Model-Based Parameter Estimation in Electromagnetic Computer Modeling

摘要

Modeling in Computational Electromagnetics (CEM) can be a numerically demanding exercise. There are essentially two factors that contribute to this situation. One is the need to describe the propagation of the electromagnetic field via the Maxwell curl equations, Green's function, mode expansions, or ray and geometrical optics. It is in this part of the problem that a source-field relationship is quantitatively developed. The other is the subsequent need to invert the source-field relationship to proceed from prescribed existing fields and known sources to the induced sources that result and the fields they consequently produce. A moment-method solution, based on an integral equation formulation, embodies both of these factors. There are basically two paths by which the computer times involved in CEM applications might be reduced. One would be the development of alternate formulations that reduce the time required for either of the activities listed above, or that eliminate the need for it completely. The geometrical theory of diffraction is one example of this path. The other would be the development of more efficient numerical approaches for implementing the moment-method model. Under this contract we have investigated several means of reducing the computation time involved in the applications of integral equation, moment-method modeling. Keywords: Mathematical models. (kr)