目的 基于理想解排序（technique for order preference by similarity to an ideal solution，TOPSIS）法的前馈控制实现不同批次丹参醇提液的均化。方法 首先采用积分球漫反射收集各批次丹参药材的近红外光谱数据；然后在不同批次丹参8种有效成分（丹参素、原儿茶醛、丹酚酸B、丹酚酸A、丹参酮I、隐丹参酮、丹参酮II_A、迷迭香酸）提取工艺单因素和含量测定实验的基础上，以提取时间、提取温度、乙醇体积分数为数值因素、批次为分类因素，8种有效成分含量经TOPSIS法综合评价后的得分为响应值，进行I-最优实验设计；最后建立以近红外数据，工艺参数为输入变量，提取液有效成分含量的综合得分为输出变量的遗传神经网络模型，以各批次丹参的近红外数据调整相应的醇提工艺参数，使得不同批次丹参醇提液中8种有效成分的含量差异减小。结果 通过建立的模型调整相应的提取工艺使得4个批次丹参醇提液中各成分含量RSD值的平均值24.9%降低至13.5%。结论 基于TOPSIS法的前馈控制能提升丹参醇提液质量的一致性，为中药提取液的均化研究提供了参考。

Objective The feedforward control based on the technique for order preference by similarity to an ideal solution (TOPSIS) method is used to homogenize alcohol extracts of different batches of Danshen (Salvia miltiorrhiza). Methods Firstly, the near-infrared diffuse reflectance spectra of S. miltiorrhiza were obtained. Then a high-performance liquid chromatography method had been developed for the simultaneous quantification of eight active components (danshensu, protocatechualdehyde, salvianolic acid B, salvianolic acid A, tanshinone I, cryptotanshinone, tanshinone II_A, rosmarinic acid) contained in alcohol extracts of S. miltiorrhiza.On the basis of single factor experiment and content determination experiment, extraction time, extraction temperature, ethanol concentration were take as numerical factors, the batch was take as categorical factors, eight control targets in the alcohol extract which were comprehensively evaluated by TOPSIS method was taken as response of I-optimal experimental design for the experiment. Finally, a genetic neural network model which reflected the relationship between near-infrared data, process parameters and the effective components was established. Based on the established model, the extraction process was adjusted according to the near-infrared data to decrease the difference in eight active components caused by batch change. Results Based on the established models, extraction conditions were optimized according to near-infrared data. The RSD value of each component content in alcohol extract of S. miltiorrhiza was reduced from 24.9% to 13.5%. Conclusion The feedforward control based on the TOPSIS method has ability to improve the quality and consistency of the alcohol extract of S. miltiorrhiza.It provides a reference for research on the homogenization of Chinese medicine extraction process.

R283.6

国家自然科学基金资助项目（81873226）；国家科技重大专项重大新药创制（2019ZX09301101）；浙江省“万人计划”科技创新领军人才项目（2019）