基于贝叶斯网络的复杂系统可靠分析方法研究与应用

发布时间：2018-01-05 15:28

本文关键词：基于贝叶斯网络的复杂系统可靠分析方法研究与应用　出处：《天津工业大学》2016年博士论文　论文类型：学位论文

【摘要】：随着现代工业技术的迅速发展,产品和设备系统日趋复杂化。系统的复杂性,一方面体现为其子系统或部件间相互藕合,另一方面也体现为系统的工作环境变化等外部影响因素的纷繁众多。并且,由于受到物质、空间和时间上的限制,很难获得足够的数据信息对系统的状态、特征和行为做出明确和精准的判据,这些因素导致系统包含着大量的不确定性。传统的可靠性分析方法在解决实际问题中暴露出明显的不足和局限。本文针对复杂系统可靠性分析中的主客观不确定性的问题,以贝叶斯网络理论作为不确定性分析的理论基础,结合模糊理论和传统的可靠性分析理论,以电池生产线作为研究对象,分析了目前可靠性分析理论在复杂系统可靠性分析中存在的问题和不足,提出了相应的解决方法,并建立了分析模型,其主要内容如下：(1)分析了传统故障树和贝叶斯网络的可靠性分析方法的局限性,提出了基于故障树的模糊贝叶斯网络的可靠性分析方法。该方法采用贝叶斯网络建模方法进行基本建模,用贝叶斯网络理论的节点多态表达特性来描述复杂系统的事件多态性,用贝叶斯网络理论的节点条件概率表来描述复杂系统的事件之间的不确定性逻辑关系。在贝叶斯网络模型的框架下,引入模糊集合理论,用模糊数来描述专家对事件概率的模糊评估。在对不确定权重的专家评估信息的集结过程中,提出了用依赖不确定性有序加权平均算子综合不确定权重的专家们的评估信息,来实现专家权重的客观定权。(2)分析了传统FMECA分析方法的局限性,提出了基于FMECA的模糊贝叶斯网络的可靠性分析方法。该方法采用了模糊理论中的模糊数来表示专家对RPN属性参数的模糊评级,提出了用带置信结构的模糊规则库替代传统的模糊规则库,并用来描述模糊输入数据不完备的条件下模糊规则的前提和结论之间的不确定性；提出了利用贝叶斯网络推理技术合成置信结构的模糊规则,实现模糊规则的推理,并给出了详细的建模方法和步骤；提出了利用加权平均去模糊方法,实现故障危害等级的清晰化、明确化。(3)分析了当前复杂系统的可靠性分析中存在的问题,研究了不同可靠性分析方法的结合应用。论文以电池生产线系统的可靠性分析为例,通过对在多种主客观不确定信息的条件下电池生产线系统的可靠性分析,来研究可靠性分析方法的结合应用。在电池生产线系统的可靠性分析实例中,提出了采用基于FMECA的模糊贝叶斯网络的可靠性分析方法确定系统的关键重要子系统,采用基于故障树的模糊贝叶斯网络的可靠性分析方法确定影响系统的主要部件及其故障模式,定性定量地实现系统的可靠性分析。
[Abstract]:With the rapid development of modern industrial technology, the product and equipment systems are becoming more and more complicated. The complexity of the system is reflected in the coupling between subsystems or components on the one hand. On the other hand, it is also reflected in the variety of external factors, such as the change of the working environment of the system. Moreover, due to the material, space and time constraints, it is difficult to obtain enough data and information to the state of the system. Characteristics and behaviors provide clear and precise criteria. These factors lead to a large number of uncertainties in the system. The traditional reliability analysis method has exposed obvious shortcomings and limitations in solving practical problems. This paper aims at subjective and objective uncertainty in reliability analysis of complex systems. About sex. Taking Bayesian network theory as the theoretical basis of uncertainty analysis, combined with fuzzy theory and traditional reliability analysis theory, the battery production line is taken as the research object. The problems and shortcomings of the reliability analysis theory in complex systems are analyzed, and the corresponding solutions are put forward, and the analysis model is established. The main contents are as follows: (1) the limitations of the traditional fault tree and Bayesian network reliability analysis methods are analyzed. The reliability analysis method of fuzzy Bayesian network based on fault tree is proposed, which is based on Bayesian network modeling method. The characteristics of node polymorphism in Bayesian network theory are used to describe the event polymorphism of complex systems. The uncertain logic relationship between events of complex systems is described by using the nodal conditional probability table of Bayesian network theory. In the framework of Bayesian network model, fuzzy set theory is introduced. Fuzzy numbers are used to describe experts' fuzzy evaluation of event probability. In this paper, the objective weight determination of expert weight is realized by synthesizing the evaluation information of uncertain weights by using the weighted average operator which depends on the uncertainty.) the limitation of traditional FMECA analysis method is analyzed. The reliability analysis method of fuzzy Bayesian network based on FMECA is proposed. The fuzzy number of fuzzy theory is used to express the fuzzy rating of RPN attribute parameters. In this paper, the fuzzy rule base with confidence structure is proposed to replace the traditional fuzzy rule base, and it is used to describe the uncertainty between the premise and conclusion of fuzzy rule under the condition of incomplete fuzzy input data. The fuzzy rules of confidence structure are synthesized by using Bayesian network reasoning technology, and the reasoning of fuzzy rules is realized, and the detailed modeling method and steps are given. In this paper, a weighted average de-fuzzy method is put forward to realize the clarity of fault damage grade, and the problem of reliability analysis of complex system is analyzed. This paper takes the reliability analysis of battery production line system as an example and analyzes the reliability of battery production line system under the condition of various subjective and objective uncertain information. In the example of reliability analysis of battery production line system. A reliability analysis method based on FMECA fuzzy Bayesian network is proposed to determine the key subsystems of the system. The reliability analysis method of fuzzy Bayesian network based on fault tree is used to determine the main components and fault modes of the system, and the reliability analysis of the system is realized qualitatively and quantitatively.
【学位授予单位】：天津工业大学
【学位级别】：博士
【学位授予年份】：2016
【分类号】：TP18;N945.17

【参考文献】