IOP Conference Series: Earth and Environmental Science

PAPER • OPEN ACCESS You may also like

- Numerical simulation of air injection in part
Numerical simulation of air injection in Francis load operating point
P Shcherbakov, D Chirkov, V Skorospelov
turbine et al.

- Detection and analysis of part load and full

load instabilities in a real Francis turbine
To cite this article: D Chirkov et al 2019 IOP Conf. Ser.: Earth Environ. Sci. 240 022043 prototype
Alexandre Presas, David Valentin, Eduard
Egusquiza et al.

- Overview of verification tests on AC loss,

contact resistance and mechanical
View the article online for updates and enhancements. properties of ITER conductors with
transverse loading up to 30 000 cycles
K A Yagotintsev, W A J Wessel, A Vostner
et al.

This content was downloaded from IP address 46.2.234.154 on 17/09/2023 at 21:56

29th IAHR Symposium on Hydraulic Machinery and Systems IOP Publishing
IOP Conf. Series: Earth and Environmental Science 240 (2019) 022043 doi:10.1088/1755-1315/240/2/022043

Numerical simulation of air injection in Francis turbine

D Chirkov1, P Scherbakov1, V Skorospelov2, S Cherny1, A Zakharov3

1
Institute of Computational Technologies SB RAS, Novosibirsk, Russia
2
Sobolev Institute of Mathematics SB RAS, Novosibirsk, Russia
3
OJSC "Power Machines" LMZ, St-Petersburg, Russia

chirkov@ict.nsc.ru

Abstract. Operation of Francis turbines in part load and sometimes in full load operating
points is associated with increased pressure pulsations. One of the practical ways to reduce
these pulsations is to add atmospheric air into the flow. It was shown previously that air
injection/admission through the center of the runner cone can significantly reduce the
amplitude of pressure pulsations both in part load and in full load operating points. Up to now
the effect of flow aeration has been investigated mainly experimentally. In the present paper
we performed CFD simulations of this phenomenon. Computations have been carried out in
frames of homogeneous three-phase “liquid – vapor – non-condensable gas” mixture model.
Air phase has been assumed incompressible. Both part load and full load operating points have
been considered with different air flow rates. Computations have shown that even at small flow
rate the air changes the structure of the swirling flow downstream the runner and considerably
reduces pressure pulsations, caused by vortex rope rotation. The obtained results are in
agreement with corresponding experimental data for model turbine.

1. Introduction
It is well known, that operation of Francis turbines in part load and sometimes in full load operating
points is associated with increased level of pressure pulsations. One of the most effective practical
solution to reduce these pulsations consists in addition of atmospheric air into the turbine flow passage
[1]. Different methods of aeration have been suggested in the literature, including aeration through the
runner shaft, through blade trailing edges, draft tube (DT) peripheral aeration, and aeration in the
vaneless space between the guide vanes and the runner [2, 3]. Air is either injected into the flow
passage by a compressor (forced air injection) or sucked naturally through a special valve when the
pressure in the flow drops below the specified level (natural air admission). Aeration through the
center of the runner shaft proved to be the most effective to suppress the vortex rope pulsations in part
load [4]. In [1, 5] it was shown that air injection/admission can also reduce power oscillations and unit
vibrations in high load operating points.
Injected air affects the pressure pulsations in different ways. First, the airflow changes pressure and
velocity field, increasing the pressure and thus reducing cavitation in the center of draft tube cone.
Second, it alters the cavity compliance C Vc / H of the gaseous cavity of volume Vc including
water vapor and air. It should be noted, that C is one of the key parameters determining the stability
and the frequency of the natural oscillations of the power plant [6].
Despite recent progress in development of incompressible and multiphase CFD methods, as well as
increased computer power, there are only few papers devoted to numerical simulation of air injection

