Cell Centered Finite Volume Schemes for Multiphase Flow Applications - - PowerPoint PPT Presentation

cell centered finite volume schemes for multiphase flow
SMART_READER_LITE
LIVE PREVIEW

Cell Centered Finite Volume Schemes for Multiphase Flow Applications - - PowerPoint PPT Presentation

Outline Applications and meshes Flux formulation Discrete variational framework Numerical Experiments Cell Centered Finite Volume Schemes for Multiphase Flow Applications L. Agelas 1 , D. Di Pietro 1 , J. Droniou 2 , I. Kapyrin 1 , R. Eymard 3


slide-1
SLIDE 1

Outline Applications and meshes Flux formulation Discrete variational framework Numerical Experiments

Cell Centered Finite Volume Schemes for Multiphase Flow Applications

  • L. Agelas1, D. Di Pietro1, J. Droniou2, I. Kapyrin1, R. Eymard3,
  • C. Guichard1, R. Masson1.

1Institut Français du Pétrole 2 Université de Montpellier 3 Université Paris Est

june 22-24th 2009

  • L. Agelas1, D. Di Pietro1, J. Droniou2, I. Kapyrin1, R. Eymard3, C. Guichard1, R. Masson1.

Cell Centered Finite Volume Schemes for Multiphase Flow Applications

slide-2
SLIDE 2

Outline Applications and meshes Flux formulation Discrete variational framework Numerical Experiments

1

Applications and meshes

2

Flux formulation The L and G schemes

3

Discrete variational framework The GradCell scheme The O scheme

4

Numerical Experiments

  • L. Agelas1, D. Di Pietro1, J. Droniou2, I. Kapyrin1, R. Eymard3, C. Guichard1, R. Masson1.

Cell Centered Finite Volume Schemes for Multiphase Flow Applications

slide-3
SLIDE 3

Outline Applications and meshes Flux formulation Discrete variational framework Numerical Experiments

Applications Basin Modeling Reservoir simulation CO2 geological storage

  • L. Agelas1, D. Di Pietro1, J. Droniou2, I. Kapyrin1, R. Eymard3, C. Guichard1, R. Masson1.

Cell Centered Finite Volume Schemes for Multiphase Flow Applications

slide-4
SLIDE 4

Outline Applications and meshes Flux formulation Discrete variational framework Numerical Experiments

Meshes: corner point geometries with faults

  • L. Agelas1, D. Di Pietro1, J. Droniou2, I. Kapyrin1, R. Eymard3, C. Guichard1, R. Masson1.

Cell Centered Finite Volume Schemes for Multiphase Flow Applications

slide-5
SLIDE 5

Outline Applications and meshes Flux formulation Discrete variational framework Numerical Experiments

Meshes: corner point geometries with erosions

  • L. Agelas1, D. Di Pietro1, J. Droniou2, I. Kapyrin1, R. Eymard3, C. Guichard1, R. Masson1.

Cell Centered Finite Volume Schemes for Multiphase Flow Applications

slide-6
SLIDE 6

Outline Applications and meshes Flux formulation Discrete variational framework Numerical Experiments

Meshes: basin geometries

  • L. Agelas1, D. Di Pietro1, J. Droniou2, I. Kapyrin1, R. Eymard3, C. Guichard1, R. Masson1.

Cell Centered Finite Volume Schemes for Multiphase Flow Applications

slide-7
SLIDE 7

Outline Applications and meshes Flux formulation Discrete variational framework Numerical Experiments

Meshes: nearwell meshes

  • L. Agelas1, D. Di Pietro1, J. Droniou2, I. Kapyrin1, R. Eymard3, C. Guichard1, R. Masson1.

Cell Centered Finite Volume Schemes for Multiphase Flow Applications

slide-8
SLIDE 8

Outline Applications and meshes Flux formulation Discrete variational framework Numerical Experiments

Difficulties: Geometry Degenerated cells due to erosion Dynamic mesh (basin models): the scheme must be recomputed at each time step Faults in basin models: geometry not always available (overlaps and holes) Conductive Faults in basin models General polyhedral cells Submeshes (dead cells) Local Grid Refinement Adaptive Mesh Refinement Boundary conditions

  • L. Agelas1, D. Di Pietro1, J. Droniou2, I. Kapyrin1, R. Eymard3, C. Guichard1, R. Masson1.

Cell Centered Finite Volume Schemes for Multiphase Flow Applications

slide-9
SLIDE 9

Outline Applications and meshes Flux formulation Discrete variational framework Numerical Experiments

Difficulties: complex physics Heterogeneous anisotropic media Dispersion (full tensor,time and space dependent) Multiphase Darcy Flows Complex closure laws: thermodynamical equilibrium, geochemistry, Kinetics Thermics Geomechanics

  • L. Agelas1, D. Di Pietro1, J. Droniou2, I. Kapyrin1, R. Eymard3, C. Guichard1, R. Masson1.

