地震油气储层的小样本卷积神经网络学习与预测

地震储层预测是油气勘探的重要组成部分,但完成该项工作往往需要经历多个环节,而多工序或长周期的研究分析降低了勘探效率.基于油气藏分布规律及其在地震响应上所具有的特点,本文引入卷积神经网络深度学习方法,用于智能提取、分类并识别地震油气特征.卷积神经网络所具有的强适用性、强泛化能力,使之可以在小样本条件下,对未解释地震数据体进行全局优化提取特征并加以分类,即利用有限的已知含油气井段信息构建卷积核,以地震数据为驱动,借助卷积神经网络提取、识别蕴藏其中的地震油气特征.将本方案应用于模型数据及实际数据的验算,取得了预期效果.通过与实际钻井信息及基于多波地震数据机器学习所预测结果对比,本方案利用实际数据所演算结果与实际情况有较高的吻合度.表明本方案具有一定的可行性,为缩短相关环节的周期提供了一种新的途径.     

Mineral classification from quantitative X-ray maps using neural network: Application to volcanic rocks

We developed a mineral classification technique of electron probe microanalyzer (EPMA) maps in order to reveal the mineral textures and compositions of volcanic rocks. In the case of lithologies such as basalt that include several kinds of minerals, X-ray intensities of several elements derived from EPMA must be considered simultaneously to determine the mineral map. In this research, we used a Kohonen self-organizing map (SOM) to classify minerals in the thin-sections from several X-ray intensity maps. The SOM is a type of artificial neural network that is trained using unsupervised training to produce a two-dimensional representation of multi-dimensional input data. The classified mineral maps of in situ oceanic basalts of the Juan de Fuca Plate allowed us to quantify mineralogical and textural differences among the marginal and central parts of the pillow basalts and the massive flow basalt. One advantage of mineral classification using a SOM is that relatively many minerals can be estimated from limited input elements. By applying our method to altered basalt which contains multiple minerals, we successfully classify eight minerals in thin-section.     

一种神经网络改进小波的地震数据随机噪声去除方法

地震资料的有效信号反射弱,且易受多次波的影响,不可避免地存在随机噪声干扰。提出一种基于神经网络改进小波的地震数据随机噪声去除方法,采用神经网络模型,识别出随机噪声信号,对该信号进行小波包分解,获取多类别随机噪声信号,采用级联BP神经网络模型提取出多类别随机噪声信号,实现地震数据的随机信号压制。实验结果显示,这种改进小波方法对地震数据随机噪声信号的去噪效果较好,在复杂沉积地质结构被探测介质的地震数据随机噪声压制方面具有较强的适用性。     

Self-Organizing Maps for identification of zeolitic diagenesis patterns in closed hydrologic systems on the Earth and its implications for Mars

We have conducted a survey of zeolite occurrences in saline-alkaline paleolake deposits on Earth to identify the most prominent zeolite alteration patterns and to characterize the most common authigenic minerals and their paragenetic relationships. We collected the bulk mineral assemblages (from previous and our studies) as identified by X-ray diffraction from zeolitic tuff beds and associated sedimentary beds from thirteen paleolake deposits from the USA, Mexico, Greece, and Tanzania. We applied the Kohonen Self-Organizing Maps (SOM) to look for interesting patterns in the tuff bed mineral assemblages without prescribing any specific interpretation, and for information reduction and classification. Decision Tree (DT) method was applied to characterize these clusters. We were able to define clear class boundaries between fresh glass, non-analcime zeolites, analcime, and K feldspar. The non-analcime zeolites were further grouped into several classes based on mineral type. We also discuss the potential implications for Mars, showing that the mineral assemblages of diagenetic facies identified by SOM and DT can be used to test or validate the orbital, in situ, or modeling results, while the trained SOM provides a robust generalized ability to classify the new mineral assemblage data into the most common diagenetic facies identified in saline-alkaline paleoenvironments that contain zeolites. The study concludes that generalizing the complex geochemical behaviors using unsupervised statistical learning methods can help to identify the most prominent geochemical behaviors.     

