Number of boundary layers for DBC (and mDBC)
Hi, I'm a bit confused with number of boundary particle layer for DBC (as well as for mDBC). In reference paper to DualSPHysics, DualSPHysics wiki and in English et al. 2021 is stated:
"The boundary particles of mDBC are arranged in the same way as the boundary particles in the original DBC, ..."
For exmaple even in test cases /mdbc/01_StillWedge and /mdbc/03_Sloshing (which are presented in the paper English et al. 2021) is used the same number of boundary particle layers (i.e. 3 resp. 4) for DBC and mDBC, which is the same configuration.
On the other hand, in DualSPHysics guide to mDBC states:
"However, creating the initial condition for mDBC is more complicated since: ... Several boundary particles layers are necessary (three layers or more). It depends on the SPH smoothing length (h)."
Moreover for testcases with DBCs (for example /main/01_DamBreak, /main/01_DamBreak2D, I guess it's same for almost all testcases withou mDBC), there is only one layer of boundary particles. At least when I check CaseName_Bound.vtk, the same is for imported .stl files. (*)
Can please someone explain to me, how it actually is with the number of boundary layers and clear, why is sometimes used only one layer and sometimes more? Because based on resources mentioned above, the situation is a bit confusing. Besides that, I somehow can't imagine DBC with single layer of boundary particles.
____
*) I figured out, that if I want to have multiple layer of boundary particles with imported stl, I need prepare thicker model, with the thick of face equals to multiple dp, for example with solidify modifier in blender. Another option is use multiple .stl files scaled up or down by dp for each boundary layer. But I'm still curious. Moreover, this approach (at least I think) is necessary when I want to use mDBC with imported stl. If I want to create mDBC with imported file as states in Alex's answer ( in https://forums.dual.sphysics.org/discussion/2159/mdbc-with-an-stl-geometry#latest ), besides the resized model for generation of normals I need to use thicker stl for generation of boundary particles or multiple models, each for one boundary layer, otherwise I obtain only one layer of boundary particles. At least that was something I was facing. If there is any other way how to generate multiple layer of boundary particles with imported .stl or what is the proper way to solve this, please let me know.
Comments
Hi
Regarding your *), correct, seems to be bestwise, other than using "drawpoint(s)" to draw your own distribution. See CaseTemplate.xml under examples/others.
In regards to your first part of the message, you are right that the particle distribution can be the same for mDBC and DBC AND BOTH need more than one layer. Less than 3 layers (in general) is not professional, when you want to avoid kernel truncation at boundary. One layer of boundary particles is just an example, not something you should follow.
Kind regards
Thank you for answer. Yes, I know that there is need for multiple layers to full the compact support but I'm curious where the layers actually are? Definitely not in
CaseName_Bound.vtkand it's the same for almost all examples, not only that I mentioned above.Because on the other hand, all the examples work suspiciously well...
Using one layer is not correct, no matter what they do in the examples. There is no "hidden layers"
Kind regards
Not sure if it's helpful but following this study
Fuchs, S. L., Meier, C., Wall, W. A., & Cyron, C. J. (2021). An SPH framework for fluid–solid and contact interaction problems including thermo-mechanical coupling and reversible phase transitions. Advanced Modeling and Simulation in Engineering Sciences, 8(1), 1-33.
I take number of layers to be floor(kernel radius/dp).
Regardless of the geometry, I explicitly create the additional layers.
I understand this. I'm just curious how the test cases work, because I can't find where the another layers are. And each of the testcases works fine.
All cases included in DualSPHysics_v5.0\examples\mdbc use several layers when mDBC is applied.
You have to create a number of layers higher than 2*h
Regards