engleski

Modelling and Adaptive Control of Bioreactors by Neural Networks

Computer modelling in biotechnology plays essential role in process engineering applications such as: analysis, design, scale up, monitoring, control, quality insurance and optimisation. Modelling of bioreactors is the most complex due to intricate dependence of physical, chemical and biological interactions. The classical representation of mass, energy and momentum balances in form of ordinary differential equations requires analytical knowledge of all relevant process interactions and parameters. Such models are the most complete, but are very expensive to develop. Applications of artificial intelligence methods, such as artificial neural networks and fuzzy logic modelling have become very popular in biotechnology engineering community. Fuzzy logic models incorporate human reasoning while neural networks mimic biological structures in living organisms. In this work are applied neural networks designed on principal curve method which integrates results from nonlinear chemometrics and nonlinear autoregression models with exoneous inputs. Presented are results from fed batch operation in industrial production of baker's yeast and continuous production of mandelic acid in membrane enzyme reactor with coenzyme regeneration.

enzyme reactor control; neural networks; fuzzy logic models

nije evidentirano