Skip to Main content Skip to Navigation
Theses

Building Reduced Order Models for Dynamic Simulation and Optimization of Numeric Electromagnetic Models

Abstract : There are many available methods capable of producing High-Fidelity Models (HFM) of electromagnetic systems. If the required precision is very high or the nature of the phenomenon that is being modeled is complex, a very large number of equations may have to be solved. If the model is used for applications where many different geometries or parameters must be considered, as is the case in design optimization, solving these equations many times can be very time consuming.To avoid the burden of this computation, Model Order Reduction (MOR) algorithms have been developed. They consist in procedures of finding Reduced Order Models (ROMs) that accurately describe the input/output behavior of the High-Fidelity Model but using only a very small number of equations.In this work, MORs techniques are analyzed and improved. Special attention is paid to Moment Matching. Problems like placement of expansion points, stability and numerical robustness are investigated. This has allowed the simulation and optimization of complex electromagnetic device. As examples of application we have presented a laminates bus bar and a wave scattering problem.In addition to that, a smart adaptive way of sampling the design space to allow fast optimization has been developed. The sampled points are used to perform interpolation and approximate the objective function in a very fast manner. Classic optimization techniques have also been coupled with the Reduced Order Models, accelerating the computations. The proposed approaches have been tested mainly in electromagnetic problems obtained by the Partial Equivalent Circuit Element (PEEC) method.
Document type :
Theses
Complete list of metadatas

Cited literature [123 references]  Display  Hide  Download

https://tel.archives-ouvertes.fr/tel-01886827
Contributor : Abes Star :  Contact
Submitted on : Wednesday, October 3, 2018 - 11:29:20 AM
Last modification on : Wednesday, October 14, 2020 - 4:21:33 AM
Long-term archiving on: : Friday, January 4, 2019 - 1:52:27 PM

File

ANTUNES_OLIVEIRA_LEITE_2018_di...
Version validated by the jury (STAR)

Identifiers

  • HAL Id : tel-01886827, version 1

Collections

Citation

Mateus Antunes Oliveira Leite. Building Reduced Order Models for Dynamic Simulation and Optimization of Numeric Electromagnetic Models. Electric power. Université Grenoble Alpes; Universidade federal de Minas Gerais, 2018. English. ⟨NNT : 2018GREAT034⟩. ⟨tel-01886827⟩

Share

Metrics

Record views

222

Files downloads

133