Professional Documents
Culture Documents
Renewable Energy
journal homepage: www.elsevier.com/locate/renene
a r t i c l e i n f o a b s t r a c t
Article history: Sediment particles, especially quartz which has very high hardness factor, flowing along with water
Received 23 June 2015 erodes turbo machinery parts such as guide vanes, runner, draft tube etc. Among them runner is very
Received in revised form crucial part, so its design should be optimum for minimum erosion by maintaining the highest possible
15 October 2015
efficiency.
Accepted 15 October 2015
Available online xxx
This paper reports on methodology for designing Francis runner blade. This involves finding best
outlet angle (b2) and blade angle distribution (b-distribution) causing minimum possible erosion for
given volume flow rate (Q ¼ 14.34 m3/s), head (H ¼ 40 m), rpm (N ¼ 333.33 rpm) and eroding particle
Keywords:
Francis turbine
flow rate (0.08 kg/s). At first, the outlet angle b2 (designing parameter) was varied from 14 to 32 using
Erosion linear blade angle distribution for all models. Then these models were simulated to find b2. In second,
Efficiency with that b2, blade angle distribution was varied and simulated to find the best blade angle distribution
b-distribution having minimum erosion rate with considerable efficiency. By using hydrodynamic theory for given Q
and H, main dimensions were found out and 3D model was generated using b-distribution. Relation of
designing parameters with erosion and efficiency was made. Optimum blade was obtained from pro-
posed methodology and was compared with the reference blade in terms of erosion and efficiency.
© 2015 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.renene.2015.10.023
0960-1481/© 2015 Elsevier Ltd. All rights reserved.
308 K. Khanal et al. / Renewable Energy 87 (2016) 307e316
Nomenclature
the application of CFD, FEM and FSI at different stages [4]. Neo-
pane et al. presents the laboratory study of particle velocity
measurement in highly swirl conditions similar to turbine flow in Fig. 1. Axial view of runner.
curved path [5]. Thapa et al. studied the relation between blade
angle distribution of runner blade and the erosion rate [6].
Gaungjie et al. investigated the relation between the wear rates
on the surface of runner blade and guide vane and the sediment
concentration, and analyzed the distribution of wear rates for
normal turbine operating condition [7].
The main objective of the research is to propose a method-
ology which minimizes the erosion in Francis runner blade by
considering efficiency. This research also aims to obtain the
relation of designing parameters with erosion and efficiency so
that it can help in prediction of designing parameters for any
analogous case. Based on the methodology, this paper intends to
propose an optimum blade and compare result with reference
blade. For this, Devighat Hydropower Plant (DHP) with head
(H) ¼ 40 m, volume flow rate (Q) ¼ 14.34 m3/s, speed
(N) ¼ 333.33 rpm situated at Nuwakot district of Nepal was taken
as the reference site. This study focuses on the erosion of the
Francis runner blade for design condition (not for partial and full
load condition) without considering the effect of guide vanes and
draft tubes. Coalesced effect of erosion and cavitation was also
not included in this study.
Fig. 2. Velocity triangle.
D2, U2, U1, D1, B and b1 are calculated from following relations [3].
2.1. Francis turbine design
qffiffiffiffiffiffiffiffiffiffi
4:Q
Francis turbine is operated by utilizing the potential energy of D2 ¼ P:Cm2
the storage water partially into the kinetic energy and partially into N ¼ 60.U2/(P.D2)
pressure energy. The water enters the turbine radially, imparts Cm2 ¼ U2.tan (b2)
energy to the runner blades and leaves axially. It works above at- h ¼ (U1Cu1 U2Cu2)/(gH)
mospheric pressure and will be fully immersed in water. If there are Rearranging we have
sediment particles in the water flow during energy transfer process, qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
240:Q
D2 ¼ 3 p2 :N:tanðb Þ
then it will strike turbine and erode the surface to the turbine. 2
2.2. Wear mechanism and erosion model divided by the mass of particle), Vp is the particle impact velocity, g
is the impact angle in radians between the approaching particle
Damages in hydro power turbines are mainly caused by cavi- track and the wall, go being the angle of maximum erosion, k1 to k4,
tation problems, sand erosion, material defects and fatigue. In k12 and go are model constants and depend on the particle/wall
general, wear mechanisms can be classified in three categories; material combination.
