State estimation and error diagnosis in industrial fed-batch yeast fermentation



Measurements provide the basis for monitoring and control of industrial processes as well as model development and validation. Therefore, systematic approaches are of great value to increase accuracy and reliability of measurements. In bioprocesses, linear conservation relations such as elements and enthalpy can be employed to relate conversion rates. In this work, a systematic approach has been applied to production scale fed-batch yeast fermentations. The six data sets obtained from two industrial size bubble columns, one with 25 m3 volumes and the other with 100 m3 volumes, are analyzed for state estimation and error diagnosis. A statistical test is employed for error diagnosis. The serial elimination method is used to analyze and locate the source of errors. The conversion rates are calculated from primary measurements such as flow rates, temperatures, and concentrations. When available measurements are more than the degrees of freedom of the system, it is said that the system is redundant. The redundancy is, therefore, used for error detection and data reconciliation for the six data sets in this work. In addition to elemental balances, heat balance has been set up for the bubble columns, and metabolic heat production rate is employed as an additional measurement. The redundancy is employed for state estimation, and biomass concentration and specific growth rate have been estimated with great accuracy. The estimations can be further used for process monitoring and control. © 2006 American Institute of Chemical Engineers AIChE J, 2006