Flow behind a 3D cube
 

Hi,

I'm repeating the simulations of flow behind a 3D cube from the paper:

Sedighi, K. & Farhadi, M. (2006), Three-Dimensional Study of Vortical Structure Around a Cubic Bluff Body in a Channel,
Facta universitatis - series: Mechanical Engineering, Vol. 4, Iss. 1, pp. 1-16
http://scindeks.ceon.rs/article.aspx...01001S&lang=en

This paper is a LES study so to start I'm running a steady RANS-SST simulation which I'll use for initializing an unsteady RANS-SST run. Then I'll move onto a SAS-SST run and maybe a LES run - if my hardware will do it.

I've made the geometry and multiblock mesh in ICEM to the same criteria as in the paper (it's gives all the geom and mesh details). The geom is shown in the figure although the paper does not give a valve of H I have used 1 m. The Reynolds number based on H is 40,000. Also the inlet velocity profile is the 1/7 power law. The paper also states all the domain boundary condition types used.

So I've made all this but get an error in CFX and it stops after 3 iterations - I've attached the out file (as a txt file).

My first thought is that the downstream boundary needs to be moved further away. However, if this is the case why would CFX not be robust enough to be able to repeat this simulation if it can be done by these other authors? Secondly, it says the maximum Mach number gets incorrectly high (even though inlet velocity = 0.6 m/s) but due to the CFX failure I cannot see in CFD-Post where in the domian this occured.

But wanting to keep the geom the same I have set the outlet face to an opening but I get a CFX falure as well (see the out file)

Any thoughts? Thanks

You have the physical time step set as a CEL expression. This is a bad idea. For new simulations you always need to adjust it bigger and smaller according to how the simulation is responding. And given that you are diverging rapidly you clearly need a much smaller time step. Once it is starting to converge then you can increase the time step (using edit run in progress is good for this so you don't have to stop and restart).

Thanks Glenn,

I've deleted the CEL for the physical timescale. Now I have tried the Auto Timescale > Conservative and Auto Timescale > Aggressive so that CFX can choose what it needs.

Also, I have changed the outlet to an entrainment opening when I ran CFX for 2 iterations (it crashes on the 3rd) all the flow on the outlet face was flowing into the fluid domain (as seen in CFD-Post). To make things a bit simpler still I have also set the top geom boundary from a wall to an entrainment opening.

But I still get CFX crashing on the 3rd iteration with those crazy high Mach numbers - which I have also seen in CFD-Post. Whilst in CFD-Post I plotted the velocity vectors at the inlet face just to check that they where as set in CFX-Pre and all was correct there. Yet, the rest of the flowfield is unphysical.

I don't generally use entrainment openings. Just the default is usually OK, and seems to be more stable for me.

You need to change it to physical time scale, but make it a lot smaller time step than it is currently using. Alternately give it a better initial condition from a coarse mesh or first order upwinding simulation.