mechanical, chemical and thermal actions. In hydraulic turbine
mechanical wear of main concern. 3. CFD analysis
There are three types of mechanical wear; abrasive, erosive and
cavitation wear. Abrasive and erosive are due to particles on the Computational Fluid Dynamics (CFD) is one of the branches of
fluid flow, while cavitation is caused by the collapse of bubbles on fluid mechanics that uses numerical methods and algorithms to
the surface. Abrasive wear is defined as the loss of material by the solve and analyze problems that involve fluid flows. The numerical
passage of hard particles over a surface. Erosive wear is caused by solution of NaviereStokes (NS) equations in CFD usually implies a
the impact of particles against a solid surface. Fig. 3 shows the discretization method: it means that derivatives in partial differ-
schematic images of four representative wear modes. ential equations are approximated by algebraic expressions which
CFX solver has two inbuilt erosion model in it, namely: Finnie can be either obtained by means of the finite-difference or the
and Tabakoff model. Tabakoff and Grant Erosion Model [9] was finite-element method or finite-volume method. The result is a set
preferred. Erosion rate E is determined from the following relation: of algebraic equations through which mass, momentum, and en-
h i ergy transport are predicted at discrete points in the domain. The
E ¼ k1 $f ðgÞ$Vp2 $Cos2 ðgÞ 1 R2T þ f ðVPN Þ governing equations are non-linear and coupled; several iterations
of the solution loop must be performed before a converged solution
where is obtained. Millions of calculations are required to simulate the
interaction of fluids and gases with complex surfaces. However,
p 2
2
even with simplified equations and high speed supercomputers,
f ðgÞ ¼ 1 þ k1 k12 sin g only approximate solutions can be achieved in many cases [10].
go
The CFD software includes the package to model the fluid flow
phenomena under the turbulent models. Usually turbulent nu-
RT ¼ 1 k4 Vp sin g
merical simulation consists of two main parts, namely: Direct Nu-
4 merical Simulation (DNS) and Indirect Numerical Simulation (INS).
f ðVPN Þ ¼ k3 Vp sin g DNS has a precise calculated result, but the whole range of spatial
and temporal scales of the turbulence must be resolved which re-
quires a very small time step size. So, this is not suitable for CFD
1:0 if g 2go
k2 ¼ simulation. There are three different types of simulated methods
0:0 if g > 2go
under the Indirect Numerical Simulation which are large eddy
Here, E is the dimensionless mass (mass of eroded wall material simulation (LES), Reynolds-averaged NaviereStokes (RANS) and
Table 1
Rate of erosion.
1 40 0.28
2 80 0.21
3 125 0.27
4 170 0.24
5 215 0.32
6 260 0.30
7 305 0.26
8 350 0.35
Average 0.28
" #
vðrkÞ v v mt vk 0
þ rUj k ¼ mþ þ Pk b rku þ Pkb
vt vxj vxj sk3 vxj
Fig. 4. Head versus mesh elements.
" #
vðruÞ v v mt vu
þ rUj u ¼ mþ
vt vxj vxj su3 vxj
detached eddy simulation (DES). RANS is the oldest and most
common approach to turbulence modeling. The equation of 1 vk vu u
þ ð1 F1 Þ2r þ a3 Pk
Reynolds-averaged NaviereStokes (RANS) is defined as: su2 u vxj vxj k
b3 ru2 þ Pu b
" ! #
DUi vP v vUi vUj The proper transport behavior can be obtained by a limiter to
r ¼ þ m þ ru0i u0j
Dt vXi vXj vxj vxi the formulation of eddy-viscosity:
The left hand side of the equation describes the change in mean a1 k
momentum of fluid element and the right hand side of the equation vt ¼
maxða1 u; SF2 Þ
is the assumption of mean body force and divergence stress. ru;i u;j is
an unknown term and called Reynolds stresses. Due to the aver- where
aging procedure information is lost, which is then feed back into
the equations by turbulence model [9]. vt ¼ mt =r
The shear-stress transport (SST) k-u model [9] was developed by
Menter to effectively blend the robust and accurate formulation of
the k-u model in the near-wall region with the free-stream inde- 3.1. Mesh independent test
pendence of the k-ε model in the far field. To achieve this, Baseline
(BSL) k-w model, which combines advantages of Wilcox k-w and k- Since the results would be numerical approximation using CFX
ε model, is provided with the proper transport behavior to limit the solver, the numerical analysis was done to check the numerical
over prediction of eddy viscosity. The equations are: stability and accuracy of the simulation.
BSL model: Mesh independence test was done for the selection of number
Fig: 5. (a) Test specimen before test. (b) Test specimen after 350 min. (c) Result of CFD analysis d) Runner of JHC [12].
K. Khanal et al. / Renewable Energy 87 (2016) 307e316 311
were compared.
