摘要： 提出了基于正反向混合推理的故障诊断方法，运用基于动态模糊综合评判的正向推理生成候选故障集，作为反向推理的基础。基于诊断任务的多层次混合分解策略，提出了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，给出了并行推理过程中的冲突消解算法。

关键词: 多Agent；混合推理；故障诊断；冲突消解

Abstract: A new fault diagnosis method based on hybrid inference is put forward. The forward inference process of dynamic fuzzy comprehensive evaluation provides a candidate fault set as the premise for the backward inference process. Based on multi-layer decomposition strategy of diagnosis task, a multi-agent synchronous diagnosis federation integrated different knowledge expression modes and inference mechanisms are presented. The functions of management agent, diagnosis agent and decision agent are analyzed. And the corresponding conflict resolution algorithm is given.

Key words: Multi-agent; Hybrid inference; Fault diagnosis; Conflict resolution