【摘要】 目的 基于中药指纹图谱技术，结合CRITIC赋权，采用Box-Behnken响应面法结合BP(back-propagation)神经网络多指标优化汉桃叶总黄酮提取工艺。方法 采用Box-Behnken响应面法考察乙醇浓度（%）、回流提取时间（简称提取时间，min）和料液比（g∶m L）对汉桃叶总黄酮提取工艺的影响，制备汉桃叶提取物，计算干膏得率（%），测定DPPH·清除率（%），建立汉桃叶提取物指纹图谱。以槲皮素峰面积、指纹图谱共有峰总峰面积（总峰面积）、干膏得率（%）、DPPH·清除率（%）为评价指标，基于CRITIC客观赋值法计算综合评价结果（R），获得最佳提取工艺。建立BP神经网络模型，选取合理数据进行学习和验证，并预测汉桃叶的最佳提取工艺。结果 采用Box-Behnken响应面法及BP神经网络预测的汉桃叶总黄酮的最佳提取工艺均为乙醇浓度85%、提取时间50 min、料液比1∶6(g∶mL)。验证试验的结果表明，Box-Behnken响应面法的综合评价结果均值与响应面优化理论值的绝对误差为0.045 3、相对误差为3.50%。结论 基于指纹图谱及CRITIC赋权的Box-Behnken响应面法结合BP神经网络多指标优化汉桃叶总黄酮提取工艺，结果可靠、稳定，可用于汉桃叶总黄酮的提取。更多还原

【Abstract】 Objective: To optimize the extraction method of total flavonoids in Hantaoye by using Box-Behnken response surface method combined with BP neural network based on fingerprint technology of traditional Chinese medicine and CRITIC weighting method. Methods: Effects of ethanol concentration(%), reflux extraction time(abbreviated as extraction time, min), and solid-liquid ratio(g∶mL) on the extraction process of total flavonoids in Hantaoye were studied by using Box-Behnken response surface method, Hantaoye extract was prepared and its dry extract yield(%) was calculated, and its DPPH· clearance rate(%) was measured. HPLC fingerprint of Hantaoye extract was established. The optimal extraction process was obtained by calculating comprehensive evaluation results(R) based on CRITIC objective weigting method taking peak area of quercetin, total peak area of common peaks in fingerprint(hereinafter referred to as total peak area), dry extract yield(%) and DPPH· clearance rate(%) as evaluation indicators. A BP neural network model was established, reasonable date was selected for learning and validation, and the optimal extraction process for Hantaoye was predicted. Results: The optimal extraction process for total flavonoids in Hantaoye was obtained by using Box-Behnken response surface method and BP neural network kwas as follows: ethanol concentration of 85%, extraction time of 50 min, and solid-liquid ratio of 1∶6(g∶mL). The results of the validation test indicated show that the absolute error between the mean value of the comprehensive evaluation results of the Box-Behnken response surface method and the theoretical value of the response surface optimization was 0. 045 3, and the relative error was 3. 50%. Conclusion: Box-Behnken response surface method based on fingerprint and CRITIC weighting combined with BP neural network optimization for the extraction process of total flavonoids in Hantaoye is reliable and stable, and can be used for the extraction of total flavonoids from Hantaoye.更多还原