An approach to fault detection is described which uses neural-network
pattern classifiers trained using data from a rigorous differential-eq
uation-based simulation of a pilot-plant column. Two cases studies are
presented, both considering only plant data. For two classes of proce
ss data, a neural network and a K-Means classifier both produced excel
lent diagnoses. Extending the study to include three additional classe
s of plant operation, a neural network again gave accurate classificat
ions, while a K-Means classifier failed to correctly categorise the da
ta. Principal components analysis is used to visualise data clusters.
The robustness of the neural networks was found to be generally good.
Copyright (C) 1997 Elsevier Science Ltd.