Neural network and adaptive neuro-fuzzy inference system modeling of the hot air-drying process of orange-fleshed sweet potato

The primary objective of this study is to determine the hot air drying characteristics and nutritional quality of orange-fleshed sweet potato (OFSP) in a convective dryer. Three temperatures (323.15, 333.15, and 343.15 K) and fan speed levels (0.5, 0.9, and 1.3 m/s) were used. A rehydration study of dried OFSP was also carried out. Modeling and prediction of experimental moisture data were done using artificial neural networks (ANNs) and adaptive neuro-fuzzy inference system (ANFIS) models. The result showed that the drying rate and rehydration ratio were significantly (p < .05) affected by drying temperature and fan speed levels. The effective diffusivity (D-eff) of the samples ranged from 2.5 x 10(-9) to 4.25 x 10(-9) m(2)/s, and it was found to increase with temperature and fan speed. Protein and fat content appeared to be strongly influenced by drying processing variables, whereas other properties appeared to be insignificant. ANFIS showed better modeling ability than ANNs in predicting the experimental moisture data of OFSP with R-2 and RMSE values of .99786 and 0.0225 respectively. In conclusion, the findings from this research will be useful in product optimization and process monitoring of hot air drying of OFSP, in establishing its drying temperature and fan speed. Practical applications Dried orange-fleshed sweet potato (OFSP) is utilized as a precursor to many industrial goods and feedstock in the food industry. Establishing the process conditions for drying of OFSP is very important for product adaptability by industries. Modeling the drying kinetic data is useful for developing controls for industrial dryers. Mathematical models have been used in time passes, although they lack the robustness to combine several process variables at time. Therefore, this study applied robust artificial intelligence tool; artificial neural networks, and adaptive neuro-fuzzy inference system (ANFIS) in the prediction of the drying curve of OFSP. Also, the study shows how the process variables affect the quality of the chips. ANFIS showed better prediction ability, and thus can be used in developing robust control systems for industrial drying of OFSP.