PP wave & PS wave joint inversion based on GPU high performance computation platform
Cui Dong1, Zhang Yujie2, Song Jianguo3, Shao Li4
1. Research Institute of Petroleum Exploration and Development, CNPC, Beijing 100083, China;
2. Institute of Geology and Geophysics, The Chinese Academy of Sciences, Beijing 100029, China;
3. China University of Petroleum (East China), Qingdao, Shandong 266555, China;
4. NO.1 Gas Production Plant, Xinjiang Oilfield company, PetroChina, Karamay, Xinjiang 834000, China
Abstract:This paper presents our work on GPU computation acceleration for generalized linear inversion (GLI). Both theoretical analysis and model tests indicate that the data quality has been significantly improved with sparse spike L1 norm deconvolution. The proposed method is not only applicable to multi-wave data, but also to PP data and PS data alone. The following observations are obtained based on inversion results: ①L1 sparse spike deconvolution can eliminate wavelet effects before inversion, so that the GLI plays a better role in the reflection coefficient levels. ②GPU parallel computing technology can greatly improve the computation speed and efficiency. GPU speed may increase hundreds of times for a single ADCIG. Dealing with prestack data, this advantage will be more obvious. ③Coarse-grained design would be adopted in parallel computing inversion rather than fine-grained design to minimize delays and improve efficiency.
Cooke D A,Schneider W A.Generalized linear inversion of reflection seismic data.Geophysics,1983,48(6): 665-676.
[2]
Macdonald C,Davis P,Jackson D.Inversion of reflection traveltimes and amplitudes.Geophysics,1987,52(5): 606-617.
[3]
Russell B H.Introduction to Seismic Inversion Methods.SEG Books,1988.
[4]
Rijssen E,Herman G.Resolution analysis of band-limited and offset-limited seismic data for plane-layered subsurface models.Geophysical Prospecting,1991,39(1):61-76.
[5]
De Haas J,Berkhout A.On the information content of PP,P-SV,SV-SV,and SV-P reflections.SEG Technical Program Expanded Abstracts,1988,7:1190-1194.
[6]
Micikevicius P.3D finite difference computation on GPUs using CUDA.Proceedings of 2nd Workshop on General Purpose Processing on Graphics Processing Units.ACM,2009,79-84.
[7]
张兵,赵改善,黄骏等.地震叠前深度偏移在 CUDA平台上的实现.勘探地球物理进展,2009,31(6):427-432] Zhang Bing,Zhao Gaishan,Huang Jun et al.Seismic prestack depth migration on the CUDA platform.Progress in Exploration Geophysics,2009,31(6):427-432.
Liu Weifeng,Wang Yongsheng,Zhang Tianlei et al.Performance study of seismic wave propagation simulation using GPU.Journal of System Simulation,2009,21(1):170-174.
[9]
李博,刘国峰,刘洪.地震叠前时间偏移的一种图形处理器提速实现方法.地球物理学报,2009,52(1): 245-252] Li Bo,Liu Guofeng,Liu Hong.A method of using GPU to accelerate seismic pre-stack time migration.Chinese Journal of Geophysics,2009,52(1):245-252.
Liu Hongwei,Li Bo,Liu Hong et al.The algorithm of high order finite difference pre-stack reverse time migration and GPU implementation.Chinese Journal of Geophysics,2010,53(7):1725-1733.