CoilSolver for tall coils

[vencels]

Hi,

Let's continue this thread. To my understanding CoilSolver uses CoilNormal, CoilTangent and CoilTangent2 as orthonormal basis.

There is one interesting coil type - slant (inclined) coil.

Slant (inclined) short coil.

short.png (81.99 KiB) Viewed 946 times

The feature of it is non-orthonormal basis. To my understanding the most challenging part here is coil length because "winding normal" points through the coil's wall. Any ideas how to treat such coils?

Coil parameters (V1)

Coil Normal(3) = Real 0. 0. 1.
Coil Tangent(3) = Real 1. 0. 0.
Coil Geometry Tall = Logical True

Slant (inclined) tall coil. Normal (0 0 1)

normal.png (111.55 KiB) Viewed 946 times

Coil parameters (V2)

Coil Normal(3) = Real 0. -1. 1.
Coil Tangent(3) = Real 1. 0. 0.
Coil Geometry Tall = Logical True

Slant (inclined) tall coil. Normal (0 -1 1)

inclinedNorm.png (106.82 KiB) Viewed 946 times

[raback]

Hi Juris,

The "coil normal" and "coil tangent" are only used to create two opposite cuts in a closed mesh. These cuts are then used for setting BCs for a Poisson equation. As the cut is not smooth it causes local disturbances. Hence we use two different solutions of Poisson equation and combine them to obtain a smooth gradient everywhere.

The gradient field may then be normalized which is fitting for "stranded" coils. If not normalized it is like a "masive" coil. These do not represent all coil forms. Additionally there is a routine by Eelis which is aimed for "foil winding" coils where 2 depth fields are solved and then then a cross product is used to model the direction of the winding,

Maybe you could give a schematic view of the slant coil. I see here fishy looking fields but how should the desired one look like?

-Peter

[vencels]

Hi Peter,

It is a stranded coil that has inclined turns. It is wrapped as a regular coil around an oval core, then the oval core is substituted by a round core and winding gets this inclined shape.

Screenshot from 2022-02-21 13-22-47.png (10.75 KiB) Viewed 940 times

[raback]

Hi Juris,

I can safely say that none of the methods implemented so far can deal with this. The problem is that if we use Poisson equation to define directions it favors the shortest routes. Here the wires do not take that so we should invent something better. I guess currently the only option would be analytical functions. Do you have any suggestion how to span a suitable vector field?

-Peter

[vencels]

Hi Peter,

Analytic expression exists for current and it would look like this

Current Density im 1 = Variable coordinate
Real MATC "sin(atan2(tx(1),tx(0)))"

Current Density im 2 = Variable coordinate
Real MATC "-cos(atan2(tx(1),tx(0)))/sqrt(2)"

Current Density im 3 = Variable coordinate
Real MATC "-cos(atan2(tx(1),tx(0)))/sqrt(2)"

This could be sufficient for coil excitation, but..
The actual problem is to model it as a secondary coil like a sensor where external B-field induces currents.

It seems that

Electric Conductivity 1 = Variable coordinate
Real MATC "sin(atan2(tx(1),tx(0)))"

Electric Conductivity 2 = Variable coordinate
Real MATC "-cos(atan2(tx(1),tx(0)))/sqrt(2)"

Electric Conductivity 3 = Variable coordinate
Real MATC "-cos(atan2(tx(1),tx(0)))/sqrt(2)"

is not working straight away.
Is there a way to set Electric conductivity similarly like we set Current density in the body force?

[raback]

Hi Juris,

"Electric Conductivity" may be a tensor of size (3,3) or if diagonal of size (3). You could look examples on unisotropic material parameters in test cases "HeatUniso*". Probably the one with UDF would work best for you here.

-Peter