云计算环境下震前震源异常次声波自动识别方法

云计算下采用三点阵次声源定位方法,在自动识别震前震源次声波过程中不能自动筛选识别大量的异常次声波数据,导致震前监测准确度不高且效率低下。因此提出一种云计算环境下震前震源异常次声波自动识别方法,构建JNS异常次声波数据采集筛查模组,全天候实时扫描访问端口,快速反馈异常次声波数据,采用NDS异常次声波数据序列异常检测算法快速识别错误序阵,准确回查、定位和锁定异常次声波数据;利用震前震源异常次声波自动识别方法分类识别异常次声波信号,判断该信号是否是地震可疑信号。实验结果表明,所提方法可有效自动识别震前震源异常次声波信号类型,信号分类准确率最大值达到99.99%;多次识别耗时最大均值仅为1.3min,具有准确率高和效率快的优势。     

General seismic wave and phase detection software driven by deep learning

We developed an automatic seismic wave and phase detection software based on PhaseNet, an efficient and highly generalized deep learning neural network for P- and S-wave phase picking. The software organically combines multiple modules including application terminal interface, docker container, data visualization, SSH protocol data transmission and other auxiliary modules. Characterized by a series of technologically powerful functions, the software is highly convenient for all users. To obtain the P- and S-wave picks, one only needs to prepare three-component seismic data as input and customize some parameters in the interface. In particular, the software can automatically identify complex waveforms (i.e. continuous or truncated waves) and support multiple types of input data such as SAC, MSEED, NumPy array, etc. A test on the dataset of the Wenchuan aftershocks shows the generalization ability and detection accuracy of the software. The software is expected to increase the efficiency and subjectivity in the manual processing of large amounts of seismic data, thereby providing convenience to regional network monitoring staffs and researchers in the study of Earth's interior.     

基于SOM和PSO的非监督地震相分析技术

地震相分析技术是储层预测的一种重要方法,可以用来描述有利沉积相带的分布规律.传统的地震相聚类分析方法对大数据的处理运算速度较慢,且容易陷入局部极小值,造成聚类分析的结构不准确.本文提出基于自组织神经网络(SOM)和粒子群优化方法(PSO)相结合的地震相分析技术,利用自组织神经网络能够保持原始地震数据的拓扑结构特性的特点,将大量冗余样本压缩为小样本数据,再通过粒子群的全局寻优能力改善K均值聚类的效果.理论模型和实际应用表明该方法能既有效实现数据压缩,又能提供较为准确的全局解,在地震相预测中兼顾计算效率和计算精度.     

Implementation aspects of eigendecomposition‐based high‐resolution velocity spectra

In this paper, we discuss high‐resolution coherence functions for the estimation of the stacking parameters in seismic signal processing. We focus on the Multiple Signal Classification which uses the eigendecomposition of the seismic data to measure the coherence along stacking curves. This algorithm can outperform the traditional semblance in cases of close or interfering reflections, generating a sharper velocity spectrum. Our main contribution is to propose complexity‐reducing strategies for its implementation to make it a feasible alternative to semblance. First, we show how to compute the multiple signal classification spectrum based on the eigendecomposition of the temporal correlation matrix of the seismic data. This matrix has a lower order than the spatial correlation used by other methods, so computing its eigendecomposition is simpler. Then we show how to compute its coherence measure in terms of the signal subspace of seismic data. This further reduces the computational cost as we now have to compute fewer eigenvectors than those required by the noise subspace currently used in the literature. Furthermore, we show how these eigenvectors can be computed with the low‐complexity power method. As a result of these simplifications, we show that the complexity of computing the multiple signal classification velocity spectrum is only about three times greater than semblance. Also, we propose a new normalization function to deal with the high dynamic range of the velocity spectrum. Numerical examples with synthetic and real seismic data indicate that the proposed approach provides stacking parameters with better resolution than conventional semblance, at an affordable computational cost.     

