A hybrid lumped parameter/neural network model for spouted bed drying of pastes with inert particles

Abstract

The current study analyzed the suitability of a hybrid CST/neural network model to describe the highly coupled heat and mass transfer during paste drying in a spouted bed. In the present approach, the main information was the moisture content predictions in the powder. The model was based on global energy and water mass balances in the liquid and the gaseous phases. In this model, the inter-phase coupling term r, which reflects both water evaporation and particle coating, was described by an artificial neural network. Artificial neural networks are efficient computing models which are extensively used whenever theoretical models fail to properly represent a given phenomena and reliable data basis of the main variables involved is available. Simulations were done in MatLab. The drying experiments for model verification were carried out in a conical semi-pilot scale spouted bed, from which measurements of gas and solid phase moisture were done. The good agreement between calculated and measured powder moisture content suggested that the well-mixed hypothesis could be applied for paste drying in a spouted bed. The robustness of the model with respect to changes in feed flow rates and other operating conditions showed the merits of using a trained neural network.