Sie sind auf Seite 1von 12

A benckmark test

case for swirling


flows
design of the swirl apparatus,
experimental data and numerical
challenges

Romeo SUSAN-RESIGA
Constantin POPESCU IAHR2018-
Introduction (our experience)
 For the past decade we focused on the
fundamentals of turbomachinery swirling flows, with
emphasis on self-induced instabilities in the
discharge cone;
 We have extensively explored a jet injection
approach to mitigate the vortex rope and improve
the stability of the decelerated swirling flows.
 A swirl apparatus has been used to mimic the swirl
exiting a Francis runner at partial discharge: use
non-rotating guide vanes to set the swirl level, then
re-distribute the swirl from hub to shroud using a
rotor at runaway speed.
 Although a surrogate for the real Francis turbine
model, the swirl apparatus is still has intrinsic
complexity in terms of rotor-stator interaction
  build a simpler experimental setup, with a more
controlablehttps://openfoamwiki.net/index.php/Sig_Turbomachinery_/_Timisoara_Sw
environment for investigating swirling
Introduction (motivation & goals)
 Preliminary design and optimization of hydraulic turbomachines
(including the assessment of novel technologies for flexible,
efficient and smooth operation) still benefit from the
axisymmetric swirling flow models (so-called through-flow
approach)
 Even if ultimate accuracy is not the selling point of through-flow
methods, robustness must be insured for reliable use.

 We introduce a novel swirl apparatus designed, built and tested


at Politehnica University Timișoara;
 Simple yet not trivial geometry for the swirl generator (easy
to model);
 Simple diffuser (step cylindrical diffuser);
 LDV and pressure measurements;
 Modular construction, easy to change the configuration;
 Geometry and experimental data available upon request.
SWAP design (meridian profiling)
 Swirl Apparatus (SWAP) has a swirl generator and a test section;
 The swirl generator has a convergent annular section (Ø 150 mm  Ø
100 mm )
 The hub shape is computed for a constant geometrical flow cross-
section
 Blades
80 (lean flat plates)
63.50 mm are provided in the convergent regions
bladed region blade ring
radial coordinate [mm]

60

44.64 mm

40
swirling flow
20

0
-60 -40 -20 0 20 40 60 80 100
axial coordinate [mm]
SWAP design (meridian profiling)
 Swirl Apparatus (SWAP) has a swirl generator and a test section;
 The test section is modular, with interchangeable cylindrical sections
of Ø 100 mm or Ø 120 mm, discharging in a pipe of Ø 160 mm.
 We opted for a step diffuser (piecewise cylindrical) configuration
instead of a conical difusser.
 Each cylindrical section has optical window for LDV measurements
and three unsteady pressure sensors.
SWAP design (swirl generator
blades)
 Swirl Apparatus (SWAP) has a swirl generator and a test section;
 The swirl generator has straight leaned blades; 36 blades, 1 mm
thick.
 Swirl is generated by inducing a radial velocity component through
contraction in the bladed region, thanks to the flow tangency
condition.

R* = 39.46
mm
Rhub = 55.6
mm
λhub = 45°
Rpipe = 75.0
mm
λpipe = 31.87°
SWAP design (ensemble)
 Swirl Apparatus (SWAP) has a swirl generator
and a test section;
 LDV system and wall unsteady pressure
sensors.
 Typical flow discharge: 30 l/sec.
 Geometry available on demand.
Through-flow model
(axisymmetric turbomachinery swirling flow
 The computation)
through-flow model remains the backbone of preliminary turbomachinery design
and analysis, including bladed and bladeless regions.
 A blade body force is added to the momentum eqs. to account for the flow turning in
the bladed region; suitable for large number of blades, valid in the limit of infinite
number of infinitely thin blades.
Through-flow model
(FLUENT implementation with User Defined
 The Function)
through-flow model is implemented in the FLUENT turbulent axisymmetric swirling
flow solver.
 Use the porous flow assumption in the bladed region to account for the blade
blockage (thickness).
 User Defined Function (UDF) is used to include the blade body force into the FLUENT
solver.
DEFINE_SOURCE(blade_body_force,c,t,dS,eqn) DEFINE_PROFILE(blockage,t,iv)

