论文详情

基于CPU/GPU异构平台的全波形反演及其实用化分析

石油物探

2014年 53卷第No. 4期

阅读：92

查看详情

Title

Full waveform inversion on the CPU/GPU heterogeneous platform and its application analysis

单位

(1.同济大学海洋与地球科学学院波现象与反演成像研究组,上海200092；2.中国石油化工股份有限公司胜利油田分公司物探研究

院,山东东营257022；3.浙江大学地球科学系,浙江杭州310027)

Organization

(1.Wave Phenomena and Inversion Imaging,School of Ocean and Earth Science,Tongji University,Shanghai 200092,China; 2.Shengli

Oilfield Geophysical Research Institute,SINOPEC,Dongying 257022,China; 3.Department of Earth Sciences,Zhejiang

University,Hangzhou 310027,China)

摘要

全波形反演(Full Waveform Inversion,FWI)在理论上是当前精度最高的速度估计方法。通过分析FWI的计算特点,使用图形处理器(Graphic Processing Unit,简称GPU)进行算法加速,形成了基于CPU/GPU异构平台的时空域声波方程全波形反演算法实现流程。理论模型测试结果表明,该算法不仅对速度模型具有高精度刻画能力,而且计算效率比基于CPU集群的FWI算法大幅提升。对胜利探区某陆上区块实际地震资料进行全波形速度反演试处理,取得了初步的应用效果。在此基础上,讨论了FWI对实际地震资料质量的要求,就FWI在陆上地震资料的生产性应用提出了相应的策略。

Abstract

Full waveform inversion(FWI) is theoretically the most accurate velocity estimation method nowadays.By analyzing the calculation characteristics of FWI and using Graphic Processing Unit(GPU) for FWI algorithm acceleration,we design and implement a FWI workflow of acoustic equation in time-space domain based on CPU/GPU heterogeneous platform.Compared with CPU cluster-based FWI algorithm workflow,the trial of synthetic Marmousi datasets proves that our FWI workflow has the capacity to describe the high-precision velocity model with a substantial raise of computational efficiency.We apply the FWI workflow based on

CPU/GPU heterogeneous platform for a land seismic dataset in the Shengli Oilfield and get its performance.Based on this,we discuss the requirements of FWI for the quality of seismic data and give our opinions on the productive application of FWI for land seismic data.

关键词：

地震反问题; 全波形反演; 速度估计; CPU/GPU异构平台; 陆上地震资料;

Keywords:

seismic inverse problem; full waveform inversion; velocity estimation; the CPU/GPU heterogeneous platform; land seismic data

;

基金项目

DOI

10.3969/j.issn.1000-1441.2014.04.012