石油地球物理勘探2026，Vol.61Issue(3)：584-594,11.DOI:10.13810/j.cnki.issn.1000-7210.20250245
神经切线核自适应权重物理信息网络走时层析
Neural tangent kernel adaptive weight physics-informed neural network for traveltime tomography
摘要
Abstract
As a crucial approach for constructing subsurface velocity models,traveltime tomography derives the velocity distribution of the subsurface medium by solving an inverse problem constrained by observed traveltime data.However,in physics-informed neural network(PINN)traveltime tomography,the weighting between the data misfit term and the physical constraint term often relies on manual experience,making adaptive balancing difficult.This can lead to slow convergence or even cause the network to become trapped in local optima.There-fore,a neural tangent kernel(NTK)based adaptive weight optimization method for PINN traveltime tomogra-phy is proposed.First,to address the challenge of optimizing multi-objective loss functions,a dynamic weight adjustment mechanism is constructed based on the NTK theory.Second,the trace of the NTK is used to charac-terize gradient flow and to adaptively balance the contributions from the data and physical constraint terms.Fi-nally,this mechanism optimizes the training process by addressing gradient imbalance,accelerating conver-gence of the PINN,and enhancing inversion stability.Numerical experiments and real-data applications demon-strate that the proposed method improves training stability for complex velocity models and yields superior inver-sion results compared to traditional fixed-weight PINNs.关键词
走时层析/物理信息网络/神经切线核/物理约束/数据约束Key words
traveltime tomography/physics-informed neural network/neural tangent kernel/physical constraint/data constraints分类
天文与地球科学引用本文复制引用
唐杰,产嘉怡,文郑鑫,潘登,彭婧妍..神经切线核自适应权重物理信息网络走时层析[J].石油地球物理勘探,2026,61(3):584-594,11.基金项目
本项研究受国家自然科学基金项目"强吸收介质中面波和P-导波频散—衰减特征及近地表参数一体化反演"(42574157)资助. （42574157）
