Hello everyone, I encountered issues while using MFIX for simulations on a Linux system. My geometric model, built in SolidWorks and imported as an STL file, consists of a cylindrical empty chamber with six uniformly distributed inlet pipes around it, and a pressure outlet below the chamber. Using a simple pure gas model, I noticed two problems:
1.A few face elements are missing in the mesh (Figure 1). I found that models generated within the software don’t have this issue, but imported STL files consistently exhibit similar problems.
2.Some of the six inlet pipes are not perpendicular to the computational domain boundary. To make the gas enter the central chamber along the pipe direction, I tried setting x and z component velocities at the mass inlets—for example, 10 in the z-direction and 10√3 in the x-direction to ensure a resultant velocity of 20 with the correct direction. However, this setup causes uneven gas velocity distribution in these pipes, which differs significantly from the velocity profiles in pipes perpendicular to the boundary. The figure below is a gas velocity cloud map, where the vertical pipe pointed by the yellow arrow is different from the non-vertical pipe pointed by the blue arrow. I’m unsure if this relates to mesh quality or how component velocities are set. Could you advise on how to ensure uniform gas inflow along the pipes?Thank you for your help!
I’ve uploaded the case file for a 0.2-second simulation run.
case.zip (4.2 MB)
Hello,
I allow the support experts answer and correct me if I am wrong, but regarding the first question, in my experience, making the .STL longer and then cropping it with the domain extents usually provides better element quality. You can always inspect the boundary mesh to check and adapt your mesh and geometry if it is doing it correctly.
Concerning the second issue, I would say that it is problably some problem with the cartesian grid and the boundary conditions at the walls. I am sorry, but I would do exactly the same as you are doing. You can probably check changing some mesh parameters to see if you get something more uniform.
Cheers
Are you worried about the graphics, or have your tried and get errors running simulations on this as well? In my experience the mesh graphics for funky designs like yours are not always the same as the real mesh.
Regarding 2): I think you might be limited by the outer “box” where boundaries are defined. Could one alternative be to bend your 6 small pipes so that the all enter through y-axis?
Also, I would add more resolution, you seem to have only 3 regular cells in those small pipes. In my experience that is too coarse.
Hello,
Thank you for your response. Are you suggesting that despite apparent missing surface patches in the visualization, the actual grid configuration may not contain errors? In practice, while encountering computational anomalies during simulations, I found no conclusive evidence linking these to surface patch discontinuities.
The persistent challenge lies in the second issue. Regarding your proposed approach of applying curving pipes to align all pipes vertically:The airflow will collide with the wall at the bends, causing airflow disorder – contradicting my objective of achieving uniform airflow ingress into the central cavity. Additionally, even with tenfold grid density refinement, the velocity profile discrepancies remained unresolved.
Without STL, I constructed a very simple geometric model in MFIX and compared two pipes - one perpendicular and the other non-perpendicular to the computational domain (as shown in the figure below) - which still exhibited significant discrepancies. I think this may relate to structured grid generation: the perpendicular pipe employs uniform square grid elements, while the non-perpendicular configuration results in irregular grid alignment (figure below). What methodologies can be employed to adjust grid generation parameters for preserving structured grid regularity?
Hello everyone,I tried adjusting the Mesher and the number of meshes, but still no effect. I would like to know what should be adjusted? Or is there a way to set the mesh at the boundaries not as Cut cells but staircase steps.
You shouldn’t be too focused on the mesh regularities here, as you perhaps have with CFD meshes from before. If you want to improve any of this, you should make the mesh finer, not focus only on the geometry edge quality. The cut-cell method deals with all the occurences of non-cubic cells.
About your inlets, I think you are limited to the sides of the overall “boundary box”, which has exactly 6 sides, and you might probably use 4 of them to have horizontal inlets. So this is an obvious limitation when you want 6 or more horizontal inlets. I don’t see any other easy way to accomplish what you want, other than introducing bends. But first and foremost, you should increase the numbers of cells inside you small pipes even further, so that the majority of cells are whole cubic cells, and not cut.
Thank you for your suggestion. However, I find that this problem not only affects the quality of the edge mesh, but also the calculation of complex models, such as turbulence models.
I attempted to conduct unidirectional flow simulation in a simple.STL geometric structure and added the turbulence model, but the following error occurred:
Error: Unexpected solver message:
Program received signal SIGFPE: Floating-point exception - erroneous arithmetic operation.
Backtrace for this error:
0 start_thread
at ./nptl/pthread_create.c:442
Floating point exception (core dumped)
Under exactly the same conditions, the same geometric model established directly in the mfix software did not have errors. I think there might be problems with the meshwork of the STL model in the software, which might be related to the resolution of the model.For complex geometric models, I can only import stl files, and such models report errors more frequently.Therefore, the main problem now might be how to adjust the model resolution and the parameters in mesher to improve the overall mesh quality of STL model.
Well, turbulence can be notoriously difficult to define on BC and initialize, this is not unique to MFiX. If you want to evaluate this properly you should start with a simple 2D model and get your BC’s correct there first. So do not jump straight to an STL setup, which is prone to a wider array of errors, and not just turbulence related - that would probably make your life easier.
