller inlet, directly affect the impeller
performance as it does in most radial turbines (Yang, 1991; Baskharone, 2006). In previous
experimental studies, impeller design was the main focus and very little attention was given
to the influence of nozzle design or inlet flow conditions, while de-emphasizing the penalty
that a poor nozzle design can cause on the impeller performance. To quantify the effects of
inlet flow conditions or the nozzle design on the impeller performance and develop methods
113
to design a good nozzle, a simple two-dimensional (2-D) analytical model for nozzle design
is developed in this study.
In Chapter 4, the nozzle performance of the 7 kW turbine was found to be poor, par-
ticularly on converting the head H into kinetic energy at the impeller inlet. In the case of
0.53 kW turbine, all the head H was approximately converted into kinetic energy. However,
the velocity and the flow angles were found to vary azimuthally in both turbines. Thus it
is important to define a nozzle geometry, which turns approximately the entire head H into
kinetic energy and creates a suitable flow angle β1 before it enters the impeller. Thus the
nozzle performance and matching with the impeller were identified as the main limitations
in the baseline turbine designs investigated in Chapter 4, and it is essential to address these
issues for improving the maximum efficiency.
The assumption of 2-D inviscid irrotational flow is appropriate for analyzing the flow in
the nozzle because power extraction does not take place in the nozzle and there is negligible
viscous loss. The key is that the nozzle is converging so there will be no flow separation at the
walls and the boundary layers are likely to decrease with distance along the nozzle. Since
the flow is predominantly constrained in the tangential and radial directions, Bernoulli’s
equation can be used to calculate nozzle dimensions with reasonable accuracy to achieve its
objectives. As the nozzle flow is predominantly two-dimensional, the analysis is carried out
on the r − θ plane using the radial and tangential velocity components in the cylindrical
coordinate system as shown in Figure 5.3. This introductory analysis assumes the throat h0
is aligned radially so the inlet flow is tangential. In addition, uniform radial and tangential
velocities are assumed at the impeller inlet so that power extraction from the first stage can
be maximized.
To compute the nozzle rear-wall shape R(θ), and obtain a uniform flow distribution and
maximize the angular momentum at the impeller inlet, continuity in the radial and tangential
directions is applied. Referring to Figure 5.3, a uniform radial velocity ur is desired at the
impeller inlet for the best performance of the turbine. It is also assumed that uθ is uniform
114
Y
flow h0 U0
uθ
ur h(θ)
dθ
θ
R1
θs
Z
X
Figure 5.3: Schematic illustration of the geometry of the tangential entry nozzle used in the
analysis.
along the entry to the impeller and is equal to the velocity at the nozzle throat, i.e. uθ ≈ U0 .
Consider a small angle dθ of the nozzle shown in Figure 5.3, associated with this turning
are the changes in total velocity du, tangential velocity duθ , and radial velocity dur in the
tangential and radial directions respectively. Continuity applied to the radial and tangential
velocities using the requirement that the radial velocity ur must be independent of θ gives:
rur dθ = −h(θ + dθ)uθ (θ + dθ) + h(θ)uθ (θ) (5.3)
so that
rur = −d(h(θ)uθ (θ))/dθ (5.4)
115
Integration of Equation (5.4) gives:
ur R1 θ = −h(θ)uθ (θ) + h0 U0 (5.5)
Imposing the boundary conditions: h(θ) = h0 and uθ (θ) = U0 at θ = 0, and using h0 U0 =

ur R1 θs as all the flow entering the nozzle must enter the impeller, Equation (5.5) reduces to
h(θ)uθ (θ) = h0 U0 (1 − θ/θs ) (5.6)
The nozzle rear wall h(θ) from the inlet of the impeller can be computed from Equation
(5.6) as:
h(θ) = h0 (1 − θ/θs ) (5.7)
The radial distance of the nozzle rear-wall from the center of the impeller R(θ) can be thus
expressed as:
R(θ) = R1 + h0 (1 − θ/θs ) (5.8)
where R1 is the impeller outer radius and h(θ) is the radial distance of the nozzle wall from
the inlet of the impeller. This is the special equation for the tangential entry of the nozzle.
A more generalized equation for computing R(θ) of an arbitrary nozzle orientation, as shown
in Figure 5.4, is obtained as:
p θ
h(θ0 + γ + θ) = ( (R1 sin θ0 + h0 )2 + (R1 cos θ0 )2 − R1 )(1 − ) (5.9)
θs − γ
where θ0 = orientation of the left nozzle lip. The derivation is provided in Appendix B. This
equation is useful for designing any nozzle, and will be used in Chapter 6 for analyzing the
nozzles of both the 0.53 kW and 7 kW turbines.
116
Y
C E
flow h0 U0
uθ
B F h(θ)
ur
dθ
γ θ D
R1
θs
θ0
0 Z
X
Figure 5.4: Schematic illustration of the generalized nozzle geometry used in the analysis.
Assuming that H is entirely converted into kinetic energy at the impeller inlet at atmo-
spheric pressure requires
1
ρgHQ = ρQ (u2θ + u2r ) (5.10)
2
For the simple tangential nozzle, using continuity at the nozzle throat h0 and impeller entry
θs :
U0 h0
ur = (5.11)
R1 θs
Equation (5.10) reduces to

U02 h2
gH = (1 + 20 2 ) (5.12)
2 R1 θs
Equation (5.12) indicates that a perfect nozzle converts all H into kinetic energy at the
impeller inlet.
117
The angle of the streamlines entering the impeller is given by
ur
β1 = tan−1 ( ) (5.13)
uθ − R1 ω
Good turbine design requires that β1 is equal to the blade entry angle β1b in order to avoid
flow separation from the blades. Therefore, the conditions that give the best nozzle perfor-
mance also give essential information for the impeller design.
The angular momentum flux at the inlet of the impeller is given by
ρW R12 uθ ur θs = ρW R1 U0 h0 U0 (5.14)
and the impeller shaft power Ẇ is given by
Ẇ = ρW U0 h0 R1 ωU0 (5.15)
This equation can be expressed as
Ẇ = ρQU0 R1 ω (5.16)
where ω is the angular velocity of the impeller. Assuming all the entering angular momentum
at the impeller inlet is converted into useful torque (i.e. no angular momentum at the impeller
outlet or Ẇ = ρQgH), we get
ρQU0 R1 ω = ρQgH (5.17)
Combining Equations (5.12) and (5.17) gives
R1 ω 1 h2
= (1 + 20 2 ) (5.18)
U0 2 R1 θs
118
Equation (5.18) shows that maximum efficiency occurs at around a tip-speed ratio of 0.5,
provided that H is entirely converted into kinetic energy. It is noted that Equation (5.18)
is the same equation used for optimizing the Pelton turbine when viscous losses in the
jet and elsewhere are ignored. Invoking this condition for the maximum efficiency, ω at
the maximum efficiency point can be approximately computed. Furthermore, this equation
shows that there is only one Q for which the head H can be fully converted into kinetic
energy to obtain the maximum efficiency. Similarly, for a particular H at the maximum
efficiency point, there is a specific ω that gives the maximum efficiency. From control point
of view, this is a crucial result because it means that flow control must be used for part-load
operation and it is not possible to rely on electronic control of the generator to maintain high
efficiency. It is worth noting that maximum efficiency point is obtained at slightly higher
tip-speed ratios than given by the Equation (5.18) because the increase is partly due to
h20 /(R1 θs )2 and the impeller cannot practically extract all the angular momentum available
at the impeller inlet due to losses. This fact is also observed in the performance maps of
the two turbines considered in the study. It is also noted that in practice, it is desirable
to keep an impeller speed constant to match the gearbox and generator speed for optimal
performance.
5.3 Impeller Design
The important design parameters for the impeller are outlined in Section 5.1. The impeller
design is simplified by a) needing to match blade inlet angle and flow angle and b) using
circular arc blades, so only inner to outer diameter ratio D2 /D1 , number of blades Nb , and
blade outlet angle can be adjusted. The key element of the design is that the blade angles
have to match the flow angles at both stages, which can not be achieved without the help of
CFD simulations. In addition, the inlet flow angle β1 is defined by the nozzle design. For the
preliminary calculations, the outer blade angle β1b can be selected between 30 - 39◦ and the
119
inner blade angle is usually fixed at 90◦ as suggested in the literature. Desai (1994) showed
that the maximum efficiency is obtained at β1b = 39◦ and β2b = 90◦ , which can be taken as
a reference in the preliminary design of the impeller.
As the flow field is highly complex and the inlet flow conditions may vary azimuthally at
the inlets of both stages, analytical approaches (i.e. kinematic relations using velocity trian-
gles) for the impeller design do not capture the flow dynamics, and thus are not attempted
here. Moreover, as shown by the example in Appendix A, their predictive capability is very
limited due to lack of information of the local flow field. Any attempt on design improvement
will therefore require high-fidelity computational simulations.
The impeller design begins with the specification of the operating speed, which can often
be selected based on the generator speed, the working head H, and the flow rate Q. As
discussed in the previous section, maximum efficiency occurs at around the tip-speed ratio
of 0.5. Based on the impeller speed N and the tip-speed ratio of 0.5, the outer diameter
D1 can be computed. Then using the empirical value of D2 /D1 = 0.64 - 0.68 suggested
by previous experimental studies, the inner diameter D2 and the blade radius Rb can be
computed using the Equation (5.2). After the size of the impeller and blade radius are
computed, the number of blades are selected in such a way that separation on the impeller
blades is minimized. It was revealed in Chapter 4 that flow separation is significant in the
7 kW turbine, but negligible in the 0.53 kW turbine. These results suggest that at least
30 blades are needed to minimize separation in the 7 kW turbine as the outer diameters of
both turbines are comparable. The choice of the number of blades is crucial because power
is extracted within a narrow region in the second stage, and there are sharp rise and fall
in power extraction, accompanied with negative power. An optimum number of blades will
increase the opportunity to extract the power more efficiently from this narrow flow region
of the second stage, but at the same time, extraction from the first stage will also increase.
120
5.4 Inlet-flow Control Mechanism
While hydro turbines will likely run at design conditions for the majority of their operational
lifetime, part-flow operation may occur when the power demand reduces or the water supply
is reduced by lack of rain which in turn may affect the performance. This is the only
off-design operation condition, called part-flow condition because head is always constant.
Reductions in the flow-rate at the same impeller speed N lower the velocity and increase
the flow angles to larger than optimal, which negatively impact the impeller performance.
Being able to achieve high part-load efficiency for a manufacturer is also beneficial as this
reduces the number of turbine models to cover the possible range of customers’ Q and H.
Turbines are installed with variable inlet-flow control systems to operate efficiently at
part-flow conditions. In conventional designs, a guide vane is installed in the nozzle passage
to control the flow during part-flow conditions as shown in Figure 2.2. The objective is to
keep the high flow velocity and maintain an optimum inlet flow angle by changing the flow
direction. This is important because the performance of the impeller at part-flow conditions
with the same operating head is significantly reduced if the flow is not controlled because
the inlet flow does not match the impeller design. This will be examined in Section 6.2.1.
Inlet flow is controlled (velocity and flow) to ensure efficient part-load operation. For this,
guide vanes are usually put in the nozzle passage. However, putting guide vanes in the nozzle
passage induces extra losses due to blockage effects and increased viscous drag. Moreover,
it cannot precisely control the flow as desired, which incurs extra losses. The Table 2.1
in Chapter 2 shows that turbines with guide vanes have low efficiency and that the high
efficiency 0.53 kW turbine does not have a guide vane. A promising mechanism to reduce
these losses and precisely control the inlet flow during part-flow conditions is to use the slider
type control mechanism shown in Figure 5.5, which can be rotated azimuthally to adjust the
nozzle throat and impeller inlet area (Sinagra et al., 2014). If it can maintain high efficiency
at part-load, then the slider is probably cheaper to build and can be accurately activated.
This design was first proposed by Miroslav Cink (Sinagra et al., 2014). A promising design
121
feature of this device is that it can regulate the inlet flow rate by changing the throat h0
and entry arc angle θs . A detailed analysis on the influence of this control mechanism on
the impeller performance is not available in the literature, but Sinagra et al. (2014) and
Sammartano et al. (2016) carried out a basic CFD analysis, which has shown that maximum
efficiency very close to the design flow-rate can be maintained when Q is reduced to 20% of
its maximum. The simple operational design clearly shows that this control mechanism may
be a better design than commonly used guide vanes for crossflow turbines.
The analytical model for nozzle design shows that there exists a unique flow Q which gives
a maximum efficiency η at a given head H for a particular nozzle design without any control
mechanism for the inlet flow. So for lower flow rates at the same head, the slider control
mechanism can change the nozzle throat and the nozzle entry arc. In other words, by using
the slider, the shape of the nozzle changes at lower flow rates. In this situation, the analytical
model for the nozzle design is very useful to study the effects of slider control mechanism.
In this thesis, guide vanes are not considered, instead the slider type control mechanism is
investigated in Chapter 6 after the quality of the basic nozzle design is established.
Figure 5.5: Slider control mechanism for controlling the inlet flow
122
5.5 Chapter Summary
This chapter has presented an integrated nozzle and impeller design framework for design-
ing the crossflow turbine. A simple 2-dimensional analytical model for nozzle design was
presented. The main focus on designing the nozzle was put on converting approximately
all the head to kinetic energy and directing the flow at a suitable angle to match with the
impeller outer blade angle. To keep the efficiency high, inlet flow must be controlled at
part-flow conditions. For this, flow regulator, such as the one first introduced by Miroslav
Cink, can be used. The analytical model also allows design and control of this flow regulator
to maintain high efficiency at part-flow operations. The design principle presented in this
chapter will be applied to investigate the design improvement in Chapter 6.
123
Chapter 6
Design Improvement
This chapter applies the insights gained in Chapter 4 within the design framework presented
in Chapter 5 to improve the maximum efficiencies of the 0.53 kW (88%) and 7 kW (69%)
turbines. Toward this goal, two design concepts are investigated. In the first part, the
focus is on designing the nozzle to meet its basic functions, which are 1) all the head is
approximately converted into kinetic energy and the angular momentum flux is maximized
at the impeller inlet, and 2) the flow is directed to the impeller at a suitable angle. Toward
this goal, a sensitivity analysis is carried out to quantify the influence of nozzle design on the
performance of the impeller. In the second part, the influence of different impeller design
parameters on the impeller performance is assessed to develop methods for optimal matching
of the nozzle and the impeller designs, particularly 1) matching the inlet blade angle with the
inlet flow angle and 2) matching the impeller speed to extract maximum angular momentum.
This is crucial because mismatching of the nozzle and the impeller designs can outweigh any
benefits achievable from individual component’s optimal design. For this, computations are
carried out on a number of modified designs of the 0.53 kW and the 7 kW turbines. All the
computational efforts on design improvement will evolve from these two baseline designs. The
rationale for using different parameters, e.g. conversion of head into kinetic energy, inlet flow
angle, and flow matching between the first and the second stages, to evaluate the efficiency
124
of the candidate turbine designs is explained. Finally, optimal design configurations are
suggested to improve the efficiency of the turbines, which will address unanticipated factors
not taken into account during the turbine design process.
In the sensitivity analysis, the first investigations are for maximum flow rate Q and then
part-load conditions will be investigated for maintaining high efficiency.
6.1 Key Design Concepts for Improvement
In Chapter 4, the key design problem was described and the potential areas for improvement
were outlined for both turbines. Chapter 5 described an integrated framework for designing
the turbine, which rules out a simple low-order design model for conducting extensive high-
fidelity simulations to further improve the design. It was concluded that an understanding
of the influence of different design parameters on the performance of both the nozzle and the
impeller, and examining the methods for optimal matching of the nozzle and the impeller are
important. The enabling idea is the development of a design framework for the nozzle and the
impeller and their matching criteria to account for the total conversion of head into kinetic
energy, create a suitable flow angle at the impeller inlet, and improve the performances of the
impeller stages. The analysis is applied to the 0.53 kW turbine because the efficiency of this
turbine is already high (88%). The high efficiency means that the design already embodies
at least some of the principles outlined in Chapter 5, implying the validity of the method
followed, and so it is important and useful to see if further improvements in efficiency are
possible. More importantly, for the 7 kW turbine the redesign must produce a substantial
improvement with the principles used in the analysis.
The results presented in Chapter 4 revealed two aspects of importance for improving the
efficiency. The first is nozzle performance as it influences the flow conditions at the impeller
inlet, particularly the conversion of head into kinetic energy and directing the flow to the
impeller at a suitable flow angle. The second is the problem of matching the inlet flows at
125
the first and second stages of the impeller, so that power is efficiently extracted from both
stages. Mismatching between the nozzle and the impeller designs appears to be the main
cause for poor performance because flow separation is the key indicator of mismatching as
found in Chapter 4. These are the fundamental concepts behind the design improvement ap-
proach. Toward this goal, a sensitivity analysis provides quantitative results for determining
the key design parameters and their optimal combinations for higher performance. Varia-
tions of the geometric parameters serve to perturb the flow, thus permitting identification
and quantification of major performance limiting mechanisms and their relative significance
in contributing to the performance loss associated with the design. Quantitative relation-
ships between the flow parameters, i.e. velocity and flow angles, are the basic information
necessary to assess the impeller performance. The analytical model for the nozzle design is
used to investigate the nozzle performance, its influence on the impeller performance, and
enhance the flow matching between the nozzle and the impeller. More importantly, CFD
results allow to determine whether the analytic model is sufficiently accurate. The above
approach of design analysis is sufficiently general to investigate the entire design space.
6.2 Sensitivity Analysis: 0.53 kW Turbine - What
Matters Most?
This work seeks to identify the key performance limiting flow mechanisms in the nozzle and
the impeller. Matching a good nozzle design with the impeller is emphasized because the
penalty of poor nozzle design on the impeller performance can outweigh any benefits achieved
from improved impeller design. In order to determine how different geometric parameters
change the performance of the turbine and identify the key design parameters that matter
most on the impeller performance, sensitivity analysis is carried out on the 0.53 kW turbine
within the design framework described in Chapter 5. The key motivation for investigating
the influence of different design parameters is to see whether the high efficiency 0.53 kW
126
turbine (88%) embodies the design principles developed in Chapter 5 and determine the
potential for improved performance by improving the inlet flow conditions and the impeller
design. Moreover, the flow mechanisms limiting the performance are not currently studied
and well-understood. This analysis seeks to investigate the relationship between the geo-
metric parameters, flow features and performance through sensitivity analysis, focusing on
characterizing the relative importance of different design parameters. Sensitivity analysis
enables one to iterate through different turbine designs and configurations, analyze how the
turbine efficiency responds to various design changes, and synthesize those informations to
improve the design. Since the limitation on computational resources does not allow too
many simulations for a complete search of optimal parameters, the parameter changes are
done within the vicinity of the design space of the 0.53 kW turbine considering only the key
design parameters because the efficiency of this turbine is already high.
To analyze turbine efficiency, flow velocities and flow angles at the impeller inlet and
the power or the torque produced at the first and second stages are the basic parameters
of interest. These parameters are the local as well as the global parameters governing the
internal flow and performance. The flow angles, velocities and torque capture the local effects
of changes in flow conditions and geometry, and are used as indicators of local and global
changes in turbine efficiency. The physical relationships between the geometric parameters,
the flow parameters, and ultimately the turbine’s performance need to be established. The
operating conditions considered in the sensitivity analysis are the same as that of the original
0.53 kW turbine, which are presented in Chapter 4.
The conventions used in the analysis are schematically illustrated in Figure 6.2.
6.2.1 Influence of Nozzle Design on the Efficiency of Impeller
To quantify the influence of nozzle design on the impeller performance and identify the
key performance limiting flow mechanisms, computations were performed on two different
127
Y
left nozzle lip flow
nozzle throat
first stage
Ψ
θs
Z
right nozzle lip
second stage
Figure 6.1: Schematic illustration of the conventions used in the analysis. The azimuthal
position ψ, nozzle throat, left nozzle lip, right nozzle lip, first stage, and second stage will
be used. ψ will be used to refer to a specific position or a region in the nozzle entry arc and
the first and second stages. It is noted that for the 0.53 kW turbine, the left nozzle lip is at
the azimuthal position 120◦ and the right nozzle lip at 210◦ .
nozzle designs for the same impeller of the 0.53 kW turbine. To design the nozzle, the new
analytical model for nozzle design presented in Chapter 5 was used.
In the case of 7 kW turbine, it was shown in Chapter 4 that the inlet flow angle β1 was
significantly greater than the outer blade angle β1b of the impeller, and varied azimuthally
over the impeller inlet. More importantly, H was only partially converted into kinetic energy.
These were identified as the main reasons for the poor impeller performance of the 7 kW
turbine. In the case of 0.53 kW turbine, it was found that the nozzle design is approximately
optimal, particularly in terms of converting H into kinetic energy or maximizing the angular
momentum flux at the impeller inlet and maintaining β1 close to β1b over a larger portion
of the impeller inlet. However, a considerable azimuthal variation of radial velocity ur ,
128
tangential velocity uθ and inlet angle β1 was found, particularly in the impeller inlet section
near the left nozzle lip. In this region, β1 was found to be significantly greater than β1b .
Thus, there are still two issues that were not resolved in Chapter 4. The first is whether the
azimuthal non-uniformity of the inlet flow limits the impeller performance, provided that
H is approximately converted into kinetic energy. This issue is examined here in detail;
however, the influence of this non-uniform inlet flow can not be quantified precisely unless
a uniform inlet flow design is developed with a new nozzle design, which is attempted here.
The second is the extent to which the difference between β1 and β1b limits the impeller
performance, provided that H is approximately converted into kinetic energy. This issue is
examined in Section 6.2.3 by performing a sensitivity analysis on varying β1b . Therefore, the
previous results give a strong hint that the nozzle design can still be improved, particularly
in terms of matching the inlet flow with the blade angle near the nozzle throat.
For the tangential nozzle, which is the simplest, the design parameters are the throat h0 ,
the width W , the entry arc angle θs , and the rear wall shape R(θ). More complex nozzles
have angle of attack α etc. Since all the head H is approximately converted into kinetic
√
energy (ut = 2gH) in the existing nozzle, this implies that the nozzle throat area was
properly sized. This can be approximately checked using the Equation (5.12), which gives
a simple criterion for the total conversion of head into kinetic energy at the impeller inlet.
For this, tangential velocity U0 was computed as ut cos22, h0 = 0.8944, R1 = 0.152, and
θs = π/2. It is noted that α = 22◦ was used as used by the author of the turbine (Desai,
1994). The calculation shows that the head is approximately converted into kinetic energy.
So the same values of h0 , W and θs as that of the original design were used for both new
nozzle designs. The orientation of the inlet of the original nozzle is at 30◦ to the vertical as
illustrated in Figure 6.2.
By following the above procedure for nozzle design, two series of computational simula-
tions were performed without altering h0 , W , θs and the impeller of the original turbine. In
the first series of simulations, a new nozzle, hereafter called new nozzle, was designed for the
129
Y
left nozzle lip 220 h0 = 89.44 mm
0
30 R(θ)
0
90
X
right nozzle lip
Figure 6.2: Schematic of the new nozzle with the original nozzle orientation.
0
h0= 89.44 mm
90
0
4
left nozzle lip
R(θ) right nozzle lip

