Artificial Neural Network Modeling of Hot-air Drying Kinetics of Mango Kernel

Abstract

Large quantities of mango seeds are generated as waste during extraction of mango pulp. The mango kernels are nutritionally rich and can be used as food in the form of flour and starch. Present study was undertaken to investigate the effect of blanching and convective drying air temperature of 50, 60 and 70°C on drying characteristics of mango kernel in splitted and shredded form. The drying characteristics of prepared samples were studied in terms of moisture ratio, drying time, and effective moisture diffusivity. The colour  parameters (‘L’, ‘a', ‘b’) of dried samples, were also estimated separately. Drying kinetics (moisture ratio vs drying time) of mango kernels modelled using three transfer functions (Tansig, Logsig and Purelin) of Artificial Neural Network (ANN). A reduction in the total drying time was observed with decrease in size of kernel but with rise in drying air temperature. The splitted and shredded kernels took about 450 to 840 min and 210 to 600 min respectively to be dried to final moisture content of 9 ± 1% (d.b.). Blanching did not show any significant influence on drying time. The drying process of mango kernels for all the conditions was observed to follow the falling rate. Modeling of drying kinetics of mango kernels was carried out using experimental results through artificial neural network. Results showed that the developed ANN model using logsig transfer function could predict the moisture ratio with high coefficient of determination (R² = 0.99) and low root mean square error (0.01) within the range of tested operating conditions. The established ANN model can be used for online prediction of moisture content of splitted and shredded mango kernels during hot air drying process which has relevance to the food and pharmaceutical industry to produce dried mango kernels at desired moisture content.