CFD Online Logo CFD Online URL
www.cfd-online.com
[Sponsors]
Home > Forums > Main CFD Forum

Solver suggestion?

Register Blogs Members List Search Today's Posts Mark Forums Read

Reply
 
LinkBack Thread Tools Display Modes
Old   February 5, 2001, 08:49
Default Solver suggestion?
  #1
Oleg Melnik
Guest
 
Posts: n/a
Dear All, I am solving 1D multiphase flow in a pipe. The system of equations looks as follows:

d rho_i/dt + d rho_i V_j/dx =G_i (i=1..4, j=1,2) discontinuity equation

0 = d p(rho_i)/dx +d (mu(rho_i) d V1/dx)/dx +Fmas momentum for main phase, inertia is small.

0= d p(rho_i)/dx K(V2-V1) - momentum for dispersed phase.

Boundary conditions: rho_i (0) are given, V1(0)=V2(0), -p(L)+2*(mu*dV1/dx) (L)=-pa.

Without viscous terms the system could be discretized and linearized on 2 points. This brings to equation:

A_i*deltaU_i B_i*deltaU_i-1=F_i (1)

Also substituting deltaU_i with deltaU_i-1 repetitively one can get

deltaU_n=Phi*deltaU_0+theta.

These condition multiplied by discretized BC gives additional condition at x=0. Then using (1) U_i could be found. Without viscous terms (both in equations and BC) the method I use gives good convergence. If I add viscous term to BC convergence slows down dramatically and sometimes it diverges. In the method I use it's not possible to include viscous terms in the equations.

The questions is how to include viscous term (implicitly!) and how to force the convergence.

Best regards Oleg
  Reply With Quote

Old   February 5, 2001, 11:53
Default Re: Solver suggestion?
  #2
John C. Chien (Alter Ego)
Guest
 
Posts: n/a
Have you tryed using a deferred correction for your viscous terms. (or part of your viscous term to keep your stencil compact.)
  Reply With Quote

Old   February 6, 2001, 03:37
Default Re: Solver suggestion?
  #3
Oleg Melnik
Guest
 
Posts: n/a
Dear John, I am not strong in CFD terminology. Do you mean taking viscous term from previous time-step or previous iteration. Yes I have done both. It leads to divergence in the computation. The problem also is that viscous term is small in most of the flow and becomes dominant near the boundary x=L. This makes(?) possible to include it only into a boundary conditions. But having derivative there slows convergence dramatically (from 2 to 30 iterations). Could it be so because equations are not properly linked with boundary conditions?

Thank you, Oleg
  Reply With Quote

Old   February 6, 2001, 19:15
Default Re: Solver suggestion?
  #4
John c. Chien (Alter Ego)
Guest
 
Posts: n/a
What is the solver you are using??
  Reply With Quote

Old   February 7, 2001, 03:50
Default Re: Solver suggestion?
  #5
Oleg Melnik
Guest
 
Posts: n/a
Dear John,

I have briefly discussed the method that I use to solve the problem in my first message. I have an equation in the form: G(U_i,U_i-1)=0 without viscous terms. After linearisation it comes to

A_i*deltaU_i B_i*deltaU_i-1=F_i or deltaU_i = P_i* deltaU_i-1-Q_i (P=A^-1B,Q=A^-1F) (1)

Removing intermediate terms I can calculate Phi and theta so that:

deltaU_n=Phi*deltaU_0+theta. (2)

I also have a boundary condition which after linearisation comes to

alpha* deltaU_n+beta=0. (3)

in both cases pressure or stress is equal to atmospheric. (For the last case I also use (1) to calculate deltaU_n-1 as a function of delta_U_n). I multiply then (2) by (3) and get additional relation for deltaU_0. Then I solve (1) with all parameters known at the left boundary.

That's the method I use. It converges very fast in case of given pressure and slow in case of stress biundary condition.

Oleg
  Reply With Quote

Old   February 7, 2001, 13:53
Default Re: Solver suggestion?
  #6
John c. Chien (Alter Ego)
Guest
 
Posts: n/a
If your using an implicit method then you give the final linear equation to a linear solver to be solved. What is this linear solver??? CGSTAB ?? SIP ??? SOR??? If your solution is diverging then it may be a good idea to check whether your matrix is positive definite and diagonally dominate, but these types of things really depend on the solver you are using.

1) Check what solver you are using 2) find out what conditions this solver requires

for convergence 3) check to see if these conditions are met.

  Reply With Quote

Old   February 8, 2001, 03:34
Default Re: Solver suggestion?
  #7
Oleg Melnik
Guest
 
Posts: n/a
Dear John,

I just do exactly what I describe in previous message no special solver. Just compute matrixes, inverse it and multiply as described. Possibly it's better to use standard solver for linear equations but I do it strait forward. I'll check diagonal dominance and positive definition of matrixes.

Thank you for comments

Oleg
  Reply With Quote

Reply

Thread Tools
Display Modes

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are On
Pingbacks are On
Refbacks are On


Similar Threads
Thread Thread Starter Forum Replies Last Post
A New Solver for Supersonic Combustion nakul OpenFOAM 6 March 30, 2014 03:00
Quarter Burner mesh with periosic condition SamCanuck FLUENT 2 August 31, 2011 11:34
Working directory via command line Luiz CFX 4 March 6, 2011 21:02
Suggestion for the usage of appropriate solver gandesk OpenFOAM 5 November 10, 2010 13:00
why the solver reject it? Anyone with experience? bearcat CFX 6 April 28, 2008 14:08


All times are GMT -4. The time now is 22:53.