Are you sure?
This action might not be possible to undo. Are you sure you want to continue?
G. Degrez ULB
Collège Belgique
Nouvelles méthodologies multidisciplinaires pour la simulation des écoulements 25 Novembre 2009
Introduction
What is magnetohydrodynamics or (probably better) magneto ﬂuid dynamics?
2
Introduction
What is magnetohydrodynamics or (probably better) magneto ﬂuid dynamics? Magneto ﬂuid dynamics (MFD) deals with ﬂows of electrically conducting ﬂuids, such as electrolytes, liquid metals or plasmas in presence of an electromagnetic ﬁeld Interaction between ﬂowﬁeld (velocity & thermodynamic variables) and electromagnetic ﬁeld.
2
Examples
S N
F
http://web.mit.edu
The liquid jet is deviated by the electromagnetic (Lorentz) force.
3
Examples
S N
F
http://web.mit.edu
The liquid jet is deviated by the electromagnetic (Lorentz) force.
3
Examples (cont.)
electromagnetic agitation
N
liquid metal solidiﬁed steel
S
magnetic ﬁeld
Steel production
N
S
electromagnetic pump
Geomagnetism
4
Examples (cont.)
Nuclear fusion Nuclear fusion reactor
Basic principle
Magnets
Blanket
Plasma
ITER
5
Governing equations
• Electromagnetics
·E 1 µ0 ·B ∂E ×B− 0 ∂t ×E = ηc
0
= 0 = J ∂B = − ∂t
with J = σ E + u × B
6
Approximations:
• Quasineutral medium ( • Displacement current (
ηc ≈ 0 ) fp )
electrostatic oscillations neglected ( f
0 ∂ E/∂t
) negligible
electromagnetic waves neglected With these approximations, Maxwell’s law reduces to Faraday’s law
1 J= µ0
7
×B
• Flowﬁeld ∂ρ
∂t ∂ρu + ∂t ∂ρE + ∂t +
· ρu =
0 ·τ +J ×B ·q+J ·E · u) + µv I ·u
· ρu ⊗ u = − p + · ρuH =
· (τ · u) − 2 ⊗u − I 3
T
with
τ q = µ( = −λ T ⊗u+
8
Dimensionless parameters • Deﬁne dimensionless quantities
• Then,
E xi xi = , ˜ L = Ut ˜ t= , L ui ui = , ˜ U B ˜ B= B0 J 1 −u×B = σ µ0 σ ×B−u×B
= U B0
˜ ×B ˜ 1 ˜ ˜ B×u+ σr µ0 U L σ ˜
and Rm = σr µ0 U L is the magnetic Reynolds number
9
• Lorentz force
FL = J × B = =
2 σr U B0 σ ˜
˜ ˜ ˜ ˜ E+u×B ×B
2 σr U B0 ˜ ˜ ˜ ( × B) × B Rm
∂ρu ˜ + ˜ · ρu ⊗ u ˜ ˜ ˜ ∂t
• momentum equation
1 ˜ ˜ ˜ ˜ ˜ = ˜p + ˜ τ +N E+u×B ×B ˜ Re N ˜ 1 ˜ ˜ ˜ τ+ ˜ ( × B) × B = ˜p + ˜ Re Rm 2 σr B0 L where N = is the magnetic ρr U
interaction parameter
10
• Induction equation [☜,☚]
˜ ∂B +˜× ˜ ∂t
˜ 1 ˜ ×B ˜ ˜ B×u+ Rm σ ˜
=0
• Joule heating
PJ = J · E = =
2 B0 U
˜ ˜ ×B ˜ ∂B 1 ˜ ˜ ˜ + ˜ · (u ⊗ B − B ⊗ u) = − ˜ ˜ × ˜ Rm σ ˜ ∂t
2 B0 U ˜ ˜ ( × B) · Lµ0
˜ ×B ˜ ˜ ˜ B×u+ Rm
˜ × B)2 ˜ ( ˜ ˜ ˜ ˜ ˜ ((B ⊗ u − u ⊗ B) · · ˜ ⊗ B + ) Lµ0 Rm
11
• energy equation [☜,☚]
∂ ρ(˜ + ˜e ˜ ∂t
u ·u ˜ ˜ Ec 2 )
u·u ˜ ˜ Ec + ˜ · ρu(h + Ec ˜˜ ˜ )= 2 Re
· (τ · u)− ˜ ˜
˜ × B)2 ˜ 1 ˜ N Ec ˜ ( ˜ ·q+ ˜ ((B ⊗ u − u ⊗ B) · · ˜ ⊗ B + ˜ ˜ ˜ ) Re Rm Rm
where Ec = U 2 /∆hr is the Eckert number. The reference enthalpy variation ∆hr is at least equal to U 2, so that Ec ≤ 1.
