arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2510.04459eess.AScs.SD
收藏

用于声场重建的可微物理方法

Differentiable physics for sound field reconstruction

Department of Electrical and Photonics Engineering, Technical University of Denmark (DTU)(电气与光子工程系,丹麦技术大学)

Samuel A. Verburg, Efren Fernandez-Grande, Peter Gerstoft

AI总结:

本研究提出一种可微物理声场重建方法,用神经网络近似波动方程初始条件、可微数值求解器计算微分算子,结合稀疏约束,在极端数据稀缺下实现了比物理信息神经网络更高的精度与更好的收敛性。

摘要:

声场重建旨在从有限数量的空间分布观测值中估计声场。本研究提出了一种用于声场重建的可微物理方法:用神经网络近似波动方程的初始条件,并通过可微数值求解器计算微分算子。与将物理规律作为损失函数约束的传统物理信息神经网络(PINN)不同,该方法使用数值求解器在将物理规律作为强约束的同时实现稳定的网络训练。研究还引入了额外的稀疏促进约束,即使在严重欠采样条件下也能得到有意义的解。实验表明,该方法可在极端数据稀缺情况下重建声场,相比物理信息神经网络,精度更高、收敛性更好。

英文摘要:

Sound field reconstruction involves estimating sound fields from a limited number of spatially distributed observations. This work introduces a differentiable physics approach for sound field reconstruction, where the initial conditions of the wave equation are approximated with a neural network, and the differential operator is computed with a differentiable numerical solver. The use of a numerical solver enables a stable network training while enforcing the physics as a strong constraint, in contrast to conventional physics-informed neural networks, which include the physics as a constraint in the loss function. We introduce an additional sparsity-promoting constraint to achieve meaningful solutions even under severe undersampling conditions. Experiments demonstrate that the proposed approach can reconstruct sound fields under extreme data scarcity, achieving higher accuracy and better convergence compared to physics-informed neural networks.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