ABSTRACT
This chapter demonstrates that advanced modeling techniques, based on a hybrid neural approach, may be exploited to efficiently predict the behavior of two different processes regarding either food transformation or food-waste valorization. In the first of the two case studies, a hybrid neural model (HNM) based on the combination of a theoretical model and two artificial neural networks was developed so as to predict the behavior of the convective drying of vegetables. In particular, the HNM was aimed at determining the effect of operating conditions on both color degradation—chosen as a reference quality parameter—and microbial population abatement occurring during potato drying. A transport model, accounting for the simultaneous transfer of momentum, heat, and mass occurring in the drying air and in the food sample, was formulated and coupled to two different artificial neural network (ANN) models describing, respectively, the microbial inactivation kinetics of Listeria monocytogenes and the kinetics of color changes.
In the second of the considered case studies, an HNM was developed to predict the behavior of glyceride enzymatic transesterification aimed at biodiesel production from waste olive oil. The formulated HNM, based on the combination of a rigorous kinetic model and an ANN, was aimed at identifying the mutual relationships existing between the inputs (enzyme/substrate, glycerides/alcohol feed mass ratios, and some of the most important operating conditions) and the outputs (biodiesel composition and reaction yield). The developed hybrid neural model incorporated both an a priori knowledge of enzymatic transesterification kinetics and an artificial neural network aimed at identifying the difficult-to-model (uncertain) part of process dynamics.