0
90
X
Figure 6.3: Schematic of the tangential entry nozzle.
130
90
Original nozzle
89.5 Tangential nozzle
New nozzle
89
88.5
Efficiency [%]
88
87.5
87
86.5
86
85.5
85
185 190 195 200 205 210 215
Speed [RPM]
Figure 6.4: Efficiency vs speed for different nozzle designs for the 0.53 kW turbine [H =
1.337 m and Q = 46 lps].
same orientation as that of the 0.53 kW turbine, whereas the second nozzle was designed
for the tangential entry to the impeller, hereafter called tangential nozzle. As illustrated in
Figure 6.2, the inlet of the first nozzle is oriented vertically as that of the original nozzle,
and the left nozzle lip was inclined at 30◦ to the vertical axis. In the case of tangential entry
nozzle, the inlet is approximately horizontal1 and the left nozzle lip is oriented at 90◦ to the
horizontal. In both nozzles, h0 , W and θs were kept the same as that of the original nozzle
because the nozzle design is already approximately optimal. Only the rear wall shape R(θ)
was calculated in both cases using the new nozzle design equations presented in Chapter 5.
R(θ) of the first new nozzle was calculated using Equation (5.9), whereas R(θ) of the tan-
gential entry nozzle was calculated using the Equation (5.8). Thus the differences between
these two new nozzles are R(θ) and the orientation. Simulations were performed only at
three different speeds: 186.4, 199.1 and 211.8 RPM as these speeds are within the vicinity
of the maximum efficiency of the reference turbine.
1
The lower wall of the inlet was inclined at 4◦ in order to create the computational domain, which is
otherwise impossible.
131
6
4 Absolute velocity − original nozzle

Radial Velocity − original nozzle
Velocity [m/s]
Theoretical jet velocity
2 Absolute velocity − new nozzle
Radial velocity − new nozzle
−2
−4
120 130 140 150 160 170 180 190 200 210
Figure 6.5: Comparison of velocities at the impeller inlet for the new nozzle and existing
nozzle for the 0.53 kW turbine [H = 1.337 m and Q = 46 lps].
The influence of the nozzle designs on the turbine efficiency is summarized in Figure 6.4.
With the first new nozzle, marginal improvement in maximum efficiency was obtained (from
88% to 89.45%). Similarly, the tangential nozzle gave a very similar maximum efficiency
(89.21%). It is also noted that the maximum efficiency for all nozzles has occurred at the
same impeller speed. This optimum speed is slightly greater than the one predicted in
the analytical equation for nozzle design. These results indicate that R(θ) is important for
efficiency because it modifies ur , uθ and β1 . However, the nozzle orientation did not alter the
turbine efficiency. In these nozzles, β1 and the percentage power productions in the impeller
stages are discussed below.
The plots of computed total velocity u0 , ur and β1 at the impeller inlet for the first
new nozzle and the original nozzle are shown in Figures 6.5 and 6.6. It can be seen from
Figure 6.5 that computed total velocity u0 at the impeller inlet for the new nozzle is slightly
greater than for the original nozzle. In addition, β1 in the new nozzle is slightly lower and
more uniform than in the original nozzle. This demonstrates that the new nozzle is better
132
90
Original nozzle
80 New nozzle
Outer blade angle
70

60
50
40
30
20
10
0
120 130 140 150 160 170 180 190 200 210
Figure 6.6: Comparison of inlet flow angle β1 for the new nozzle and the existing nozzle at
maximum efficiency of the 0.53 kW turbine [H = 1.337 m, Q = 46 lps and N = 199.1 RPM].
Figure 6.7: Contour plot of the magnitude of mean water velocity for the original nozzle at
maximum efficiency of the 0.53 kW turbine [H = 1.337 m, Q = 46 lps and N = 199.1 RPM].
at converting H into kinetic energy and creating a more uniform inlet flow. These are the
reasons for the increased impeller performance as the impeller is the same in both cases.
A very interesting feature can be observed in the case of tangential nozzle. H is ap-
proximately converted into kinetic energy because the computed total velocity u0 is nearly
√
equal to the total velocity ut = 2gH. More importantly, β1 is almost equal to β1b and
133
Figure 6.8: Contour plot of the magnitude of mean water velocity for the new nozzle at the
maximum efficiency point of the 0.53 kW turbine [H = 1.337 m, Q = 46 lps and N = 199.1
RPM].
Figure 6.9: Contour plot of the magnitude of mean water velocity for the tangential nozzle
at the maximum efficiency point of the 0.53 kW turbine [H = 1.337 m, Q = 46 lps and N
= 199.1 RPM].
134
80
Computed flow angle: tangential nozzle
70 Computed flow angle: original nozzle
Outer blade angle

60
50
40
30
20
10
0
120 130 140 150 160 170 180 190 200 210
Figure 6.10: Comparison of the inlet flow angle β1 for the tangential nozzle and existing
nozzle at the maximum efficiency point of the 0.53 kW turbine. The azimuthal position ψ
for the tangential nozzle has been increased by 30◦ for the purpose of comparison [H = 1.337
m, Q = 46 lps and N = 199.1 RPM].
0.2
Original Nozzle
Tangential Nozzle
0.15 New Nozzle
power loss
0.1
Power [kW]
0.05
first second
−0.05 stage stage
−0.1
120 140 160 180 200 220 240 260 280
Figure 6.11: Comparison of power outputs in the impeller for different nozzle designs at the
maximum efficiency point [199.1 RPM]. Note that the changes are due to the change in the
inlet flow angle β1 . The azimuthal position ψ for the tangential nozzle has been increased
by 30◦ for the purpose of comparison.
135
90
Tangential nozzle
80 Original nozzle