Cell Centered Finite Volume Schemes for Multiphase Flow Applications

slide-10
SLIDE 10

Outline Applications and meshes Flux formulation Discrete variational framework Numerical Experiments

Motivations of cell centered schemes for compositional multiphase Darcy flow applications

Nc primary unknowns per cell for multiphase compositional flows Explicit linear fluxes Easier to combine TPFA and MPFA Existing Efficient Preconditioners like CPR-AMG

Adapted to fully or semi implicit discretizations of multiphase compositional Darcy flows

But “compact” MPFA VF schemes are non symmetric on general meshes

Possible lack of robustness due to mesh and diffusion coefficients dependent coercivity (linear solver, pressure convergence) and monotonicity (non linear solver)

  • L. Agelas1, D. Di Pietro1, J. Droniou2, I. Kapyrin1, R. Eymard3, C. Guichard1, R. Masson1.

Cell Centered Finite Volume Schemes for Multiphase Flow Applications

slide-11
SLIDE 11

Outline Applications and meshes Flux formulation Discrete variational framework Numerical Experiments

ArcGeoSim platform Based on Arcane Platform co-developped by CEA-DAM and IFP

  • L. Agelas1, D. Di Pietro1, J. Droniou2, I. Kapyrin1, R. Eymard3, C. Guichard1, R. Masson1.

Cell Centered Finite Volume Schemes for Multiphase Flow Applications

slide-12
SLIDE 12

Outline Applications and meshes Flux formulation Discrete variational framework Numerical Experiments

Cell centered schemes currently implemented in ArcGeoSim L and G schemes O scheme GradCell scheme

  • L. Agelas1, D. Di Pietro1, J. Droniou2, I. Kapyrin1, R. Eymard3, C. Guichard1, R. Masson1.

Cell Centered Finite Volume Schemes for Multiphase Flow Applications

slide-13
SLIDE 13

Outline Applications and meshes Flux formulation Discrete variational framework Numerical Experiments The L and G schemes

Model problem

Let Ω ⊂ Rd be a bounded polygonal domain For f ∈ L2(Ω), consider the following problem:

  • −div(ν∇u) = f in Ω,

u = 0 on ∂Ω Let a(u, v) =

  • Ω ν∇u · ∇v. The weak formulation reads

Find u ∈ H1

0(Ω) such that a(u, v) =

fv for all v ∈ H1

0(Ω)

(Π)

  • L. Agelas1, D. Di Pietro1, J. Droniou2, I. Kapyrin1, R. Eymard3, C. Guichard1, R. Masson1.

Cell Centered Finite Volume Schemes for Multiphase Flow Applications

slide-14
SLIDE 14

Outline Applications and meshes Flux formulation Discrete variational framework Numerical Experiments The L and G schemes

Model problem

Let {Ωi}1≤i≤NΩ be a partition of Ω into bounded polygonal sub-domains ν|Ωi smooth and ν(x) is s.p.d. for a.e. x ∈ Ω

  • L. Agelas1, D. Di Pietro1, J. Droniou2, I. Kapyrin1, R. Eymard3, C. Guichard1, R. Masson1.

Cell Centered Finite Volume Schemes for Multiphase Flow Applications

slide-15
SLIDE 15

Outline Applications and meshes Flux formulation Discrete variational framework Numerical Experiments The L and G schemes

Polyhedral admissible meshes

Th: set of cells K Eh = Ei

h ∪ Eb h : set of inner and boundary faces σ

mσ: surface of the face σ mK: volume of the cell K

  • L. Agelas1, D. Di Pietro1, J. Droniou2, I. Kapyrin1, R. Eymard3, C. Guichard1, R. Masson1.

Cell Centered Finite Volume Schemes for Multiphase Flow Applications

slide-16
SLIDE 16

Outline Applications and meshes Flux formulation Discrete variational framework Numerical Experiments The L and G schemes

Discrete function space Vh

Vh: space of piecewise constant functions on Th vh(x) = vK for all x ∈ K Equip Vh with the following discrete H1

0 norm:

∀vh ∈ Vh, vhVh =  

K∈Th

  • σ∈EK

mσ dK,σ |γσ(vh) − vK|2  

1/2

using the following trace reconstruction at the faces σ      γσ(vh) = if σ = vK dL,σ+vLdK,σ

dL,σ+dK,σ

if σ = EK ∩ EL ∈ Ei

h,

γσ(vh) = 0 if σ ∈ Eb

h .

  • L. Agelas1, D. Di Pietro1, J. Droniou2, I. Kapyrin1, R. Eymard3, C. Guichard1, R. Masson1.

