【作者】 母丽华；

【导师】 沈继红；

【作者基本信息】 哈尔滨工程大学 ， 系统工程， 2010， 博士

【摘要】 煤炭作为我国重要的能源与燃料,占能源消费总量的比例大约75%左右。由于煤炭工业的基础地位,实现安全、高效地可持续发展是中国实现经济重大战略目标的可靠保证。近几年,我国重大煤矿事故的不断发生,说明我国的煤矿安全形势没有得到根本性的好转。煤矿安全是煤炭工业目前最突出、最需要解决的关键问题。本文将利用突变、再生核神经网络、支持向量机等基本原理给出矿井安全系统中的煤与瓦斯突出、地下煤层面重构及地表沉陷等方面进行安全预测的方法,为矿井安全预测提供有力的理论依据。1、突变理论是奇点理论和分岔理论研究不连续变化现象的理论,是一个新兴的数学分支。煤与瓦斯突出是发生在矿井煤岩体中的灾害动力现象,是含瓦斯煤岩体的一种剧烈的能量释放形式。基于力学或能量的观点,突出从孕育到启动和从发展到结束都具有明显的突变特征。本文将根据煤与瓦斯突出产生的条件,对煤与瓦斯突出的综合作用过程和突出过程做出系统的理论分析,运用突变学理论,建立燕尾突变模型,实现安全预测。2、再生核源于不同学科分支,目前已成为函数逼近的重要工具。将再生核与神经网络有机地结合起来,提出一种新型的再生核神经网络。它将网络的训练归结为求解线性方程组问题,建立了一种既有足够精度又能表现系统行为的稀疏解数学模型。这种模型对于断层面的重构方法与样条方法比较,更符合实际情况。本文将利用再生核神经网络对断层面进行重构,及时了解采煤面的变化,预防井塌事故的发生,实现安全预测。3、煤矿开采所产生的沉降受多个因素的影响,并且每个因素对其作用的函数关系很难界定。因此煤矿开采的地表沉降预测是属于复杂的非线性系统问题。支持向量机理论是对数据挖掘和处理的新方法。从统计学的概率角度考虑比神经元网络更精确,是解决煤矿开采问题的更为有效方法。本文将运用支持向量机方法,通过分析矿区水文地质条件和沉降基本因素,处理地表沉降数据。得到更符合实际的地表沉降的预测模型。给出矿区沉降的预测方法,实现安全预测。更多还原

【Abstract】 Coal mine, as China’s important energy source and fuel, has the proportion of the total energy demand, about 75% or so. Since the fundamental position, secure implementation, and efficient and sustainable development of coal industry are the reliable guarantee for attaining great economic strategic objectives of China. In recent years, the repeated occurrence of major coal accidents indicates that the coal security situation has not been improved fundamentally. The problem of coal mine security is the most remarkable and urgent one in China that should be solved at present.This article will give some methods to predict coal and gas outburst, reconstruction of underground coal delamination, and surface subsidence in mine security system, which utilizes the basic principles of mutation, reproducing kernel neural system and support vector machines. It provides the strong evidences for the safe prediction of mine based on theory. Catastrophe theory is a new subdivision of mathematics, a theoretical research studied by singularity theory and bifurcation theory about the discontinous change phenomenon. Coal and gas outburst is a catastrophic dynamic phenomena happening in the mine, coal and rock mass, also a form of violent energy releasing of coal-rock with gas. Based on the opinions of mechanics and energy, distinct mutation features are stressed from gestation to sartup and from development to the end. According to the conditions of producing coal and gas outburst, theroretical analysis is made systematically in this article to the synthetically effects and outburst process in the coal and gas outburst, on the other hand,catastrphe theory is applied, which are both used to build swallowtail catastrophe model and to fulfill secure prediction.Reproducing kernel originating from different branches, which has become an important tool of approximation currently. Combining reproducing kernel with neural networks organically aims to put forward a new type of reproducing kernel-neural network. his new type contributes network training to the problem solving linear equations, and set up a sparse solution of mathematical model with sufficient accurancy and performing of system behavior. Compared with the reconstruction of fault plane and spline, this model is more in line with the actual situation. This article will take advantage of reproducing kernel-neurual network in order to reconstruct the fault plane, learn changes on surface coal mining, prevent the accidents of well collapse and achieve the secure prediction.Precipitation produced by coal mining is effected by a number of factors. The relationship between each factor and the function it plays is hard to define. Therefore, the prediction of surface coal mining subsidence belongs to the complicated nonlinear system problems. Support vector machine theory is a new method to data mining and processing. Taking the probability viewpoint of statistics into consideration, it is more accurate than neurual network, and the best way to solve the coal mining problem. This article applies the methods for support vector machines and the analysis of hydro-geological conditions in mine area, settlement and other basic factors lement data, get the prediction model of surface settlement that is more realistic, give the prediction of mine area settlement and achieve security forecast.更多还原