基于粗集—神经网络的矿井通风系统可靠性理论与方法研究

【作者】 王洪德；

【导师】 马云东；

【作者基本信息】 辽宁工程技术大学 ， 采矿工程， 2004， 博士

【摘要】 矿井通风系统可靠性理论与技术研究，主要是针对该领域存在的一些问题而提出的，其目的是提高矿井通风系统可靠性水平，降低通风系统的建设和维护成本，防止和减少灾害事故发生，保障矿井高产高效的实现。因此，进一步完善和改进现有矿井通风系统可靠性理论机制，更新技术和管理手段，是提高通风系统可靠性研究的合理性、实效性和推广应用性的关键；也是优化矿井通风设计、促进安全生产的一项重要任务。 本文针对矿井通风系统可靠性研究实际，在分析总结已有成果基础上，将网络流理论、人工神经网络技术和粗集理论与系统可靠性理论相结合，分别基于网络流理论和神经网络理论，建立了矿井通风系统的可靠性评价的解析模型和仿真模型；基于粗集-神经网络理论建立了通风系统可靠性预警仿真模型；同时，基于系统单元特性，进行了矿井通风系统可靠性设计，并对可靠性设计工作中的通风系统故障特征提取和可靠性分配方法进行了深入研究。主要研究工作如下： a．基于网络流理论的矿井通风系统可靠性研究 根据通风系统的自身特点，将其看成是由通风网络系统和主通风机系统两个子系统构成的可修系统。应用网络流理论，针对通风网络拓扑结构及通风构筑物特性，建立了功能型通风网络可靠性评价模型，并给出了求解网络可靠性指标的不交化最小路集算法；针对矿井主通风机子系统的可修特性，应用Markov过程理论，建立了基于冷储备可修理论的矿用主通风机子系统可靠性模型。 b．基于人工神经网络理论的矿井通风系统可靠性研究 矿井通风系统是一个多环节、动态、随机、时变的大系统，影响该系统可靠性运转的因素十分复杂和繁多，各影响因素之间往往又相互关联，具有很强的耦合特性。因此，作者在对影响通风系统可靠性的因素进行全面分析后，建立了基于BP神经网络的通风系统可靠性评价仿真模型；采用Weibull过程模型和自适应神经网络技术对通风系统的故障过程进行了分析，给出了基于影响因素属性的通风系统故障过程改善和劣化的定量描述；论证了通风系统故障过程的改善(劣化)博士学位论文荃于粗集一神经网络的通风系统可靠性评价理论与方法研究辽宁工程技术大学与故障强度、累积故障强度之间的关系;推导出了平均故障间隔时间和首次故障时间等特征参数模型。 c.基于粗集一神经网络的矿井通风系统可靠性预替研究 通过采用神经网络技术对矿井通风系统进行可靠性预警仿真时发现，由于该系统的特征参数过多，造成了网络规模过大、训练时间过长、以及系统规则库存在冗余等现象，导致了可靠性预警仿真模型实用性能的降低。为此，作者将粗集方法作为BP神经网络的前置系统，通过对矿井通风系统属性特征的提取和影响因素的约简，优化了神经网络中的输入节点个数，降低了神经网络结构的复杂性，从而形成了一个精确度更高、解算速度更快的基于粗集一神经网络的通风系统可靠性预普仿真模型。在此基础上，论文又提出了分层发掘基于粗集一神经网络的通风系统可靠性仿真的概念，使得用户可以根据对所研究系统的不同需要，在不同的层次对通风系统进行可靠性评价。 d.基于单元特性的矿井通风系统可靠性设计 在已知矿井通风系统可靠性综合指标基础上，建立了基于单元相对易损度、单元重要度和复杂度、以及单元相对故障频度的矿井通风系统可靠性分配模型，并用于主通风机子系统研究中。通过对主通风机风量变化规律及故障类型的统计分析，建立了基于效率可靠度的主通风机子系统可靠性设计模型，并运用该模型求解出矿用主通风机运行各阶段的可靠度指标值，及其对主通风机合理工况范围的影响。 以上解析模型和仿真模型的建立和特征参数的求取均编制了具体实现的计算机源程序，并在VC++6.0和Matlab6.0环境下运行通过。 本文基于网络流理论、粗集理论和神经网络技术所建立的矿井通风系统可靠性评价、可靠性预替及可靠性设计决策仿真系统，为矿井通风系统设计、管理和维护部门提供更加实用的决策支持工具，也为进一步提高矿井通风系统可靠性水平提供了新的理论依据。更多还原

【Abstract】 The research on reliability theory and technique in mine ventilation system is put forward, and it mainly aims at resolving the problems in this field. The purpose is to increase the reliability level of mine ventilation system, to decrease the building cost and maintenance cost of it, to prevent and reduce the happening of the disaster and accident in it, and to ensure the fulfilling high yield and high efficiency of mine. So it is the key to advance the rationality, actual effect quality, expanding quality and applying quality that more perfecting and improving reliability theory mechanism of mine ventilation system in existence, renovating the technical and administrant means. And so is one of the important tasks to optimize mine ventilation design and to accelerate safe production.In this text, the author aiming at the reliability research practice about mine ventilation system, and basing on the analysis and summary of the predecessor’s production, combined the theory of network stream, ANN technique and the rough set theory with system reliability theory, set up the analytic model and imitated model about the reliability evaluation of the mine ventilated system basing on the theory of network stream and ANN respectively, and set up the reliability early-warning model based on rough set and artificial neuron network theory. At same time, basing on the system cell’s speciality, the author designed the mine ventilation reliability, and investigated the method of the failures characteristic’s picking and the reliability distribute in reliability design deeply. Primary research work as follows:a. Mine ventilation system reliability research based on network stream theoryBasing on the characteristics of mine ventilated system oneself, it can be considered as a repairable system made up of two subsystem of the ventilated network system and main fanner system. Making use of the network stream theory, aiming at the topological structure of the mine ventilated network and the speciality of ventilated buildings, the author set up the ventilated network reliability evaluation model with functional style, and brought forward the minimizing path sets arithmetic to solve network reliability indexes; in allusion to the repairable specialities of the main fanner subsystem, applying Markov process theory, basing on the repairable theory of cold reserve, the author set up reliability model of the main fanner subsystem used in mine.b. Research on mine ventilation system reliability based on artificial neuron network theoryBecause of the mine ventilated system is a more taches, dynamical, random, and time-changing large system, the factors affecting the system reliability are very complicated and numerous, and sometimes the influence factors relations one another and taking on very strong coupling specialty. So after completely analyzing factors that influence the reliability of mine ventilated system, the author set up an evaluating imitated model about ventilation system reliability based on BP neuron network technique; The Weibull process model is adopted to analyze the process of the ventilated system’s failures. The fixed amount describe of the mine ventilated system failures’ improvement and worsen basing on the influence factors’ attributes is broughtup; Demonstrating the relations about the improvement or worsen with the failures intensity and the cumulated failures intensity during the ventilated system failures’ process; The characteristic parameter model about the mean time between failures and the first failures time was deduced.c. Research on ventilation system reliability early-warning based on rough set and artificial neuron network theoryForm the reliability early-warning imitating to mine ventilated system with artificial neuron network technique, we have found that the much more characteristic parameters make the ANN scale too large, train time too long, the redundancy of the system rule databases and so on, and it result in the practicability function of the imitated model lower. So the author p更多还原