【作者】 王肖霞；

【导师】 杨风暴；

【作者基本信息】 中北大学 ， 信号与信息处理， 2014， 博士

【摘要】 尾矿坝作为整个矿山企业最大的人造危险源，其安全性对矿区人民的生命、财产以及社会的安全稳定造成了严重影响，引起了社会各界的高度重视。但由于外界环境的干扰性、传感器的局限性、监测手段的不完善性以及坝体结构的动态变化性等因素的影响，使得监测信息和风险模态信息常常是不完备的，且具有随机、模糊、不完整等不确定性，进而造成了两者间映射关系的复杂多变性，严重制约了监测系统的评估功效。因此，如何对坝体的监测信息和风险模态信息进行不确定性分析，建立两者间的复杂多变性映射关系是解决评估效果差、评估难问题的关键。可能性理论作为一种新的不确定性处理方法，其描述的是事物未来发生的一种可能性，不需要先验知识，且具有不确定性表征的多样性、不确定性处理的并存性、映射关系的多变性及适合小样本信息的处理等优点，这与尾矿坝安全监测信息处理的需求一致。因此，本文以上游式细粒尾矿坝为研究对象，以可能性理论为手段，并结合其他不确定性处理方法来解决坝体的风险评估问题。本文的主要研究内容包括以下几个方面：（1）论述了尾矿坝安全监测信息的不确定性及其处理方法。现有的不确定性信息处理方法一般由不确定性的种类所决定，对于不同的不确定性来说，处理方法各有差异。找到一种适合处理各类不确定性的方法是多源信息处理领域中的一大突破。本文在阐述不确定性信息来源及分类的基础上，深入分析影响尾矿坝监测信息不确定性的因素及不确定性的分类；研究各类不确定性的表征方法，解决现有不确定性信息处理方法在处理多种不确定性时存在的局限，并以此为研究基础，展开后面的工作。（2）构造了监测指标与风险模态信息的可能性分布。根据库水位、浸润线、干滩长度及坝体位移等监测指标的特点，利用小样本、可能性中值等研究各指标信息的可能性分布构造方法；对坝体的风险模态进行界定，利用流变-突变理论研究尾矿坝的失效机理，并结合坝体本身的固有频率，对风险模态进行表征；研究相邻模态边界的模糊性，为并存风险模态的表征提供解决方法。（3）建立了监测指标与风险模态间的复杂多变性映射关系。利用截集将监测指标和风险模态信息的分布转化为落影集合，通过研究落影集合间的模糊推理揭示两者间存在的一对多、多对一、多对多的复杂映射规律，建立两者间的复杂多变性映射关系；对不同截集下指标与风险模态间的可能性集值映射进行仿真。结果表明，分布间截集的相互变化可为监测指标与风险模态信息间复杂多变性映射关系的建立提供理论依据。（4）提出了基于分布匹配度集中模态的多属性风险评估模型。以指标与风险模态信息间的映射为评估指标，将坝体的风险评估作为模态等级的选取问题，以解决现有评估方法难以对模态等级较集中的情况进行有效判断的问题。对评估指标与各模态等级间的关系进行研究，确定评估指标与模态等级间的匹配程度，建立匹配矩阵；利用综合赋权法确定各评估指标的权值，对匹配矩阵进行加权融合，获得各模态等级的匹配度，并进行排序选出最优匹配等级；通过仿真验证所提方法的有效性。（5）提出了基于柔性相似测度并存模态的信息融合风险评估模型。将分布间相似测度和证据理论相结合对坝体进行融合评估，以解决模态等级并存的问题。在广义模糊数的基础上，定义了广义非线性模糊数，给出了柔性相似测度的计算方法，并对其基本性质进行了证明，与其它相似测度进行比较验证其有效性和合理性；利用诱导函数生成基本信任分配函数进行融合评估；通过仿真验证所提方法的有效性。（6）以火谷都尾矿坝和黄梅山尾矿坝为例验证所提模型的有效性。分析两个尾矿坝的基本概况以及溃坝事故发生时的征兆、损失等，深入分析溃坝发生前1个月内的监测信息和坝体的风险模态信息，对其进行表征；建立两者间可能存在的复杂多变性映射关系，利用提出的风险评估模型对坝体的模态等级进行反演。结果证明两种风险评估模型都可以对坝体的模态等级做出准确判断。更多还原

