石油物探2017，Vol.56Issue(1)：107-115,9.DOI:10.3969/j.issn.1000-1441.2017.01.013
基于GPU/CPU和震源随机编码技术的混合域全波形反演
Hybrid domain full waveform inversion based on GPU/CPU and source random coding technique
摘要
Abstract
For the conventional waveform inversion we use the common shot gathers,which consumes too much computation capacity.The conventional phase coding technique to carry out the full waveform inversion will induce the problem of cross talk between the shots.In order to solve these problems,we propose a hybrid domain full waveform inversion based on GPU/CPU and source random coding technique.This method randomly combines the shot gathers and divides the shot gathers into groups with the same number of shot gather.Then,the common shot gathers are replaced by the combined shot gathers to carry out inversion.Compared with the traditional phase coding technique,source random coding technique can effectively increase the efficiency and the convergence rate and reduce the cross talk noise.With the acceleration of the GPU computation,the full waveform inversion efficiency of the hybrid domain is effectively improved further.The testing result with Marmousi model shows that the common shot gathers and the combined shot gathers have the same inversion result,but the computational efficiency of the combined shot gathers is significantly higher than that of the common shot gathers.Compared with the traditional phase encoding technique,crosstalk is effectively suppressed by the combined shot gathers.关键词
混合域/全波形反演/GPU/CPU/组合炮集/震源随机编码Key words
hybrid domain/full waveform inversion/GPU/CPU/combined shot gathers/source random coding分类
天文与地球科学引用本文复制引用
冯海新,刘洪,孙军,胡婷,刘志伟..基于GPU/CPU和震源随机编码技术的混合域全波形反演[J].石油物探,2017,56(1):107-115,9.基金项目
国家重点研发计划(2016YFC0601101)和中国石油集团“弹性波地震成像技术合作研发课题”(2015A-3613)联合资助.This research is financially supported by the National Key Basic Research Program of China (Grant No.2016YFC0601101) and the Collaboration Project of Seismic Imaging in Elastic Medium of China National Petroleum Corporation (Grant No.2015A-3613). （2016YFC0601101）
