基于大数据分析矿山灾害孕育、演化及突发的全过程，是感知矿山物联网服务矿山的紧迫任务。将大数据分析处理的最新技术“深度学习”引入到矿山灾害感知数据处理领域，探讨多源海量前兆信息的大数据分析技术，对矿山安全生产具有重要意义，也是感知矿山物联网亟待前瞻性研究的课题。.针对矿山灾害感知数据的特点和处理的实际要求，在矿山灾害认知领域知识的引导下，融合深度学习与支持向量机模型，研究顾及时、空相关性的矿山灾害感知多维数据分析的深度学习模型，研究基于主题增强的可解释性深度学习模型，研究深度学习、支持向量机的深度融合模型。在此基础上，综合分析处理时空分布的多源、多维灾害感知序列数据，提取灾害演变过程的深层次特征，研究特征集的优选模型，研究矿山灾害演变过程的预测模型。最后，分析矿山灾害的时、空演变规律，实现对灾害演变过程的智能识别与自动、半自动检索，开发矿山灾害实时监测与预警系统，直接服务于矿山安全生产。

Based on big data, analysing the whole process of disaster gestation,evolution and occurrence is an urgent task for perception mine Internet of Things to serve the mine production.Introducing the latest technology of big data analysis and processing "Deep Learning" into the field of mine disaster perception data processing, discussing the big data analysis technology for multi-source and massive precursory information, is of great significance to safety production in mines, and is also an urgent and prospective research topic of perception mine Internet of Things..According to the characteristics of mine disaster perception data and the practical requirements of its processing, guided by the knowledge in the field of mine disaster perception, this project will study the fusion theory and technology of the Deep Learning and Support Vector Machine model，research the Deep Learning model for multi-dimensional data analysis of mine disaster perception considering the spatio-temporal correlation, research the interpretable Deep Learning model based on thematic enhancement, research the deep fusion model of Deep Learning and Support Vector Machine. On all above basis, the project will analyze and process the multi-source and multi-dimensional disaster perception sequence data with spatio-temporal distribution, extract the deep-seated features of the disaster evolutionary process, research the optimization model of the features, and research the prediction model of mine disaster evolutionary process. Finally, the project will analyze the spatio-temporal evolution law of mine disasters, realize intelligent identification and automatic and semi-automatic retrieval of the evolutionary process of disasters, develop real-time monitoring and Early warning system of mine disasters, and directly serve mine safety.