贝叶斯网络在起重机故障诊断中的应用

发布时间：2018-01-03 04:06

本文关键词：贝叶斯网络在起重机故障诊断中的应用　出处：《湖南大学》2011年硕士论文　论文类型：学位论文

【摘要】：近年来，针对功能日益强大和结构愈加复杂的大型起重机的故障诊断已成为一个重要的研究内容。大型起重机的部件以及部件之间存在很多关联耦合的相互关系，，不确定性因素和信息充斥其间，导致出现的故障一般都为不确定性故障。因此，解决不确定问题是目前大型起重机故障诊断的重要问题。本文在研究了起重机故障诊断和贝叶斯网络理论的基础上，采用了一种用于起重机故障诊断的贝叶斯网络诊断模型，并以某大型起重机已有的故障诊断系统为应用对象，将贝叶斯诊断模型应用于该系统中。改进后的系统在保留了原有的基于规则库的诊断方式基础上,融入了基于贝叶斯网络的诊断方式，使新系统可以进行多种形式的诊断推理，从而达到解决起重机不确定性问题的目的。应用于起重机故障诊断的贝叶斯网络建模分为三个步骤：第一是针对起重机故障领域的贝叶斯网络知识库的构建。运用因果调查问卷以及概率刻度等方法分别获取贝叶斯网络结构和贝叶斯网络参数，从而完成了对贝叶斯知识库的构建；第二是针对贝叶斯网络参数的学习。采用贝叶斯网络参数学习中的最大似然估计方法对起重机贝叶斯网络的参数进行学习修正；第三是模型的推理机制构建。采用贝叶斯精确推理算法中的基于簇树传播的算法作为本模型的推理算法，能完成对故障的量化推理。在实现方面，本文对系统中的贝叶斯网络诊断方式进行了总体设计。在设计中以模块为单位，对网络生成模块、推理模块、贝叶斯知识库管理模块、网络参数学习模块进行了设计，为改进后的整个起重机诊断系统的软件开发奠定了坚实的基础。通过分析比较原系统和采用贝叶斯网络诊断模型后的系统对同一故障的诊断结果，表明改进后的系统解决了原系统面对不确定性故障时无法快速定位故障原因的问题，是对原有系统的一种扩展和改进。从而验证了本文构建的故障诊断模型及算法等工作的有效性和应用价值。
[Abstract]:In recent years. Fault diagnosis for large cranes with increasingly powerful functions and more complex structures has become an important research content. There are many interrelated coupling relations between the components of large cranes and their components. The faults caused by uncertainty are usually uncertain faults. Therefore, solving the uncertain problem is an important problem in fault diagnosis of large cranes at present. Based on the study of crane fault diagnosis and Bayesian network theory, a Bayesian network diagnosis model for crane fault diagnosis is proposed in this paper. Taking the existing fault diagnosis system of a large crane as the application object, the Bayesian diagnosis model is applied to the system. The improved system retains the original diagnosis method based on rule base. The diagnosis method based on Bayesian network is integrated, so that the new system can carry out many kinds of diagnostic reasoning. The Bayesian network model used in crane fault diagnosis is divided into three steps:. The first is the construction of Bayesian network knowledge base in the field of crane fault. The Bayesian network structure and Bayesian network parameters are obtained by using causality questionnaire and probability scale. Thus, the construction of Bayesian knowledge base is completed. The second is the learning of Bayesian network parameters. The maximum likelihood estimation method of Bayesian network parameters learning is used to modify the parameters of crane Bayesian network. The third is the reasoning mechanism of the model. Using the cluster-tree propagation algorithm in the Bayesian exact reasoning algorithm as the reasoning algorithm of this model, the quantitative reasoning of fault can be completed. In this paper, the Bayesian network diagnosis mode in the system is designed. In the design, the network generation module, the reasoning module, the Bayesian knowledge base management module are taken as the unit. The network parameter learning module is designed, which lays a solid foundation for the software development of the improved crane diagnosis system. Through the analysis and comparison of the original system and the Bayesian network diagnosis model of the system for the same fault diagnosis results. It shows that the improved system solves the problem that the original system can not locate the fault quickly when it is confronted with uncertain fault. It is an extension and improvement to the original system, thus validating the validity and application value of the fault diagnosis model and algorithm constructed in this paper.
【学位授予单位】：湖南大学
【学位级别】：硕士
【学位授予年份】：2011
【分类号】：TH21;TP18;TH165.3

【参考文献】