摘要
目的利用BP神经网络-遗传算法结合信息熵理论优化凉粉草煎煮提取工艺。方法采用HPLC测定咖啡酸、紫云英苷、迷迭香酸、丹酚酸B的含量,UV法测定总黄酮含量,以4种指标成分含量、总黄酮含量及干膏收率为综合评分指标,采用信息熵确定各指标的客观权重,实现对提取工艺的正交试验优选;以正交试验结果作为初始种群,以加水倍数、煎煮时间、提取次数为输入值,提取综合得分作为其函数的输出值,通过BP神经网络模型结合遗传算法对煎煮提取工艺进行全局寻优。结果根据正交试验结果,最佳提取工艺为加16倍量水、煎煮3次、每次1.5 h,综合评分均值为92.08;建立结构为3-7-1的三层BP人工神经网络,结合遗传算法确定最优提取方案为加17倍量水、煎煮3次、每次1.5h,综合评分均值为92.55,优于正交试验,3批验证结果与预测值相对标准偏差均小于5%。结论BP神经网络-遗传算法结合信息熵理论稳定性及重复性好,可用于凉粉草提取工艺的预测和优选。
Objective To optimize the extraction process parameters of Mesona chinensis Benth.by BP neural network-genetic algorithm combined with information entropy theory.Methods HPLC was employed to determine the contents of caffeic acid,astragalin,rosmarinic acid and salvianolic acid B,and the content of total flavonoids was determined by UV method.Based on the four component contents,total flavonoid content,and dry paste yield as comprehensive scoring indicators,information entropy was used to determine the objective weight of each indicator to achieve the optimization of the extraction process by orthogonal experiments.Based on the orthogonal test as the initial population,the water addition multiple,decoction time and extraction times as the input values,extract comprehensive score as the output value,the global optimization of decocting extraction process was carried out by BP neural network model combined with genetic algorithm.Results According to the results of orthogonal experiment,the best extraction process conditions were as follows:extracted three times with 16 times of water quantity,each for 1.5 h;the comprehensive score was 92.08;a three-layer BP artificial neural network with a structure of 3-7-1 was established;combined with genetic algorithm to determine the optimal extraction plan was to add 17 times the amount of water,decocting 3 times for 1.5 hours each time;the comprehensive score was 92.55,which was better than orthogonal experiment;the RSD of the three batches of verification results and the predicted value was less than 5%.Conclusion BP neural network-genetic algorithm combined with information entropy theory has good stability and repeatability,and can be used for prediction and optimization of Mesona chinensis Benth.extraction process.
作者
谢平
魏海峰
温仁华
沈金海
郝春莉
陈良华
XIE Ping;WEI Haifeng;WEN Renhua;SHEN Jinhai;HAO Chunli;CHEN Lianghua(College of Environment and Public Health,Xiamen Huaxia University,Xiamen 361024,China;Biochemical Pharmacy Engineering Research Center of Fujian Province University,Xiamen 361024,China;Xiamen Key Laboratory of Food and Drug Safety,Xiamen 361024,China;Fujian Institute of Subtropical Botany,Xiamen 361006,China)
出处
《中国中医药信息杂志》
CAS
CSCD
2022年第2期86-92,共7页
Chinese Journal of Information on Traditional Chinese Medicine
基金
中央引导地方科技发展专项(YDZX20193502000001)
福建省教育厅中青年教师教育科研项目(JAT190982、JAT200879、JAT200874)
福建省大学生创新创业训练计划项目(202012709029)
厦门华厦学院“育苗基金”项目(HJYM2020012)。
关键词
凉粉草
煎煮提取工艺
正交试验
信息熵理论
BP神经网络-遗传算法
Mesona chinensis Benth.
decoction and extraction process
orthogonal experiment
information entropy theory
BP neural network-genetic algorithm