11. Parametric Modeling of EM Behavior Using Neural Networks

A parametric EM model represents the EM behavior not only with respect to frequency or time, but also with respect to physical/geometrical variables of the EM components. The use of physical/geometrical variables for EM model is important for design purpose such as sensitivity analysis, optimization, and statistical design. When the values of the physical/geometrical variables are changed, the EM behavior will change. Using conventional EM simulation methods, the EM simulation has to be performed again each time the physical/geometrical parameters change, multiplying the computational time. In this chapter, we describe a neural network-based method for parametric modeling. The neural network is first trained to learn the EM behavior versus various values of physical/geometrical parameters, and trained neural network can be used to provide fast estimation of EM behavior during EM optimization, sensitivity analysis, and statistical design.