Negative Density
 

Hi, I have a problem when trying to run transient case for combustion process. I have finished modelling our IC engine by using es-ice and created the es-ice panel inside PROSTAR. And then after setting some parameters, I perform to run the transient anaylsis using pmovlink and it runs succesfully. But when I continue to perform "star" with 100 events at the time step number 86, there is a error message as follows:

FINISH TIME STEP NO. 86 CPU TIME IS 4457.14 ELAPSED TIME IS 4468.64

*** ERROR #076 *** NEGATIVE DENSITIES FOUND AT MORE THAN 100 CELLS. RUN STOPPED

Please let me know soon regarding this error message. Thanks for your attention. Look forward to getting your response.

wendy

Re: Negative Density
 

Re: Negative Density
 

Hello Wendy,

I am not familiar with es-ice or moving meshes but I have dealt with negative densities alot. For me negative densities usually occur because I have bad cells from trimming. You can get the cell number by doing a grep on the info file (if you are on Linux)

grep NEG casename_000*/*info*

You can also use the awk command to put it in an easy form "cset add clist $9", sorry I do not have this command memorized.

With steady state on can lower the relaxation factors to help get a converged solution but since you are transient, have you tried lowering the time step?

Good Luck, Kevin

Re: Negative Density
 

Dear Kevin,

Thanks for your sharing information to me. After running the solution using STAR with the transient case, the error message appeared, which stating that the negative density found in more than one hundred cells. I have already tracked down the number of cells, which have the negative density from the info that es-ice produced during running transient analysis.

Considering that I am working under SGI Origin, I am not so sure whether the command you suggested to me is quite similar or not.

About reducing the time step, I already done this by reducing the time step size. Unfortunately, the solution will be diverge or not converge so that the run stopped suddenly. I will try my best to know my problem case at this moment.

Anyway, could you please inform to me the email address from someone in CD Adapco or only Adapco who has the expert experience in utilising es-ice. es-ice is the abbreviation of Expert System for Internal Combustion Engine that produced for simulation the in-flow cylinder analysis quite user-friendly.

I really hope that you could share and give me the info for the one who responsible for the enquiries in es-ice. I would appreciate for your help.

regards,

wendy

Re: Negative Density
 

Dear Wendy,

I do not think I can give out an e-mail address on this forum, I have you tried going through the cd-adapco website for support, technical and sending an e-mail there.

You said that you can see the bad cells, Does that mean you can put them in your cset? What do they look like, (concave, warped)? Are they close to a boundary? For example the interface between moving and non moving, is it too coarse?

Kevin

Re: Negative Density
 

Try to change the turbulent viscosity definition.

When the turbulence viscosity is too low (both with respect to physics and for a too coarse grid and/or large time step), gradients (as temperature) are sharp. This causes oscillations which are carried over the density field as spurious waves. An UD or TVD scheme is not always sufficient to damp this waves since they may be solved as acoustics invariants.

Re: Negative Density
 

hi i also had a similar problem while working on an ic engine problem.this can be solved by reducing the time step,but it will take more convergencec time