A new predictive neural architecture for solving temperature inverse problems in microwave-assisted drying processes

In this paper, a novel learning architecture based on neural networks is used for temperature inverse modeling in microwave-assisted drying processes. The proposed design combines the accuracy of the radial basis functions (RBF) and the algebraic capabilities of the matrix polynomial structures by using a two-level structure. This architecture is trained by temperature curves, Tc(t), previously generated by a validated drying model. The interconnection of the learning-based networks has enabled the finding of electric field (E) optimal values which provide the Tc(t) curve that best fits a desired temperature target in a specific time slot.