Deep Artificial Neural Networks as a Tool for the Analysis of Seismic Data

The number of seismological studies based on artificial neural networks has been increasing. However, neural networks with one hidden layer have almost reached the limit of their capabilities. In the last few years, there has been a new boom in neuroinformatics associated with the development of third-generation networks, deep neural networks. These networks operate with data at a higher level. Unlabeled data can be used to pretrain the network, i.e., there is no need for an expert to determine in advance the phenomenon to which these data correspond. Final training requires a small amount of labeled data. Deep networks have a higher level of abstraction and produce fewer errors. The same network can be used to solve several tasks at the same time, or it is easy to retrain it from one task to another. The paper discusses the possibility of applying deep networks in seismology. We have described what deep networks are, their advantages, how they are trained, how to adapt them to the features of seismic data, and what prospects are opening up in connection with their use.     

基于深度学习的地震震级分类

为了探索地震加速度时程记录的震级信息,训练卷积神经网络基于地震震级大小对地震记录进行分类,将K-NET和KiK-net中将近12万个地震记录作为样本,对其进行信息筛选和归一化,之后将地震加速度时程记录用作输入,训练卷积神经网络模型以M5.5为分类界限来区分大震和小震。结果显示,在训练集中基于该模型的分类准确率为93.6％,在测试集中的准确率为92.3％,具有良好的分类效果,这表明大震记录与小震记录之间存在一些根本的区别,即可通过地震动加速度时程记录获取一定的震级信息。      

Rock Properties and Seismic Attenuation: Neural Network Analysis

—Using laboratory data, the influence of rock parameters on seismic attenuation has been analyzed using artificial neural networks and regression models. The predictive capabilities of the neural networks and multiple linear regresssion were compared. The neural network outperforms the multiple linear regression in predicting attenuation values, given a set of input of rock parameters. The neural network can make complex decision mappings and this capability is exploited to examine the influence of various rock parameters on the overall seismic attenuation. The results indicate that the most influential rock parameter on the overall attenuation is the clay content, closely followed by porosity. Though grain size contribution is of lower importance than clay content and porosity, its value of 16 percent is sufficiently significant to be considered in the modeling and interpretation of attenuation data.     

基于深度学习卷积神经网络的地震波形自动分类与识别

发展高效、高精度、普适性强的自动波形拾取算法在地震大数据时代背景下显得越来越重要.波形自动拾取算法的主要挑战来自如何适应不同区域的不同类型地震事件的分类与筛选.本文针对地震事件-噪音分类这一问题,使用13839个汶川地震余震事件建立数据集,应用深度学习卷积神经网络(CNN)方法进行训练,并用8900个新的汶川余震事件作为检测数据集,其训练和检测准确率均达到95%以上.在对连续波形的检测中,CNN方法在精度和召回率上优于STA/LTA和Fbpicker传统方法,并能找出大量人工挑选极易遗漏的微震事件.最后,我们应用训练好的最优模型对选自全国台网的441个台站8天的连续波形数据进行了识别、到时挑取及与参考地震目录关联,CNN检出7016段波形,用自动挑选算法拾取到1380对P,S到时,并与540个地震目录事件成功关联,对1级以上事件总体识别准确率为54%,二级以上为80%,证明了CNN模型具有泛化能力,初步展示了CNN在发展兼具效率、精度、普适性算法,实时地震监测等应用上具有巨大潜力.     