12
Additional remarks :
•
2 2 N B0 B0 pr 4 = = = for P. G., Rm ρr U 2 µ0 pr µ0 ρr U 2 γM 2 β where M is the Mach number and β is the socalled magnetic β .
•
number. It typically takes very small values for liquid metal, electrolytes and plasmas, which implies Rm Re .
Rm = νσr µ0 = P rm is the magnetic Prandtl Re
13
Flow regimes & simpliﬁcations • Re, R 1 : ideal MHD (astrophysics)
The dissipative terms ( ∝ 1/Re or 1/Rm ) are neglected (but not the terms ∝ N/Rm ). Induction equation [☞]
∂B + ∂t · (u ⊗ B − B ⊗ u) = 0
m
Taking the scalar product of this equation by B/µ0 and adding to the energy equation [☞], one gets
14
∂ρE +
B2 2µ0
∂t B · µ0
+
· ρuH+ ⊗B =0 µ0
· (u ⊗ B − B ⊗ u) + (u ⊗ B − B ⊗ u) · ·
B 1 = ·( µ ·(u⊗B−B⊗u))= µ 0 0
·(B 2 u−(u·B)B)
√ or, deﬁning ˆ = B/ µ0 , b ∂ρE + ∂t
ˆ2 b 2
+
or else (with ρE = ρE
∂ρE + ∂t
· (u(ρH + ˆ2 ) − (u · ˆ ˆ = 0 b b)b)
ˆ2 + b2 ˆ2
)
b · (u(ρE + p + ) − (u · ˆ ˆ = 0 b)b) 2
15
Since
( × B) × B = (B · ∂ρu + ∂t )B −
the momentum equation can be rewritten as so that the complete system reads ﬁnally
ˆ2 b · (ρu ⊗ u − ˆ ⊗ ˆ + (p + )I) = 0 b b 2
B2 = 2
·B⊗B−
B2 2
ρ ∂ ρu + ∂t ˆ b ρE
ρu ˆ2 ρu ⊗ u − ˆ ⊗ ˆ + (p + b2 )I b b · u⊗ˆ−ˆ⊗u b b ˆ2 u(ρE + p + b2 ) − (u · ˆ ˆ b)b
16
=0
• constant density ﬂows (liquid metals &
·u = 0
electrolytes) with Re 1, Rm ≥ O(1) . Assuming constant viscosity & electrical conductivity,
∂u + ∂t ∂b + ∂t · (u ⊗ u − b ⊗ b) · (u ⊗ b − b ⊗ u) = − P +ν = νm
2 2
u
b
with
ˆ b b= √ , ρ p b2 P = + , ρ 2
17
νm
1 = . σµ0
• constant density ﬂows with Re
1, Rm
1.
Decomposing the magnetic ﬁeld into a uniform externally imposed ﬁeld B0 and a ﬂuctuation B much smaller than B0 , the system simpliﬁes to (quasistatic approximation)
∂u + ∂t · (u ⊗ u − b0 ⊗ b ) − ·u = 0
2
= − P +ν = νm
2
u
p with P = + b0 · b ρ
· (b0 ⊗ u)
b
18
• low speed (small Ec), very small R
Induction equation [☞]
∂B + ∂t ×B × =0 σµ0
variable density ﬂows (e.g. IC thermal plasmas): fully resistive MHD.
m
or alternatively (taking the curl of this eqn),
∂σµ0 E + ∂t ×( × E) = 0
☞
19
Energy equation [☞] With Ec 1 , the kinetic energy contribution and the power of the viscous stresses can be neglected. But the Joule heating contribution 2 scaling as N Ec/Rm cannot.
