[zx-81]

Hi,

I implemented a solver for incompressible two-phase flow with high density contrast (rho_gas/rho_liquid = 1/1000).
It works but the pressure solve (Multigrid Preconditioned CG) takes about 10x as much time than the single-phase solver (free surface, rho_gas=0).
The reason is the highly ill conditioned system matrix for the combined air-liquid pressure solve.

So I would like to discouple the pressure correction in two separate poisson solves for each phase. But then, how to best "couple" the two system ?

Unfortunately, for this approach the literature I found so far is scarce(most papers deal with using the appraoch "single-fluid" with variable density for each of the phases)

So I would ask the forum for any links, hints.

Thanks in advance for any advice.

[arjun]

Originally Posted by zx-81

Hi,

I implemented a solver for incompressible two-phase flow with high density contrast (rho_gas/rho_liquid = 1/1000).
It works but the pressure solve (Multigrid Preconditioned CG) takes about 10x as much time than the single-phase solver (free surface, rho_gas=0).
The reason is the highly ill conditioned system matrix for the combined air-liquid pressure solve.

So I would like to discouple the pressure correction in two separate poisson solves for each phase. But then, how to best "couple" the two system ?

Unfortunately, for this approach the literature I found so far is scarce(most papers deal with using the appraoch "single-fluid" with variable density for each of the phases)

So I would ask the forum for any links, hints.

Thanks in advance for any advice.

The problem with de-coupling of pressure equation is that it could be very unstable because of something what i call islands in matrix. This is group of equations that have no connection to other groups. Which leads to multiple pure neumann sub matrices and all ill conditioned.



This you do not get in single pressure approach where whole matrix is single island.