地震黄土滑坡滑距预测的BP神经网络模型

地震滑坡的滑距与重力滑坡的滑距有着显著的不同,科学预测地震发生时黄土地区滑坡的滑动距离是合理评估黄土地区滑坡风险和减轻滑坡灾害的有效方式之一。基于海原特大地震诱发黄土滑坡的400组野外调查数据,通过引入BP神经网络算法,论证了BP神经网络模型用于预测黄土地震滑坡滑距的适宜性和可行性;建立了地震诱发黄土滑坡滑距的BP神经网络预测模型,并通过67组数据进行了验证。BP神经网络算法和传统多元线性回归、多元非线性回归结果的对比显示,BP神经网络的预测更接近真实情况,具有较为理想的预测效果,可以用于黄土地震滑坡滑距的预测,并为圈定较为可靠的致灾范围提供依据。     

地震与测井数据综合预测裂缝发育带

章针对前新生代海相残留盆地碳酸盐岩裂缝识别这一难题提出一种利用地震与测井数据综合预测裂缝发育带的方法.用测井资料标识出裂缝发育层段,同时利用地震资料具有空间上数据点多、分布均匀的特点通过地震多属性的人工神经网络方法将测井与地震数据结合起来综合预测裂缝发育带,充分利用测井资料和地震资料的各自优势,达到在剖面上和区域上预测裂缝发育带的目的.本方法经过实际资料的处理与预测证实比常规方法预测精度高.     

Enforced self‐organizing map neural networks for river flood forecasting

Self‐organizing maps (SOMs) have been successfully accepted widely in science and engineering problems; not only are their results unbiased, but they can also be visualized. In this study, we propose an enforced SOM (ESOM) coupled with a linear regression output layer for flood forecasting. The ESOM re‐executes a few extra training patterns, e.g. the peak flow, as recycling input data increases the mapping space of peak flow in the topological structure of SOM, and the weighted sum of the extended output layer of the network improves the accuracy of forecasting peak flow. We have investigated an ESOM neural network by using the flood data of the Da‐Chia River, Taiwan, and evaluated its performance based on the results obtained from a commonly used back‐propagation neural network. The results demonstrate that the ESOM neural network has great efficiency for clustering, especially for the peak flow, and super capability of modelling the flood forecast. The topology maps created from the ESOM are interesting and informative. Copyright © 2007 John Wiley & Sons, Ltd.     

基于深度卷积神经网络的剪切波分裂质量检测

剪切波分裂是分析地震各向异性的一种重要手段,常规方法是利用网格搜索获取分裂参数,再通过不同方法的测量结果对比测量结果进行质量检测,这一过程会耗费大量计算时间。本文针对这一问题提出了一种利用深度卷积神经网络对剪切波分裂进行质量检测的新方法,对使用了Resnet残差结构的深度神经网络进行训练,直接对二分量剪切波波形数据的质量进行分类。整个过程为:神经网络通过卷积层提取波形特征,计算损失函数后反向传播训练模型参数,完成迭代训练后的模型对输入波形数据正向计算自动输出类型。本文利用川西台站接收到的实际数据以及随机生成的合成数据分别对该网络进行训练,均可以获得准确的分类结果。相比于通过多种剪切波分裂方法对比测量结果的质量检测方法,基于神经网络的方法可以省略网格搜索的计算过程直接判断质量类型,在运算速度上的优势明显,并可继续通过训练提高模型的精度,为提升剪切波分裂方法在数据处理过程中的操作效率提供帮助。     

混沌噪声背景下检测微弱信号的神经网络方法分析

地震勘探资料的噪声许多呈现混沌现象，利用传统的去噪方法效果并不理想，如何根据混沌固有的性质，对地震勘探资料中的有效信号进行提取是许多科学工作者极为关注的问题，针对这种混沌噪声下的微弱信号检测，本文提出三种神经网络方法并对此进行比较，理论分析及仿真实验表明这三种神经网络在信噪比达到—37dB时，均能检测混沌噪声背景中的微弱信号。     

Active Source Seismic Identification and Automatic Picking of the P-wave First Arrival Using a Convolutional Neural Network

