【作者】 孙晋非；

【导师】 岳建华；

【作者基本信息】 中国矿业大学 ， 地球信息科学， 2012， 博士

【摘要】 煤矿突水事故严重威胁着矿工人身安全，并对国家造成重大的经济损失。目前突水预测预报的决策多依据决策者对于问题的主观分析，这种决策方式有一定的局限性，一方面决策者对于现实情况的认识未必清楚，对于知识的掌握未必全面，另一方面采场突水是一个具有不确定性的、非线性的复杂概率事件，本文主要研究在计算机的支持下如何对于突水预测预报做出决策，考虑到与已经发生突水事故地点条件相近的地方突水的概率最大，研究将已经发生突水事故的典型案例数据作为出发点，从已知到未知，利用过去的案例或经验进行推理来求解新问题。本文主要研究成果如下：（1）分析了煤矿突水典型案例数据库设计的必要性，设计了案例数据库的概念结构、逻辑结构，使用Microsoft SQL Server关系数据库管理系统，将搜集整理的案例入库，实现煤矿历史突水资料数字化；（2）研究了三维空间数据模型，提出适用于煤矿水文地质体的基于混合结构的三维数据模型，并设计了三棱柱、TEN等数据模型的数据结构；（3）使用平行坐标方法可视化突水数据，在标准化、变换、平移数据的基础上，绘制了平行坐标图，发现灵敏属性的存在；（4）在研究模糊自适应神经网络的结构和学习算法的基础上，对于标准化方法的选用、隶属函数的选择进行比较分析，设计用于突水量预测的ANFIS模型；（5）研究了支持向量机模型的数学推理过程，设计用于底板突水量预测的SVM模型，提出参数C和γ的优化选择算法和交叉验证算法；（6）设计了基于本体的突水预测预报知识库，包括突水本体库和突水规则库，并提出了模糊规则的形式化定义。更多还原

【Abstract】 The coal mine water inrush seriously threat to the miners’ personal safety, andthe national significant economic losses. Forecasting water burst the decision-makingand more subjective analysis of the basis for decision-makers for the problem, thisdecision-making has some limitations. The one hand, the decision makers’understanding of the situation may not be clear, comprehensive know-how may not be,on the other hand the the water inrush is one of uncertainty, nonlinear complexprobability event.This study is in support of the computer about how to make decisions for waterinrush prediction, taking into account with the probability of water inrush locationsimilar water inrush. The study would have taken place in a typical case of waterinrush data as a starting point, from known to unknown, using of past cases orexperience to carry out the reasoning used to solve new problems.In this paper, research results are as follows:(1) The coal mine water inrush typical case of database is designed, includingdesign of the conceptual structure and logical structure of the case database, usingMicrosoft SQL Server relational database management system.225cases arecollected. Coal mining history of water inrush data is converted to digital data;(2) The three-dimensional spatial data model is proposed for the minehydrogeological body based on the3D data model of the hybrid structure, and thedata structure of data models are designed, including TEN, TIN and Octree;(3) The use of parallel coordinates visualization of water inrush data and Iris datafrom UCI, on the basis of standardization, transformation, translation data, drawinga parallel plot, and the existence of sensitive attributes is found;(4) Based on fuzzy adaptive neural network structure and learning algorithms,the selection of standardized methods, choice of input attributes, the choice ofmembership function for comparative analysis, building ANFIS model for waterinrush prediction;(5) Research support vector machine model for mathematical reasoning processto build the SVM model for the inrush from floor predicting water optimizationselection algorithm and cross-validation algorithm proposed parameters C and γ;(6) The water inrush prediction knowledge base based on ontology is designed,including water inrush ontology and rule base of water inrush, and a formal definitionof fuzzy rule is made.更多还原