摘要
在单因素试验基础上,采用Box-Behnken设计响应面和遗传算法-神经网络两种方式对龙牙百合总黄酮提取工艺条件进行优化。结果表明:各因素对提取结果均呈现先上升后下降的趋势,响应面法和遗传算法-神经网络模型法相对误差和决定系数R2分别为1.19%、0.9554和0.72%、0.9947。经验证,遗传算法-神经网络模型优化结果高于响应面,表明前者具有更强优化能力。最终采用遗传算法-神经网络优化获得提取龙牙百合总黄酮最佳工艺条件为:提取温度73℃、提取时间50 min、液固比43∶1(mL/g)、乙醇体积分数53%,在此条件下总黄酮含量为47.17 mg/g,高于响应面预测值46.63 mg/g。
Response surface methodology(RSM)of Box-Behnken design and genetic algorithm-neural network(GA-NN)were used on the basis of the single-factor test to optimize the extraction of total flavonoids from Lilium brownii.The results showed that the amount of each extracted factor showed a rising trend followed by a decline.GA-NN showed better prediction and optimization abilities and had a lower relative error rate(0.72%versus 1.19%)and higher determination coefficient R2(0.9947 versus 0.9554)than RSM.Accordingly,the optimal conditions determined using GA-NN were as follows:50 min extraction time,53%ethanol concentration,43∶1(mL/g)liquid-solid ratio,73℃temperature.Under these optimized conditions,the total flavonoid content was 47.17 mg/g,which was higher than the amount obtained using RSM(46.63 mg/g).
作者
尹乐斌
邓鹏
何平
刘桠丽
李乐乐
YIN Le-bin;DENG Peng;HE Ping;LIU Ya-li;LI Le-le(College of Food and Chemical Engineering,Shaoyang University,Shaoyang 422000,Hunan,China;Hunan Provincial Key Laboratory of Soybean Products Processing and Safety Control,Shaoyang 422000,Hunan,China)
出处
《食品研究与开发》
CAS
北大核心
2021年第7期105-113,共9页
Food Research and Development
基金
湖南省教育厅优秀青年项目(18B427)
邵阳学院研究生科研创新项目(CX2019SY057)
邵阳学院“双一流”建设产学研合作平台(邵院通〔2018〕50号)。
关键词
龙牙百合
总黄酮
响应面
人工神经网络
遗传算法
Lilium brownii
total flavonoids
response surface methodology
artificial neural network
genetic algorithm