In seismic data processing, picking of the P-wave first arrivals takes up plenty of time and labor, and its accuracy plays a key role in imaging seismic structures. Based on the convolution neural network (CNN), we propose a new method to pick up the P-wave first arrivals automatically. Emitted from MINI28 vibroseis in the Jingdezhen seismic experiment, the vertical component of seismic waveforms recorded by EPS 32-bit portable seismometers are used for manually picking up the first arrivals (a total of 7242). Based on these arrivals, we establish the training and testing sets, including 25,290 event samples and 710,616 noise samples (length of each sample:2s). After 3,000 steps of training, we obtain a convergent CNN model, which can automatically classify seismic events and noise samples with high accuracy (> 99%). With the trained CNN model, we scan continuous seismic records and take the maximum output (probability of a seismic event) as the P-wave first arrival time. Compared with STA/LTA (short time average/long time average), our method shows higher precision and stronger anti-noise ability, especially with the low SNR seismic data. This CNN method is of great significance for promoting the intellectualization of seismic data processing, improving the resolution of seismic imaging, and promoting the joint inversion of active and passive sources.     

基于多尺度注意残差网络的地震波形分类研究

选用2010年2月—2016年12月发生在北京顺义及河北三河等首都圈邻近区域的117个地震事件(包括54个天然地震事件和63个非天然地震事件——爆炸事件)作为研究对象,利用文章所提出的多尺度注意残差网络对其中的天然地震事件和爆炸事件波形进行二分类。首先,对原始地震波形进行简单预处理并截取成相同长度的地震时序数据,直接将其作为网络模型的输入;其次,选用含有残差模块的深度神经网络作为基础网络,利用深度神经网络对特征的自动提取能力,省略了传统波形分类需要提前提取时域波形的特征作为分类算法输入的步骤;然后,融合通道注意力机制(ECA)并对其进行改进,将空间维度的信息融入通道信息,优化了网络对关键信息的关注,更好地聚焦重要特征;最后,使用空间金字塔池化代替最大池化进行多尺度特征融合,得到更多的特征信息,构成多尺度注意残差网络。实验结果表明,最高分类准确率为97.11%,平均分类准确率为96.53%,证明了多尺度注意残差网络在地震波形分类任务中的有效性,为震源类型识别工作提供了一种新的方法。     

Clustering spatial–temporal precipitation data using wavelet transform and self-organizing map neural network

A data analysis method is proposed to cluster and explore spatio-temporal characteristics of the 22 years of precipitation data (1982–2003) for Taiwan. The wavelet transform self-organizing map (WTSOM) framework combines the wavelet transform (WT) and a self-organizing map (SOM) neural network. WT is used to extract dynamic and multiscale features of the non-stationary precipitation time-series, and SOM is applied to objectively identify spatially homogeneous clusters on the high-dimensional wavelet-transformed feature space. Haar and Morlet wavelets are applied in the data preprocessing stage to preserve the desired characteristics of the precipitation data. A two-level SOM neural network is applied to identify clusters in the wavelet space in the clustering stage. The performance of clustering is evaluated using silhouette coefficients. The results indicate that singularities or sharp transitions are more significant than changes in the periodicity or data structure in the spatial–temporal precipitation data. The WTSOM results show that six clusters are optimal for both Haar and Morlet wavelet functions, but their corresponding geographic locations are different. The geographic locations of clusters based on the Haar wavelet, which captures the occurrence of extreme hydrological events, appear in blocks while those classified by the Morlet wavelet, which indicates periodicity changes and describes fine structures, appear in strips that cross the island of Taiwan. Principal component analysis is applied to the precipitation data of each cluster. The first principal components explain 62–90% of the total variation of data. Characteristics of precipitation data for each cluster are explored using scalogram analysis. The results show that both extreme hydrological events and periodicity changes appear in the spatial and temporal precipitation data but with different characteristics for each cluster. Recognizing homogeneous hydrologic regions and identifying the associated precipitation characteristics improves the efficiency of water resources management in adapting to climate change, preventing the degradation of the water environment, and reducing the impact of climate-induced disasters. Measures for countering the stress of precipitation variation for water resources management are provided.