On-Line Estimation of Biomass Concentration Based on ANN and Fuzzy C-Means Clustering
A multiple-model modeling method has been proposed for soft-sensing in a complex non-linear biochemical process for years. In this study, a multi-fuzzy-neural network model (MFNN), which is combined by multiple-model modeling method based on neural network and fuzzy c-means clustering algorithm (FCM), is presented to estimate the biomass concentration in fermentation process. Low dimensional sample data is achieved through principal component analysis (PCA).FCM is used for the analysis of the distribution of principal data and grouping them into overlapping clusters with different membership degrees. Then, a soft-sensing model is developed using multi- fuzzy-neural network to fit the different hierarchic property of the process. The biomass concentration is estimated by computing the sum of outputs of local models weighed by the corresponding degrees of membership. The model is applied to an erythromycin fermentation process, and case studies show that the approach has better performance compared to the conventional global model.
Keywordssoft-sensing fuzzy c-mean clustering algorithm neural network fermentation
Unable to display preview. Download preview PDF.
- 1.Leigh, J.R.: Modeling and Control of Fermentation Processes. Peter Peregrinus Ltd (1987)Google Scholar
- 2.Bastin, G., Dochain, D.: On-line Estimation and Adaptive Control of Bioreactors. Elsevier science, Amsterdam (1990)Google Scholar
- 4.Bastin, G., Van Impe, J.F.: Nonlinear and Adaptive Control in Bioreactors. Eur. Journal of Control, 37–53 (April 1991)Google Scholar
- 9.Bakker, B., Heskes, T.: Model Clustering by Deterministic Annealing. In: Proc. ESANN, pp. 87–92. D-Facto Public, Bruges (1998)Google Scholar
- 15.Adilson, J., Rubens, M.: Soft Sensors Development for On-line Bioreactor State Estimation. Computers and Chemical Engineering 24(2), 1099–1103 (2000)Google Scholar