[en] Data reconciliation and principal component analysis are tno recognised statistical methods used for plant monitoring and fault detection. We propose to combine them for increased efficiency. Data reconciliation is used in the first step of the determination of the projection matrix for principal component analysis (eigenvectors). principal component analysis can then be applied to raw process data for monitoring purpose. The combined use of these techniques aims at a better efficiency in fault detection. It relies mainly in a lower number of components to monitor. The method is applied to a modelled ammonia synthesis loop. (C) 2001 Elsevier Science Ltd. All rights reserved.
Cameron, D., Fault detection and diagnosis (1999) Model Based Manufacturing-Consolidated Review of Research and Application, , http://capenet.chemeng.ucl.ac.uk/
Heyen, G., Maréchal, E., Kalitventzeff, B., Sensitivity calculations and variance analysis in plant measurement reconciliation (1996) Computers and Chemical Engineering, 20 S, pp. 539-544
Kresta, J.V., MacGregor, J.F., Marlin, T.E., Multivariate statistical monitoring of process performance (1991) The Canadian Journal of Chemical Engineering, 69, pp. 35-47
(1999) Belsim, VALI III Users Guide, , BELSIM sa, Allée des Noisetiers 1, 4031 Angleur (Belgium)
Snedecor, G., (1956) Statistical methods (5th ed.), , The Iowa State College Press