BGK Model Standard method Presentation of the method (1D) 2D method Conclusions A new adaptative numerical method for kinetic equation Stéphane Brull, Louis Forestier-Coste, Luc Mieussens MNMCFF2014 - Beijing May 21-27, 2014 Forestier-Coste local velocity grid 1/ 49 BGK Model Standard method Presentation of the method (1D) 2D method Conclusions Introduction rarefied gas flow Deterministic simulations (DVM) of Boltzmann equation (BGK) A global discrete velocity grid in space and time For practical applications in aerodynamics (atmospheric re-entry problems), the grid is so large that the computational ressources (memory storage and CPU time) require by the simulation are huge Project : construction of local velocity grids (unsteady flows) Forestier-Coste local velocity grid 2/ 49 BGK Model Standard method Presentation of the method (1D) 1 BGK Model 2 Standard method 3 Presentation of the method (1D) 4 2D method 5 Conclusions Forestier-Coste local velocity grid 2D method Conclusions 3/ 49 BGK Model Standard method Presentation of the method (1D) 1 BGK Model 2 Standard method 3 Presentation of the method (1D) 4 2D method 5 Conclusions Forestier-Coste local velocity grid 2D method Conclusions 4/ 49 BGK Model Standard method Presentation of the method (1D) 2D method Conclusions f (t, x, v ) distribution function : mass of particles at time t, at position x ∈ Ω ⊂ RD with a velocity of v ∈ RD . Forestier-Coste local velocity grid 5/ 49 BGK Model Standard method Presentation of the method (1D) 2D method Conclusions f (t, x, v ) distribution function : mass of particles at time t, at position x ∈ Ω ⊂ RD with a velocity of v ∈ RD . Z mass density ρ(t, x) = f (t, x, v )dv RD Forestier-Coste local velocity grid 5/ 49 BGK Model Standard method Presentation of the method (1D) 2D method Conclusions f (t, x, v ) distribution function : mass of particles at time t, at position x ∈ Ω ⊂ RD with a velocity of v ∈ RD . Z mass density ρ(t, x) = f (t, x, v )dv RD velocity Forestier-Coste u(t, x) = 1 ρ(t, x) Z v f (t, x, v )dv RD local velocity grid 5/ 49 BGK Model Standard method Presentation of the method (1D) 2D method Conclusions f (t, x, v ) distribution function : mass of particles at time t, at position x ∈ Ω ⊂ RD with a velocity of v ∈ RD . Z mass density ρ(t, x) = f (t, x, v )dv RD velocity u(t, x) = 1 ρ(t, x) Z energy E (t, x) = RD Forestier-Coste Z v f (t, x, v )dv RD kv k2 f (t, x, v )dv 2 local velocity grid 5/ 49 BGK Model Standard method Presentation of the method (1D) 2D method Conclusions f (t, x, v ) distribution function : mass of particles at time t, at position x ∈ Ω ⊂ RD with a velocity of v ∈ RD . Z mass density ρ(t, x) = f (t, x, v )dv RD velocity 1 u(t, x) = ρ(t, x) Z energy E (t, x) = RD temperature Forestier-Coste 1 T (t, x) = DR Z v f (t, x, v )dv RD kv k2 f (t, x, v )dv 2 2E (t, x) − ku(t, x)k2 ρ(t, x) local velocity grid 5/ 49 BGK Model Standard method Presentation of the method (1D) 2D method Conclusions f (t, x, v ) distribution function : mass of particles at time t, at position x ∈ Ω ⊂ RD with a velocity of v ∈ RD . Z mass density ρ(t, x) = f (t, x, v )dv RD velocity 1 u(t, x) = ρ(t, x) Z v f (t, x, v )dv RD Z energy temperature Forestier-Coste kv k2 f (t, x, v )dv 2 RD Z 1 kv − u(t, x)k2 f (t, x, v )dv T (t, x) = DRρ(t, x) RD E (t, x) = local velocity grid 5/ 49 BGK Model Standard method Presentation of the method (1D) 2D method Conclusions Notations collisional invariants : m(v ) = 1, v , 21 kv k2 moments : ρ = (ρ, ρu, E )T Z < g >= g dv T RD ρ =< mf > Forestier-Coste local velocity grid 6/ 49 BGK Model Standard method Presentation of the method (1D) 2D method Conclusions BGK model ∂t f + v .