Hi

1- you can decrease The number of mesh but it must be Appropriate. 

2-I think you can use parallel processing, even it is much complex when UDF is adopted. Parallel in Ansys CFX, you can 3 or 4 core of system,

3- yes, your system is weak for simulation. my system has 24 RAM and 8 core, you change your system, Laptop is weak. you use Pc.

4- Do you use symmetry boundary??If your geometry is Symmetric you should modeling half the geometry, so your speed is better.

Good luck!

Hi

1- you can decrease The number of mesh but it must be Appropriate. 

2-I think you can use parallel processing, even it is much complex when UDF is adopted. Parallel in Ansys CFX, you can 3 or 4 core of system,

3- yes, your system is weak for simulation. my system has 24 RAM and 8 core, you change your system, Laptop is weak. you use Pc.

4- Do you use symmetry boundary??If your geometry is Symmetric you should modeling half the geometry, so your speed is better.

Good luck!

oh great, thank you.. I can try to choose option 2 and option 3. let me try then I would contact you back !

you are welcome, Good Luck!

Hi Muhammad,

Another approach for speeding up your simualtion could be to simplify your case. For example, if you incorporate turbulence modelling in your simulation switch off the solution of turbulence equation at the beginning until a steady solution is found and then add it again. The same could be used for a multi-phase flow. Of course, this is only feasible for some cases.

In terms of UDFs it should not make a difference between parallel and serial computation, because the UDF code is shared between all nodes. You have to be careful only if you want access flow data of the total domain, then the individual nodes typically have to pass their data to the host node.

Depending on your setup, you could also change some settings to improve convergence: when convergence per time step is achieved with fewer iterations your simulation time should decrease, too. Usually FOU schemes show better convergence behaviour then higher order schemes (but FOU are less accurate too) and you could also modify the underrelaxation factors to improve convergence.

Good luck.

Regards,

Stephan

Dear Stephan C. Kaiser,

Thank you for your suggestions. To click off the turbulence models , I think it is not appropriate for my case.

Do you mean that computation does not speed up if I use Parallel processing ??

Nowadays, I am using parallel processing and experienced little speed up of my simulation. Because my laptop 4-processors are just of 1.7 GHz, so did not improve more speed. Recently, I am going to upgrade my laptop processors speed up to 2GHz with parallel 8-processors and 16GB RAM. I am sure that computation will boost up.. WHAT IS YOUR OPINION ??

Please guide.. 

Regards..

Dear Muhammad,

I totally agree that you will speed up your simulation with the parallel processing as well as with the increased processeor power and memory. In previous projects, we could often cut the turnaround time into halves by doubling the number of CPU cores, but it also depends on your case. In general, the more cores you can use for your simulation the faster it should get, but keep the following in mind: the gain in speed typically decrease beyond a certain number of cores because then the communication between the cores can become a limiting factor (especially if you use a bunch of PCs within a network). According to my own experiences and advices from the ANSYS support, a good rule of thumb is to use about one core per 100-200k control volumes. Perhaps, you can find some more helpful information on ANSYS support website: http://www.ansys.com/Solutions/Solutions-by-Role/IT-Professionals/Platform-Support

Good luck.

Regards,

Stephan

Thank you Dear Stephan C. Kaiser, the above link has much useful information..

Regards