Content from this work may be used under the terms of the Creative Commons Attribution 3.0 licence. Any further distribution
of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI.
Published under licence by IOP Publishing Ltd 1
29th IAHR Symposium on Hydraulic Machinery and Systems IOP Publishing
IOP Conf. Series: Earth and Environmental Science 240 (2019) 022043 doi:10.1088/1755-1315/240/2/022043

in hydraulic turbines [7, 8, 9]. These papers adopted different mathematical models for simulation of
the air transport. However all these papers considered 2-phase “water-air” mixture and did not take
into account cavitation. It is known that cavitation plays an important role in the development of
pressure pulsations, especially in full load operating points.
In [10, 11] for the first time we tried to simulate the effect of air injection on self-excited
oscillations in full load. For that we considered a 3-phase “water-vapor-air” mixture model for the
turbine domain and attached a 1D hydro-acoustic model of the penstock, as suggested in [12].
Computations were performed for a prototype turbine with gravity acceleration taken into account. It
was shown that even small air flow rate of about 0.4% of water flow rate completely eliminates
cavitation in the DT, the source of self-excited oscillations, and thus stabilizes the turbine operation.
These results are briefly shown in figure 1.

a b c
Figure 1. The effect of air injection on full load operation of prototype turbine.
a – vapor volume fraction with no air injection, b – axial velocity with air injection Qair=0.4%,
c – efficiency and pressure pulsation as function of air flow rate.
In the present paper the above 3-phase model is applied for simulation of forced air injection in a
scale model turbine with specific speed nq=73.6, figure 2. The computed results are compared to
experimental measurements, performed in the Laboratory of Hydraulic Turbines, “Power Machines”
LMZ (St-Petersburg).

2. Governing equations
Cavitating flow in turbine flow passage is described as isothermal homogeneous three-phase mixture,
consisting of water, vapour, and non-condensable gas (air). It is assumed that all the phases share
common pressure and velocity fields. Distribution of liquid volume fraction L is governed by
transport equation with source terms, responsible for evaporation and condensation. The convection of
a gas volume fraction G is governed by the same transport equation with zero source term [13]:

div( v) 0, (1)
t
v
div( v v) pˆ div( τ) f, (2)
t
1
L
div( L v) (m m ), (3)
t L

G
div( G v) 0. (4)
t
Here L L G G (1 L G) V is the mixture density; v is the velocity vector;
pˆ p 2 k / 3 ; p is the static pressure; k is the turbulence kinetic energy. Absolute reference frame is
used for static components of turbine flow passage, while rotating reference frame is used for the

2
29th IAHR Symposium on Hydraulic Machinery and Systems IOP Publishing
IOP Conf. Series: Earth and Environmental Science 240 (2019) 022043 doi:10.1088/1755-1315/240/2/022043

runner, rotating with angular velocity around the Oz axis. Thus for runner sub-domain
f ( x1 2
2u2 , x2 2
2u1 , g ) . Gravity acceleration g is necessary to account the buoyancy of the
air phase. In (2) τ is the tensor of viscous stresses. Dynamic viscosity of the three phase mixture is
computed as
L L G G (1 L G) V ,

where L , G , V are the dynamic viscosity coefficients of liquid, gas (air) and vapour, respectively.
Kim-Chen k- [14] and SST k- turbulence models with log-law wall functions were used to evaluate
k and turbulent viscosity μT needed to close the mean flow equations (1)-(4).
Condensation (m+) and evaporation (m–) terms in (3) are evaluated using the model of Zwart,
Gerber, and Belamri [15]:
2 max[0, p pV ] , 2 max[0, pV p]
m C prod (1 L G) V m Cdest L V , (5)
3 L 3 L

where C prod 3 104 , Cdest 7.5 104 .

According to (4) air is treated as incompressible fluid with constant density G < L. It should be
noted that since all phases share the same velocity, the present model is unable to describe the effect of
separation of premixed phases.
Boundary conditions are total specific flow energy fixed in the inlet cross section and total specific
flow energy kept in the DT outlet. With these inlet/outlet conditions discharge Q is not known a priory
and is found iteratively in the process of computation.
In case of air injection, constant volumetric air flow rate is specified in the DT inlet section, in cells
where R Rv , see figure 3.

