For simulating a time periodic flow, I increased simulation times up to 100s and use adaptive time steps.
But I didn't get desire result...I checked simulation setup several times...I don't know why CFX results are different from Analytical solution?
CFX simulation take about 20 days.
*may be by increase the simulation time get a better result!
what do you think?
Hello
Depend on your problem you have to make sure about the following points:-
1- make sure you have good mesh (bad mesh will lead to bad results)
2- the model you used and discritization for your governing equations since you need to solve partial differential equations you need to choose the best order of magnitude please see CFX or FLUENT manual.
3- the residual setting and (pressure, momentum, energy damping factors )
4- try to check the convergence criteria using UDF or set any output (heat transfer coefficient, average area mass, CD, Cl ...ect)
hope these tips will help you a lot.
There is something which is called "The sampling Theorem". This essentially states, that if you have a periodic time signal, than accuracy of its representation is getting lost, if a period of the transient signal is not resolved by at least a minimum of "sampling points". In this case the frequency of te signal cannot be any longer be represented accurately. This theorem comes initially from measurement work. But it almost equally applies to CFD results as well, where the "sampling points" are your time-steps of the solution.
I guess, that your criterion for the transient adaptive time-stepping is too weak. Try to make the time-steps smaller (by factor of at least 2) and see in which direction the results develop. Mesh resolution of course needs to be able to represent the transient wave with desired accuracy as well.
Regards,
Dr. Th. Frank.
Diaelhag Khalifa Thanks for your attention...I consider all items you said...but I think I need a method or analysis type that find Boundary Value Problem for a specified frequency....A hormonic or periodic problem....Do you have any idea?...CFX can solve some problem like this?
Thanks for your answer
Thomas Frank Thanks Dr. Th. Frank. for your attention....I understand what you mean...I think I need a method or analysis type that solve a periodic or harmonic solution...in fact I need to solve the Boundary value problem part of solution....I know the frequency...CFX able to solve some problem like this?....how?...If as you said I make time step smaller...simulation take more time to complete...I should reduce simulation duration time...As a transient analysis I need to simulate for 100 seconds or more to reach the desire answer... I used a sinusoid function for wave maker displacement...now I find that there are an option in wall boundary condition for a periodic displacement....just we enter the displacement amplitude in each direction and the frequency...I want to know if I use a periodic displacement, it's affect simulation duration time? or I can use steady state analysis type to reach the solution?....I know the solution is time dependence...I mean the time periodic solution....I'm waiting for your answer...
thanks for your answer
Saeed Pashazanousi : To be honest, I cannot see how this problem should be able to be solved in a steady-state simulation. From the information given by you so far it is a time-dependent flow simulation, which needs to be properly resolved in space and time to reach an accurate solution of the problem.
If you talk about "solution", which physical quantity is your desired solution quantity? I.e. what is your target, which you try to predict form this type of simulation (if not the transient solution of the flow in it's entirety)?
Thomas Frank Thanks for your attention. I'm working on a wave maker linear theory. it's my master thesis. my main purpose is validation of an analytical solution which obtain from specified boundary conditions and frequency. Actually I want to compare a time periodic solution which obtain from an analytical solution with CFX results which obtain from a numerical simulation. However I need control the CFX simulation time to get desire result. But because of long duration of simulation time(a transient result for 100 seconds or even more) I'm looking for an other solution if exist. With this analysis configuration it took more than 200 days with my desktop Computer. I did it one time for a fixed time step(0.01s) for 100s transient simulation but I didn't get the desire result. It takes about 2 weeks. I know solution is a time dependence and need a transient analysis type. I've just hope may be exist a shorter method.
Thanks for the time you spent
Probably no other method exists with CFD methods (to my knowledge). You should consider to use a parallel Linux Cluster, where probably the flow calculation can be crunched to a percentage of your given calculation time.
Regards,
Dr. Th. Frank.
- Badges
- Science topic