煤炭是我国的主要能源资源，科学、合理的开采一直被煤炭企业所重视。了解井下煤炭质量（以下简称煤质）变化趋势及煤层结构分布情况对于煤炭企业进行生产计划、开采方案等的制定具有重要指导意义。 井下煤质预测采用的方法是根据样本点煤质信息对待预测区域做插值处理，从而达到对范围或任意点的煤质进行预测的目的。传统的插值方法没有考虑样本点之间的空间自相关性，且容易受样本点变化的影响。本文在对样本数据进行离群值分析和正态校验的基础上，给出了Kriging进行煤质预测的具体方法，并针对变差函数拟合前的步长选择问题，提出平均最邻近距离进行优化选择。实验证明，优化后的步长相比于传统的经验法确定的步长能拟合出更优的变差函数模型。同时在该步长下，分别以球状模型、指数模型、高斯模型为基础，进行了煤质预测实验，得出最适合本文数据的为指数模型。 在井下工作面三维可视化研究中，以面片表示工作面空间结构时，一般多项式曲面的数学解析式对于复杂曲面无法准确表示，甚至无法表示。因此本文将B样条引入工作面的分层模型中。通过分析煤层煤样的数据特征和存储结构，基于B样条曲面给出了工作面单层模型和多层模型的统一表达，并通过反求控制点，使得曲面可以精确的通过已知样本点。对多层模型建立过程中的缺层现象给出了虚拟煤样的解决方案，实验结果表明，B样条曲面计算简单稳定，可以取得较好的建模效果。 在深入分析浏览器端三维建模优势和Three.js三维建模原理的基础上，本文实现了浏览器端工作面的三维可视化。为了降低编码耦合，设计了面向对象的三维数据组织模型，给出了基于数据驱动的三维建模方案。并优化了数据请求时延和建模脚本文件安全性的问题，同时基于模型拾取实现了煤质指标与三维工作面的动态关联。在进行旋转、缩放、平移等交互操作时，页面均可以做出流畅的响应。最终，通过采用J2EE标准，开发了具有基础信息管理、报表自动生成、煤质信息预测以及工作面三维可视化功能的Web系统，经过实验测试效果良好，从而证明了本文研究成果的可行性与实际应用价值。

Coal is China's main energy resources. Scientific and rational mining has been the importance of coal enterprises. Understanding of coal quality changes in the trend and the distribution of coal structure for coal enterprises have important guiding significance on production plans and mining programs. The method of coal quality prediction is to interpolate the prediction area according to the sample coal quality information, so as to achieve the purpose of predicting the coal in the range or at any point. The traditional interpolation method does not consider the spatial autocorrelation between sample points, and it is easily affected by the change of sample points. Based on the outlier analysis and normal verification of the sample data, this paper gives the specific method of Kriging for coal quality prediction. In order to select the step size before the fitting of variogram function, the average nearest neighbor distance is used. The experimental results show that the optimal step size is better than the traditional empirical method to fit the better variogram model. Under the optimal step size, the coal quality prediction experiment is carried out based on the spherical model, the exponential model and the Gaussian model respectively, and the most suitable data for this model are exponential model. In the study of 3D visualization of underground working face, the mathematical expression of the general polynomial surface can’t be expressed accurately or even impossible for the complex surface. Therefore, this paper introduces the B-spline into the layered model of the working face. By analyzing the data characteristics and storage structure of coal seam samples, the uniform expression of the single-layer model and the multi-layer model are given based on the B-spline surface. And through the reverse control point, the surface can be accurately passed through the known sample points. In this paper, the virtual coal sample solution is given for the phenomenon of missing layer in the process of multi-layer model. The experimental results show that the B - spline surface calculation is simple and stable, and the better modeling effect can be obtained. Based on the deep analysis of the advantages of three-dimensional modeling of the browser and the three-dimensional modeling principle of Three.js, this paper realizes the 3D visualization of the browser-side working face. In order to reduce the coding coupling, an object-oriented three-dimensional data organization model is designed and a data - driven 3D Modeling Scheme is given. The problem of the data request delay and the security of modeling script file is optimized, and the dynamic relationship between the coal quality and the 3D working face is realized based on the model picking. In the rotation, zoom, translation and other interactive operations, the page can make a smooth response Finally, through the use of J2EE standards, developed with the basic information management, automatic report generation, coal quality information prediction and working face 3D visualization of the Web system, and the experimental results are good, which proves the feasibility and practical application value of the research results.