Distorted mesh elements in FEKO

Sra

Hello,

 

Could anyone please explain step by step process of removing distorted elements in FEKO. It was explained in the FEKO user manual but it is not clear for me.

 

I have an antenna model and when I simulated, I used to get a warning saying results are not converged. I found the problem that there are lot of distorted mesh elements. So kindly request you to explain step by step process of solving this issue.

 

Regards,

Sravan

JIF

Hello Sravan321,

 

Your description is not quite clear enough for me to provide step by step instructions for removing distorted elements. Is your model CAD or an imported mesh?

Although distorted elements can cause problems during the solution, difficulties with convergence can be caused by many factors. In order for us to help you, please provide as much info as possible (providing a model would be the easiest). We need to know what solution method you are using (FEM, MoM, MLFMM or a combination of these) and if it is a CAD model or a mesh model. Did FEKO complain about distorted elements or are you guessing that distorted elements are the cause of the problems?

Sra

Hello JIF,

 

It is my CAD model, not an imported mesh. Actually FEKO did not complain about distorted elements, but I get a warning: the results are not converged. I am guessing distorted elements is the problem for convergence. I am using MOM . Even though my model is small enough, FEKO is executing (simulating) the model for long hours. So I was guessing distorted elements is the problem

 

I am using fine mesh which is required. As I change the mesh to standard or coarse the results are not reliable. I couldn't attach the model here due to some confidential issues. Am sorry for that 

 Could you please mention the reason for convergence problem and how can i solve the distorted elements problem.

 

Thanks

 

 

 

JIF

The MoM results in a dense matrix and thus uses a direct solver and not an iterative solver (at least by default). Thus, the convergence issues that you are talking about is probably the adaptive frequency sampling used for continuous frequency simulations that has not converged (resulting in the warning). This can sometimes happen when the sampling algorithm tries to resolve small discontinuities due to different approximations (such as changing the number of integration terms). This can also be caused by trying to resolve a null (in the far field as an example) since the value is actually zero, but due to numerical precision it fluctuates around zero causing problems for the AFS algorithm. The best I can suggest without seeing the model would be to change to discrete frequency samples instead of continuous frequency sampling. If I had access to the model, I could try to make other suggestions.

 

I hope that helps. It will definitely get rid of the warning and should make the simulation time more predictable.

Altair Forum User

Hi Sra,

 

Your original question was about how you would go about removing distorted elements.  The comments afterwards indicated that you have the CAD models and have identified where the distorted elements are.  To help with these, you can try:

• Local mesh sizes on the parts that result in distorted elements.  This will result in more triangles (which comes with longer simulation times and more memory) which might not be supported by student versions of FEKO.
• Curvilinear triangles might also help here.  If you activate 'higher order basis functions' on the solver settings dialogue, then remeshing will result in curved triangles.  These might still be 'distorted' in a sense due to the underlying technique, you should get more accurate results.

That said - Jif's suggestion of setting your model to a list of discrete frequencies might help lower the runtimes.  You can also reduce the simulation range.  For instance, if you simulate from 1kHz to 1GHz, try reducing it to a smaller subset (e.g. 300-500MHz).

 

Regards,

Andries

Sra

Hello JIF and Andries,

 

Thanks a lot for your valuable suggestions. It really helped me.

 

Thanks,

Sra