A floating point exception - SEM Model
 

Dear experts,

I am doing a LES of fully developed turbulent plane channel flow. I employed the inherent Synthetic Eddy Model to specify the realistic inflow boundary conditions. However, when I start to run the simulation, it shows a floating point exception as shown in the attached file below. Actually, I have read a support article "What do I do about a floating point exception run time error" and carried out a RANS K-Epsilon with the same setup as that in LES before, which gives a good result. Moreover, I have used "remove invalid cells" features to check the mesh quality which indicates no invalid cells. I think the problem arises from the SEM model as described from the STAR--CCM+ warning shown in the attached pic. I was wondering if you could give me any advice on removing the floating point exception problem. And if you need more information about my simulation, please tell me.

Best regards,

Yansheng

Dear experts,

To continue from the thread, the aim of my work is to carry out a LES simulation which solves the problem of Floating Point Exception (FPE) caused by SEM setup. Actually, I have analysed the problem by considering several fundamental questions like “is SEM necessary for LES?”, “Is precursor RANS result possible to replace the role of SEM to give desired results?” Also, I have carried out relevant simulations to verify my assumptions, from which I find that LES without SEM cannot give an accurate result, and precursor RANS input cannot replace the role of SEM. Thus it seems that we cannot ignore the effect of SEM when doing a LES turbulence modelling.
As a result, I try to go through every single setup inside a LES simulation and finally found the reason why I cannot even initiate the simulation under the use of SEM. Since there is a specific requirement – “Maximum number of eddies”, with a default number of 1000, which is far less than my cell number which is over one hundred thousand. So I modify this value to 150000, which is big enough to include all the cells. And then the LES simulation could be initiated and run. However, the result I get is around two times as big as the desired one. Based on the post-processing, I found out the velocity field of the channel flow in the LES is similar to that of RANS without periodic boundary conditions, and also the value of the skin friction is pretty close. This result is really positive because it means that LES give a good simulation of the turbulent channel flow. However, the problem appears when I try to implement the periodic boundary conditions on the LES simulation. Since I found the Floating Point Exception appears again when I select the periodic boundary conditions in the streamwise and spanwise directions. So I was wondering is there any confliction here between periodic boundary conditions and SEM implementattion.