GlobalTomo
收藏资源简介:
GlobalTomo是由北京大学人工智能研究所创建的全球首个3D综合地震波场建模和全波形反演数据集。该数据集包含三个不同复杂度的层级,从基础的声波传播到复杂的全球尺度弹性波传播,覆盖了从地球表面到地核的范围。数据集通过先进的正演模拟优化,生成了高分辨率的地震模拟数据,适用于机器学习社区。GlobalTomo不仅支持快速正演建模和灵活的反演策略,还通过神经算子学习和自动微分技术,显著提高了效率。该数据集的应用领域包括地球内部结构的精确建模和反演,旨在解决全球地震波场建模和全波形反演中的计算挑战。
GlobalTomo is the world's first 3D comprehensive seismic wavefield modeling and full-waveform inversion dataset developed by the Institute of Artificial Intelligence, Peking University. This dataset comprises three hierarchical levels of varying complexity, ranging from basic acoustic wave propagation to complex global-scale elastic wave propagation, covering the entire span from the Earth's surface to the core. Generated via advanced forward modeling optimization, it produces high-resolution seismic simulation data tailored for the machine learning community. GlobalTomo not only supports rapid forward modeling and flexible inversion strategies, but also significantly enhances efficiency through neural operator learning and automatic differentiation technologies. Its application fields include accurate modeling and inversion of Earth's internal structure, aiming to address computational challenges in global seismic wavefield modeling and full-waveform inversion.

- 1GlobalTomo: A global dataset for physics-ML seismic wavefield modeling and FWI北京大学人工智能研究所 · 2024年



