@article{open954,
          volume = {82},
          number = {3},
          author = {P R Patnaik},
            note = {Copyright of this article belongs to  Canadian Society for Chemical Engineering.},
           title = {Neural and hybrid neural modelling and control of fed-batch fermentation for streptokinase: Comparative evaluation under nonideal conditions.},
       publisher = {wiley},
            year = {2004},
         journal = {Canadian Journal of Chemical Engineering},
           pages = {599--607},
        keywords = {streptokinase, fed-batch fermentation,
nonideal conditions, hybrid neural network.},
             url = {http://crdd.osdd.net/open/954/},
        abstract = {Fermentations involving competition between two or
more kinds of cells under nonideal conditions show
complex profi les that are sensitive to the extra-cellular
environment. These fermentations therefore require
accurate and rapid on-line data acquisition and control.
However, both on-line measurements and modelling
are diffi cult and expensive for large bioreactors, thus
limiting the usefulness of model-based control. While
neural networks offer an alternative, they require
extensive training and can be diffi cult to optimize
for large arrays. Hybrid networks combining
a few neural networks with some mathematical
equations offer a good compromise. The possibility of
using a hybrid model for simulation-cum-control has
been examined here for the fed-batch production of
streptokinase. Under noideal conditions, hybrid neural
models outperformed both mathematical models
and arrays of neural networks, thus suggesting their
viability for large-scale fermentation monitoring and
control.
Les fermentations provoquant une comp{\'e}tition entre
deux ou plusieurs sortes de cellules dans des conditions
non id{\'e}ales montrent des profi ls complexes qui
sont sensibles {\`a} l?environnement extra-cellulaires. Ces
fermentations n{\'e}cessitent donc une acquisition et un
contr{\^o}le en continu des donn{\'e}es qui soient pr{\'e}cis et
rapides. Toutefois, les mesures et la mod{\'e}lisation en
continu sont diffi ciles et co{\^u}teuses pour les grands
bior{\'e}acteurs, ce qui limite l?utilit{\'e} du contr{\^o}le bas{\'e} sur
des mod{\`e}les. Les r{\'e}seaux neuronaux sont une autre
possibilit{\'e}, mais ceux-ci n{\'e}cessitent un entra{\^i}nement
pouss{\'e} et peuvent {\^e}tre diffi ciles {\`a} optimiser pour de
grands dispositifs. Les r{\'e}seaux hybrides combinant
r{\'e}seaux neuronaux et {\'e}quations math{\'e}matiques
offrent un bon compromis. La possibilit{\'e} d?utiliser
un mod{\`e}le hybride pour la simulation et le contr{\^o}le
a {\'e}t{\'e} examin{\'e}e dans ce travail pour la production
{\`a} alimentation discontinue de streptokinase. Dans
des conditions non id{\'e}ales, les mod{\`e}les neuronaux
hybrides offrent une meilleure performance que les
mod{\`e}les math{\'e}matiques ou les dispositifs de r{\'e}seaux
neuronaux, et il pourrait donc s?av{\'e}rer viable pour la
surveillance et le contr{\^o}le de fermentation {\`a} grande
{\'e}chelle}
}