优化卷积神经网络在道编辑中的应用

doi:10.6038/pg2019BB0387

地球物理学进展 ›› 2019, Vol. 34 ›› Issue (1): 214-220.doi: 10.6038/pg2019BB0387

• 应用地球物理学Ⅰ（油气及金属矿产地球物理勘探） • 上一篇下一篇

优化卷积神经网络在道编辑中的应用

王文强¹,孟凡顺^1,^2,^*(),孙文亮¹

^1. 中国海洋大学海洋地球科学学院,山东青岛 266100
^2. 海底科学与探测技术教育部重点实验室,山东青岛 266100

收稿日期:2018-05-19 修回日期:2018-10-23 出版日期:2019-02-20 发布日期:2019-04-15
通讯作者: 孟凡顺 E-mail:mengfsh@ouc.edu.cn
作者简介:王文强,男,1991年生,山东临沂人,硕士研究生,主要从事地震波场数值模拟及地震数据机器学习研究.(E-mail: oudywang@gmail.com)
基金资助:
国家科技重大专项“东海深层大型气田勘探评价技术”资助.(2016ZX05027-002)

Trace editing based on optimized convolutional neural network

WANG Wen-qiang¹,MENG Fan-shun^1,^2,^*(),SUN Wen-liang¹

^1. College of Marine Geosciences, Ocean University of China, Shandong Qingdao 266100, China
^2. Key Lab of Submarine Geosciences and Prospecting Techniques, Shandong Qingdao 266100, China

Received:2018-05-19 Revised:2018-10-23 Online:2019-02-20 Published:2019-04-15
Contact: Fan-shun MENG E-mail:mengfsh@ouc.edu.cn

摘要/Abstract

摘要：

由于激发、接收及工区现场等导致野外采集的地震数据出现异常道,这时需要对地震记录道编辑处理.当数据量比较大时,人工进行道编辑工作量庞大.非人工做法主要是利用计算机将异常道剔除,没有对异常道细致分类,由此造成了大量的原始数据损失,异常道的产生原因也无从得知.随着计算机性能的提高,深度学习发展迅猛,卷积神经网络(CNN)在深度学习领域起着至关重要的作用.CNN避免了前期很多工作,可以直接输入数据训练模型,将模型用于分类预测.作为一种快速高效的识别算法,可以广泛应用到各个研究领域.本文对极性反转、单频信号、强振幅噪声、空道四种常见的异常道和正常道进行细致分类编号,利用优化的深度卷积神经网络算法识别坏道并进行有效分类,不仅有利于后续对相应道的特殊处理,而且有利于推断产生异常道的原因,在以后的工作中针对产生原因做相应的工作调整.

关键词: 深度学习, 反向传播, 卷积神经网络, 随机梯度下降, 道编辑, 正常道, 异常道

Abstract:

Due to shooting, receiving, work area and so on, the abnormal traces of the seismic data often occur during field acquisition. At the meanwhile, the seismic record needs trace editing. Manual editing is a huge task when the amount of the seismic data is large. Non-artificial approach mainly removes the abnormal traces with computers. However, there is no detailed classification of abnormal traces, so this causes a loss of a large amount of original data. At the same time, we don’t even know the reason of the abnormal traces. With the improvement of computer performance, deep learning is developing rapidly and the Convolutional Neural Network (CNN) plays a great role in deep learning domain. CNN can avoid a lot of early work and we can directly input data and then use the model to predict. As a fast and efficient recognition algorithm, it can be widely applied to various research fields. In this paper, polarity reversal, single frequency signal, high amplitude noise and void trace, these four abnormal trace and normal trace will be painstakingly classified and numbered. Using the algorithm to identify and classify the bad traces effectively, not only can be conducive to the follow-up corresponding processing work, but also can benefit for engineers to learn about the reasons of the abnormal traces and we can make corresponding adjustment.

Key words: Deep learning, Backpropagation, Convolutional neural network, Stochastic gradient descent, Trace editing, Normal traces, Abnormal traces

中图分类号:

P631

王文强, 孟凡顺, 孙文亮. 2019. 优化卷积神经网络在道编辑中的应用. 地球物理学进展, 34(1): 214-220. doi: 10.6038/pg2019BB0387.

WANG Wen-qiang, MENG Fan-shun, SUN Wen-liang. 2019. Trace editing based on optimized convolutional neural network. Progress in Geophysics. 34(1): 214-220. doi: 10.6038/pg2019BB0387.

