Professional Documents
Culture Documents
Sailesh Chitrakar
July, 2013
Abstract
Sediment erosion is one of the key challenges in hydraulic turbines from a design and maintenance
perspective in Himalayas and Andes. Past research works have shown that the optimization of the
Francis turbine runner blade shapes can decrease erosion by a significant amount. This study conducted as a Masters Thesis has taken the proposed designs from past works and conducted a CFD
analysis on a single passage of a Francis runner blade to choose an optimized design in terms of erosion
and efficiency. Structural analyses have been performed on the selected design through one-way and
two-way FSI to compare the structural integrity of the designs.
Two types of cases have been considered in this thesis work to define the boundary condition of the
structural model. In the first case, a runner blade is considered to have no influence of the joint and
other stiffer components. In the second case, a sector of the whole runner has been modeled with
necessary boundary conditions. Both one-way and two-way FSI have been performed on the cases
for the designs. Mesh independent studies have been performed for the designs, but only for the first
case, whereas in the second case, a fine mesh has been used to make the analysis appropriate.
The loads have been imported into the structural domain from the fluid on the interfaces for one-way
FSI. In the case of two-way FSI, the Multi-Field Solver (MFX) supported by ANSYS has been used
to solve the coupled field analysis. A fully coupled FSI in ANSYS works by writing an input file in
the structural solver containing the information about the interfaces in the structural domain, which
is imported in the fluid solver. The interaction between the two domains is defined in ANSYS-CFX,
including the mesh deformation and solver setups. The results have been post-processed in CFX-Post,
where the results from both the fields are included. It has been found that the structural integrity of
the optimized design is better than the reference design in terms of the maximum stress induced in
the runner. The two-way FSI analysis has been found as an inevitable part of the numerical analysis.
However, with the advancement of the computational capability in the future, there could be a great
scope in the research field to carry out a fully-coupled transient simulation for the whole runner to
get a more accurate solution.
Keywords: Sediment erosion, one-way FSI, two-way FSI, Francis turbine
Acknowledgements
I would like to express my gratitude to Professor Michel Cervantes for the continuous support, supervision, useful comments, remarks and engagement throughout this master thesis. Furthermore I
would like to thank Mr. Biraj Singh Thapa for providing me with all the necessary inputs in this
thesis work. His constant support, encouragement and belief towards me and my work made me do
the work effectively and punctually. I would also like to thank my program coordinator and lecturer,
Professor Damian Vogt for accepting my proposal of doing the thesis in Nepal.
Also, I would like to express my appreciation to all the members of the Turbine Testing Lab, whose
continuous care and support made my stay a pleasant one. I would also like to thank Professors Bhola
Thapa and Hari Prasad Neopane for their continuous motivation during the project.
Finally, I would like to thank my parents who motivated me and helped me complete my KTM works
while I was away in KU.
Contents
List of abbreviations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
List of symbols . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
10
1 Introduction
15
1.1
15
1.2
16
1.3
16
1.4
Study methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
17
1.5
Scope of study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
17
1.6
17
2 Hydro Turbines
18
2.1
Hydropower in Nepal . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
18
2.2
19
2.3
Cavitation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
20
2.4
21
2.4.1
Pelton turbines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
22
2.4.2
Kaplan turbines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
22
2.4.3
Francis turbines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
23
2.4.4
24
2.4.5
25
3 Sediment Erosion
27
3.1
28
3.2
29
3.3
30
3.4
Erosion models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
31
3.4.1
32
34
4.1
CFD works . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
34
4.2
FSI works . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
37
4.3
37
5 FSI review
5.1
5.2
40
Coupled-Field Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
40
5.1.1
41
5.1.2
41
42
5.2.1
42
5.2.2
43
5.3
5.2.3
43
5.2.4
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
43
43
6 CFD analysis
45
6.1
Sensitivity study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
46
6.2
46
6.3
47
6.4
48
6.4.1
48
6.4.2
49
6.4.3
50
51
6.5.1
51
6.5.2
52
6.5.3
52
54
6.5
6.6
7 Structural analysis
57
7.1
Geometry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
57
7.2
Boundary condition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
60
7.2.1
Case I . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
60
7.2.2
Case II . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
60
7.3
62
7.4
63
7.5
Case-I . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
63
7.6
Case-II . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
63
8 FSI analysis
68
8.1
Mesh deformation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
68
8.2
Interface setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
69
8.3
Solver Setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
71
8.4
Post processing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
71
8.5
73
8.5.1
Case-I . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
73
8.5.2
Case-II . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
73
9 Conclusion
77
79
Bibliography
80
83
84
86
88
90
List of Figures
2.1
. . . . . . . . .
20
2.2
21
2.3
. . . . . . . . . . . . . . . . . . . .
21
2.4
22
2.5
23
2.6
24
3.1
27
3.2
28
3.3
30
3.4
Sediment erosion wear in the Francis turbine guide vane and runners in Jhimruk [19] .
31
4.1
35
4.2
CFX-pre setup file showing the blade passage and the mesh . . . . . . . . . . . . . . .
35
4.3
35
4.4
38
LIST OF FIGURES
4.5
38
4.6
39
5.1
42
5.2
44
6.1
Mesh convergence study for the factor ratio of 1.15, RMS of 1E-6 and y+ value on the
blade . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
47
6.2
48
6.3
49
6.4
The erosion pattern for the particle diameter of 0.01 mm on both Pressure and Suction
side . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
49
6.5
Average and maximum erosion rate density on the blade for various particle sizes . . .
50
6.6
. . . . . . . . . . . . . . . . . . . .
50
6.7
51
6.8
52
6.9
53
54
56
7.1
57
7.2
Comparison of the two domains with two cases (right one shows better mapping) . . .
58
7.3
59
7.4
61
7.5
61
7.6
62
7.7
64
LIST OF FIGURES
7.8
65
7.9
66
67
8.1
69
8.2
70
8.3
. . . . . . . . . . .
72
8.4
74
8.5
75
8.6
. . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . .
76
11.1 Boundary Vector at the inlet with the given flow direction . . . . . . . . . . . . . . . .
84
11.2 Boundary Vector at the inlet with the given flow direction . . . . . . . . . . . . . . . .
85
11.3 Boundary Vector at the inlet with the given flow direction . . . . . . . . . . . . . . . .
85
11.4 Result of the two flow directions, unmodified (top) and modified(below) . . . . . . . .
87
11.5 Discrepancy between the results when only the runner and the full stage is modeled .
89
12.1 Choices of imposing cyclic symmetry property to the sector of the runner . . . . . . .
91
List of Tables
2.1
19
2.2
26
3.1
33
4.1
36
List of abbreviations
KU
Kathmandu University
TTL
FEM
FSI
NORAD
NEA
N P SHa
N P SHr
SST
CFD
CSD
CMD
FEA
MFS
MFX
SST
Shear-Stress Transport
RANS
APDL
List of symbols
C1
Cx
C1
C2
U2
U1
Power [W]
Density [kg/m3 ]
10
List of symbols
FSI Parameters
Pressure [Pa]
ui
Ff (t)
Fs (t)
Km
A pseudo-structural stiffness matrix which is defined for the whole domain [N/m]
dm
unew
uold
min
disp
Cstif f
Size of the mesh or the distance from the nearest boundary [-]
11
List of symbols
Erosion Parameters
E, W, Er
Erosion Rate
N
mp
Finnie
Vp
f ()
Value of exponent
Tabakoff
0
k1 k4 , k12
Model constants
V1 V4
Reference velocity
RT , VP N
Bardal
C
Concentration of particles
Kmat
Material constant
Kenv
Environment constant
Vp
12
List of symbols
Erosion Parameters
Tsuguo
Value of exponent
k1 , k2
k3
Thapa
Km
Material factor
Khardness
Hardness factor
Kf
Flow factor
Kshape
Shape factor
Loss of material
a, b
Empirical constants
13
List of symbols
Units
m
Meter
mm
Milli meter
rad
Radian
MW
Mega Watt
Pa
Pascal
MPa
Mega Pascal
Celsius
kg
Kilogram
deg
Degree
rev
Revolution
min
Minute
sec
Second
atm
Atmosphere
mol
Mole
Newton
14
1
Introduction
15
16
CHAPTER 1. INTRODUCTION
fully-coupled analysis of the Francis runner exposed to sediment erosion will therefore, be a major
challenge in this project in terms of carrying out the simulation and validating the results. Thus, this
report will also contain some of the basic principles behind FSI and an example of conducting FSI in
ANSYS.
