stress tensor in SPS turbulent model
Dear DSPH developers,
The SPS stress tensor in SPS turbulent model that is defined in Sphysics Manual (page 12 , 3rd line) https://wiki.manchester.ac.uk/sphysics/images/SPHysics_v2.2.000_GUIDE.pdf
is different from what it is defined Eq. (57) in this Review paper by same authors http://www.iahr.org.cn/iahr/rootfiles/2010/06/03/1273896626666468-1273896626669911.pdf
and this
http://www.sciencedirect.com/science/article/pii/S0378383905001304
Which one is correct? DualSphysics code follows the former one. In addition, what's the formula for SPS turbulence kinetic energy k?
My second question is that it looks like the SPS stress tensor Tau used at time t in momentum equation is actually computed by using strain S (or Csph in the code) saved at t-dt, which means there is one time step delay. I understand the reason that it is designed this way is to avoid nested neighbor-searching sum (strain contribution to stress) within a neighbor-searching sum (stress contribution to dvdt) which cost time for twice neighbor searching. But I just want to make sure I understand the code correctly. And is it safe to always delay stress tensor by one time step although each time step is small?
Thanks very much,
Kai
The SPS stress tensor in SPS turbulent model that is defined in Sphysics Manual (page 12 , 3rd line) https://wiki.manchester.ac.uk/sphysics/images/SPHysics_v2.2.000_GUIDE.pdf
is different from what it is defined Eq. (57) in this Review paper by same authors http://www.iahr.org.cn/iahr/rootfiles/2010/06/03/1273896626666468-1273896626669911.pdf
and this
http://www.sciencedirect.com/science/article/pii/S0378383905001304
Which one is correct? DualSphysics code follows the former one. In addition, what's the formula for SPS turbulence kinetic energy k?
My second question is that it looks like the SPS stress tensor Tau used at time t in momentum equation is actually computed by using strain S (or Csph in the code) saved at t-dt, which means there is one time step delay. I understand the reason that it is designed this way is to avoid nested neighbor-searching sum (strain contribution to stress) within a neighbor-searching sum (stress contribution to dvdt) which cost time for twice neighbor searching. But I just want to make sure I understand the code correctly. And is it safe to always delay stress tensor by one time step although each time step is small?
Thanks very much,
Kai
Comments
the correct one is the one that DualSPHysics follows,
however if you still think that we have a bug there, please report that to us
about the use of strain S (Csph in the code) there is no problem of using the values computed at t-dt. the error is minimal and we can avoid and extra loop of interaction which is an important saving in time. so you are right in your conclusions.
we checked in the past and there is no problem with the delay in the stress tensor only by one small time step
best regards
Alex
I have problems in putting stress tensor code in my case ( analyzing particles in a pump for gaining stress tensor in each particle) ...
Please tell me how should I do it step by step? ( I didn't catch it up in manuals and papers...)
and Is it possible to have a graph of tensors in particles at the end? How?
Thank you so much
I'll appreciate it if you answer me quickly!
And using the new code ToVTK you can add tensor as variable to be plotted along time.
Regards
Alex
Please be informed I am also struggling with the same problem.If it's possible for DualSPHysics team please give us more information about how to correctly assign CSPH and TAU variables and actually how can we putting stress tensor and pressure in my case(pump case) ?Could you also give me a copy of the example xml file?
Modifying CSPH and TAU or creating new variables in the code is a question of your skills in coding with C++ or CUDA. We can not help with that.
Regards
Alex