Staged Soft-Sensor Modeling for Batch Fermentation Process

Crucial biochemical parameters are hard to be measured on-line in real-time in batch fermentation process, which brings difficulties to control and optimization of this process. To solve this problem, soft-sensor technique is applied to implement the on-line estimation of crucial biochemical parameters, and a staged modeling method is presented. Firstly, the modeling data are classified according to their stages by using fuzzy c-means clustering algorithm. Then, a model for the on-line identification of fermentation stages and some local soft-sensor models corresponding to each stage are developed by using neural network. Lastly, simulation is performed based on the production data from Nosiheptide batch fermentation process. The results show the effectiveness of the presented soft-sensor modeling method.