The unmatched grids in CFX4.2
 

Hi everybody

I have some problems in the unmatched grids used in CFX4.2.

1. How does CFX4.2 codes calculate the orientation

change to interpret the gradient vector in the new

block? Can somebody give me the formula used in

this calculation?

2. How does CFX4.2 correct the estimates of convection

coefficients calculated from both side of the

umatched grid interface to enforce mass

conservation ? I have check the mass conservation

on the unmatched grids in my computation. I found

that the mass conservation is not archieved. Why?

Isn't it really enforced in all conditions?

Darcy

Re: The unmatched grids in CFX4.2
 

Hi Darcy,

I have been doing some sliding-mesh simulations (in CFX 4.3). When using the sliding-mesh option, you also have to deal with unmatched grids. The manual gives quite some useful information about unmatched grid issues. Have you read all this? My experience is (in compliance with the manual) that the "AMG General" solver is very good in conserving mass.

As far as I understood from the manual the convection through the unmatched interface is calculated through simple 1st order interpolation, using the first layer of cells directly connected to the unmatched interface only. That is, higher order schemes should not improve your solution.

I hope this helps a bit. If not, please be a little bit more specific about your problem: do you use sliding-mesh, what solver do you use, what discretization do you use, etc.

Cheers,

Niels

Re: The unmatched grids in CFX4.2
 

Thank you, Neil

I use the BLOCK STONE solver. I do not use AMG because it takes much more time. However I'll try it again for your suggestion.

The descretization I used is higher order upwind scheme in convection term. Since only first order accurace scheme can be used between two sets of unmatched cells, does it matter whether I use higher order scheme or not?

Darcy

Excuse me!
 

I forget some thing,Neil.

I do use sliding grids. There is a rotating compartment(propeller) and stationary parts (wing and nacelle)in my problem.

Re: Excuse me!
 

Hi Darcy,

My name is Niels and not Neil, but hey nevermind.

When you use higher order upwind, I think you'll only obtain higher order accuracy some cells away from the unmatched grid. To me it is not clear at all what cells are used in the higher order discretisation near the unmatched grid. If I were you I would stick to the hybrid scheme.

The AMG solver is only useful for the pressure equation, because it is much more accurate. If your convergence criteria are based on mass conservation, it will actually be faster than the Stone solver, because it needs fewer iterations.

With respect to the sliding grid, it is recommended to choose your timesteps in such way that you slide exactly one grid cell at a time. In that way you minimize the effect of interpolation.

By the way, in CFX 4.4 the "multiple frames of reference" method is implemented. This method might speed up your calculations considerably. This may be useful to you.

Kind regards,

Niels

Re: Excuse me!
 

Thank you, Niels

First I'm very sorry for incorrectly typing your name. Secondly I think that a lower order accurace scheme is applied only on those cell face neighbor to the unmatched grid interface as long as the higher order upwind is employed for convection term. It's because a dummy cell can be the second cell in the higher order upwind scheme when the scheme is applied on the face of the second cell from unmatched grid interface.

Thank's for your regard. I always use AMG to solve all the system equations. Therefore it takes me too much time. I will try to used it only on pressure equation and see if I'll get better mass conservation.

Darcy