ilmenauer mhd-woche - vom 20. bis 24. september 2004 7th mhd-days (20. - 21. september 2004)...
TRANSCRIPT
Ilmenauer MHD-Woche - vom 20. bis 24. September 2004Ilmenauer MHD-Woche - vom 20. bis 24. September 20047th MHD-Days 7th MHD-Days (20. - 21. September 2004) (20. - 21. September 2004)
Numerical modelling
of an MHD flow in the presence
of transverse non-uniform magnetic field
F. DUBOIS*, J. ETAY*, O. WIDLUND** and Y. DELANNOY**EPM-Madylam ENSHMG BP95 38402 S t Martin d'Hères Cedex (FRANCE)
** CEA DER/SSTH/LDAS (Bat. 10.05) Rue des Martyrs 38000 Grenoble (FRANCE)
Tool :Tool :Fluent + MHD module + wall functions
Goals:Goals:Compare numerical results with Ilmenau experimental results
At each iteration of the flow calculation
with boundary conditions
ϕ = ϕo
∂ϕ
∂ n
= U × B( ) ⋅ n
on chosen
on -o
11- solve the problem
2 2 - then calculate j and F
MHD modulusMHD modulus
MHD modulusMHD modulus - tricky points1 - calculation of gradients is done using a self-developed subroutine <-> sensitivity to the mesh
2 - calculation of near the wall <-> term is reconstructed
U × B( )
MHD modulusMHD modulus - validationanalytical solution of a Hartman flow : Ha = 10, 30 and 100
discrepancy 2%, 2.2% and 1.2%
axial velocity current density
parallellayer
Wall functions for laminar MHD flowWall functions for laminar MHD flow
wall
B
U n
Hyp : - elec. insulating wall - solid wall -
p ⊥ n
Xn
= X ⋅ n( ) n = Xn
n
Xp
= X − X ⋅ n( ) n
Ref : O.Widlund/Eur. J. Mech B/Fluids 22 (2003)
U pi n( ) =τi δH sinhnδH( )
ρν
+ C i +Ei( )
δH2
ρνcosh
nδH
⎛
⎝ ⎜ ⎜
⎞
⎠ ⎟ ⎟ −1
⎛
⎝ ⎜ ⎜
⎞
⎠ ⎟ ⎟
+dJ DBi δH
3
ρνsinh
n
δH
⎛
⎝ ⎜ ⎜
⎞
⎠ ⎟ ⎟ −
n
δH
⎛
⎝ ⎜ ⎜
⎞
⎠ ⎟ ⎟
δH
=
1
Bn
ρν
σ
τ
i
= ρν
∂ Up
i
∂ n
0( )
C
i
= ∇ P − n ∇ P ⋅ n( )( )
i
E
i
= σ Bn
∇ ϕ × n( )
i
DB
i
= B × n( )
i
dJ
= σ ∇p
2
ϕ
GeometryGeometry
Ilmenau loop
Honey comb20 mm
100 mm
30 mm
x
y
z
Galinstan flow
GridGrid
Refined gridRefined grid (no wall function)
•Hartman layer : first cell = 8.82m 10 cells
•// layer : first cells = 16.07m 10 cells
•total = 79 650
Coarse gridCoarse grid (+ wall function on Ha)
•Hartman layer : first mesh = 1 mm 1 cells
•// layer : first mesh = 16.07m 10 cells
•total = 31 860
100 mm 30 mm 200 mm100
mm
x
y
magnetTop view
Input Input parametersparameters
magnetic field
B ≈ Bz
x , y( )
x y
ResultsResultshyp : laminarhyp : laminar
particles path
e.m. forcese.m. forces
xelec. currentselec. currentson symmetry face
x
y
x
ResultsResults
Are identical for both grids, thanks to the use of wall functions
Are in conformity with phenomenology
Magnetic field acts as a “semi-permeable body”
2 wall jets developed at the vertical walls
Elec. Currents paths are different upstream and
downstream of the magnet
Comparison with Comparison with experimental results 1/4experimental results 1/4
voltagevoltage on x=0, on symmetry plane
shapes and level are similar
difference near the // walls
built with ϕ Δϕy num.
Comparison with Comparison with experimental results 2/4experimental results 2/4
voltagevoltage on symmetry plane
Comparison with Comparison with experimental results 3/4experimental results 3/4
on ∆ϕy - in the centre : identical level - near the side wall : num M / exp no M
proposed explanation in experiments
∂ϕ
∂ y
= −
jy
σ
+ ux
Bz
⎛
⎝
⎜
⎞
⎠
⎟
in the center ux weak jy high
∂ϕ
∂ y
≈ −
jy
σ
in the M ux high jy weak
∂ϕ
∂ y
≈ − ux
Bz
when B->0 then ∆ϕy->0
Comparison with Comparison with experimental results 4/4experimental results 4/4
velocity
- in the centre : same level- in the M-jet : different level and
width- lower velocity at the border of
the jetproposed explanation- calibration ? - size of the probe ?- in exp. no flow conservation ux(z)?
calculations
experiments
z
ux (m/s) - symmetry plane - x=19,5 mm
probe
Comparison of numerical results with Ilmenau experimental results are satisfactory.
Wall functions effective in reducing the cost of gridding in boundary layers
Discrepancies can be explained
Turbulence is weak => the turbulence model can not be checked in optimal conditions.
ConclusionsConclusions
ResultsResults
x
elec. potential