Cell Centered Finite Volume Schemes for Multiphase Flow Applications

slide-17
SLIDE 17

Outline Applications and meshes Flux formulation Discrete variational framework Numerical Experiments The L and G schemes

Finite Volume Scheme

Let FK,σ(uh) denote a conservative linear approximation of

  • σ

ν∇u · nK,σ conservativity: FK,σ(uh) + FL,σ(uh) = 0, σ = EK ∩ EL ∈ Ei

h.

The finite volume scheme reads find uh ∈ Vh such that −

  • σ∈EK

FK,σ(uh) =

  • K

f for all K ∈ Th.

  • L. Agelas1, D. Di Pietro1, J. Droniou2, I. Kapyrin1, R. Eymard3, C. Guichard1, R. Masson1.

Cell Centered Finite Volume Schemes for Multiphase Flow Applications

slide-18
SLIDE 18

Outline Applications and meshes Flux formulation Discrete variational framework Numerical Experiments The L and G schemes

Discrete variational formulation

For all uh, vh ∈ Vh, let ah(uh, vh) =

  • K∈Th
  • σ∈EK

FK,σ(uh)(γσ(vh) − vK) =

  • σ=EK ∩EL∈Eh

FK,σ(uh)(vL − vK) −

  • K∈Th
  • σ∈EK ∩Eb

h

FK,σ(uh)vK The finite volume scheme is equivalent to: find uh ∈ Vh such that ah(uh, vh) =

fvh for all vh ∈ Vh.

  • L. Agelas1, D. Di Pietro1, J. Droniou2, I. Kapyrin1, R. Eymard3, C. Guichard1, R. Masson1.

Cell Centered Finite Volume Schemes for Multiphase Flow Applications

slide-19
SLIDE 19

Outline Applications and meshes Flux formulation Discrete variational framework Numerical Experiments The L and G schemes

Assumptions

Flux consistency in Q: for all ϕ ∈ Q with ϕh = {ϕ(xK)}K∈Th, limh→0

K∈Th

  • σ∈EK

dK,σ mσ |FK,σ(ϕh) − mσ ν∇ϕK · nK,σ|2 = 0 Coercivity of the bilinear form ah: ∀v ∈ Vh, ah(v, v) > ∼ v2

Vh

  • L. Agelas1, D. Di Pietro1, J. Droniou2, I. Kapyrin1, R. Eymard3, C. Guichard1, R. Masson1.

Cell Centered Finite Volume Schemes for Multiphase Flow Applications

slide-20
SLIDE 20

Outline Applications and meshes Flux formulation Discrete variational framework Numerical Experiments The L and G schemes

Well-posedness

Using the coercivity of ah and the discrete Poincaré inequality: uh2

Vh <

∼ ah(uh, uh) =

  • Ω fuh

< ∼ fL2(Ω)uhL2(Ω) < ∼ fL2(Ω)uhVh Stability of the scheme in Vh norm: uhVh < ∼ fL2(Ω)

  • L. Agelas1, D. Di Pietro1, J. Droniou2, I. Kapyrin1, R. Eymard3, C. Guichard1, R. Masson1.

Cell Centered Finite Volume Schemes for Multiphase Flow Applications

slide-21
SLIDE 21

Outline Applications and meshes Flux formulation Discrete variational framework Numerical Experiments The L and G schemes

Convergence

Consider a sequence of admissible FV discretizations with h → 0. Discrete Rellich theorem: there exist a subsequence and ˜ u ∈ H1

0(Ω) s.t.

uh → ˜ u in L2(Ω) and ∇huh ⇀ ∇˜ u weakly in (L2(Ω))d with ( ∇hvh)K = 1 mK

  • σ∈EK

mσ(γσ(vh) − vK)nK,σ Using the coercivity and the flux consistency assumptions and the density of Q in H1

0(Ω) we can deduce that

˜ u = u.

  • L. Agelas1, D. Di Pietro1, J. Droniou2, I. Kapyrin1, R. Eymard3, C. Guichard1, R. Masson1.

Cell Centered Finite Volume Schemes for Multiphase Flow Applications

slide-22
SLIDE 22

Outline Applications and meshes Flux formulation Discrete variational framework Numerical Experiments The L and G schemes

The L and G schemes

Consistent vertex-based gradient reconstruction to obtain fluxes L scheme (see [Aavatsmark et al., 2007], [Aavatsmark et al., 2008])

monotonicity enhancing tuning

G scheme: (see [Agélas et al., 2009])

coercivity-enhancing tuning

Convergence analysis: see [Agélas et al., 2009]

  • L. Agelas1, D. Di Pietro1, J. Droniou2, I. Kapyrin1, R. Eymard3, C. Guichard1, R. Masson1.

Cell Centered Finite Volume Schemes for Multiphase Flow Applications

slide-23
SLIDE 23

Outline Applications and meshes Flux formulation Discrete variational framework Numerical Experiments The L and G schemes

Groups

G = {G ⊂ EK ∩ Es s.t. cardG = d}

  • L. Agelas1, D. Di Pietro1, J. Droniou2, I. Kapyrin1, R. Eymard3, C. Guichard1, R. Masson1.

