Hello dear online comunity,
i am new to CFD and we are currently involved in a project work for our masters degree.
Since the problem we decided to simulate is quite complex, we cannot rely on the basic ANSYS fluent functionalitys being taught.
We want to avoid simplyfying and discretising the problem to do a (exhausting) manual steady investigation of our transient problem.
I need help to evaluate the possibility of implementing a udf, aswell as basic tips to how i can get into udf programming.
First i will present the problem:
We are simulating a tube bundle heat exchanger, where the pipes are filled with so called "phase change material" short PCM.
Its supposed to be used as a mobile heating unit. Melting point of PCM is around 42,5°C inlet temperature of our working fluid (air) is 18°C.
The aim is to find out what heating performance is like over time.
We ruled out using the "solidification and melting" feature provided by fluent, scince it takes too long.
So we created several models of a single pipe with a coat of hardened PCM on the inside varying in thickness.
All of these Models where simulated in steady conditions for different inlet temperatures.
The results were exported into matlab to get function fits for 1. the heat flux at a pipes outer surface and 2. the temperature at the pipes outer surface.
Functions are dependend on the percentage of hardened PCM (which i will call PHI) and the Temperature coming at the Pipe.
Now i want to implement this into ANSYS fluent, i dont know if it is doable or if I´m on the right way.
Please give me your assesment if my idea is doable. Remember im new to all of this so my idea might be quite naive.
Thank you very much in advance!!!
Greetings Philipp
The attachement contains a pseudo code i came up with...
Hi Philipp,
interesting project :). From my point of view, the subroutine you implemented in Matlab should be easily transferred into a UDF, especially if you use the third option given in the attached pseudo UDF. My advice would be to go through the examples in the manual. There is a special chapter for programming UDFs. You should hereby also consider the parallelization of your UDF, which could make things very complicated in case you need a lot of communication or synchronization of the computational nodes. However, the task is surely doable.
Since you didn't use the Solidification and Melting option, I am wondering if you considered the amount of latent heat of fusion. Depending on the material this will be not negligible. One way to do that would be to use the apparent heat capacity approach.
Best of luck!
Antoni Artinov
Hi there,
thank you very much for your response Antoni Artinov , you certainly motivated me to try implementing the udf!!
So first i´d like to point out that i followed given examples from the udf-manual and cfd-related forums. I came past the compilation of my code. Now scince youve already mentioned the parallelization complications i wonder if you can help me at my current state.
I will briefly explain the way i implemented my idea:
1. Im working with a simple model to test udfs consisting of two rows of pipes with a cutting plane in between - to measure flow temperature coming at the top pipe row. (reminder: i have a function fit for the pipe surface temperature, dependent on gas temperature and percentage of hardened pcm (PHI))
2. the first udf uses the execute_at_end macro to calculate the current PHI-Values and print it to a txt file as memory.
3. the second udf (used twice with a different name) uses a define_profile macro to set the boundary temperature of the pipe-rows. It reads the corresponding Phi.txt file and gets the gas-temperature by calculating the average temperature at the corresponding cutting plane.
So i have a working model wich i tested in a transient simulation using constant boundary conditions. (btw: ive done this twice- for parallel + double precision; and serial + No dp)
After compiling, i can hook the first udf as execute_at_end. As soon as i set the second udf as boundary condition for one of the pipe-rows i get error messages in the TUI, aswell as graphical bugs in the initialization and run-calculation windows. I will attach the fully commented code.
The error i´m getting reads (latest run in serial) :
1. Node 0: Process 9896: Received signal SIGSEGV.
2. Error: Cortex received a fatal signal (SEGMENTATION VIOLATION). Error Object: ()
3. 999999: mpt_accept: error: accept failed: No such file or directory
4. The fl process could not be started.
There is also a .txt file appearing stating most notably:
5.Node 0 Fatal signal raised sig = Segmentation fault
Do you know hot to resolve this problem? I started fluent2020 via the VS2019 command prompt -> loaded the mesh data via file->read->mesh -> did the setup and tested it -> compilled my source files -> hooked 1.st udf -> set 2.nd as bounday condition -> error.
Thank you again Antoni for the motivation now i really want so see if my udfs work. I would appreciate your help verry much!!
P.S. as you mentioned the latent heat : it yould be possible to adjust the total energy reservoir after succesful implementation.
Greetings from Wuppertal :)
- Badges
- Science topic