论文部分内容阅读
目的优化同时提取甘草中皂苷和总黄酮的工艺条件。方法在单因素试验的基础上,氨浓度(A)、乙醇浓度(B)、回流时间(C)、液料比(D)为自变量,选用响应面法中的中心组合设计(CCD)进行4因素5水平试验,采用紫外分光光度法进行测定,以甘草中皂苷和总黄酮含量作为检测指标,检测波长分别为252 nm和510 nm。运用R语言环境下的熵权法对上述2指标进行权重赋值,建立3层结构的BP神经网络模型,对不同隐层神经元(size)的个数进行模型的测试训练,最后采用R语言的实数编码程序,建立并优化遗传算法数学模型,对提取工艺进行目标寻优,最终得到最佳提取工艺。结果皂苷和总黄酮分别在质量浓度0.008~0.056 g/L和0.024~0.080 g/L与吸光度具有良好的线性关系,方法学考察符合测定要求。选取隐层神经元个数为5的神经网络模型,优化遗传算法的各项参数后对甘草中皂苷和总黄酮进行提取工艺的目标优化,最终得到的最佳提取条件为:氨浓度0.62%,乙醇浓度64%,回流时间1.8 h,液固比12∶1。模型综合评价预测值为191.65,而按照上述最佳提取条件试验所得的平均综合评价值为188.90,两者相对误差为1.43%,证明神经网络和遗传算法具有较好的预测性。结论建立的数学模型寻求同时提取甘草皂苷和总黄酮最佳提取条件是科学可行的,为实现中药化学成分乃至药效物质基础多目标寻优提供了新的参考和思路。
Objective To optimize the extraction of saponin and total flavonoids from Glycyrrhiza uralensis Fisch. Methods Based on the single factor test, ammonia concentration (A), ethanol concentration (B), reflux time (C) and liquid to liquid ratio (D) were used as independent variables. 4 factor 5 level test, using UV spectrophotometry determination of licorice saponin and total flavonoids as a detection index, the detection wavelength were 252 nm and 510 nm. We use entropy weight method in R language environment to carry on the weight assignment to the above two indexes, build a BP neural network model with 3 layers structure, test the number of different hidden layer neurons (size), finally use R language Real number encoding program, establish and optimize the mathematical model of genetic algorithm, optimize the target of the extraction process, and finally get the best extraction process. Results Saponin and total flavonoids had good linearity between absorbance and concentration of 0.008 ~ 0.056 g / L and 0.024 ~ 0.080 g / L, respectively. The methodological study accorded with the determination requirements. The neural network model with 5 neurons in hidden layer was selected to optimize the extraction of saponin and total flavonoids in Glycyrrhiza glabra after optimization of the parameters of genetic algorithm. The optimal extraction conditions were: ammonia concentration 0.62% Ethanol concentration of 64%, reflux time of 1.8 h, liquid to solid ratio of 12: 1. The prediction value of model comprehensive evaluation is 191.65, while the average comprehensive evaluation value obtained according to the above optimal extraction condition test is 188.90, the relative error between the two is 1.43%, which proves that neural network and genetic algorithm have good predictability. Conclusion It is scientifically feasible to establish the mathematical model to extract the optimal extraction conditions of glycyrrhizin and flavonoids at the same time. It provides a new reference and ideas for realizing the multi-objective optimization of the chemical constituents of traditional Chinese medicine and even the medicinal substances.