[priyankap]

Hello Everyone,


I am new to OpenFOAM and CFD. However, I have been reading about OpenFOAM and basics of CFD from a while now. I know the flow of solving a case in OpenFOAM which starts from making a geometry, then mesh and then solving it using a solver.



What I don't understand is,
1. How do we know that the solver has already solved my problem?



2. Where can I see the parameters changing?



3. I read that we know our problem is solved when solver converges. How can I check that the solver has converged already?


I have tried to understand the log which is generated by the solver but could really understand what it is happening inside it.

[RobertHB]

Originally Posted by priyankap

1. How do we know that the solver has already solved my problem

There a two ways in which you can end a simulation. a) by reaching the endTime specified in your controlDict (mostly transient flows) or b) if the Initial residual (part of the console or log output) of a variable falls below the value defined in the fvSolution residualControl entry (steadyState flows). When looking for a converged steadyState solution, a solution can be concidered convered if the residual is below 1e-4. With lower values meaning better convergence and a residual of 1e-6 usually being enough. Meaning you won't gain much from running your simulation any longer.

2. Where can I see the parameters changing?

You can observe the residuals in your console or log output. If you want the see a graphical representation of what is happening you will have to open your simulation results in paraview.

3. I read that we know our problem is solved when solver converges. How can I check that the solver has converged already? I have tried to understand the log which is generated by the solver but could really understand what it is happening inside it.

I'll attach a snippet of one of my own logs and try to explain.

Time = 19999

smoothSolver:  Solving for Ux, Initial residual = 0.00476975, Final residual = 8.58411e-09, No Iterations 36
smoothSolver:  Solving for Uz, Initial residual = 0.0665207, Final residual = 9.32058e-09, No Iterations 42
[...]
GAMG:  Solving for p, Initial residual = 0.115453, Final residual = 8.04117e-09, No Iterations 39
[...]

 ExecutionTime = 566.98 s  ClockTime = 577 s

Ux: variable being solved.
Init. res.: Residual value at the beginning of the timestep.
Final res.: Residual value at the beginning of the timestep. If the final residual falls below the residualControl value your simulation end.
ExecutionTime is the time spend by your computer actually doing something and ClockTime being the time since you started your simulation.


You can also use runtime post-processing by including #includeFunc residuals in your control dict. This will write a file with all residuals for every timestep, which then can be plotted during runtime using the foamMonitor utility. As such you can graphically observe your residuals decreasing.

[priyankap]

Hello RobertHB,


Thanks a lot for your reply, it cleared many of my doubts. I was trying to understand the physics behind the calculations.


Now my question is:


Is it possible to verify a simple hand calculated problem on OpenFOAM.For example, I want to realise the heat transfer in a metal rod attached by an aluminium plate on both sides. I know the temperature of both plates. So, can I realise this case in OpenFOAM? If yes, do I get the value of heat flow rate in the log file?



Please excuse me if it was a silly question.



Thank you

[RobertHB]

Originally Posted by priyankap

Now my question is:


Is it possible to verify a simple hand calculated problem on OpenFOAM.For example, I want to realise the heat transfer in a metal rod attached by an aluminium plate on both sides. I know the temperature of both plates. So, can I realise this case in OpenFOAM?

Sorry, i cant help you with this as i'm not dealing with this kind of simulation myself. But it sounds awfully complex to set up a simulation of you can calculate the results by hand.

If yes, do I get the value of heat flow rate in the log file?

No. You would need to conduct some kind of post-processing e.g. using Paraview/Matlab. But i cant help you further since i dont know anything about heatflow.