BP神经网络结合遗传算法优化玫瑰茄火龙果固体饮料工艺条件

Optimization of solid beverage process of rosette and dragon fruit by BP neural network combined with genetic algorithms

以火龙果与玫瑰茄干花提取液为芯材,麦芽糊精、可溶性淀粉、阿拉伯胶为壁材,采用喷雾干燥法制备固体饮料。首先使用单因素实验确定提取液用量和壁材使用量,采用正交设计实验探讨进料温度、进料速度、风机速度3个工艺参数对玫瑰茄火龙果固体饮料喷雾干燥效果的影响;在正交实验的基础上,通过BP神经网络确定预测模型后结合遗传算法对喷雾干燥工艺参数进行寻优,获得火龙果玫瑰茄固体饮料的最佳工艺。喷雾干燥最佳工艺条件为玫瑰茄火龙果提取液质量浓度600 g/L,进风温度131 ℃,进料速度22 mL/min,风机速度3.1 m 3/min,在此条件下得到的玫瑰茄火龙果固体饮料的出粉率为47.85%,相比正交实验的优化结果47.12%,提高了0.73%。BP神经网络结合遗传算法优化喷雾干燥工艺制备玫瑰茄火龙果固体饮料是可行的,可为农产品的深加工工艺优化提供参考。

Using the extract of the dried rosette and dragon fruit as the core material, and the maltodextrin, the soluble starch and the Arabic gum as the wall material, the solid beverage was prepared by spray drying. The amounts of extract and wall material were determined by single factor experiment, followed by orthogonal design experiment to explore the influences of feed temperature, feed speed and fan speed on spray drying effect of the solid beverage. On the basis of the orthogonal experiment, the parameters were optimized by BP neural network combined with the genetic algorithm and the optimum conditions were: the concentration of rosette and dragon fruit extract of 600 g/L, the inlet air temperature of 131 ℃, the feed speed of 22 mL/min, and the wind speed of 3.1 m3/min. The flour yield of the fruit solid beverage could reach 47.85%, which was 0.73% higher than that of orthogonal experiment. It is feasible to optimize spray drying process with BP neural network combined with genetic algorithm to prepare rosette and dragon fruit solid beverage, which provides a reference for the optimization of deep processing technology of agricultural products.