Kaveh Mohammad, Chayjan Reza Amiri
Department of Biosystems Engineering, Faculty of Agriculture, Bu-Ali Sina University, Hamedan, Iran.
Acta Sci Pol Technol Aliment. 2014 Jan-Mar;13(1):65-78. doi: 10.17306/j.afs.2014.1.6.
Drying of terebinth fruit was conducted to provide microbiological stability, reduce product deterioration due to chemical reactions, facilitate storage and lower transportation costs. Because terebinth fruit is susceptible to heat, the selection of a suitable drying technology is a challenging task. Artificial neural networks (ANNs) are used as a nonlinear mapping structures for modelling and prediction of some physical and drying properties of terebinth fruit.
Drying characteristics of terebinth fruit with an initial moisture content of 1.16 (d.b.) was studied in an infrared fluidized bed dryer. Different levels of air temperatures (40, 55 and 70°C), air velocities (0.93, 1.76 and 2.6 m/s) and infrared (IR) radiation powers (500, 1000 and 1500 W) were applied. In the present study, the application of Artificial Neural Network (ANN) for predicting the drying moisture diffusivity, energy consumption, shrinkage, drying rate and moisture ratio (output parameter for ANN modelling) was investigated. Air temperature, air velocity, IR radiation and drying time were considered as input parameters.
The results revealed that to predict drying rate and moisture ratio a network with the TANSIG-LOGSIG-TANSIG transfer function and Levenberg-Marquardt (LM) training algorithm made the most accurate predictions for the terebinth fruit drying. The best results for ANN at predications were R2 = 0.9678 for drying rate, R2 = 0.9945 for moisture ratio, R2 = 0.9857 for moisture diffusivity and R2 = 0.9893 for energy consumption.
Results indicated that artificial neural network can be used as an alternative approach for modelling and predicting of terebinth fruit drying parameters with high correlation. Also ANN can be used in optimization of the process.
对笃耨香果实进行干燥处理是为了确保微生物稳定性，减少因化学反应导致的产品变质，便于储存并降低运输成本。由于笃耨香果实对热敏感，因此选择合适的干燥技术是一项具有挑战性的任务。人工神经网络（ANN）被用作非线性映射结构，用于对笃耨香果实的一些物理和干燥特性进行建模和预测。
在红外流化床干燥器中研究了初始含水量为1.16（干基）的笃耨香果实的干燥特性。采用了不同水平的空气温度（40、55和70°C）、空气流速（0.93、1.76和2.6 m/s）以及红外（IR）辐射功率（500、1000和1500 W）。在本研究中，研究了应用人工神经网络（ANN）预测干燥水分扩散率、能量消耗、收缩率、干燥速率和水分比（ANN建模的输出参数）。将空气温度、空气流速、IR辐射和干燥时间视为输入参数。
结果表明，对于预测干燥速率和水分比，具有TANSIG-LOGSIG-TANSIG传递函数和Levenberg-Marquardt（LM）训练算法的网络对笃耨香果实干燥做出了最准确的预测。ANN预测的最佳结果为：干燥速率的R2 = 0.9678，水分比的R2 = 0.9945，水分扩散率的R2 = 0.9857，能量消耗的R2 = 0.9893。
结果表明，人工神经网络可作为一种替代方法，用于对笃耨香果实干燥参数进行具有高度相关性的建模和预测。此外，ANN可用于该过程的优化。
