Академический Документы
Профессиональный Документы
Культура Документы
1 Introduction
Nowadays, the Francis turbines are required to operate within a large range of
discharge values, with significant losses in efficiency, as well as increased flow
unsteadiness, as the operating point moves away from the best efficiency one. For
medium-high specific speed hydraulic turbines the draft tube losses are practically
shaping the efficiency hill-chart, as a result of the changes in the ingested swirling
flow. Moreover, at part load the decelerated swirling flow in the discharge cone
develops self-induced instabilities with severe pressure fluctuations that hinder the
turbine operation. In this paper we investigate a century old solution [1,2]
corresponding to a splitter band installed between the crown and band in the runner
bladed region. This solution has been reconsidered in [3,4] for mitigating the interblade vortices or to improve the off-design operation.
Since the draft tube hydrodynamics is essentially influenced by the ingested
swirling flow, we have focused on developing mathematical models for computing the
flow exiting the turbine runner within a range of operating regimes. We started by
considering analytical representations of the axial and circumferential velocity profiles
on a cross-section in the discharge cone [5], then we developed mathematical
models for computing this swirling flow without actually computing the runner [6]. The
improved understanding of the evolution of the decelerated swirling flow led to a
novel approach for mitigating the vortex rope [7] by injecting a water jet from the
crown along the machine axis. Also, this mathematical model was successfully
employed for optimizing the swirling flow at the design operating point for minimum
draft tube losses within a range of operating points [8,9]. Further developments
considered a cross-section as close as possible to the runner blades trailing edge
[10], with a complete mathematical formulation and validation presented in [11]. This
model, summarized in Section 2, is employed in this paper to investigate the
influence of the splitter band on the draft tube hydrodynamics.
Section 3 presents the axial, radial and circumferential velocity profiles at
runner outlet for variable discharge, as well as the procedure for finding the optimum
location of the splitter band. The influence of the splitter band on the swirl exiting the
runner is also presented. In Section 4 we examine the results of the three-
dimensional, turbulent unsteady flow computation in a draft tube, without and with
splitter band, and the paper conclusions are summarized in the Section 5.
line
ban
d
dge
oud
shr
ge
hub
-to-
tra
ilin
lead
band (shroud)
tter
(hu
b)
ing
edge
wn
spli
cro
machine axis
Fig. 1. Meridian cross-section through a Francis turbine runner. The splitter band is displayed only in
the right-hand half-plane.
= di .
a()
+b()
+c()
(1)
d
r
Equation (1) is a second order differential equation with variable coefficients for the
streamfunction () , where is the arc-length coordinate from hub to shroud. All
quantities are made dimensionless with respect to the runner outlet radius Rref , and
the characteristic velocity Vref Rref , with the runner angular speed. For the
Francis turbine model considered in this paper we have Rref = 0.2m
and
1
0.8
0.6
q=0.34
0.31
3726
1.2
=0.73617
1.4
/q
*=
1.6
0.4
0.2
0
0
q=0.22
0.2
0.4
0.8
1
0.6
dimensionless arc-length coordinate
= 0.73167 .
Although we are not aiming in this paper at actually designing the meridian shape of
the splitter band, we still have to choose the design discharge fraction. This is done
by actually choosing a design operating point. For example, as shown in Fig. 2, if we
consider the design point at q = 0.28 , we should have = 0.313726 . In order to
account for the splitter blade in the model underlined in Section 2, we simply add a
constraint on the solution as ( ) = (q / 2) , which is a minor modification of the
finite element algorithm presented in [11].
Figure 3 shows the axial, radial and circumferential velocity profiles computed
without (dashed lines) and with (solid lines) splitter band at three discharge values.
At the design operating point, q = 0.28 , the splitter band does not alter the velocity
profiles, Fig. 3b. At part load, Fig. 3a, the splitter band accelerates the meridian flow
near the crown, with the benefit of reducing the size of the central quasi-stagnant
region that develops when the flow is further decelerated in the discharge cone. At
full load, Fig. 3c, the splitter band has an opposite effect, by decelerating the
meridian flow near the crown. Obviously, the splitter band also introduces a jump in
the axial and circumferential velocity profiles, while the jump in radial velocity is
negligible thanks to the negligible radial velocity at the band location.
0.1
0
-0.1
-0.2
0
axial velocity
radial velocity
swirl velocity
0.2
0.4
0.8
1
0.6
dimensionless arc-length coordinate
0.2
0.1
0
-0.1
-0.2
0
axial velocity
radial velocity
swirl velocity
0.2
0.4
0.8
1
0.6
dimensionless arc-length coordinate
velocity components
0.2
velocity components
velocity components
0.3
0.3
0.3
axial velocity
radial velocity
swirl velocity
0.2
0.1
0
-0.1
-0.2
0
0.2
0.4
0.8
1
0.6
dimensionless arc-length coordinate
a) q = 0.22
b) q = 0.28
c) q = 0.34
Fig. 3. Axial, radial and circumferential velocity profiles on the hub-to-shroud line from Fig. 1.
