基于SVM算法的地震小断层自动识别

发布时间：2018-02-03 10:24

本文关键词： 支持向量机断层自动识别地震属性　出处：《煤炭学报》2017年11期 　论文类型：期刊论文

【摘要】：为了提高小断层解释的准确率,构建基于支持向量机(SVM)算法的断层自动识别方法。通过分析构造部位和非构造部位地震属性特征,建立SVM两分类的断层识别模型。首先,研究支持向量机两分类算法的基本原理和结构,表明支持向量机在两分类问题上具有准确率高的优点;然后建立断层正演模型,分析不同地震属性的断层响应特征,表明断层分布与属性值变化趋势相关;通过对支持向量机算法和正演模型的分析,表明利用地震属性作为支持向量机的输入,预测小断层具有可行性。从叠后地震数据中提取方差、曲率等与断层相关的属性集合;通过相关性分析和聚类分析评估属性,确定4种互相关性低的地震属性。利用14口钻井、3条巷道的地震属性和断层信息共606个数据,选取400个作为训练样本,构造SVM断层识别模型;206个数据作为测试样本,进行断层识别,识别正确率达到98%。利用地震属性建立的SVM断层自动识别模型,能够有效识别小断层,降低人为主观因素的影响,缩短了解释周期;钻孔分布越均匀、数目越多,解释精度就越高。
[Abstract]:In order to improve the accuracy of small fault interpretation, an automatic fault recognition method based on support vector machine (SVM) algorithm is proposed. The fault recognition model of SVM two-classification is established. Firstly, the basic principle and structure of SVM two-classification algorithm are studied, which shows that SVM has the advantages of high accuracy in the two-classification problem. Then the fault forward model is established, and the fault response characteristics of different seismic attributes are analyzed, which shows that the fault distribution is related to the change trend of attribute value. Through the analysis of support vector machine algorithm and forward model, it shows that it is feasible to predict small fault by using seismic attribute as input of support vector machine, and extract variance from post-stack seismic data. Curvature and other attributes related to faults; Through correlation analysis and cluster analysis, four kinds of seismic attributes with low correlation are determined. 606 seismic attributes and fault information of 14 boreholes and 3 roadways are used. The SVM fault recognition model was constructed by selecting 400 samples as training samples. 206 data are used as test samples for fault recognition, and the recognition accuracy is 98. The automatic recognition model of SVM fault based on seismic attributes can effectively identify small faults. Reducing the influence of human subjective factors and shortening the period of explanation; The more uniform the borehole distribution and the more the number, the higher the interpretation accuracy.
【作者单位】：中国矿业大学(北京)煤炭资源与安全开采国家重点实验室;
【基金】：国家自然科学基金青年基金资助项目(41402143) 国家重点研发计划资助项目(2016YFC0501102) 国家科技重大专项资助项目(2016ZX05066001-001)
【分类号】：P631.4
【正文快照】： 常规的断层解释通过观察地震剖面上振幅、相位和时差等特征识别断层,由于小断层在时间剖面上的变化微小,难以肉眼识别[1],解释结果受到解释人员主观因素影响较大。地震属性是地震数据通过数学计算得到的运动学、动力学、几何学及统计学特征。在构造解释,地层岩性解释等领域得

【相似文献】