基于色度-化学成分关联的蜜糠炒白芍质量控制研究

目的 建立30批白芍和蜜糠炒白芍的指纹图谱，研究白芍经蜜糠炒制后颜色的变化与内在成分的关联性，并构建相关的人工神经网络（artificialneuralnetwork,ANN）模型。方法 利用色差仪对白芍及蜜糠炒白芍色度进行测定，通过HPLC建立白芍及蜜糠炒白芍的HPLC指纹图谱，结合多元统计学方法对白芍及蜜糠炒白芍样品色度值与共有成分进行关联分析。以色度和相关成分建立ANN模型。结果 白芍经蜜糠炒制后，饮片粉末的L*、a*、b*值呈上升趋势；相关分析结果表明没食子酸（峰1）、芍药苷（峰6）、没食子酰芍药苷（峰7）、五没食子酰葡萄糖（峰8）、苯甲酸（峰9）、苯甲酰芍药苷（峰10）与色度值存在显著相关；当隐含层神经元个数为8时，ANN模型的均方根误差最低（RMSE=0.124 0）,R2最高（R2=0.9835），模型整体拟合程度最好。因此选择4个输入层、8个隐含层和6个输出层组成ANN模型为最佳拓扑结构。结论 色度与成分结合能更加全面科学地评价蜜糠炒白芍整体质量，可将色度指标纳入蜜糠炒白芍饮片质量评价体系中，为蜜糠炒白芍饮片在线质量控制提供参考。

Quality control of honey bran-fried Paeoniae Radix Alba based on color-chemical composition association

Objective To establish the fingerprints of 30 batches of Baishao (Paeoniae Radix Alba, PRA) and honey bran-fried PRA (hb-fPRA), study the correlation between the color change and the internal components of PRA after being fried with honey bran was studied, and construct the relevant artificial neural network (ANN) model. Methods The chroma of PRA and hb-fPRA was determined by colorimeter. The HPLC fingerprints of PRA and hb-fPRA were established by HPLC, and the correlation analysis between the chromaticity value of PRA and hb-fPRA and the common components was analyzed by multivariate statistical method. ANN model was established by chromaticity and correlation components. Results After the PRA is fried with honey bran, the L^*, a^* and b^* value of the decoction piece powder showed an upward trend. The correlation analysis results showed that gallic acid (peak 1), paeoniflorin (peak 6), galloylpaeoniflorin (peak 7), pentagalloylglucose (peak 8), benzoic acid (peak 9) and benzoylpaeoniflorin (peak 10) are significantly correlated with the chromaticity values. When the number of neurons in the implicit layer is eight, the root mean square error of the ANN model is the lowest (RMSE = 0.124 0) and the highest R² (R² = 0.983 5), and the overall fit of the model is the best. Therefore, the ANN model is selected as the best topology for four input layers, eight hidden layers, and six output layers. Conclusion The combination of color and ingredients can more comprehensively and scientifically evaluate the overall quality of hb-fPRA, and the color index can be included in the quality evaluation system of hb-fPRA decoction pieces, providing a reference for the online quality control of hb-fPRA.