基于正交设计和BP神经网络-遗传算法多指标综合优化茶叶提取工艺

Study on Multi-index Comprehensive Optimization of Tea Extraction Process Based on Orthogonal Design and BP Neural Network Genetic Algorithm

关键词:  茶叶  BP神经网络  遗传算法  多指标综合评价法  提取工艺

Key words:  tea  BP neural network  genetic algorithm  multi-index comprehensive evaluation method  extraction process

摘要:

目的 用正交设计及BP神经网络-遗传算法对茶叶提取工艺进行多指标综合优化。方法 以咖啡因、表没食子儿茶素没食子酸酯（epigallocatechin gallate，EGCG）、表儿茶素没食子酸酯（epicatechin gallate，ECG）为考察指标，在单因素实验的基础上，采用正交设计及BP神经网络-遗传算法优选超声辅助提取茶叶中有效成分的工艺，并对2种方法优选所得的工艺进行验证。结果 正交设计得到的最佳提取条件为乙醇浓度85%、浸提温度80℃、超声时间10 min。工艺验证评分为99.050。BP神经网络-遗传算法得到的最优提取方案为乙醇浓度89%、浸提温度88℃、超声时间13 min，网络预测评分为100.758，工艺验证评分为99.651，相对误差为1.099%。结论 BP神经网络-遗传算法数学模型可用于茶叶中有效成分提取工艺预测和优选，且略优于正交设计。

Abstract:

OBJECTIVE To optimize the extraction process of tea by orthogonal design and BP neural network-genetic algorithm. METHODS Using caffeine, EGCG and ECG as the indexes, based on the single factor experiment, orthogonal design and BP neural network-genetic algorithm were used to optimize the ultrasonic-assisted extraction process of effective components in tea, and these process by the two methods were validated. RESULTS The optimum extraction conditions were 85% ethanol concentration, 80℃ and 10 min ultrasonic time. The validation score was 99.050. The optimum extraction scheme obtained by BP neural network-genetic algorithm was ethanol concentration 89%, extraction temperature 88℃, ultrasonic time 13 min, network prediction score 100.758, process verification score 99.651, relative error 1.099%. CONCLUSION BP neural network-genetic algorithm mathematical model can be used to predict and optimize the extraction process of effective components in tea, and slightly better than orthogonal design.