Cell Centered Finite Volume Schemes for Multiphase Flow Applications

slide-24
SLIDE 24

Outline Applications and meshes Flux formulation Discrete variational framework Numerical Experiments The L and G schemes

L-interpolation

G = {σ, σ′} TG = {K, L1, L2} Piecewise linear interpolation of uh on K, L1, L2 Full continuity of potential and normal fluxes at the faces σ and σ′ Amounts to solve a linear system AGX = b of size d with X the vector

  • f the d components of (∇huh)G

K.

  • L. Agelas1, D. Di Pietro1, J. Droniou2, I. Kapyrin1, R. Eymard3, C. Guichard1, R. Masson1.

Cell Centered Finite Volume Schemes for Multiphase Flow Applications

slide-25
SLIDE 25

Outline Applications and meshes Flux formulation Discrete variational framework Numerical Experiments The L and G schemes

Fluxes

Gσ = {G ∈ G|σ ∈ G} F G

K,σ = mσνK(∇huh)G K · nK,σ

FK,σ =

  • G∈Gσ

θG

σ F G K,σ with

  • G∈Gσ

θG

σ = 1, θG σ ≥ 0

L scheme: θG

σ are fixed, choice of the groups G to enhance the

monotonocity of the scheme ([Aavatsmark et al., 2007]) G scheme: keep all the groups and choose the θG

σ to enhance the

coercivity of the scheme

  • L. Agelas1, D. Di Pietro1, J. Droniou2, I. Kapyrin1, R. Eymard3, C. Guichard1, R. Masson1.

Cell Centered Finite Volume Schemes for Multiphase Flow Applications

slide-26
SLIDE 26

Outline Applications and meshes Flux formulation Discrete variational framework Numerical Experiments The L and G schemes

L-Groups (exemple for an half edge in 2D)

  • L. Agelas1, D. Di Pietro1, J. Droniou2, I. Kapyrin1, R. Eymard3, C. Guichard1, R. Masson1.

Cell Centered Finite Volume Schemes for Multiphase Flow Applications

slide-27
SLIDE 27

Outline Applications and meshes Flux formulation Discrete variational framework Numerical Experiments The L and G schemes

G-Weights I

G = {σ, σ′} TG = {K, L1, L2} Let Hh,G = {uK, uL1, uL2} Set ah,G(u, v) = F G

K,σ(u)(vL1 − vK) + F G K,σ′(u)(vL2 − vK)

Define γG = inf{u∈HTh,G , uVh =1} ah,G(u, u)

  • L. Agelas1, D. Di Pietro1, J. Droniou2, I. Kapyrin1, R. Eymard3, C. Guichard1, R. Masson1.

Cell Centered Finite Volume Schemes for Multiphase Flow Applications

slide-28
SLIDE 28

Outline Applications and meshes Flux formulation Discrete variational framework Numerical Experiments The L and G schemes

G-Weights II

For a given ǫ > 0 define

  • fǫ(x) =

ǫ2 ǫ−x

if x < 0, fǫ(x) = x + ǫ

  • therwise,

The weights are defined as θG

σ =

fǫ(γG)

  • G′∈Gσ fǫ(γG)

∀G ∈ G, ∀σ ∈ G. The larger γG, the more the group G contributes!

  • L. Agelas1, D. Di Pietro1, J. Droniou2, I. Kapyrin1, R. Eymard3, C. Guichard1, R. Masson1.

Cell Centered Finite Volume Schemes for Multiphase Flow Applications

slide-29
SLIDE 29

Outline Applications and meshes Flux formulation Discrete variational framework Numerical Experiments The L and G schemes

Convergence analysis of L and G schemes

Fluxes consistency in Q:

require to assume the stability of the L-interpolation i.e. (AG)−1 ≥ β

Coercivity assumption: mesh and ν dependent

sufficient computable condition using the submatrix around each vertex.

  • L. Agelas1, D. Di Pietro1, J. Droniou2, I. Kapyrin1, R. Eymard3, C. Guichard1, R. Masson1.

Cell Centered Finite Volume Schemes for Multiphase Flow Applications

slide-30
SLIDE 30

Outline Applications and meshes Flux formulation Discrete variational framework Numerical Experiments The GradCell scheme The O scheme

GradCell scheme: using a discrete variational framework [Agélas et al., 2008]

Discrete variational formulation ah(uh, vh) =

fvh for all vh ∈ Vh. with ah based on two cellwise constant gradients and stabilized by residuals a(uh, vh) =

  • K∈Th

mKνK(∇huh)K · ( ∇hvh)K +

  • K∈Th

ηK

  • σ∈EK

mσ dK,σ RK,σ(uh)RK,σ(vh)

  • L. Agelas1, D. Di Pietro1, J. Droniou2, I. Kapyrin1, R. Eymard3, C. Guichard1, R. Masson1.

