Active Diagnosis by Self-Organization: An Approach by the Immune Network Metaphor

Abstract

We propose a concept of active diagnosis that differs from the conventional passive (i.e. event-driven) diagnosis in temporal (diagnosis is carried out by always monitoring normal condition as opposed to identifying faulty only when abnormal condition is detected) sense as well as spatial (diagnosis is carried out by agents distributed in the sensor network) sense. As one way of realizing active diagnosis, we present immunity-based agents approach based on the self creating, monitoring, and maintaining feature of immune systems. We apply the approach to process diagnosis where the agents are defined on the sensor network. Each agent corresponding to sensor or process constraint evaluates a kind of reliability by communicating other agents. System level recognition of sensor/process fault can be attained by continuously and mutually monitoring and maintaining consistency among sensor values and process constraints. 1