New nozzle
70
60
50
40
30
20
10
0
1 2
Impeller stage
Figure 6.12: Comparison of percentage power production in the impeller stages for the
different nozzles at the maximum efficiency point of the 0.53 kW turbine [199.1 RPM]. Note
that the changes are due to the change in the inlet flow angle β1 .
azimuthally uniform over the entire inlet of the impeller as shown in Figure 6.10. β1 is
slightly less than β1b near the right nozzle lip: 200◦ ≤ ψ ≤ 215◦ . This implies that the
tangential entry nozzle gives nearly uniform inlet flow to the impeller, and that the simple
analytic model for the optimum nozzle is remarkably accurate. The impact of uniform inlet
flow on the performance of the downstream impeller is shown in Figures 6.11 and 6.12. It
is observed that power extraction in the first stage is almost uniform azimuthally, except
for 190◦ ≤ ψ ≤ 210◦ , where power extraction has sharp rise and fall. The percentage power
production in the first stage is significantly greater (about 81%) than the other two designs,
where the first stage contribution is about 69%. This means that a uniform inlet flow with β1
approximately equal to β1b produces more power from the first stage, so there is less energy
for the second stage to extract. It is interesting to note that even with such a huge change
in stage performance, the efficiencies in these two cases are similar. Similarly, comparing
the original nozzle and the new nozzle, the first stage performance increased only marginally
with the new nozzle. It is more likely due to the increased conversion of H into kinetic
energy. The velocity contour maps at the maximum efficiency point for the existing and new
136
nozzles reveal that they are similar without major flow separation on the blades in either
stage. The results show that tangential nozzle provides a uniform β1 to the impeller although
it does not have much influence in increasing the efficiency compared to the other new nozzle
because it was nearly as efficient at producing kinetic energy and the second stage of the
impeller has picked up the missing power. In other words, the impeller in a crossflow turbine
can perform efficiently even if the inlet flow is azimuthally varying, provided that all H is
converted into kinetic energy and β1 is close to β1b over a large extent of the impeller inlet.
It can be concluded that the original nozzle was good but not perfect at converting H into
kinetic energy and the second stage has produced power that was missed in the first stage
due to the poorer design of the original nozzle.
6.2.2 Influence of Number of Blades (Nb )
The number of blades Nb affects a number of fluid dynamic processes, primarily boundary
layer separation and power extraction. It is reasonable to assume that higher Nb would tend
to reduce susceptibility to separation. From the design point of view, separation was not
found in the second stage in both turbines as revealed in Chapter 4. Increasing Nb does not
alter the ratio of outlet to inlet area for the blade passages, however, it can help keep the flow
attached because the displacement effect of the blade boundary layers will effectively reduce
the ratio. As shown in Chapter 4, power extraction at the second stage takes place within a
very narrow region, and its azimuthal variation is significant, characterized by a sharp rise
and drop in power extraction followed by power loss into the flow. This variation can also be
closely connected to the Nb . It was pointed out in chapter 4 that increasing Nb would give
more second stage blades and this may increase the efficiency of its power extraction. The
flow processes leading to boundary layer separation are of particular interest in the selection
of an optimum Nb because a highly separated flow at the first stage leads to poor quality of
the flow at the second stage, where about 25 - 30% of total power production takes place. As
revealed in Chapter 4 in the case of 7 kW turbine, a significant flow separation was found on
137
95
20 blades
30 blades
35 blades
40 blades
Efficiency [%] 90
85
80
185 190 195 200 205 210 215
Figure 6.13: Influence of blade number on the efficiency.
the first stage blades, which was primarily attributed to significantly higher β1 than β1b and
insufficient Nb ( = 20). In contrast to this, the flow in the 0.53 kW turbine was almost fully
attached, which was attributed to a better match between β1 and β1b and larger Nb ( = 30).
In this study, the influence of Nb on the impeller performance is studied by computing the
flows at Nb = 20, 30, 35 and 40 without altering the original nozzle and any other aspects
of the impeller (0.53 kW turbine). In an experimental study carried out by Totapally and
Aziz (1994) on the same 0.53 kW turbine, Nb = 35 was found to give a maximum efficiency
of 90%. They also found that Nb = 40 decreased the maximum efficiency to about 86%.
A comparison of the efficiencies for different Nb is shown in Figure 6.13. Amongst the
cases considered, the maximum efficiency of 89.87% was obtained for Nb = 35, which is close
to the experimentally obtained efficiency of 90% by Totapally and Aziz (1994). At Nb =
40, the efficiency was found to be 88.23%, which is only marginally lower than at Nb = 35,
but greater than the experimentally obtained efficiency of 86% by Totapally and Aziz (1994)
at Nb = 40. It is thus concluded that about 35 blades would give the maximum efficiency
138
of about 90%. Similarly, when Nb was reduced to 20, the maximum efficiency dropped
significantly from 88% to 85.87%. It is noted that the impeller speed for maximum efficiency
has not changed amongst the cases, which is important. The changes in the flow field in the
impeller and the reasons for the performance changes are explained as the following.
The quantitative differences in the impeller performance were investigated by computing
the azimuthal variation of torque production by the blades. Figure 6.14 summarizes the
torque distributions at the maximum efficiency point for different Nb . As Nb is increased
from 20 to 30, more power is extracted from the first stage and less power is extracted from
the second stage. When Nb = 20, the second stage extracts more power compared to 30
or 35 blades because the first stage blades miss more power. Of particular note is that
the blades passing through ψ = 260◦ produce negative power. It is evident that the power
production in the first stage is more uniform in all the cases, whereas the power extraction
in the second stage is confined to a narrow region, and thus changes sharply compared to 30
and 35 blades. This is due to the fact that the significant azimuthal variation of the β2i as
explained in Chapter 4 remains when Nb is altered.
The influence of Nb on the stage performance is shown in Figure 6.15. It can be observed
that the power production at the first stage has increased with the increase in Nb from 20
to 30. When Nb was increased from 20 to 30, the first stage performance or the percentage
of total power production increased from 66.40 to 69.20%. Similarly, when Nb was increased
from 30 to 35, the power production at the first stage decreased. The results reveal that
an optimum number of blades increases the performance of both stages. The differences
between the flow fields in the impeller when Nb = 20, 30, 35 and 40 were examined at
the maximum efficiency points by comparing the velocity contours, velocity vectors and
streamline patterns.
The velocity contour plots for different Nb depicted in Figures 6.16 - 6.21 elucidate the
differences in overall flow patterns in the impeller at different Nb . The key distinguishing
feature is the increased flow separation on the blades as Nb is decreased. At Nb = 20, there
139
7
20 blades
6 30 blades
35 blades
5
40 blades
4
Torque [Nm]
0 first second
stage stage
−1
−2
120 140 160 180 200 220 240 260 280
Azimuthal blade positions [deg]
Figure 6.14: Influence of blade number on the power extraction in the impeller at the
maximum efficiency point (199 RPM). The data are computed at each blade position and
normalized with the torque for 30 blades.
70 20 blades
30 blades
60 35 blades
40 blades
50
40
30
20
10
0
1 2
Impeller stage
Figure 6.15: Influence of number of blades on percentage power production in the impeller
stages.
140
Figure 6.16: Contours of the magnitude of the mean water velocity in the impeller with
20 blades at the maximum efficiency point. Note the suction side separation near the right
nozzle lip, where the power extraction is lower compared to 30 or 35 blades.
Figure 6.17: Water velocity vectors at the second stage of the impeller with 20 blades at the
maximum efficiency point illustrating the regions of main flow separations in the impeller of
the 0.53 kW turbine.
141
Figure 6.18: Streamlines superimposed on the water velocity contours at the second stage of
the impeller with 20 blades at the maximum efficiency point. Note that the inlet flow angle
at the entry to the second stage in the lower portion of the flow is less than or almost equal
to the inner blade angle in the impeller of the 0.53 kW turbine.
Figure 6.19: Contours of the magnitude of the mean water velocity in the impeller with 30
blades at the maximum efficiency point.
142
143
is some flow separation on the blades of the first stage for 120◦ ≤ ψ ≤ 150◦ , and significant
separation on the suction side of the blades at ψ = 190◦ . At the second stage, blades passing
through ψ = 228◦ have suction side separation. Around this azimuthal position at the second
stage, the inlet flow angle is very small as is evident from the plots of velocity vectors and
streamlines as shown in Figures 6.17 and 6.18. In addition, there is water-air free-surface,
the exiting flow has not covered fully the blades. Therefore, there is a significant suction
side separation all along the blade length. As Nb is increased to 30, the flow is almost fully
attached at both stages as shown in Figure 6.19. Negligible flow separation is present near
the nozzle throat: 120◦ ≤ ψ ≤ 150◦ at the first stage, where β1 is greater than the outer
blade angle. Similarly, the amount of separation has decreased considerably in the second
stage. This implies that the separation in the second stage is strongly dependent on Nb .
As shown in the Figure 6.20, the impeller with Nb = 35 has produced fully attached flow,
as well as more power extraction from both stages. Water velocity contour and velocity
vector plots reveal that flow separation around the uppermost blade of the second stage, is
significantly intensified as Nb is decreased from 30 to 20. It is also observed that there is a
large recirculation region around the uppermost blade at the second stage mainly due to the
trapped flow from the first stage. It is important to note that although fully attached flow
is the main design goal, only an optimum Nb produce higher performance. For example, the
flow is fully attached for Nb = 40 as shown in Figure 6.21, but the power extraction from
both stages has decreased. In contrast to this, the impeller with 35 blades has produced
almost fully attached flow, as well as more power extraction from both stages.
In all the cases considered above, it is observed that the magnitude of streamwise velocity
at the entry to the second is increased as dictated by continuity. Therefore, higher velocity
contours can be seen on the suction sides of the blades from the Figures. This is a desired
transition to favourable pressure gradient, which reduces the tendency to separation. It is
worth noting that most of the power is produced in this flow region. As the flow in the
second stage is confined to a narrow tangential region, a large number of blades tend to
144
extract power more efficiently. This is because inlet blade angle β2i at the second stage
varies significantly, and the opportunity to extract power is improved by increasing Nb . This
emphasizes the importance of optimum Nb on limiting the flow separation on the blades and
improving the opportunity to capture power from the flow in both stages. At greater than
optimum Nb , it is likely that boundary layer effects, such as blockage in the passage area
due to growth in boundary layer thickness, may be significant, which reduces the overall
efficiency. The details of the effect of boundary layer development at Nb was not pursued in
this study. In addition, the increased weight and frictional effects may be important.
6.2.3 Influence of Outer Blade Angle (β1b )
The outer blade angle β1b is the inlet angle at the first stage and the exit angle at the
second stage. As discussed in Section 5.1, for the maximum efficiency, the inlet flow angle
β1 must be aligned to β1b and the flow is expected to exit radially from the second stage
when viewed by a stationary observer, to avoid efflux of angular momentum. Thus β1b is
important as it can affect the performance of both stages. As given by Equation (5.2),
decreasing β1b decreases the blade radius Rb for the same D1 and D2 and vice versa. The
value of β1b cannot change the ratio of outlet to inlet area for the blades, so all it can do is
affect separation by any difference from the flow angle. Although some degree of separation
is inevitable, if an optimum Nb is selected and β1 and β1b are matched, separation can be
minimized in the impeller, particularly in the first stage. Separation in the second stage is
found negligible, so may not be a critical consideration. The influence of Nb on minimizing
separation was demonstrated in the previous section.
One of the key design principles is the importance of matching β1b to β1 . The results
of Chapter 4 revealed that β1 is significantly greater than β1b near the left nozzle lip, but
decreases azimuthally toward the right nozzle lip below β1b . As such, it is essential to examine
the influence of β1b on the impeller performance at both above and below the designed value:
β1b = 39◦ . For this, computations were performed at β1b = 35, 39 and 43◦ , where 39◦ is that
145
Table 6.1: Impeller design parameters at different outer blade angle β1b
Outer blade angle Blade radius Diameter ratio Inner blade angle
β1b [deg] Rb [mm] D2 /D1 β2 [deg]
35 50 0.68 90
39 52.71 0.68 90
43 56 0.68 90
of the existing design. In the following computations, the nozzle and other impeller design
parameters were kept the same as that of the original turbine. Similarly, the same H and Q
were used. The summary of design calculations for the new impellers are presented in Table
6.1. It is noted that inlet flow angle β1 for these cases is the same for a given impeller speed
because the same nozzle is used. This was given in the previous section.
The influence of β1b on the turbine efficiency is shown in Figure 6.22. When β1b was
increased from 39 to 43◦ , the maximum efficiency decreased significantly from 88 to 85%,
and occurred at a lower impeller speed N = 186.4 RPM than at N = 199.1 RPM for β1b =
39◦ . For β1b = 35◦ , almost the same maximum efficiency was obtained as that of β1b = 39◦
at the same N . Nevertheless 39◦ is likely in the vicinity of the optimal β1b . To explain the
differences in efficiency, the changes in the flow structures in the impeller and the behaviour
of power extraction at the maximum efficiency points were examined.
Figure 6.23 shows the azimuthal variation of power production in the impeller at different
β1b at the maximum efficiency points. Similarly, the percentage of total power production in
the stages is shown in Figure 6.24. In terms of total power production in the first stage, the
performance of the first stage is almost same for all the cases, but the power productions
in the second stage are different among the cases. Second stage power production is only
marginally different for β1b = 35 and 39◦ , whereas the performance has decreased for β1b =
43◦ . It is seen from Figure 6.24 that the first stage contribution is highest at β1 = 43◦ among
the cases. It is concluded that the values greater than β1b = 39◦ reduce the second stage
contribution.
146
90
89
88
87 35 degree
Efficiency [%]
39 degree
86
43 degree
85
84
83
82
81
80
185 190 195 200 205 210 215
Figure 6.22: Influence of outer blade angle β1b on the efficiency of the turbine [H = 1.337
m, Q = 46 lps].
0.12
43°: 186 RPM
0.1 °
39 : 199 RPM
35°: 199 RPM
0.08
Power [kW]
0.06
0.04
0.02
0
first second
−0.02
stage stage
−0.04
120 140 160 180 200 220 240 260 280
Figure 6.23: Influence of outer blade angle β1b on the power production in the first and
second stages of the impeller at the maximum efficiency points [H = 1.337 m, Q = 46 lps].
147
80
°
43

70 °
39
60 35°
50
40
30
20
10
0
1 2
Impeller stage
Figure 6.24: Influence of outer blade angle β1b on the percentage power production at the
first and second stages at the maximum efficiency points [H = 1.337 m, Q = 46 lps].
The velocity contours in the impeller at the maximum efficiency points for different β1b are
shown in Figures 6.25 - 6.28. As shown in the Figures, β1 in 126◦ ≤ ψ ≤ 140◦ is significantly
greater than β1b = 35◦ compared to β1b = 39 and 43◦ . As a result, more flow separation
has occurred on the pressure sides of the blades passing through this region. Similarly, the
amount of separation on the suction sides of the blades around the right nozzle lip, i.e. in
200◦ ≤ ψ ≤ 210◦ , has increased at β1b = 43◦ than 39 and 35◦ . In overall, there is not much
flow separation between the cases in the first stage, except near the left nozzle lip and right
nozzle lip. Therefore the changes in power production in these two regions can be observed
for different β1b .
Most of the power production in the second stage takes place in the azimuthal region:
230◦ ≤ ψ ≤ 250◦ as shown in Figure 6.23. Also in this region, there is a noticeable variation
in power production between the cases. This is the region at the second stage, where the
flow coming from the first stage: 126◦ ≤ ψ ≤ 150◦ produces power. At β1b = 43◦ , slightly
more power is produced near the nozzle throat compared to other β1b . This is the reason
that β1b = 43◦ produces less power from the second stage. Following the same argument,
at β1b = 35◦ , more power is produced from the second stage compared to 39◦ . At β1b =
148
Figure 6.25: Water velocity contours in the impeller at β1b = 43◦ . Note the suction side
separation near the right nozzle lip, responsible for decrease in power extraction.
35 and 39◦ , the blades produce negative power after ψ = 260◦ , whereas the blades for β1b
= 43◦ start producing negative power after they pass through ψ = 250◦ . β1b = 43◦ thus
produces slightly less power than β1b = 39◦ . The conditions for negative power and positive
power production are explained in Section 4.1. As shown in Figure 6.26, for the last two
blade positions in the second stage, looking at the velocity vectors at the leading and trailing
edges gives strong hints that the blade rotation is opposed by the flow (counter swirl), which
means that the blades passing through these positions lose power to the flow. As seen by a
stationary observer, as the flow is in the direction of the rotation which means loss of power.
The tangential velocity at the leading edges of these blades is negative or opposite to the
impeller motion. Similarly, at the trailing edges, where the ideal flow is expected to exit
radially, the flow has deflected by more than 150◦ . This implies that the flow has opposed
the impeller rotation, due to which the impeller power is lost.
In summary, it is concluded that increasing the outer blade angle β1b from 39◦ is not
favourable because it produces less power from the second stage. Also β1b = 35◦ is not
149
Figure 6.26: Water velocity vectors in the impeller at β1b = 43◦ , illustrating the flow sepa-
ration on the suction sides of the blades near the right nozzle lip and the second stage.
Figure 6.27: Water velocity contours in the impeller for β1b = 39◦ at maximum efficiency.
150
Figure 6.28: Water velocity contours in the impeller for β1b = 35◦ at maximum efficiency.
a good choice as it produces slightly lower power from the first stage although producing
slightly more power from the second stage compared to β1b = 39◦ . Thus it is concluded that
β1b = 39◦ is approximately optimal to obtain maximum efficiency.
6.2.4 Influence of Inner Blade Angle (β2b )
In practice, the inner blade angle β2b is generally set at 90◦ . The close proximity of the
first and second stage may relate to outlet blade angle being 90◦ . As β2b impacts the inlet
flow angle at the second stage and hence its performance, it is one of the parameters that
can be adjusted to improve the second stage performance. For circular arc blades, reducing
β2 increases the blade radius Rb , whereas increasing β2b reduces Rb . Since β2b changes the
direction of the flow and the inlet flow angle at the second stage, it is an important parameter
for the performance of the second-stage.
The influence of β2b on the impeller performance was investigated by performing com-
putations at β2b = 85◦ . Totapally and Aziz (1994) experimentally showed that β2b = 55◦
151
Table 6.2: Impeller design parameters at different β2b
Inner angle Blade radius Diameter ratio Outer angle

