Improved fully-connected neural network approach for decoupling microstrip antenna array design

Abstract

This article presents an improved fully-connected neural network (FCNN) approach to design a decoupling microstrip antenna array. Through combing the advantage of waveguide (WG) structure and complementary double split ring resonator (CDSRRs), a novel decoupling structure is proposed, which can effectively suppress the mutual coupling between two close patches, and our improved FCNN can predict the geometry of the proposed antenna array to obtain the desired isolation. To demonstrate the validity and reliability of this method, the improved FCNN approach is used to predict the geometric parameters of the proposed antenna array when the distance of patch is reduced to 8 mm, 7mm and 6 mm. The measurement results of the predicted antenna arrays show that, their corresponding isolations are 51.16 dB, 41.99 dB and 42.43 dB respectively. The improved FCNN approach is proved to obtain the optimal isolation with lower computational complexity.

KEYWORDS:

• Improved fully-connected neural network
• decoupling microstrip antenna array
• waveguide structure
• complementary double split-ring resonator

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

This work was supported by the National Natural Science Foundation of China: [grant number 61631007,62161017, 61701208]; Tianyou Youth Talent Lift Program of Lanzhou Jiaotong University: [grant number 1520260111]; The Natural Science Foundation of Gansu Province: [grant number 20JR10RA604,21JR7RA283].