An advanced method of modeling radio-frequency (RF) devices based on a deep learning technique is proposed for accurate prediction of S parameters. The S parameters of RF devices calculated by full-wave electromagnetic solvers along with the metallic geometry of the structure, permittivity and thickness of the dielectric layers of the RF devices are used partly as the training and partly as testing data for the deep learning structure. To implement the training procedure efficiently, a novel selection method of training data considering critical points is introduced. In order to rapidly and accurately map the geometrical parameters of the RF devices to the S parameters, deep neural networks are used to establish the multiple non-linear transforms. The hidden-layers of the neural networks are adaptively chosen based on the frequency response of the RF devices to guarantee the accuracy of generated model. The Adam optimization algorithm is utilized for the acceleration of training. With the established deep learning model of a parameterized device, the S parameters can efficiently be obtained when the device geometrical parameters change. Comparing with the traditional modeling method that uses shallow neural networks, the proposed method can achieve better accuracy, especially when the training data are non-uniform. Three RF devices, including a rectangular inductor, an interdigital capacitor, and two coupled transmission lines, are used for building and verifying the deep neural network. It is shown that the deep neural network has good robustness and excellent generalization ability. Even for very wide frequency band (0–100 GHz), the maximum relative error of the coupled transmission lines using the proposed method is below 3%.

One major difficulty of parameter extraction for an equivalent circuit lies in applying an optimization algorithm to efficiently approach the global minimum. Traditional genetic algorithm (GA) is able to find the global minimum but it has very low convergence rate. On the other hand, direct search method may be easily trapped in local minima. In this work, an improved parameter extraction method which is capable of effectively performing the global minimum search is proposed. This method combines GA and Powell’s method to efficiently determine circuit elements of an equivalent circuit model. A spiral inductor and an interdigital capacitor are used as design examples to illustrate the extraction procedure and validate the proposed hybrid method.

A compact high-selective band-pass frequency selective surface (FSS) with the unit cell less than λ/7 is presented. For the simulation of the structure, the multilevel Green’s function interpolation method (MLGFIM) using Floquet theory is adopted to accelerate the calculation of the complex unit cell. The radial basis function (RBF)-QR method is used in the interpolation, which makes the shape parameter in the RBF function not required to be retested for different periodicity. In this design, with an aperture coupling structure between the top and bottom layers patterned by triangular patches and meander lines, the FSS has two transmission zeros (TZs) on both sides of the pass-band and achieves a steep roll-off rate of 192 dB/GHz. Consequently, the FSS has high selectivity and out-of-band suppression, besides profiting from the low profile and symmetric geometry, this FSS exhibits good angular and polarization stabilities. The prototype of the proposed FSS is fabricated and good performance is obtained.