β2b [deg] Rb [mm] D2 /D1 β1b [deg]
85 57.00 0.68 39
0.14
90 °
0.12
85 °
0.1
0.08
Power [kW]
0.06
0.04
0.02
−0.02
first second
stage stage
−0.04
−0.06
120 140 160 180 200 220 240 260 280
Figure 6.29: Azimuthal variation of power production in the impeller at the maximum
efficiency point for β2b = 90 and 85◦ .
would give an efficiency of more than 90%, whereas it is 88% at β2b = 90◦ . However, it is not
clear whether β2b was changed by keeping all other parameters (β1b and D2 /D1 ) the same
or not. In this study, computations were performed at β2b = 85◦ with D2 /D1 = 0.68 and β1b
= 39◦ . A summary of the design calculation for the impeller is presented in Table 6.2. The
key findings are summarized below.
At β2b = 85◦ , efficiency was significantly decreased from 88 to 83%, in contrast to the
experimental results of Totapally and Aziz (1994), who obtained a maximum efficiency of
about 86%. However, further simulations were not performed at lower values of β2b . The
behaviour of power extraction is shown in Figure 6.29. The changes in the flow field and the
reasons for the efficiency drop are explained as follows.
It is evident from Figure 6.29 that the power extraction from the first stage has slightly
increased for β2b = 85◦ , but the power production from the second stage has decreased by a
large amount. This is better illustrated in the plot of water velocity streamlines and velocity
152
Figure 6.30: Close view of water velocity streamlines superimposed on the velocity contour
map in the second stage for β2b = 85◦ at maximum efficiency, illustrating that a considerable
portion of the flow at the entry of the second stage is misaligned with the blades [inlet flow
angle β2i greater than β2b = 85◦ ] that results into significant performance loss in the second
stage.
vectors in Figures 6.30 and 6.31. The fact that the inlet flow angle β2i at the second stage is
greater than β2b = 85◦ , resulting into performance loss early on than at β2b = 90◦ . As can be
seen, the blades lose power after they pass through ψ = 250◦ . It is noted that a significant
amount of flow is passing through this region. Thus it is concluded that β2b = 90◦ is nearly
optimal value.
6.2.5 Influence of Impeller Diameter Ratio (D2 /D1 )
Inner to outer diameter ratio D2 /D1 affects the blade radius Rb if the outer diameter D1 , the
inner blade angle β2b and the outer blade angle β1b are kept the same. For the same outer
diameter D1 , β1b and β2b , decreasing D2 /D1 increases Rb and the chord-length of the blade,
whereas increasing D2 /D1 reduces Rb and the chord-length. Therefore, D2 /D1 ratio has an
153
Figure 6.31: Velocity vectors in the impeller at the maximum efficiency point for β2b =
85◦ , illustrating that a considerable portion of the flow at the entry of the second stage is
misaligned with the blades [inlet flow angle β2i greater than β2b = 85◦ ] that results into
significant performance loss in the second stage.
influence on the boundary layer separation, flow deflection and the stage performance. More
precisely, D2 /D1 ratio affects the curvature and length of the blade passage, i.e. nozzle-like at
the first stage and diffuser-like at the second stage. For example, higher D2 /D1 ratio means
shorter and less converging blade passages at the first stage and less diverging passages at
the second stage and vice-versa.
The influence of D2 /D1 on the impeller performance was investigated by computing the
flow at D2 /D1 = 0.64, 0.68, and 0.72, where D2 /D1 = 0.68 is that of the 0.53 kW turbine.
In all computations, D1 , β1b , β2b and the nozzle design were kept the same. The calculations
of the design parameters are summarized in Table 6.3. The influence of D2 /D1 was studied
by Desai (1994)), who experimentally found that D2 /D1 = 0.68 gave the highest efficiency.
The influence of D2 /D1 on the impeller performance are summarized in Figures 6.32
and 6.33. When D2 /D1 was increased from 0.68 to 0.72, the maximum efficiency decreased
154
Table 6.3: Impeller design parameters at different D2 /D1
Diameter ratio Blade radius Outer angle Inner angle

D2 /D1 Rb [mm] β1b [deg] β2b [deg]
0.64 57.88 39 90
0.68 52.71 39 90
0.72 47.22 39 90
0.14
0.72
0.12 0.68
0.64
0.1
0.08
Power [kW]
0.06
0.04
0.02
0 first second
stage stage
−0.02
−0.04
120 140 160 180 200 220 240 260 280
Figure 6.32: Influence of diameter ratio D2 /D1 on the impeller performance at the maximum
efficiency point.
significantly from 88% to 83%. Similarly, when D2 /D1 was decreased to 0.64, no appreciable
change in the maximum efficiency was found, which indicates that there must be an optimum
D2 /D1 between 0.64 and 0.68. A comparative analysis of the flows in the impeller explains
how these differences in impeller performance arise. It is of interest to examine the changes
in flow regimes in the impeller at the maximum efficiency points for each D2 /D1 ratio.
It was shown in Section 4.3 that the β1 is higher than β1b near the nozzle throat. Figure
6.32 shows that the power production of the first stage for 126◦ ≤ ψ ≤ 210◦ has decreased
considerably for D2 /D1 = 0.72 compared to the case with D2 /D1 = 0.68. This implies that
decrease in blade length and increase in Rb is strongly related to blade loading. It can be
seen that the power production in the azimuthal range 190◦ ≤ ψ ≤ 210◦ is relatively lower
155
0.72

70
0.68
60 0.64
50
40
30
20
10
0
1 2
Impeller stage
Figure 6.33: Influence of diameter ratio D2 /D1 on the percentage of total power production
at the first and second stages at the maximum efficiency point.
compared to the case D2 /D1 = 0.68. It is worth noting that this is the azimuthal range
where the first stage most efficiently extracts the power because the inlet flow angle β1 is
matched with the blade angle. Although separation is reduced over a larger extent of the
impeller inlet, a significant separation on the suction sides of the blade for 198◦ ≤ ψ ≤ 210◦ is
observed. As a result of this separation, the blade loading is significantly reduced because the
amount of separation is directly related to the blade loading or power production. Therefore,
the power production has reduced considerably.
At the second stage, it can be observed that there is a sharp increase in power production,
in 220◦ ≤ ψ ≤ 238◦ compared to 0.68, where the blades produce most of the power. Then,
in 238◦ ≤ ψ ≤ 250◦ , there is a sharp drop in power production. After ψ = 250◦ , the blades
lose power to the flow. However, the power loss is relatively less compared to the 0.68 case.
Since the second stage is extracting power in a relatively narrow region, in overall D2 /D1 =
0.72 produces less power than at 0.68.
At the second stage, as the blades pass through 238◦ ≤ ψ ≤ 250◦ , there is a sudden drop
in power production. This is the region where the high velocity fluid coming from the first
stage, i.e. 126◦ ≤ ψ ≤ 150◦ must produce most of the second stage power, but this is not
156
the case. Therefore, D2 /D1 = 0.72 produces less power from the second stage. Similarly, the
blades produce negative power after ψ = 250◦ , whereas the blades for D2 /D1 = 0.68 produce
negative power after they pass through ψ = 250◦ . Overall, D2 /D1 = 0.72 thus produces less
power. The conditions for negative power production has been explained in Section 4.1. For
the two last blade positions, the velocity vectors at the leading and trailing edges reveal that
the blade rotation is opposed by the flow, resulting into a loss of power in the flow. The
tangential velocity at the leading edges of these blades is negative or opposite to the impeller
rotation. Similarly at the trailing edges, where the flow is expected to exit radially with no
angular momentum, the flow is deflected by more than 150◦ .
When D2 /D1 was reduced to 0.64, the performance of the first stage is only marginally
improved and follows the same trend as that of D2 /D1 = 0.68 over the entire azimuthal
positions of the first stage: 126◦ ≤ ψ ≤ 210◦ . Within 238◦ ≤ ψ ≤ 250◦ , the performance
of the blades with D2 /D1 = 0.64 has sharply increased whereas that of D2 /D1 = 0.68 has
remained uniform, which is better. After that the performances of both cases have dropped
with the same trend. This gives a strong hint that there exists an optimum D2 /D1 that
gives higher efficiency than either 0.64 or 0.68.
In summary, it was found that increasing D2 /D1 from 0.68 is not favourable as it produces
less power from the first stage. In addition, the second stage performance also deteriorated.
However, decreasing D2 /D1 to 0.64 showed promising results, in the sense that first stage
performance was marginally improved while the performance of the second stage is similar to
that of the reference case D2 /D1 = 0.68. This indicates that optimal D2 /D1 exists between
0.64 and 0.68, but the variation in efficiency over this range is small. To state more precisely,
the beneficial effects of decreasing D2 /D1 from 0.68 to 0.64 on impeller performance come
mainly from the increased power extraction from the first stage. Although counter torque was
not reduced, the findings provide a rigorous explanation for searching an optimum D2 /D1
between 0.64 and 0.68. In this study, 0.68 is taken as an optimum value.
157
6.2.6 Performance at Part-Load - Slider Control Mechanism
For a particular nozzle design without any control of the inlet flow, Equation (5.18) shows
that there exists a unique flow rate Q at a given head H, which gives a maximum efficiency.
Further, it was shown in Section 5.2 that for each H, there exists a unique impeller speed
N that maximizes the efficiency. This is because at this Q and N , the inlet flow angle β1
matches the outer blade angle β1b . However, the performance of the impeller deteriorates
at part-flow conditions if the inlet flow is not controlled to match the velocity and the flow
angle with the impeller design. This is the main reason for using inlet flow control devices,
such as guide vanes in the nozzle. However, as a limitation, such guide vanes in the nozzle
passage cannot precisely control the nozzle throat h0 (does not change at all) and the impeller
inlet area or the entry arc angle θs , which are the important design parameters governing the
inlet flow conditions. In addition, putting guide vanes inside the nozzle also results into extra
losses due to viscous dissipation and trailing-edge wakes. It is noteworthy that the turbines
with highest efficiency do not have guide vanes (e.g. Desai (1994)) and the turbines that
do have guide vanes have low efficiency (e.g. Choi et al. (2008) and Acharya et al. (2015)).
Ability to control precisely h0 and entry arc angle θs , equivalently the velocity and inlet flow
angle, is crucial to keep the turbine efficiency at maximum level during part-flow operations.
One mechanism that has potential to precisely control the inlet flow and minimize losses is
the slider shown in Figure 5.2, which can be installed at the inlet periphery of the impeller
and rotated circumferentially to adjust the nozzle throat h0 and the impeller inlet arc angle
θs during part-flow operations (Sinagra et al., 2014). However, it will modify the original
nozzle shape, the impact of which on efficiency needs to be characterized.
To investigate the impeller performance of the 0.53 kW turbine at part-flow conditions
with and without the slider, computations were performed at three different flow rates: Q =
40, 30, and 20 lps with the same design head: H = 1.33 m. It is noted that the design-point
flow rate Q is 46 lps. The nozzle and the impeller were kept the same as that of the 0.53
kW turbine. For modeling the slider, only the portion of the circular slider segment covering
158
Table 6.4: Operating parameters for the slider control
Q [lps] h00 [mm] θs0 [deg]

40 77.4 78.4
30 58.0 58.7
20 38.7 39.1
the entry arc was modelled to reduce the size of the computational mesh as well as the
complexity of mesh generation around the thin circular section. The typical slider modelled
in the computational domain looks like the one shown in Figure 6.34. When the slider is
used to regulate Q, both the nozzle throat h0 and the nozzle entry angle θs will change. The
slider should be rotated in such a way that the new nozzle throat h00 and the new entry arc
angle θs0 will conform to the condition that all the head H is approximately converted into
kinetic energy. The analytical model for nozzle design presented in Chapter 5 was used to
calculate approximately h00 and θs0 . For example, at the design point Q = 46 lps, h0 = 89.44
mm and θs = 90◦ . It is necessary to compute h00 for the part-flow Q0 = 40 lps. Using the
continuity, for these two conditions in order to get the same tangential velocity U0 at the
entry to the impeller, h00 is computed as 78.26 mm. Then for this value of h00 , the entry arc
angle θs0 = 78◦ . The design calculations for h00 and θs0 for different Q are summarized in Table
6.4. It is also possible to determine the impeller speed required for maximum efficiency. The
speed is calculated by using Equation (5.19). This is important because only by maintaining
a particular impeller speed, maximum efficiency can be achieved. So for the same head H,
maximum efficiency will be obtained at the same speed, provided that all H is converted
into kinetic energy.
The efficiencies of the turbine at different Q with and without the flow control mechanism
are summarized in Figures 6.35 and 6.36. Simulations for the case without the inlet flow
control were performed only at Q = 40 lps, where the efficiency dropped significantly. This
is the first documentation of poor part-load efficiency without flow control. The reason is
that H is not fully converted into kinetic energy because the nozzle throat area is too large
for that particular Q. The maximum efficiencies of 86%, 88.10% and 88.7% were obtained
159
Figure 6.34: Close view of the slider at the inlet of the impeller [Q = 20 lps, H = 1.337 m].
Note that only the inlet section of the slider is modelled.
90
20 lps
89 30 lps
40 lps
88
87
Efficiency [%]
86
85
84
83
82
81
80
185 190 195 200 205 210 215
Figure 6.35: Part-flow efficiencies at different impeller speeds of crossflow turbine with slider
control mechanism for the inlet flow.
160
90
88
With slider control
86
Without slider control
84
Efficiency [%] 82
80
78
76
74
72
70
20 25 30 35 40 45 50
Flow rate [lps]
Figure 6.36: Part-flow efficiencies of crossflow turbine with slider and without slider control
mechanism.
respectively at Q = 20, 30 and 40 lps. As suggested by the analytical model, maximum

efficiency occurred at the same impeller speed as that of the original design. The results
showed that the impeller performance did not deteriorate during part-flow operations from
that of the maximum Q when the inlet area is controlled. These results are consistent
with the results obtained by Sammartano et al. (2016) for the slider at part-flow operating
conditions. They showed that by using the slider, maximum efficiency was maintained at
about 84% down to 30% part-flow operation, whereas the maximum design-point efficiency
of the turbine was reported to be about 87%.
The contribution of power extraction from each stage is depicted in Figure 6.37. An in-
teresting behaviour of the impeller can be observed at lower Q. The percentage contribution
of the first stage on total power output is significantly higher for part-flow conditions than at
the design condition. The contribution of the first stage is highest at Q = 20 lps and gradu-
ally decreases as Q increases. The first stage contribution to the total power can be as high
as 86%. In Section 6.2.1, it was found that if β1 is uniform and matches with the outer blade
angle, the percentage contribution would be high at the first stage. The performance of the
161
90
20 lps
80 30 lps

40 lps
70
60
50
40
30
20
10
0
1 2
Impeller stage
Figure 6.37: Stage performances at various part-flow operating conditions.
impeller at part-flow operations is explained with the help of nozzle velocities and flow angle.
The computed total velocity u0 at different part-flow operations is shown in Figures 6.38 -
6.40. Similarly, the corresponding inlet flow angles β1 at impeller inlet are shown in Figure
6.41 - 6.43. The contour maps of the magnitude of the mean water velocity in the impeller
at the maximum efficiency points are shown in Figures 6.44 - 6.46. At part-flow conditions,
it is seen from the Figures that the β1 is more uniform at the inlet at Q = 20 lps than at
40 lps. However, the computed absolute velocity at the inlet is approximately close to total
√
inlet velocity (ut = 2gH) in all the cases, implying that the slider mechanism is good at
converting head into kinetic energy. It is noteworthy that the simple analytical Equation
(5.18) for nozzle design implies that maximum efficiency occurs at the same impeller speed
for all part-load operations, provided that all the head H is converted into kinetic energy.
The present results of the slider analysis also agree with this simple analysis.
It is thus concluded that the capability to maintain high efficiency during part-load
conditions is the key feature of the slider control mechanism compared to the guide vanes.
This suggests that the slider is a promising simple flow controller for future turbines.
162
6
4
Absolute velocity
Radial velocity
Velocity [m/s]
3
Tangential velocity
2 Total velocity
−1
−2
−3
170 180 190 200 210
Entry arc azimuthal position [deg]
Figure 6.38: Velocity at the impeller inlet [Q = 20 lps, H = 1.337 m].
4
Absolute velocity
Velocity [m/s]
3 Radial velocity
Tangential velocity
2 Total velocity
−1
−2
−3
150 160 170 180 190 200 210
163
60
Inlet flow angle
Outer blade angle
50
Flow angle [deg] 40
30
20
10
0
130 140 150 160 170 180 190 200 210
60
Inlet flow angle
Outer blade angle
50
Flow angle [deg]
40
30
20
10
0
165 170 175 180 185 190 195 200 205 210
Figure 6.41: Inlet flow angle at the impeller inlet [Q = 20 lps, H = 1.337 m and N = 199.1].
164
60
Inlet flow angle
Outer blade angle
50
Flow angle [deg]

