In this work, we investigated the feasibility of applying deep learning techniques to solve 2D Poisson's equation. A convolutional neural network is set up predict distribution electric potential in 2D. With training data generated from a finite difference solver, strong approximation capability allows it make correct prediction given information source and permittivity. Numerical experiments show that predication error can reach below one percent, with significant reduction CPU time...

Fast and efficient computational electromagnetic simulation is a long-standing challenge. In this article, we propose data-driven model to solve Poisson's equation that leverages the learning capacity of deep techniques. A convolutional neural network (ConvNet) trained predict electric potential with different excitations permittivity distribution in 2-D 3-D models. With careful design cost function proper training data generated from finite-difference solvers, proposed enables reliable...

Many electromagnetic devices such as reflectarray antennas, metasurfaces, etc., can be modeled quasi-periodic arrays. In these devices, similar array elements with a few varying geometrical parameters are positioned in periodic lattice. Compared to modeling of arrays, efficient arrays very challenging due the loss periodicity array. this work, we applied reduced basis method integral equation solvers for modeling. scheme, new set based on is constructed through an offline process. Because...

In this work, we investigated the feasibility of applying deep learning techniques to solve Poisson's equation. A convolutional neural network is set up predict distribution electric potential in 2D or 3D cases. With proper training data generated from a finite difference solver, strong approximation capability allows it make correct prediction given information source and permittivity. applications L2 regularization, numerical experiments show that predication error cases can reach below...

In this study, we investigate the feasibility of applying deep learning technique to build a 3D electrostatic solver. A convolutional neural network (CNN) is proposed take advantage power CNN in approximation highly nonlinear functions and prediction potential distribution field. Compared with traditional numerical solvers based on finite difference scheme, method uses data-driven end-to-end model. Numerical experiments show that error can reach below 3 percent computing time be...

Quasi-periodic arrays have appeared in many electromagnetic devices, such as reflectarray antennas, metasurfaces, and nanoparticle arrays. In these elements with similar configuration are positioned on a periodic lattice. The entire array is usually large multiscale structure. Therefore, efficient modeling of quasi-periodic can be very challenging due to the loss periodicity feature array. this work, based similarity among elements, we apply reduced basis method integral equation solvers...

With the development of electromagnetic theory and microwave engineering, researchers have proposed many novel quasiperiodic arrays to realize various functions in controlling wave propagation. They similar elements located on periodic lattices. Full-wave simulation quasi-periodic is necessary challenging, because whole array usually electrically large multiscale. In this communication, we study a fast algorithm for full-wave modeling arrays. It constructs reduced basis set based geometric...

Polarization conversion metasurface has been widely studied from microwave to optical spectrum and utilized for many practical applications. In this paper, a dual-band high-efficiency cross polarization is proposed investigated. Cross unit based on perforated symmetric circular double split ring resonator which deposited dielectric backed by metal sheet. The can efficiently convert linearly polarized incident wave its reflection direction with 3 dB at Ku band (11.85 GHz ~ 18.42 GHz) K (20.76...

Planar quasi-periodic arrays are widely used in many electromagnetic devices. They can be modeled as elements with geometrical control parameters on periodic locations. In this work we apply the reduced basis method to integral equation solvers for 2D array modeling. A new set based varying is constructed through an offline process. Owing similarities among elements, number of functions each element much less than one from original mesh. Numerical examples show that both computational and...

Metasurfaces provide an exciting platform to control and manipulate incident electromagnetic wave both in transmission reflection mode. Up now, reflect with its origin direction (i.e. planar retroreflector) wide angle bandwidth is still difficult challenging. In this paper, we report a reflective metasurface unit element of circular split ring resonator combined ground metal plate sandwiched by dielectric. The numerical simulation shows that the proposed cell can produce huge phase variation...

Using simple geometrical structures, a group of novel three-dimensional (3-D) cubic loop antennas for operation at around 915 MHz ISM Band are proposed to achieve the omnidirectional patterns compact electrical size (about 36% smaller than square antenna), due fact that or meander have sizes planar ones. The feasible as receiving applications in wireless sensing base stations, RFID, multiple-input-multiple-output (MIMO) communications, and electromagnetic energy harvesting mostly...

In this study, we investigate the barycentric subdivision method for numerical integration in three-dimensional surface integral equation. This allows a uniform treatment of both singular and non-singular integrals by avoiding overlap between quadrature points source field integral. We studied convergence integration. Numerical examples also show that could achieve same level accuracy moments. Moreover, can reduce time matrix setup half hence increase computational efficiency

Quantum computer has attracted much attention from academy and industry due to the marvelous acceleration compared with classical computers. A quantum computing algorithm for Poisson equations inhomogeneous media is presented here. The original equation discretized square grids transformed into linear using finite difference method. Then solving sparse matrix applied equation. can achieve O(log N) complexity. Numerical simulations verify accuracy of such algorithm.

A symmetry relation of the integral operators in a layered medium proved by reciprocal theory can be applied to improve accuracy magnetic-field equation (MFIE). Based on symmetric Green’s function, normal vector transferred into inner keep field as elements impedance matrix. When discretized divergence-conforming this field-extracted scheme shows better performance than classic expression accuracy. Meanwhile, form combined-field (CFIE). Several numerical results are provided validate...

The dielectric coating defects existing on the surface of low detectable targets have nonnegligible impacts their electromagnetic (EM) scattering characteristics. Due to large electrical dimension target, traditional EM analysis methods are inefficient and cannot meet requirement real-time analysis. To address this issue, article proposed a U-net-based deep neural network (DNN) perform efficient center (SC) prediction for with from input 2-D geometric image it. Furthermore, facing difficulty...

The radar absorbing material (RAM) coating defects existing on the surface of stealth aircraft have important effects their electromagnetic (EM) scattering characteristics. Considering complexity and randomness large electrical dimension platform, it is necessary to analyze characteristics through a series complex time-consuming processes including geometric modeling, meshing, EM simulation, ISAR imaging, thus cannot meet requirement real-time analysis. To address this issue, paper proposed...

Based on Huygens' principle, the equivalence principle algorithm can decompose a large computational domain into several subdomains and model electromagnetic phenomenon efficiently. This paper analyses numerical projection error of operator (EPO), i.e., scattered equivalent currents surfaces projected from inside surface. Numerical studies show that accuracy magnetic current surface its far field radiation depend shape scatterer, convergence rate is usually faster for spherical than cubic...

The reduced basis method is a model-order reduction technique that converts the original system equations into smaller ones using set. This set usually derived from of solutions equation with different values control parameters. If model changes affinely parameter values, reduced-order can be derived. In this abstract, we apply to moments for quasi-periodic array modeling. With method, new single element constructed through an offline process. Because similarities among elements, number...

The equivalence principle algorithm is a domain decomposition scheme based on integral equation formulation. One of the key building blocks in this operator that computes scattered equivalent current virtual closed surfaces. In abstract, we studied numerical error inside-out procedure operator. This projects smooth electromagnetic fields three dimensions onto two-dimensional currents may contain geometrical discontinuities. We observe field projection related with geometry surfaces as well...

Planar quasi-periodic arrays are widely used in many electromagnetic devices, such as reflectarray antennas, metasurfaces, nano particle arrays, and etc.. They can be modeled elements with geometrical control parameters located at periodic locations. Modeling more challenging compared arrays. The entire array is required during the modeling process due to difference of each element. Moreover, a may include large number elements, element contains fine details. Therefore unknown very when an...