【作者】 朱世松；

【导师】 汪云甲；

【作者基本信息】 中国矿业大学 ， 地图制图学与地理信息工程， 2013， 博士

【摘要】 瓦斯灾害防治仍然是我国煤矿安全工作的重中之重。综合利用布设在井下空间的各类非接触式传感设备动态采集的相关数据，对具有突出危险的工作面实现实时跟踪监测和早期诊断预警，为煤矿及时采取针对性措施，提高监控系统可靠性，防范和抑制瓦斯突出、瓦斯积聚和瓦斯爆炸等事故提供决策依据，是目前煤矿瓦斯安全监测系统亟待增强的功能目标。论文根据煤与瓦斯突出、瓦斯爆炸等危险性预测技术和预警理论，采用多传感器信息融合方法，充分挖掘瓦斯、风速、电磁辐射、声发射等各类传感数据所蕴涵的规律性知识，发挥各类传感器的优势，按照“手段多样、优势互补、相互验证、短中长期搭配”的思路，着力构建基于多传感器信息融合的瓦斯安全监测预警系统，实现对井下工作面瓦斯危险的“实时感知、准确辨识、快速响应、有效控制”。论文取得的主要研究成果如下：全面总结分析了国内外煤矿瓦斯安全动态监测手段和突出危险性评价指标的研究成果，包括瓦斯浓度、电磁辐射、声发射等传感监测技术及突出预测方法，为发挥各自优势，实现煤与瓦斯突出多传感器融合预警奠定了坚实的理论基础。通过现场调研，分析了煤矿瓦斯灾害防治实际需求，本着提高监测系统效能，降低系统资源消耗的理念，提出了瓦斯监测多传感器信息融合的目标体系、闭环工作流程、传感器选用与组织以及各种瓦斯安全动态监测传感信息融合分析理论的合理运用，从而最终确定了瓦斯监测多传感器信息融合体系总体结构，重点研究了基于模糊专家系统的瓦斯突出预测多传感器信息决策融合方法。提出了基于时间序列相似性度量的瓦斯超限报警信号辨识方法。基于DTW距离对煤矿采掘工作面瓦斯超限报警时间序列进行了聚类分析。对所获得的7种典型的时间序列模式，基于分段形态度量方法，提取了15个特征指标，从中筛选出5个分类效能较强指标，建立了瓦斯超限报警时间序列形态特征表。在此基础上提出了一种瓦斯报警信号快速辨识算法。提出了基于时空相关分析的煤矿采掘工作面瓦斯监测数据异常自动识别技术。定性分析了工作面顺风流方向瓦斯运移存在的时空异步相关特性；确定了相关系数计算过程中涉及的异步相关最优滞后步长的计算方法和瓦斯气体涌出后在回风巷道中体积分数随时空变化的预测和反演公式；统计计算了8种原因导致的瓦斯数据异常存在的相关系数值变化区间；提出了基于时空相关分析的工作面瓦斯监测数据异常识别算法；为提高相关分析效率，提出了能表达空间拓扑信息的井下瓦斯传感器层次编码方法。提出将工作面瓦斯安全监测问题归类为专家诊断范畴。研究了瓦斯监测信息知识发现方法，提出了瓦斯时间序列聚类分析与知识提取方法；针对瓦斯监测多传感器信息决策融合专家知识库系统的设计需求，提出了瓦斯监测知识学习算法和瓦斯监测专家知识的组织存储策略。最后举例说明了基于专家系统的工作面瓦斯超限原因识别推理应用过程。更多还原

【Abstract】 In coal mine area, the most crucial safety issue is still gas disaster prevention and control.Based on coal-gas outburst and gas explosion prediction technology and early warning theory, this thesis aimed at using MSIF method to mine regular knowledge from data about gas concentration, wind speed, electromagnetic radiation, and acoustic emission, et al. In addition to that, following the principle of diverse methodologies, complementary advantages, cross validation and short-long term integration, this thesis also has built a coal-gas safety monitoring and early-warning system based on fusion of multi-sensors’data. This system could accurately monitor the mining face’s coal-gas safety in real time and rapidly provide effective control measures. The main research achievements of this dissertation are as follows:A solid foundation in understanding theory of coal-gas outburst warning by using multi-sensors’data was established based on literature review of coal-gas safety dynamic monitoring methods and outburst risk evaluation indexes which include gas concentration, electromagnetic radiation and acoustic emission, et al.After field investigation and analysis of the system actual demands, the final structure of coal-gas monitoring sensors’data fusion system was proposed based on the consideration of enhancing efficiency of the monitoring system and reducing its resources consumption, and the research focused on coal-gas outburst prediction fuzzy expert system which makes decision based on fusion of multi-sensors’data.Gas monitoring warning signal identification method based on time series similarity measure is presented. The representative gas warning time series (GWTS) patterns are gained by clustering analysis method based on DTW distance. With the piecewise morphological measure methods, the character value table is established.Based on spatio-temporal correlation analysis method, the automatic identification techniques for coal-gas monitoring data anomaly are presented.The coal-gas safety monitoring problem was classified as an expert diagnosis issue here. The methods of knowledge discovery and expert knowledge base system design based on cognitive model are provided.更多还原