[bladomas]

I will be modelling tube flow with the use of SPH (smoothed-particle hydrodynamics). My question was, which open source SPH code would be the best to use? Or which ones have clear advantages (some disadvantages as well) over others? I have found a website which lists several of them but I'm not sure if I can post a link here (for now).

[beer]

The most known and used one if probably DualSPHysics from

http://dual.sphysics.org/


I personally think that the best one right one is probably SPHlisHSPlasH
https://github.com/InteractiveComput...s/SPlisHSPlasH
They implemented almost every model I know for SPH.

[bladomas]

Thanks for your reply, I will check out SPHlisHSPlasH. After looking into some free codes, I found GPUSPH to be decent. How comparable is it to SPHlisHSPlasH?

[beer]

I think I tried the GPU-Version of GPUSPH at some point. My concern was that it had only constant material properties for water and WCSPH as a solver.

I am not 100% sure though.

[bladomas]

My end goal is model blood flow in arteries with moving walls. I will be starting out quite simple, normal tube flow with water. So I'm not sure whether having only constant material properties is an issue for me. I think I'm more interested in the ease of use and how easy it is to learn using the code.

[specializedSPH]

Originally Posted by bladomas

I will be modelling tube flow with the use of SPH (smoothed-particle hydrodynamics). My question was, which open source SPH code would be the best to use? Or which ones have clear advantages (some disadvantages as well) over others? I have found a website which lists several of them but I'm not sure if I can post a link here (for now).

I can send you my developed ISPH code that I wrote about 8 papers with it:
1. Incompressible SPH simulation of landslide impulse-generated water waves. Natural Hazards, 82(3), 1779-1802.
2. Modified variable mass incompressible SPH method for simulating internal fluid flows. Journal of the Brazilian Society of Mechanical Sciences and Engineering, 38(7), 2009-2027.
3. Eulerian ISPH Method for Simulating Internal Flows. Journal of Applied Fluid Mechanics, 9(3).
4. On the numerical simulation of the nonbreaking solitary waves run up on sloping beaches. Computers & Mathematics with Applications, 70(9), 2270-2281.
5. Comparative study on the accuracy of solitary wave generations in an ISPH-based numerical wave flume. Applied Ocean Research, 54, 115-136.
6. Simulating Solitary Wave Generation Using Incompressible SPH. Ann Limnol Oceanogr 1(1): 013-021.
7. On the Generation, Propagation and Runup of Solitary Water Waves on Steep Beaches, in 6th International Offshore Industries Conference (Tehran, Sharif University of Technology: 2015).
8. ISPH Numerical Simulation of Tsunami Generation by Submarine Landslides, Arabian Journal of Geosciences

[bladomas]

I would love to have a look at your code. I'm currently busy learning C++ and several different programs for compiling codes. Soon I will have a good look at the codes and compare them.

[specializedSPH]

Originally Posted by bladomas

I would love to have a look at your code. I'm currently busy learning C++ and several different programs for compiling codes. Soon I will have a good look at the codes and compare them.

mail to futureidea@chmail.ir to send code.

[pbrady2013]

Originally Posted by beer

I think I tried the GPU-Version of GPUSPH at some point. My concern was that it had only constant material properties for water and WCSPH as a solver.

I am not 100% sure though.

Hi @beer,


From my recollection you were right from a while ago.


I use GPUSPH now and am getting good results. There are different kernels now and flexible equations of state.


I'm focussed on free surface problems and have not tried enclosed flows with moving walls as the OP has later added. Would be worth a look.



Cheers,
-pete

[meandcfd]

This is not free code. specializedSPH is selling it. His/her comment via email when asked about it being free: "And you figured it would be free of charge? Extraordinary Impression!!!!!!!!!!"

Originally Posted by specializedSPH

I can send you my developed ISPH code that I wrote about 8 papers with it:
1. Incompressible SPH simulation of landslide impulse-generated water waves. Natural Hazards, 82(3), 1779-1802.
2. Modified variable mass incompressible SPH method for simulating internal fluid flows. Journal of the Brazilian Society of Mechanical Sciences and Engineering, 38(7), 2009-2027.
3. Eulerian ISPH Method for Simulating Internal Flows. Journal of Applied Fluid Mechanics, 9(3).
4. On the numerical simulation of the nonbreaking solitary waves run up on sloping beaches. Computers & Mathematics with Applications, 70(9), 2270-2281.
5. Comparative study on the accuracy of solitary wave generations in an ISPH-based numerical wave flume. Applied Ocean Research, 54, 115-136.
6. Simulating Solitary Wave Generation Using Incompressible SPH. Ann Limnol Oceanogr 1(1): 013-021.
7. On the Generation, Propagation and Runup of Solitary Water Waves on Steep Beaches, in 6th International Offshore Industries Conference (Tehran, Sharif University of Technology: 2015).
8. ISPH Numerical Simulation of Tsunami Generation by Submarine Landslides, Arabian Journal of Geosciences

[beer]

Originally Posted by pbrady2013

Hi @beer,


From my recollection you were right from a while ago.


I use GPUSPH now and am getting good results. There are different kernels now and flexible equations of state.


I'm focussed on free surface problems and have not tried enclosed flows with moving walls as the OP has later added. Would be worth a look.



Cheers,
-pete

Hi Pete


So they added a few features, nice. Especiall the moving and floating objects seem to perform well.

However they still only have WCSPH implemented. You should have a look at PCISPH and IISPH. Those are iterative methods for incompressible SPH but do not need the solution of a PPE. The possible speed up is crazy and the incompressbility is acutally enforced. Those are implemented in SPHlisHSPlasH.

I haven't had a look into these codes lately because I decided to write my own OpenCL code (don't do that...) and now Cuda code (much better).



Regards

[meandcfd]

If the problem is dealing with incompressible then look at MPS (moving particle semi-implicit) which has been around for years and addresses incompressible.

[pbrady2013]

Originally Posted by beer

Hi Pete


So they added a few features, nice. Especiall the moving and floating objects seem to perform well.

However they still only have WCSPH implemented. You should have a look at PCISPH and IISPH. Those are iterative methods for incompressible SPH but do not need the solution of a PPE. The possible speed up is crazy and the incompressbility is acutally enforced. Those are implemented in SPHlisHSPlasH.

I haven't had a look into these codes lately because I decided to write my own OpenCL code (don't do that...) and now Cuda code (much better).



Regards

Hey @beer,


Thanks for the clarification and update. I should have a look at your code. Its that the following link on github?


https://github.com/InteractiveComput...s/SPlisHSPlasH



GPUSPH released an updated solver yesterday but still WCSPH.


Cheers,
-pete

[beer]

Oh, it is not my code. I have a differnt one with different features. I just think it is really cool.

[hancock]

Originally Posted by bladomas

I will be modelling tube flow with the use of SPH (smoothed-particle hydrodynamics). My question was, which open source SPH code would be the best to use? Or which ones have clear advantages (some disadvantages as well) over others? I have found a website which lists several of them but I'm not sure if I can post a link here (for now).

For mesh free methods PySPH is an opensource tool which provides a wide range of schemes for simulating the various type of physics. and the good part about it is that a python code is used to change or create a new scheme or simulation. The link for the software is

https://github.com/pypr/pysph

[AlexKaz]

Free DualSPHysics, I think. It works on cuda cards and CPUs. In my tests Ryzen 1800x (~160-180 GFlops in double) was about 3 times slower than NVIDIA Tesla C2050 (~500 GFlops in double).

https://dual.sphysics.org/downloads/