Learning Forum

[HoomanTavasso]

Hello everyone,

I have been trying to simulate a 'simple' coil consisting of 4 connected springs with a thin circular (D=0.072mm) in Maxwell 3D. These four connected springs make a rectangle (top view) I have tried several attempts on modeling the coil in other CAD softwares and importing them into Ansys Electronics Desktop. Here is the list of attempts I made:

1- Modeling in Maxwell 3D and uniting the springs together to generate one single body - Result: current leaks to the air

2- Modeling in Solidworks and importing as .stp .stl .igs into EDT - Result: Cannot handle thin geometry (the coil will no longer have ciruclar geometry so cannot even apply a current to a terminal)

3- Modeling in NX and importing as .stp. .stl .igs into EDT - Same problem

Modeling one single spring and 2 of them connected to each other (attempt nr 1) did not give me error and 'Validation and Analyse all' was a success. But where I connected 3 or 4 of them, Ansys Electronic Desktop is giving me an error when 'Analyse all' is happening about current leak to the air which is not that straight forward to understand what's wrong, although the Validation goes well.

I also tried to model it in Design Modeler, it's interesting that first spring gets imported very well but the ones connected to it will have messed up geometry and at the end this one body geometry cannot be verified in Maxwell 3D (2 steps for simulation: Validation and Analyse all)

I would be extremely gladto be helped with this very problem and my main question is: does Ansys have hard time understanding and simulating such thin and fine geometry? is this the main reason? and eventually, what's the way out. I believe the main problem here is that I don't know what's exactly wrong. i.e. What does the current leak to the air mean? and where this current leak is taking place so I can modify my model. As a final comment, I tried both ways, open loop and closed loop.

Thanks, any comment is appreciated!

[Mark Solveson]

Please inspect the mesh.   
Select the coil objects and create a mesh plot on these objects.  Check for good mesh resolution of the geometry. 

Are these coil objects created with curved surfaces? 
You may consider going to Mesh Operations -> Initial Mesh Settings, and use the slider bar to increase the Curved Surface Meshing.  

[HoomanTavasso]

Hi Mark,

Thanks a lot for your advises. I performed mesh operation and obtained the following results with mesh operation successful completion. Yes it does contain curved surfaces. I used Resolution size 2 (one step larger than smallest). (Auto mesh method). Is there a rule of thumb number for choosing Curved Surface Meshing value on slide bar? for thin materials/wires as 0.072mm in diameter? As a second question, is such mesh plot result normal?

 

Thank you!

[Mark Solveson]

That mesh does not look great, but may provide an adequate solution.  
It can be viewed as a trade-off between simulation time and accuracy as to what is suitable for your simulation goals. 
Another approach is instead of using the slider, you can check the radio button for Manual Settings, this gives the user additional settings for the Surface Deviation and the Normal Deviation for the curved surfaces. 

[Mark Solveson]

Since this geometry looks as though it is created with "true" surfaces, that is, a circle for the cross section that is revolved around a circular circumference, the mesher creates a mesh that tries to represent these curves as best as possible, with a reasonable number of elements.  

Usually, Mesh Operations -> Initial Mesh settings gives the user adequate control of the mesh on these surfaces using either the slider, or the above mentioned Surface Deviation and Normal Deviation settings.  But, another approach would be to create the geometry with segmented objects rather than "true" surfaces.  If a segmented polygon is used for the cross section, and that is revolved around a segmented circumference (for example, revolved at 15 deg segments) then the geometry would have these geometric facet points in both the cross section of the geometry and along the circumference that the mesher will use.  The drawback of this approach is that the model only has this segmented geometry representation of the coil, and if further discretization is required, the geometry must be recreated with more segmentation.  Thus, I prefer using Mesh Operations -> Initial mesh settings to obtain a desired mesh on a "true" surface type of geometry and resort to segmentation only if necessary.  

[HoomanTavasso]

Thanks @Mark for the suggestions. I tried to first change the mesh setting as you said, considering that it takes a significant amount of time to do the trial and error to see if the parameters changed are working or not, it did not seem very trivial for my case, and I could not try this approach fully. 
However, I tried to change my wire's profile to a polygon as well as changing the revolving pattern to a polygon instead of circular to decrease my geometry's curviness. I eventually could have the simulation finishing without an error but the calculated values are not close enough to validation test I ran. (~50% error) I will try to make a more accurate geometrical approximation and hopping I could get a better validation accuracy. Thanks again!