3. Numerical method
Governing equations of the above three-phase model are solved numerically using our in-house
developed CADRUN solver, based on finite volume artificial compressibility approach. Dual time
stepping is used for unsteady calculations. In pseudotime equations are marched using implicit finite
volume scheme. Third order accurate MUSCL scheme is used for discretization of inviscid fluxes
through cell faces. In order to prevent numerical oscillations, artificial dissipation is added on liquid-
gas interfaces, as suggested in [13]. Second order backward scheme is applied for physical time
derivatives. Linearized system of discrete equations is solved using LU-SGS iterations. All equations
(1)-(4) are solved for (p,v, L, G) in a coupled manner. Details of the solver can be found in [11].
Periodic stage approach is used for turbine flow analysis, requiring computations only in one
wicket gate (WG) channel, one runner channel, and the whole draft tube. Mixing plane boundary
condition is applied on “WG – runner” and “runner – DT” interfaces with circumferential averaging of
flow variables (p, Cr, Cu, Cz, L, G, k, , where Cr, Cu, Cz are the radial, circumferential and axial
components of the velocity vector, respectively.

4. Simulation results
Computations were performed for scale model turbine (runner diameter D1=0.46m), operated at
H=21.077m, unit speed n11=71.32. Computational domain is shown in figure 3. It consists of one
guide vane channel, one runner channel, and the whole draft tube. Basic block structured mesh
consists of 260’000 cells, see figure 3. The value of y+ on the solid walls is in the range 15 to 230. The
time step t is taken equivalent to 7.5 degrees of runner rotation. Total flow energy in the draft tube
outlet is set corresponding to Thoma number of the scale model tests.

3
29th IAHR Symposium on Hydraulic Machinery and Systems IOP Publishing
IOP Conf. Series: Earth and Environmental Science 240 (2019) 022043 doi:10.1088/1755-1315/240/2/022043

Figure 2. Monitoring points in the draft tube. Figure 3. Computational domain and mesh.

4.1. Part load operating point

Part load operating point corresponds to guide vane opening (GVO) a0=20mm. First, two-phase
cavitational liquid-vapor computations were performed with no air injection. Figure 4 shows pressure
pulsations in the DT points D3 and D4, obtained using Kim-Chen turbulence model. It can be seen
that the peak-to-peak pressure pulsation in point D4 is higher than that in D3. Figure 5 shows that both
Kim-Chen and SST turbulence models gave similar frequency and amplitude of pressure pulsations,
that are in good agreement with experimental data. However in case of SST model no cavitation was
observed in the core of the vortex rope, figure 6.
Then a series of computations was carried out for different volumetric air flow rates, ranging from
Qair 0.25% to Qair 2% of nominal turbine discharge. Figure 7 compares frequency and amplitude
of pressure pulsations in point D3, obtained in computation and experiment for the case Qair 0.5%
(see also figure 9 for peak-to-peak values). Experimental data clearly indicate that air injection with
Qair 0.5% reduces the amplitude of pressure pulsations. At the same time the frequency of pulsations
f / f n shifts from 0.22 (no air) to approximately 0.3 in case of air injection. As for numerical
simulation, the result, obtained in the basic mesh, depends on the turbulence model used. In case of
Kim-Chen model, the helical vortex rope almost completely disappeared. In case of SST simulation
with air injection the vortex rope remained. However, the amplitude and frequency of the pulsations
did not alter significantly. Figure 8 shows that air flow rate has a small influence on pressure
pulsations. Figure 9, left shows the aerated vortex rope, obtained using SST model, visualized as iso-
surface of the air volume fraction G 0.1 . Figure 9, right shows the evolution of air flow rate in the
DT inlet and elbow outlet cross sections. The frequency of air flow rate pulsations corresponds to
vortex rotation frequency.

Figure 4. Pressure pulsations in points D3 and D4. No air injection. Kim-Chen model.

4
29th IAHR Symposium on Hydraulic Machinery and Systems IOP Publishing
IOP Conf. Series: Earth and Environmental Science 240 (2019) 022043 doi:10.1088/1755-1315/240/2/022043

Kim-Chen SST
Figure 5. Spectrum of pressure pulsations in point D3. No air injection.

Kim-Chen SST
Figure 6. Vortex rope in the DT. No air injection.

Kim-Chen SST
Figure 7. Spectrum of pressure pulsations in point D3. Air flow rate Qair=0.5%.

5
29th IAHR Symposium on Hydraulic Machinery and Systems IOP Publishing
IOP Conf. Series: Earth and Environmental Science 240 (2019) 022043 doi:10.1088/1755-1315/240/2/022043

Figure 8. The influence of air flow rate on pressure pulsation in point D3. SST model.

