[Aden]

...spanning over partition


Dear everbody,
thanks for all the support.

I am doing an unsteady DPM-Analysis of emboli inside an aterial tree model.
For comparison of inert and massless particles I would like to do two calculations and release particles at the same injection locations.

Therefore I created a first injection with the following properties:
- Injection type: Surface, Number of streams: 10, inject using face normal, Randomize starting points
- Particle type: Inert, uniform diameter (2e-04 m) Particle in parcel: 1.

After initializing I was able to list the injection-properties in the console (see below). I copy-pasted them into a newly created ".inj" file, which I want to use for the massless particles (see attachment).

Particle streams from thrombus-injecton: STREAM ...........................INITIAL VALUES...........................
NO TYP (X) (Y) (Z) (U) (V) (W) (T) (DIAM) (MFLOW)
0 IN 1.67e-03 -2.77e-03 -7.24e-04 1.69e-05 -5.31e-02 2.03e-01 3.00e+02 2.00e-04 4.52e-06
1 IN 2.42e-03 4.57e-04 1.19e-04 1.69e-05 -5.31e-02 2.03e-01 3.00e+02 2.00e-04 4.52e-06
2 IN -1.90e-03 1.07e-03 2.81e-04 1.69e-05 -5.31e-02 2.03e-01 3.00e+02 2.00e-04 4.52e-06
3 IN -4.55e-04 1.50e-03 3.92e-04 1.69e-05 -5.31e-02 2.03e-01 3.00e+02 2.00e-04 4.52e-06
4 IN -1.41e-03 -3.67e-03 -9.59e-04 1.69e-05 -5.31e-02 2.03e-01 3.00e+02 2.00e-04 4.52e-06
5 IN -7.64e-04 1.66e-03 4.33e-04 1.69e-05 -5.31e-02 2.03e-01 3.00e+02 2.00e-04 4.52e-06
6 IN -3.05e-04 2.91e-03 7.59e-04 1.69e-05 -5.31e-02 2.03e-01 3.00e+02 2.00e-04 4.52e-06
7 IN -3.78e-05 -1.99e-03 -5.21e-04 1.69e-05 -5.31e-02 2.03e-01 3.00e+02 2.00e-04 4.52e-06
8 IN 8.97e-04 2.17e-03 5.68e-04 1.69e-05 -5.31e-02 2.03e-01 3.00e+02 2.00e-04 4.52e-06
9 IN -1.64e-04 2.34e-03 6.12e-04 1.69e-05 -5.31e-02 2.03e-01 3.00e+02 2.00e-04 4.52e-06

Then I created a massless particle (ml_injection) and read the .inj-file. When i am now trying to list the second particle in the console I recieve this message:

Injection ml_injection: 6 particles lost outside the domain while spanning over partitions.

Particle streams from ml_injection:
STREAM ...........................INITIAL VALUES...........................
NO TYP (X) (Y) (Z) (U) (V) (W) (T) (DIAM) (MFLOW)
2 ML -1.90e-03 1.07e-03 2.81e-04
3 ML -4.55e-04 1.50e-03 3.92e-04
8 ML 8.97e-04 2.17e-03 5.68e-04
9 ML -1.64e-04 2.34e-03 6.12e-04

I checked and all 10 injection points are on the inlet surface.
I tried using diffrent numbers of CPU-Cores.
I tried to create and load files with 5, 10 and 21 streams, on two different geometries.
The problem always remains, just diffrent number of streams are lost.

Intrestingly, although i could not make out a pattern of which particle streams are lost, for one file the same streams are lost when loaded into a new case (.cas).


for me, this does not make sense, since fluent itself created these points on the same geometry.


I appreciate any thoughts on this issue.
Cheers, Aden


PS: I am using Ansys Fluent 2022 R1
Different question, will the randomized surface injection have the same injectionpoints at every timestep?

[gwqsz]

Dear Aden,
Have you solved this problem yet? I'm having this problem now too. I need your help very much.

[Aden]

Hi gwqsz,
unfortunatly no, I didn't. I tried to work around the issue by:


1. create twice as many injections in Fluent as necessary.
2. after initialisation print the particles locations to the console and copy-pasted these locations into a text editor.
3. delete the redundant streams and saved it as a .inj-file and

4. in the Fluent-case, delete the original injection and load the streams needed from the files.



slightly pathetic and a lot of extra work I know, but it did the job.


Still I'd have use for a solution to this problem.

cheers

[gwqsz]

Dear Aden，
Thank you for getting back to me. I'm sorry to hear that you didn't find a direct solution to the issue, but I'm glad to hear that you were able to work around it by creating extra injections in Fluent and manually copying the particles locations.
I understand that it may have been a lot of extra work, but it's great that you found a way to get the job done.
I wanted to share with you that I have found an alternative solution to the problem. I changed the "parallel" option from "hybrid" to "shared memory" in the "Discrete Phase Model" section of Fluent. However, this method significantly slowed down the computation.
Best regards，Weiqi Guo.