4. Conclusion and outlook

After an overview of the various parameters calculated for electromagnetic device performance analysis, methods for establishing equivalent models were presented. These models can replace, with varying degrees of accuracy and over a limited range, the rigorous analysis of devices that is normally required. For increasingly complex devices, the use of neural networks seems to offer some potential, although their educational stage is still time-consuming, since it generally requires the use of rigorous models.

With regard to device design and optimization, we have seen how equivalent models are inserted and we have briefly presented the various techniques that enable convergent correction of the various input parameters to achieve the objectives (specifications). While so-called local algorithms, based on the quasi-Newton or gradient technique, have long been used,...