Figure 9. Air flow rate Qair=0.5%. SST turbulence model. Left: iso-surface of air volume fraction
G 0.1 . Right: air flow rate in DT inlet and elbow outlet cross sections.

Figure 10. Peak-to-peak pressure pulsations in point D3 as function of GVO.

4.2. Full load operating point

Figure 11 shows the computed pressure pulsations in the DT (point D3) for full load operating point
(GVO a0=48mm). Both “no air” and “air injection” cases are shown. It can be seen that Kim-Chen and
SST models give similar results for both cases. In case of air injection the amplitude of pressure
pulsations decreased more than twice in agreement with experimental data, see figure 10. The reason
of this reduction is the elimination of cavitation in the DT with aeration. In case of no aeration a
cavitation cavity is observed in the centre of the drat tube just below the runner (figure 12, left). This
cavity is proved to be the source of synchronous pressure pulsations of frequency 0.08fn, shown in
figure 11 (left), by analogy with self-excited oscillations in prototype turbines, investigated in [11,12].
In case of air injection the vapour cavity in the DT is completely eliminated. Figure 12, right shows
the instantaneous vortex-like distribution of the air volume fraction in the DT.

6
29th IAHR Symposium on Hydraulic Machinery and Systems IOP Publishing
IOP Conf. Series: Earth and Environmental Science 240 (2019) 022043 doi:10.1088/1755-1315/240/2/022043

No air Qair=0.5%
Figure 11. Computed pressure pulsations in the DT.

No air Qair=0.5%
Figure 12. Gaseous cavities in the DT. Full load operating point, SST model. Left: no air, vapor cavity
is shown as iso-surface V 0.5 . Right: air injection, air cavity is shown as iso-surface of G 0.1 .

5. Discussion and conclusion

Three phase “liquid – vapor – non-condensable gas” RANS model is used for numerical simulation of
air injection in model Francis turbine. A homogeneous mixture model is used meaning that all phases
share the same velocity and pressure field. The air phase is assumed incompressible, and distribution
of its volume fraction is governed by a simple transport equation. Kim-Chen k- and SST k- models
are used for turbulence. The model is applied for simulation of turbine aeration though the central hole
in the runner shaft. Both part load and full load operating points are considered. Draft tube pressure
pulsations obtained in “warer – vapor” and “warer – vapor – air” simulations are compared to
experimental data.
Computations show that air injection at flow-rate Qair =0.5% effectively eliminates cavitation in the
draft tube, both in part load and full load operating points. In full load operating point air injection
significantly reduced pressure pulsation (more than twice), in good agreement with experimental data.
In part load the situation is more difficult. For the case of SST model the air injection does not take
а significant effect on the frequency and amplitude of pressure pulsations. Moreover, the amplitude is
even a little bit larger than for no-aeration case, and does not depend on the air flow rate, in contrary to
experimental data. Similar result was obtained in [7]. The case of Kim-Chen turbulence model needs
further investigation, since in the present computations air injection suppressed the formation of a
rotating helical vortex rope, and thus eliminated the associated pressure pulsations.

7
29th IAHR Symposium on Hydraulic Machinery and Systems IOP Publishing
IOP Conf. Series: Earth and Environmental Science 240 (2019) 022043 doi:10.1088/1755-1315/240/2/022043

Cavitation in the draft tube can be the source of pressure pulsations (as in the case of full load
instability). In this case aeration through the runner cone removes cavitation, and as the result, reduces
the pressure pulsations. This effect is well captured by the present 3-phase model.
In part load pressure pulsations are caused by the rotation of the helical vortex rope. Usually the
amplitude of these pressure pulsations does not depend significantly on Thoma number. From the
other hand, flow aeration considerably reduces the pulsations. Therefore it seems that aeration affects
the flow-field in a more complex way than simply to prevent cavitation in the vortex core. Evidently,
this mechanism is still not captured by the present 3-phase numerical model. Accurate simulation of
air injection in part load operating point requires some issues that should be refined.
First, it should be noted that the absolute value of pressure pulsations both with and without air
injection is lower than experimentally measured. One of the reasons is the lack of grid resolution in the
draft tube. The other reason is the use of periodic stage approach for the runner, requiring
circumferential averaging of flow parameters on the mixing plane in draft tube inlet section. Ideally,
full 360° runner domain should be used for vortex rope computations. Then, in the present
computations the hydro-acoustics of the water conduit of the test rig was not taken into account. In
order to check its influence a 1D-3D computation should be carried out, similar to [12]. Another
important improvement is to account the compressibility of the air phase. These are the topics of
ongoing research.