17
CHAPTER 1. INTRODUCTION
Introduce the FSI based simulations of the Francis runner through one-way and two-way coupling
techniques to establish the mechanical integrity of the design, for both the conventional and the
optimized designs.
Make a comparative analysis of the results between CFD, one-way FSI and two-way FSI and
identify the level of significance of FSI in the field of Francis turbines.
2
Hydro Turbines
Hydropower machineries are the machines that convert hydraulic power from the water to the mechanical power on the machine shaft. Like any other machines, these machines involves various losses
that arise partly in the machine itself and partly in the water transfer into and out of the machine
such as pipe friction losses, losses due to bends in pipes, gates, valves and losses due to abrupt and
gradual expansion and contraction of the pipes [4]. Some of the basic components of a hydropower
plant is listed below [4] :
A water diversion structure like a dam or a weir creating a gross head of water.
A penstock, which intakes the water from the dam and transports it to the turbines. Screening
is done in the intake, to prevent unwanted objects (debris and aquatic animals) entering into
the turbine.
Turbines and governing system.
Electrical generators, electrical control and switching equipment, equipment housing, transformers and electricity transmission lines.
Some of the other complementary components are the penstock gates, surge tank and a tail race
if the turbine exhaust water cannot be discharged directly (through the draft tubes) into the
river. Draft tubes are used to utilize the kinetic energy of the water leaving the turbine and
allows the turbine to be installed above the tailwater level without decreasing the available head
and hence, the available power.
18
19
annual precipitation of approximately 1700 mm. The total annual average run-off from the nations
600 rivers flowing from high mountains is over 200 billion m3 [1].
Most of the power plants in Nepal are run-of-river type with energy available in excess of in-country demand during the monsoon season and deficit during the dry season [1]. Some of the major hydropower
stations of Nepal along with the organization and their capacity is given in Table 2.1.
Table 2.1: Major hydropower plants in Nepal
Station
Organization/company
Capacity
Kaligandaki A
144 MW
Middle Marsyangdi
NEA
70 MW
Marsyangdi
NEA
69 MW
Kulekhani 1
NEA
60 MW
Khimti
60 MW
Bhotekoshi
36 MW
Kulekhani 2
NEA
32 MW
Trishuli
NEA
24 MW
Chilime
22 MW
Gandaki
NEA
15 MW
Jhimruk
12 MW
(2.1)
Where,
C2 : circumferential absolute velocity component at the outlet
C1 : circumferential absolute velocity component at the inlet
U2 : tangential velocity of the runner at the outlet
U1 : tangential velocity of the runner at the inlet
This Euler equation implies that in order to have a change in the total head, two ingredients are
necessary : tangential speed of the rotor and the change in the circumferential velocity component or
variation of the circulation between the inlet and outlet of the turbine.
20
x(m)
C2
U1
C1
W1
U2
W2
r
x
Flow passage
Relative streamline
Absolute streamline
(2.2)
Depending upon the sign of C , the sign of Htot can be determined. When this value is positive,
it means that the energy is added to the fluid and such kinds of hydraulic devices are called pumps.
When this value is negative, it means that the energy is extracted from the fluid and such kinds of
devices are called turbines. The velocities triangles of a typical turbomachinery rotor is shown in
Figure 2.1. The absolute and the relative velocities at the inlet and the outlet can be split into axial
and circumferential components. From the figure, C2 < C1 i.e. Htot < 0, so this is a case of a turbine
where the energy is extracted from the fluid.
In the case of radial-axial turbines, the radius at the inlet is not identical to the radius at the outlet.
This means that the tangential speed of the rotor is different at the inlet and the outlet. Also, the axial
and the meridional co-ordinates are not the same and instead of the axial co-ordinate (x), meridional
co-ordinate (m) has to be referred.
2.3 Cavitation
Cavitation is one of the principle challenges in hydro turbines, which occurs when the local pressure
falls below the vapor pressure of the water. This happens due to an increase in velocity or an ambient
drop in pressure. The water vapor forms at the area of low pressure in the form of bubbles, which
when carried to areas of higher pressure, can collapse violently. This collapse induces high pressures
and sets up fatigue stresses in nearby bodies.
A general cavitation phenomenon is shown in Figure 2.2, assuming a section of a pipe where a fluid is
flowing at a certain temperature and pressure with a velocity C1 . By entering the contracted region,
from the conservation of mass, the velocity of the fluid C1 increases to C2 . From the Bernoullis
21
principle, when the velocity of the fluid increases, the pressure in the fluid will decrease to maintain
the constant total pressure. This figure also shows the dependency of the saturation pressure on
the fluid (water) temperature. As the pressure in the fluid is decreased, the fluid will evaporate
at lower temperature, which means the formation of bubbles. When the fluid reaches the normal
situation (uncontracted condition) again, the fluid starts to decelerate and the vapor bubbles start to
disappear. Because of the increase of the fluid pressure, the vapor bubbles implode sending out small
but very high pressure micro jets. These micro jets, when close to the material surfaces, blast away
the material. This process is shown in Figure 2.3. Cavitation in the system can be checked through
a measure provided by the manufacturers called as N P SHr (Net Positive Suction Head required).
This N P SHr is compared with N P SHa (Net Positive Suction Head available), which is a system
parameter indicating the head surplus at inlet before the saturation pressure is reached.
[0 ]
100
30
1
Saturation pressure
4.2
101.3
[]
Figure 2.2: Cavitation due to contraction of a pipe and saturation pressure vs fluid temperature
1
Blast of material
22
impulse stage with all the pressure drop occurring across the stationary components and no pressure
drop across the runner. The reaction stages such as Francis and Kaplan turbines have a proportion
of the pressure drop in the rotor and a proportion of the pressure drop in the stator.
Figure 2.4: Jet impingement into a bucket with corresponding velocity triangles
23
24
+
+z
(2.3)
.g 2.g
.g 2.g
penstock
draf t tube
25
The general expression for the work done according to Euler momentum equation is given by,
work done = .Q(C1 .u1 C2 .u2 )
(2.4)
Where,
Q = Discharge through the runner, m3 /s
When C2 = 0, the maximum output is obtained.
Hydraulic efficiency, h is given by the total power developed by the runner over the power supplied to
the turbine. If H is the net head, then input to the turbine is given by .g.Q.H. Hence, the following
equation can be achieved:
.Q(C! .u1 )
.g.H.Q
(2.5)
C1 .u1
g.H
(2.6)
Shaf t power(P )
P ower developed by the runner
(2.7)
Shaf t power
P
=
W ater power
.g.Q.H
(2.8)
h =
or,
h =
Mechanical efficiency,
m =
Overall efficiency,
0 =
0 = h m
(2.9)
Hence, the overall efficiency of the Francis turbine can be deduced as a product of hydraulic and
mechanical efficiencies.
