Recent Advances in ANN-Based Methods for Yield-Driven Design of EM Structures

The design of microwave passive components should take into account the tolerances in the manufacturing process. This makes yield estimation and yield-driven design optimization essential steps in electromagnetic (EM) simulation-based microwave component design. Brute-force methods based on Monte Carlo (MC) analyses are not computationally feasible in practice due to the demand for a large number of EM simulations. Artificial neural networks (ANNs) hold a great potential to address this challenge. This paper reviews recent advances in ANN-based methods for the yield-driven design of EM structures, including a neuro-TF surrogate-assisted yield estimation method and the corresponding yield optimization algorithm. These advanced approaches are demonstrated by two microwave examples.