基于多色集合的贝叶斯诊断网络构造方法研究

Construction of Bayesian Diagnostic Network Based on Polychromatic Sets

为了解决贝叶斯诊断网络知识表达不一致的问题，构建了故障模式和物理组件之间的映射框架，并提出了一种基于结构分解的层级式多重故障结构模型．引入多色集合理论，对这种结构模型的底层框架进行形式化描述，使得模型便于计算机表达和操作．通过对多色集合的推理矩阵进行迭代搜索实现贝叶斯诊断网络（BDN）自动生成，并给出了具体的推理步骤．以汽轮机设备为例对提出的方法进行了有效性验证，结果表明该方法有助于BDN知识表达的规范化和构造过程自动化程度的提高．

To solve the consistency problem of knowledge representation of the Bayesian diagnostic network,the mapping framework between the failure modes and the physical components are established,and a kind of hierarchical structural model of the multiple failures is proposed combined with the structural decomposition of the complex system.Following the polychromatic sets theory,the rock-bottom frame of this model is described formally,which facilitates constructing the computational expression and leading to a more convenient operation.Bayesian diagnostic network(BDN) is generated automatically via the iterative search process on the reasoning matrices of the polychromatic sets,and the specific procedure of the reasoning is presented.A case study for a steam turbine is reported to illustrate the validity of the proposed method,which is helpful to the standardization of the DBN knowledge representation and promotes the automation of the BDN construction process.