A comparative study was carried out between a large number of mathematical models and artificial neural networks to estimate the drying curves. Diamente et al. model and modified two‐term exponential‐V model were determined as the best ones describing drying curves for natural and forced drying air systems, respectively. The ANNs with 4‐9‐9‐1 and 3‐9‐1 topologies, log‐sigmoid, and hyperbolic tangent sigmoid transfer functions, and Levenberg–Marquardt training algorithm presented the best results for the former and latter systems, respectively. Furthermore, it was found that ANN modeling had much better performance in prediction of drying curves with respect to statistical analyses. Moisture diffusivity values were obtained in a range of 0.932×10-10-11.976×10-10normalm2/s and 2.209×10-10-9.848×10-10normalm2/s for the systems, respectively. Activation energies were determined as 88,509, 60,344, and 44,806 W/kg for the former system and 36.88, 29.66, and 18.59 kJ/mol for the latter system with an increase in the thickness of samples.
Practical applications
The main aim of drying food products is the reduction of moisture content up to a certain level to achieve the smallest possible amount of the microbiological spoilage. This process has a considerable effect on the drying kinetics and quality of the dried product as well. Infrared (IR) heating, as an alternative drying method for agricultural products, is efficient to preserve the main characteristics and to shorten drying time. It is important to investigate the effect operating parameters of IR dryer on the mentioned cases. Our studies have clearly displayed that IR heating of pumpkin samples can result in high heating rate and fast drying. Therefore, this alternative approach could be employed as an energy saving drying method along with improved drying efficiency and better product quality in comparison with the common drying methods. Study results are useful for producers of not only dried pumpkins but also other agricultural products.