Hi,all,

I have met a problem on realizing UDF on parallel cluster. I used the No.17 example in Fluent 6.3.26 tutorial guide. It is a VOF model simulation. The UDF defines the initial velocity of the ink. I interpreted the UDF on serial PC, it works well.

I tested it on a cluster, which I have to write a script to run Fluent. That script works fine for unsteady flow. The structure of the script is below:

************************************************ ~sofu/bin/fluent_qs Enter Fluent case name (inkjet):INKJET Is this a 2D or 3D model (3D/2D): 2D Double or single precision run? (d/s): D Enter number of nodes requested (10): 10 Assuming that Fluent's auto-partitioning scheme is satisfactory for this run. If not, modify batch.jou file for other partitioning options before submitting the run.

Steady-state or time-dependent solution? (ss/td): TD Enter maximum number of iterations for this run (1): 100

Do you want use only the 2GB big memory nodes? (n/y): N Enter maximum wall-clock in HH:MM:SS format: 25:00:00 You may include other qsub options such as '-m bea' for e-mail notification, or '-q shared' for quick access to 8 shared nodes for debugging Enter other queuing options, if any:

Ready to submit your job to queueing system. Would you like to proceed? (y/n): Y **********************************************

Things after ":" is needed to be entered. After this process, the job is submitted to calculate. However, I can not use this script to calculate the case including the UDF. Even I put 'C file' and 'mesh file' in my folder on the cluster and configure the 'case and data files' by interpreting the UDF. I could not get any result except for 10 (equal to the number of the nodes) 'H files'.

I am wondering what the problem is. The UDF is absolutely right. Maybe it is because the script, any suggestion about this is appreciated.

Haoyin

The above script is not very clear, I resent it:

**************************************************

~sofu/bin/fluent_qs Enter Fluent case name (inkjet):INKJET

Is this a 2D or 3D model (3D/2D): 2D

Double or single precision run? (d/s): D

Enter number of nodes requested (10): 10

Assuming that Fluent's auto-partitioning scheme is satisfactory for this run. If not, modify batch.jou file for other partitioning options before submitting the run.

Steady-state or time-dependent solution? (ss/td): TD

Enter maximum number of iterations for this run (1): 100

Do you want use only the 2GB big memory nodes? (n/y): N

Enter maximum wall-clock in HH:MM:SS format: 25:00:00 You may include other qsub options such as '-m bea' for e-mail notification, or '-q shared' for quick access to 8 shared nodes for debugging Enter other queuing options, if any:

Ready to submit your job to queueing system. Would you like to proceed? (y/n): Y

****************************************

Thanks