Error prediction in electromagnetic simulations using machine learning

2019-07-01
KARAOSMANOGLU, BARISCAN
Ergül, Özgür Salih
© 2019 IEEE.We present a novel approach of using deep convolutional neural networks (CNN) to predict electromagnetic scattering errors in iterative solutions of electrically large three-dimensional objects. Deep CNN models are constructed and trained by using surface current images to predict far-zone scattering errors. Numerical experiments demonstrate successful predictions with more than 95% accuracy. The constructed models can be useful to quickly assess the accuracy of candidate solutions of current distributions via their images.

Suggestions

Electromagnetic interaction complexity reduction using deep learnin
Karaosmanoğlu, Barışcan; Ergül, Özgür Salih; Department of Electrical and Electronics Engineering (2019)
In this thesis, we present a novel approach to accelerate electromagnetic simulations by the multilevel fast multipole algorithm (MLFMA). The strategy is based on a progressive elimination of electromagnetic interactions, resulting in trimmed tree structures, during iterative solutions. To systematically perform such eliminations, artificial neural network (ANN) models are constructed and trained to estimate errors in updated surface current coefficients. These column eliminations are supported by straightf...
Parallel-MLFMA Solutions of Large-Scale Problems Involving Composite Objects
Ergül, Özgür Salih (2012-07-14)
We present a parallel implementation of the multilevel fast multipole algorithm (MLFMA) for fast and accurate solutions of large-scale electromagnetics problems involving composite objects with dielectric and metallic parts. Problems are formulated with the electric and magnetic current combined-field integral equation (JMCFIE) and solved iteratively with MLFMA on distributed-memory architectures. Numerical examples involving canonical and complicated objects, such as optical metamaterials, are presented to...
Measurement of chi(c1) and chi(c2) production with root s=7 TeV pp collisions at ATLAS
Aad, G.; et. al. (2014-07-01)
The prompt and non-prompt production cross-sections for the chi(c1) and chi(c2) charmonium states are measured in pp collisions at root s = 7TeV with the ATLAS detector at the LHC using 4.5 fb(-1) of integrated luminosity. The chi(c) states are reconstructed through the radiative decay chi c -> J/psi gamma ( with J/psi -> mu(+)mu(-)) where photons are reconstructed from gamma -> e(+)e(-) conversions. The production rate of the chi(c2) state relative to the chi(c1) state is measured for prompt and non-prompt...
Sensor Fusion of a Camera and 2D LIDAR for Lane Detection
Schmidt, Klaus Verner (null; 2019-04-26)
This paper presents a novel lane detection algorithm based on fusion of camera and 2D LIDAR data. On the one hand, objects on the road are detected via 2D LIDAR. On the other hand, binary bird’s eye view (BEV) images are acquired from the camera data and the locations of objects detected by LIDAR are estimated on the BEV image. In order to remove the noise generated by objects on the BEV, a modified BEV image is obtained, where pixels occluded by the detected objects are turned into background pixels. Then,...
Visual Result Prediction in Electromagnetic Simulations Using Machine Learning
Karaosmanoglu, Bariscan; Ergül, Özgür Salih (Institute of Electrical and Electronics Engineers (IEEE), 2019-11-01)
In this letter, we present a novel approach based on using convolutional neural networks (CNNs) to visually predict solutions of electromagnetic problems. CNN models are constructed and trained such that images of surface currents obtained at the early stages of an iterative solution can be used to predict images of the final (converged) solution. Numerical experiments demonstrate that the predicted images contain significantly better visual details than the corresponding input images. The developed approac...
Citation Formats
B. KARAOSMANOGLU and Ö. S. Ergül, “Error prediction in electromagnetic simulations using machine learning,” 2019, Accessed: 00, 2020. [Online]. Available: https://hdl.handle.net/11511/56764.