The reliable operation of flow-injection analysis systems (FI systems)
demands a high degree of knowledge and experience. This a priori know
ledge is of great importance, especially for a fast detection and diag
nosis of faults. Transferring it into the form of a knowledge-based sy
stem and combining it with information received on-line from the recor
ded detector signal enables automatic operational supervision of FI sy
stems. This contribution presents a real-time knowledge-based system f
or the supervision of FI systems applied in on-line bioprocess monitor
ing. The special conditions of real-time systems are explained and the
basic structure of the knowledge base is illustrated. Examples of typ
ical faults of the FI system are given to explain how symbolic knowled
ge processing can be combined with numerical analysis of data to perfo
rm a fast and reliable fault detection and fault diagnosis.