故障诊断系统中基于多Agent的并行推理机制研究

基于诊断任务的多层次混合分解策略，提出了一种融合不同知识表达方式和推理机制的多Agent并行诊断结构，分析了管理Agent、诊断Agent和决策Agent的基本职能；并将证据理论应用于多诊断Agent的联合决策过程，指出了Dempster合成规则及Yager改进方法的不足，提出了基于证据支持度的冲突信息加权分配方法，并以一个算例证明了该方法的有效性。

Research on Multi- Agent Based Synchronous Inference Mechanism in Fault Diagnosis System

Based on muhi----layer decomposition strategy of diagnosis task, a multi agent synchronous diagnosis federation integrated different knowledge representation modes and inference mechanisms is presented. The functions of management agent, diagnosis agent and decision agent are analyzed. Evidence theory is adopted into the joint decision--making process of multl--agent system. The deficiencies of Dempster combination rule and Yager modification melhod are pointed out. A new weighted distribution method for conflict information based on evidence support degree is put forward. And a numerical example shows this new combination rule can combine conflict evidences with good results and convergence efficiency.