Artificial neural networks to model kinetics and energy efficiency in fixed, fluidized and vibro-fluidized bed dryers towards process optimization

Drying of inorganic particulate compounds is recognized as an energy-intensive process. Due to this fact, optimization studies through modeling and simulation of experimental data play a critical role towards process economic benefits. The main drawback of some models from literature is their dependence on operating parameters and dryer type, which can be a key issue for data generalization. Artificial neural networks may provide the first step to solve this problem. This study analyses the feasibility of neural models to fit and estimate drying kinetics data, cumulative and instantaneous energy efficiency indices from fixed, fluidized and vibro-fluidized bed dryers at different operating conditions. The networks were trained considering different scenarios for both drying kinetics and energy analysis. It was shown that the neural model is consistent to estimate new patterns not addressed in the trainings for the case in which the database is regarded for a single type of dryer. Simultaneous training considering multiple datasets of each dryer resulted in predictions with poor accuracy, but considering the complex hydrodynamic conditions of the moving beds, there is room for improvement when efficient data is used. In this way, neural models can be considered an interesting tool to predict parameters also for energy analysis.