Cell Centered Finite Volume Schemes for Multiphase Flow Applications

slide-31
SLIDE 31

Outline Applications and meshes Flux formulation Discrete variational framework Numerical Experiments The GradCell scheme The O scheme

Discrete gradients reconstructions

The cellwise constant gradients are obtained via the Green formula and trace reconstructions (∇hvh)K =

1 mK

  • σ∈EK mσ(IK,σ(vh) − vK)nK,σ

( ∇hvh)K =

1 mK

  • σ∈EK mσ(γσ(vh) − vK)nK,σ

Residuals: RK,σ(vh) = IK,σ(vh) − vK − (∇hvh)K · (xσ − xK)

  • L. Agelas1, D. Di Pietro1, J. Droniou2, I. Kapyrin1, R. Eymard3, C. Guichard1, R. Masson1.

Cell Centered Finite Volume Schemes for Multiphase Flow Applications

slide-32
SLIDE 32

Outline Applications and meshes Flux formulation Discrete variational framework Numerical Experiments The GradCell scheme The O scheme

Fluxes

Fluxes can be derived from the bilinear form using: ah(uh, vh) =

  • σ=EK ∩EL∈Eh

FK,σ(uh)(vL − vK) −

  • K∈Th
  • σ∈EK ∩Eb

h

FK,σ(uh)vK Stencil of the scheme: neighbours of the neighbours Example for topologicaly cartesian grids

13 cells in 2D 21 cells in 3D

  • L. Agelas1, D. Di Pietro1, J. Droniou2, I. Kapyrin1, R. Eymard3, C. Guichard1, R. Masson1.

Cell Centered Finite Volume Schemes for Multiphase Flow Applications

slide-33
SLIDE 33

Outline Applications and meshes Flux formulation Discrete variational framework Numerical Experiments The GradCell scheme The O scheme

Convergence analysis

Consistency of ∇huh for piecewise smooth functions with normal flux continuity: stability of the L-interpolation (AG)−1 ≥ β Coercivity assumption: a(vh, vh) > ∼ vh2

Vh

  • L. Agelas1, D. Di Pietro1, J. Droniou2, I. Kapyrin1, R. Eymard3, C. Guichard1, R. Masson1.

Cell Centered Finite Volume Schemes for Multiphase Flow Applications

slide-34
SLIDE 34

Outline Applications and meshes Flux formulation Discrete variational framework Numerical Experiments The GradCell scheme The O scheme

Remarks

Remark 1: a linear interpolation IK,σ can be used for smooth diffusion tensors ν Remark 2: choosing ∇huh = ∇huh yields a symmetric coercive scheme but at the price of a larger stencil (21 in 2D and 81 in 3D) see [Eymard and Herbin, 2007].

  • L. Agelas1, D. Di Pietro1, J. Droniou2, I. Kapyrin1, R. Eymard3, C. Guichard1, R. Masson1.

Cell Centered Finite Volume Schemes for Multiphase Flow Applications

slide-35
SLIDE 35

Outline Applications and meshes Flux formulation Discrete variational framework Numerical Experiments The GradCell scheme The O scheme

O scheme: notations [Agélas and Masson, 2008]

Choose xK ∈ K ⇒ unknown uK Choose xs

σ ∈ σ ⇒ unknown us σ

Choose ms

σ ≥ 0 such that s∈Vσ ms σ = mσ ⇒ subcell K s

ms

K = 1

d

  • σ∈EK ∩Es

ms

σdK,σ

Es: set of faces connected to s EK: set of faces of the cell K Vσ: set of vertices of the edge σ VK: set of vertices of the cell K

  • L. Agelas1, D. Di Pietro1, J. Droniou2, I. Kapyrin1, R. Eymard3, C. Guichard1, R. Masson1.

Cell Centered Finite Volume Schemes for Multiphase Flow Applications

slide-36
SLIDE 36

Outline Applications and meshes Flux formulation Discrete variational framework Numerical Experiments The GradCell scheme The O scheme

Discrete gradient

Piecewise constant gradient on each K s: (∇hu)s

K =

  • σ∈EK ∩Es

(us

σ − uK) gs K,σ

with gs

K,σ ∈ Rd such that

  • σ∈EK ∩Es

v · (xs

σ − xK) gs K,σ = v

for all v ∈ Rd.

  • L. Agelas1, D. Di Pietro1, J. Droniou2, I. Kapyrin1, R. Eymard3, C. Guichard1, R. Masson1.