Fig. 6. 3D model obtained in BladeGen. The observation for wear pattern was made with painted sur-
face. After running the apparatus for half an hour, it was observed
of elements in a domain so that the result does not vary signifi- that paint in some location of blade surface was removed. Fig. 5(a)
cantly with increase in the mesh size. This method helps to obtain and (b) shows the wear pattern in test specimen. The painted
minimum number of mesh elements which saves the computation surface has been found to be scratched severely in the outlet region
time without deviating from the accuracy. of blade while some minor scratches have also been observed
Head was chosen as observant parameter. Fig. 4 shows the throughout the blade surface.
relation of head and mesh elements. Since, the change in head is It can be seen that the erosion damage is mostly located in the
not so different for more than 600,000 elements. But for the con- far outlet region near to the edge of the blade. The location of paint
venience and fast computation study was carried out at removal is identical to the pattern of wear observed in the turbines
300,000 mesh elements with an error of only 0.2 m (with respect to operating in real cases. Fig. 5(d) shows the wear observed in the
600,000 elements). runner blades of JHC, which is also observed most severe in the
outlet region of the blades. The wear pattern observed during the
experiment is also quite similar to the pattern which has been
3.2. CFD validation [12] predicted from the CFD analysis.
Table 1 shows the rate of erosion in the specimens for each test.
CFD validation was made by referencing with the experiment The average rate of erosion was found to be 0.28 mg/gm/min, which
performed at Turbine Testing Lab, Kathmandu University, Nepal. A explains that in average, 0.28 mg of material was worn out of 1 g of
test rig called rotating disc apparatus has been developed and test specimen in a minute of operation. It can also be observed that
installed in Turbine Testing Laboratory for carrying out sediment the overall trend of the erosion rate is increasing with the total
erosion test in Francis runner blades. The main purpose of this test duration of the test run.
rig was to compare the wear pattern appearing in test specimens CFD results of JHC and DHP will be compared in result and
with the result of the CFD as well as with the wear pattern observed discussion section so as to confirm the validation of this study.
in turbine operating in real case.
For experiment, hydraulic parameters of Jhimruk Hydropower 3.3. Methodology
Center (JHC) were taken and blade profile of the Francis runner
blade has been modeled. The CFD results, test results and real case In general there are two approaches to runner design; the direct
method and the inverse method. This study uses the direct method,
which begins by setting Q, H and N. Main dimensions such as inlet
diameter, outlet diameter, inlet height, inlet blade angle and outlet
blade angle of runner geometry were obtained by using basic hy-
drodynamic theory [3]. The modeling of the turbine include
creating 2D view in the BladeGen software, which is then converted
into the 3D view by applying the beta distribution. Fig. 6 shows a 3D
model blade obtained in BladeGen.
The 3D Models are created in two ways in two parts:
Table 2
Detail of runner inlet.
Table 3
Detail of runner outlet without swirl.
The hub and shroud curve were chosen arbitrarily and also the
leading edge and trailing edge of blade were made as shown in
Fig. 12. Each of the stream line has different inlet blade angle b1 and
different outlet blade angle b2. Table 2 and Table 3 summarize the
inlet and outlet blade angle at hub, mean and shroud of the blade.
All the dimension of geometry such as leading edge, trailing
edge, hub, shroud, b1 and b2 were kept same but blade angle dis-
tribution (b-distribution) from leading to trailing edge were varied.
Such variation was done to produce 91 blade models.
91 models were simulated and their erosion and efficiency
pattern was observed. Erosion was not the absolute measurement
but only the relative one. The relation of erosion and efficiency with
blade profiles was obtained from these simulations. Also the opti-
mum blade having minimum erosion relative to other blades
without compromising the efficiency was selected and was then
compared to the reference one.
Fig. 11. Efficiency Vs b2.
Fig. 13 shows the relation between erosion and curvature per-
centage for all curvature position. As maximum erosion take place
efficiency of b2 ¼ 14 and b2 ¼ 16 were nearly same and were best at single dot point on the blade and contribute more than 60% of
than others. It can be noticed from the graph that erosion tends to total erosion, so, y-axis is set as total erosion minus maximum
increase as outlet angle increases and efficiency tends to decrease erosion and this compares the blade models effectively. The figure
with increase in outlet angle with some irregularities. By consid- reveals that erosion is relatively high at 35% curvature and higher
ering the manufacturing complexity, b2 ¼ 14 is relatively difficult values of curvature position also has higher erosion. For example,
to manufacture than b2 ¼ 16 . So b2 ¼ 16 was selected as best b2 0.9 curvature position has highest erosion at 35%, 55% curvatures.
whose corresponding b1 ¼ 43 . Although there is no exact regular pattern of erosion according to
Different blade profile of selected outlet angle (b2 ¼ 16 ) was curvature percentage, it can be seen that the trend of erosion de-
produced and then the result was plotted. Fig. 12 shows meridional creases as we shift from negative curvature to zero curvature but
view of Francis runner including the main dimension of the blade positive curvatures has almost same value of erosion with small
used in the simulation. fluctuation for all curvature position. It was observed that the
erosion increased with the increase of curvature towards the
trailing edge; that is at 0.9, 0.85, 0.8 have generally higher erosion
than other curvature positions.