26
Table 2.2: Technical specification of Francis turbines installed in Hydro-power stations in Nepal
Station
Head [m]
No. of blades[-]
Kaligandaki A
3 x 48 MW
115
13
2.306 -2.564
Middle Maryangdi
2 x 38 MW
96.5
13
2.256 max.
Marsyangdi
3 x 26 MW
13
1.93 -2.234
Bhotekoshi
2 x 22 MW
135.5
Jhimruk
3 x 4.2 MW
201.5
17
0.540 - 0.890
The turbines shown in the table above are continuously facing the problem of sediment erosion. More
about the erosion problem these turbines are discussed in Chapter 3.
3
Sediment Erosion
Erosion in general, is one of the many categories of wear caused by the impact of particles of solid
or liquid against the surface of an object. The mechanism of the erosive wear is quite similar to the
abrasive wear, but in the case of the abrasive wear, the eroding agent is much bigger in size and the
angle of impingement is lower. The erosive wear on the other hand, is accompanied with relatively
small particles with several number of wear mechanisms. These mechanisms are differentiated based
on the impingement angle, size, shape and speed of the particles and the mechanical properties of the
base material. The pictoral representation of these mechanisms are shown in the Figure 3.1.
Impingement angle
Base material
Abrasive/cutting
erosion at Low impact
angle
Sa
Plastic deformation,
flakes at high impact
angle for ductile
material
The figure explains how the erosion takes place depending on the orientation and the properties of the
27
28
particles and the base material. These parameters also give the quantitative measure of the erosive
wear. For example, a low angle of impingement is favorable for the wear process as the particles are
drawn across the surface after the impact. Similarly, if the speed is low, then stresses at impact are
insufficient for plastic deformation or brittle fracture. In such cases, the wear by surface fatigue is
more probable depending upon the endurance limit of the base material. If the shape of the eroding
particle is blunt or spherical, the plastic deformation is more likely to occur, whereas, if the particles
are sharp, the cutting wear is more common. It has been seen that for the ductile mode, the maximum
erosive wear is generally found close to an angle of 30 whereas, for the brittle mode, the maximum
erosive wear is found around 90 of impingement angle [5].
Figure 3.2: Erosive wear for various materials at different impingement angles
It can be seen from the figure that some materials such as cobalt having a very good erosion resistance
at a low impingement angle but one of the worst materials for high impingement angles. According
to a study made between a martensitic (13Cr4Ni) and an austenitic (21Cr4Ni) steels, it was seen that
the erosion resistance of 21Cr4Ni strengthened with Nitrogen is higher than the former one due to the
distribution of hard carbides in the matrix of stabilized austenite [9].
In applications where the working temperature is high, ceramics are gaining a particular interest due
29
to their excellent high temperature properties. However, these materials are brittle which might result
in the brittle fracture.
The prevention of the turbine components can be done by applying a coating on the surface. These
coating materials depend upon the exposed environment, for example whether the surrounding is
wet or dry (hot). The most common type of coating seen in the hydro turbines is the Tungstencarbide(WC-Co) coating which typically uses 86-88% WC and 6-13% Co [10]. These coatings have
excellent hardness, with better adhesion and large toughness.
30
Runner
blades
Guide
vanes
Stay
vanes
Figure 3.3: Areas exposed to sediment erosion wear in Francis turbines [19]
31
Figure 3.4: Sediment erosion wear in the Francis turbine guide vane and runners in Jhimruk [19]
(3.1)
The expression for erosion was simplified in [11], which is given in Equation 3.2.
Erosion (velocity)m
(3.2)
Where m is the exponent of velocity. According to [12], the most general formula for the pure erosion
is give by Equation 3.3.
W = Kmat .Kenv .C.Vpm [mm/year]
(3.3)
Where W is the erosion rate in mm/year, Kmat is the material constant and Kenv is the environment
constant, C is the concentration of the particles and Vp is the velocity of the particle.
An erosion prediction was done based on 8 years of erosion data of 18 hydro-power plants in [13]
suggested Equation 3.4 to calculate erosion in turbines.
W = .C x .ay .k1 .k2 .k3 .V m [mm/year]
(3.4)
Where W is loss of thickness per unit time, is turbine coefficient at eroded part, V is relative flow
velocity, a is the average grain size coefficient on the basis of unit value for the grain size 0.05 mm.
The terms k1 and k2 are shape and hardness coefficient of sand particles and k3 is the abrasion resistant coefficient of the material. The exponent values x and y are for the concentration and the size
coefficient respectively.
32
According to [18], the erosion rate was estimated through laboratory tests of various turbine materials
under different test conditions. Equation 3.5 gives an empirical relation to predict the erosion rate for
16Cr5Ni, which is the most widely used turbine material.
y = 6E 5x3.13 [mg/kg]
(3.5)
Where x(m/s) is the velocity of eroding particles impinging at the angle of 45 and y is the loss of the
material in mg per kg of eroding particles striking the surface.
Recently, an erosion model was proposed in [14] that could estimate both absolute erosion rate
(mm/year) and corresponding reduction in efficiency (% per year) of Francis runners due to suspended particles. This model was termed as the improved version of the two former models. The final
equation yielded by this model was given by Equation 3.6 and Equation 3.7.
Er = C.Khardness .Kshape .Km .Kf .a.(size)b [mm/year]
(3.6)
r = a.(Er )b [%/year]
(3.7)
Where Km is the material factor, Kf is the flow factor, Kshape is the shape factor and Khardness is
the hardness factor. a and b are the empirical constants defined as :
a = 351.35, b = 1.4976 for quartz content of 38%,
a = 1199.8, b = 1.8025 for quartz content of 60%, and
a = 1482.1, b = 1.8125 for quartz content of 80%.
Model of Finnie
This model shows that the erosion is affected by the impact angle and the velocity given by:
E = kVpn f ()
(3.8)
Where,
E is a dimensionless mass,
Vp is the particle impact velocity and
f () is a dimensionless function of the impact angle which is in radian
n is the value of exponent which is usually in the range of 2.3 to 2.5 for metals.
33
(3.9)
Where,
f () = [1 + k2 .k1 .2sin(
/2 2
)]
0
(3.10)
RT = 1 k4 .VP sin()
(3.11)
(3.12)
(3.13)
Where,
0 is the angle of maximum erosion
k1 to k4 , k12 and 0 are model constants and depend on the particle/wall material combination.
The Tabakoff model requires the specification of five parameters : k12 constant, 3 reference velocities
and the angle of maximum erosion 0 . An example of these parameters for Quartz-Aluminum is shown
in Table 3.1.
Table 3.1: Coefficients for Quartz-Aluminum using Tabakoff Erosion Model
Variable
Coefficient
Value
k12
k12
0.585
Ref velocity 1
V1
159.11 [m/s]
Ref velocity 2
V3
194.75 [m/s]
Ref velocity 3
V4
190.5 [m/s]
25[deg]
4
Recent works - Review
The damage on hydraulic machineries due to sand erosion was initially studied in [17] and [18] through
various design aspects such as material selection, mechanics of material and hydraulics. This research
work led the path to carry out further investigations numerically and experimentally, which has now
become an integral aspect of the machinery design.
