Neuro‐expert approach to characteristics recognition for process monitoring and performance evaluation

Abstract

A process behavior feature extraction and recognition method based on neuro-expert approach is proposed for process monitoring and control. This approach provides an effective way to the analysis and automatic extraction of different types of process behavior features during operation without precise description of the process model. By combining use of symbolic manipulation, fuzzy membership function, rule-based reasoning and neural computation, the characteristic patterns which can capture system behavior features are deduced from on-line data and subsequently used for process monitoring and control. A multi-mode information extraction and representation mechanism is provided to meet the needs of different tasks of system supervision. This leads a more general and effective way to make full advantage of on-line information. The basic methodology and a general application of this approach to process monitoring and performance evaluation are discussed in this article. © 1997 John Wiley & Sons, Inc.