∇x f = Q(f ) Q(f ) = 1 (M[ρ,u,T ] − f ) τ + boundaries conditions Forestier-Coste local velocity grid 7/ 49 BGK Model Standard method Presentation of the method (1D) 2D method Conclusions Maxwellian Maxwellian M[ρ,u,T ] (t, x, v ) : solution of an entropy minimisation problem (H(f ) =< f ln(f ) >). H(M[ρ,u,T ] ) = min{H(g ), g ≥ 0 s.t. < mg >= ρ} M[ρ,u,T ] (t, x, v ) = Forestier-Coste ρ(t, x) D (2πRT (t, x)) 2 |v − u(t, x)|2 exp − 2RT (t, x) local velocity grid 8/ 49 BGK Model Standard method Presentation of the method (1D) 1 BGK Model 2 Standard method 3 Presentation of the method (1D) 4 2D method 5 Conclusions Forestier-Coste local velocity grid 2D method Conclusions 9/ 49 BGK Model Standard method Presentation of the method (1D) 2D method Conclusions Discretization Physical model: ∂t f + v · ∇x f = Q(f ) Velocity space RD velocities + BCs V = vk , k ∈ ND set of discrete Discrete kinetic equation: ∂t fk + vk · ∇x fk = Qk (f ) Forestier-Coste local velocity grid + BCs 10/ 49 BGK Model Standard method Presentation of the method (1D) 2D method Conclusions Velocity grid? Cartesian grid V: 0 8 1 0 1 0 19 l∆vx l 0 : lmax = < V = vl,m,n = vmin + @m∆vy A , @mA = @0 : mmax A : ; n∆vz n 0 : nmax Constraints: bounds and grid step : the grid must be large enough fine enough Forestier-Coste local velocity grid 11/ 49 BGK Model Standard method Presentation of the method (1D) 2D method Conclusions How to design a cartesian velocity grid? p 2 RT (x) Information on the support of f (x, .): assuming f close to its local Maxwellian, then centered p on u(x) with standard deviation RT (x) p u(x) − c RT (x) u(x) p u(x) + c RT (x) The grid contains all the distributions if the bounds are at least p p vmin = min{u(x) − c RT (x)} vmax = max{u(x) + c RT (x)} x∈Ω x∈Ω (c around 4) At least 3 points into the ”core” of each distribution. The grid step should be: p ∆v ≤ min{ 2RT (x)} x∈Ω Forestier-Coste local velocity grid 12/ 49 BGK Model Standard method Presentation of the method (1D) 2D method Conclusions Estimation of the bounds and the step of the grid Temperature and velocity fields are a priori unknown! several solutions : use the boundaries values: upstream flow (u∞ , T∞ ) and velocity and temperature of the body (uwall , Twall ) Flow parameter inside the shock: ushock and Tshock estimated by Rankine-Hugoniot relations. Even better: A compressible Navier-Stokes pre-simulation : u NS and T NS in Ω let vmin = min{u x∈Ω NS (x) − c p RT NS (x)} vmax = max{u NS (x) + c x∈Ω p ∆v = a min{ 2RT NS (x)} p RT NS (x)} x∈Ω Forestier-Coste local velocity grid 13/ 49 BGK Model Standard method Presentation of the method (1D) 2D method Conclusions Drawback of a global grid For reentry problem : the velocity is very large (≈ 6.000 m/s), the temperature is very large in the shock (∼ 105 K), and very small in the upstream and at the boundary (∼ 102 K) Consequently: very large grid bounds, and very small step, so very large number of discrete velocities Example: Mach 20, altitude 90 km: V contains 52 × 41 × 41 points. Around 350 GB memory requirements with a coarse 3D mesh in space!... Forestier-Coste local velocity grid 14/ 49 BGK Model Standard