Estimation of drying rate constant from static bed moisture profile by neural network inversion
Keywords:
Static bed drying, Non-equilibrium model, Drying rate constant, Bed moisture profile, Neural network inversion, Sum squared errorAbstract
This study aims at extracting a mathematical expression for describing the moisture loss kinetics from grains dried in the form of a static bed, based on a measured grain moisture profile across the bed, and validating its reliability in predicting the drying times. The target expression for moisture transfer is the Lewis equation with Arrhenius type dependence of the drying rate constant on temperature, thereby reducing the problem to the determination of two coefficients (i.e., Ea and K0) for the drying rate constant. The scheme of numerical solution of the non-equilibrium model of the deep bed drying process is represented as a trained neural network, with the values of the coefficients as inputs and the sum squared error (SSE) in the prediction of moisture content at various bed depths as the output. Training data for the neural network were generated for static bed drying of barley at an airflow rate of 638 kg/m2·h. The two coefficients were estimated by inversion of the trained neural network. The derived expression for drying rate constant was found to give a better prediction of the drying time and drying air temperature profiles at different experimental runs with air flow rates close to 638 kg/m2·h. It underlines the fact that grain moisture loss kinetics, extracted from a known moisture profile across the static bed can reliably be used to predict the batch drying time.
Keywords: static bed drying, non-equilibrium model, drying rate constant, bed moisture profile, neural network inversion, sum squared error