Enhanced surrogate models for statistical design exploiting space mapping technology

Abstract

. Abstract-We present advances in microwave and RF device modeling exploiting Space Mapping (SM) technology. We propose new SM modeling formulations utilizing input mappings, output mappings, frequency scaling and quadratic approximations. Our aim is to enhance circuit models for statistical analysis and yield-driven design. We illustrate our results using a capacitively-loaded two-section impedance transformer, a single-resonator waveguide filter and a six-section H-plane waveguide filter.

Space Mapping With Multiple Coarse Models for Optimization of Microwave Components

“…We enhance using and a standard SM modeling methodology [1], [2]. We define an enhanced coarse model as (5) with parameters and found as (6) while , are base points, e.g., the so-called star-distribution [6] with center at an optimal solution of the model , i.e., . Typically, the number of base points is between 1 to 2 1 with being the number of design variables.…”

“…Typically, the number of base points is between 1 to 2 1 with being the number of design variables. Optionally, in order to find a better starting point for SM optimization, one can perform one or more SM iterations using as a coarse model and as a fine model and use the optimization outcome as a center of the base set in (6). In practice, we often use one of the special cases of model (5).…”

“…In practice, we often use one of the special cases of model (5). If necessary, model (5) can be enhanced by other mappings, e.g., a frequency scaling [6].…”

“…The enhanced has been created using as described in Section II with input and frequency SM and the star-distribution base set (nine evaluations of model ). The evaluation time for is virtually the same as for , i.e., 1.5 s. SM optimization used the input SM surrogate enhanced by frequency SM [6]. Table I shows that the multicoarse-model SM produces a better solution than the standard SM with a smaller number of fine model evaluations.…”

Fast Space Mapping with Variable Weight Coefficients for Microwave Device Modeling

“…The standard SM modeling methodology [5] provides reasonable accuracy while using a small amount of fine model data. Increasing the amount of fine model data has, however, little or no effect on the accuracy due to parameter extraction being independent of the evaluation point of the surrogate [7].…”