40
30
20
10
0
150 160 170 180 190 200 210
60
Inlet flow angle
Outer blade angle
50
Flow angle [deg]
40
30
20
10
0
130 140 150 160 170 180 190 200 210
165
Figure 6.44: Contour maps of the magnitude of the mean water velocity at the maximum
efficiency point for Q = 20 lps with slider.
efficiency point for Q = 30 lps with slider.
166
efficiency point of Q = 40 lps with slider.
6.2.7 Synthesis of Results
The analysis of the previous sections in this chapter revealed that the nozzle design, the
number of blades, the outer blade angle, and the diameter ratio are important design pa-
rameters for the turbine performance. Since it was shown that efficiencies of over 90% could
be achieved by suitable choices of these parameters, it is likely that they are the parame-
ters of first order for the determination of crossflow turbine efficiency. As the problem is
multi-dimensional, however, it cannot be ruled out that a combination of the best param-
eter values found above would yield the most efficient turbine design or that even higher
efficiencies could be achieved by considering additional parameters. Nevertheless, the results
show that the design goal of obtaining efficiencies of over 90% has been been achieved. To
make the design analysis more tractable, the turbine design with new nozzle and the impeller
with 35 blades would be a starting point in the computational simulations. Then another
important design parameter, the outer blade angle can be altered to match the inlet flow
167
conditions for further design improvement. Finally, D2 /D1 can be chosen 0.68 as it is already
close to an optimum value.
6.3 Improved Design: 0.53 kW Turbine
In the sensitivity analysis, the influence of varying a particular design parameter on the
impeller performance was investigated, but what combination of those parameters give the
best impeller performance still remains to be studied. It was shown that there exists a local
optimal value for each design parameter, which improved the maximum efficiency of the
turbine. As the problem is multi-dimensional, it can not be ruled out that a combination of
the optimal parameters found above would yield the most efficient turbine. To state more
precisely, the use of optimal parameters may not be simply additive. However, the sensitivity
analysis indicates that the design with a new nozzle and the impeller with 35 blades will
improve the efficiency as well as guide towards the optimal design. This is because of the fact
that the new nozzle improved the efficiency of the impeller with 30 blades. It’s also likely
that matching β1 with the outer blade angle would further improve the performance. Thus
a design strategy taken here is to perform computational simulations with the new nozzle
design, Nb = 35, β1 = 39◦ , and D2 /D1 = 0.68. Thus the simulations were performed for this
design. The operating parameters used in the analysis are the same as that of the original
0.53 kW turbine, and are given in Table 4.4. The major findings of the computations are
summarized as follows.
Figure 6.47 shows the comparison of efficiencies of the existing design and the new nozzle
design with the existing impeller of 35 blades. By improving the nozzle design and increasing
Nb from 30 to 35, a maximum efficiency of 90.61% was obtained. It is noted that by re-
designing the nozzle without altering the impeller, 89.45% efficiency was obtained in Section
6.2.1. By increasing Nb from 30 to 35 without altering the original nozzle, maximum efficiency
of 89.87% was obtained. Thus the performance gain can be attributed to increased angular
168
100
New nozzle:35 blades
98 Original design: 30 blades
96
Efficiency [%] 94
92
90
88
86
84
82
80
185 190 195 200 205 210 215
Figure 6.47: Comparison of the efficiency of the new nozzle and the existing design. The
improved design has new nozzle and the impeller with 35 blades.
120
New nozzle: 35 blades
100 Original design: 30 blades
80
60 power loss
Power [W]
40
20
0
first second
stage stage
−20
−40
120 140 160 180 200 220 240 260 280 300
Figure 6.48: Comparison of azimuthal variation of power production in the impeller of the
new design and the existing designs. The improved design has the new nozzle and the
impeller with 35 blades.
169
Figure 6.49: Streamlines superimposed on the contour map of the magnitude of the mean
water velocity at the maximum efficiency point. Note that the flow is perfectly aligned with
the blade in both stages. The improved design has the new nozzle and the impeller with 35
blades.
Figure 6.50: Water velocity vectors in the impeller at the maximum efficiency point, illus-
trating the main flow in the impeller.
170
momentum flux at the impeller inlet, more uniformity in β1 and its better match with the
outer blade angle, and enhanced power extraction with Nb = 35. The performance gain can
be elucidated by examining the power extraction and the flow field in the impeller.
The comparison of power production in the impeller for the existing design and new
design is shown in Figure 6.48. It is noteworthy that in the azimuthal range 200◦ ≤ ψ ≤
215◦ at the first stage, the power production in the case of new nozzle with 35 blades is
relatively significant than the original design. As a result the power extraction from the first
stage has improved. However, at the second stage, the trend is almost similar for both cases.
The only difference is that the flow is more deflected to upward region of the second stage.
The power production/loss in the second stage is similar in both cases. It is observed that the
flow is fully attached in both stages. The water velocity contour plots and velocity vectors
in the improved design are shown in Figures 6.49 and 6.50. The improved performance gain
comes mainly from the first stage as shown in Figure 6.48.
6.4 Improved Design: 7 kW Turbine
This section presents the design strategy for improving the performance of the 7 kW turbine
followed by the results of computational simulations. The design features of the 0.53 kW
turbine, its key flow patterns, and the results of sensitivity analysis provide the key design
concepts for improving the design of 7 kW turbine.
The computational analysis presented in Chapter 4 showed that the nozzle performance
of the original turbine was very poor, which is identified as the main cause for the poor
impeller performance. Firstly, the existing nozzle design of the 7 kW turbine does not
fully convert the head H into kinetic energy, which results into lower angular momentum
flux at the impeller inlet. In contrast to this, the nozzle of the 0.53 kW turbine (88%)
converts significantly more H into kinetic energy. Secondly, the inlet flow angle β1 of the
7 kW turbine is significantly greater than the outer blade angle β1b = 30◦ and the impeller
171
has insufficient number of blades to extract efficiently the power from the flow, which has
resulted into considerable flow separation on the blades at the first stage. In addition, counter
torque was found at both stages after the maximum efficiency point (N = 450 RPM), which
significantly reduced the impeller performance. These are the primary performance limiting
flow mechanisms. The insights gained from the sensitivity analyses of the 0.53 kW turbine
also suggest that both the nozzle and the impeller of the 7 kW turbine are poorly designed.
The operating parameters for the design analysis presented here are the same as that of the
original turbine, and are given in Table 4.2. The major findings of this study are presented
as follows.
6.4.1 Influence of Nozzle Design on the Efficiency of Impeller
To quantify the influence of nozzle design on the performance of the impeller, a new nozzle
was designed using the analytical model presented in Section 5.2, whose design characteris-
tics follow the analysis of Section 6.2.1 and are similar to that of the new nozzle of 0.53 kW
turbine. Computations were performed without altering the original impeller design. This
analysis allows the identification of the major changes in the impeller flow field, primarily
the flow separation and flow deflection to the second stage, and the performance. Multi-
ple computational evaluations can be avoided or very poor solutions can be eliminated by
following this procedure.
It was shown in Chapter 4 that the computed total velocity u0 at the impeller inlet is
√
significantly lower (≈ 12 m/s) than total inlet velocity 2gH = 14 m/s, implying that there
is only partial conversion of H into kinetic energy. In the existing design, for the design
flow rate Q = 105 lps and head H = 10 m, the continuity shows that the nozzle throat area
(h0 W ) was larger than needed. Similarly, the impeller inlet area R1 θs W was smaller than
needed, resulting in higher ur and β1 . As a result, H was not fully converted into kinetic
energy. Thus the throat h0 , the width W , and the nozzle entry arc (θs ) were computed using
the criterion for total conversion of H into kinetic energy [Equation (5.12)]. Following the
172
Y
h0 = 83 mm
flow
220
0
0
80
45
Z
X
Figure 6.51: Schematic illustration of the new nozzle design for the 7 kW turbine [Q = 105
lps and H = 10 m].
Table 6.5: New nozzle design parameters

Nozzle throat: h0 [mm] 83
Nozzle width: W [mm] 94.34
Nozzle entry arc: θs [deg] 80
Angle of lower inlet wall: δ [deg] 22
nozzle configuration of the 0.53 kW turbine, the aspect ratio W/h0 was selected as 1.14. As
θs needs to be increased, so θs = 69 was increased to 80◦ . Then using the criterion for total
conversion of H into kinetic energy [Equation (5.12)], throat h0 was computed as 83 mm [Q
= 105 lps and H = 10 m]. The nozzle and the impeller configuration is schematically shown
in Figure 6.51. A summary of design calculations for the new nozzle is presented in Table
6.5.
The influence of the new nozzle design on the performance of the original impeller was
found significant. By re-designing the nozzle without altering the impeller, the maximum
173
15
10 Computed total inlet velocity

Theoretical: total inlet velocity
Velocity [m/s] Tangential velocity
Radial velocity
5
−5
140 150 160 170 180 190 200 210

at the impeller inlet with the new nozzle [Q = 105 lps and H = 10 m].
efficiency was improved from 69 to 87%. This implies that the influence of the nozzle design
or the inlet flow conditions is significant for the downstream impeller performance. The
reasons for this performance gain are now examined in more detail.
The computed total, radial and tangential flow velocities at the impeller inlet for the
new nozzle is shown in Figure 6.52, the same for the original nozzle is shown in Figure
4.15. In the case of new nozzle, it can be observed that the computed total velocity u0 is
√
approximately equal to the total velocity ( 2gH = 14 m/s), implying that all the head H is
approximately converted into kinetic energy. This also means that the angular momentum
flux at the impeller inlet has increased compared to the original nozzle. It is observed that
computed total velocity u0 is azimuthally uniform but not ur and uθ , which has resulted into
azimuthal variation of β1 . This can be observed in the case of both nozzles. A comparison
of β1 and β1b = 30◦ in the new nozzle and the original nozzle is shown in Figure 6.53. It
is observed that the difference between β1 and β1b has decreased considerably in the new
nozzle. β1 is also more uniform over a large extent of the impeller inlet: 155◦ ≤ ψ ≤ 210◦ ,
except there is still some azimuthal variation in β1 near the left nozzle lip (135◦ ≤ ψ ≤ 155◦ ),
174
80
Computed flow angle: new nozzle
Computed flow angle: existing nozzle
70
Designed blade angle
60
Flow angle [deg]

50
40
30
20
10
0
140 150 160 170 180 190 200 210
Figure 6.53: Comparison of the inlet flow angle β1 and the blade angle at the maximum
efficiency points for the existing impeller with the new nozzle and the existing nozzle. [Q =
105 lps and H = 10 m].
but this variation is considerably lower than in the original nozzle. It is noted that β1 is
about 10◦ greater than β1b over a large extent of the impeller inlet, which implies that β1b
must be increased to about 40◦ to match the inlet flow with the blades. It is also interesting
to note that the azimuthal variation of β1 is very similar to that of the 0.53 kW turbine.
With the new nozzle, there is negligible flow separation in the first stage of the impeller
compared to the original nozzle, except some separation on the pressure sides of the blades
passing through 135◦ ≤ ψ ≤ 155◦ (near the left nozzle lip), where β1 is considerably greater
than β1b (Figures 4.17, 6.54 and 6.55). The separation is less because the difference between
β1 and β1b is lower than in the original nozzle. In the second stage, however, there is a
considerable flow separation on the suction side of the blade passing through ψ = 235◦ . In
this region, β2i is significantly lower than β2b , which results into suction side separation. It is
noted that there is still a considerable difference between β1 and β1b near the left nozzle lip in
the new nozzle. Based on the 90% efficiency of the 0.53 kW turbine, this is not a major issue
for the impeller performance because power is extracted in two stages. The power missed
out at the first stage due to high β1 is captured at the second stage as was found for the
175
Figure 6.54: Streamlines superimposed on the contour map of the magnitude of the mean
water velocity at the maximum efficiency point.
0.53 kW turbine in Section 4.4. To improve power extraction and avoid flow separation on
the blades, Nb must be increased. These aspects will be investigated in the next section.
To investigate the changes in flow structures at the exit of the first stage and the entry
of the second stage due to change in β1 , the flow angles at these locations are plotted in
Figures 6.56 and 6.57 at the maximum efficiency points. It is observed that β2 is about 100◦
in 140◦ ≤ ψ ≤ 195◦ , but has increased gradually to 120◦ in 195◦ ≤ ψ ≤ 220◦ . This reveals
that at lower values of β1 (near the right nozzle lip), β2 increased sharply, which implies
that the flow is highly deflected with about β2 = 120◦ . This is a favourable condition for
efficient power extraction from the first stage. Therefore there is a difference of about 10◦
between β2 and the inner blade angle 90◦ over a large extent of the first stage exit, and upto
20◦ over a small section (near the right nozzle lip) of the first stage exit. This difference is
slightly lower compared to the original nozzle design. It is also observed that there is almost
negligible separation near the blade trailing edges at the first stage, so this is a significant
improvement in reducing the size of wakes compared to the original design.
176
Figure 6.55: Water velocity vectors at the maximum efficiency point, illustrating the flow
separation and flow angle at the entry of the second stage for the existing impeller (20 blades)
with the new nozzle.
150
Computed flow angle
140 Inner blade angle
130
120
Flow angle [deg]
110
100
90
80
70
60
50
140 150 160 170 180 190 200 210 220
Figure 6.56: Comparison of the computed flow angle and the inner blade angle at the exit
of the first stage of the existing impeller with new nozzle at the maximum efficiency point
[N = 500 RPM].
177
100
Computed flow angle
90 Inner blade angle
80
70
Flow angle [deg]

60
50
40
30
20
10
0
230 235 240 245 250 255 260 265
Figure 6.57: Comparison of the computed flow angle and the inner blade angle at the inlet of
the second stage of the existing impeller with new nozzle at the maximum efficiency point.
As shown in Figure 6.56, it is observed that the entry angle β2i to the second stage
has significant azimuthal variation, which is almost linearly increasing from 20 to 80◦ . It is
evident that β2i is less than the inner blade angle, which is favourable in terms of extracting
more power from the flow as long as there is no separation on the blades. The velocity
contours with streamlines and velocity vectors of the resulting flow field in the impeller at
the maximum efficiency point shown in Figures 6.54 and 6.55 reveal that the flow is mostly
aligned with the blades, and there is no major flow separation in the blade passages as found
in the case of original nozzle. Thus the first stage performance has improved significantly.
At the entry to the second stage, the flow is converging and is slightly accelerated. The only
major flow separation in the second stage is on the suction side of the blade passing through
ψ = 235◦ . In this region, β2i is significantly lower than β2b , and the flow has free-surface,
which results naturally into some separation. Furthermore, this separation is not in that
region, where maximum power is extracted from the flow.
A comparison of the impeller in terms of power production with the new and original
nozzles is shown in Figure 6.58. Similarly, a comparison of percentage power production in
178
2.5
Existing nozzle
New nozzle
2
Power [kW] 1.5

