CFD Online Discussion Forums (https://www.cfd-online.com/Forums/)
-   CFX (https://www.cfd-online.com/Forums/cfx/)
-   -   time steps in transient diffusion problem (https://www.cfd-online.com/Forums/cfx/21071-time-steps-transient-diffusion-problem.html)

 Anurag March 10, 2005 17:41

time steps in transient diffusion problem

Dear all,

i m doing a transient simulation of drug diffusion in tissues, which acts as a porous media. drug is injected from a point source and injection velocity is zero.

linear resistance coefficient of the porous media is very high so the process of drug distribution is very slow. total time of injection is 24 hrs. (86400 secs) what time step should i select (considering memory and time of run)? i m starting with small time step of 10 sec and finally going up to 1000 sec. is that ok ? with this selection, RMS residual of mass fraction converges (1e-5). but pressure does not come below 1e-4.

thanks a lot for any suggestion.

Anurag

 Jeff March 10, 2005 22:41

Re: time steps in transient diffusion problem

Look in the FORTRAN examples at the end of the Solver Modeling manual. There's some handy user Fortran for an adaptive time stepping scheme. If your coefficient loops are few (say less than three) you can set a multiplier to increase the time step size. If it's not converging in say 8 or 9 coefficient loops, it will reduce the time step.

I've modified it a bit to limit by a maximum time step, and not to increase until at least 10 steps in a row have met the minimum coefficient loops target (for stability) and it works pretty good. My problems have convection as well, but your diffusion problem aught to be able to work its way up to some large time steps as the gradients are reduced.

Jeff

 All times are GMT -4. The time now is 21:23.