An efficient model development strategy for bioprocesses based on neural networks in macroscopic balances

Abstract

In the serial gray box modeling strategy, generally available knowledge, represented in the macroscopic balance, is combined naturally with neural networks, which are powerful and convenient tools to model the inaccurately known terms in the macroscopic balance. This article shows, for a typical biochemical conversion, that in the serial gray box modeling strategy the identification data only have to cover the input‐output space of the inaccurately known term in the macroscopic balances and that the accurately known terms can be used to achieve reliable extrapolation. The strategy is demonstrated successfully on the modeling of the enzymatic (repeated) batch conversion of penicillin G, for which real‐time results are presented. Compared with a more data‐driven black box strategy, the serial gray box strategy leads to models with reliable extrapolation properties, so that with the same number of identification experiments the model can be applied to a much wider range of different conditions. Compared to a more knowledge‐driven white box strategy, the serial gray box model structure is only based on readily available or easily obtainable knowledge, so that the development time of serial gray box models still may be short in a situation where there is no detailed knowledge of the system available. © 1997 John Wiley & Sons, Inc. Biotechnol Bioeng 53: 549–566, 1997.

Design Considerations in Hybrid Neural Optimization of Fed-Batch Fermentation for PHB Production by Ralstonia eutropha

“…The optimized performance based on neural models is better than that achieved through model-based control. However, neural networks have poor physiological fidelity and extrapolation capability, they require carefully screened data, and they may require considerable amount of training for complex systems (Hassoun 1995;Haykin 1998;van Can et al 1997). These weaknesses may be reduced by using combinations of neural networks and mathematical models.…”

“…While native PHB is rather brittle and has a melting point (180°C) close to its thermal degradation temperature (200°C), these properties improve significantly on copolymerization with other polymers, notably polyhydroxyvalerate (van der Walle et al 2001). Then, PHB has many properties comparable to those of large-scale chemically synthesized polyethylene (PE) and polypropylene (PP ;Doi 1990;Steinbuchel 1996).…”

“…Designs of hybrid neural models therefore require explorations of plausible configurations. While most designs allow some flexibility in the decisions referred to above (Galvanauskas et al 2004;Patnaik 2004;van Can et al 1997), this flexibility itself sometimes results in two or more comparably good models. Then, the choice of the most suitable hybrid model depends on the objective.…”

“…Then, the choice of the most suitable hybrid model depends on the objective. Nevertheless, hybrid models in general elicit better performances from bioreactors than methods based on mathematical models alone or purely neural descriptions (Galvanauskas et al 2004;James et al 2002;van Can et al 1997). The present study therefore explores the effectiveness of some hybrid designs for PHB fermentation optimization in a non-ideal bioreactor.…”

Hybrid filtering of feed stream noise from oscillating yeast cultures by combined Kalman and neural network configurations

“…Applications to a number of monotonic fermentations [3,19,20] have shown their superiority over their components in stand-alone mode. A recent application [13] has shown that a hybrid filter can also outperform algorithmic filters for oscillating cultures of S. cerevisiae.…”