Intelligent software sensors for fed-batch fermentation processes

Software sensors have attracted great research interests due to the problem of lacking suitable and robust online sensors for key fermentation variables in fed-batch fermentation processes. In this paper, intelligent software sensors have been developed based on multivariate statistical process control methods. The software sensors not only provide real time estimation of key variables but also have the facilities of self-diagnosis, self-validation and self-calibration using available lab assay data. An application of the software sensors to a fed-batch penicillin fermentation process is presented, and significant improvements over ordinary methods have been shown in the simulation results.