34
35
Figure 4.1: Hub, shroud and the blade passage from Turbogrid
Figure 4.2: CFX-pre setup file showing the blade passage and the mesh
36
268455
Factor ratio
Sediment
Quartz density
2.65 gm/cm3
1 kg/kmol
Particle Diameter
0.1 mm
0.586
Reference velocity 1
159.11 m/s
Reference velocity 2
194.75 m/s
Reference velocity 3
190.5 m/s
25 deg
Particle coupling
Turbulence model
SST
No Slip Wall
Inlet components
Mass flow rate
138.235 kg/s
0, 0.214349, 0.976757
Turbulence
Medium(Intensity = 5%)
0.07 kg/s
Particle position
Uniform injection
Uniform injection
Outlet
Relative pressure
1 atm
0.05
Solver control
Max. Iterations
100
Residual tolerance
1E-4
37
k 2 . 2 2
(ri x2 ))[P a]
2.g
(4.1)
k 2 . 2 2
(rp x2 ))[P a]
2.g
(4.2)
The concept of a fully coupled FSI in Francis turbines has been introduced in a study [22] where a
strongly coupled partitioned equations are solved separately using different solvers, but are coupled
implicitly into one single module based on a reduced-order model. The proposed model is used to
predict the unsteady flow fields of a 3D complete passage, involving in stay, guide vanes, and runner
blades of a Francis turbine. Such reduced-order model is based on only a few displacement and
stress modes, which not only saves computing time but also enlarge the range of applications in
engineering [22]. This study has also shown that the numerical results when considering FSI shows
better concordance with the experimental results than when not considering FSI.
A two-way coupled FSI of a propeller turbine is seen to have been made in the premises of ANSYS to
determine the mechanical integrity of the turbine blades by varying the stiffness of the blades [23]. A
multi-field simulation has been used in this study as CFD and FEA solvers to exchange information
at the interface.
38
Figure 4.5: Boundary conditions of the runner used in the study [7]
39
1
Shape 1
Shape 2
Shape 3
Shape 4
Shape 5
5
FSI review
Since the work is intended to be done in ANSYS, ANSYS coupled-field guide [24] is mainly referred
in this chapter.
40
41
of the different fields, providing more flexibility and efficiency when the coupled-interaction does not
have a high degree of nonlinearity. Coupling can be sequentially done either by a physics file or by
the multi-field solver (In the case of ANSYS, ANSYS-Multi field solver).
MFS-Single code
MFS code is used when the small models are used with all the physics field contained within a single
product executable. It uses an iterative coupling where each physics is solved sequentially and each
matrix equation is solved separately. The solver iterates between each physics until loads transferred
across the physics interface converge.
MFX-Multiple code
MFX code is used when much larger models are needed to be simulated compared to the MFS. It is
the enhanced version of ANSYS multi-field solver used for simulations with physics field distributed
between more than one product executable (eg. between ANSYS multiphysics or Mechanical and
CFX). A field solver runs different codes involved in the coupled-interaction. These fields are then
coupled using a form of iteration called stagger iteration. The solution loop now consists of two
loops. The multi-field time loop and the multi-field stagger loop. An example of the ANSYS multifield solver process for Fluid-Structure interaction between ANSYS mechanical and ANSYS CFX is
shown in Figure 5.2. The solution is divided into two different solvers, of which one is called master and
the other one is called slave. The master performs the coupling setup (reads all the MFX commands,
collects the interface meshes from the slave codes, does the mapping) and sends instructions (time
and stagger loop controls) to the slave executable. Contrarily, the slave code receives the coupling
information from the master code and sends the interface meshes to the master. In MFX, the ANSYS
code is always the master, and CFX code is always the slave. In the current study, MFX code will be
42
ANSYS Master
CFX Slave
Do Mapping
Time controls
Time loop
Stagger loop
Time loop
Stagger loop
Stagger controls(ANSYS to
CFX), Load transfers, Stagger
controls(bidirectional)
ANSYS
CFX
solver
solver
Time controls
End Time
loop
End Time
loop
43
(5.1)
Where,
u = velocity,
= density,
= dynamic viscosity,
p = pressure,
ui = cartesian component of velocity u in direction xi ,
Ff (t) = transient load vector defined for the fluid.
The equation of continuity for the incompressible flow is given by,
.
u =0
(5.2)
(5.3)
44
Where,
X = displacement,
M = Mass matrix,
K = Stiffness matrix,
C = Damping matrix,
Fs (t) = Transient load vector defined for the solid.
And finally, the mesh movement in the fluid domain may be modelled as a pseudo structural problem
with its own dynamics with a spring based mesh movement, governed by [25],
Km .dm = fm (t)
(5.4)
Where,
Km = a pseudo-structural stiffness matrix which is defined for the whole domain,
dm = displacement of the mesh,
CFD
CSD
+1
CMD
Figure 5.2: Schematic of Fluid structure interaction
The schematic diagram of the representation of FSI is shown in Figure 5.2. The figure shows the
interaction between the three fields and boundary information shared between each domain. From
the fluid dynamics, using the Navier-Stokes equation for the incompressible flow, the pressure on the
boundaries of the structure is calculated. This pressure is exported to the structural dynamics and
using the equation of motion, the deflection of the structure is known. The deflection of the structure
results in the distortion of the mesh of the flow field surrounding the structure, which affects the
computation of the fluid dynamics. The displacement of the mesh is transfered to the fluid dynamics
and the pressure field is calculated for the next time step. Other than Computational Fluid dynamics
(CFD), CSD and CMD in the above diagram represents Computational Structural Dynamics and
Computational Multi-body Dynamics respectively.
6
CFD analysis
The objective of the CFD analysis in the context of this project is to build up a base for the FSI
analysis. This base is made from a detailed CFD analysis, such that the model could be used further
in the FSI studies. A CFD model independent of the mesh density will be chosen from the mesh
convergence study. Then, the same mesh will be used in the sensitivity study, where various physical
and numerical parameters affecting the solution will be tested. This is because most of the parameters
used in CFD are based on various assumptions to simplify the solution. The sediment passing through
the turbine could have various physical properties. By carrying out the sensitivity study, the effect
of the variation of the input parameters on the results could be studied. Also, by increasing the
tolerance of the residual criteria, the solution could be significantly affected. Following parameters
will be investigated in the sensitivity analysis :
Effect of the particle size (particle diameter)
Effect of the particle shape (particle shape factor)
Effect of the particle behavior (mass flow rate, concentration)
Effect of the erosion model
Effect of the numerical parameters such as convergence criteria and turbulence model
These parameters will undergo independent variations, which will be performed on a baseline case.
After performing the sensitivity study, the reference design will be compared with the 4 other optimized
designs in terms of blade shape and the blade showing the best erosion resistance without influencing
the efficiency will be chosen for the further studies.
45
46
(6.1)
where mp is the mass of the particle and N is the number rate. The overall erosion of the wall is then
the sum over all the particles. The erosion rate is in kg/s, but in the CFX-Post Processor, the results
are in the form of kg/s/m2 and named as Erosion Rate Density.
47
This problem of mesh convergence for the other cases was treated by decreasing the residual criteria,
the RMS value from 1E-4 to 1E-6. The computational time by doing this increased a lot, but it
improved the convergence behavior, as shown in Figure 6.1. Here, the factor ratio was chosen to be
1.15 in order to have a better quality of mesh than that of the previous studies. After the target
mesh of 0.75 million node, the solution did not change much even at 2 million nodes. The better
convergence of the mesh than the previous studies was achieved due to:
The residual criteria for convergence was decreased from the RMS value of 1E-4 to 1E-6.
The modification made in the CFX parameter for the inlet angle. This is defined in Appendix-I
in Chapter 11.
This figure also shows the y+ values for various mesh nodes under study. For the most refined mesh,
the average value of y+ around the blade was about 37.