Cell Centered Finite Volume Schemes for Multiphase Flow Applications

slide-37
SLIDE 37

Outline Applications and meshes Flux formulation Discrete variational framework Numerical Experiments The GradCell scheme The O scheme

Usual O scheme I

Assume that {xs

σ − xK}σ∈Es∩EK defines a basis of Rd

then (∇hu)s

K is uniquely defined by the gradient of the linear

interpolation of (uK, xK), (us

σ, xs σ)σ∈Es∩EK .

Subfluxes: F s

K,σ(uh) = ms σνK(∇hu)s K · nK,σ

  • L. Agelas1, D. Di Pietro1, J. Droniou2, I. Kapyrin1, R. Eymard3, C. Guichard1, R. Masson1.

Cell Centered Finite Volume Schemes for Multiphase Flow Applications

slide-38
SLIDE 38

Outline Applications and meshes Flux formulation Discrete variational framework Numerical Experiments The GradCell scheme The O scheme

Usual O scheme II: Hybrid Finite volume scheme

    

  • σ∈EK

s∈Vσ

F s

K,σ(uh)

  • =
  • K

f(x)dx for all K ∈ Th, F s

K,σ(uh) = −F s L,σ(uh)

for all s ∈ Vσ, σ = EK ∩ EL ∈ Ei

h.

(us

σ)σ∈Es are eliminated around each vertex s

in terms of the cell centered unknowns around s.

  • L. Agelas1, D. Di Pietro1, J. Droniou2, I. Kapyrin1, R. Eymard3, C. Guichard1, R. Masson1.

Cell Centered Finite Volume Schemes for Multiphase Flow Applications

slide-39
SLIDE 39

Outline Applications and meshes Flux formulation Discrete variational framework Numerical Experiments The GradCell scheme The O scheme

Usual O scheme: discrete hybrid variational formulation

Bilinear form on Hh: Hh =

  • (uK)K∈Th, (us

σ)σ∈Es,s∈Vh, s. t. us σ = 0 for all σ ∈ Eb h

  • .

ah(u, v) =

  • K∈Th
  • s∈VK

ms

K(∇hu)s K · νK(

∇hv)s

K

with ( ∇hu)s

K =

1 ms

K

  • σ∈EK ∩Es

ms

σ(us σ − uK)nK,σ.

Finite volume scheme: find uh ∈ Hh such that ah(uh, v) =

  • K∈Th

vK

  • K

f(x)dx for all v ∈ Hh.

  • L. Agelas1, D. Di Pietro1, J. Droniou2, I. Kapyrin1, R. Eymard3, C. Guichard1, R. Masson1.

Cell Centered Finite Volume Schemes for Multiphase Flow Applications

slide-40
SLIDE 40

Outline Applications and meshes Flux formulation Discrete variational framework Numerical Experiments The GradCell scheme The O scheme

Generalization: discrete hybrid variational formulation

Bilinear form on Hh: ah(u, v) =

  • K∈Th
  • s∈VK
  • ms

K(∇hu)s K · νK(

∇hv)s

K

+ αs

K

  • σ∈EK ∩Es

ms

K

(dK,σ)2 Rs

K,σ(u)Rs K,σ(v)

  • with

Rs

K,σ(u) = us σ − uK − (∇hu)s K · (xs σ − xK).

Finite volume scheme: find uh ∈ Hh such that ah(uh, v) =

  • K∈Th

vK

  • K

f(x)dx for all v ∈ Hh.

  • L. Agelas1, D. Di Pietro1, J. Droniou2, I. Kapyrin1, R. Eymard3, C. Guichard1, R. Masson1.

Cell Centered Finite Volume Schemes for Multiphase Flow Applications

slide-41
SLIDE 41

Outline Applications and meshes Flux formulation Discrete variational framework Numerical Experiments The GradCell scheme The O scheme

Hybrid Finite Volume scheme

ah(u, v) =

  • K∈Th
  • s∈VK
  • σ∈Es∩EK
  • σ′∈Es∩EK

(T s

K)σ,σ′(us σ′ − uK)(vs σ − vK),

Let us define the following subfluxes: F s

K,σ(u) =

  • σ′∈Es∩EK

(T s

K)σ,σ′(us σ′ − uK),

Hybrid Finite volume scheme:       

  • σ∈EK

s∈Vσ

F s

K,σ(uh)

  • =
  • K

f(x)dx for all K ∈ Th, F s

K,σ(uh) = −F s L,σ(uh)

for all s ∈ Vσ, σ = EK ∩ EL ∈ Ei

h.

  • L. Agelas1, D. Di Pietro1, J. Droniou2, I. Kapyrin1, R. Eymard3, C. Guichard1, R. Masson1.