∂ρe + ∂t · ρuh = − ·q+J ·E
=σE 2
20
ρ ∂ ρu + ∂t ˆ b ρE
• Ideal (compressible) MHD
MHD waves
ρu b ˆ ⊗ ˆ + (p + ˆ2 )I ρu ⊗ u − b b 2 · u⊗ˆ−ˆ⊗u b b ˆ2 u(ρE + p + b2 ) − (u · ˆ ˆ b)b 1 ωe = ∂s ρ − 2 ∂s p a
21
Hyperbolic system with 7 eigen modes: entropy wave
λe = u s ,
=0

bs λA = us ± √ , ρ
 ˆAlfvén waves 
ˆo b ωA = −ˆo ∂s un + ˆn ∂s uo ± √ ∂sˆn b b b ρ
ˆn b b √ ∂sˆo ρ
Fast/slow magnetoacoustic waves λf,s =us ± cf,s with
cf,s = ρa2 + ˆ2 ± b (ρa2 + ˆ2 )2 − 4ρa2ˆ2 b bs 2ρ ρcf,sˆsˆn b b ∂s un ρc2 − ˆ2 b
f,s s
ωf,s =ρ∂s ρ ± ρcf,s ∂s us
ρc2 ˆn b ρc2 ˆo f,s f,s b ∂sˆn + b ∂sˆo + ∂s p b ρc2 − ˆ2 bs ρc2 − ˆ2 bs f,s f,s
22
ρcf,sˆsˆo b b ∂s uo + ρc2 − ˆ2 b
f,s s

the 8th wave is degenerate. It has a 0 wave speed ( the jacobian matrix is singular) and its amplitude is
ω8 = ∂sˆs = 0 b
since for a 1D (plane) wave, the b solenoidal constraint imposes ∂sˆs = 0 . This is a source of numerical difﬁculties since numerically, the solenoidal constraint is not exactly satisﬁed.
23
The solenoidal constraint
Numerical problem & solution techniques
• At the continuous level, the induction
∂B + ∂t ∂(
equation ensures that ∂t ( · B) = 0 . Indeed,
· B) ×E =0 ⇒ = − · ( × E) = 0 ∂t At the discrete level though, · ( × E) is not exactly 0. As a result, · B errors can
grow unboundedly unless some mechanism to control them is implemented.
24
• Constrained transport technique
consists in applying a special discretization ensuring that · ( × E) = 0 exactly at the discrete level. For FDM/FVM, this generally implies the use of a staggered mesh with hydrodynamic variables, electric ﬁeld and magnetic ﬁeld evaluated at different locations. For FEM, this implies the use of mixed or hybrid elements such as Nédélec’s element.
25
• Powell’s source term technique
S=− ˆ · b 0 ˆ b u
Add to the RHS of the ideal MHD system the source term In this way, the 8th degenerate wave transforms into
u·ˆ b
so that ﬂow (but may buildup in stagnation regions).
26
∂ ·ˆ b ·ˆ b ( )+u·( )=0 ∂t ρ ρ · ˆ errors are swept away by the b
• Divergence cleaning techniques
∂B + ∂t
The basic idea of divergence cleaning techniques is to add a Lagrange multiplier to the induction equation
· (u ⊗ B − B ⊗ u) + ψ=0
and a differential equation for the Lagrange multiplier
D(ψ) +
• D(ψ) = 0 : This is the classical projection
27
·B =0
⇔
∂D(ψ) − ∂t
2
ψ=0
scheme. Requires solving a Poisson eqn at each time step: expensive!
• D(ψ) = ψ/α : Then, the differential problem
for the Lagrange multiplier ψ reads i.e. a diffusion equation.
