Enhanced CPML Based on the Autoformer Network for 2D WCS-FDTD Method.

Saved in:

Bibliographic Details
Title:	Enhanced CPML Based on the Autoformer Network for 2D WCS-FDTD Method.
Authors:	Wu, Yumeng^1,2 23040531@hdu.edu.cn, Xu, Ning^1,2 232040214@hdu.edu.cn, Li, Yexin¹, Xu, Kuiwen¹ kuiwenxu@hdu.edu.cn, Chen, Juan³ chen.juan.0201@mail.xjtu.edu.cn
Source:	Applied Computational Electromagnetics Society Journal. Jan2026, Vol. 41 Issue 1, p19-31. 13p.
Subjects:	Finite difference time domain method, Perfectly matched layers (Mathematical physics), Time complexity, Electromagnetic wave absorption, Artificial neural networks
Abstract:	This paper proposes a novel convolutional perfectly matched layer (CPML) for the weakly conditionally stable finite-difference time-domain (WCSFDTD) method. The Autoformer neural network is introduced to replace the conventional multi-layer CPML. Employing only a single-layer structure, the Auto-former-driven CPML considerably reduces both the computational domain scale and algorithmic complexity. By leveraging sequence decomposition and sparse attention mechanisms, the wave-absorption performance of this method is significantly improved. Integrated into the 2D WCS-FDTD framework, the proposed method overcomes Courant-Friedrichs-Lewy (CFL) stability constraints for FDTD intelligent absorbing boundaries, with its time step size independent of fine grid sizes in any direction. Numerical results demonstrate that the proposed method can achieve excellent wave-absorption performance with high computational efficiency, while maintaining satisfactory robustness in complex scenarios. [ABSTRACT FROM AUTHOR]
	Copyright of Applied Computational Electromagnetics Society Journal is the property of River Publishers and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)
Database:	Engineering Source

Description
Abstract:	This paper proposes a novel convolutional perfectly matched layer (CPML) for the weakly conditionally stable finite-difference time-domain (WCSFDTD) method. The Autoformer neural network is introduced to replace the conventional multi-layer CPML. Employing only a single-layer structure, the Auto-former-driven CPML considerably reduces both the computational domain scale and algorithmic complexity. By leveraging sequence decomposition and sparse attention mechanisms, the wave-absorption performance of this method is significantly improved. Integrated into the 2D WCS-FDTD framework, the proposed method overcomes Courant-Friedrichs-Lewy (CFL) stability constraints for FDTD intelligent absorbing boundaries, with its time step size independent of fine grid sizes in any direction. Numerical results demonstrate that the proposed method can achieve excellent wave-absorption performance with high computational efficiency, while maintaining satisfactory robustness in complex scenarios. [ABSTRACT FROM AUTHOR]
ISSN:	10544887
DOI:	10.13052/2026.ACES.J.410103