method Presentation of the method (1D) 2D method Conclusions Need of dynamics grids : example two interacting blastwaves: à t = 0 ρ = 1, u = 0, T = [1000, 0.01, 100] √ max (u + 4 RT ) = 126 √ at t = 0.02 (after the interaction): max (u + 4 RT ) = 236: larger bounds! the optimal velocity grid get 2551 points! Forestier-Coste local velocity grid 15/ 49 BGK Model Standard method Presentation of the method (1D) 2D method Conclusions Main Idea 1 BGK Model 2 Standard method 3 Presentation of the method (1D) Main Idea Illustration Extension of the LVG 4 2D method 5 Conclusions Forestier-Coste local velocity grid 16/ 49 BGK Model Standard method Presentation of the method (1D) 2D method Conclusions Main Idea Local velocity grids idea: define a velocity grid for each t and x Forestier-Coste local velocity grid 17/ 49 BGK Model Standard method Presentation of the method (1D) 2D method Conclusions Main Idea Local velocity grid: problems 1 how to define the Local Velocity Grid: bounds and step? 2 how to exchange informations between two grids? Example: ∂x f (t, x, v ) ≈ f (t, x + ∆x, v ) − f (t, x, v ) ∆x but f (t, x, .) and f (t, x + ∆x, .) are not defined on the same velocity grid... Forestier-Coste local velocity grid 18/ 49 BGK Model Standard method Presentation of the method (1D) 2D method Conclusions Main Idea Definition of a LVG: conservation laws at tn : space cell [xi− 1 , xi+ 1 ] 2 2 n ≈ f (t , x , v n ), where fi,k n i i,k n o n n n n vi,k ∈ Vin = vi,1 , vi,2 , . . . , vi,K n i local velocity grid (with Kin points) Uin ≈ U(tn , xi ) with a quadrature on the LVG: Z m(v )f (t n , xi , v ) dv = hmf (t n , xi , .)i U(t n , xi ) = (ρ, ρu, E )(t n , xi ) = R ↓ Kin Uin = (ρni , ρni uin , Ein ) = X n n n m(vi,k )fi,k ωi,k = hmfi n iV n i k=1 where m(v ) = (1, v , 12 |v |2 ). Forestier-Coste local velocity grid 19/ 49 BGK Model Standard method Presentation of the method (1D) 2D method Conclusions Main Idea Definition of a LVG: conservation laws approximation of U(t n+1 , xi ) by discrete conservation laws: conservation laws ∂t U + ∂x hvmf i = 0 finite volume upwind scheme n n Uin+1 − Uin Φi+ 21 − Φi− 12 + =0 ∆t ∆x where the numerical fluxes are defined by n Φni+ 1 = v + mfi n V n + v − mfi+1 Vn 2 Forestier-Coste i local velocity grid i+1 20/ 49 BGK Model Standard method Presentation of the method (1D) 2D method Conclusions Main Idea Definition of a LVG approximations of u(tn+1 , xi ) and T (tn+1 , xi ): uin+1 and Tin+1 Forestier-Coste local velocity grid 21/ 49 BGK Model Standard method Presentation of the method (1D) 2D method Conclusions Main Idea Definition of a LVG approximations of u(tn+1 , xi ) and T (tn+1 , xi ): uin+1 and Tin+1 bounds of Vin+1 set to q n+1 vmin,i = uin+1 − 4 RTin+1 and n+1 vmax,i = uin+1 + 4 q RTin+1 remark: corrects bounds if f close of its local Maxwellian Forestier-Coste local velocity grid 21/ 49 BGK Model Standard method Presentation of the method (1D) 2D method Conclusions Main Idea Definition of a LVG approximations of u(tn+1 , xi ) and T (tn+1 , xi ): uin+1 and Tin+1 bounds of Vin+1 set to q n+1 vmin,i = uin+1 − 4 RTin+1 and n+1 vmax,i = uin+1 + 4 q RTin+1 remark: corrects bounds if f close of its local Maxwellian new discrete velocity grid Vin+1 : uniform with a constant number of points (Kin+1 = 10 à 30) remark: variables step and number of points could be necessary (discontinuous f ). Forestier-Coste local velocity grid 21/ 