A novel model-based technique to improve design processes for microstrip antennas

Abstract

The present work aims to prove the concept of a novel approach to designing microstrip antennas with desired radiation patterns without time-consuming trials and simulations. While it is pretty straightforward to design a microstrip antenna operating at a specific frequency, it requires repetitive trials to design an antenna with a specific radiation pattern. For this purpose, a unique model-based design technique is applied using the cavity model expressions in the present work. The behaviors and mutual influences of various parameters in the intended design problem are described by two graph models. The “employer model” oversees the limitations and freedoms of the antenna's parameters, while the “employee model” uses graph theory and machine learning to define the relationships between graph nodes. These two models have a dynamic structure that changes every calculation step to minimize design error. After two models are constructed, these models suggest physical parameter values according to the cavity model for the desired antenna radiation pattern. Then, the results for examples are demonstrated, and the validity of the proposed technique is proven. Finally, the developability of the method and its further works are discussed.