Cell Centered Finite Volume Schemes for Multiphase Flow Applications

slide-42
SLIDE 42

Outline Applications and meshes Flux formulation Discrete variational framework Numerical Experiments The GradCell scheme The O scheme

Convergence analysis

Consistency of ∇hϕh for ϕ ∈ C∞

c (Ω) with

ϕh =

  • ϕ(xK), ϕ(xs

σ)

  • K∈Th,σ∈Es,s∈Vh

Coercivity assumption: a(vh, vh) > ∼ vh2

Hh

vHh =

  • K∈T
  • σ∈EK
  • s∈Vσ

ms

K

(dK,σ)2 (vs

σ − vK)21/2

.

  • L. Agelas1, D. Di Pietro1, J. Droniou2, I. Kapyrin1, R. Eymard3, C. Guichard1, R. Masson1.

Cell Centered Finite Volume Schemes for Multiphase Flow Applications

slide-43
SLIDE 43

Outline Applications and meshes Flux formulation Discrete variational framework Numerical Experiments The GradCell scheme The O scheme

Symmetric unconditionaly coercive cases

Simplicial cells Parallepipedic cells

for an ad hoc choice of xs

σ one has

(∇huh)s

K = (e

∇huh)s

K i.e. gs K,σ nK,σ

Symmetrization (C. Lepotier)

Choose (∇hu)s

K := (e

∇hu)s

K

Fluxes are not consistent except for the above cases

  • L. Agelas1, D. Di Pietro1, J. Droniou2, I. Kapyrin1, R. Eymard3, C. Guichard1, R. Masson1.

Cell Centered Finite Volume Schemes for Multiphase Flow Applications

slide-44
SLIDE 44

Outline Applications and meshes Flux formulation Discrete variational framework Numerical Experiments

Hexahedral meshes, Identity permeability tensor, u(x, y, z) = sin(πx) sin(πy) sin(πz) The grid

  • L. Agelas1, D. Di Pietro1, J. Droniou2, I. Kapyrin1, R. Eymard3, C. Guichard1, R. Masson1.

Cell Centered Finite Volume Schemes for Multiphase Flow Applications

slide-45
SLIDE 45

Outline Applications and meshes Flux formulation Discrete variational framework Numerical Experiments

Randomly distorted hexahedral meshes I

  • L. Agelas1, D. Di Pietro1, J. Droniou2, I. Kapyrin1, R. Eymard3, C. Guichard1, R. Masson1.

Cell Centered Finite Volume Schemes for Multiphase Flow Applications

slide-46
SLIDE 46

Outline Applications and meshes Flux formulation Discrete variational framework Numerical Experiments

ν = I L2 error on pressure L2 error on fluxes

  • L. Agelas1, D. Di Pietro1, J. Droniou2, I. Kapyrin1, R. Eymard3, C. Guichard1, R. Masson1.

Cell Centered Finite Volume Schemes for Multiphase Flow Applications

slide-47
SLIDE 47

Outline Applications and meshes Flux formulation Discrete variational framework Numerical Experiments

Randomly distorted hexahedral meshes II

  • L. Agelas1, D. Di Pietro1, J. Droniou2, I. Kapyrin1, R. Eymard3, C. Guichard1, R. Masson1.

Cell Centered Finite Volume Schemes for Multiphase Flow Applications

slide-48
SLIDE 48

Outline Applications and meshes Flux formulation Discrete variational framework Numerical Experiments

ν = I L2 error for pressure L∞ error for pressure

  • L. Agelas1, D. Di Pietro1, J. Droniou2, I. Kapyrin1, R. Eymard3, C. Guichard1, R. Masson1.

Cell Centered Finite Volume Schemes for Multiphase Flow Applications

slide-49
SLIDE 49

Outline Applications and meshes Flux formulation Discrete variational framework Numerical Experiments

Hybrid meshes Slice of the mesh Zoom in

  • L. Agelas1, D. Di Pietro1, J. Droniou2, I. Kapyrin1, R. Eymard3, C. Guichard1, R. Masson1.

Cell Centered Finite Volume Schemes for Multiphase Flow Applications

slide-50
SLIDE 50

Outline Applications and meshes Flux formulation Discrete variational framework Numerical Experiments

Hybrid meshes, Isotropic tensor ν. Pressure L2 error Flux L2 error

  • L. Agelas1, D. Di Pietro1, J. Droniou2, I. Kapyrin1, R. Eymard3, C. Guichard1, R. Masson1.

Cell Centered Finite Volume Schemes for Multiphase Flow Applications

slide-51
SLIDE 51

Outline Applications and meshes Flux formulation Discrete variational framework Numerical Experiments

Hybrid meshes, Isotropic tensor ν, performance with AMG Number of GMRES iterations vs grid size

  • r

r0 < 10−9

. No of cells O L G GradCell 2-point 12723 7 — 16 8 7 40847 7 — 85 9 7 254645 11 — — 15 8 813368 8 — — 11 8

  • L. Agelas1, D. Di Pietro1, J. Droniou2, I. Kapyrin1, R. Eymard3, C. Guichard1, R. Masson1.