49 BGK Model Standard method Presentation of the method (1D) 2D method Conclusions Main Idea Computation of fi n+1 kinetic equation: ∂t f + v ∂x f = Q(f ) approximation of f (tn , xi , v ): fi n (v ) Forestier-Coste local velocity grid 22/ 49 BGK Model Standard method Presentation of the method (1D) 2D method Conclusions Main Idea Computation of fi n+1 kinetic equation: ∂t f + v ∂x f = Q(f ) approximation of f (tn , xi , v ): fi n (v ) finite volume - upwind scheme (continuous v ): n n (v ) − fi+1 (v ) − fi n (v ) fi n+1 (v ) − fi n (v ) + fi n (v ) − fi−1 +v +v = Q(fi n )(v ) ∆t ∆x ∆x n and f n not defined on the same problem: fi n+1 , fi n , fi−1 i+1 grids. Forestier-Coste local velocity grid 22/ 49 BGK Model Standard method Presentation of the method (1D) 2D method Conclusions Main Idea Computation of fi n+1 reconstruction procedure: for each i, reconstruction of fi n on its LVG Vin fin ( f¯i n (v ) = R(fi n )(v ) 0 n n if vmin,i ≤ v ≤ vmax,i f¯in (v) else, n vmin,i n vi,k n vmax,i numerical scheme: n f n (v ) − fi−1 (v ) fi n+1 (v ) − fi n (v ) f n (v ) − fi n (v ) + v+ i + v − i+1 = Q(fi n )(v ) ∆t ∆x ∆x Forestier-Coste local velocity grid 23/ 49 BGK Model Standard method Presentation of the method (1D) 2D method Conclusions Main Idea Computation of fi n+1 reconstruction procedure: for each i, reconstruction of fi n on its LVG Vin fin ( f¯i n (v ) = R(fi n )(v ) 0 n n if vmin,i ≤ v ≤ vmax,i f¯in (v) else, n vmin,i n vi,k n vmax,i numerical scheme: n f n (v ) − fi−1 (v ) fi n+1 (v ) − fi n (v ) f n (v ) − fi n (v ) + v+ i + v − i+1 = Q(fi n )(v ) ∆t ∆x ∆x n+1 for every vi,k of Vin+1 : n+1 n+1 n+1 n+1 ¯n n+1 ¯n ¯n ¯n n+1 fi,k − f¯i n (vi,k ) n+1 + fi (vi,k ) − fi−1 (vi,k ) n+1 − fi+1 (vi,k ) − fi (vi,k ) + vi,k + vi,k ∆t ∆x ∆x n+1 n ¯ = Q(fi )(vi,k ) Forestier-Coste local velocity grid 23/ 49 BGK Model Standard method Presentation of the method (1D) 2D method Conclusions Main Idea Advantages of this scheme 1 advantage 1: the local velocity grid adapt in time and space to the local temperature T and velocity u 2 advantage 2: only initial values for u and T are required Forestier-Coste local velocity grid 24/ 49 BGK Model Standard method Presentation of the method (1D) 2D method Conclusions Main Idea Properties of the numerical scheme 1 if R positive, then the scheme is positive, provided that ∆t satisfies a standard CFL condition Tin+1 positive Forestier-Coste local velocity grid 25/ 49 BGK Model Standard method Presentation of the method (1D) 2D method Conclusions Main Idea Properties of the numerical scheme 1 if R positive, then the scheme is positive, provided that ∆t satisfies a standard CFL condition Tin+1 positive 2 under this slight modification on the quadrature: hfi n iV n := hf¯i n i i Tin+1 is positive Forestier-Coste local velocity grid 25/ 49 BGK Model Standard method Presentation of the method (1D) 2D method Conclusions Main Idea Properties of the numerical scheme 1 if R positive, then the scheme is positive, provided that ∆t satisfies a standard CFL condition Tin+1 positive 2 under this slight modification on the quadrature: hfi n iV n := hf¯i n i i Tin+1 is positive 3 Forestier-Coste in practice: even with non positive R, the scheme generally preserves the positivity local velocity grid 25/ 49 BGK Model Standard method Presentation of the method (1D) 2D method Conclusions Illustration 1 BGK Model 2 Standard method 3 Presentation of