Soft-Sensing Modeling Based on Multi-Phases for Fermentation Process

For the process of Nosiheptied fermentation, a new method for soft-sensing modeling by phases is presented. By using the state equations established for the Nosiheptied fermentation process, the secondary variables are determined according to the inverse system theory. Then, fuzzy c-means clustering algorithm and neural network are used for phase identification, and for each phase, a local neural network model for soft sensing is developed. Finally, the estimation is implemented by computing the sum of outputs of the developed local models weighted by the corresponding degrees of membership from the phase identification. The testing result shows the effectiveness of the approach to the development of the soft-sensing model.