Is DualSPHysics suitable for pipeflow?

Asalih3d

Hello!

This might be a simple question to answer for some, but I would like an explanation if possible. I was thinking about simulating some kind of pipe like this:



But I've been reading a little more about DualSPHysics and as far as my limited understanding, I believe that DualSPHysics wouldn't be suited for this currently, since it a Free Surface Flow approach?

So the fluid (water) wouldn't be able to interact with the solid boundary properly and I wouldn't get any useful results like friction between solid/fluid ect?

My question would probably be better suited for the version in October 2018 with inlet/outlet (hopefully), but there is no reason not to ask now. I hope you can help me understand this topic a little more.

Kind regards, Ahmed

NWRichmond

Hi Ahmed,

I'm not an authority on the issue, but you could absolutely fill the pipe with fluid particles using fillbox (see CasePump for example) and get some meaningful information about the interaction of the fluid and boundary particles, e.g., forces and pressure. You can use the periodic boundary condition to reintroduce particles leaving the pipe's outlet back to its inlet. With the coordinate system you show above, it looks like you would want to use a periodic increase in both Y and X.

I am also eagerly awaiting inlet / outlet conditions (essential for rivers), but the periodic boundary condition is a very useful stop-gap measure.

I lay out a basic procedure for working with fillbox here: http://dual.sphysics.org/vanilla/discussion/comment/1681/#Comment_1681

...but this also seems like a problem that is well-suited to DesignSPHysics, which has the added benefit of visualizing the fill point fillbox.

Good luck,
Nick

Alex

Pipeflow is not the best example to be simulated with DualSPHysics.. for several reasons:

1) The first one is that pipeflow is a problem with fixed boundaries where no complex geometries are moving fast and where there is no need to detect the free surface ... so that the problem should be solved using mesh-based methods (finite element, finite volume) and not with a mesh-less particle method... These can of problems have been succesfully solved with mesh-based methods in the last year... So that SPH is not going to add value or improvement... You can try the open code OpenFOAM for example

2) Even mesh-based methods are more suitable to deal with these problems, DualSPHysics can be used to simulate the case of pipeflow (as NWRichmond explains), however our code is not designed for internal flows, but for free-surface problems. In fact our equation of state (Batchelor equation) is not the best option to reproduce the pressure gradient in the flow inside a pipe.... So several changes should be done in the current version of DualSPHysics to simulate this problem with accuracy.

3) By the end of the year we will release a code version with inlet/outlet, however most of our problems are focused on free-surface problems, wave propagation, etc. We will include some example of internal flows (no free-surface problem), but once again let me highlight that our code is more focused on coastal engineering applications

Regards

Asalih3d

Thank you very much both of you! I will answer both in my comment here.

@NWRichmond , your approach seems very logical to me, so that is probably what I would end up using, thanks. I've actually seen your fillbox comment before, it really helped me out earlier, this is one of the reasons I really like these forums. I am also using DesignSPHysics I love it, wish it would work directly in Blender though.

@Alex , thanks for your in depth explanation. I have some follow up questions, which I hope you'd like to answer

1. So basically the case you are making is that SPH is much more suited to more complex problems with moving parts and that the advantages of SPH doesn't "shine" as much when the obstacles are stationary?

2. Ah okay that is a very valid point - when you say "not best option" it means that it gives a result which is 10% off or does it mean 50%, or maybe that isn't possible to say? So if I wanted to simulate a pipe flow I should "just" change the equation of state and then I'd be able to get realistic results or would I have to change the whole simulation setup?

3. Yes, I am looking very forward to these boundary conditions. And I understand your point with primary focus on coastal applications ect.

I might be a little "stubborn" but I really enjoy the SPH approach to problems, since you guys have been able to get it working by GPU and also the idea of using particles to approximate means that a lot of trouble (meshgeneration ect.) gets bypassed with ease.

Thanks for your time.

Regards, Ahmed

Sergey

Numerical modelling of supercritical flow in circular conduit bends using SPH method
Journal of Hydrodynamics
2017,29(2):344-352
DOI: 10.1016/S1001-6058(16)60744-8

1. Numerical model
The DualSPHysics code based on the standard SPH method is used in this study. This version of the SPHysics code is developed for GPU computations. A more detailed explanation of this code can be found in Refs.[5,8,9].

Asalih3d

Thanks @Sergey ! Will look for more of those pdfs.

Sergey

but Alex is right:
Pipeflow is not the best example to be simulated with DualSPHysics