the method (1D) Main Idea Illustration Extension of the LVG 4 2D method 5 Conclusions Forestier-Coste local velocity grid 26/ 49 BGK Model Standard method Presentation of the method (1D) 2D method Conclusions Illustration Illustration 1: Sod test case 0,6 0,5 0,4 velocity profile: 0,3 0,2 0,1 0 0 0,1 0,2 0,3 0,4 0,5 ... : global velocity grid (wrong bounds, with initial data) - : global velocity grid (correct bounds, 100 points: bounds and step estimated by an Euler computation, then a grid convergence study) .- : local velocity grid (10 points) Forestier-Coste local velocity grid 27/ 49 BGK Model Standard method Presentation of the method (1D) 2D method Conclusions Illustration Illustration 2: two interracting blast waves before the interaction (top), after (bottom) T u 3 p 20 400 2,5 15 300 2 10 1,5 200 5 1 100 0 0,5 0 0 0,2 0,4 0,6 0,8 1 4 -5 0 0,2 0,4 0,6 0,8 1 0 20 600 18 550 0 0,2 0,4 0,6 0,8 1 0 0,2 0,4 0,6 0,8 1 3 16 500 14 450 2 12 400 1 10 350 8 0 0 0,2 0,4 0,6 0,8 1 0 0,2 0,4 0,6 0,8 1 300 • global velocity grid: 2531 points (bounds and step estimated by an Euler computation , then grid convergence study) • local velocity grid: 30 points Forestier-Coste local velocity grid 28/ 49 BGK Model Standard method Presentation of the method (1D) 2D method Conclusions Illustration Reconstruction procedure piecewise polynomial interpolation high order interpolation is necessary discontinuous distributions ( large Knudsen number), a non oscillatory method is necessary (ENO interpolation used) Forestier-Coste local velocity grid 29/ 49 BGK Model Standard method Presentation of the method (1D) 2D method Conclusions Illustration Influence of the reconstruction step: Sod test ρ u T 1 2 0,8 1,8 0,8 1,6 0,6 1,4 0,6 1,2 0,4 0,4 1 0,2 0,2 0,8 0,6 0 -1 -0,5 0 0,5 1 0 -1 -0,5 0 0,5 1 -1 -0,5 0 0,5 1 -0,5 0 0,5 1 3 1 0,8 2,5 0,8 0,6 2 0,6 1,5 0,4 0,4 1 0,2 0,2 0,5 0 -1 -0,5 0 0,5 1 0 -1 -0,5 0 0,5 1 0 -1 Free transport regime (no collisions): top: global grid (30 points) bottom: local velocity grid (30 points): affine interpolation, ENO3, ENO4 Forestier-Coste local velocity grid 30/ 49 BGK Model Standard method Presentation of the method (1D) 2D method Conclusions Illustration Boundaries conditions − diffuse reflection: f (t, x = 0, v > 0) = − hvhv+ Mfwi i Mw ghost cell approach Mw on its own local velocity grid (based on the wall temperature and velocity) however, such reflection can generate very discontinuous and non-symmetric distributions! Forestier-Coste local velocity grid 31/ 49 BGK Model Standard method Presentation of the method (1D) 2D method Conclusions Illustration Boundaries conditions example: Heat transfert problem ρ0 t = 0 u0 = 0 T0 = TL TL x=0 x=0 Forestier-Coste local velocity grid TR x=1 32/ 49 BGK Model Standard method Presentation of the method (1D) 2D method Conclusions Illustration Boundaries conditions example: Heat transfert problem ρ0 t = 0 u0 = 0 T0 = TL TL x=0 x=1 x=0 for small t, the support of f (t, x ≈ 1, .) is non symmetric: Forestier-Coste p local velocity grid TR RTR p RTL 32/ 49 BGK Model Standard method Presentation of the method (1D) 2D method Conclusions Illustration Boundaries conditions example: Heat transfert problem ρ0 t = 0 u0 = 0 T0 = TL TL x=0 TR x=0 x=1 for small t, the support of f (t, x ≈ 1, .) is non symmetric: p RTR p RTL local velocity grid (symmetric by construction) is not large enough for v < 0 (−1900 instead of −2200). 