基于Hadoop的地震属性处理系统研究与实现

发布时间：2018-07-03 02:44

本文选题：Hadoop架构 + 地震属性分析　；参考：《中国石油大学(北京)》2016年硕士论文

【摘要】：多地震属性综合分析技术在地质勘探中有着重要的作用。通常多地震属性综合分析计算具有计算复杂度高、计算量大的特点。使用单机进行计算,计算速度受到限制,需要消耗大量的计算时间。为解决单机计算耗时长的问题,提出利用Hadoop进行分布式计算的方法。单一地震属性以SEGY格式独立保存,多种地震属性上传至HDFS时被分发至不同节点,而计算时使用多种地震属性共同参与计算,增加了节点之间数据传递的开销。通过改变地震属性保存格式,抽取同一坐标位置的不同地震属性数据组合在一起,减少计算时集群节点间不必要的数据传递。设计基于Hadoop的地震属性计算方案,将计算过程分解成地震属性数据格式转换、数据上传、地震属性计算、数据下载、计算结果格式转换5个阶段。其中地震属性数据格式转换和计算结果格式转换在单机上完成,数据上传和数据下载是单机和Hadoop集群数据交互的过程,地震属性计算使用Hadoop的Map Reduce框架在Hadoop集群上完成。通过实验验证该方案可行性,与单个节点计算相比,将系统部署在8个DataNode节点的Hadoop集群上能节约70%左右的计算时间。
[Abstract]:Multi-seismic attribute comprehensive analysis technique plays an important role in geological exploration. In general, multi-seismic attribute analysis and calculation have the characteristics of high computational complexity and large amount of computation. The calculation speed is limited by using a single machine, which requires a lot of computing time. In order to solve the time-consuming problem of single computer, a method of distributed computing using Hadoop is proposed. Single seismic attributes are stored independently in set format, and many seismic attributes are distributed to different nodes when they are uploaded to HDFS. In calculation, multiple seismic attributes are used to participate in the calculation, which increases the cost of data transfer between nodes. By changing the preservation format of seismic attributes, different seismic attribute data of the same coordinate position are extracted together to reduce the unnecessary data transfer between cluster nodes. The seismic attribute calculation scheme based on Hadoop is designed. The calculation process is decomposed into five stages: data format conversion, data upload, seismic attribute calculation, data download and result format conversion. The transformation of seismic attribute data format and computing result format is completed on a single computer. Data upload and data download are the process of data interaction between single computer and Hadoop cluster. Hadoop's Map reduce framework is used to calculate seismic attributes on Hadoop cluster. The feasibility of the scheme is verified by experiments. Compared with the calculation of single node, the system deployed on Hadoop cluster with 8 DataNode nodes can save about 70% computing time.
【学位授予单位】：中国石油大学(北京)
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TP311.52

【参考文献】