Learning Forum

[Marcin Fisior]

General question: What would be the proper simulation steps to simulate thermal depended core/winding losses of an inductor?

Description:

You have an inductor (let say coupled - flyback). It works in DCM mode at specific Duty Cycle. From your inductor core/winding geometric design you need to get inductance matrix, core/winding losses and steady state temperature.

1. Since core losses will be different (i.e much higher at low duty cycle) with triangual current vs sine excitation we can not use EddyCurrent Solver to calculate core losses (or we can?) and Transient solver seems to be an option. Therefore

example can't be directly applied.

2. Switching to transient simulation would require different core loss model and thermal modifier. For Relative permeability looks we can use B-H loop with some thermal modifier equation based on datasheet values. However for core loss model with B-P curve the thermal modifier can't be applied so Power Ferrite model should be selected ? If yes, you need to calculate Cm,X,Y and thermal modifiers for that, however manufacturers give thermal modifier equation for whole Steinmetz equation, not for individual Cm,X,Y. Is there better option for that?

3. To make a proper transient simulation you need excitation waveforms. To get them you need inductance values. I assume that we need to make a EddyCurrent simulation first to get inductance matirx. Looks like 20 Frequency swepps is needed for that?. Then create Simplorer/Twin Builder circuit and put the inductor model from EddyCurrent solver to our circuit to get current waveforms. Then Pushback it to transient solver to get timedependent core/winding losses.

4. Can average value of time dependent core/winding losses be coupled to Icepak to make a automated Two-Way Coupling like with EddyCurrent example? 

5. Whole proces starting from getting inductance matrix with updated temperature need to be repeted. That 20 Frequency swepps in EddyCurrent solver looks like will be most time consuming, especially for 3d simulation. Can that process be accelereted somehow? Without loosing accuracy.

 

[HDLI]

Hello Marcin,

        1. I suggest to use Maxwell transient solver and provide time averaged field losses to Icepak, because thermal time constant is much bigger that electrical time constant. We could just calculate average loss and ignore duty cycle effect.

         2. Linking Maxwell transient solver would be similar with using eddy current solver. Please review this  https://courses.ansys.com/index.php/courses/etm-using-ansys-maxwell-and-icepak/ and manual. 

         I think thermal modifier provided by manufacture might look like k*equation. If Yes, I suggest to use "electrical steel" as core loss model, and this k will be for each coefficient.  

         3. No need to calculate indutance first, because current, field and loss results are included indctance calculation in either transient and eddy current solver. Indutance would be calculated by material properties and geometry automatically in FEA.

         4.  Yes.

         5. Just need coupling between Maxwell and Icepak. 

HDLI