多Agent诊断求解方法研究

Multi-agent Diagnosis Methods

该文首先建造了一种包括多诊断Agent的诊断系统结构。为提高诊断A gent求解效率 ,对诊断Agent内部的故障模式进行聚类 ,将故障模式聚集成多个故障原型 ,使诊断只在特定空间上进行求解 ,避免搜索过程的盲目性和无关性 ;在此基础上 ,设计了一种基于故障原型的诊断Agent内部序贯诊断算法。另外考虑到诊断对象固有的关联性特征造成的多Agent间的诊断耦合 ,提出了以故障原型为依据建立Agent间关联模型的方法。基于故障原型的序贯诊断算法和Agent关联模型使Agent的内部诊断和Agent间的协作行为得以有效地联系。文中最后通过一个化工过程诊断实例对提出的多Agent诊断求解方法进行了验证。

The structure of multiagent diagnosis system is established. In order to improve agent's solution efficiency,the fault patterns of each diagnosis agent are clustered into several fault prototypes.A diagnosis procedure is carried out only in a specific space and avoids blind searching. Then the fault prototype based sequential diagnosis algorithm is adopted to perform agent's internal diagnosis.Given the diagnosis coupling in diagnosis process arising from the inherent relationship among different diagnosis objects, the relationship models for agents are brought forward to integrate agent's internal action and agent's external cooperation. In the end,an example of a chemical process diagnosis is presented that verifies the method's feasibility and efficiency.