7
x 10
1.2
2000
1.15
Average y+ on the blade
1.25
1.1
1.05
1
0.95
1500
1000
0.9
500
0.85
0.8
X: 2.04e+006
Y: 37.44
0
0
0.2
0.4
0.6
0.8
1
1.2
1.4
Number of Nodes
1.6
1.8
2.2
6
x 10
0.5
1
1.5
Number of Nodes
2.5
6
x 10
Figure 6.1: Mesh convergence study for the factor ratio of 1.15, RMS of 1E-6 and y+ value on the
blade
The second reason can be explained from Figure 6.2, where the erosion pattern on the blade was
converging for the finer mesh. This contradicts the previous study, where the erosion almost vanishes
after the mesh was made very fine [21]. From these figures, it can be concluded that with the residual
target of 1E-6, the target mesh of 0.75 million node is sufficient for further study (FSI). However,
in order to study the erosion pattern better, the mesh of 1.25 million node was chosen to study and
compare the erosion behavior of the optimized blades.
48
Target mesh
20000
100000
500000
1250000
250000
750000
2000000
49
0.1 mm
0.3 mm
0.5 mm
0.7 mm
1 mm
0.9 mm
Figure 6.3: Effect of the size of the particle on the erosion pattern
except when the diameter is 0.01 mm as shown in Figure 6.4. This shows that the size of the particle
has a significant but uneven influence on the erosion pattern.
PS
SS
Figure 6.4: The erosion pattern for the particle diameter of 0.01 mm on both Pressure and Suction
side
50
x 10
x 10
1
0.8
1.5
0.6
0.4
0.5
0.2
0.1
0.2
0.3
0.4
0.5
0.6
0.7
Particle diameter [mm]
0.8
0.9
2.5
0
1
Figure 6.5: Average and maximum erosion rate density on the blade for various particle sizes
account. The non-uniformity of the shape was not considered in this study.
The result of this analysis is shown in Figure 6.6. The margin in the legend was made wider in order
to have a better picture of the differences in the results. The spherical particle was seen to have the
least effect on the erosion compared to the elliptical ones. However, there was not much difference in
the overall erosion pattern.
51
The direction of the particle is made identical to the direction of the fluid flow. The particle mass
flow rate was varied between 1 to 50 kg/s per machine. The particle diameter was kept constant (0.1
mm). The particles are uniformly injected with 1000 particles at the inlet. This number can also be
chosen as Proportional to mass flow rate where the number of particles per unit mass flow should be
specified.
The effect of the mass flow rate of the particle on the average erosion rate density on the blade is
shown in Figure 6.7. The erosion increases linearly with the mass flow rate. Since other parameters
like the injection position, number of particles and the diameter distribution were made constant, the
erosion patterns between these mass flow rate were the same.
4.5
x 10
4
3.5
3
2.5
2
1.5
1
0.5
0
0.5
1
1.5
2
Mass flow rate per passage (kg/s)
2.5
52
study, the RMS of 1E-6 shows better results in terms of the mesh independence study. Hence, the
RMS of 1E-6 was used as the erosion is more of the interest than other parameters in this project.
1.2
1
Pressure (MPa)
0.8
0.6
0.4
0.2
0
0.2
0.4
0
0.1
0.2
0.3
0.4
0.5
0.6
Streamwise(01)
0.7
0.8
0.9
Figure 6.8: Effect of the residual criteria for convergence on the blade loading
53
SST
k-
54
Tabakoff
Quartz-Aluminum
Average Sediment
erosion rate density on
the Blade:
8.77E-8 kg/m2s
Tabakoff
Quartz-Steel
Average Sediment
erosion rate density on
the Blade:
1.25E-7 kg/m2s
Finnie
Velocity power
factor, n = 2
Reference velocity = 1
m/s
Average Sediment
erosion rate density on
the Blade:
17.68 kg/m2s
55
It can be seen from this comparative analysis that the efficiency of the blades does not vary much
with the change in the shape. However, the erosion rate density varies a lot. The blade shape 2 and 4
are the ones which show a decrease in the average erosion rate density on the blade compared to the
reference shape. The decrease in this quantity is as much as 21% for the case of shape-4. Hence, the
blade shape 4 is the most optimized blade in terms of erosion and efficiency and it is selected for the
FSI analysis.
However, it could be inferred from the mesh study that though the mesh convergence study was
performed on the reference blade, and the results were very sensitive to the type of the mesh generated,
it could be interesting to see how the types of mesh vary the results of other blades as well. This will
make the comparison more reliable as each of the CFX model will be independent of the mesh density.
It can be concluded that the erosion predicted by ANSYS CFX is mesh sensitive and requires a high
quality and high density to have a converging result. The mesh needs to be refined significantly
without disturbing the overall quality of the mesh. This was done in this project by reducing the
factor ratio, but making the mesh density large, so that both the criteria are met.
0,00
0,20
0,40
0,60
0,80
1,00
1,20
8,97
98,29
7,07
8,11 97,81
6,40
97,40
8,51
98,01
Figure 6.11: Sediment erosion results for various shapes of the blade
97,30
(E-8 kg/m2s)
Erosion pattern of
the blades
(%)
Efficiency
56
7
Structural analysis
The structural analysis was performed in Static Structural in Workbench. This chapter deals with
the modeling of the 3D geometry of the runner for the structural analysis and the first structural
analysis setup and results. This chapter primarily focuses on carrying out one-way FSI in ANSYS. The fully coupled analysis will be carried out in the next chapter. For both the cases in
this report, Structural Steel was chosen as the structural material having following properties :
Density
7850 kg/m3
Youngs Modulus
2E11 P a
Poissons Ratio
0.3
7.1 Geometry
The geometry of the blade was made with the help of the curve files in Pro Engineer. The leading
edge and the trailing edge were made from the instruction file of NTNU [27]. The leading edge was
designed in such a way that on the pressure side, the edge was rounded with quarter a circle whereas
on the suction side, it was rounded with an arc of thrice the radius of the leading edge. The trailing
edge was designed with a 30 degree angle on suction side with the end part chopped off to prevent the
breaking off due to the pressure fluctuation in the system.
LE
TE
Figure 7.1: Leading edge and trailing edge design for FEM [27]
The hub and the shroud were made from the respective curve files, providing the necessary thickness
57
58
and the place for labyrinths. Before the assembly was made between these three parts, the hub was
cut into a single sector containing 1 out of the 17 blades. This was done in Pro-E by the following
sequence,
Only (360/17) of the total part was modeled. This was done by extruding out the unwanted
portion from the hub taking the reference of the blade profile. The blade profile at the first
section was imported. This was done at a separate datum plane such that the curves could be
projected to and away from the surface of the hub.
The profile section consists of curve made of points at the pressure side and the suction side of
the blade. This curve was patterned along the main axis with the value of (360/17 4) and
4 on the direction as shown in Figure 7.3, so the blade lies in the position close to the CFD
domain. By placing the blade in this position gave better mapping with the CFD domain than
when placed in the mid position. It can be verified from Figure 7.2, where cases with proper and
improper mappings have been shown. The first figure to the left shows that the fluid domain is
shifted upwards compared to the structural domain, whereas in the second figure, the fluid and
the structure are aligned more closely. Important thing to consider is the amount of decimal
places that needs to be provided in Pro-E, to maintain a consistency with ANSYS. The value of
(360/17) is not good in Pro-E because it will round-off the value in lower decimal places than
needed in ANSYS. The more precise value of 21.17646 was used.
The patterned curves were used to create a closed curve so that the extrude could be made as
shown in the figure 7.3. This was done by choosing two curves on the same side (Pressure side
in this case). A circle was made of arbitrary radius, but bigger than the hub diameter. This
circle was connected to the two curves tangentially through a line on the inlet side and through
a line connecting the center and the final point on the outlet side.
