【作者】 张嘉桐；

【导师】 路来君；

【作者基本信息】 吉林大学 ， 数字地质科学， 2013， 博士

【摘要】 矿产资源是国民经济发展所需的重要物质基础，国家“十二五”规划对资源战略部署提出了明确要求。矿产资源预测是资源发现与勘察中的指导性工作。由于矿产资源预测属于系统工程，近四十年来，国内外众多学者与科研团体在矿产资源预测研究领域的不懈努力与投入，形成了以矿产资源预测理论、方法及技术为主要内容的数学地质学重要分支。进入二十一世纪以来，随着矿产资源预测理论的不断进步，以及与地学信息（计算机技术、3S技术，SAR技术）的不断融合，以矿产资源预测理论为内核，以空间数据库为基础，以精细化、智能化及三维可视化功能为目标的地学软件系统—矿产资源预测系统不断出现。由于矿产资源预测系统属于复杂的地学信息工程，涉及到跨学科数据集成及地学大数据融合分析、地质空间建模、矿产资源定位定量预测等一系列过程，其中高性能计算或云计算为地学大数据快速处理提供了高效手段，因此基于云计算条件下的矿产资源预测系统已成为当代学科的前沿研究方向；而云计算下的地学G~4I系统是导师几年来科研主攻方向之一，作者作为课题的主要参加人员承担了该系统的结构设计工作。复杂系统的结构设计是实现系统研制的先导性工作，本文所做的工作是在地学G~4I系统基础上就云计算架构的从新整合及其系统的更新设计，目的是使系统成为嵌入式的云计算重要结点，为系统的云计算升级提供理论与技术准备。我国地学软件系统开发起步较晚，自上世纪九十年代后期因加大科技投入，科技成果明显增多，功能多为地学测量及制图系统；以矿产资源预测及地质灾害预警为目标的地学软件系统在我国并不多见。由于矿产资源预测系统的技术层面涉及地学基础理论与信息领域中的许多高新技术，我国地学软件系统在质与量竞争力上均有待提高，矿产资源预测系统的水平在某种程度上代表我国地学科技竞争力，目前本领域的国内外研究趋势具有如下特点：系统赖以支撑的空间数据链多元化，数据类型涉及地质学、地球物理学、地球化学、地质遥感等众多学科，地学空间数据库的设计涉及到空间数据集合、空间数据管理、空间数据应用等若干子系统，多学科地学数据库融合与图形图像数据互操作技术成为系统的基本功能。在可能条件下，地学数据的元数据设计以成为不可缺少的内容，空间数据挖掘与空间数据仓库设计日益成为地学空间数据库研究的重要方向。目前，国内外矿产资源预测在系统集成、图形图像表达、数据库及互操作技术方面日益成熟，主要的技术热点是系统的内核理论支持与地学模型数值计算方面；同时，由于矿产资源预测中的关键环节—地质找矿模型在很大程度上依赖于找矿专家头脑的知识结构，此前，曾出现过不少地质找矿专家系统，但应用效果均不理想，甚至曾一度中断过地质成矿专家系统的开发工作。实践证明，矿产资源预测中，机器始终不能代替找矿专家的知识水平，其原因在于机器对人脑智能仿真程度尚处低级阶段，使得目前预测过程完全自动化产生的效果不尽如意。采用人机交互式操作或机器学习相结合方式效果更好。所谓人机交互是指在系统运行过程中将人工智能作为一个独立操作模块，采用人工解释与机器学习交错并行的方式执行系统设定的技术路线。由于互联网技术、虚拟现实技术及近年来云计算技术的兴起，使得基于Internet/Intranet的信息资源共享平台研制成为可能，尤其是云计算技术为复杂地学计算提供了远程服务条件，同一地学问题并行计算可扩展到无限空间范围内进行，因此将云计算技术融进矿产资源预测系统开发成为一种发展趋势。综上所述，针对矿产资源系统研制中所面临的各项技术问题，本文结合地学G~4I初级版本的技术积累，在融入矿产资源预测新理论和创新方法条件下，开展地学G~4I系统的技术集成与结构设计研究。其中重要的模块是将有关地质找矿专家的知识结构作为系统的有机部分，采用人机交互方式的执行方式构成系统一大特色，即在功能模块设计中，将机器学习理论及知识推理作为系统的智能模块，与地质专家人工操作形成互为一体式的操作模式，优化地质成矿模型与地质找矿模型的分析过程。同时，将云计算技术与G~4I系统相融合，提供云计算数据存储和计算服务，力求达到地学数据资源共享和快速运算，极大提高系统运行效率。本系统结构设计中重点包括矿产资源预测理论模型模块、空间数据库集成、机器学习及云计算技术架构等内容。更多还原

