A Deep Learning Application for Dolph-Tschebyscheff Antenna Array Optimisation

Abstract

This paper proposes a design for a Dolph-Tschebyscheff-weighted microstrip antenna array using a deep learning application. For this purpose, a multilayer perceptron and a deep learning model, both created using the same data set generated by a genetic algorithm, were compared. The antenna array population is initially generated randomly and then optimised with a genetic algorithm. The data produced by this model becomes a data set used for training in the deep learning application. The dimensions and specifications of the antenna array are obtained from this application, ensuring precision and optimisation in the design process. A new microstrip antenna array structure is employed for the proposed method, taking advantage of this design technique. The Dolph-Tschebyscheff weights are applied to achieve better characteristics for the microstrip antenna array, thus obtaining low side lobe levels, which are crucial for enhancing signal clarity and reducing interference. The results demonstrate that the proposed algorithm significantly improves the specifications of the structure. This improvement highlights the potential for integrating deep learning with traditional optimisation algorithms for advanced antenna design.