Similar process was done for the shroud. It was seen that when the sharp edges are included in
ANSYS while applying the cyclic boundary, the program could not recognize the boundary and
it terminated with an error. Hence, the sharp edges had to be cut in both hub and shroud.
Finally, the blade, the hub and the shroud were assembled together. Also, it was seen that a
single merged part was needed in ANSYS rather than three different parts. This was done by
importing the geometry file of each part as an independent file and saving as a part file rather
than the assembly file.
Fluid domain
Structural domain
Figure 7.2: Comparison of the two domains with two cases (right one shows better mapping)
59
Full Hub
Blade Profile
Profile duplicate at 4 degree
about the centre
Full Shroud
Hub Sector
Shroud Sector
Arbitrary radius circle outside the
boundary
Figure 7.3: The geometry modeling in Pro-E as described in the above procedure
60
7.2.1 Case I
This case consists a single blade of the runner with the following conditions :
Zero displacement (fixed support) of the surface connecting blade-hub and blade-shroud.
Rotational velocity about the z-axis with 104,7 rad/s.
Acceleration due to gravity (g).
Imported pressure load on the blade surface.
The reason behind conducting this analysis is to check the structural integrity of the blade without
the influence of the other stiffer components. This will be easier when the comparison has to be made
with the optimized blade. According to similar studies mentioned above, this type of analysis gives a
reasonable estimation of the stress induced by the flow on the blade. However, the maximum stress
could be less than the actual value because of the joint between the blade-hub and the blade-shroud.
This analysis will be used as a starting solution to see the stress distribution and the deflection of
the blade due to the pressure load from the flow field, without the influence of the other structural
components and the neighboring blades. The boundary conditions imposed in ANSYS are shown in
Figure 7.4.
7.2.2 Case II
This case consists of the following boundary conditions :
Zero displacement (fixed support) of the surface connecting the shaft and the hub.
61
62
12
Converged solution
12
10
10
8
6
6
4
0
0
50000
100000
150000
200000
250000
300000
Reference design
350000
2
0
100000
200000
300000
400000
500000
Optimized design
63
7.6 Case-II
The result of the Case-II is shown in Figure 7.8, Figure 7.9 and Figure 7.10. The maximum stress
occurs at the connecting surface of the shaft and the hub, because of the fixed constraint defined on
this surface. The second maximum stress occurs at the leading edge, on the blade-hub region. On
comparing this result with Case-I, it can be seen that the maximum stress is not on the outlet region,
but it is on the inlet blade-hub region. The loads on the hub and the shroud are also imported on
the structure, due to which stress values are more than the previous case. The values of the pressure
load is maximum at the leading edge regions, due to which, the stress values are also maximum.
Besides, the connection between the blade and the other components induces stress concentrations,
which makes these regions prone to failure.
On comparing the results between the reference and the optimized design, it can be seen that the value
of the maximum stress is bigger in the reference design than in the optimized design. The maximum
stress on the reference design is around 179.3 MPa whereas 153.8 MPa for the optimized design.
Similarly, when only the stress on the blade is considered as shown in the figure 7.9, the maximum
stress on the reference design is 123.7 MPa whereas it is 114.9 MPa on the optimized design.
Similarly, in the figure 7.10, the deformation of the runner and the blade is compared. The maximum
overall deformation is seen in the hub towards the inlet due to high pressure load in that region, for
both the designs. The value of the maximum deformation is 0.087 mm for the reference design and
0.093 mm for the optimized design. When only the blades are considered, the maximum deformation
is towards the shroud-blade connection for both the designs. This value is also higher in the case of
optimized design, with 0.078 mm than in the case of reference design, with the value of 0.066 mm.
64
Reference design
Optimized design
65
Reference design
Optimized design
66
Reference design
Optimized design
Figure 7.9: Result of one-way coupling (Stress distribution on the blade) for Case-II
67
Reference design
Optimized design
8
FSI analysis
ANSYS muti-field solver (MFX) was used to do the two-way FSI. This can be done either in the
Workbench or standalone applications. One example of the project schematic in Workbench for
conducting two way FSI is shown in Figure 8.1. Like the one-way FSI technique, two independent
models are needed for the two domains in this context. An input file is created from ANSYS structural,
which is imported in CFX, where all the solver parameters are set. One of the inevitable things while
conducting the two-way FSI is the deformation of the CFD mesh. The deflection of the structure
results in the deformation of the fluid mesh, which changes the flow field surrounding it. Hence, an
appropriate model of the mesh displacement and a sufficient value of the mesh stiffness have to be
chosen to avoid the folding of the mesh, which is very common in two way FSI. Some of the important
aspects to consider while conducting two-way FSI are discussed below:
.(disp .) = 0
(8.1)
Where, is the displacement relative to the previous mesh locations and disp is the mesh stiffness.
This equation is solved at the start of each outer iteration. It can be inferred from this equation that
the relative mesh distribution of the initial mesh has been preserved. This means that the refinement
of the mesh at the boundary will remain fine after the deformation at the relative position.
68
69
ds.dat
disp =
1
a
Cstif f
(8.2)
Here a represents either the size of the control mesh volumes, or the distance from the nearest
boundary, depending upon the type of the option chosen. In any case, the stiffness of the mesh will
increase when this value will decrease. The rate at which this stiffness increases depends upon the
value of Cstif f , which represents the model exponent. The default value of this exponent is 10, which
can be altered according to the need of the problem. In the case when ANSYS Multi-field is chosen
for FSI, the mesh motion can be imposed on the wall boundaries. In this case, the mesh motion has
to be imposed on the blade, hub and shroud regions.
70
(1,2...) and CFX surfaces are flagged by interface name (FSIN 1, FSIN 2...), these numbers and the
name on the two domains should represent the same boundary. A typical procedure of setting up the
interface for the blade is listed below :
In ANSYS Structural, choose Fluid Structure Interface and select the blade surface. This
interface will be given a number (Starting from 1)
When the input file is written for the structural case, or when the setup of the structural and
CFX are linked, the input from the structural can be read in the CFX in the Analysis type.
In the domain definition, choose the mesh motion to Regions specified. This will enable the
selection of the mesh motion option on the wall boundaries.
On blade, select the mesh motion to be ANSYS multi-field. Here, the selection of the appropriate
interface number can be selected. This should correspond to the number provided in structural.
The ANSYS Multi-field solver automatically transfers mesh-based quantities across dissimilar meshes.
However, it is better to have similar size meshes between the two fields, to ensure the correct mapping
between the fields. The quality of the mesh of the two fields is shown in Figure 8.2. The distribution
of the mesh is finer near the edges in the case of fluid, as this mesh was created from highly optimized
topology option from Turbo-grid. Hence, the quality of the mesh is better in the fluid domain.
CFD mesh
Structural mesh
71
log(/min )
log(10/min )
(8.4)
When < min , e < 0 in the plot and the convergence occurs. A convergence plot used in this
study is shown in Figure 8.3. The stagger iteration starts with the start of the simulation. The next
iterations are done as a normal CFD-solver, where the convergence occurs after reaching the RMS
residual criteria, or the maximum number of iterations set in the solver control. After the end of the
first stagger iteration, a new monitor plot appears, which shows the load transfer from CFD to the
structural. After the load is transferred, the flow field is disturbed and the solver tries to converge the
momentum and mass for the new deflected position of the blade. For each iteration, the value of e is
calculated from the Equation 8.4. The stagger iteration continues till the value of the parameter e is
negative. The figure shows that the convergence of the coupled-field analysis occurs after 6 stagger
iterations.