∂ψ =α ∂t
2
ψ
•
1 ∂ψ and the differential problem D(ψ) = 2 V ∂t for the Lagrange multiplier ψ reads ∂2ψ 2 2 =V ψ ∂t2
i.e. a wave equation. This is analogous to the artiﬁcial compressibility technique for incompressible ﬂows (ACA).
28
The ideal MHD system then transforms to
∂ ∂t ρ ρu ˆ b ρE ψ + ρu ρu ⊗ u − ˆ ⊗ ˆ + (p + ˆ2 )I b b b 2 u ⊗ ˆ − ˆ ⊗ u + ψI b b · ˆ2 b u(ρE + p + 2 ) − (u · ˆ ˆ b)b V 2ˆ b ωd.c. = ∂sˆs ± b =0
This hyperbolic system has 2 additional eigen modes, the divergence cleaning waves with 1
λd.c. = ±V,
2
V Note that since the velocity V is a constant,
∂s ψ
errors may not build up.
29
Sample applications
I. Space weather (ideal MHD) M.Yalim, D.Vanden Abeele & H. Deconinck
• Discretization: cellcentered ﬁnite volume, • Timediscretization: (implicit) backwardEuler (steady cases), (implicit) 3point backward or 2stage RK (unsteady test cases).
30
2nd order TVD scheme or LF with reduced dissipation, BarthJespersen limiter with limiter freezing.
•
Validation 1: steady magnetic nozzle ﬂow Flow conditions:
u M = = 2.54, a u MA = = 3, vA β=2
• Unstructured grid with 4972 triangles
31
Validation 1: steady magnetic nozzle ﬂow (cont.)
Density contours with magnetic ﬁeld lines (ACA + TVDLF)
Density contours with magnetic ﬁeld lines (Powell + TVDLF)
32
Validation 1: steady magnetic nozzle ﬂow (cont.)
Density distributions along wall/symmetry line
33
Validation 1: steady magnetic nozzle ﬂow (cont.)
Numerical scheme 1st order ACALF (∇⋅b)min (∇⋅b)max 6.90 108 2.84 107
→ →
idem with reduced (.1) dissipation 1.49 107 9.67 107 ACATVD/LF ACATVD/Roe PowellTVD/LF
→
7.49 107 2.34 107 4.13 107 5.08 107 0.37666 0.37173
(∇⋅b) errors with several schemes
34
Validation 1I: supersonic OrszagTang vortex Unsteady test case: subiterations are b performed in order to achieve · ˆ = 0 at each time step.
Pressure contours at ﬁnal time: (a) ACA, (b) const. transport
35
Validation 1I: supersonic OrszagTang vortex (cont.)
Pressure distribution along y=1.93
→ →
Iterative scheme 3point backward with subiterations RK2 without subiterations
(∇⋅b)min (∇⋅b)max 0.0771 0.536 0.0785 0.675
36
Solar wind/earth magnetosphere interaction
• For this application, the global magnetic
ﬁeld is decomposed into the contribution of the earth magnetic ﬁeld (B0 ) and a perturbation ( B1 ). steady dipole ﬁeld
• The earth magnetic ﬁeld is modelled as a
1 B0 = 3 (3(m · er )er − m) r
37
Solar wind/earth magnetosphere interaction (cont.)
• The ideal MHD system is rewritten in
ρ ∂ ρu + ∂t ˆ1 b ρE1 ·
terms of the perturbation ﬁeld B1 (Tanaka).
ρu ˆ2 ˆ1 ⊗ ˆ1 + (p + b1 )I ρu ⊗ u − b b 2 · u ⊗ ˆ1 − ˆ1 ⊗ u b b ˆ2 b1 u(ρE1 + p + 2 ) − (u · ˆ1 )ˆ1 b b 0 ˆ0 · ˆ1 )I − (ˆ0 ⊗ ˆ1 + ˆ1 ⊗ ˆ0 ) (b b b b b b =0 ˆ0 − ˆ0 ⊗ u u⊗b b u(ˆ0 · ˆ1 ) − (u · ˆ1 )ˆ0 b b b b
+
• Powell’s source term approach and ACA
are appropriately adapted.
