Monitoring estimation and predictive control based on statistical techniques

In this paper the principal component analysis (PCA) is used to monitor and diagnose via the online measurements any fault that might occur in the fed batch fermentation process. Predicting the unmeasurable state variables is achieved using the Karhunen-Loeve (K-L) approach. The possibility of using a predictive control technique to control the states around a desired trajectory by controlling the amount of nitrogen fed inside the fermenter is described and illustrated.