72
SI = 2
SI = 3
SI = 4
Figure 8.3: Convergence plot in CFX-solver for the FSI analysis in this study
73
8.5.2 Case-II
The stress distribution of the case-II for two way coupling is shown in Figure 8.5. The region of the
maximum stress is towards the blade-hub inlet region. A high stress distribution was found towards
the joint between the blade-hub and the blade-shroud. The value of the maximum stress is 916.8 MPa
for the reference design and 823.9 MPa for the optimized design. Compared to the one-way FSI, this
value is about 7 times big. This difference reflects the significance of conducting two way FSI in this
application.
In two way FSI, the deflection of the structure is represented by the deformation of the mesh of the
flow field surrounding it. Figure 8.6 shows the magnitude of the total mesh displacement around the
blade at three different stream-wise positions. It also shows the overall deflection of the CFD mesh
surrounding the blade. In the case of the reference design, the deformation of the mesh is maximum in
the shroud region. This figure only shows the local displacement value on the legend. On the runner
blade, the maximum deformation was found to be 0.37 mm. In the case of the optimized design, the
pattern of the deformation on the blade is quite different. Maximum deflection occurs towards the
inlet and the midspan of the outlet region. Though the maximum deformation on the blade is less
than the reference design (0.12 mm), the overall deformation of the CFD mesh is similar.
74
8.30E6
max
3.9E6
7.7E6
Reference design
1.25E7
1.25E7
max
6.12E6
max
6.12E6
9.61E6
9.61E6
Optimized design
75
Reference design
Optimized design
76
Reference design
Optimized design
Figure 8.6: Mesh deformation in the fluid domain from two-way FSI
9
Conclusion
The objective of this project work was to carry out the FSI analysis of reference and optimized
Francis runners exposed to sediment erosion. An optimized runner was chosen between 5 shapes
of the blade proposed by earlier studies based on the reduced erosion effect without affecting the
efficiency. The erosion effect was imposed in ANSYS-CFX with Tabakoff Erosion Model by building
a mesh independent model. Some assumptions were chosen to build the CFD parameters, such as
shape, size and concentration of the quartz particle in the flow. However, a sensitivity study was
made, where the effect of changing these parameters were observed. To do this, a baseline case was
initially established and a single parameter was varied at a time keeping other parameters constant.
The results of CFD were very sensitive to the size and distribution of the mesh. The CFD meshing
was done with the help of the ATM optimized topology in Turbogrid, which is known for generating
a high quality mesh but less freedom of choices. By decreasing the factor ratio, it was seen that the
convergence could be obtained in the results. A target mesh node count of 0.75 million node was
chosen for the FSI study, but in order to study the erosion pattern better, a mesh of 1.25 million node
was chosen for the CFD study. Among the five different runner blades, shape-4 (the physical meaning
of the shape was described in the figure 4.6) was seen to have reduced erosion effect by 21% without
affecting the efficiency. Hence, it was chosen for the further study.
The structural model of the blade and a sector of the runner was made in Pro-E, with the help of
the curve files generated from the Matlab Program. While generating the section of the runner, the
hub and the shroud regions were trimmed so as to map with the fluid model closely. Two types of
boundary conditions were considered, where the FSI analysis was performed on each of the cases. In
the first case, a single blade without the hub and the shroud were created, whereas in the second case,
the full geometry consisting of a sector with 1/17th of the total runner was developed. The results
showed that the stresses, which were plotted as the equivalent Von Mises Stress, were larger in the
second case than in the first case. This is due to the effect of the joints and the loads imposed on the
hub and the shroud, which were not considered in the first case.
One way FSI was conducted in Chapter 7 on both the cases and both the designs. A mesh independent
study was performed on the first case by considering a mapped tetrahedral mesh around the blade.
For the second case, a fixed size mesh was used in both the designs. FSI was imposed by importing
the loads from CFD on the surfaces of the structure. In one way FSI, the cyclic symmetric property
77
CHAPTER 9. CONCLUSION
78
of the blade was established by defining the high and the low periodic boundaries. Other boundary
conditions, such as rotational velocity, gravity, and fixed support on the surface connecting the hub
and the shaft were imposed. The results of one way FSI showed that the maximum stress on the
runner is less in the optimized design compared to the reference design, in both the cases. The value
of the maximum stress on the runner decreased by around 14%. On the contrary, the maximum
deflection of the runner increased by around 6%.
Two way FSI was conducted in Chapter 8 on both the cases and both the designs. This was done by
writing an input file from the structural case consisting of the same boundary condition as one way
FSI, but importing it into CFX on the specified boundaries. The deformation of the mesh on these
boundaries was selected with the provided mesh deformation model. The maximum number of stagger
iteration was chosen to be 10 but it was seen that the results converged after maximum 8 iterations.
The post processing for both the fields were done in CFX-Post. In the first case of boundary condition,
the maximum value of the stress increased by around 29% for the optimized design than the reference
design. However, in the second case, the value of the maximum stress decreased by around 10%.
Compared to one-way FSI, the value of the maximum stress was around 7 times bigger than when two
way FSI was performed. The difference in the results with the same mesh and boundary conditions
show the significance of conducting the fully coupled analysis in this field. This vast difference also
resulted because of the different ways of imposing the cyclic symmetry boundary condition to the
runner. It was seen that when the APDL command was used to make the cyclic property of the
sector, it resulted in mapping problems. This problem could be solved in Workbench Static Structural
by using the model > symmetry > cyclicsymmetry in the outline tree directly and selecting the
higher and the lower periodic boundaries. However, this option was not possible for the two way FSI
till the current release of ANSYS. Hence, for the two way FSI, APDL command was used to impose
the cyclic symmetry property. It was seen that using the APDL command not only shows the mapping
problem, but also overestimates the stresses by some amount. In that way, the results of the two-way
FSI could have escalated a bit more than expected, for the second case. A demonstration of the
comparison between the two ways of imposing the cyclic symmetry condition is shown in Appendix-II
in Chapter 12.
One more limitation in the ANSYS release was the restriction in selecting the type of the mesh. It
was seen that a model with a periodic symmetry property was unable to generate a hexagonal mesh.
In order to maintain the consistency in the type of the mesh, tetrahedral mesh was used for all the
cases. Creating hexagonal mesh could result in better mapping with the fluid interface because of the
similar type of the mesh.
10
Future scope in the related field
In the case of the CFD analysis, the mesh convergence study was performed on the reference
blade only. However, performing the study on all the blades will ensure the correctness of the
solutions as all the blades will be mesh independent.
In the case of the structural analysis, the mesh study was performed on the first case only, where
only a single blade was considered. It was seen that for the second case, about 2 million mesh
was needed to have a mesh independent solution. In order to perform the mesh study for this
case, a very high computational time is needed. Also, the hexagonal mapped mesh could be
generated from other meshing tools or softwares to have better results.
A single blade of the runner was considered in this study. The Matlab program which were used
to generate the curve files of the blade can also generate curve files and input parameters of the
guide and stay vanes. This can be useful to find erosion information on these regions, along with
their structural integrity.
This project still assumed the results to be steady. However, by considering the guide vanes,
the transient calculations could be done and the unsteady forces on the runner blades could be
known.
The material of the structure in this study was chosen to be Structural Steel. However, the
analysis has to be carried out for the exact material from which the manufacturing is done. The
most common materials that are chosen are stainless steel and titanium- and nickel- alloys.
Perfect mapping of the two domains (fluid and structure) has to be carried out. This will increase
the accuracy of the solution.
