Design centering of microwave systems using normed distances and kriging surrogate models

Abstract

A new geometrical design centering-based approach for microwave systems is proposed. In this approach, a norm related to the system parameters probability distribution is used to determine distances from a point to the boundary of the parameter feasible space (feasible region) that it belongs to. Building on the knowledge on the obtained normed distances, a max-min optimization problem is formulated to describe the design centering problem at hand. To reduce the expected expensive computational overhead associated with the microwave system design cycle, computationally efficient kriging surrogates are utilized to construct an iterative sequence of updated feasible region approximations. In each of which, the normed distances-based centering process is applied to develop a better center point. This process is repetitively executed till an optimal design center is achieved. Practical microwave systems are provided to demonstrate the effectiveness of the proposed design centring technique namely: the band-stop micro-strip filter employing Sonnet’s em and the ultra-wideband (UWB) multiple-input multiple-output (MIMO) antenna employing CST Studio Suit. The UWB MIMO antenna final design is also fabricated where experimental results show consistency with simulations.