0 -10 -20 -30 -40 -50 0 0,1 0,2 0,3 0,4 0,5 0,6 (Kn = 0.01, global velocity grid=100 points, LVG=30 points) Forestier-Coste local velocity grid 32/ 49 BGK Model Standard method Presentation of the method (1D) 2D method Conclusions Extension of the LVG 1 BGK Model 2 Standard method 3 Presentation of the method (1D) Main Idea Illustration Extension of the LVG 4 2D method 5 Conclusions Forestier-Coste local velocity grid 33/ 49 BGK Model Standard method Presentation of the method (1D) 2D method Conclusions Extension of the LVG Extension of the LVG: algorithm n+1 n+1 compute fi,k for every vi,k of Vin+1 n+1 w = vi,1 (leftmost point) loop w = w − ∆vin+1 compute fi n+1 (w ) by the numerical scheme: n f¯n (w ) − f¯i−1 (w ) fi n+1 (w ) − f¯i n (w ) f¯n (w ) − f¯i n (w ) +w + i +w − i+1 = Q̄(fi n )(w ) ∆t ∆x ∆x if fi n+1 (w ) is too large then add w to the grid and continue the loop else stop Forestier-Coste local velocity grid 34/ 49 BGK Model Standard method Presentation of the method (1D) 2D method Conclusions Extension of the LVG Extension of the LVG: illustration Heat transfert problem: T 1000 u p 3 0 900 -10 2,5 800 -20 700 2 600 -30 500 1,5 -40 400 300 0 0,1 0,2 0,3 0,4 0,5 0,6 900 -50 0 0,1 0,2 0,3 0,4 0,5 0,6 1 0 0,1 0,2 0,3 0,4 0,5 0 0,1 0,2 0,3 0,4 0,5 0,6 3 0 800 -10 2,5 700 -20 600 2 -30 500 1,5 -40 400 300 0 0,1 0,2 0,3 0,4 0,5 0,6 -50 0 0,1 0,2 0,3 0,4 0,5 0,6 1 no extension, symmetric LVG (top) with extension (bottom) Kn = 0.01, global velocity grid=100 points, LVG=30 points Forestier-Coste local velocity grid 35/ 49 BGK Model Standard method Presentation of the method (1D) 2D method Conclusions 2D preliminary results 1 BGK Model 2 Standard method 3 Presentation of the method (1D) 4 2D method 2D preliminary results Embedded local velocity grid Illustration Mesh Refinment 5 Conclusions Forestier-Coste local velocity grid 36/ 49 BGK Model Standard method Presentation of the method (1D) 2D method Conclusions 2D preliminary results To the 2D reduct model : F (t, x, y , vx , vy , vz ) f (t, x, y , vx , vy ) = hF iz g (t, x, y , vx , vy ) = hF |vz |2 iz 2 cartesian velocity grid reconstruction : 2D ENO interpolation Forestier-Coste local velocity grid 37/ 49 BGK Model Standard method Presentation of the method (1D) 2D method Conclusions 2D preliminary results Heat transfert problem on a cylinder: outside wall temperature : 1000K inside wall and domain temperature : 300K gas at equilibrium Forestier-Coste local velocity grid 38/ 49 BGK Model Standard method Presentation of the method (1D) 2D method Conclusions 2D preliminary results Heat transfert problem on a cylinder: outside wall temperature : 1000K inside wall and domain temperature : 300K gas at equilibrium Forestier-Coste local velocity grid 38/ 49 BGK Model Standard method Presentation of the method (1D) 2D method Conclusions 2D preliminary results computational time too long Each velocity of each cell interpolated at least 8 times. Forestier-Coste local velocity grid 39/ 49 BGK Model Standard method Presentation of the method (1D) 2D method Conclusions 2D preliminary results computational time too long Each velocity of each cell interpolated at least 8 times. ⇒ new definition of local velocity grids Forestier-Coste local velocity grid 39/ 49 BGK Model Standard method Presentation of the method (1D) 2D method Conclusions Embedded local velocity grid 1 BGK Model 2 Standard method 3 