Application of ANN models in biochemical and fermentation industries concentrate mostly on soft sensor development for estimating infrequently measured quality variables such as biomass concentration using process variables that are frequently measured. Developing such empirical models is almost similar to developing statistical regression models. There are numerous applications for different types of fermentations in the literature. Applications include use of ANN models to estimate biomass concentration in continuous mycelial fermentation [28, 649], improve yield in penicillin fermentations , and develop on-line estimators for batch and cell-recycle systems . Applications in general for soft sensor technology [25, 75, 98], optimization and fault diagnosis [92, 345, 587] and experimental design based on ANNs  have also been reported. There are also hybrid neural network-first principles models that incorporate fundamental models with ANN models for better accuracy on predictions [170, 481, 588]. One approach is to use a first principle model to explain as much variation as possible in data. The remaining significant variation is modeled by an ANN. This hybrid structure is inspired from the first principle-time series model combinations that rely on the same philosophy. In another approach, the parameters of a first principles model are estimated by ANN in a hybrid structure . Another hybrid structure  creates series parallel k
Figure 4.19. A hypothetical feedforward ANN with one hidden layer for estimating substrate and biomass concentrations in fed-batch penicillin fermentation (OUR: Oxygen uptake rate, S: Substrate conc., X: Biomass conc., wij and weight vectors associated with interconnected layers).
Was this article helpful?