Artificial neural network for prediction of dielectric properties relevant to microwave processing of fruit juice

M. Kubo (1, 2) , S. Curet (1, 3) , P. Augusto (2) , Lionel Boillereaux (1, 3)

1 GEPEA-OSE - Optimisation - Système - Energie
2 USP - Universidade de São Paulo = University of São Paulo
3 ONIRIS - École nationale vétérinaire, agroalimentaire et de l'alimentation Nantes-Atlantique

Dielectric properties are fundamental parameters to determine the heating behavior of food products using microwaves. Determining and modeling these properties as a function of process conditions and product composition is required to be implemented in Multiphysics modeling, in which heat transfer and electromagnetic aspects are solved, allowing the characterizing and optimization of microwave processing. This work evaluates the dielectric properties of model fruit juice solutions at concentrations ranging from 5 to 65°Brix. Measurements were carried out at frequencies between 200 and 3,000 MHz in the temperature range of 20–80 °C. The results show that the dielectric properties are non-linearly influenced by temperature, frequency and sugar content. A comparison between the dielectric properties of model solutions and of real juices was performed, and similarities were found. Furthermore, based on Artificial Neural Networks, the modeling and prediction of the dielectric properties at different conditions were achieved.

Domaines: Génie des procédés

M. Kubo, S. Curet, P. Augusto, Lionel Boillereaux. Artificial neural network for prediction of dielectric properties relevant to microwave processing of fruit juice. Journal of Food Process Engineering, 2018, 41 (6), pp.e12815. ⟨10.1111/jfpe.12815⟩. ⟨hal-01959514⟩