Vacuum pulse drying characteristics and moisture content prediction of pumpkin slices

Bai Junwen,Zhou Cunshan,Cai Jianrong,Xiao Hongwei,Gao Zhenjiang,Ma Haile

DOI: https://doi.org/10.11975/j.issn.1002-6819.2017.17.038

2017-01-01

Abstract:Pumpkin is one of the most important vegetable crops grown in the world because of its nutritional qualities. Pumpkin is a kind of seasonal vegetable, which is generally harvested from August to September. Due to its high moisture content, pumpkin cannot be stored for a long time even in low temperature environment. Drying is one of the most important processing processes for pumpkin, which can greatly prolong the shelf life. At present, the most common drying method of pumpkin is hot air drying. Hot air drying is relatively simple and easy to control, but it also has some problems such as long drying time, browning seriously and loss of nutrient. Vacuum pulse drying is a new type of drying technology, which has high drying efficiency and is very suitable for heat-sensitive materials. The aim of this study was to investigate the drying characteristics and to predict the moisture content of pumpkin slices during the vacuum pulse drying process. The effect of atmospheric pressure duration (AD), vacuum duration (VD), drying temperature and slice thickness on the drying rate and drying time of pumpkin slices was studied. The effect of pressure pulsation process on heat and mass transfer of pumpkin slices was expounded on the basis of core temperatures which were gathered by temperature sensor during drying process. The BP (back propagation) neural network model was established with the architecture of "5-11-1" which included 5 input layers of AD, VD, drying temperature, slice thickness and drying time, 11 hidden layers and single output layer of moisture content. The results demonstrated that AD and VD showed a significant impact on drying time. In a certain range, the application of longer AD time and shorter VD time not only can improve the drying efficiency, but also can significantly reduce vacuum pump running time during the drying process. The shortest drying time was about 352 min with the AD of 10 min and the VD of 9 min under drying temperature of 60 ℃ and slice thickness of 7 mm. The drying temperature and slice thickness both showed a significant impact on drying rate and drying time, and higher drying temperature and thinner thickness would result in higher drying rate. Unlike the hot air or infrared drying method, the core temperature of pumpkin presented an alternate'high-low-high-low' phenomenon. During the VD, the core temperature decreased rapidly due to the endothermic process of moisture evaporation, and after the drying chamber pressure returned to atmospheric pressure, the core temperature was raised rapidly due to the heating effect by electrical heating plate. In the early period of drying process, the change range of core temperature between AD and VD phase was greater than that of the late drying period. The BP neural network model was trained for finite iteration calculation with Levenberg-Marquardt (LM) algorithm as the training function and tansig-purelin as the network transfer function. The determination coefficient (R2) and root mean squared error (RMSE) between the predicted and measured values were 0.996 8 and 0.017 3, respectively. The results will provide theoretic reference and technical supports for the application of vacuum pulse drying and the on-line prediction of moisture content in pumpkin drying process.