Learning Forum

[Ramesh Chandra Mishra]

Hi

I am trying to run a simulation both on my PC and HPC but velocity profile at the inlet is comletely different in both the results. My simulation involves a UDF. Can someone point out how this can be fixed?

I am attaching my journal file for reference.

 

define/user-defined/compiled-functions load "libudf"
/file/read-case STD_k-epsilon_2.cas.h5
/define/boundary-conditions/velocity-inlet inlet no no yes yes yes yes "udf" "velocity_profile::libudf" no 0 no 300 yes yes yes "udf" "k_profile::libudf" yes yes "udf" "dissip_profile::libudf" yes no 0
solve/initialize/compute-defaults/velocity-inlet/inlet
solve/initialize/initialize-flow
/solve/iterate 10000
quit

/file

write-case-data

y

quit
quit
quit
quit
quit
quit
quit
quit
quit
quit
quit
quit
quit

exit

 

[Rob]

Some images may help. Have you opened both cases to see what's going on? 

As an aside missing the first / on the line opens you up for all sorts of fun if you happen to be in the wrong menu level when you trigger the journal. 

[Ramesh Chandra Mishra]

Case for the files is same. I have  checked for velocity profile at the inlet and its showing absolutely 0 for the inlet for HPC results whereas when run on PC , its showing the correct profile. I am attaching some pictures

[Rob]

So, the UDF profile didn't load on the HPC. Did you recompile the .c file on the cluster?

[Ramesh Chandra Mishra]

Yes I did and it did compile because I can see velocity at other portions of my domain but just at the inlet its zero. 

[Ramesh Chandra Mishra]

I think I am writing boundary conditions wrong in the jornal file. Can you somehow confirm whether the command in journal file is correct for booundary conditions.

[Rob]

If the inlet is zero how is flow happening elsewhere? 

The command varies based on which models are on, open a case locally and see what happens when you try the command.  One tip, in TUI journals the first /  for /define forces the full command to run from root, define/whatever if launched from non-root level will fail. 

[Ramesh Chandra Mishra]

I am attaching my contours for reference. 

And I didn't get what you said about the commands. Can you please elaborate it?

[Rob]

As a guess you've got a pressure boundary and buoyancy? 

You need to check the command. The exact form of the TUI commands (the yes/no bits) is very dependent on what other models are active. So, the turbulence model may add an extra yes/no and some number inputs. 

In the TUI the command /define/something can be run from any level of the TUI menu.  define/something will only work if run from root (top level). 

[Ramesh Chandra Mishra]

No buoyancy. I have got velocity inlet and pressure outlet. In velocity inlet, I have specified UDF.

[Rob]

Looking at the image it's not hooked up correctly, so you need to check the command. I'd also be looking at the HPC result to understand how you got the flow pattern you're seeing. 

[Ramesh Chandra Mishra]

Hi Rob

Will it be okay if I send my case file and you could just verify if the command is right. I have struggling with this for quite a long time. I have tried but can't figure out what is wrong.

[Rob]

Run the journal line by line locally. Does it work? Does Fluent think it's looking for the local path for libudf, or the full path? Have you recompiled on the HPC noting that Linux/Windows have different libraries, as do each release of Fluent. 

We're not able to review files, it's one of the rules we need to comply with for Export Control. 

[Ramesh Chandra Mishra]

Ok Rob. i will try what you suggested. Just a query.

In order to run the simulations on HPC, I do not load UDF in my case file. So in boundary conditions, velocity magnitude remains zero. I am attaching a screenshot for the same. Could that be a reason?

[Rob]

Yes, if the inlet is set as zero that's the velocity that Fluent will use. 

[Ramesh Chandra Mishra]

Okay. But I cannot run my simulations with UDF loaded in case file. It gives errors. I have tried that.

Can you tell me the correct way?

 

[Ramesh Chandra Mishra]

PS: My velocity profile is specified in the UDF.

[Rob]

If you don't load the UDF then you won't get the same velocity profile and so aren't modelling the same conditions. 

What errors are you getting? 

[Ramesh Chandra Mishra]

Hi Rob.

I got the correct results. The problem was with the inlet boundary conditions . Like you said, velocity magnitude zero will give the zero results. I corrected it.

Thanks.