Feature-Based Uncertainty Quantification

This chapter demonstrates how the response feature approach can be incorporated into uncertainty quantification procedures, especially for estimation of the fabrication yield and its optimization. We start by recalling formulation of the yield estimation and enhancement task. Subsequent sections describe algorithms for yield optimization, where response features are employed to reformulate the statistical analysis process, thereby accelerating it, as well as discuss more involved approaches where robust design procedures are further expedited by incorporating variable-resolution EM models. Furthermore, we investigate an alternative formulation of robust design task, which is tolerance optimization along with the feature-based methods of solving at low computational expenses. Finally, the use of inverse surrogates is discussed in the context of rapid yield optimization. The presented design techniques are illustrated using a number of high-frequency components such as antennas and microwave structures (couplers, filters, etc.), and compared to non-feature-based methods such as direct EM-driven statistical analysis and surrogate-assisted algorithms.