基于熵权-层次分析法及反向传播神经网络多指标优化地黄水提物提取工艺

Optimization of Extraction Process of Aqueous Extract of Rehmannia glutinosa Based on Entropy Weight Method in Cooperation with Analytic Hierarchy Process and Back Propagation Neural Network with Multiple Indicators

目的:基于熵权法(Entropy Weight Method, EWM)-层次分析(Analytic Hierarchy Process, AHP)和反向传播神经网络(Back Propagation Neural Network, BPNN)对地黄水提物超声提取工艺进行优化。方法:以超声温度、超声时间、料液比为正交试验考察因素,梓醇、地黄苷D、益母草苷,地黄多糖含量以及水溶性浸出物得率为指标,采用EWM-AHP法确定各指标权重,从而得到多指标综合得分;再以正交试验工艺条件为输入,综合得分为输出,建立并训练BPNN模型,寻找地黄提取最佳工艺。结果:BPNN优选工艺为在60℃下,加33倍水,提取70 min,梓醇、地黄苷D、益母草苷,地黄多糖含量以及水溶性浸出物得率平均值分别为17.04、3.75、10.57、24.86 mg/g、0.82 g/g,综合得分为97.74;正交试验最佳工艺为提取温度50℃,加25倍水,提取1 h,梓醇、地黄苷D、益母草苷,地黄多糖含量以及水溶性浸出物得率平均值分别为12.72、2.58、8.20、25.02 mg/g、0.80 g/g,综合得分为97.26。结论:本研究确定BPNN优选工艺为最优,为EWM-AHP法结合BPNN在提取工艺中的应用提供了参考。

Objective: To optimize the ultrasonic extraction process of aqueous extract of Rehmannia glutinosa based on entropy weight method(EWM) together with analytic hierarchy process(AHP) and back propagation neural network(BPNN). Methods: Taking ultrasonic temperature, ultrasonic time, solid-liquid ratio as orthogonal test factors, catalpa,rehmannioside D, ajugol, rehmannioside polysaccharide content, and the yield of water-soluble extract as indexes, EWMAHP method was used to determine the weight of each index, to obtain the comprehensive score of multiple indexes. Then,the experimental process conditions were used as the input and the synthesis score was used as the output to build and train a BPNN model to find the best process for the extraction process of Rehmannia glutinosa. Results: The preferred process of BPNN was extraction at 60 ℃ with 33 times of water for 70 min. The mean values of catalpa, rehmannioside D, ajugol,rehmannioside polysaccharide content, and the yield of water-soluble extract were 17.04, 3.75, 10.57, 24.86 mg/g and 0.82 g/g, respectively, with a combined score of 97.74;The best process of the orthogonal test was extraction at 50 ℃ with 25 times of water for 1 h. The mean values of catalpa, rehmannioside D, ajugol, rehmannioside polysaccharide content, and the yield of water-soluble extract were 12.72, 2.58, 8.20, 25.02 mg/g and 0.80 g/g, respectively, with a combined score of 97.26. Conclusion: This study determined that the BPNN preferred process was optimal and provided a reference for the application of the EWM-AHP method combined with BPNN in the extraction process.