power loss
0.5
first second
0 stage stage
−0.5
140 160 180 200 220 240 260 280
Figure 6.58: Comparison of power production in the impeller at the maximum efficiency
points between the existing and the new nozzle designs.
70 Original nozzle
New nozzle
60
50
40
30
20
10
0
1 2
Impeller stage
Figure 6.59: Comparison of percentage power production in the impeller with the original
and new nozzles at the maximum efficiency points.
the impeller between the two cases is shown in Figure 6.59. It is observed that with the new
nozzle design, the power production in the first stage has increased significantly. However,
the performance of the second stage is relatively similar to that of the original nozzle. The
improvement in the first stage performance is due to the increased flow velocity or the angular
momentum flux at the impeller inlet, decrease in the difference between β1 and the outer
179
blade angle, and reduced azimuthal variation of β1 compared to the original nozzle design.
It is observed that the power production in the first stage has increased linearly toward the
right nozzle lip. It is maximum where β1 is small, and minimum where β1 is high. This
observation is similar to that of the 0.53 kW turbine. It is remarkable to note that the
percentage power production in the impeller with the new nozzle is also similar to that of
the 0.53 kW turbine.
In the second stage, it can be observed that there is a sharp increase in power production
in 235◦ ≤ ψ ≤ 255◦ , where the flow is fully attached. Remarkably, in this region β2i is close
to β2b =39◦ , which is the inlet flow angle at the impeller inlet. This is the region where the
flow with high β1 missed at the first stage, i.e. in 135◦ ≤ ψ ≤ 150◦ , contributes to second
stage power. It is seen that β2i is less than the inner blade angle (90◦ ) through the entire
inlet of the second stage, which avoids negative power. This is evident from the plot of
velocity streamlines and vectors. Any portion of the flow that has β2i greater than the outer
blade angle would result in power loss. Looking at the lower portion of the flow at the second
stage, the velocity vectors at the leading and trailing edges reveal that there is only a very
small region where the blade rotation is opposed by the exiting flow. The tangential relative
velocity at the leading edge of the blade is negative or opposite to the impeller rotation,
where β2i is marginally greater than 90◦ . Similarly at the trailing edges, where the flow is
expected to exit radially, the flow is deflected at more than 150◦ . This signifies that the flow
is dragged by the impeller rotation or losing power to the flow. Nevertheless, it is evident
from Figure 6.58 that the power loss in the new design is very small.
It is thus concluded that there is a significant influence of inlet flow conditions on the
first stage performance, particularly in terms of flow separation and power extraction. The
results revealed that the inlet flow is not perfectly aligned with the blades at both stages.
This indicates that by increasing the outer blade angle to match β1 and increasing the number
of blades, the impeller performance can be improved. The results of the sensitivity analysis
of 0.53 kW turbine suggest that the outer blade angle of about 39◦ and 30 - 35 number of
180
blades combined with the new nozzle design most likely give the efficiency above 90%. These
design aspects will be examined in more details in the next section.
6.4.2 Matching of Nozzle and Impeller Designs
From the results presented so far, it is concluded the nozzle performance and mismatching
between the nozzle and the impeller are the major cause for poor turbine performance,
rather than the impeller design itself. The sensitivity analysis presented in Section 6.2 and
the design improvement of 0.53 kW turbine presented in Section 6.3 conclude that the key
to improving the performance is to convert all H into kinetic energy and match the nozzle
and impeller designs, mainly the inlet flow angle β1 and the outer blade angle β1b . This
section will examine the major issue of matching the nozzle and the impeller designs to
enable further improvement of the turbine presented in the previous section.
Given a good nozzle design using the criteria of Section 5.2, flow separation on the
blades is the key indicator for the mismatching and performance loss. The mismatching
was examined in the previous section by changing the nozzle design for the original impeller
and observing the improved flow in the impeller and performance. It was found that β1 is
significantly lower than β1b , an indicator of flow mismatching between the two components.
For a good nozzle, a number of impeller designs can be evaluated to determine how the
flow field/impeller performance responds to changes in the impeller design parameters. The
following computational simulations have been undertaken to elucidate how the optimal
matching between the nozzle and the impeller can improve the efficiency. In all simulations,
the nozzle design is fixed and the impeller design is altered.
The most important parameters identified in Chapter 4 and the previous section for the
7 kW turbine are β1b and the number of blades Nb . Similarly, the key parameters identified
in the sensitivity analysis of the 0.53 kW turbine for improving the impeller performance
are β1b , D2 /D1 and Nb . Since Nb = 30 - 35 was found to give the maximum efficiency for
181
Table 6.6: Design parameters of the new impeller
Design parameter Original New

impeller impeller
External diameter: D1 [mm] 316 316
Diameter ratio: D2 /D1 0.67 0.67
Nozzle lower wall inclination: δ [deg] 16 22
Outer blade angle: β1b [deg] 30 39
Inner blade angle: β2b [deg] 90 90
Blade radius: Rb [mm] 52.14 56.02
Number of blades: Nb 20 35
Nozzle width: W [mm] 150 94.34
Nozzle thickness or throat: h0 [mm] 65 83
Nozzle entry arc: θs [deg] 69 80
the comparable size of the impeller of 0.53 kW turbine 2 , the impeller was designed with 35
blades. Then β1b was calculated based on the inlet flow angle β1 , and the calculation follows
from the nozzle design presented in Section 5.2. Using the Equations (5.13) and (5.18) to
obtain total conversion of head into kinetic energy and match the flow angle with the blade
angle, maximum efficiency occurs after 461 RPM, where the optimum flow angle β1 is about
41◦ . From the analysis of previous section, the maximum efficiency occurred at 500 RPM
and the flow angle varies linearly from 75 to 39◦ as shown in Figure 6.53. As shown in the
Figure, β1 is about 39◦ over a larger portion of the impeller inlet. So β1b = 39◦ is used here.
The design parameters chosen for the new impeller are presented in Table 6.6. The major
findings are summarized as follows.
The performance of the new impeller design is shown in Figure 6.60. A maximum effi-
ciency of 91% was obtained for the new impeller design. The maximum efficiency occurred
at 500 RPM, which is slightly greater than the impeller speed 461 RPM predicted by the
nozzle analysis for optimal operation. Therefore, in a detailed optimization study, the simple
analytical model for nozzle design provides a remarkably good guidance for high-fidelity sim-
ulations. This efficiency gain was achieved through new nozzle design, increasing β1b from 30
to 39◦ , and increasing Nb from 20 to 35. The important flow features and the performance
2
The 0.53 kW turbine has an outer diameter of 304 mm, whereas that of 7 kW turbine is 316 mm.
182
100
95
90
Efficiency [%]
85
80
75
70
250 300 350 400 450 500 550 600
Impeller speed (RPM)
Figure 6.60: Performance characteristics of the improved design [H = 10 m, Q = 105 lps].
behaviour of the new design leading to 91% efficiency will be explained below in more detail
with reference to the original turbine and the 0.53 kW turbine.
A comparison of β1 and β1b = 39◦ at the maximum efficiency point is shown in Figure
6.61. It is observed from 6.61 that β1 is closely matched with β1b over a large extent of the
impeller inlet: 155◦ ≤ ψ ≤ 215◦ . In 135◦ ≤ ψ ≤ 155◦ , β1 varies from 70 to 40◦ . However,
compared to the original design, there is less flow separation in this region. The main reasons
for less separation are 1) relatively smaller difference between β1 and β1b and 2) large number
of blades (Nn = 35) in the new impeller. Similarly, comparisons of flow angles at the exit of
the first stage and the inlet of the second stage with the inner blade angle (90◦ ) are shown in
Figures 6.62 and 6.63 respectively. It is observed the difference between β2 and β2b is about
10◦ over half of the exit region and is becoming more significant (20◦ ) near the right nozzle
lip. This indicates that the first stage performance is good. Similarly, the variation of β2i at
the entry of the second stage is very similar to that of the 0.53 kW turbine.
To compare the behaviour of the impeller in extracting the angular momentum for the
existing design, new nozzle with existing impeller and new nozzle with improved impeller,
183
80
Computed inlet flow angle
70 Outer blade angle
Relative flow angle [deg]

60
50
40
30
20
10
0
140 150 160 170 180 190 200 210
Figure 6.61: Comparison of the computed inlet flow angle β1 and the outer blade angle at
the inlet of the impeller at the maximum efficiency point: N = 500 RPM.
torque per blade was plotted as shown in Figure 6.64. The Figure shows that the first
stage performance has significantly improved by improving both the nozzle and the impeller
designs, whereas the second stage performance has only slightly improved compared to the
existing design and the new nozzle with existing impeller. This also implies that first stage
performance is important.
As a result of azimuthal variation in β1 at the impeller inlet, the power production at
the first stage has varied azimuthally as shown in Figure 6.66, which is similar to that of the
0.53 kW turbine. The power production near the left nozzle lip, 135◦ ≤ ψ ≤ 150◦ , is low due
to high β1 compared to β1b = 39◦ ), and is gradually increasing toward the right nozzle lip,
where β1 is approximately matched with β1b . An important implication of this behaviour is
that small β1 may be beneficial for better blade performance, but its influence on the onset
of suction side separation and the second stage performance is not fully understood. At very
high β2i at the entry of the second stage, greater than β2b = 90◦ , the blades start losing
power to the flow as is evident from Figure 6.66. This has occurred after the impeller speed
525 RPM. Therefore, the second stage performance is poor after 525 RPM. As a result,
184
120
100
Flow angle [deg] 80 Computed flow angle

