Hydraulic turbine distributor simulation

using OpenFOAM

Fifth OpenFOAM Workshop, June 21-24 2010, Gothenburg, Sweden

1

F.Guibault, C. Devals and J-F DubéÉcole Polytechnique de Montréal, Canada

T. Vu and B. NennemanAndritz Hydro, Pointe-Claire, Canada

Context

Viscous flow simulations for hydraulic turbomachinery design

Spiral

Stator

Rotor

2

Runner

Draft-tube

Spiral casing

Distributor

Objectives

• Validate OpenFOAM RANS simulations for distributor blade passage in tandem mode on hybrid meshes

• Validate periodic boundary conditions• Predict torque on guide vanes for the full range of blade

openings• Compare torque prediction with experimental data

3

• Compare torque prediction with experimental data

Model turbine test – Medium head Francis turbine

� Throat diameter: Dth = 0.35 m� Test case: 24 wicket gates, 24 stay vanes� Casing type: piguet� Tested turbine head: H = 30 m� Wicket gate opening angles: 14 deg, 20 deg, 25 deg, 30 deg and 42 deg.� Distributor inflow angles: 35 ± 5deg

Test case description

4

� Steady state Reynolds averaged Navier-Stokes equations

Physical properties and Models

1kg.m997 −=ρ

turblam

T

U

UUIp

UU

ννν

νρ

+=

=⋅∇

∇+∇+−⋅∇=⊗⋅∇

0

))(()(

r

rrrr

5

�Newtonian fluid

� steady-state case (5000 iterations)

� turbulence model

126

1

10.89257.0

kg.m997

−−

−

==

=

smlamρ

µν

ρ

ε−k

Outlet

cyclicGgi

empty

wall

Boundary conditions

Inlet velocity profile

� Constant cylindrical values →

implemented using profile1DfixedValuePeriodicity boundary conditions

� Implemented using GGI (General Grid

Interface)

Turbulence boundary conditions

� Turbulent kinetic energy intensity and

6

Inlet

cyclicGgi

empty

� Turbulent kinetic energy intensity and

mixing length

Boundary conditions

[Name]Profil radial

[Spatial Fields]R

[Data]R [ m ], Velocity Axial [ m s^-1 ], Velocity Radial [ m s^-1 ], Velocity Circumferential [ m s^-1 ]0.0000000, 0.0, -1.959806, 3.3944840.2865129, 0.0, -1.959806, 3.3944840.2865130, 0.0, -1.959806, 3.3944840.2865160, 0.0, -1.959806, 3.3944840.2865161, 0.0, -1.959806, 3.394484

Velocity.cvs file:

Inlet for U

type profile1DfixedValue;fileName "velocity.csv";fileFormat "turboCSV";interpolateCoord "R";fieldName "Velocity";fieldScaleFactor 1;value uniform (0 0 0);

7

0.2865161, 0.0, -1.959806, 3.394484

Inlet for epsilon

type turbulentMixingLengthDissipationRateInlet;mixingLength 0.000849; value $internalField;

Inlet for k

type turbulentIntensityKineticEnergyInlet;intensity 0.02;value $internalField;

Solvers

For p: GAMG� tolerance 1e-06� relTol 0.01� smoother GaussSeidel� cacheAgglomeration true� nCellsInCoarsestLevel 100

SIMPLE� nNonOrthogonalCorrectors 0� pRefCell 0� pRefValue 0� pMin pMDin [1 -1 -2 0 0 0 0] 100

relaxationFactors� p 0.3

FvSolution

8

� nCellsInCoarsestLevel 100� agglomerator faceAreaPair� mergeLevels 1

For U k epsilon: PBiCG� preconditioner DILU� tolerance 1e-05� relTol 0.1

� p 0.3� U 0.7� k 0.7� epsilon 0.7

FvSchemes

ddtSchemes� default steadyState;gradSchemes� default Gauss linear� grad(p) Gauss linear� grad(U) Gauss linear

divSchemes� default none

laplacianSchemes� default none� laplacian(nuEff,U) Gauss linear corrected� laplacian((1|A(U)),p) Gauss linear corrected� laplacian(DkEff,k) Gauss linear corrected� laplacian(DepsilonEff,epsilon) Gauss linear corrected� laplacian(DREff,R) Gauss linear corrected� laplacian(DnuTildaEff,nuTilda) Gauss linear corrected

9

� div(phi,U) Gauss GammaV 1� div(phi,k) Gauss upwind� div(phi,epsilon) Gauss upwind� div(phi,R) Gauss upwind� div(R) Gauss linear� div(phi,nuTilda) Gauss upwind� div((nuEff*dev(grad(U).T()))) Gauss linear

corrected

interpolationSchemes� default linear� interpolate(U) linear

snGradSchemes�default corrected

fluxRequired�default no

p

Structured/unstructured (hybrid) mesh

� Automatic near-blade domain partitioning

� Structured mesh generation in near-blade

region

� Periodic surface optimization based on

minimum and maximum opening angles

� Delaunay-based unstructured mesh generation

Mesh generation

10

� Delaunay-based unstructured mesh generation

� Unstructured mesh adaptation in flow passage

� OpenFOAM import through CGNS standard file

forma

Small opening Large opening

Opening of the wicket gate

11

Exemple of convergence: wg42in35

12

Experimental-CFD Comparison

13

Conclusions and Perspectives

Conclusions

� “Relatively universal” numerical setup that provides robust results over the range of distributor operating conditions,

� Convergence stagnation to a level that provides satisfactory numerical results

� Validation of several types of boundary conditions, including cyclicGGI,� Results correlate very well with experimental measurements

14

Perspectives

� Complete automation of case setup and post-processing� Unsteady flow simulations� Shape optimization

Special Thanks:

Maryse Page from IREQ/Hydro QuébecYing Zhang from from École Polytechnique de Montréal

15

Ying Zhang from from École Polytechnique de Montréal

for their help