Presentation of the method (1D) 4 2D method 2D preliminary results Embedded local velocity grid Illustration Mesh Refinment 5 Conclusions Forestier-Coste local velocity grid 40/ 49 BGK Model Standard method Presentation of the method (1D) 2D method Conclusions Embedded local velocity grid Embedded LVG : definiton √ u ± 4 RT reference velocity step. Forestier-Coste local velocity grid 41/ 49 BGK Model Standard method Presentation of the method (1D) 2D method Conclusions Embedded local velocity grid Embedded LVG : definiton √ u ± 4 RT reference velocity step. reference point for the grids : (0, 0). Forestier-Coste local velocity grid 41/ 49 BGK Model Standard method Presentation of the method (1D) 2D method Conclusions Embedded local velocity grid Embedded LVG : definiton √ u ± 4 RT reference velocity step. reference point for the grids : (0, 0). take into account the accuracy wished. Forestier-Coste local velocity grid 41/ 49 BGK Model Standard method Presentation of the method (1D) 2D method Conclusions Embedded local velocity grid Need of auto extension cellule 1130 -43.4373 cellule 1130 -12.412 -41.4344 -12.4799 2954.9 2482.1 v_y v_y -2482.1 -2009.3 3427.7 -2954.9 v_x -2363.9 without Forestier-Coste v_x 3427.7 with local velocity grid 42/ 49 BGK Model Standard method Presentation of the method (1D) 2D method Conclusions Illustration 1 BGK Model 2 Standard method 3 Presentation of the method (1D) 4 2D method 2D preliminary results Embedded local velocity grid Illustration Mesh Refinment 5 Conclusions Forestier-Coste local velocity grid 43/ 49 BGK Model Standard method Presentation of the method (1D) 2D method Conclusions Illustration Shock wave on a cylinder high velocity wave coming from left. Knudsen number around 1. Forestier-Coste local velocity grid 44/ 49 BGK Model Standard method Presentation of the method (1D) 2D method Conclusions Illustration T 2567 2000 1000 236 Forestier-Coste local velocity grid 45/ 49 BGK Model Standard method Presentation of the method (1D) 2D method Conclusions Illustration Error 0.057 0.04 0.02 0 Forestier-Coste local velocity grid 45/ 49 BGK Model Standard method Presentation of the method (1D) 2D method Conclusions Illustration N 63 60 50 40 30 17 Forestier-Coste local velocity grid 20 45/ 49 BGK Model Standard method Presentation of the method (1D) 2D method Conclusions Mesh Refinment 1 BGK Model 2 Standard method 3 Presentation of the method (1D) 4 2D method 2D preliminary results Embedded local velocity grid Illustration Mesh Refinment 5 Conclusions Forestier-Coste local velocity grid 46/ 49 BGK Model Standard method Presentation of the method (1D) 2D method Conclusions Mesh Refinment If extrem rarefied then far from equilibrium. Distributions can be multimodal. last method : good boundaries but lot of points to see the thinest mode AMR idea : take the boundaries from last method. refine only where the mode are thin. Forestier-Coste local velocity grid 47/ 49 BGK Model Standard method Presentation of the method (1D) 1 BGK Model 2 Standard method 3 Presentation of the method (1D) 4 2D method 5 Conclusions Forestier-Coste local velocity grid 2D method Conclusions 48/ 49 BGK Model Standard method Presentation of the method (1D) 2D method Conclusions Conclusions smaller velocity grid time-consuming high order ENO interpolation Embedded LVG Perspectives better extension of 2D LVG 3D Forestier-Coste local velocity grid 49/ 49