Surrogate modeling of microwave circuits using polynomial functional interpolants

A new formulation for developing surrogate models using polynomial-based functional interpolants is proposed in this work. Our formulation starts from a zero-order model that can be as simple as a fixed fine model response (for cases where a continuous coarse model is not available), or it can also be a simple linear input mapped coarse model. This zero-order model is enhanced by multidimensional polynomial interpolants around a central base point in the design space. The polynomial approximation is a low-order function of the design variables, and it is used to interpolate highly accurate electromagnetic responses in a region of interest around the selected central base point. Global optimal values for the surrogate model weighting factors are efficiently obtained in closed form, using compact formulas. Our technique is illustrated by a substrate integrated waveguide interconnect with CPW transitions.