AHP-CRITIC混合加权法、灰色关联度分析与反向传播人工神经网络在芪志方提取工艺优化中的综合应用  

Comprehensive Application of AHP-CRITIC Hybrid Weighting Method,Grey Correlation Analysis and BP-ANN in Optimization of Extraction Process of Qizhi Prescription

作　　者：兰群 程怡 李子安 吴冰雨 王锦玉[1] 刘德文[1] 仝燕[1] LAN Qun;CHENG Yi;LI Zian;WU Bingyu;WANG Jinyu;LIU Dewen;TONG Yan(Institute of Chinese Materia Medica,China Academy of Chinese Medical Sciences,Beijing 100700,China)

机构地区：[1]中国中医科学院中药研究所,北京100700

出　　处：《中国实验方剂学杂志》2025年第8期176-186,共11页Chinese Journal of Experimental Traditional Medical Formulae

基　　金：国家重点研发计划项目(2023YFC3504000);中国中医科学院科技创新工程项目(CI2021A04303)。

摘　　要：目的:基于层次分析法(AHP)-指标相关性的指标权重确定方法(CRITIC)混合加权法、灰色关联度分析与反向传播人工神经网络(BP-ANN),优化芪志方的水提取工艺,为芪志方制备工艺优化及质量标准的建立提供实验依据。方法:采用L_(9)(3^(4))正交试验,结合AHP-CRITIC混合加权法确定各指标成分黄芪甲苷、细叶远志皂苷、毛蕊异黄酮葡萄糖苷、远志酮Ⅲ、3,6′-二芥子酰基蔗糖质量分数和干膏得率的权重系数,计算正交试验中各因素水平组合的综合评分,作为评价指标以选择较优工艺参数,通过直观分析、方差分析、灰色关联度分析考察提取次数、提取时间、溶剂用量对芪志方水提工艺的影响;同时,建立BP-ANN分析模型,反向预测该复方的最优提取工艺因素水平,并对优化的工艺参数进行验证。结果:AHP-CRITIC混合加权法确定5个指标成分黄芪甲苷、细叶远志皂苷、毛蕊异黄酮葡萄糖苷、远志(口山)酮Ⅲ、3,6′-二芥子酰基蔗糖质量分数、干膏得率的权重系数分别为25.7%、20.82%、16.41%、12.45%、15.96%、8.67%。优化的提取工艺参数为分别加8、6、6倍量水提取3次,每次1 h。BP-ANN检测样本的网络预测结果与正交试验结果一致,网络预测值和实际测量值的均方误差(MSE)<1%。通过相关数学模型分析和预测所得的芪志方水提取工艺稳定可行,黄芪、远志有效成分的提取率明显升高,验证试验平均综合评分90.85分,相对标准偏差(RSD)1.55%。结论:该研究建立了复方芪志颗粒的水提工艺,优选的提取工艺可有效提高黄芪、远志有效成分的提取效率,可为其他临床经验方的制备工艺优化及质量标准建立提供有益借鉴。Objective:Based on analytic hierarchy process(AHP)-criteria importance through intercriteria correlation(CRITIC)hybrid weighting method,grey relational analysis and backpropagation artificial neural network(BP-ANN),to optimize the water extraction process of Qizhi prescription,so as to provide an experimental basis for optimization of the preparation process of this prescription and the establishment of quality standards.Methods:L_(9)(3^(4))orthogonal test was employed,and the AHP CRITIC hybrid weighting method was utilized to determine the weight coefficients of the quality fractions of various components,including astragalosideⅣ,polygalaxanthoneⅢ,calycosin-7-O-β-D-glucoside,tenuifolin,and 3,6′-disinapoylsucrose,as well as the dry extract yield.The comprehensive score of each factor level combination in the orthogonal test were calculated as evaluation indicator to select the optimal extraction process parameters.The effects of extraction times,extraction time,and solvent dosage on the aqueous extraction process of the formula were investigated through intuitive analysis,variance analysis,and grey relational analysis.Meanwhile,a BP-ANN model was established to reverse-predict the optimal extraction process parameters of Qizhi prescription,and the optimized process parameters were validated.Results:The weight coefficients of the five index components(astragalosideⅣ,tenuifolin,calycosin-7-O-β-D-glucoside,polygalaxanthoneⅢ,and 3,6′-disinapoylsucrose)and dry extract yield were 25.7%,20.82%,16.41%,12.45%,15.96%and 8.67%,respectively.The optimized extraction process parameters were extracted 3 times with 8,6,6 times the amount of water,each time for 1 h.The network prediction results of BP-ANN test samples were consistent with the orthogonal test results,and the mean square error(MSE)of the predicted and measured values of the network was<1%.The water extraction process of Qizhi prescription analyzed and predicted by relevant mathematical models was stable and feasible,which could effectively improve the

关 键 词：芪志方 层次分析法(AHP) 指标相关性的指标权重确定方法(CRITIC) 正交试验 灰色关联度分析 反向传播人工神经网络(BP-ANN) 提取工艺 

分 类 号：R277[医药卫生—中医学] R943[理学—分析化学] O657[理学—化学]