Antenna Modeling by Nested Kriging with Automated Domain Thickness Determination

Abstract

Fast surrogate models may alleviate the difficulties pertinent to high computational cost of electromagnetic (EM)-driven design procedures. Approximation surrogates are by far the most popular but their applicability to antenna modeling is severely limited by the curse of dimensionality. Domain confinement, as in the recently proposed nested kriging approach, offers a viable workaround this issue, in particular, enables the construction of reliable surrogates over wide ranges of antenna operating conditions and geometry parameters. Unfortunately, the original nested kriging method requires the user to set up the domain thickness (the ratio of its lateral to tangential size). The value of this parameter is critical for achieving a proper balance between the model predictive power and the cost of training data acquisition. This paper proposes a procedure for automated a priori determination of the domain thickness and highlights the computational benefits associated with the employment of the presented approach using a dual-band dipole antenna example.