FSS Derived Using a New Equivalent Circuit Model Backed Deep Neural Network

An equivalent circuit model backed deep neural network (DNN) is introduced to develop a broadband frequency selective surface (FSS). Shielding effectiveness (SE) and resonant frequency are inputs and shielding structure parameters are the outputs of the DNN. Moreover, three different configurations of DNN are investigated to get a minimum mean square error of 2.99 ×
$10^{-6}$
between targeted and observed outputs. The size of FSS is 0.30
$\lambda _0$
× 0.30
$\lambda _0$
and the total thickness of the shielding structure is only 0.1 mm. The variation in resonance frequency is only 0.08 GHz when the angle of incidence changes from
$\theta$
=
$0^\circ$
to
$\text{80}^\circ$
for transverse electric polarization. Furthermore, a fractional bandwidth of 80
${\%}$
is achieved with a resonance at 10 GHz. The broad shielding bandwidth, ultrathin, polarization-insensitivity, wide angular stability attained by using a new, very simple, and miniaturized geometry are a few key features of the proposed structure. The results obtained from DNN are validated using full-wave simulation and by measurement of the fabricated prototype. SE performance concurs well among all demonstrate the effectiveness of the aforementioned technique for electromagnetic interference shielding application.