38
Solar wind/earth magnetosphere interaction (cont.)
• earth magnetic ﬁeld: m = −3e • Far ﬁeld conditions: M = 6.2, M
u = uex
z A
= 8, β = 2 ,
• •
ˆ = −ˆ z (interplanetary magnetic ﬁeld b be
aligned with earth dipole) opposite to earth dipole)
ˆ = ˆ z (interplanetary magnetic ﬁled b be
• inner boundary = magnetosphere/ionosphere ˆ
boundary (at r = 3Rearth ): ρ = 1, p = 8, b · n = 0
39
Solar wind/earth magnetosphere interaction (cont.)
• Unstructured grid with 1.220.758
tetrahedra
Mesh in the plane of symmetry (xz)
40
Solar wind/earth magnetosphere interaction (cont.)
Schematic of ﬂow features (aligned IMF)
41
Solar wind/earth magnetosphere interaction (cont.)
(a)
(b)
Pressure ﬁeld contour & magnetic ﬁeld lines in symmetry plane (aligned IMF) (a) ACA, (b) Powell’s source term (Powell et al.)
42
Solar wind/earth magnetosphere interaction (cont.)
(a)
(b)
Pressure ﬁeld contour & magnetic ﬁeld lines in symmetry plane, closeup (aligned IMF) (a) ACA, (b) Powell’s source term (Powell et al.)
43
Solar wind/earth magnetosphere interaction (cont.)
(a)
(b)
Pressure ﬁeld contour & magnetic ﬁeld lines in symmetry plane (opposite IMF) (a) ACA, (b) Powell’s source term (Powell et al.)
44
Solar wind/earth magnetosphere interaction (cont.)
(a)
(b)
Pressure ﬁeld contour & magnetic ﬁeld lines in symmetry plane, closeup (aligned IMF) (a) ACA, (b) Powell’s source term (Powell et al.)
45
II. Incompressible MHD turbulence M. Kinet, B. Knaepen & D. Carati
•
Application 1: homogeneous turbulence very low magnetic Reynolds number quasistatic approximation
• Discretization: spectral (Fourier), pseudo• Timeintegration: combination of analytic
integration (viscous terms) and 3stage RungeKutta method (nonlinear terms)
46
spectral treatment of nonlinear terms and Rogallo’s phaseshift dealiasing method
Application 1: homogeneous turbulence (cont.)
without magnetic ﬁeld with magnetic ﬁeld
B
zero vorticity
large vorticity
B
Main effects 1) “less turbulent” 2) anisotropy development
B
t
47
Application 1: homogeneous turbulence (cont.)
B
B
B
t
vertical vortices
Energy spectrum
48
Application 1: homogeneous turbulence (cont.)
E(k⊥ , k )
80 70 60 50 40 !8 30 20 10 0 0 k 0 !2 !4 80 70 60 50 !6 40 !10 30 !10 20 10 0 0 k 0 !2
Joule dissipation
14.4
!4 !6 !8
12.6 9.8 7
10.8 7.2
20 40 k⊥ 60 80
!12 !14 !16 !18
!12 20 40 k⊥ 60 80 !14
without magnetic ﬁeld
with magnetic ﬁeld
49
Application II: Instabilities in duct ﬂow
• Discretization: 2nd order cellcentered
ﬁnite volume
• Timeintegration: fractional step method
for incompressible ﬂows, CrankNicolson discretization of viscous terms, semiimplicit discretization of convective terms, explicit Euler discretization of Lorentz force. pressure correction), BiCG stab (momentum equations).
50
• Linear solvers: AMG (electric potential /
Application II: Instabilities in duct ﬂow (cont.)
• Flow conﬁguration & phenomena
dp d2 u 2 = −σB u + ρν 2 dx dy
In 2D (inﬁnite width), J = σ(u × B) directed 2 spanwise, and FL = −σB u .
