Abstract: This paper proposes a multi-layered immune diagnosis model for network fault diagnosis that is inspired by the multi-layered defense mechanism and incorporates the co-stimulation in the nature immune system.The model is comprised of inherent diagnosis layer，fault propagates recognition layer and adaptive diagnosis layer.Inherent diagnosis layer inherits available knowledge that a fault could not cause influence to other nodes.Fault propagates recognition layer adopt the structure of the B-lymphocyte network based on the fault propagation network for the fault localization.Adaptive diagnosis layer learn the unknown pattern.The method can detect not only known fault，but also unknown fault.The system could find the best combination of test points in network by using the fault propagates model and the results of fault localization algorithm to diagnose all the faults.

Key words: artificial immune system, fault diagnosis, fault propagates model, inherent diagnosis, adaptive diagnosis

摘要： 借鉴生物免疫系统的分层防御机理以及层次间的相互刺激作用，提出了用于网络故障诊断的多层免疫诊断模型。模型采用三层结构，包括固有诊断层、故障传播识别层和适应性诊断层。固有诊断层考虑故障在发生概率上相互独立的已有知识，故障传播识别层采用B细胞免疫网络作为故障定位的故障传播模型，适应性诊断层学习和概括未知故障中发现的模式。多层故障诊断方法既能检测出已知故障，又能检测出未知故障。采用网络故障传播模型及定位算法，能找出最优的测点组合诊断所有的故障，可以减少需测测点的数量。

关键词: 人工免疫系统, 故障诊断, 故障传播模型, 固有诊断, 适应性诊断