Learning Forum

[576650550]

Regarding the coupling process between Icepak and HFSS, HFSS outputs surface and volume loss density. According to the finite volume method, is the surface loss density transferred to the centroid of the Icepak volume element? The HFSS interpolates the loss density at the centroid through mesh nodes. Therefore, it is necessary to ensure that the mesh is similar to avoid interpolation errors or other inaccuracies caused by excessive differences in the mesh, right?

[Rabindra Paul]

Hi, Thanks for using Ansys learning forum. You are right about the interpolation inaccuracies for excessive differences in the mesh. However, if you have required mesh resolution in HFSS to represent the correct power losses then Icepak should get the correct power losses as an input. It is not required to use the same mesh as long as power losses are correctly represented. You can  display the EM losses in Icepak before solving to confirm the correct power distribution. In Icepak, you need to make sure that the mesh resolution is good to capture the power losses. As for example, you need more mesh for the area of high power gradient compared to the area where power is uniform.

Hope this answers your question.

 

[576650550]

Yes, I actually found that too thick or too dense icepak mesh would reduce the mapping accuracy. I think it is necessary to adjust the HFSS or icepak mesh to make them similar (scale factor close to 1) when the mesh quality meets the requirements, rather than the exact same mesh. I have tried to further improve the mapping accuracy factor≈1 of key parts by adjusting HFSS mesh (total loss is basically unchanged) on the premise of keeping icepak mesh unchanged and with good quality. Therefore, I want to verify whether my idea of adjusting the mesh in this way is correct. Instead of encrypting one side of the grid, the two are adjusted to the appropriate value or close to each other, provided that the grid quality is at least good enough. Forgive me for sending it a few times because of the network delay.

[576650550]

Yes, I actually found that too thick or too dense icepak mesh would reduce the mapping accuracy. I think it is necessary to adjust the HFSS or icepak mesh to make them similar (scale factor close to 1) when the mesh quality meets the requirements, rather than the exact same mesh. I have tried to further improve the mapping accuracy factor≈1 of key parts by adjusting HFSS mesh (total loss is basically unchanged) on the premise of keeping icepak mesh unchanged and with good quality. Therefore, I want to verify whether my idea of adjusting the mesh in this way is correct. Instead of encrypting one side of the grid, the two are adjusted to the appropriate value or close to each other, provided that the grid quality is at least good enough

Yes, I actually found that too thick or too dense icepak mesh would reduce the mapping accuracy. I think it is necessary to adjust the HFSS or icepak mesh to make them similar (scale factor close to 1) when the mesh quality meets the requirements, rather than the exact same mesh. I have tried to further improve the mapping accuracy factor≈1 of key parts by adjusting HFSS mesh (total loss is basically unchanged) on the premise of keeping icepak mesh unchanged and with good quality. Therefore, I want to verify whether my idea of adjusting the mesh in this way is correct. Instead of encrypting one side of the grid, the two are adjusted to the appropriate value or close to each other, provided that the grid quality is at least good enough.

Yes, I actually found that too thick or too dense icepak mesh would reduce the mapping accuracy. I think it is necessary to adjust the HFSS or icepak mesh to make them similar (scale factor close to 1) when the mesh quality meets the requirements, rather than the exact same mesh. On the premise that icepak mesh is unchanged and the quality is good, adjusting HFSS mesh (the total loss is basically unchanged) can further improve the mapping accuracy factor≈1 of key parts, so I want to verify whether my idea of adjusting the mesh is correct. Instead of encrypting one side of the grid, the two are adjusted to the appropriate value or close to it, provided that the grid quality is at least good enough.

 

 

[Rabindra Paul]

Thanks for the detail explanations. I agree with your idea as that would make sure both output from HFSS and input in Icepak are modeled with good resolution and hence will have better accuracy. 

[576650550]

What is the scaling Factor a ratio of? I originally thought it was a loss ratio, but sometimes the temperature of a model with a scaling Factor greater than 1 is not higher than that of a more accurate model.

[Rabindra Paul]

Hi, Thanks for the comment. The scaling factor is 1 by default, which means the power values ( EM loss) from HFSS will be used as is for the Icepak simulation. However, to model worse case scenerio, users may use a factor more than 1. More power ( or scaling factor >1) should provide higher temperature than default 1, if everything else in the simulations remained the same. Please feel free to contact support through support portal if there are any concerns with the results.

Hope this answers the question. Please let us know any comments.

 

[576650550]

Yes, I found it difficult to keep the scaling factor higher or lower for each object in both unidirectional and bidirectional coupling, but looking at the loss distribution cloud map, it seems that the temperature of the better mapping is sometimes higher, probably because I adjusted the HFSS and icepak to improve the mapping accuracy while also changing the total loss value. In addition, I used the field calculator to integrate the loss power of icepak objects, and the results of some objects could not exactly correspond to the scaling factor. But in theory it's the process you're talking about.

[Rabindra Paul]

You are right, thank you. 

[576650550]

So scaling factor is not the exact ratio of losses, is it?

[Rabindra Paul]

Hi - As I mentioned before, the power losses from HFSS is the input bundary conditions for Icepak. So input power loss for Icepak is calculated by power loss from HFSS and multiplying with the factor. So it should be the exact ratio. However, Icepak results ( temperature) will depend on many other factors. Please feel free to create support cases if needed.

 

[576650550]

Yes, I think so, but some people have raised questions about this, and the help is not clearly defined, that is most likely the field calculator calculation is not accurate, the actual icepak loss should be calculated according to the scaling factor

[576650550]

Because the loss density is integrated with the field calculator in icepak, the loss pairs will not match. I did not know which should be the criterion for the results calculated by scaling factor. Now I know, thank you very much