Local models are slowly entering the mainstream of modelling activities as they allow one to grasp the behavior of a class of systems in a single model. An example of the usefulness of such models is the design of transmission line filters. A local model will then describe the behavior of the filter for a given range of dimensions of the widths and lengths of the transmission lines.
Such models can assist a designer in the process of optimizing the design parameters in a fast and accurate way. The backside of the medal is that their extraction requires a significant amount of electromagnetic (EM) simulations for a number of possible geometric parameters. In order to compete with a massive EM-simulation based approach, the required number of simulations has to be kept as low as possible.
In this work, we will try to optimize the selection of the set of simulated systems to obtain a model with a maximum accuracy at a minimal numerical cost. Starting from the currently used heuristic methods, we will apply a combination of data and model based validation to reduce the number of required simulations. To further increase the gain in calculation cost, we will adapt the principles of optimal excitation design to cope with this multi-variable modelling framework.
Simulations are conducted using Matlab and/or ADS. Model extraction and experiment design will be handled in Matlab. The final goal is to verify the model efficiency on a real example, taken from a practical and challenging design problem.