Systematic Order Fitting Algorithm in Neuro-TF for Parametric Modeling of Microwave Components

This letter proposes a novel systematic order fitting technique for parametric modeling using a combined neural network and transfer function (neuro-TF) technique for electromagnetic (EM) behaviors of microwave components. The proposed technique can not only systematically determine the minimum order of the transfer function, but also systematically change the orders. By using the proposed technique, the problem of pole/residue discontinuity due to the improperly fitting can be solved, which results in a better neuro-TF modeling accuracy over the existing order-changing technique. The proposed method is demonstrated by two microwave examples.

Automated summary

This letter proposes a novel systematic order fitting technique, using a combined neural network and transfer function technique, for simulating electromagnetic behaviors of microwave components. The technique not only determines the minimum order of the transfer function, but also allows for systematic order changes, solving the issue of pole/residue discontinuity caused by improper fitting.