 We
{
found that the standard k-ε modelDELTA;
real x[ND_ND],R,DENS,TANLEAN,VA,VR,DVRDZ,DVRDR,BBFA,BETA, with enhanced
{
/* porositywall
of the function
media is definedgives the
as the ratio of best

agreement with/*measurements
get cell centroid radius */ (other models were
DELTA=NB*TN/TWOPI/RT; /* blade blockage parameter */
C_CENTROID(x,c,t); R=x[1]; not found
/* Thread *t - pointerrobust for
to thread, int the index */
the volume occupied by the fluid to the total volume */
iv - variable

axisymmetric
VA = C_U(c,t); /* get axialswirling
velocity */ flow solver; bugs!?)
TANLEAN=RT/sqrt(R*R-RT*RT); BETA=1.0-DELTA*TANLEAN; cell_t c; /* cell index */
real x[ND_ND]; /* cell centroid coordinates */
VR = C_V(c,t); /* get radial velocity */ real R; /* cell centroid radius */
DVRDZ = C_V_G(c,t)[0]; /* dVr/dZ */ begin_c_loop(c,t)
DVRDR = C_V_G(c,t)[1]; /* dVr/dR */ {
BBFA = VA*DVRDZ+VR*DVRDR-VR*VR/R*TANLEAN*TANLEAN; C_CENTROID(x,c,t); /* get cell centroid coordinates */
DENS = C_R(c,t); /* get fluid density */ R=x[1]; /* get cell centroid radius */
BBFA = -DENS*BBFA*TANLEAN; C_PROFILE(c,t,iv)=1.0-NB*TN/TWOPI/sqrt(R*R-RT*RT);
dS[eqn] = 0.0; /* explicit implementation */ }
if(eqn==EQ_Z_MOM){return BBFA;} /* circumferential bbf */ end_c_loop(c,t)
if(eqn==EQ_Y_MOM){return BBFA*TANLEAN;} /* radial bbf */ }
}
Through-flow model (numerical
results)
 FLUENT 16.2 turbulent axisymmetric swirling flow solver, standard k-ε model, porosity
 2D problem in the meridian half-plane (computationally tractable for parametric
studies)
 Discretization with 85631 cells, check refinement for 342524 cells.
2
tan(l), Eq. (1)
-Vq/Vr, numerical
1.5

0.5

0
0.045 0.05 0.055 0.06
radius [m]

Check that the blade body force


correctly enforces the flow Meridian flow field (Stokes’ streamlines and total
tangency condition Vθ=-Vr tan(λ); pressure map) for axisymmetric flow without swirl
The axial flow is not kinematically (lower half plane) and axisymmetric swirling flow
constrained by the straight leaned (upper half plane).
blades. Swirl improves flow attachment both at hub and
Through-flow model
(comparison with experimental data)
 Axial and circumferential velocity LDV
measurements in two survey sections of Ø 100 mm
and Ø 120 mm, respectively.

6 6

4 4
velocity components [m/s]

velocity components [m/s]


2 2

0 0

-2 -2
axial velocity, experimental axial velocity, experimental
circumferential velocity, experimental circumferential velocity, experimental
-4 -4
axial velocity, computed axial velocity, computed
circumferential velocity, computed circumferential velocity, computed
-6 -6
-0.04 -0.02 0 0.02 0.04 -0.06 -0.04 -0.02 0 0.02 0.04 0.06
radial coordinate [m] radial coordinate [m]
Conclusions
 The new swirl apparatus has a simple geometry as
o Simple swirl generator with straight leaned blades in
the convergent region
o Simple test section, with piecewise cylindrical step
diffuser
 The setup provides an benchmark case for
testing/assessing simplified flow models useful in
turbomachinery hydrodynamics
 A simple example of through-flow model derivation and
implementation is provided (can also be used for
didactical purposes before tackling complicated radial-
axial blading)
 The turbulence model is the “Achilles heel” for 2D
axisymmetric swirling flow computations, particularly with
strong swirl, separation, recirculation, etc. The present
results suggest that only the standard k-ε model is robust
and consistent with measured velocity profiles.

Das könnte Ihnen auch gefallen