【作者】 孙忠林；

【导师】 谭允祯；

【作者基本信息】 山东科技大学 ， 安全技术及工程， 2009， 博士

【摘要】 论文以现代预测理论和技术为基础,以“安全第一,预防为主”为研究前提,深入分析了矿山安全生产系统的安全管理和安全控制的现状,针对煤矿安全生产系统的安全状态预测理论和技术进行研究,给出了煤矿安全生产系统、子系统、区域、安全状态、安全度和安全等级等基本概念;提出了煤矿安全状态转换模型——SAFU模型,利用SAFU模型对煤矿区域、组合区域和功能子系统进行安全状态的预测,以预测的安全状态类型确定引起煤矿区域安全状态变化的因素,为煤矿安全的预测预报和安全管理提供决策信息和依据。论文对回归分析预测方法、BP神经网络束集预测模型方法、灰色预测方法和支持向量机预测模型进行了分析研究,讨论了预测模型在煤矿安全生产预测的应用方式,给出了参数数量不同的情况下统一使用回归预测方法的方式;束集模型减少了神经网络的关联数量,提高了网络模型的迭代速度和预测效率;用较少的分类样本矢量实现了支持向量机的煤矿安全状态预测。使用SAFU模型,描述了煤矿五种安全状态的变化模式,提出了以安全度值逆向确定煤矿安全生产危险因素的算法。对定义的煤矿区域和组合区域分析了区域安全状态的相关性,给出了确定连通区域的算法,提出了区域安全度和基于连通区域计算安全度的算法。给出了单参数区域的安全度预测方法和多参数区域的预测方法。证明了煤矿区域安全生产整体预测的有效性、煤矿区域安全度的存在性和可量化性,推导出煤矿区域安全度是可以预测的结论。给出了利用样本数据训练预测模型、选择预测模型的算法。提出了安全状况的经验模式发现、描述和进行预测的最长匹配序列模式,使无法用函数描述的区域安全状况,实现了模式识别的预测。提出了预测模型的三种基本组合模式,引入参数映射实现了预测模型的组合构建。给出了煤矿安全生产区域预测模型的选择、组合和模型调度方法。分析设计了煤矿安全生产安全状态的预测系统的业务流程分析、预测功能和数据处理。设计了预测系统的系统体系结构;提出了监测时序数据同步的处理方法,应用噪声处理技术,给出了监测数据的清洗算法,给出了样本数据库的更新算法;设计了预测系统信息定制的两种方式:模型参数的定制和用户信息的定制。更多还原

【Abstract】 Taken the modern forecasting theory and technology as the foundation and adopted the policy of "safety first, precaution crucial", the dissertation has an in-depth analysis of the current situation of the safety management and control within the mine safety production system. The present research focuses on the mine production system, safety state forecasting theory and technology, giving some basic concepts such as mine safety production system; the subsystem; the region; the safety state; the safety degree; the security rating and so on. In addition, the mine safety state transition model—SAFU, is put forward, which can be adopted to forecast the safety state of the mine region, the combination region and the function subsystem. The forecast safety state types can determine the factors which cause the changes of the mine region safety states. And this will eventually provide the policy-making basis for the forecast and the safety control of the mine safety.The present dissertation has carried out a series analyses on the following aspects: the regression analysis forecasting method, BP neural network binding forecasting model method, the grey forecasting method and the support vector forecasting model. Meanwhile the dissertation also discusses the ways the forecasting model adopts in the prediction of the mine safety in production, presenting the means to apply the regression forecasting method unify under different parameter numbers. The binding forecasting model reduces the relative number of the neural network, enhances the iterative speed and the forecasting efficiency of the neural model and accomplishes the mine safety state forecasting with less classified sample vectors. With the SAFU model, it not only describes five changing patterns of the mine safety states but also proposes the algorithm for danger factors of the mine safety in production, which is determined by the reversal safety value. As to the defined mine region and combined region, the paper analyzes the relevance of the regional safety state, suggesting the algorithm for the recognition of the connected region, putting forward the concept of the regional safety degree and the algorithm for the safety degree on the basis of the connected region. In addition, the methods to forecast the safety degree of the single-parameter and multi-parameter regions are proposed as well, which proves the efficiency of the unity forecasting in the mine regional safety in production and the existence and quantization of the mine safety degree and draws the conclusion that the mine regional safety degree is predictable. What’s more, in the paper the author has set up the forecasting model based on the sample data drills and suggested the algorithm for the selection of the forecasting models. The longest match pattern presented in the paper, also known as the experienced model of the safety state, can be used to do the jobs of finding, description and prediction, which enables the regional safety state indescribable by the function to conduct the forecasting in model recognition. The paper also puts forward three basic combined patterns for the forecasting model, introducing the parameter projection so as to accomplish the combined construction of the forecasting model. The selection, combination as well as the model dispatching method of the mine safety in production region is also discussed in the paper. The author analyzes and designs the forecasting system of the mine production safety state for the analysis of the procedure, the prediction function and the data processing. The system is made further designation in its configuration which brings about the proposed processing method of monitoring sequential data synchronization. With the application of the noise processing technique, the present research proposes two algorithms: one for the cleaning of the monitoring data; the other is for the transaction of the sample data. Meanwhile, two modes for the information custom in the forecasting system have been postulated: the model parameter custom and the client information custom.更多还原