Faults diagnosis in industrial processes with a hybrid diagnostic system
Chapter in Scopus
- Additional Document Info
- View All
© Springer-Verlag Berlin Heidelberg 2002.In this paper we present an approach to detect and diagnose multiple faults in industrial processes with a hybrid multiagent diagnostic system. We integrate artificial intelligence model-based diagnosis with control systems Fault Detection and Isolation (FDI) techniques. We adapt a probabilistic logic framework to perform fault detection and diagnosis tasks. The whole diagnosis task is performed by agents and is executed in two phases. In first phase, the Alarm Processor (AP) agent processes the discrete observations and alarms, and outputs a set of most likely faulted components. In second phase, Fault Detection (FD) agents discard the components not participating in the failure, by analyzing a set of continuous signals, that have a different behavior in normal and in faulty state. The FD agents include dynamic probabilistic models, able to deal with noise, nonlinear behavior and missing data. The output of the diagnostic system includes the components with abnormal behavior and the type of faults. We have tested our approach by diagnosing faults in a simulated electrical power network.