Premium
Response features for low‐cost statistical analysis and tolerance‐aware design of antennas
Author(s) -
Koziel Slawomir,
Bekasiewicz Adrian,
Cheng Qingsha S.
Publication year - 2017
Publication title -
international journal of numerical modelling: electronic networks, devices and fields
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.249
H-Index - 30
eISSN - 1099-1204
pISSN - 0894-3370
DOI - 10.1002/jnm.2297
Subject(s) - antenna (radio) , computer science , monte carlo method , point (geometry) , feature (linguistics) , tolerance analysis , surrogate model , planar , reflection (computer programming) , reliability (semiconductor) , algorithm , nonlinear system , mathematical optimization , electronic engineering , reliability engineering , mathematics , statistics , machine learning , telecommunications , engineering , power (physics) , engineering drawing , geometry , physics , linguistics , philosophy , computer graphics (images) , quantum mechanics , programming language
Abstract In this work, we discuss a methodology for rapid statistical analysis of antenna structures. The approach exploits response features, which are appropriately selected characteristic points of the antenna response that are important to determine satisfaction/violation of given performance requirements. As the dependence of the feature point coordinates on antenna geometry parameters is much less nonlinear than for original responses (here, reflection versus frequency), a reliable surrogate can be constructed at the level of the features. The surrogate is further used to conduct fast yield estimation through Monte Carlo analysis and, when coupled with sequential approximate optimization, allows for rapid tolerance‐aware design. The purpose of the latter is to maximize the probability of satisfying given performance requirements under the assumed distribution of manufacturing tolerances. Our methodology is demonstrated using a planar dual‐band antenna and compared with direct EM‐based statistical analysis as well as yield estimation using data‐driven surrogates.