⇒ 1 dp u=− σB 2 dx
with δH =
ρν , and the Hartmann number σB 2 √ h σ Ha = = Bh = N Re δH ρν
51
cosh(y/δH ) 1− cosh(h/δH )
Application II: Instabilities in duct ﬂow (cont.) In 3D, the ﬂow behaviour near the side walls strongly depends on the electrical properties (conductor/insulator) of the walls.
Insulating walls
Conducting walls
52
Application II: Instabilities in duct ﬂow (cont.)
Conducting Hartmann walls, insulating side walls
Strong side jets for large Hartmann numbers (intense magnetic ﬁeld), more prone to the development of instabilities.
lent rapide
•When do they appear? •How do they modify ﬂow properties?
53
Application II: Instabilities in duct ﬂow (cont.)
B
54
Application II: Instabilities in duct ﬂow (cont.)
B
54
Application II: Instabilities in duct ﬂow (cont.)
B
54
Application II: Instabilities in duct ﬂow (cont.)
Laminar regime
“First” unstable regime
55
Application II: Instabilities in duct ﬂow (cont.)
“Second” unstable regime
turbulent regime
56
III. RF inductively coupled thermal plasmas D.Vanden Abeele, T. Magin, P. Rini, G. Degrez
RF inductively coupled plasma torch
57
III. RF inductively coupled thermal plasmas D.Vanden Abeele, T. Magin, P. Rini, G. Degrez
RF inductively coupled plasma torch
•Gas at 0.01 atm or higher is injected in a
quartz tube surrounded by a copper inductor, •A radiofrequency electrical current runs through the inductor,
57
III. RF inductively coupled thermal plasmas D.Vanden Abeele, T. Magin, P. Rini, G. Degrez
RF inductively coupled plasma torch
57
III. RF inductively coupled thermal plasmas D.Vanden Abeele, T. Magin, P. Rini, G. Degrez
RF inductively coupled plasma torch
•and induces a secondary current through
the gas inside the quartz tube, which heats up by means of ohmic dissipation to a partially ionized plasma state with peak temperatures around 10,000 K.
57
58
Because of their electrodeless heating, IC plasmas are of high chemical purity, which makes them popular for a variety of applications:
58
Because of their electrodeless heating, IC plasmas are of high chemical purity, which makes them popular for a variety of applications:
•
deposition of metal coatings,
58
Because of their electrodeless heating, IC plasmas are of high chemical purity, which makes them popular for a variety of applications:
• •
deposition of metal coatings, synthesis of ultraﬁne powders,
58
Because of their electrodeless heating, IC plasmas are of high chemical purity, which makes them popular for a variety of applications:
• • •
deposition of metal coatings, synthesis of ultraﬁne powders, generation of high purity silicon,
58
Because of their electrodeless heating, IC plasmas are of high chemical purity, which makes them popular for a variety of applications:
• • • •
deposition of metal coatings, synthesis of ultraﬁne powders, generation of high purity silicon, testing of thermal protection materials for atmospheric (re)entry vehicles.
58
Speciﬁc modelling features
• Axisymmetric geometry and ﬂow, • Steady hydrodynamic ﬁeld (velocity/
thermodynamic variables),
• purely monochromatic azimuthal electric
poloidal current),
ﬁeld, E = (ER + i EI ) exp(iωt)eθ = E exp(iωt)eθ
• steady ambipolar poloidal electric ﬁeld (zero
whence the induction equation simpliﬁes to
nc 2
E − iωµ0 σE = −iωµ0 Ic
59
i=1
δ(r − ri )
Speciﬁc modelling features (cont.)
•
Steady ﬂow at low Eckert number, hence
· ρu ⊗ u = − p + · ρuh = − · ρu = 0 ·q+J ·E ·τ +J ×B
where the Lorentz force and Joule heating terms are timeaveraged.
