博士论文推荐: 时频分析及其在地震勘探中的应用
#1
blueocean
发表于 11-04-2006 - 13:49
精准.石油.论坛 forum.petro-china.com
----------------------------------------------------------------------------------------------------------
本主题是专题讨论"时频分析及其在地震勘探中的应用"的一部分,欢迎访问该专题讨论的主贴:
http://forum.petro-c...?showtopic=4585
----------------------------------------------------------------------------------------------------------
此博士论文为休斯顿大学Jianlei Liu所著, 主要介绍Jianlei所研究的时频分析方法以及在地震解释中的应用, 包括, 河道的探测, 薄层厚度的预测, 时频属性的彩绘, 并涉及一点低频的应用等. 如果你对时频分析较感兴趣, 可以看一看此论文.
The title is:
Spectral Decomposition and its Application in Mapping Stratigraphy and Hydrocarbons
by Jianlei Liu
University of Houston
2006
http://www.agl.uh.ed...issertation.pdf
ABSTRACT
In this dissertation, I develop a new spectral decomposition method and apply it to mapping stratigraphy and hydrocarbons. This new spectral decomposition method combines matching pursuit concepts with a least-squares solution, providing an accurate time-frequency decomposition. The output spectral attributes include single frequency, peak frequency, peak amplitude, and peak phase volumes. Based on these spectral attributes, I can generate the composite volume of either peak frequency or phase with peak amplitude. I also show how I can plot spectral attributes which include a newly developed Red-Green-Blue display method.
I demonstrate the value of spectral decomposition through these applications. First I use spectral attributes to predict thin bed thickness of a Pennsylvanian age limestone verified by equations and a synthetic model. Next I show how to use composite volume of peak frequency, peak amplitude and coherence to detect channels in a Tertiary Gulf of Mexico as well as in a Paleozoic West Texas survey. Coherence can detect the lateral variation, while peak frequency can show the vertical thickness variation. Finally, I show how the new Red-Green-Blue display technique, which is based on three different basis functions, can be used to detect low frequency zones.
精准.石油.论坛 forum.petro-china.com本主题是专题讨论"时频分析及其在地震勘探中的应用"的一部分,欢迎访问该专题讨论的主贴:
http://forum.petro-c...?showtopic=4585
----------------------------------------------------------------------------------------------------------
此博士论文为休斯顿大学Jianlei Liu所著, 主要介绍Jianlei所研究的时频分析方法以及在地震解释中的应用, 包括, 河道的探测, 薄层厚度的预测, 时频属性的彩绘, 并涉及一点低频的应用等. 如果你对时频分析较感兴趣, 可以看一看此论文.
The title is:
Spectral Decomposition and its Application in Mapping Stratigraphy and Hydrocarbons
by Jianlei Liu
University of Houston
2006
http://www.agl.uh.ed...issertation.pdf
ABSTRACT
In this dissertation, I develop a new spectral decomposition method and apply it to mapping stratigraphy and hydrocarbons. This new spectral decomposition method combines matching pursuit concepts with a least-squares solution, providing an accurate time-frequency decomposition. The output spectral attributes include single frequency, peak frequency, peak amplitude, and peak phase volumes. Based on these spectral attributes, I can generate the composite volume of either peak frequency or phase with peak amplitude. I also show how I can plot spectral attributes which include a newly developed Red-Green-Blue display method.
I demonstrate the value of spectral decomposition through these applications. First I use spectral attributes to predict thin bed thickness of a Pennsylvanian age limestone verified by equations and a synthetic model. Next I show how to use composite volume of peak frequency, peak amplitude and coherence to detect channels in a Tertiary Gulf of Mexico as well as in a Paleozoic West Texas survey. Coherence can detect the lateral variation, while peak frequency can show the vertical thickness variation. Finally, I show how the new Red-Green-Blue display technique, which is based on three different basis functions, can be used to detect low frequency zones.
----
来源: 精准石油论坛 - 推进信息共享,提升科技水平
#4
blueocean
发表于 11-05-2006 - 00:01
精准.石油.论坛 forum.petro-china.com
做S-transform, 你可以把frequency的间隔取的小一点, 例如0.5Hz, 做出来的就比较平滑, 另外, S-transform有两类计算方法, 一类是频率域, 一类是时间域计算方法, 如果你用的是原作者写的频率域方法(1996, Stockwell), 那你可以试试时间域方法, 计算出来的效果要好一些.
至于Jianlei提出来MPD+LS, 只是采取了MPD的概念, 真正计算还是主要依赖与Least-squares.
共同探讨
精准.石油.论坛 forum.petro-china.com至于Jianlei提出来MPD+LS, 只是采取了MPD的概念, 真正计算还是主要依赖与Least-squares.
共同探讨
----
来源: 精准石油论坛 - 推进信息共享,提升科技水平
#7
blueocean
发表于 11-11-2006 - 10:32
精准.石油.论坛 forum.petro-china.com
引用框(jacklove @ 2006-11-07, 06:38 AM)
确实在一些情况下,sample freq给小点,平滑了。
你所说的时间域求法?我不明白。
原作者的st程序,是在频率域进行相乘。如果是时间域那就直接convolution?是这意思吗?
你所说的时间域求法?我不明白。
原作者的st程序,是在频率域进行相乘。如果是时间域那就直接convolution?是这意思吗?
对, 就是时间域直接做cross-correlation, 具体公式看第一章节.
----
来源: 精准石油论坛 - 推进信息共享,提升科技水平
#13
blueocean
发表于 04-03-2007 - 10:42
精准.石油.论坛 forum.petro-china.com
引用框(kine @ 2007-04-02, 01:06 AM)
想问个问题,最近看到有文献在使用RADON-WIGNER变换做时频分析,可是自己去做却无法获得相应的结果,想问下是什么原因?
特别是在使用后得到的剖面根本就不是原来得到的时频剖面?
特别是在使用后得到的剖面根本就不是原来得到的时频剖面?
我看看了RADON-WIGNER变换, 猜测是在WIGNER-VILLE变换后, 直接做RADON TRANSFORM, 也就是地球物理上所说的倾斜叠加, 对于时频分布为线性的, 此方法有一定的用处, 理想情况下得到的剖面也就是一个点, 因为数据的有限性, RADON TRANSFORM总是有假频, 注意此剖面和原始的时频剖面并不一致, 如果再做一个滤波和反RADON TRANSFORM, 才能得到提高的时频剖面.
因为地震数据的时频分布并不象CHIRP一样, 此方法应用应该不大适合.
以上仅为我的浅析
----
来源: 精准石油论坛 - 推进信息共享,提升科技水平

登录
注册
帮助

多重回复