韩令贺^1,2,,
狄帮让^1,,,
胡自多^2,3,
刘威^2,3,
王国庆^2,3,
徐中华^2,3
1. 中国石油大学(北京)油气资源与探测国家重点实验室, 北京 102249
2. 中国石油勘探开发研究院西北分院, 兰州 730020
3. 中国石油天然气集团有限公司油藏描述重点实验室, 兰州 730020

基金项目: 国家重点研发项目"面向E级计算的能源勘探高性能应用软件系统与示范"（2017YFB0202905）资助


详细信息

作者简介: 韩令贺, 男, 1987年生, 博士研究生, 主要从事波动方程数值模拟及成像方法研究.E-mail: han_lh@petrochina.com.cn

通讯作者: 狄帮让, 男, 1961年生, 教授, 主要从事地震物理模型和地震采集技术研究.E-mail: wdibr@126.com

中图分类号: P631

收稿日期:2021-01-06
修回日期:2021-05-28
上线日期:2021-09-10

Time-reversal scatterer detection method using machine learning

HAN LingHe^1,2,,
DI BangRang^1,,,
HU ZiDuo^2,3,
LIU Wei^2,3,
WANG GuoQing^2,3,
XU ZhongHua^2,3
1. China University of Petroleum(Beijing), State Key Laboratory of Petroleum Resources and Prospecting, Beijing 102249, China
2. Research Institute of Petroleum Exploration&Development-Northwest Petrochina, Lanzhou 730020, China
3. CNPC Key Laboratory of Reservoir Description, Lanzhou 730020, China


More Information

Corresponding author: DI BangRang,E-mail:wdibr@126.com

MSC: P631

--> Received Date: 06 January 2021
Revised Date: 28 May 2021
Available Online: 10 September 2021


摘要
摘要:地下小尺度散射体的检测和识别对于提高地震勘探的分辨率具有重要意义，目前业界普遍采用绕射波分离及成像方法检测地下散射体，而绕射波成像的质量主要取决于绕射波和反射波波场分离的程度.本文将被动源震源定位问题中常用的时间反转原理引入到地下散射体检测中，首先通过分析被动源和主动源模型反传波场的聚焦状态，验证了时间反转原理应用于地下散射体检测中的可行性；并引入机器学习中的朴素贝叶斯分类算法，给出适用于时间反转散射体检测的分类算法框架，计算模型中每个点成为散射体的概率，最终检测出地下散射体最有可能存在的位置.散射体模型和Sigsbee2a模型的试算结果证实了本文方法在不需对反射波和绕射波分离的情况下，即可实现对地下散射体的检测和定位，同时由于考虑了多次散射的影响，检测结果能准确反映地下散射体的位置.
关键词: 散射体检测/
时间反转/
机器学习/
朴素贝叶斯分类

Abstract:The detection and identification of subsurface small-scale scatterers is of great significance to improve the resolution of seismic exploration. At present, diffraction wave separation and imaging methods are widely used to detect subsurface scatterers, and the quality of diffraction imaging mainly depends on the degree of separation of diffraction wave and reflection wave. In this paper, the time-reversal principle, which is commonly used in passive source location, is introduced into the subsurface scatterer detection. Firstly, by analyzing the focusing status of the back propagation wavefield of the passive source and the active source, the feasibility of applying the time-reversal principle to the subsurface scatterer detection is verified. Then the naive Bayes classification algorithm in machine learning is introduced, and the classification algorithm framework suitable for time-reversal scatterer detection is given. The probability of each point in the model to become a scatterer is calculated, and finally the most likely position of subsurface scatterer is detected. The experimental results of scatterer point model and Sigsbee2a model show that the proposed method can detect and locate subsurface scatterers without separating reflection wave and diffraction wave. Meanwhile, considering the influence of multiple scattering, the detection results can accurately reflect the location of subsurface scatterers.
Key words:Scatterer detection/
Time-reversal/
Machine learning/
Naive Bayes classification



PDF全文下载地址:

http://www.geophy.cn/data/article/export-pdf?id=7419f283-17df-40dd-921b-faddc1655bda