文摘
A process data driven procedure has been developed that allows auniversal time-efficient bioprocess analysis. The procedure is particularly suitedfor industrialproduction processes which have not yet been comprehensivelyinvestigated. It makesuse of artificial neural networks in combination with mass balanceequations torepresent the process dynamics on a commercial workstation. Theessential conceptbehind the procedure is to start with the already available knowledgeformulated bya very simple process representation which includes only thosevariables that are firmlyknown to be essential. Then, stepwise, additional variables areadded to the basicrepresentation after they passed a test procedure in which they provedto enhancethe model's performance. The result of the procedure is anumerical representationof the important process relationships that immediately allows todetermine improvedset points and/or profiles for the manipulated variables with respectto processperformance. It may be used to improve state estimation andcontrol. The procedurehas already been tested in industrial applications. In this paper,a validation of theprocedure with simulated bioprocess data is presented.