search

ElmerCSC / elmerfem

Official git repository of Elmer FEM software
http://www.elmerfem.org
Other
1.2k stars 319 forks source link

Post processing for stranded homogenization model loss in harmonic 3D models #422

Open ettaka opened 1 year ago

ettaka commented 1 year ago

Homogenization for harmonic 3D stranded coils exists, but the loss distribution processing is not yet implemented.

In principle, we could already replace the homogenized "sigma" in material parameters and get the correct "skin effect" loss. Thus, we would only need to implement the "proximity effect" loss based on the homogenized parameters.

However, the "homogenized" sigma is given in component. So we could also use that so that it is less confusing. In that case, we could implement a single homogenization loss field. Or then we could also separate the two in to proximity loss and skin loss components.

ettaka commented 1 year ago

What do you think @jvencels, should we separate the homogenization into skin loss and proximity loss fields or just a single homogenization loss field?

jvencels commented 1 year ago

Orthogonality between skin effect and proximity effects are assumed (without going into details, many articles mention that). With this, superposition can be applied and losses summed.

Separately estimated skin and proximity losses will give an engineer more insight to focus on one or another loss type.

ettaka commented 1 year ago

FYI @jvencels I implemented a first draft https://github.com/ElmerCSC/elmerfem/tree/feature/homogenization-post

It has an issue with showing the corrent fields. Shows only "Skin Loss" nodal field, other fields are empty. Also, that is really the "Proximity Loss" field. I need to check it.

The way it works, you say "Calculate Homogenization Loss = Logical True" in CalcFields section.

ettaka commented 1 year ago

@jvencels I got the "proximity loss" now working: image image image

I modified the existing homogenization test

However, see the commit message

What do you think? Should we just put the losses in "harmonic loss" and "joule heating" fields?

jvencels commented 1 year ago

@ettaka Regarding putting proximity losses into harmonic loss, I see no difference as it is just a naming.

Regarding Joule losses, in 3D, they are not considering the fill factor. In other words, as a joule loss, we get dot("current density re e","current density re e")/coilMaterialConductivity/2

For homogenized cases, we should use modified coilMaterialConductivity, likely Sigma 33?

ettaka commented 1 year ago

Ok I will put the proximity loss in harmonic loss and modify joule loss to take sigma 33 into account.