Artificial Neural Network Modelling and Process Optimization of Tray Drying for Taro Starch Powder Production

Abstract

Objective: Investigation of drying kinetics of taro starch powder through artificial neural network (ANN) modelling and optimization of its tray drying operational conditions for improved performance, reduced operational time, and quality retention. Methodology: The taro was pre-processed with sequential steps like cleaning, peeling, cutting, blending with distilled water (1:2 w/v), filtration, and centrifugation. In order to reduce the time with maximum quality retention, the slurry was dried at different tray drying conditions, such as drying temperature (50, 60, and 70 °C) and slurry thickness (2, 4, and 6 mm). An ANN model was employed to predict the drying kinetics of the taro slurry. Furthermore, process optimization was done by considering the parameters like drying time, color, starch content, resistant starch content, swelling power, solubility, and bulk density. Results and Discussion: Results indicated that the ANN model was well fitted with the experimental data, showing greater fit with higher R2 value (0.9944 to 0.9995), lower RMSE (0.0073 to 0.0240), and lower 𝜒2 values (6.2 x 10-4 to 5.8 x 10-5) at different drying conditions. The optimal condition was found at 64 °C and 4 mm thickness, and optimized readings are also found well aligned with experimental responses. Therefore, it can be concluded that the optimized tray drying process significantly improved performance, reduced operational time, and retained quality.