当前位置: X-MOL 学术J. Geophys. Eng. › 论文详情
Our official English website, www.x-mol.net, welcomes your feedback! (Note: you will need to create a separate account there.)
Wave-mode separation in inhomogeneous anisotropic media based on adaptive sampling
Journal of Geophysics and Engineering ( IF 1.4 ) Pub Date : 2020-08-06 , DOI: 10.1093/jge/gxaa025
Jian Wang 1 , Hong Liu 2
Affiliation  

The reverse time migration of an elastic wave is an advanced method for an imaging velocity model, and the reverse time migration of an elastic inhomogeneous TTI medium has become a research hotspot. The amount of calculation and storage is an important factor restricting its development. To solve the storage problem, we improve the wave field separation algorithm using adaptive sampling in the process of reverse time migration. This resampling algorithm can retain the core region and resample the marginal area of a quasi-differential operator. After sampling we only need to save the sampling point and the sampling location. It is proved by model calculation that the storage capacity of the algorithm can reduce a large part of the computational time of a traditional algorithm without reducing calculation accuracy. The new algorithm is more suitable for GPU because the storage is greatly reduced. In the process of reverse time migration, we can store the quasi-differential operator in the GPU so as to avoid a lot of I/O of GPU to CPU and improve calculation efficiency.

中文翻译:

基于自适应采样的非均质各向异性介质中的波模分离

弹性波的逆时偏移是成像速度模型的一种先进方法,弹性非均质TTI介质的逆时偏移已经成为研究热点。计算和存储量是限制其发展的重要因素。为了解决存储问题,在逆时偏移过程中,我们采用自适应采样改进了波场分离算法。该重采样算法可以保留核心区域并重新采样准微分算子的边缘区域。采样后,我们只需要保存采样点和采样位置。通过模型计算证明,该算法的存储容量可以减少传统算法的大部分计算时间,而不会降低计算精度。由于大大减少了存储量,因此新算法更适合GPU。在逆向时间迁移过程中,我们可以将准微分运算符存储在GPU中,从而避免了GPU对CPU的大量I / O,提高了计算效率。
更新日期:2020-09-26
down
wechat
bug