Nomenclature
a0 Model guide vane opening [mm] Vc Cavity volume [m3]
C Cavity compliance [m2] L Liquid volume fraction [-]
E Total specific energy [m] G Air volume fraction [-]
g Gravity acceleration [m/s2] ρ Mixture density [kg/m3]
H Net head [m] ρL Water density [kg/m3]
pV Vapor pressure [Pa] ρV Vapor density [kg/m3]
Q Discharge [m3/s] ρG Air density [kg/m3]

References
[1] Dörfler P, Sick M and Coutu A 2013 Flow-induced pulsation and vibration in hydroelectric
machinery (London: Springer-Verlag)
[2] Dorfler P 1986 Design criteria for air admission systems in Francis turbines Proc. of IAHR
Symp. on Hydraulic Machinery and Systems (Montreal, Canada)
[3] March P 2011 Hydraulic and environmental performance of aerating turbine technologies
Proc. of EPRI/DOE Conference on Environmentally-Enhanced Hydropower Turbines (Palo
Alto, California)
[4] Papillon B, Kirejczyk J and Sabourin M 2000 Atmospheric air admission in hydro turbines
Proc. of Hydrovision paper 3C
[5] Türkmenoglu V 2013 The vortex effect of Francis turbine in electric power generation Turkish
Journal of Electrical Engineering and Computer Sciences 21 pp 26–37
[6] Nicolet C 2007 Hydroacoustic modelling and numerical simulation of unsteady operation of
hydroelectric systems EPFL Ph. D. Thesis No. 3751.
[7] Yu A, Luo X and Ji B 2015 Numerical simulation and analysis of the internal flow in a Francis
turbine with air admission IOP Conference Series: Materials Science and Engineering 72
042047
[8] Dekterev A A and Dekterev D A 2016 Computational investigation of the effect of air injection
into the fluid flow with precessing vortex core Journal of Siberian Federal University:
Engineering & Technologies 9 pp 1110–1119 (in Russian)
[9] Pitorac L-I 2017 Numerical simulation of air injection in Francis turbines / NTNU Master thesis
[10] Chirkov D, Scherbakov P, Cherny S, Zakharov A, Skorospelov V and Turuk P 2016 Mitigation
of self-excited oscillations at full load: CFD analysis of air admission and effects of runner

8
29th IAHR Symposium on Hydraulic Machinery and Systems IOP Publishing
IOP Conf. Series: Earth and Environmental Science 240 (2019) 022043 doi:10.1088/1755-1315/240/2/022043

design Proc. of 28th IAHR Symposium on Hydraulic Machinery and Systems (Grenoble,
France)
[11] Chirkov D V, Shcherbakov P K, Cherny S G, Skorospelov V A and Turuk P A 2017 Numerical
investigation of the air injection effect on the cavitating flow in Francis hydro turbine
Thermophysics and Aeromechanics 24 pp 691–703
[12] Chirkov D, Avdyushenko A, Panov L, Bannikov D, Cherny S, Skorospelov V and Pylev I 2012
CFD simulation of pressure and discharge surge in Francis turbine at off-design conditions
Proc. of IAHR Symp. on Hydraulic Machinery and Systems (Beijing, China)
[13] Kunz R F, Boger D A, Stinebring D A et al 2000 A preconditioned Navier-Stokes method for
two-phase flows with application to cavitation prediction Computers & Fluids 29 pp 849–75
[14] Chen Y S and Kim S W 1987 Computation of turbulent flows using an extended k- turbulence
closure model. NASA CR-179204.
[15] Zwart P J, Gerber A G and Belamri T A 2004 Two-phase flow model for predicting cavitation
dynamics Proc. of ICMF 2004 International Conference on Multiphase Flow (Yokohama,
Japan) Paper No. 152