What are typical (or promising) techniques and methods in CFD to achieve high-detailed turbulence and non-linear waves interaction? And with "active" geometry (when bodies interacts with fluid in both ways)?

Currently I've been recommended SPH, but I am very unsure, that SPH is smart choice for my task. Especially considering my attempts to achieve nice parallel behaviour.

Also, it would be interesting to see some promising and good papers about some recommended techniques.

Update on motivation: the reason to write our own code is because there is no high-order discretization with low dissipation and dispersion in most comercial codes. Also, software like ANSYS is known to be not very scalable.

  • $\begingroup$ Are you interested in DNS? RANS? LES? Compressible or incompressible? What range of Reynolds number? Are you planning to write your own code (and if so, why)? How scalable do you want it to be? $\endgroup$ Jun 20, 2012 at 21:21
  • $\begingroup$ I don't know yet which of DNS/RANS/LES is good for me. The flow will be compressible. Reynolds number is in very wide range: it can be even less than 2000, but it can easily reach 10000 and little bit more. We are writing our own code, because in most software there are problems with dissipation and dispersion at very high detalization. Scalability -- for 100-1000 CPUs. $\endgroup$
    – Mixo123
    Jun 21, 2012 at 3:56
  • 1
    $\begingroup$ I'm not sure I understand why you would not use an OpenSource library that is scalable, has all the turbulence models that you can think of, and a dynamic mesh engine that supports exactly what you are asking for. As for the two phase flow and higher order accuracy: what method are you aiming at? SPH is definitely not a higher order method. For two phase flow, you can hope for a 2nd order method in time and space at best, because of the explicit nature of the interface advection. $\endgroup$
    – tmaric
    Jun 21, 2012 at 7:52
  • $\begingroup$ What kind of wave are you talking about Is it sound, gravity, inertial, Alphen ... ? $\endgroup$
    – ucsky
    Sep 18, 2012 at 20:55

2 Answers 2


I'm working with SPH and what you describe will not be suitable for the following reasons

  • SPH is a low order method (1st at most)
  • 2 phase flow can be done with SPH but is generally very costly due to the restrictions coming from the different speed of sound
  • With 1000 cores you will not be doing a RE 10k DNS

other than that LES is probably the way forward.


This is not my area, but I would recommend never the less to take a look into interDyMFoam solver of the OpenFOAM library. If I understood you right, you need body dynamics + two phase flow.

A pseudo VoF two phase flow solver coupled with a sliding mesh interface and a 6DoF solver for the solid body motion == a box that floats on the water. Turbulence for two phase flow is still in development (the modelling part, and I mean in general: there are no rigorous turbulence models for two phase flows out there yet that I know of) but you can use the standard turbulence models with this solver without a problem.

Here is the movie, it is large so it will break when you play it in your web browser. The parent directory containing papers, PhD theses, and other results is here.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.