Abstract
Drying can prolong the shelf life of a product by reducing microbial activities while facilitating its transportation and storage by decreasing the product weight and volume. The quality factors of the drying process are among the important issues in the drying of food and agricultural products. In this study, the effects of several independent variables such as the temperature of the drying air (50, 60, and 70 °C) and the thickness of the samples (2, 4, and 6 mm) were studied on the response variables including the quality indices (color difference and shrinkage) and drying factors (drying time, effective moisture diffusivity coefficient, specific energy consumption (SEC), energy efficiency and dryer efficiency) of the turnip slices dried by a hybrid convective-infrared (HCIR) dryer. Before drying, the samples were treated by three pretreatments: microwave (360 W for 2.5 min), ultrasonic (at 30 °C for 10 min) and blanching (at 90 °C for 2 min). The statistical analyses of the data and optimization of the drying process were achieved by the response surface method (RSM) and the response variables were predicted by the adaptive neuro-fuzzy inference system (ANFIS) model. The results indicated that an increase in the dryer temperature and a decline in the thickness of the sample can enhance the evaporation rate of the samples which will decrease the drying time (40–20 min), SEC (from 168.98 to 21.57 MJ/kg), color difference (from 50.59 to 15.38) and shrinkage (from 67.84% to 24.28%) while increasing the effective moisture diffusivity coefficient (from 1.007 × 10−9 to 8.11 × 10−9 m2/s), energy efficiency (from 0.89% to 15.23%) and dryer efficiency (from 2.11% to 21.2%). Compared to ultrasonic and blanching, microwave pretreatment increased the energy and drying efficiency; while the variations in the color and shrinkage were the lowest in the ultrasonic pretreatment. The optimal condition involved the temperature of 70 °C and sample thickness of 2 mm with the desirability above 0.89. The ANFIS model also managed to predict the response variables with R2 > 0.96