Cell Centered Finite Volume Schemes for Multiphase Flow Applications

slide-52
SLIDE 52

Outline Applications and meshes Flux formulation Discrete variational framework Numerical Experiments

Hybrid meshes, Isotropic tensor ν, coercivity γh = ah(uh − πh(u), uh − πh(u)) uh − πh(u)2

Vh

No of cells O L G GradCell 2-point 12723 8.07e − 01 1.35e − 01 3.63e − 01 1.45e + 00 7.94e − 01 40847 8.29e − 01 8.52e − 02 8.65e − 01 1.51e + 00 8.02e − 01 254645 8.10e − 01 7.20e − 06 1.77e − 02 1.47e + 00 8.09e − 01 813368 8.10e − 01 – 3.40e − 02 1.47e + 00 8.21e − 01

  • L. Agelas1, D. Di Pietro1, J. Droniou2, I. Kapyrin1, R. Eymard3, C. Guichard1, R. Masson1.

Cell Centered Finite Volume Schemes for Multiphase Flow Applications

slide-53
SLIDE 53

Outline Applications and meshes Flux formulation Discrete variational framework Numerical Experiments

Anisotropic tensor ν = diag{1, 1, 0.1}. Pressure L2 error Flux L2 error

  • L. Agelas1, D. Di Pietro1, J. Droniou2, I. Kapyrin1, R. Eymard3, C. Guichard1, R. Masson1.

Cell Centered Finite Volume Schemes for Multiphase Flow Applications

slide-54
SLIDE 54

Outline Applications and meshes Flux formulation Discrete variational framework Numerical Experiments

Nearwell radial grids

  • L. Agelas1, D. Di Pietro1, J. Droniou2, I. Kapyrin1, R. Eymard3, C. Guichard1, R. Masson1.

Cell Centered Finite Volume Schemes for Multiphase Flow Applications

slide-55
SLIDE 55

Outline Applications and meshes Flux formulation Discrete variational framework Numerical Experiments

Anisotropic tensor ν = diag{1, 1, 0.05}. L2 error on pressure L2 error on fluxes

  • L. Agelas1, D. Di Pietro1, J. Droniou2, I. Kapyrin1, R. Eymard3, C. Guichard1, R. Masson1.

Cell Centered Finite Volume Schemes for Multiphase Flow Applications

slide-56
SLIDE 56

Outline Applications and meshes Flux formulation Discrete variational framework Numerical Experiments

Conclusion Case dependent results O scheme the most accurate but lacks of robustness for meshes with high aspect ratio (or anisotropy) combined with distorsion L and G schemes good on hexahedral meshes but may fail for Hybrid meshes GradCell exhibits the best robustness but requires two layers of communication in parallel

  • L. Agelas1, D. Di Pietro1, J. Droniou2, I. Kapyrin1, R. Eymard3, C. Guichard1, R. Masson1.

Cell Centered Finite Volume Schemes for Multiphase Flow Applications

slide-57
SLIDE 57

Outline Applications and meshes Flux formulation Discrete variational framework Numerical Experiments

References

Aavatsmark, I., Eigestad, G., Mallison, B., and Nordbotten, J. (2008). A compact multipoint flux approximation method with improved robustness.

  • Numer. Methods Partial Differential Equations, 24(5):1329–1360.

Aavatsmark, I., Eigestad, G., Mallison, B., Nordbotten, J., and ian, E. O. (2007). A compact multipoint flux approximation method with improved robustness.

  • Numer. Methods Partial Differential Equations, 1(31).

Agélas, L., Di Pietro, D. A., and Droniou, J. (2009). The G method for heterogeneous anisotropic diffusion on general meshes. M2AN Math. Model. Numer. Anal. Accepted for publication. Preprint available at http://hal.archives-ouvertes.fr/hal-00342739/fr. Agélas, L., Di Pietro, D. A., Eymard, R., and Masson, R. (2008). An abstract analysis framework for nonconforming approximations of anisotropic heterogeneous diffusion. Preprint available at http://hal.archives-ouvertes.fr/hal-00318390/fr. Submitted. Agélas, L. and Masson, R. (2008). Convergence of the finite volume MPFA O scheme for heterogenesous anisotropic diffusion problems on general meshes. In Eymard, R. and Hérard, J.-M., editors, Finite Volumes for Complex Applications V, pages 145–152. John Wiley & Sons. Eymard, R. and Herbin, R. (2007). A new colocated finite volume scheme for the icompressible Navier-Stokes equations on general non matching grids.

  • C. R. Math. Acad. Sci., 344(10):659–662.
  • L. Agelas1, D. Di Pietro1, J. Droniou2, I. Kapyrin1, R. Eymard3, C. Guichard1, R. Masson1.

Cell Centered Finite Volume Schemes for Multiphase Flow Applications