A simplified modelling approach for predicting shrinkage and sensitive material properties during low temperature air drying of porous food materials

Many food materials are dried to enhance shelf-life. Drying is an energy-intensive process, and accurate drying models could be used in real time process control of drying equipment to drive cost optimizations. However, most physics-based models suffer from two shortcomings: they require thermo-physical properties of the food materials to be known a priori, and they usually neglect material shrinkage due to moisture loss. In this work, we first develop a simplified physics-based transport model to predict temperatures and moisture content and corresponding spatial and temporal shrinkage during low temperature air drying process, where volumetric shrinkage is dominated by moisture loss. This model agrees well with experiments conducted by us (reported in this paper) as well as with those conducted by others (taken from the literature) on food samples. Further, using the validated modelling framework, we have developed an experimentally validated deep learning-based artificial neural network (ANN) model for properties' estimation. This ANN model is designed to estimate solid density, initial porosity, and initial water saturation of a given food material, using temperature and moisture data from a set of simple experiments with error less than 1%. Using these predicted parameters as input, the physics-based model can predict temperature and moisture for real-time drying to within 5% accuracy. The method proposed in this work could play an important role in industrial drying process optimisation and will find wide applications in the food processing industry.