Inner blade angle
60
40
20
0
140 150 160 170 180 190 200 210 220
Figure 6.62: Comparison of the computed relative flow angle and the blade angle at the exit
of the first stage at the maximum efficiency point: N = 500 RPM. Note that the azimuthal
range: 180 - 220◦ corresponds to the region of maximum power extraction.
120
Computed flow angle
Inner blade angle
100
Flow angle [deg]
80
60
40
20
0
240 250 260 270 280 290
Figure 6.63: Comparison of the computed relative flow angle and the blade angle at the
inlet of the second stage at the maximum efficiency point: N = 500 RPM. Note that the
azimuthal range: ψ = 240 - 260◦ corresponds to the region of maximum power extraction.
185
30
Improved impeller: new nozzle
Existing design
25 Existing impeller: new nozzle
20
Torque [Nm]
power loss
15
10
5
first second
stage stage
0
−5
140 160 180 200 220 240 260 280
Figure 6.64: Comparison of torque production in the impeller between the improved and
the existing impeller with the new nozzle at the maximum efficiency points. Note that the
original impeller has Nb = 20 and β1b = 30◦ , whereas the new impeller has Nb = 35 and β1b
= 39◦ . For the purpose of comparison, the data are computed at each blade position and
normalized with the torque for 35 blades.
the efficiency of the turbine is decreasing. This result is very similar to the performance
behaviour of the 0.53 kW turbine.
At the entry of the second stage, it can be observed from Figure 6.63 that there is a
significant azimuthal variation of β2i . As a result, sharp rise and drop in power extraction
can be observed, particularly at speeds greater than the maximum efficiency point. It is
noted that there is power loss at the lower region of the second stage, where β2i is greater
than the inner blade angle 90◦ , and thus the flow opposes the blades. At low impeller speeds,
the negative power is lower compared to the higher speeds because the flow is more aligned
to the blades. Therefore, power loss increased with increasing impeller speeds, particularly
after 525 RPM, at the second stage. Since the flow is confined within a very narrow region
at the second stage, flow alignment is critically important to efficiently extract power. It is
noteworthy that the flow passing through the inlet region: 135◦ ≤ ψ ≤ 180◦ has significant
186
2
400 RPM
450 RPM
475 RPM
1.5
500 RPM
525 RPM
Power output [kW]
550 RPM
1 575 RPM
0.5
first second
−0.5 stage stage power loss
−1
140 160 180 200 220 240 260 280 300
Figure 6.65: Azimuthal variation of power production in the impeller of the improved design
at different speeds. The impeller speed 500 RPM corresponds to the maximum efficiency
point (91%).
contribution at the second stage. More importantly, the flow is almost perfectly aligned to
the blades in this region of the second stage so that there is no separation on the blades as
depicted in the Figures. It is interesting to note that the flow passing through the azimuthal
range: 180◦ ≤ ψ ≤ 215◦ at the first stage has contributed less power at the second stage as
expected.
To explain the flow mechanism of power extraction in the impeller in more detail, the
flow field was examined more closely with the help of velocity contours, streamlines, velocity
vectors and flow angles. The streamlines superimposed into the water velocity contours at
different impeller speeds, which illustrate the key features of the internal flow, are depicted
in Figures 6.67 - 6.69. By comparing the velocity contours at different speeds, it is observed
that the inlet flow angle β2i at the entry of the second stage has changed. As N increases to
500 - 525 RPM, which correspond to the maximum efficiency points, the flow has become
more narrower and β2i has become lower than the inner blade angle 90◦ at the second stage,
187
Figure 6.66: Contours of the magnitude of mean water velocity at N = 500 RPM corre-
sponding to the maximum efficiency (91%).
(a) 400 RPM (b) 450 RPM
Figure 6.67: Streamlines superimposed into the water velocity contour illustrating the main
flow path in the impeller. Note the increase in the inlet flow angle at the second stage with
the increase in impeller speed.
188
(a) 475 RPM (b) 500 RPM
(a) 525 RPM (b) 575 RPM
189
except at a small lower portion of the exiting flow. It is noteworthy that maximum power has
been extracted in 235◦ ≤ ψ ≤ 260◦ . The velocity contours show that the flow is accelerated
at the entry of the second stage where the maximum power is extracted. This is because
the flow exiting from the first stage is converging toward the second stage. So the flow
velocity is higher on the suction sides. At the exit of this flow region, the flow is exiting
almost radially, implying that residual angular momentum is negligible. This is the reason
that this is the most efficient flow region in the second stage at the maximum efficiency
point. As depicted in the Figures, β2i is decreasing at the upper region with the increase
in N , and thus the performance of the second stage has changed with speeds. At the lower
portion of the exiting flow at the second stage, β2i is greater than the inner blade angle 90◦ ,
resulting into negative torque. Although inner blade angle is designed for 90◦ at the inlet of
the second stage, the inlet flow angle is significantly lower than the inner blade angle, and
the blades are performing better. More interestingly, β2i is similar to β1 at the first stage,
where most of the power is extracted. This is beneficial because the flow can extract more
power at around β1 = 39◦ , similar to the first stage. Therefore there is no separation at the
second stage, where most of the power is extracted. The streamline plots show that the flow
is approximately aligned with the blades at both stages. It is thus concluded that the high
radius of curvature of the blades, a large number of blades, and lower inlet flow angles limit
the potential for flow separations on the blades.
An important conclusion can be made with regard to the improved performance. In the
existing design, counter torque was found at both stages after the impeller speed N = 450
RPM, particularly near the nozzle throat at the first stage and at the lower portion of the
second stage as described in Chapter 4. The new design has eliminated the condition of
counter torque at the first stage, and has significantly reduced the amount of counter torque
at the second stage, i.e. at the lower region of the flow at the second stage, where the inlet
flow angle is greater than the inner blade angle 90◦ . The results show that the new design
yields appropriate inlet flow angles at both stages, so that separation has been reduced.
190
These results confirm that the new nozzle design, the number of blades, and the outer blade
angle give rise to efficiency increase up to 91%. Conversion of the head to kinetic energy and
matching the inlet flow condition with the impeller appeared to be the contributing factors
for the improved efficiency of 91%. Azimuthal variation in the flow angle was not found to
be a critical issue at least to obtain 91% efficiency because 30 - 35 blades at β1b = 39◦ and β2b
= 90◦ limit the potential for flow separation and that the second stage picks up any energy
transfer missed in the first stage. Thus the objective of achieving efficiencies above 90% was
met by massively improving the efficiency of 7 kW turbine.
6.5 Second-order Effects
The achievement of redesigning both the 0.53 kW and 7 kW turbines to achieve efficiencies
over 90% indicates that the essential requirements for improving the efficiency have been
determined. From the practical design viewpoint, the key element in the design is to develop
a nozzle design that converts all the head into kinetic energy and create a suitable inlet
flow angle to the impeller. Matching the nozzle and impeller designs, particularly the inlet
flow angle with the outer blade angle, is therefore essential in the detailed design phase for
improving the performance. Outer blade angle cannot be arbitrarily assigned as is done in
practice. The viscous losses were not analyzed as these may have only second order effects
on turbine performance. In addition, the radial clearance between the impeller blade-tip
and the nozzle-casing wall is practically unavoidable, which allows flow leakage. Moreover,
it can be a source of losses due to the dynamic interaction of the rotating blade-tip and
boundary layer on the casing wall. The effects of tip-clearance on turbine performance was
not investigated in this thesis, but it is certain that the effect is second-order. The effect
of including and not including the shaft was not investigated in detail although shaft was
included in the final design of the 7 kW turbine. As the flow is not crossing the shaft, the
191
effect should be only small. The effect of non-uniformity in the inlet flow was not assessed.
Similarly, the effect of impeller discs used to hold the impeller blades was not investigated.
6.6 Synthesis of Results on Design Improvement
A design procedure has been followed which permits the design of crossflow turbines that can
achieve efficiencies more than 90%. Two design concepts were found important: 1) conversion
of head into kinetic energy and direct the flow at a suitable angle to the impeller, which are
important considerations in the nozzle design and 2) outer blade angle, number of blades
and diameter ratio are important for the impeller design. The analytical model for nozzle
design presented in Chapter 5 is very useful in the detailed computational analysis, which
meets the goal of approximately converting all the available head into kinetic energy at the
impeller inlet, and matching the two components for the maximum efficiency. It was shown
that nozzle design has significant influence on the impeller performance and demonstrated
that the simple 2-D analytical model for nozzle design can provide a good nozzle design.
Sensitivity analysis was used to provide quantitative and qualitative descriptions of the flow
in the impeller. The low-order model for nozzle-impeller design presented in this thesis when
combined with high-fidelity simulations will allow optimization of the turbine designs, that
could achieve more than 90% efficiency.
6.7 Chapter Summary
In this chapter, sensitivity analysis was performed on the 0.53 kW turbine with the aim of
identifying key design parameters. The influence of nozzle design and the impeller param-
eters was investigated in detail using the design method developed in Chapter 5. Then the
results of sensitivity analysis was first applied to improve the efficiency of 0.53 kW turbine.
This analysis allowed the identification of the major changes in the impeller flow field and
performance. Finally, the example design was presented for the 7 kW turbine. The efficiency
192
of the 0.53 kW turbine was improved from 88% to 90%, whereas the efficiency of the 7 kW
turbine was improved from 69% to 91%.
193
Chapter 7
Conclusions
The objective of this thesis was to study the major aspects of the flow and turbine geometry
that affect the efficiency of crossflow turbines and improve the design of small-scale crossflow
turbines for maximum efficiency compared to the existing turbine design of highest maximum
efficiency reported in the literature. Steady and unsteady RANS computations were used to
capture the key flow features governing the turbine performance to enable design improve-
ment. The computational model is described in Chapter 3 and its validation is presented
in Chapter 4. An integrated design framework for the nozzle and the impeller is presented
in Chapter 5. Chapter 6 presented the detailed computational analysis on design improve-
ment. The findings of the computational results of design improvement are summarized as
conclusions and future works as follows.
7.1 Conclusions
For the purpose of improving the efficiency, two small-scale crossflow turbines reported in the
literature were considered. The first turbine is a 7 kW low-efficiency turbine with a maximum
efficiency of 69%, which was experimentally studied by Dakers and Martin (1982). However,
the details of its internal flow features and the reasons for its inefficiency are not known.
This reference provides a detailed information of the turbine geometry and test results. The
194
second turbine is a 0.53 kW high-efficiency turbine with a maximum efficiency of 88%, which
was experimentally studied by Desai (1994). A further improvement in maximum efficiency
of the same turbine was obtained by Totapally and Aziz (1994), who achieved a maximum
efficiency of 90% by increasing the number of blades from 30 to 35. Thus far, this is the
most efficient turbine design reported in the literature. The reference Desai (1994) provides a
detailed information about the turbine geometry and the test results for a number of varying
operating conditions. Furthermore, the internal flow features of these turbines are not known
in any detail. A better understanding of the key flow features and the reasons for high
efficiency or low efficiency provides the essential features of the design requirements for high
efficiency turbines and avenues for further improvement. Moreover, these two turbine models
allowed validation and performance characterization and identify the key design problem in
the design of crossflow turbines. These are the primary motivations for investigating these
particular turbines to achieve the goal of design improvement of crossflow turbines.
Important performance loss mechanisms were identified in both turbines. Based on the
insights gained from the key flow features of these turbines, an integrated design for designing
the nozzle and impeller was developed. The fundamental idea underlying the design principle
is that nozzle performance influences the impeller performance, and thus matching the best
nozzle and impeller designs is critically important. Two design concepts: conversion of head
into kinetic energy with a suitable flow angle for the nozzle design and matching the inlet flow
with the impeller blade angle were applied for improving the overall turbine performance.
The main findings are summarized below.
1. The computed results of the 7 kW turbine (69% efficiency) revealed that partial con-
version of head into kinetic energy and considerable flow separation due to mismatching of
the inlet flow angle with the outer blade angle severely limited the impeller performance,
resulting from the poor first stage performance. A remarkable finding is that the impeller
blades can lose power from both stages when the inlet flow angle and blade angle are mis-
matched, which limits the turbine performance before reaching the optimal operating speed.
195
It revealed that the enabling idea in the design improvement of such designs is to design a
nozzle that converts all the head into kinetic energy, maximizes the angular momentum flux
at the impeller inlet, and directs the flow to the impeller at a suitable angle to match the
impeller.
2. Computed results of the 0.53 kW turbine (88% efficiency) showed that azimuthal
non-uniformity of the inlet flow is not a limiting factor for the turbine efficiency, provided
that all the head is approximately converted into kinetic energy, the inlet flow angle matches
approximately the outer blade angle over a large extent of the impeller inlet, and there is
negligible separation on the blades. This revealed that the important criteria for achieving
high efficiency are: conversion of all the head into kinetic energy, matching of the inlet flow
angle with the outer blade angle, and minimizing flow separation in the impeller. This is an
important finding to enable design optimization.
3. A simple analytical model for the nozzle design was developed and applied to assess
the nozzle-impeller performance of the 0.53 kW turbine. The essential design criteria of
this model were 1) to fully convert of head into kinetic energy and 2) to provide a suitable
flow angle to the impeller. A simple tangential nozzle with horizontal inlet provided a
uniform inlet flow to the impeller. With such nozzle, the first stage performance improved
remarkably, and the improvement in efficiency from 88% to 89.21% was achieved compared
to the original nozzle design. Similarly, the nozzle with arbitrary orientation with vertical
inlet section showed better performance. In both nozzles, only the orientation and rear
wall shape were different compared to the original nozzle. With the new nozzle design with
vertical orientation, efficiency improved from 88% to 89.45%.
4. The results of a sensitivity analysis on the 0.53 kW turbine suggested that the impeller
behaviour in terms of power production is strongly dependent on the inlet flow conditions.
Two new nozzle designs were tested for the original impeller design to determine the sensi-
tivity of the inlet flow conditions. It was shown that conversion of head into kinetic energy
and uniformity in the inlet flow and matching of the outer blade angle with the inlet flow
196
angle are crucial in improving the turbine efficiency. The results of the new nozzle designs
demonstrated that the simple analytical model for designing nozzles is remarkably accurate.
The number of blades was found to be another significant parameter in improving the im-
peller performance amongst others. Highest efficiency gain was achieved by adjusting the
number of blades to 35 and improving the nozzle design. The number of blades improved
the first stage performance, which is an important consideration in improving the efficiency.
Similarly, inlet blade angle was found strongly sensitive to the inlet flow conditions. An
optimum blade angle was found to improve the first stage performance amongst the cases.
Inner to outer diameter ratio was found to be strongly related to stage performance. A
higher diameter ratio means longer blades and hence more extraction from the first stage,
provided that there is no major flow separation.
5. Following the results from sensitivity analysis on the 0.53 kW turbine, two example
turbine designs of improved efficiency were presented. For this, new nozzle was designed with
the aim of converting head into kinetic energy and matching the inlet flow to the impeller,
impeller was modified to match with the inlet flow, and number of blades were increased.
The improved design of the 0.53 kW turbine achieved more than 90% efficiency, by following
the simple design principle developed in Chapter 5. Similarly the improved design of the 7
kW turbine achieved the maximum efficiency of 91%. A systematic design procedure was
developed and followed to achieve these efficiency gains.
6. At part-load operation without any flow-control mechanism, the efficiency of the
turbine was found significantly low. To address this issue, slider control mechanism was
considered for regulating the inlet flow. At part-load operations, three different part-loads,
i.e. varying Q at constant H, were considered in the case of 0.53 kW turbine. The computed
results showed that the efficiency of the impeller didn’t deteriorate significantly. At 43% part-
load, the efficiency dropped only marginally from 88% to 86%. More importantly, maximum
efficiency occurred at the same speeds where the design flow rate attains maximum efficiency,
which is very important in the operation of hydro turbines. These results demonstrated that
197
slider control mechanism is a promising flow control mechanism for crossflow turbines. A
more remarkable feature of this design is that they are simple to design and actuate, which
allows more precise mechanical flow control.
7. Cavitation inception was observed at and above the maximum efficiency point of
the 7 kW turbine. However, no cavitation was observed in the case of 0.53 kW turbine.
The full development of cavitation bubbles and their effects on turbine efficiency were not
investigated.
Based on the results presented in this thesis, it was demonstrated that the crossflow
turbines can achieve maximum efficiency of more than 90% if the design is prescribed using
the following criteria, followed by high-fidelity computational simulations.
• As nozzle is the most important component limiting the impeller performance, more
attention must be paid on its design and matching with the impeller. Particular
emphasis must be put on conversion of head into kinetic energy, and directing the
flow at a suitable inlet flow angle to the impeller to match the outer blade angle. A
simple design follows a tangential entry nozzle, but nozzles with arbitrary orientation
can be designed that provide similar or better efficiency than the tangential entry
nozzles. The analytical model for the nozzle design presented in Chapter 5 provides a
rigorous basis for designing an optimal nozzle and the methods for matching with the
impeller.
• Impeller must be designed to match the inlet flow angle, which can be calculated using
the analytical model for nozzle design (based on the available head H and the flow rate
Q). An analytical model for computing the nozzle dimensions, the criteria for optimal
speed and efficiency, and the methods to calculate flow angle is provided in Chapter 5.
Next is to select a proper number of blades and other geometrical parameters, such as
diameter ratio. A diameter ratio in the range of 0.64 - 0.68, which is recommended by
previous researchers, provides good empirical values for searching optimal values with
high-fidelity simulations. Then a number of design parameters may be varied to match
198
the nozzle and impeller designs with the help of high-fidelity computations, such as
RANS, URANS or LES, until a desired efficiency is achieved. This study showed that
at least 90% efficiency can be achieved.
7.2 Future Work
Looking to the future, there are opportunities for improving the maximum efficiency of
crossflow turbines, as well as maintaining high part-load efficiencies. The results presented
in this thesis provide a basis for an extension to form a more complete design theory for
improving the design. Only three of these are mentioned here in the context of understanding
the key features of the flow, developing the design methods from it, and understanding the
impact of cavitation. The overwhelming need is to build a turbine to the new methodology
and then test it as far as possible to establish performance relative to CFD and suggest
further ways of improving performance.
1. Flow Physics: Any analysis which will lead to a better understanding of the flow
mechanisms influencing the power extraction in the stages would be a worthy contribution.
The current understanding of the underlying flow behaviour in crossflow turbines can be
conceivably enhanced by studying: 1) effects of azimuthal flow non-uniformity at the impeller
inlet, particularly the matching of flow angle and outer blade angle 2) effects of blade-tip
clearance or nozzle leakage flow and the methods to minimize them, 3) flow angles at the
exit of the first stage and inlet of the second stage, 4) assessment of flow separation and its
location to improve individual blade performance, and 5) loss mechanisms as a method to
quantify losses and their reduction.
2. Design improvement: To confirm the findings reported in this thesis and achieve effi-
ciency more than 90%, experimental testing of the improved designs, both the flow field and
performance measurements, are recommended as the first step before undertaking paramet-
ric studies. The design option for multiple nozzles can also be investigated with the aim of
199
balancing power productions between the two stages and improving efficiency. For this, two-
nozzles may be designed. In addition, this would improve the turbine’s structural loading,
which may improve the turbine life. This design aspect may be important at low-head, high-
flow rate designs. As the slider control mechanism shows promising results on maintaining
high efficiency at part-load operations, it must be built and tested to assess its performance
at part-load operations and other operational benefits over commonly used guide vanes.
3. Cavitation: Cavitation inception studied in this thesis provides a basis for investigating
a detailed development and influence of cavitation bubbles on turbine efficiency as well as
blade erosion. A three-phase multiphase cavitation model with URANS, DES or LES is
recommended for a detailed analysis. Experimental observations of cavitation will be of
great value in the validation of computational simulations. As the flow field contains water
and air with free-surface effects, it might be possible that cavitation bubbles are strongly
influenced by the air content in the water stream. The air volume fraction may be significant
in the water stream in the second row of blades as the flow velocity is reduced at the first
stage and the turbine operates at atmospheric condition. It would be worthy to investigate
the erosion rate of such cavitation. Thus a detailed analysis pertaining to the size and
strength of cavitation bubbles and their influence on performance loss are recommended if a
complete practical explanation or design rules for the design and operation of cavitation-free
crossflow turbines need to be developed.
200
Appendix A
Analytical Equations for Efficiency
It is relevant to discuss here that the important previous models which simplified the complex,
three-dimensional, turbulent multiphase flow with free-surface effects in crossflow turbines
to develop simple analytical models for the turbine efficiency. Such models assume a steady,
inviscid, fully attached and two-dimensional flow. Banki (1918) formulated the following
equation to compute the efficiency η as (Desai, 1994):
(cosα1 − U1 /u1 )
η = 4U1 (A.1)
u1
where U1 = R1 ω is the peripheral velocity of the impeller, α1 = tan−1 (ur /uθ ) is the angle of
attack and u1 = total velocity at the impeller inlet. By simplifying Equation (A.1), Banki
included the nozzle tip and blade frictional losses to compute the maximum efficiency as:
[cv (1 + ψ)]2 cos2 α1

ηmax = (A.2)
2
where cv is the nozzle velocity coefficient, whose empirical values are in the 0.95 - 0.98, i.e.
less than unity because of friction and Ψ is the frictional loss on impeller blades, whose
201
empirical value is around 0.98. Macmore and Merryfield (1949) extended Banki’s equation
for efficiency as follows:
U1 cosβ2 U1
η = 2C 2 (1 + Ψ )(cosα1 − ) (A.3)
u1 cosβ1 u1
where C and Ψ are the empirical constants to account for the losses in the nozzle and the
impeller blade surfaces and β1 and β2 are the flow angles at inlet and exit of the blades
at the first stage. Although Equation (A.3) includes flow angles β1 and β2 , the predictive
capability is still very limited similar to Equation (A.1). The predictive capability of the
Equations (A.1) and (A.3) are compared with the experimental results in Figure A.1 for
the 7 kW turbine studied in this thesis. It is noted that both equations are equivalent
for C= 1 and β2 = 90◦ . For the 7 kW turbine, β2 = 90◦ . The Figure clearly shows that
the these basic analytical equations are inaccurate. Equally important is that the shapes
of the curves are different and so the theoretical results could not be made to match by
adjusting the constants. In addition, these models do not consider the actual upstream flow
conditions and the influence of different geometric parameters. Thus the predicted maximum
efficiency is far from the observed results. A large number of fluid dynamic and geometric
variables that influence the turbine efficiency create a considerable degree of complexity for
the analytical models. It is thus reasonable to conclude that only numerical computations
can approximately compute the flow field and predict the turbine performance, which is
the motivation for using RANS simulations in tackling this technical problem of efficiency
improvement.
202
Experiment
90 Analytical
85
80
Efficiency (%)
75
70
65
60
55
50
0.35 0.4 0.45 0.5 0.55 0.6 0.65 0.7 0.75 0.8
Impeller speed ratio
Figure A.1: Comparison between the analytical model [Equations (A.2) and (A.3)] and the
experimental results for the efficiency η vs speed ratio U1 /u1 at α1 = 16◦ [Q = 105 lps and
H = 10 m]. The experimental data are plotted for the 7 kW turbine (Dakers and Martin,
1982).
203
Appendix B
Generalized Equation for Nozzle

