# Download Antenna Design by Simulation-Driven Optimization by Slawomir Koziel, Stanislav Ogurtsov PDF

By Slawomir Koziel, Stanislav Ogurtsov

This short stories a couple of suggestions exploiting the surrogate-based optimization idea and variable-fidelity EM simulations for effective optimization of antenna constructions. The advent of every approach is illustrated with examples of antenna layout. The authors display the ways that practitioners can receive an optimized antenna layout on the computational fee comparable to a couple of high-fidelity EM simulations of the antenna constitution. there's additionally a dialogue of the choice of antenna version constancy and its impression on functionality of the surrogate-based layout method. This quantity is acceptable for electric engineers in academia in addition to undefined, antenna designers and engineers facing computationally-expensive layout difficulties.

Best mathematics books

A p-Laplacian Approximation for Some Mass Optimization Problems

We exhibit that the matter of discovering the easiest mass distribution, either in conductivity and elasticity situations, could be approximated by way of options of a p-Laplace equation, as p→+S. This turns out to supply a variety criterion whilst the optimum recommendations are nonunique.

The Diophantine Frobenius Problem

In the course of the early a part of the final century, Ferdinand Georg Frobenius (1849-1917) raised he following challenge, often called the Frobenius challenge (FP): given particularly top optimistic integers a1,. .. ,an, locate the biggest normal quantity (called the Frobenius quantity and denoted through g(a1,. .. ,an) that's not representable as a nonnegative integer blend of a1,.

Extra resources for Antenna Design by Simulation-Driven Optimization

Example text

In each repetition, a new set of training samples is added to the existing ones. , inserting new samples at the locations where the estimated modeling error is the highest. From the antenna optimization standpoint, the main advantage of approximation surrogates is that they are very fast. Unfortunately, the high computational cost of setting up such models (mostly due to acquiring the training data) is a significant disadvantage. In order to ensure decent accuracy, hundreds or even thousands of data samples are required, and the number of training points quickly grows with dimensionality of the design space.

Relative permittivity and loss tangent of the DR core are 10 and 1e−4 respectively. 5 mm thick RO4003C material (RO4000 2010). 17 mm. Metallization of the trace and ground is with 50 μm copper. 66 GHz quadcore CPU with 4 GB RAM computer). The low-fidelity model Rc is also evaluated in CST but with coarser discretization (~15,000 mesh cells, evaluation time 24 s using the same computer). 5 shows the responses of the high- and lowfidelity model of the DRA at a certain reference design, the construction of the SPRP surrogate, and the agreement between the SPRP-predicted and the actual high-fidelity model response.

An example of a substrate integrated half-mode 5 GHz antenna shown in Fig. 1 illustrates differences in its responses evaluated with models of different fidelity as well as sensitivity of the antenna responses on the model fidelity. Both of the models are defined, discretized, and simulated using CST MWS (CST Microwave Studio 2013). 33 GHz 8 core CPU with 8 GB RAM computer. A quite dense discretization of the model, which turns in a substantial simulation time, is a result of ensuring no feasible changes of the response versus discretization density.