【Abstract】 The mineral resources prediction is the fastest growing and most active branches inmathematical geology that involves the theory of metallogenic prognosis, mathematicalgeological modeling, geology space database, geophysical and geochemical computingtechnology, remote sensing and computer technologies, and many other disciplines.Mineral resources prediction system is a computer system to solve the regionalquantitative evaluation of mineral resources, which is based on geology space databaseand supported by mineral resource prediction method and technique. As regional mineralresource prediction comes to massive geospatial data and numerous technologyintegrations, since the1990s, domestic and international researchers bring the GIS systemtechnology and high-performance computing technology to the evaluation process, andachieved remarkable results. Therefore, the development of specialized mineral resourcesprediction computer systems have become the general trend of geographical studies, inwhich professional system based on GIS has been accepted by the majority of geologists.For the realization of refinement of mineral resources, three-dimensional and intelligentprediction, the development of mineral resources prediction system has entered intointerdisciplinary-based spatial database integration and high-performance computing orcloud computing stage of development, the technique is much higher than the purelynumerical level in the1990s. In this situation, the development of high standards mineralresources prediction system has become the national strategic needs of mineral industries.The G~4I system is a computer system to provide refined mineral resources predictionof solutions, in which based on4G (Geography, Geology, Geochemistry, and Geophysics) geospatial data integration, and the advanced theory prediction of mineral resources as thekernel. The system’s main core technology derived from independent intellectual propertyresearch findings of mineral resource information system development project related tothe G~4I system of Jilin University Digital Geoscience Centre, the system’s coretechnologies include mineral resources prediction technology and system integrationtechnology.In order to solve the various technical problems in the development of mineralresources system, this paper combine with the accumulated technology of G~4I preliminaryversion carry out the G~4I systems technology integration and structural design study withintegrated new theories of mineral resources and innovative approaches prediction. Oneimportant module of system is that we set the knowledge structure of geologicalprospecting expert as an integral part of the system, and the human-computer interactionsystem was used as the operation mode that is a major feature of the system. By designingof functional module, the machine learning theory and knowledge reasoning as theintelligent modules of system, with the manual operation by geologist the analysis processof geological mineralization model and geological prospecting model were optimized.Meanwhile, the cloud computing technology was combined with G~4I systems thatproviding the cloud data storage and cloud computing services. The system strives toachieve the geoscience data resource sharing and rapid operation, which greatly improvesthe system’s operation efficiency. The key system structures include mineral resourceforecasting model module, spatial database integration, machine learning, cloudcomputing technology architecture, and so on.The study of cloud computing geoscience G~4I system architecture design included thefollowing:1. Mineral resources prediction system functional design packageGeoscience G~4I system is the computer system which provides whole process servicefunction for medium or large scale regional mineral resource prediction. The systemincludes three modules: a4G geospatial database, a collection of mineral resourcesprediction calculation methods, cloud computing environments functional design. The system is an integration of three modules which supported by multi-GIS-platform.2.4G geospatial database integration design4G spatial database is the data resources for the evaluation of mineral as the basicmaterial, but also constitute the basic module of system. Since4G geoscience databelonging to different disciplines, data of various disciplines comes from different sources,different measurement methods, different data conversion format, different requirement ofcalculation methods, different input and output graphics contents, but the execution of datainformation provided by all disciplines in system are submit to the prediction of mineralresources. Technically demanding integrated database design for4G, that is, using modulardesign concept unified and integrated the4G data in the underlying system; integrateddatabase space requires the interoperability feature, which is the main technology ofsystem design.3. Knowledge learning module design based on the machine learningIn system design, the current advanced technology of artificial intelligence whichbased on machine learning module was integrated on the4G database. Its main function isdesigned for solve the multiple solutions problem of geological structure characteristicline extraction by geophysical gravity and magnetic data inversion process. The systemintegrated of machine learning module, and continuously simulates the human brainintelligence thinking process of inversion interpretation expert, to provide a simulationalgorithm, called continuous learning process. By such combination of manuallyinterpreted and machine learning, minimizing the information extraction error wasarchived.4. The system architecture for cloud computingFacing the integration of4G data processing, a high requirement for hardware isdemanded for I/O of data and the transmission and transformation process. Execution ofsystem actually belongs to the high performance computing process, because the supercomputer resources are scarce, and there is contradiction between the computation timeand customer demand. For greatly saves the computational cost, the most advancedparallel computing technology at present that cloud computing is embedded system environment, the whole process of data analysis conducted entirely in the cloudarchitecture.5. Face of the internet self-diagnosis system and firewall designSince the cloud system is a node belongs to at least a LAN or WAN environment, thesystem’s firewall design is an important part too. Security of the system is necessary forany kind of computer system to strengthen the peripheral design. In addition, theself-diagnosis function and the self-repair ability of system is an effective means to ensurethe efficient operation of the system. These two elements in design of the systemenvironment must be considered as the important parts.In this paper, deeply developing of system which aiming at a series of technicaldesign problem of G~4I system based on the original results seek the optimization schemeof network design module design. Expected results to be achieved as follows:1. Realizing the optimization design of4G database and the interoperability modeldesign.2. The design of optimization method of system interface software package formineral resources prediction.3. Cloud computing small cluster system simulation program design.4. Machine learning module design for4G geoscience data integration analysis andinformation extraction.5. The hardware design of the touch-screen for expert interpretation operation andmachine learning parallel scheme.The expected goal of the system is through the research of system kernel and designof updated system structure to provide new theories and innovative technology ofprediction of mineral resources field. As the supporting system, interoperability andinteractive technology combined with the4G database, providing a high added value ofgeo information processing software tool. The theory study of system related with frontiertechnology of domestic and international research, proposed with independent intellectualproperty innovative mineral resources prediction theory and practical techniques forsolving the current three major problems exist in the field (geological characteristics of digital expression problem, nonlinear geological modeling, geological processesthree-dimensional simulation system).更多还原