目的 建立基于CRITIC法赋权结合Box-Behnken设计-响应面法（Box-Benhnken design-response surface methodology，BBD-RSM）与人工神经网络（artificial neural network，ANN）的酒香附Cyperi Rhizoma多指标综合权重优化工艺，基于模型预测界定工艺鲁棒操作区间，为参数固化与质量一致性控制提供参考。方法 以香附烯酮、α-香附酮、总黄酮和总挥发油含量为综合评价指标，应用CRITIC法客观赋权并计算综合评价值（overall desirability，OD）。在单因素试验基础上，以炮制温度、炮制时间、闷润时间、投药量为自变量，采用BBD进行4因素3水平试验并建立回归模型。同时构建ANN预测模型，并采用Garson算法对因素相对贡献度进行解析；结合帕累托非支配解集（Pareto non-dominated solution set）对工艺参数空间进行筛选，获得鲁棒操作区间（模型预测）。结果 CRITIC法确定香附烯酮、α-香附酮、总挥发油及总黄酮的权重分别为0.274 9、0.255 4、0.253 4、0.216 3。确定的酒香附最佳工艺条件为炮制温度140 ℃、炮制时间19 min、闷润时间6.9 h、投药量32 g/L。验证试验测得OD均值为0.655 1，RSD为2.97%，与模型预测值（0.653 3）接近。基于ANN预测与Pareto非支配解集界定的鲁棒参数区间为炮制温度140～150 ℃、炮制时间18～22 min、闷润时间6.5～7.5 h、投药量30～35 g/L；模型提示在该区间内参数波动±5%时，OD可保持相对稳定。结论 构建的BBD-ANN-Pareto耦合模型具有良好的预测能力和稳定性，可为酒香附炮制工艺优化及质量一致性控制提供参考。

Objective To establish a multi-index weighted optimization strategy for wine-processed Xiangfu (Cyperi Rhizoma) by integrating CRITIC weighting, Box-Behnken design-response surface methodology (BBD-RSM), and an artificial neural network (ANN), and define a robust operating space based on model prediction, thereby providing a reference for process parameter fixation and quality consistency control. Methods The contents of cyperotundone, α-cyperone, total flavonoids, and total volatile oil were used as comprehensive evaluation indices. Objective weights were assigned using the CRITIC method, and an overall desirability (OD) value was calculated. Based on single-factor experiments, a four-factor, three-level Box-Behnken design (BBD) was conducted with processing temperature, processing time, moistening time, and dosage as independent variables, and a regression model was established. An ANN prediction model was then developed, and the Garson algorithm was applied to interpret the relative contributions of factors. In addition, the Pareto non-dominated solution set was used to screen the process-parameter space to obtain a robust operating interval (model-predicted). Results The CRITIC-derived weights for cyperotundone, α-cyperone, total volatile oil, and total flavonoids were 0.274 9, 0.255 4, 0.253 4, and 0.216 3, respectively. The optimal processing conditions were determined as follows: processing temperature 140 ℃, processing time 19 min, moistening time 6.9 h, and dosage 32 g/L. The OD obtained from validation experiments was 0.655 1, RSD was 2.97%, which was close to the predicted value (0.653 3). The robust parameter interval defined based on ANN prediction and the Pareto non-dominated solution set was as follows: processing temperature 140—150 ℃, processing time 18—22 min, moistening time 6.5—7.5 h, and dosage 30—35 g/L. The model suggested that OD could remain relatively stable under ± 5% parameter fluctuations within this interval. Conclusion The constructed BBD-ANN-Pareto coupling model demonstrated good predictive ability and stability, which may provide a reference for optimizing the processing technology and ensuring the quality consistency of wine-processed Cyperi Rhizoma.

R283.6

河北省重点研发计划项目（23372503D）；河北省自然科学基金中医药联合基金重点项目（H2025423029）；国家中医药管理局科研项目（gzy-kjs-2023-029）；国家中医药管理局科研项目（gzy-kjs-2025-015）；河北省省级科技计划项目（252W2501D）；河北中医药大学研究生创新资助项目（XCXZZSS2025019）