支持向量机和数据融合在煤与瓦斯突出预测中的应用研究

发布时间：2018-04-22 04:06

本文选题：支持向量机 + 数据融合　；参考：《辽宁工程技术大学》2013年硕士论文

【摘要】：煤与瓦斯突出是矿井瓦斯灾害中最危险的灾害之一，也是最常见的灾害，对矿井灾害的预防和控制已成为矿井安全工作的重中之重。在深入了解瓦斯突出发生机理的基础上，实现瓦斯突出危险性的准确预测成为防治瓦斯灾害的主要技术手段。为此，本文针对矿井安全生产的需要，以影响矿井瓦斯突出主控因素为研究对象，以矿井瓦斯突出预警为目的，系统地研究了支持向量机和数据融合在瓦斯突出预测中的应用。本文通过研究数据融合与支持向量机的理论与方法，构建了基于支持向量机的矿井瓦斯数据融合技术框架，提出了瓦斯突出预测的分层融合模型，并确定了各层的融合算法及其完成的主要功能。在特征层融合上，选用SVM作为特征层融合算法，建立基于SVM的煤与瓦斯突出预测模型；首先对预测系统的数据来源进行分析，本文通过灰色关联分析法计算出灰色关联度来选择矿井瓦斯突出风险特征指标集，即确定影响煤矿瓦斯突出的主控因素，并把主控因素作为预测系统的特征指标，即输入数据；利用支持向量机对这些特征指标进行模拟训练，并选择合适的支持向量机核函数，利用调步长网格搜索与十折交叉验证组合的方法对支持向量机参数进行优化，实验结果表明，在经过特征层融合后能够得到很好的预测结果，但基于SVM固有的缺点，进一步提出用D-S证据理论作为决策层融合方法，利用SVM的预测结果和几个典型的指标的预测结果共同作为D-S证据理论的决策层融合证据体，，从而构成特征层和决策层的两层融合结构模型，增加了系统决策的可靠性。最后通过选取某矿区历史突出数据，对本文提出的分层融合预测模型进行了验证，结果表明通过决策层融合后，所得的预测结果更准确，表明了该方案具有很好的可行性和有效性。
[Abstract]:Coal and gas outburst is one of the most dangerous disasters in the mine gas disaster, and it is also the most common disaster. The prevention and control of mine disaster has become the most important part of mine safety work. On the basis of deep understanding of the mechanism of gas outburst, accurate prediction of gas outburst risk has become the main technical means to prevent gas disaster. For this reason, aiming at the need of mine safety production, taking the main control factors of mine gas outburst as the research object, and aiming at the early warning of mine gas outburst, this paper systematically studies the application of support vector machine and data fusion in gas outburst prediction. By studying the theory and method of data fusion and support vector machine, this paper constructs the technical framework of mine gas data fusion based on support vector machine, and puts forward a stratified fusion model for gas outburst prediction. The fusion algorithm of each layer and its main functions are determined. In the feature layer fusion, SVM is selected as the feature layer fusion algorithm to establish the coal and gas outburst prediction model based on SVM. Firstly, the data source of the prediction system is analyzed. In this paper, the grey relational analysis method is used to calculate the grey correlation degree to select the risk index set of mine gas outburst, that is to say, to determine the main control factors affecting the coal mine gas outburst, and take the main control factor as the characteristic index of the prediction system, that is, the input data. Support vector machine (SVM) is used to simulate and train these features, and appropriate kernel function of SVM is selected. The parameters of SVM are optimized by the combination of step size mesh search and 10% cross-validation. The experimental results show that, After feature level fusion, good prediction results can be obtained. However, based on the inherent shortcomings of SVM, D-S evidence theory is further proposed as a decision level fusion method. The prediction results of SVM and several typical indexes are used together as the decision level fusion evidence body of D-S evidence theory, thus the two-layer fusion structure model of feature layer and decision layer is constructed, and the reliability of system decision is increased. Finally, by selecting the historical outburst data of a mining area, the hierarchical fusion prediction model proposed in this paper is verified. The results show that the prediction results are more accurate after the decision level fusion. The results show that the scheme is feasible and effective.
【学位授予单位】：辽宁工程技术大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：TD713.3

【参考文献】