RELIABLE ONLINE KINETICS IDENTIFICATION FOR A SEMIBATCH PROCESS BY PARALLEL USE OF SHORT-CUT TRANSFER OF INFORMATION RULE AND RECURSIVE ESTIMATORS

On-line concentration measurements from an isothermal semi-batch syste m, i.e. the liquidphase acetoacetylation of pyrrole with diketene, are used for identification of a kinetic model. An integrated expert syst em, containing coupled estimation procedures and statistical analysis tools as well as a kinetic model data bank, chooses the most suitable structure for the kinetic model and subsequently identifies model para meters. With the increased availability of data during the run, refine ments or structural changes in the kinetic model are performed automat ically and parameter estimates are updated. A recently developed short -cut procedure (the Modified Integral transformation Procedure, MIP, M aria and Rippin, 1995, 1996; Maria, 1995), is compared against classic al direct methods: Interactive use of a kinetic model-data bank throug h a numerical transfer of information strategy allows adaption of the mechanism-based kinetic model structure. Multiple nonlinear state esti mators of the Extended Kalman Filter (EKF) type are operated in parall el, while the MIP/kinetic data-bank provides information for switching between the EKFs and concerning mechanism and model structure. This o n-line use of parallel direct (short-cut) estimation procedures is par ticularly suitable for real-time process supervision, control and opti misation purposes.