•
Reactive ﬂow, i.e. nonuniform chemical composition. Can be modelled in two ways:
• •
Chemical nonequilibrium Local (thermo)chemical equilibrium
60
Chemical nonequilibrium Additional conservation equations for the chemical species must be considered: where
· ρys u = − · Js + ωs , ˙ s = 1, . . . , ns
• •
the diffusion ﬂuxes Js are computed by solving the StefanMaxwell equations, and
˙ the chemical source terms ωs are computed using the law of mass action and an Arrhenius formulation for the reaction rates.
61
Local (thermo) chemical equilibrium In this case, additional conservation equations for the chemical elements must be considered. The elemental diffusion ﬂuxes Jα can be expressed as linear combinations of the electric ﬁeld and elemental mass fractions, pressure and temperature gradients (Rini et al., Phys. Rev. E, 2005).
Jα =
E −Dα E
· ρYα u = −
· Jα ,
α = 1, . . . , ne
−
T ρDα
T−
62
p ρDα
p − ρDαβ Yβ
Heat ﬂux For both chemical nonequilibrium & local (thermo)chemical equilibrium, the heat ﬂux must be modiﬁed to take into account the diffusion heat ﬂux q = qd − λ T . Chemical nonequilibrium: qd =
ns
Local (thermo)chemical equilibrium: the diffusion heat ﬂux can also be expressed in terms of the electric ﬁeld and the elemental mass fraction, pressure and temperature gradients.
qd = −λE E − λT T − λp p − λα Yα
63
s=1
hs Js
Numerical aspects: space discretization
V = (δp, Yα or ys , ρu, T )
t
64
Numerical aspects: space discretization
• 2nd order pressurestabilized cellcentered
ﬁnite volume solver written in terms of
V = (δp, Yα or ys , ρu, T )
t
64
Numerical aspects: space discretization
• 2nd order pressurestabilized cellcentered
ﬁnite volume solver written in terms of
V = (δp, Yα or ys , ρu, T )
t
• MUSCL upwind approach for species/
elemental mass fractions, momentum & energy,
64
Numerical aspects: space discretization
• 2nd order pressurestabilized cellcentered
ﬁnite volume solver written in terms of
V = (δp, Yα or ys , ρu, T )
t
• MUSCL upwind approach for species/
elemental mass fractions, momentum & energy, ﬁeld discretized using central ﬁnite volume discretization on a farﬁeld mesh extending beyond the torch
64
• nonsingular formulation for the electric
Numerical aspects: iterative strategies
65
Numerical aspects: iterative strategies
• The discretized equations are solved using
damped (quasi) Newton strategies and modern iterative linear solvers with various levels of ﬂowﬁeld/EM ﬁeld coupling:
65
Numerical aspects: iterative strategies
• The discretized equations are solved using • decoupled approximate Jacobian with
frozen mass ﬂuxes (Picard),
damped (quasi) Newton strategies and modern iterative linear solvers with various levels of ﬂowﬁeld/EM ﬁeld coupling:
65
Numerical aspects: iterative strategies
• The discretized equations are solved using • decoupled approximate Jacobian with
frozen mass ﬂuxes (Picard), Newton),
damped (quasi) Newton strategies and modern iterative linear solvers with various levels of ﬂowﬁeld/EM ﬁeld coupling:
• decoupled approximate Jacobian (Quasi
65
Numerical aspects: iterative strategies
• The discretized equations are solved using • decoupled approximate Jacobian with
frozen mass ﬂuxes (Picard), Newton),
damped (quasi) Newton strategies and modern iterative linear solvers with various levels of ﬂowﬁeld/EM ﬁeld coupling:
• decoupled approximate Jacobian (Quasi• exact coupled Jacobian (Newton)
65
Application I: comparison of LTE & CNEQ Geometry & operating conditions
Wall: 3 mm 1.4 W/m K R1 = 67 mm R2 = 80 mm R3 = 108.5 mm
L1 = 100 mm L2 = 250 mm L3 = 470 mm
pressure mass ﬂow frequency
0.05  0.3 atm Quartz tube 6 g/s 0.45 MHz swirl angle power
3mm 1.4 Wm1K1 45 ˚ 75 kW
66
Application I: comparison of LTE & CNEQ Physicochemical and numerical parameters
• 11species air model (N , O , NO, O, N,
2 2
N2 , O2
+
+
+, O+, N+, e), ,NO
• chemical kinetics model of Dunn & Kang, • meshes: 131 x 47 , 262 x 94 (ﬂow) • noncatalytic quartz tube surface.