One can notice from Fig. 3 that by altering the meridian velocity, the splitter band also
produces a change in the circumferential velocity since the direction of the relative
flow remains the same. As a result, one can expect a change in the flux-averaged
moment of momentum UVu. Indeed, this is confirmed in Fig. 4 which shows the
dimensionless flux averaged moment of momentum versus the dimensionless
discharge, without and with splitter band. At part load, while keeping the same
turbine head the discharge is increased by the splitter band for the same guide vane
opening. On the other hand, at full load the splitter band decreases the overall
discharge for the same guide vane opening and turbine head. These preliminary
conclusions should be quantitatively assessed with numerical simulation of the
turbine flow.
0.2
0.15
0.1
0.05
0
0.2
0.34
velocity profiles computed on the hub-to-shroud line, as shown in Fig. 3. The model
draft tube used for the present investigations is the same as the one in [10], and it is
shown in Fig. 5. The draft tube losses are evaluated as the difference between the
inlet and outlet flux averaged total pressure. Fig. 6 shows the dimensionless losses
(with respect to Vref2 ) versus the discharge coefficient. The splitter band reduces the
losses practically only at full load, up to 30%. At part load the hydraulic losses are
practically unaffected by the splitter band.
0.03
0.02
0.015
0.01
0.005
0
0.025
0.22
0.24
0.3
0.26 0.28
discharge coefficient
0.32
0.34
On the other hand, on Fig. 6 we have figured the standard deviation of the
losses as a measure of the plunging pressure fluctuations. One can see that these
fluctuations are rather large at part load without the splitter band. The splitter band
effectively mitigates the plunging fluctuations at part load. This is clearly shown in
Fig. 7, where the losses fluctuation practically vanishes thanks to the splitter band.
Note that we consider the fluctuations in Fig. 7 as being of plunging type because the
losses are computed as the difference between flux averaged total pressure values
on inlet/outlet sections.
dimensionless losses [-]
0.04
q = 0.22
0.03
with splitter band
0.02
0.01
0
0
q = 0.24
0.5
1
flow time [s]
1.5
Fig. 7. Fluctuations of hydraulic losses without (dashed lines) and with (solid lines) splitter band.
The plunging fluctuations shown in Fig. 7 originate from the interaction of the vortex
rope with the elbow of the draft tube. This is clear from Fig. 8, where the vortex rope
is shown as a vortex filament extracted from the velocity field using the eigenvalues
of the velocity gradient tensor [12]. Figs. 8a) and 8c) display the vortex filament
development without splitter band, revealing the strong interaction with the elbow
which results in the plunging fluctuations displayed with dashed lines in Fig. 7.
When the splitter band is present, Figs. 8b) and 8d), the vortex rope is
significantly shortened and as a result its interaction with the elbow is weak, with the
benefit of effective mitigation of the plunging fluctuations as shown with solid lines in
Fig. 7. One can also notice in Figs. 8b) and 8d) the vorticity on the inlet section
associated with the jump in velocity components due to the splitter band.
Fig. 8. Vortex rope filament in the discharge cone and elbow of the draft tube
5 Conclusions
We investigate the influence of a splitter band in the runner bladed region on the draft
tube flow, using a numerical demonstrator. The model for computing the swirling flow
at the runner outlet, on an arbitrary hub-to-shroud line close to the blades trailing
edge, is adapted to account for the presence of a splitter band.
Basically, the splitter band is enforcing the discharge fraction (between the
crown and the splitter blade, or between the splitter band and runner band) constant
while the overall turbine discharge is changing. As a result, the meridian flow near the
crown is accelerated at part load and decelerated at full load, mitigating the wake-like
and jet-like velocity profiles, respectively, further downstream. However, the splitter
band introduces a jump in the velocity profiles at the runner outlet. The
circumferential velocity follows from the relative flow kinematics at the runner outlet
(corresponding to the blade angles). We notice that at the same discharge the splitter
band changes the average moment of momentum at the runner outlet, thus a slight
shift in the operating point is expected.
We propose a procedure for determining the optimum location of the splitter
band. In short, the splitter band intersects the trailing edge, in the meridian halfplane, where the discharge fraction is most influenced by the turbine discharge
variation.A numerical assessment of the draft tube flow is performed using threedimensional, unsteady turbulent flow simulations. The splitter band is reducing the
hydraulic losses at full load, while at part load its influence is negligible. However, we
notice that at part load the splitter band practically eliminates the plunging
fluctuations that originate from the interaction of the vortex rope with the draft tube
elbow. This unexpected feature is worth further investigations.
No experimental data were available to the authors to validate the present
results. This is the reason for introducing this work as a preliminary numerical
demonstrator for the splitter band.
References
[1]