An Adaptive Model Order Reduction Approach for the Finite Element Method in Time Domain in Electromagnetics

Ruth Medeiros, Valentín de la Rubia · 2025

Read An Adaptive Model Order Reduction Approach for the Finite Element Method in Time Domain in Electromagnetics by Ruth Medeiros, Valentín de la Rubia (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2502-12785.txt.

Introduction

Simulating how electromagnetic fields change over time with the finite element method is accurate but slow, which limits design work and real-time analysis. Medeiros and de la Rubia present TA-ROMTD, a method that speeds these time-domain runs up while keeping them faithful.

The method alternates between the full simulation and a reduced order model, a smaller stand-in built on the fly from the full solver's own data. An error estimator judges when the small model has drifted, and the full solver then takes over briefly to refresh it, using a coarse time step to keep costs down. Tests on four antennas and structures cut run times from tens of hours to minutes, with accuracy held.

Catalog record

CREATORRuth Medeiros, Valentín de la Rubia
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE50 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2502-12785

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

An Adaptive Model Order Reduction Approach for the Finite Element Method in Time Domain in Electromagnetics Ruth Medeiros, Valentín de la Rubia Abstract Time domain simulations are crucial for analyzing transient behavior and broadband responses in electromagnetic problems. However, conventional numerical methods such as finite element method in time domain (FEMTD) and finite difference time domain, can be computationally demanding due to their high-dimensional nature, making large-scale simulations impractical for design optimization and real-time analysis. This paper introduces TA-ROMTD, a time-adaptive reduced order model (ROM) for FEMTD simulations that significantly reduces computational costs while maintaining accuracy. The method alternates between FEMTD and a reduced order model in time domain (ROMTD), using an error estimator to detect when the ROMTD solution loses accuracy and switching back to FEMTD to update the ROM with new data. Thus, TA-ROMTD does not require prior knowledge of the problem, as the ROM is constructed on the fly using FEMTD data. A key feature of this approach is the use of a coarse time step during FEMTD time intervals, capturing essential system dynamics while minimizing computational overhead. By reducing the number of degrees of freedom, this method enables efficient electromagnetic simulations, making it a powerful tool for antenna and microwave circuit design. The efficiency of the TA-ROMTD strategy is demonstrated through numerical examples, including an antipodal Vivaldi antenna, a dielectric resonator antenna, a fully metallic dual-polarization frequency selective surface, and a mushroom-type electromagnetic bandgap structure. These cases show the capability of the proposed approach to achieve accurate solutions while significantly reducing the computation time from tens of hours to minutes. I Introduction The increasing demand for fast and reliable design of antennas and microwave circuits has driven the development of advanced computational electromagnetic tools that aim to reduce simulation times without compromising the accuracy of the solution. Conventional numerical methods for solving electromagnetic problems in the time domain, such as finite difference time domain (FDTD) [ 1 , 2 , 3 , 4 , 5 ] , discontinuous Galerkin in time domain (DGTD) [ 6 , 7 , 8 , 9 , 10 , 11 , 12 , 13 , 14 , 15 ] , and finite element method in time domain (FEMTD) [ 16 , 17 , 18 , 19 , 20 , 21 , 22 , 23 ] , are widely used due to their ability to analyze complex interactions and structures. However, these methods often require substantial computational resources, especially for intricate geometries, large-scale problems, or fine computational meshes, making simulation time excessively long and complicating design optimization or real-time analysis. Model order reduction (MOR) methodologies offer a powerful alternative by transforming high-dimensional electromagnetic problems into lower-dimensional representations without compromising accuracy. By reducing the number of degrees of freedom (DoFs) required to solve the problem [ 24 ] , MOR techniques significantly decrease the computational burden [ 25 , 26 , 27 , 28 , 29 , 30 , 31 , 32 , 33 , 34 , 35 , 36 , 37 , 38 , 39 ] while preserving the essential dynamics of the system. This allows for faster simulations and more efficient exploration of diverse design possibilities. Recent efforts have focused on developing reduced order models to efficiently solve Maxwell's equations in the time domain [ 40 , 41 , 42 , 43 , 44 , 45 , 46 , 47 , 48 , 49 ] . For instance, [ 49 ] proposed a data-driven approach for solving 2D time domain Maxwell's equations for transverse magnetic waves. This method uses DGTD to generate high-dimensional solutions across multiple parameter sets, processes the data using high order dynamic mode decomposition (HODMD) [ 50 ] to create a reduced basis, and enables rapid simulations. However, this approach is computationally intensive due to the large number of simulations needed to create the database. Similar strategies are explored in [ 43 , 44 ] . In [ 47 ] , a data-driven reduced order model (ROM) was introduced to accelerate FDTD simulations of cavity problems. This approach involves solving a subset of the simulation time with FDTD to collect data for the ROM construction, which is processed with dynamic mode decomposition (DMD) [ 50 ] . The solutions in the remaining time steps are extrapolated using the computed DMD-modes. This methodology has also been applied to speed up electromagnetic particle-in-cell plasma simulations [ 51 ] . On the other hand, in [ 52 ] , the authors of this work proposed a reduced order model in time domain (ROMTD) to accelerate FEMTD simulations. The strategy uses larger FEMTD time steps during data collection to capture key dynamics efficiently, significantly reducing simulation time. However, for complex problems, even this approach can become computationally demanding because the FEMTD time interval must be sufficiently long to accurately capture the dynamics of the electromagnetic system. Another limitation of this method is its reliance on prior knowledge about the problem to determine an appropriate FEMTD time interval. If the selected interval is too short, the ROMTD approach will fail to accurately represent the solution. To address these challenges, time-adaptive MOR techniques can be employed. These approaches alternate between the high-dimensional model and the ROM, significantly reducing computational costs by limiting the number of time steps solved by the high-dimensional model. For example, [ 53 ] demonstrated the effectiveness of a time-adaptive ROM for solving the electrochemical Doyle-Fuller-Newman model [ 54 ] for lithium-ion battery cells, achieving a tenfold reduction in computational time for a 3D model. Furthermore, [ 55 ] introduced a time- and parameter-adaptive ROM strategy, yielding substantial cost reductions for battery cell design optimization. This paper proposes a time-adaptive MOR strategy for FEMTD simulations, named TA-ROMTD, which can be extended to other numerical methods, such as FDTD and DGTD. The method begins by solving a short time interval using a relatively large time step to capture the key dynamics of the electromagnetic system and generate snapshots for ROM construction via proper orthogonal decomposition (POD) [ 56 ] . The ROMTD then solves the FEMTD interval with a finer time step and continues the simulation until an error estimator indicates a loss of accuracy in the solution or until the final simulation time is reached. If the error estimator exceeds a predefined threshold, the model switches back to FEMTD for a brief period to collect additional data and update the ROM basis. Once updated, ROMTD continues solving the problem from the last ROMTD time step solved. It alternates between FEMTD and ROMTD as needed, whenever a loss of accuracy is detected. This adaptive approach reduces both the number of DoFs and the time steps solved using the time-consuming FEMTD, significantly improving computational efficiency compared to traditional methods. Additionally, the integration of an error estimator automates the resolution process, ensuring the ROM accurately captures the dynamics of the electromagnetic system without requiring prior knowledge of the problem. The structure of this paper is as follows. Section II provides an overview of the high-dimensional model and its numerical resolution. Section III describes the proposed time-adaptive ROMTD strategy and details the formulation of the reduced order model. Section IV presents numerical results demonstrating the efficiency of the method in solving electromagnetic problems. Finally, Section

Read the complete text