67
Application I: comparison of LTE & CNEQ (cont.)
Oxygen elemental fraction at 0.05 atm
68
Application I: comparison of LTE & CNEQ (cont.)
Oxygen elemental fraction at 0.05 atm
Signiﬁcant nonequilibrium effects at low pressure
68
Application I: comparison of LTE & CNEQ (cont.)
Oxygen elemental fraction at 0.05 atm Oxygen elemental fraction at 0.3 atm
69
Application I: comparison of LTE & CNEQ (cont.)
Oxygen elemental fraction at 0.05 atm Oxygen elemental fraction at 0.3 atm
• Much smaller nonequilibrium effects at high pressure
69
Application I: comparison of LTE & CNEQ (cont.)
Oxygen elemental fraction at 0.05 atm Oxygen elemental fraction at 0.3 atm
• Much smaller nonequilibrium effects at high pressure • Similar amount of elemental fraction variation
69
Application I: comparison of LTE & CNEQ (cont.)
Outlet species mole fraction proﬁles at 0.3 atm
70
Application I: comparison of LTE & CNEQ (cont.)
Outlet species mole fraction proﬁles at 0.3 atm
LTE formulation valid at high pressures
70
Application II: TCNEQ computations
71
Application II: TCNEQ computations
• Identical operating conditions, except:
Pressure = 0.05 atm, Power = 50 kW
71
Application II: TCNEQ computations
• Identical operating conditions, except:
Pressure = 0.05 atm, Power = 50 kW
• Identical air model & chemical kinetics
model + 2temperature thermal NEQ model
71
Application II: TCNEQ computations
• Identical operating conditions, except:
Pressure = 0.05 atm, Power = 50 kW
• Identical air model & chemical kinetics
model + 2temperature thermal NEQ model
• coarser 66 x 29 ﬂow mesh, 85 x 42 EM mesh
71
Application II: TCNEQ computations
• Identical operating conditions, except:
Pressure = 0.05 atm, Power = 50 kW
• Identical air model & chemical kinetics
model + 2temperature thermal NEQ model
• coarser 66 x 29 ﬂow mesh, 85 x 42 EM mesh
71
Application II: TCNEQ computations (cont.)
0.600 0.000 0.200 0.013 0.000 0.000 0.050
Flowﬁeld pattern
72
Application II: TCNEQ computations (cont.)
0.600 0.000 0.200 0.013 0.000 0.000 0.050
Flowﬁeld pattern
• Typical base ﬂow pattern, with minor inﬂuence of
Lorentz and centrifugal forces
72
Application II: TCNEQ computations (cont.)
315 1000 3000 8500 8000 5000 6000 7000
Electrovibrational temperature
315
1000
3000
5000 6200
6000
6150
6200 6240
Rototranslational temperature
73
Conclusion
74
Conclusion
• Magnetoﬂuid dynamics is a tremendously
varied ﬁeld of study with many wildly different regimes and an increasing number of applications in fundamental research & industry.
74
Conclusion
• Magnetoﬂuid dynamics is a tremendously • It poses speciﬁc numerical difﬁculties, in
varied ﬁeld of study with many wildly different regimes and an increasing number of applications in fundamental research & industry. particular the satisfaction of the solenoidal constraint & BCs for the EM ﬁeld.
74
Conclusion
• Magnetoﬂuid dynamics is a tremendously • It poses speciﬁc numerical difﬁculties, in
varied ﬁeld of study with many wildly different regimes and an increasing number of applications in fundamental research & industry. particular the satisfaction of the solenoidal constraint & BCs for the EM ﬁeld. recent studies carried out at ULB & VKI.
• Recent progress has been illustrated by a few
74
Thank you for your attention
75
This action might not be possible to undo. Are you sure you want to continue?