【作者】 李江；

【导师】 林柏泉；

【作者基本信息】 中国矿业大学 ， 安全技术及工程， 2008， 博士

【摘要】 煤炭工业是国民经济的基础产业,同时也是特殊的高安全风险的行业,煤矿安全是煤炭工业健康、可持续发展的关键问题。因此研究如何有效地控制煤矿安全风险具有十分重要的理论和现实意义。本文综述了国内外煤矿安全评价理论与技术的研究现状,建立了煤矿安全动态评价和预测模型,并结合现场数据进行了应用研究。煤矿井下生产系统是一个由人-机-环境构成的、空间极其复杂的灾害系统。虽然事故发生的机理各异,但引发事故的因素却相互关联,在时间、空间上各种灾害随时随地发生,且相互影响。因此,根据煤矿井下灾害系统的结构特点,对系统的危险程度进行评价,事先获得事故的可能后果及对整个生产系统的影响,从而使煤矿的技术和管理部门有针对性地采取措施,达到安全生产的目的。寻求并建立科学、合理的矿井安全评价模型,并与实际生产相结合,是矿井安全管理与控制的关键问题。矿井安全系统中的许多问题都是非线性的,传统的、事先设定变化规律和特性的评价方法已经显现出其局限性,且难以很好地解决从因素到结果的定权和变权问题。本文在分析我国煤矿安全现状的基础上探讨了开展煤矿安全评价工作的重要意义,综述了国内外安全评价理论和煤矿安全评价技术的发展现状,分析了人工神经网络技术的特点,提出课题研究的意义、研究思路和主要内容。根据现代事故致因理论,结合人-机-环分析法与层次分析法等分析了煤矿安全生产的主要影响因素,并总结为10大类。根据安全评价模型的要求,在遵循指标构成及其定量化处理等原则的前提下,构建了全面的煤矿安全评价指标体系,使煤矿安全生产过程中的各个重要影响因素在指标体系中得到体现。根据神经网络结构特点和己经建立的煤矿安全评价指标体系,确定以误差反向传播的前向BP网络作为煤矿安全评价算法模型,并探讨了煤矿安全评价模型的网络结构设计、训练学习流程、性能改进方法。讨论了MATLAB神经网络工具箱及其图形用户界面GUI在神经网络模型的设计和训练过程中的强大功能,为煤矿安全评价网络模型的应用奠定基础。利用神经网络工具箱GUI实现了煤矿安全评价的神经网络模型设计,并结合大量的现场实际数据实现了煤矿安全动态评价的应用研究,评价结果与实际情况基本一致。预测就是依据历史寻求事物的未来发展趋势,是对事物未来发展趋势的认识,目的就是根据事件的发展与变化趋势来采取相应的措施。煤矿安全的有效控制对生产和作业人员的安危具有重要意义。有效的管理与控制,必须有完善、可靠的过程监测,而过程控制的成功与否,取决于对煤矿安全性指标的超前把握,准确的预测是超前把握并采取有效技术和管理措施的先决条件,煤矿安全预测就是通过系统现有或过去的危险信息来预测未来的系统安全状态。本文根据宏观与微观、静态与动态的辨证关系,确定了矿井安全预测的基本原则,建立了神经网络预测、灰色系统预测的数学模型。神经网络应用于安全预测的数学模型擅长于解决具有大量的历史数据的预测问题;而灰色预测GM(1,1)模型适应于历史数据不充分的预测问题,并在其基础上提出了函数变换型GM(l,1)模型,引入了UGM模型,有效的解决了GM(1,1)模型在短期预测中的不足,从而使预测结果更具客观性和预见性。更多还原

【Abstract】 As the basic industy of Chian’s national economy, the coal industy was also a special industy with high safety risk, so the coal safety was the key-point to the continuous development of the coal industry. It was vital in both theory and practice to study how to control effectively safety risk in coal mines. In this paper, the home and abroad theories and technologies of safety evaluation were summarized, the dynamic safety evaluation and safety prediction models of coalmine were established. The models which have been established were used to deal with the coalmines that data have been collected from.The production system underground in coal mine was a disaster system which consists of the human-machinery-environment and the extremely complex horizons. The mechanism of the disasters was different, but the factors initiating the disasters were interrelated each other. The disasters may take place at any time and every where, and affected from one to others. According to the structure features of the calamity in coal mine, we have to take the assessment to the hazard degree of the system, and previously acquire the effect of the possible results of the accidents to the whole production system, so that the technology and management administrators can adopt the measures, and the aim of the safety production will be obtained. The key problem to the safety management and control was to find and establish the scientific and reasonable safety evaluation models, and to combine the models with the practice production. Many problems in the safety system in coal mine were non-linear. The traditional and the previously function-setting evaluation methods have appear their localization, and the problems of the fixing and changing weight could not also be solved perfectly.In the paper, the chinese coalmine safety status was analyzed to put forward the importance and necessity to use safety evaluation. The home and abroad safety evaluation were summarized to analyze the problems of traditional coal mine safety evaluation methods, the characteristic of artificial neural network (ANN) were analyzed, the significances, ideals, methods and main contents of this subject were proposed.Based on the accidental incidence theory and other safety theories, the analytic hierarchy process were combined with other methods to analyze the primary factors affecting coal mine safety, classified into 10 types including human, mechanical, environmental factors. Based on the requirements of the safety evaluation model, under the precondition and the principles of the constitution and quantity of the indicators, the mine safety evaluation indicators system was completely constructed, and the every mportant factor during the production processes was incarnated in the indicator system.Based on the structure characteristic of artificial neural network (ANN) and the mine safety evaluation indicators system which has been established, error back-propagation algorithm (BP) was chosen to deal with the mine safety evaluation model. The designs of network structure, the training process and the methods to improve the performance of the model were discussed. The neural network (NN) tool box and the graphical user interfaces (GUI) of the MATLAB software were introduced in order to use the powerful function of them to deal with the designs and training of the mine safety evaluation model. The foundation to use the mine safety evaluation model was laid.The mine safety evaluation model was designed by the use of the neural network (NN) tool box, trained by the means of safety sample data to prove that the model was applicable for mine safety evaluation. The results calculated were similar to the actual situation.The prediction was to search the future developing trend according to the history, also was the recognition to the future developing tendency. The aim of the prediction was that the processing measures can be done according to the developing and changing trend. It was very important to control effectively the safety in coal-mine for the mine production and operators. The effective control and management relied on the perfected and reliable process supervision, and the safety process control was rested with the pre-holding for the safety indexes of the coal-mine, therefore, the exact prediction was the precondition to pre-hold and take the effective technology and management steps. The safety prediction for the coal-mine was to forecast the future safe statue based on the past or present dangerous information of the system. According to the differentiating and analyzing relations between the macroscopical and microcosmic status, static and dynamic features, in this paper, the basic principals was determined for the safety prediction in the mine, and the mathematical models based on the artificial neural network and the non-linear gray system theory were established. The artificial neural network model was suitable to data-rich, and the GM (1, 1) was suitable to data-poor. Based on GM (1, 1), functional transformation grey model (1, 1) and UGM model were proposed and used to solution the insufficient of GM (1, 1) in short-time prediction. The models make the safety predicting results more objectivity and foreseeing.更多还原