图/表 12

图1

图2

图3

图4

图5

图6

图7

图8

图9

图10

图11

参考文献 24

[1]	Cao L L, Li H T, Han Y S , et al. 2016. Application of convolutional neural networks in classification of high resolution remote sensing imagery[J]. Science of Surveying & Mapping (in Chinese), 41(9):170-175.
[2]	Cao Y S, Chen J H, Wang X Q , et al. 2015. Automatic rejection algorithm of abnormal seismic traces via singular value decomposition[J]. Progress in Geophysics (in Chinese), 30(5):2120-2124.
[3]	Liu C Z, Wang C L, Wang X . 2009. Automatic recognition and evaluation method of abnormal seismic trace among seismic data objects[J]. Petroleum Geology and Recovery Efficiency (in Chinese), 16(4):56-57, 67.
[4]	Hilterman F J . 1975. Amplitudes of seismic waves—A quick look[J]. Geophysics, 40(5):745-762. doi: 10.1190/1.1440565
[5]	Hinton G E, Salakhutdinov R R . 2006. Reducing the dimensionality of data with neural networks[J]. science, 313(5786):504-507. doi: 10.1126/science.1127647 pmid: 16873662
[6]	LeCun Y, Bengio Y, Hinton G . 2015. Deep learning[J]. Nature, 521(7553):436-444. doi: 10.1038/nature14539
[7]	LeCun Y, Bottou L, Bengio Y , et al. 1998. Gradient-based learning applied to document recognition[J]. Proceedings of the IEEE, 86(11):2278-2324. doi: 10.1109/5.726791
[8]	Li S F, Gao F Q , 2017. Handwritten numeral recognition based on convolution neural network[J]. Journal of Zhejiang Institute Sci-Tech University (in Chinese), 37(3):438-443.
[9]	McCormack M D . 1990. Seismic trace editing and first break picking using neural networks[M]. Society of Exploration Geophysicists, 321-324.
[10]	Nielsen M A. 2015. Neural networks and deep learning [J]. URL: .
[11]	Su S L, Cai X L, Zeng Q Q , et al. 2009. Automatic rejection and quality control of abnormal seismic traces under multi-constraints[J]. Progress in Exploration Geophysics (in Chinese), 32(5):334-341.
[12]	Rumelhar D E, Hinton G E, Williams D . 1986. Learning representations by back-propagating errors[J]. Nature, 323(6088):533-538.
[13]	Yi X Y, Zhou J Y , Summary of seismic attribute optimization methods[J]. Oil Geophysical Prospecting (in Chinese), 2005,40(4):482-489.
[14]	Zhang F C, Liu H Q, Niu X M , et al. 2014. High resolution seismic inversion by convolutional neural network[J]. Oil Geophysical Prospecting (in Chinese), 49(6):1165-1169.
[15]	Zhang X G, Li Y D . 1992. Automatic editing of noisy seismic data using an artificial neural network approach[J]. Chinese Journal of Geophysics (in Chinese), 35(5):637-643.
[16]	Zhu G S, Liu R L, Wang T G . 1994. Application of neural network to reservoir lateral prediction and trace editing[J]. Geophysical Prospecting for Petrol (in Chinese), ( 1):1-9.
[17]	曹林林, 李海涛, 韩颜顺 , 等. 2016. 卷积神经网络在高分遥感影像分类中的应用[J]. 测绘科学, 41(9):170-175. doi: 10.16251/j.cnki.1009-2307.2016.09.033
[18]	曹永生, 陈金焕, 王小青 , 等. 2015. 基于奇异值分解的废道自动识别算法[J]. 地球物理学进展, 30(5):2120-2124.
[19]	刘成斋, 王成礼, 王鑫 , 等. 2009. 地震数据体中非正常地震道的自动识别与评价方法[J]. 油气地质与采收率, 16(4):56-57, 67. doi: 10.3969/j.issn.1009-9603.2009.04.017
[20]	苏世龙, 蔡希玲, 曾庆芹 , 等. 2009. 多重约束下异常地震道自动剔除与质量监控[J]. 勘探地球物理进展, 32(5):334-341.
[21]	印兴耀, 周静毅 . 地震属性优化方法综述[J]. 石油地球物理勘探, 2005,40(4):482-489. doi: 10.3321/j.issn:1000-7210.2005.04.027
[22]	张繁昌, 刘汉卿, 钮学民 , 等. 2014. 褶积神经网络高分辨率地震反演[J]. 石油地球物理勘探, 49(6):1165-1169.
[23]	张学工, 李衍达 . 1992. 用人工神经网络实现地震记录中的废道自动切除[J]. 地球物理学报, 35(5):637-643.
[24]	朱广生, 刘瑞林, 王庭阁 . 1994. 神经网络在油气层横向预测和地震道编辑中的应用[J]. 石油物探, ( 1):1-9.

优化卷积神经网络在道编辑中的应用

Trace editing based on optimized convolutional neural network

RichHTML

PDF (PC)

可视化

摘要/Abstract

引用本文

使用本文

图/表 12

参考文献 24

相关文章 5

编辑推荐

多维度评价

本文评价

异常道	向量
异常道	y₁	y₂	y₃	y₄	y₅
极性反转	1	0	0	0	0
单频信号	0	1	0	0	0
强振幅噪声	0	0	1	0	0
空道	0	0	0	1	0
正常道	0	0	0	0	1

[1]	安鹏,曹丹平,赵宝银,杨晓利,张明. 基于LSTM循环神经网络的储层物性参数预测方法研究[J]. 地球物理学进展, 2019, 34(5): 1849-1858.
[2]	杨柳青,陈伟,查蓓. 利用卷积神经网络对储层孔隙度的预测研究与应用[J]. 地球物理学进展, 2019, 34(4): 1548-1555.
[3]	奚先,黄江清. 散射波场的深度学习反演成像法[J]. 地球物理学进展, 2018, 33(6): 2483-2489.
[4]	陈润航,黄汉明,柴慧敏. 地震和爆破事件源波形信号的卷积神经网络分类研究[J]. 地球物理学进展, 2018, 33(4): 1331-1338.
[5]	安鹏,曹丹平. 基于深度学习的测井岩性识别方法研究与应用[J]. 地球物理学进展, 2018, 33(3): 1029-1034.