Fig. 14 shows the representative graph for the erosion verses
curvature position. The plot shows that erosion increases as the
curvature is shifted towards trailing edge. It can also be seen that
larger area is eroded when the curvature position is shifted toward
trailing edge as the value of total minus maximum erosion is larger
when the curvature is near trailing edge.
It is necessary to observe the efficiency of the models to deter-
mine the appropriate blade for specific hydro site. Fig. 15 shows the
efficiency verses curvature percentage for all curvature positions.
The graph shows that efficiency at negative curvatures is almost
constant with small deviation but there is significant variation to-
ward positive curvatures. Blade models with curvature position
greater than 0.45 have regular decrease in efficiency as curvature
percentage increases from zero but other curvature position show
some irregularity in efficiency as curvature percentage increases
from zero. The notable point is that the efficiency is higher at
Fig. 12. Meridional view of runner blade showing main dimensions of b2 ¼ 16o. trailing edge (near 0.9 position) for negative curvature percentage
314 K. Khanal et al. / Renewable Energy 87 (2016) 307e316
Fig. 13. Erosion verses Curvature percentage for all curvature position.
Table 4
Erosion and efficiency of reference and optimum blade.
Fig. 15. Efficiency verses Curvature percentage for all Curvature position.
K. Khanal et al. / Renewable Energy 87 (2016) 307e316 315
Fig. 16. Erosion on the pressure and suction side of reference blade respectively.
Fig. 17. Erosion on the pressure and suction side of optimum blade respectively.
optimum one respectively. Red color shows the eroded area on the 3.4.3. Comparison of DHP and JHC [3].
blade. The figures clearly shows that erosion is significantly less in CFD results of this study (DHP) were compared to the referenced
optimized blade compare to the reference one. In optimized blade, JHC results.
pressure side has only some small dots of erosion and has no more Tables 5e7 shows the comparison table of DHP blade (reference
erosion in the suction side. blade) and JHC blade. Although there are some difference in the
input parameters (like shape factors, flow rate), these do not
3.4.2. Blade profile comparison changes the result pattern significantly. So, it can be said that the
Fig. 18 shows blade profile comparison of optimum and refer- rotating disc apparatus experiment performed for JHC blade can be
ence blade. Red color profile is the optimized blade where as black taken as a reference material for DHP blade to confirm CFD results.
color profile is the reference blade.
4. Conclusion and recommendation
Table 5 [3] Biraj Singh Thapa, Mette Eltvik, Kristine Gjosaeter, Ole G. Dahlhaug,
General parameters for CFX-Pre. Bhola Thapa, Chiang Mai, Design optimization of francis runner for sediment
handling, Int. J. Hydropower Dams (2012). Thailand, 1e9.
Parameter DHP JHC [4] Krishna Prasad Shrestha, Bhola Thapa, Ole Gunnar Dahlhaug, Hari P. Neopane,
Biraj Singh Thapa, Innovative design of Francis turbine for sediment laden
Turbulence SST SST
water, in: TIM International Conference, 2012. KU, Nepal.
Flow state Steady Steady
[5] Hari Prasad Neopane, Bhola Thapa, Ole Gunnar Dahlhaug, Particle velocity
Flow type Inviscid Inviscid measurement in swirl flow, laboratory studies, Kathmandu Univ. J. Sci. Eng.
Erosion model Tabakoff Tabakoff Technol. 8 (2012). KU, Nepal.
[6] Biraj Singh Thapa, Amod Panthee, Hari Prasad Neopane, Some application of
computational tools for R&D of hydraulic turbines, Renew. Nepal (2011). KU,
Nepal, 1e5.
Table 6 [7] Peng Guangjie, Wang Zhengwei, Xiao Yexiang, Luo Yongyao, Abrasion pre-
Parameters for CFX-pre sediment data. dictions for Francis turbines based on liquidesolid two-phase fluid simula-
tions, Eng. Fail. Anal. Sci. (2013). China, 327e335.
Data DHP JHC
[8] Koji Kato, Koshi Adachi, Wear Mechanisms, 2001.
Material Quartz Quartz [9] Ansys 14.5 CFX Solver Theory Guide, 2009.
Diameter 0.12 mm 0.1 mm [10] S. Khanna, CFD Analysis of Supercritical Airfoil over Simple Airfoil, 2011.
Shape factor Off 1 Dehradun: s.n.
[11] Gjoaester, Kristine, Hydraulic Design of Francis Turbine Exposed to Sediment
Flow rate 0.08 kg/s 0.07 kg/s
Erosion, 2011. Trondheim: s.n.
[12] H.P. Neopane, B. Rajkarnikar, B.S. Thapa, Development of rotating disc appa-
ratus for test of sediment-induced erosion in Francis runner blades, Wear,
ScienceDirect (2013) s.l. 119e125.
Table 7
Result from CFX-post erosion analysis.