Hi together,
I’m currently working on Simulations of Microstrip components as for example Directional Coupler as seen in the picture. The simulation can be done with commercial (FEM, FDM, MM) tools. As I am interested in optimal design of such elements I would like to use Modulus for the simulations and analyse different shape parameters such as do the inverse optimization in the model space.
Is there anyone with experience how to use Modulus for simulation of HF electronic systems?
Thanks
Hi, interesting topic 👍
Yes, HF / RF structures like microstrip lines and directional couplers can be explored with physics-informed ML approaches, but usually not in a fully “replace FEM tool” way. Most people use frameworks like NVIDIA Modulus for surrogate modeling or inverse design, where you train the model on simulation data first (coming from tools like HFSS / CST / FEM solvers).
For something like a microstrip coupler, a common workflow is:
Generate dataset using a traditional EM solver (geometry → S-parameters, field distribution)
Train a neural operator / PINN in Modulus on that dataset
Then use the trained model for fast parameter sweeps or inverse optimization
Direct full-wave EM solving purely inside Modulus is still quite limited compared to dedicated RF solvers, so hybrid approach is usually the practical route.
If your goal is inverse design, Modulus can be useful, but the key challenge is good training data quality and coverage of geometry space.
