Learning Forum

[Fatemeh Karami]

Hi all,

 

I am optimizing a structure similar to the Y-branch example (https://optics.ansys.com/hc/en-us/articles/360042305274-Inverse-design-of-y-branch) using the inverse design method. I faced an issue where many simulations are run for each iteration (for example, 11 simulations were performed for the 3rd iteration), which increases the simulation time. Another problem is that the simulation freezes after, for example, the 18th iteration (I set the maximum iteration count to 100). I would appreciate any information you might have about the source of these problems.

 

Thank you.

 

[anna.wirth-singh]

Hi Fatemeh,

It will be helpful if you can provide some more details about your simulation, perhaps a screenshot of the geometry or a plot of the Figure of Merit at each iteration. Other than the issues described, is the optimization proceeding as you would expect? Is the FOM progressing towards the desired outcome prior to freezing?

If you have modified the optimization structure, one possible explanation for these issues is that the geometry that is being optimized is extending outside the field monitor. While the structure may initially be contained within the field monitor region, it is possible that the updated structure is extending beyond the monitor region, and in this case the optimizer cannot calculate the shape derivatives. Additionally, the geometry should not exceed the mesh refinement region. A mesh refinement region should be placed over the region where the geometry may go to ensure that the mesh does not change during the optimization.

If you are using the setup script in the example, I recommend increasing the parameters size_x and/or size_y  to increase the size of the monitor and mesh override regions, and see if that resolves the issue.