Design
The derivation of the analytical equation for the nozzle design of arbitrary orientation, as
illustrated in Figure 5.4, is presented here. The derivation follows the same assumptions and
principle used for the tangential entry nozzle. Compared to the tangential nozzle, in this
nozzle configuration, the left nozzle lip is oriented at an angle θ0 with the horizontal axis
as shown in the Figure. The nozzle throat h0 is parallel to the vertical axis. To design the
nozzle, it is necessary to determine the rear-wall shape represented by C-D. A small radial
element O-C-E is taken for the analysis, and the continuity in the radial and tangential
directions are applied. So the total flow passing through O-C-D is obtained by subtracting
the flow through the radial section O-B-F.
Referring to the Figure 5.4, it is assumed that O-C = R(θ0 + γ), F-C = h(θ0 + γ), and
θs ’ = θs - γ. It follows that
p
R(θ0 + γ) = (R1 sin θ0 + h0 )2 + (R1 cos θ0 )2 (B.1)
So
p
h(θ0 + γ) = (R1 sin θ0 + h0 )2 + (R1 cos θ0 )2 − R1 (B.2)
204
Similarly,
h0 + R1 sin θ0
γ = tan−1 ( ) − θ0 (B.3)
R1 cos θ0
Now it follows from the tangential nozzle analysis that
h(θ0 + γ + θ) = h(θ0 + γ)(1 − θ/θs0 ) (B.4)
By substituting θs0 = θs - γ, Equation (B.4) reduces to
p θ
h(θ0 + γ + θ) = ( (R1 sin θ0 + h0 )2 + (R1 cos θ0 )2 − R1 )(1 − ) (B.5)
θs − γ
It is noted that Equation (B.5) reduces to the equation of tangential nozzle for θ0 = 90◦ and
γ = 0◦ .
205
Appendix C
Error and Uncertainty
This appendix presents the calculations of the error and uncertainty in the computational
simulations presented in Chapter 4. The method used here is based on the relative error
estimation by comparing the computed and experimental results. Since the validation of
the computational analysis is built around comparison of global parameter, i.e. power out-
put or efficiency, to determine the predictive capability of RANS/URANS computations to
improve the turbine design, systematic errors due to computational modeling are assumed
to be negligible. The American Institute of Aeronautics and Astronautics (AIAA) defines
uncertainty as “a potential deficiency in any phase or activity of the modeling process that
is due to lack of knowledge” (Oberkampf and Trucano, 2000). It is noted that the deficiency
may or may not exist in the computed results. In computations, numerical uncertainty refers
to variations in the solution, in this case the power output or the efficiency, once the solu-
tion is considered converged as these numerical errors give the uncertainty in the predicted
turbine efficiency. Thus, all error is associated with model’s predictive capability relative to
the experimental measurements, and the uncertainty is associated with the numerics aris-
ing mainly from mesh resolution, turbulence model, multiphase flow model, inlet boundary
condition, and the interface modeling (general grid interface) between the rotating and the
stationary domains. Uncertainties due to turbulence and multiphase models and lack of pre-
206
7.5
7.4
7.3
Power output [kW]

7.2
7.1
6.9
6.8
6.7
6.6
6.5
0 1 2 3 4 5 6
No of mesh elements [million]
Figure C.1: Results of grid convergence test of the 7 kW turbine at maximum efficiency [Q
= 105 lps, H = 10 m and N = 450 RPM].
cise inlet boundary condition data were not assessed in detail, but their cumulative effects
may be only minor. This is demonstrated by a good agreement between the computed and
experimental results for the power outputs of the turbines presented in Chapter 4.
Numerical uncertainty in quantity u was estimated by:
|ur1 − ur2 |
≤ (C.1)
ur2
In Equation (C.1), r1 is one level of grid refinement and r2 is a more refined grid and is the
accuracy required for the solution to be considered satisfactory. The refined grid was con-
sidered converged when was less than 0.1% in the power output. The total computational
error was estimated by computing the maximum relative percentage difference between the
computed results and the experimental results. The results of mesh convergence for the two
turbines studied in this thesis are shown in Figures C.1 and C.2.
207
540
530
520
Power output [Watts]
510
500
490
480
470
460
450
2 4 6 8 10 12 14 16
No of mesh elements [million]
Figure C.2: Results of grid convergence test of the 0.53 kW turbine at maximum efficiency
[Q = 46 lps, H = 1.337 m and N = 199.1 RPM].
208
Bibliography
Acharya, N., Kim, C.-G., Thapa, B., and Lee, Y.-H. (2015). Numerical analysis and perfor-
mance enhancement of a cross-flow hydro turbine. Renewable energy, 80:819–826.
Adhikari, R. C., Vaz, J., and Wood, D. (2016). Cavitation inception in crossflow hydro
turbines. Energies, 9(4):237.
Ali, Z. and Tucker, P. G. (2014). Multiblock structured mesh generation for turbomachin-
ery flows. In Proceedings of the 22nd International Meshing Roundtable, pages 165–182.
Springer.
ANSYS (2016). Academic research.
Baskharone, E. A. (2006). Principles of turbomachinery in air-breathing engines. Cambridge
University Press.
Batchelor, G. K. (2000). An introduction to fluid dynamics. Cambridge university press.
Benzon, D., Aggidis, G. A., and Anagnostopoulos, J. (2016). Development of the turgo
impulse turbine: Past and present. Applied Energy, 166:1–18.
Blazek, J. (2015). Computational fluid dynamics: principles and applications. Butterworth-
Heinemann.
Bourgeois, J. A., Martinuzzi, R. J., Savory, E., Zhang, C., and Roberts, D. A. (2011).
Assessment of turbulence model predictions for an aero-engine centrifugal compressor.
Journal of Turbomachinery, 133(1):011025.
Brennen, C. E. (2005). Fundamentals of multiphase flow. Cambridge University Press.
Choi, Y.-D., Lim, J.-I., Kim, Y.-T., and Lee, Y.-H. (2008). Performance and internal flow
characteristics of a cross-flow hydro turbine by the shapes of nozzle and runner blade.
Journal of Fluid Science and Technology, 3(3):398–409.
Cornelius, C., Biesinger, T., Galpin, P., and Braune, A. (2014). Experimental and compu-
tational analysis of a multistage axial compressor including stall prediction by steady and
transient cfd methods. Journal of Turbomachinery, 136(6):061013.
Dakers, A. and Martin, G. (1982). Development of a simple cross-flow water turbine for
rural use. In Agricultural Engineering Conference 1982: Resources, Efficient Use and
Conservation; Preprints of Papers, page 35. Institution of Engineers, Australia.
209
De Andrade, J., Curiel, C., Kenyery, F., Aguillón, O., Vásquez, A., and Asuaje, M. (2011).
Numerical investigation of the internal flow in a banki turbine. International Journal of
Rotating Machinery, 2011.
Denton, J. and Dawes, W. (1998). Computational fluid dynamics for turbomachinery design.
Proceedings of the Institution of Mechanical Engineers, Part C: Journal of Mechanical
Engineering Science, 213(2):107–124.
Denton, J. D. (2010). Some limitations of turbomachinery cfd. In ASME Turbo Expo 2010:
Power for Land, Sea, and Air, pages 735–745. American Society of Mechanical Engineers.
Desai, V. R., . (1994). A parametric study of the Cross-Flow Turbine performance. PhD
thesis.
Dixon, S. L. and Hall, C. (2013). Fluid mechanics and thermodynamics of turbomachinery.

Butterworth-Heinemann.
Dubbioso, G., Muscari, R., and Di Mascio, A. (2014). Analysis of a marine propeller oper-
ating in oblique flow. part 2: Very high incidence angles. Computers & Fluids, 92:56–81.
Durali, M. (1976). Design of small water turbines for farms and small communities.
Durbin, P. A. and Reif, B. A. P. (2001). Statistical theory and modeling for turbulent flows.
Wiley, Chichester; New York, 2 edition.
Durgin, W. and Fay, W. (1984). Some fluid flow characteristics of a cross-flow type hydraulic
turbine. Small Hydro Power Fluid Machinery, pages p77–83.
Elbatran, A., Yaakob, O., Ahmed, Y. M., and Shabara, H. (2015). Operation, performance
and economic analysis of low head micro-hydropower turbines for rural and remote areas:
a review. Renewable and Sustainable Energy Reviews, 43:40–50.
Escaler, X., Egusquiza, E., Farhat, M., Avellan, F., and Coussirat, M. (2006). Detection
of cavitation in hydraulic turbines. Mechanical systems and signal processing, 20(4):983–
1007.
Fiuzat, A. and Akerkar, B. (1989). The use of interior guide tube in cross flow turbines. In
Waterpower’89, pages 1111–1119. ASCE.
Fiuzat, A. A. and Akerkar, B. P. (1991). Power outputs of two stages of cross-flow turbine.
Journal of energy engineering, 117(2):57–70.
Greitzer, E. M., Tan, C. S., and Graf, M. B. (2007). Internal flow: concepts and applications,
volume 3. Cambridge University Press.
Hothersall, R. (1985). A review of the cross-flow turbine. Waterpower ’85, ASCE, New
York, N.Y.
IEA (2013). Energy access database. accessed July 07, 2016, from
http://www.worldenergyoutlook.org/resources/energydevelopment/energyaccessdatabase/.
210
Johnson, W., E. R. and White, F. (1982). Design and testing of an inexpensive crossflow
turbine. Small Hydropower Fluid Machinery, ASME, New York, N.Y.
Kalitzin, G., Medic, G., Iaccarino, G., and Durbin, P. (2005). Near-wall behavior of rans
turbulence models and implications for wall functions. Journal of Computational Physics,
204(1):265–291.
Khosrowpanah, S. (1984). Experimental study of the crossflow turbine.
Khosrowpanah, S., Fiuzat, A., and Albertson, M. L. (1988). Experimental study of crossflow
turbine. Journal of Hydraulic Engineering, 114(3):299–314.
Lakshminarayana, B. (1996). Fluid Dynamics and Heat Transfer of Turbomachinery. Wiley

Online Library.
Larsson, J. and Wang, Q. (2014). The prospect of using large eddy and detached eddy
simulations in engineering design, and the research required to get there. Philosophical
Transactions of the Royal Society of London A: Mathematical, Physical and Engineering
Sciences, 372(2022):20130329.
Logan Jr, E. (2003). Handbook of turbomachinery. CRC Press.
Macmore, C. and Merryfield, F. (1949). The banki water turbine. Engineering Experiment
Station, 25:3–25.
Menter, F. R. (1994). Two-equation eddy-viscosity turbulence models for engineering appli-

cations. AIAA journal, 32(8):1598–1605.
Michell, A. (1904). Impulse-turbine.
Montomoli, F., Hodson, H., and Lapworth, L. (2011). Rans–urans in axial compressor,
a design methodology. Proceedings of the Institution of Mechanical Engineers, Part A:
Journal of Power and Energy, 225(3):363–374.
Nakase, Y., Fukutomi, J., Watanabe, T., Suetsugu, T., Kubota, T., and Kushimoto, S.
(1982). A study of cross-flow turbine (effects of nozzle shape on its performance). In
Proceedings of the Winter Annual Meeting ASME, Phoenix, AZ, USA, volume 1419.
Nichols, R. H. (2008). Turbulence models and their application to complex flows. University
of Alabama, Birmingham.
Oberkampf, W. L. and Trucano, T. G. (2000). Validation methodology in computational

fluid dynamics. AIAA paper, 2549:19–22.
Opgenoord, M. M., Allaire, D. L., and Willcox, K. E. (2016). Variance-based sensitivity

analysis to support simulation-based design under uncertainty. Journal of Mechanical
Design.
Ott, R. F. and Chappell, J. R. (1989). Design and efficiency testing of a cross-flow turbine.
In Waterpower’89, pages 1534–1543. ASCE.
211
Roache, P. J. (1997). Quantification of uncertainty in computational fluid dynamics. Annual
Review of Fluid Mechanics, 29(1):123–160.
Sammartano, V., Aricò, C., Carravetta, A., Fecarotta, O., and Tucciarelli, T. (2013). Banki-
michell optimal design by computational fluid dynamics testing and hydrodynamic anal-
ysis. Energies, 6(5):2362–2385.
Sammartano, V., Morreale, G., Sinagra, M., and Tucciarelli, T. (2016). Numerical and
experimental investigation of a cross-flow water turbine. Journal of Hydraulic Research,
pages 1–11.
Shepherd, D. G. (1956). Principles of turbomachinery. Macmillan.
Sinagra, M., Sammartano, V., Aricò, C., Collura, A., and Tucciarelli, T. (2014). Cross-flow
turbine design for variable operating conditions. Procedia Engineering, 70:1539–1548.
Stern, F., Wilson, R. V., Coleman, H. W., Paterson, E. G., et al. (1999). Verification and
validation of CFD simulations. Iowa Institute of Hydraulic Research, University of Iowa.
Totapally, H. G. and Aziz, N. M. (1994). Refinement of cross-flow turbine design parameters.

Journal of energy engineering, 120(3):133–147.
Tucker, P. (2011). Computation of unsteady turbomachinery flows: Part 1—progress and

challenges. Progress in Aerospace Sciences, 47(7):522–545.
Tucker, P. (2013). Trends in turbomachinery turbulence treatments. Progress in Aerospace

Sciences, 63:1–32.
UNDP (2015). Sustainable Development Goals (SDGs). accessed August 07, 2016, from
http://www.undp.org/content/undp/en/home/sdgoverview/post-2015-development-
agenda.html.
Varga, J. (1959). Tests with the banki water turbine. Acta Technica Academicae Hungaricae,
XXVI(1):79–102.
Weide, E., K. G. S. J. . (2004). Large-scale turbomachinery computations. Annual Research

Briefs, Center for Turbulence Research , Stanford University, USA.
White, F. M. and Corfield, I. (2006). Viscous fluid flow, volume 3. McGraw-Hill New York.
Wilcox, D. C. (2006). Turbulence Modeling for CFD. DCW Industries, Incorporated.
Yang, Y.-L. (1991). A design study of radial inflow turbines in three-dimensional flow. PhD
thesis, Massachusetts Institute of Technology.
212

2 PhDThesis Design Improvement of Crossflow Hydro Turbine

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

2 PhDThesis Design Improvement of Crossflow Hydro Turbine

Uploaded by

Copyright:

Available Formats

University of Calgary

PRISM: University of Calgary's Digital Repository

Graduate Studies The Vault: Electronic Theses and Dissertations

Design Improvement of Crossflow Hydro Turbine

Adhikari, R. (2016). Design Improvement of Crossflow Hydro Turbine (Unpublished doctoral

Design Improvement of Crossflow Hydro Turbine

Ram Chandra Adhikari

GRADUATE PROGRAM IN MECHANICAL AND MANUFACTURING ENGINEERING

List of Abbreviations, Symbols, and Nomenclature xxiv

4 Validation and Performance Characterization 49

6 Design Improvement 124

B Generalized Equation for Nozzle Design 204

C Error and Uncertainty 206

4.1 Design parameters of 7 kW turbine . . . . . . . . . . . . . . . . . . . . . . . 57

2.1 Schematic illustration of the basic design features of a crossflow turbine. . . . 9

4.1 Steady flow through the control volume of the impeller. . . . . . . . . . . . . 52

C.1 Results of grid convergence test of the 7 kW turbine at maximum efficiency

D rate of diffusion of turbulence kinetic energy

ε rate of dissipation of turbulence kinetic energy

2.1 Motivation and Background

2.2 Problem Statement

A longstanding problem in the design of crossflow turbines is the improvement in efficiency. A

2.3 Research Questions and Goals

2.4 Flow Field Characteristics in Crossflow Turbines

2.5 Literature Review

2.5.1 Experimental Studies

2.5.2 Numerical Studies

Flow First Stage

2.7 Organization of the Thesis

The remainder of the thesis is set out as follows:

An incompressible isothermal flow through a turbomachine is fully described by the conti-

where U = time-averaged mean velocity, ui 0 = fluctuating velocity due to turbulence and

3.2 Multiphase Flow

where µef f = effective viscosity accounting for turbulence given by µef f = µ + µt , µ =

3.3.1 Computational Mesh

complex geometries, such as in a crossflow turbine. Recently unstructured hybrid meshes

The transformation between models is accomplished by the blending function, F1 , applied

where CDkω is written as:

Inlet boundary condition

Outlet boundary condition

Wall boundary conditions

3.5 CFD Model Validation

3.7 Chapter Summary

Validation and Performance

4.1 Turbine Efficiency and Performance Metrics

The performance of a turbine is measured in terms of maximum efficiency over a range of

(a) power extraction across (b) power loss across

where ur = radial velocity, uθ = tangential velocity, ω = angular velocity of the impeller,

4.2 Results of Validation

4.2.1 Computational Details and Results

D2=211.72 mm O = Impeller centre

Table 4.1: Design parameters of 7 kW turbine

Design parameter Value

Operating conditions Value

Design parameter Value

Table 4.4: Operating conditions of 0.53 kW turbine

Operating conditions Value

Mesh No of elements Power output Numerical

Mesh No of elements Power output Numerical

4.2.2 Unsteady RANS Computations

of this non-uniformity on the impeller performance and 2) influence of impeller motion on

4.3 Key Flow Features and Performance Characteris-

left nozzle lip ﬂow

right nozzle lip