79
Bibliography
80
81
BIBLIOGRAPHY
82
BIBLIOGRAPHY
[22] Wang, W.Q., He, X.Q., Zhang, L.X., Liew, K.M., Guo, Y.,
Strongly coupled simulation of fluid-structure interaction in a Francis hydro turbine,
International journal for numerical methods in fluids, Wiley Interscience, 2008, pg: 515-538.
[23] Schmucker, H., Flemming, F., Coulson, S.,
Two-way coupled fluid structure interaction simulation of a propeller turbine,
25th IAHR Symposium on Hydraulic Machinery and Systems, Voith Hydro GmbH and Co. KG,
Germany, 2011.
[24] ANSYS Coupled-Field Analysis Guide,
ANSYS Release 10.0, 2005.
[25] Slone, A.K., Pericleous, K., Bailey, C., Cross, M.,
Dynamic fluid-structure interaction using finite volume unstructured mesh procedures,
Computers and Structures, pg: 371-390, 2001.
[26] Eltvik, M., Thapa, B.S., Dahlhaug, O.G., Gjosaeter, K.,
Numerical analysis of effect of design parameters and sediment erosion on a Francis runner,
Proceedings of Fourth International Conference on Water Resources and Renewable Energy Development in Asia, Thailand, 2012.
[27] Wei, Z., Finstad, P.H., Olimstad, G., Walseth, E., Eltvik, M.,
High Head Hydraulic Machinery, compendium,
Water Power Laboratory, NTNU, 2009.
[28] Negru, R., Marsavina, L., Muntean, S.,
Analysis of Flow induced stress field in a Francis turbine runner blade,
Buletinul Institutului Politehnic Din Iasi (Bulletin of the Technical Institute of Iasi), 2011.
[29] Negru, R., Muntean, S., Marsavina, L., Susan-Resiga, R., Pasca, N.,
Computation of stress distribution in a Francis turbine runner induced by fluid flow,
Proceedings of the 21st International Workshop on Computational Mechanics of Materials, 2011.
[30] Saeed, R.A., Galybin, A.N., Popov, V., Abdulrahim, N.O.,
Modeling of the Francis turbine runner in power stations. Part II:stress analysis,
WIT Transactions of the Built Environment, Vol 105, 2009.
[31] Lais, S., Liang, Q., Henggeler, U., Weiss, T., Escaler, X., Eguasquiza, E.,
Dynamic Analysis of Francis Runners-Experiment and Numerical Simulation,
International Journal of Fluid Machinery and Systems, Vol 2, 2009.
11
Appendix-I - Some discrepancies with the design program (Khoj)
The curve files of the blades along with the boundary conditions needed for the CFX simulations
were taken from the Matlab design program called Khoj. More information about the theories and
executing the program can be found in the earlier papers [19] and [21]. In this thesis project, these
parameters and files were needed as inputs to conduct the CFD and FSI in later part of the project.
The output CFX parameters was provided in the following way:
Blades
17.000000
Flow rate
2.350000
Rotational speed
-1000.000000
Velocity components:
Between runner and guide vanes:
C_theta
0.976757
C_r
0.214349
C_z
0.000000
Between guide vanes and stay vanes:
C_theta
0.834840
C_r
0.550493
C_z
0.000000
At stay vane inlet:
C_theta
0.855737
C_r
0.517410
C_z
0.000000
These parameters provide information about the inflow conditions, rotational speed of the runner and
the direction of the flow towards the stay vane, guide vane and the runner inflow separately. These
parameters provide the freedom of modeling the runner independently or together with the stationary
components when needed. There have been few concerns raised during the use of these inputs, which
are discussed in this chapter. The results generated in the previous studies and the conclusions have
been modified accordingly.
83
CHAPTER 11. APPENDIX-I - SOME DISCREPANCIES WITH THE DESIGN PROGRAM (KHOJ)
84
Figure 11.1: Boundary Vector at the inlet with the given flow direction
The explanation of the direction of the flow is explained in Figure 11.2. The co-ordinate axes defined
here are according to the convention used in CFX. It can be seen that the tangential component of
the velocity should be in the direction of the rotation, which is negative. Also, the radial component
should be towards the inward direction, which is also negative in this case.
This misalignment was not only seen for the runner inlet, but even when the stationary components
were modeled, the flow vector was in the opposite direction, as shown in Figure 11.3. Here, both the
guide vane and the stay vane have been modeled and the inflow condition was given according to the
same file, but for the stay vane inlet. This can be more clearly understood for the stationary domain,
as there is no rotational component and with that, the absolute velocity should be aligned towards
the direction of the flow. It can be seen that the angle of the inflow looks acceptable, but the direction
has been misaligned.
85
CHAPTER 11. APPENDIX-I - SOME DISCREPANCIES WITH THE DESIGN PROGRAM (KHOJ)
Inlet
Outlet
Shroud
Hub
Figure 11.2: Boundary Vector at the inlet with the given flow direction
Figure 11.3: Boundary Vector at the inlet with the given flow direction
CHAPTER 11. APPENDIX-I - SOME DISCREPANCIES WITH THE DESIGN PROGRAM (KHOJ)
86
A significant effect was seen in the sediment erosion density plot. Major difference was seen when the
stationary components were modeled, as shown in Figure 11.4. With the provided information, no
erosion effect was seen on the turbine components, which indicates some errors in the CFX parameters.
On changing the direction as discussed above, the erosion pattern was seen profoundly and at the
expected places (Guide vane, stay vane inlet and runner outlet). Even when only the runner is
modeled, the erosion was seen in the bigger amount when the directions were modified than when
they were unmodified.
87
CHAPTER 11. APPENDIX-I - SOME DISCREPANCIES WITH THE DESIGN PROGRAM (KHOJ)
Before
No erosion!
Modified
Figure 11.4: Result of the two flow directions, unmodified (top) and modified(below)
CHAPTER 11. APPENDIX-I - SOME DISCREPANCIES WITH THE DESIGN PROGRAM (KHOJ)
88
CHAPTER 11. APPENDIX-I - SOME DISCREPANCIES WITH THE DESIGN PROGRAM (KHOJ)
Head : 205.6 m
Head : 153.22 m
89
Figure 11.5: Discrepancy between the results when only the runner and the full stage is modeled
12
Appendix-II - Imposing cyclic symmetry boundary conditions in
ANSYS
It was seen that the cyclic symmetry boundary conditions can be imposed in ANSYS Workbench by
either of the two ways :
From workbench, choosing the symmetry property and selecting the higher and the lower boundaries of the symmetric body. Doing this requires a new cylindrical co-ordinate system defined
and the exact geometry on the higher and the lower side so that the geometries and mesh are
properly mapped.
By using commands in Mechanical (APDL). This can also be done in workbench itself by writing
the following commands,
/prep7
cyclic,17
/solu
Tolerances to map the faces can be chosen as per necessity.
The solutions of the two options shown above were apparently different. In the latter case, when the
APDL command was used to impose the cyclic property to the runner, it resulted in some mapping
problems. One solution of removing this problem is to define certain tolerances for the mapped mesh.
But since this option could not get the robust solution, the former option was chosen. Here, by
defining the high and the low boundaries of the cyclic region, the problem of the mapping did not
persist. This comparison is shown in Figure 12.1. Also, it can be seen from the comparison that the
maximum stress predicted by the latter case is much more than the former case.
Although the results of the former case was adopted for one-way FSI, this option was not valid for the
two-way FSI till the current release. Hence, the latter option was chosen for that case. Though the
mapping problems could not be seen in the two way FSI, because the post processing is carried out in
CFX-Post, it is expected that the results of the analysis overestimated the stresses by some amount.
90
91
Mapping problems
Figure 12.1: Choices of imposing cyclic symmetry property to the sector of the runner