【作者】 陈玲侠；

【导师】 杨志强；

【作者基本信息】 长安大学 ， 资源与环境遥感， 2013， 博士

【摘要】 在“数字矿山”的大趋势背景下，大多数矿山企业数字化、信息化和三维可视化的建设需求非常迫切，因此矿山三维实体建模研究则成为数字矿山的重要研究内容之一。文章主要从数据准备、数据处理以及数据三维显示三个大的方面对矿山实体建模进行了研究，并将此方法应用到某矿山的三维可视化中，其研究成果对“数字矿山”的理论方法研究、研究资源勘查、矿山决策管理、矿山安全生产与矿山救援等具有重要意义。通过研究获取的主要成果如下:1、研究应用小波理论对矿山数据进行粗差探测和剔除。由于矿山三维建模的数据主要来源于钻探数据和测井资料，并通过数据综合作为建模的原始数据。在数据获取过程中，由于外界因素和人为因素的影响，采集到的数据会存在粗差和误差。首先利用小波的多分辨率特性，对原始数据进行小波分解，讨论了不同小波函数和小波的分解层数对粗差探测的影响，最终确定利用db2小波对原始数据信号进行四层分解探测；然后采用小波的三种阈值处理方法对探测到的粗差进行剔除，通过比较分析，默认阈值粗差剔除效果较好。结果表明，利用该方法进行矿山空间数据的探测具有可靠性和可行性。2、研究粒子群克里金神经网络的矿山数据插值方法。由于矿山地下数据的获取成本高，因此数据量是有限的。可采用数据插值方法挖掘出更多的信息,在地学中克里金方法应用较为广泛，其中克里金方法中变异函数模型的参数是关键，本文选择了神经网络和粒子群两者结合的优化方法对参数进行优化，避免了神经网络容易陷入局部极小的缺陷，利用了粒子群方法全局搜索的优势，提出了新的插值方法：粒子群克里金神经网络插值方法，并且通过与克里金方法、神经网络、粒子群神经网络插值方法的结果进行比较，结果表明利用该方法进行插值误差最小、结果可靠、效果好。因此利用该方法实现了研究区各地层界面的数据插值。3、研究了基于TIN和多层DEM的地层、煤层体和巷道的三维建模。首先对插值数据采用逐点插入方法构建TIN，建立不同地层地层界面的三角面片、煤层体的顶底板的三角面片以及巷道的截面和腰面的三角面片；然后采用最短对角线方法把不同层面片进行缝合，模拟地层、煤层体、巷道的真实空间分布。在三角网建立过程中利用格网索引的方法进行离散点管理以及改善了三角网的优化方法，使三角形形状更接近“肥”三角形和使建模速度提高。在巷道建模中，利用不规则三角网的建模方法实现了巷道建模；并且利用在道路研究方面圆曲线的处理方法，研究了巷道在转弯处和相交处的平滑方法。4、本文利用粗差探测方法、数据插值方法和三维实体建模方法对某矿的地层、煤层体和巷道进行了三维可视化，效果较好。更多还原

【Abstract】 Under the background of the general tendency of “digital mine”, study on3D entitymine modeling has become an increasingly important research in the field of digital mine. Inthis dissertation, it mainly studies the mine entity modeling based on data preparation, dataprocessing and3D digital display. This research achievement has an important guidingsignificance to the theory studies of “digital mine”, mineral resources exploration,miningdecision-making management,mine production and mine safety etc.The main researches ofthis dissertation are given as follows:1、The data source of the3D mine modeling is obtained mainly from the drilling data andlogging information which is established as the primary data of modeling from the aggregateanalysis. However, there exist gross error and the error inevitably in the process of the dataacquisition due to the influence of external factors and human factors. Based on the wavelettheory, it primarily deals with the gross error detecting and eliminating in the acquired data inthis dissertation. By taking advantage of the characteristic of multi-resolution analysis ofwavelet, it illustrates the decomposition of the original information and discusses the effectsof the wavelet function and the decomposition level on the gross error detection, and finallydecides that the original data signal is decomposed into the four scales by using db2wavelet.Furthermore, it explores the methods on the gross error elimination by using threshold-basedwavelet. The results show that the gross error detection of spatial data in mining based on thewavelet theory proves to be more effective and reliable.2、The acquired data for underground mines is limited due to its high costs, therefore, inorder to make the data meet the needs of modeling, the interpolation methods are researchedin the dissertation. Kriging interpolation method is widely used in the methodology, and theparameter of variation function model is key for the kringing method. In the dissertation, thisparameter is optimized by use of combination of neural network and particle swarm, whichavoids the defects of neural network falling into local minimum and makes best of searchadvantage in global optimization of the particle group method. As a result, a new interpolation method, the particle swarm Kriging neural network, is put forward, which has the smallestinterpolation error and reliable results with good effect compared with the Krigingmethod,neural network and particle swarm neural network interpolation, and will helpimprove the interpolation of the different strata interface in the research area.3、In the dissertation, it makes a thorough research on the3D modeling of strata, coalseam and laneway based on TIN and multi-DEM. Surface patches of different strata, coalbody, laneway were constructed by use of point by point method for interpolation data; then,the different levels of the TIN are sutured by using the shortest diagonal method andmulti-DEM to simulate real space distribution of strata, coal seam and laneway. The modelingspeed is boosted through building a grid index for points and improving optimization methodof TIN. In addition, laneway smoothing is further studied by use of circular curve processingmethod on the road research.3D visualization for the strata, coal seam and laneway of a mineis better achieved by using these methods.4、By use of the gross error detection method, interpolation method and3D entitymodeling method, strata, coal body and tunnel of mine was better constructed.更多还原