【Abstract】 The safety of tailings dam, as a largest artificial hazard in entire mining enterprises, has agrave impact on people’s life and property as well as social security and stability, which drawsgreat attention of the society from all walks of life. But owing to strong interference ofenvironment influence, limitations of monitoring sensor itself, imperfection of monitoringmethod and dynamic of dam structure, monitoring information and risk modal information areoften imperfect, and these information have random, fuzzy and incomplete uncertainty,leading to complex and changeable mapping relation of monitoring information and riskstates information, which severely restricts assessment effect of monitoring systems.Therefore, the ways of uncertainty analysis for monitoring information and risk statesinformation, and establishment of the mapping relationship are the key role to solve lesseffective and difficult risk assessment.Possibility theory is a new mathematical tool to deal with uncertainty and can be used todescribe the possibility of future events without any prior knowledge. It has the advantagesthat diversity of uncertainty characterization, coexistence of uncertainty processing, variety ofmapping relationship and processing for small sample information, which have common withinformation processing needs of tailings dam. So taken upstream fine grained tailings dam asobjects and possibility theory combined with other uncertainty processing methods as themeans, risk assessment problem of tailings dam is solved.The main works are introduced as follows:(i)The uncertainty of information and its processing method are discussed in safetymonitoring of tailings dam. Existing uncertainty information processing method are generallydetermined by uncertainty types, and the processing method is different as the change ofuncertainty types. So finding appropriate method to deal with all kinds of uncertainty is abreakthrough in multiple source information processing field. Based on expounding sourcesand classification of uncertainty information, this paper analyzes influential factors and typesof uncertainty for tailings dam monitoring information deeply, and researches characterization of all kinds of uncertainty to solve the limitations of existing methods in processing variousuncertainties. Based upon the above research, other works are developed as follows.(ii)Possibility distributions of monitoring indexes and risk modal information areconstructed. According to the characteristics of monitoring indexes such as reservoir waterlevel, phreatic line, length of dry, beach dam displacement and video image, small samplesconstruction and possibility median are employed to construct possibility distribution of eachindex. Risk modal of tailings dam is defined. Failure mechanism of tailings dam are studiedbased on rheology and mutation theory. Combined with inherent frequency of dam, riskmodal is characterized and the boundary fuzziness of adjacent modal is researched, whichprovides a new method to solve coexist risk modal.(iii)The complex and changeable mapping relationship between information ofmonitoring indexes and risk modal are set up. Possibility distributions are transformed intoshadow sets by cut-set, then the complex mapping law is reveal by studying fuzzy inferencebetween shadow sets and the complex and changeable mapping relationship is established.Possibility set-valued mappings are simulated under different cut-sets. The results show thatthe mutual changes of cut-sets between distributions can provide theoretical basis to establishthe complex and changeable mapping relationship between information of monitoring indexesand risk modal.(iv)A multi-attribute risk evaluation model with concentrated modal is proposed based onmatching degree of distributions. Taken the mapping between index and risk modalinformation as evaluation index, risk assessment problem is regarded as risk modal selectionin order to solve the problem that existing methods can not judge risk modal effectively whenthe modals are concentrate.The matching degrees of evaluation index and risk modal aredetermined and matching matrix is builded by researching the relationship between evaluationindex and risk modal class.The weights of each evaluation index are obtained bycomprehensive weight method then matching matrix is fused to obtain matching degrees ofeach modal. The optimal matching modal is selected. The effectiveness of the proposedmethod is validated by simulation.(v)Information fusion risk assessment model with coexistence risk modal is proposed byflexible similarity measure. Flexible similarity measure between the distributions andevidence theory are used in risk assessment to solve evaluation problem of coexist risk modaland improve the effect of evaluation precision. On the basis of generalized fuzzy numbers, the generalized nonlinear fuzzy number is defined, and the calculation method of flexiblesimilarity measure is given. Its basic properties are proved, and its effectiveness andrationality is verified compared with other similar measures. The basic belief distributionfunction of evaluation index is generated by inducing function,and then the assessment resultis obtained by D-S combination rules. The effectiveness of the proposed method is validatedby simulation.(vi)The effectiveness of the proposed methods have been validated by Huo Gudu andHuang Meishan tailings dam as two examples. The basic situation of the tailings dams areanalyzed, and the symptoms and loss in dam collapse are introduced. The information ischaracterized by analyzing the monitoring information and dam risk modal information within amonth before dam collapse.The complex mapping relationship between the both is established,and the risk modal of the dams are inverted by using the proposed risk assessment methods.The results show that two kinds of risk assessment methods can both make accurate judgmenton dam risk modal.更多还原