089 Error detection and diagnosis: R.T.J.M. van der Heijden, B. Romein, C. Hellinga, J.J. Heijnen, K.Ch.A.M. Luyben, pp 135–140

• Publisher: Elsevier Science
• Year: 1993
• Tongue: English
• Weight: 115 KB
• Volume: 1
• Category: Article
• ISSN: 0967-0661
• DOI: 10.1016/0967-0661(93)91484-e

✦ Synopsis

Fermentation in a beer brewery has been modeled in order to estimate the main process variable, the substrate concentration, used by brewers to control the process. A 'Fuzzy-Aided Extended Kalman Filter' (FA-EKF) and a 'Fuzzy-Aided Neural Network' (FA-AAN) estimator have been implemented at a production-scale fermenter. These techniques are compared in the fight of their realtime applicability at the plant. Both are of comparable performance, but the FA-EKF has greater reliability when the process runs into states which have not been observed before, whereas the FA-AAN's development time, and the time needed to adapt the software system to new process variations, are considerably smaller.