Dear colleagues.
the hydrodynamic of a system has converged via fluent in a flow time tantamount to a minute.I want to add reaction equations to system and having hydrodynamic as a initial guess instead of initializing.
while hydrodynamic may change due to viscosity changes, I must solve them simultaneously.
What do you deem to be convergence time approximately?I mean It may take 1 minute of flow time again ?or shorter?(our reaction is not time dependent)
to be honest when I look as mathematical aspect I expect 1 minute because equation must be solved again.
but by looking at physical facet I speculate it to be lower than that because of known velocity
here is what i would do when trying to solve unsteady (transient) flow for first time on application im not familiar with. i would monitor flow variables such as Temperature, pressure and velocity in two ways. fist i want to insure that at every time step the solution is converged . that means i have taken enough sub iteration to get converged solution. typically the monitors of flow variables should flatten out. when you take next time step you get a jump then they must flatten out. If this does not happen you are not convering your equations sufficient at each time step. Second , you also monitor flow but now verses time. watch for trend. does your flow eventuslly repeat or reach some expected trend. its hard to answer the last part precisely because it differ depending on the flow problem you are solving.
Dear Laith
many thanks
your answer is very cogent and cause a firm precedent to trade off between time step and iterations but convergence in a time step is vehemently depend on your desired error.so convergence in each time step seems to be a time consuming way.by this way you may set you time step small in order to lower changes by a time step or use high residuals.
I think convergence in each time step may cause an superfluous time while we it can converge faster
and my question is unanswered.
as an example our hydrodynamic has converged in 70 seconds. I have added reaction equations to this reactor without initializing. when it should be expected to converge?another 70 second or less?
Hi Mahdi,
the reason I mention the subiteration because once you turn on the reaction what you have set for convergence criterion for the hydro part of the system my not be enough with reaction. The rule of thumb is it will take more effort to converge hydro plus reaction than hydro alone. But its impossible to tell you how long because all depend on the reaction type and model and flow physics involve. But it is not unrealistic to expect doubling in time to convergence.
thanks
LZ
Convergence criteria is the limit of uncertainity you are giving for your solution. More the accuracy the time taken for solving will increase and I suppose with numerical techniques/ algorithms crunching numbers, it is imossible to get exact solution. So relaxation is given in achieving solutions. these relaxation factors when properly addressed will give you solution. Suppose if you look in terms of non adaibatic process, continuity convergence can be relaxed . by default the values are about 10e-6. Change and see ..... this is what I am doing it from last 15 years with various versions of fluent.
By the way when reactions are involved, chemical kinetics as per arhenius rate of reactions govern the reaction rate rather than convergence parameters. You can use CANTERA / STANJAN Code. More useful for reacting fluids.
All the best
- Badges
- Science topic