Hybrid Fault Diagnosis System For Large Chemical Plants Under Control

D. ZUMOFFEN; M. BASUALDO

Lyon, France.

Congreso; ESCAPE 18; 2008

In this paper, a novel monitoring system, able to support fault tolerant control, is presented. It is developed by connecting different research areas such as statistical monitoring, knowledge-based and history-based systems. Several tools such as Adaptive Principal Component Analysis (APCA), Fuzzy-Logic (FL) and Artificial Neural Network (ANN) are integrated to develop a fault detection, isolation and estimation (FDIE) system for specific faulty process elements. In this work faults such as extra dead time in actuators and bias in sensors which cause saturation are analyzed. The FDIE implementation is done on a benchmark of Waste Water Treatment Plant (WWTP). This is specially chosen in order to give a complete set of simulation results for a wide range of different abnormal events. Specific performance indexes for analyzing the effectiveness of the methodology proposed here are accounted.