CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   Main CFD Forum (http://www.cfd-online.com/Forums/main/)
-   -   Finite volume boundary condition (http://www.cfd-online.com/Forums/main/15141-finite-volume-boundary-condition.html)

shuo May 2, 2008 06:58

Finite volume boundary condition
 
I am solving the conservation form of the Euler equations using cell-centered finite volume method.

U + dF/dx + dG/dy = 0

At the wall boundaries I set the derivative of the flux to zero i.e dF/dx = 0 |y = wall. However the cell centered velocity for the node adjacent to the wall does not approach zero despite uniform grid refinement. Why is that?

Shuo


keith May 2, 2008 10:44

Re: Finite volume boundary condition
 
zero flux does not mean zero velocity. zero flux basically means no different in velocity accross the wall boundary. you may want to try a no-slip boundary instead.

no slip boundary u_i = - u_i-1

that way, you can ensure that velocity AT the boundary is zero, and hence the velocity near the boundary will approach 0.


shuo May 2, 2008 18:43

Re: Finite volume boundary condition
 
I did that for the fluxes.

i.e for

dU/dt_(i, j) = -1/delta_x(F_(i+0.5, j) - F_(i-0.5, j)) - 1/delta_y(G_(i, j + 0.5) - G_(i, j - 0.5))

at wall | (i + 0.5, j)

F_(i+0.5, j) =

[ 0 P_(i - 0.5, j) //boundary condition 0 0 ]

G_(i, j + 0.5) =

// stays the same

[rho_(i, j+0.5)*v_(i, j+0.5) rho_(i, j+0.5)*u_(i, j+0.5)*v_(i, j+0.5) rho_(i, j+0.5)*v_(i, j+0.5)*v_(i, j+0.5) + P_(i, j+0.5) (e_(i, j+0.5) + P_(i, j+0.5))*v_(i, j+0.5) ]


Rami May 4, 2008 08:13

Re: Finite volume boundary condition
 
Since you solve Euler (rather NS) equations, there is slip at walls, so you should not expect zero velocity near the wall.

Tushar May 4, 2008 13:16

Re: Finite volume boundary condition
 
no....not for the fluxes...just create a layer of dummy cells along the boundary and change the direction (reverse by 180 degrees) for the NORMAL component at that point in dummy cell. Keep rest things same.viz. in c:

for(i=1;i<=nx;i++)

{

mirror_dp[i]=dp[i][1];

mirror_pp[i]=pp[i][1];

mirror_up[i]=up[i][1];

mirror_vp[i]=-vp[i][1];

mirror_ep[i]=ep[i][1];

mirror_ap[i]=ap[i][1];

//printf("\n%lf %lf %lf %lf %lf %lf",dp[i][4],pp[i][4],up[i][4],vp[i][1],ap[i][1],ep[i][1]);

} Also, for cell centered, as the information is stored at the cell center AND NOT AT THE WALL ITSELF, it might be one of the reasons not to get zero velocity at the wall.mail me at a.patil@iitg.ernet.in if reqd.

Shuo May 4, 2008 22:08

Re: Finite volume boundary condition
 
i am modelling a supersonic impinging jet and the problems occurs for across wall jet. i have about 40- 50 grid points across this jet. Is this enough resolution?

Shuo

jinwon park May 5, 2008 11:01

Re: Finite volume boundary condition
 
For the concept of "dummy cells", I recommend you "computational fluid dynamics: principles and applications" written by J. Blazek, 2nd edition. It help you understand the implementation of boundary conditions.


All times are GMT -4. The time now is 05:58.