基于BP神经网络和遗传算法优化莪术超临界萃取工艺

 目的以莪术醇含量为响应指标,用BP神经网络和遗传算法优化莪术有效成分的超临界CO₂萃取工艺。方法采用气相色谱法测定莪术醇的含量,建立神经模型,通过均匀试验设计,利用遗传算法(GA)对网络模型进行优化,并对优化后的网络进行寻优,获得最佳提取工艺。结果莪术的超临界CO₂最佳萃取工艺为萃取压力20 MPa,萃取温度45℃,动态萃取时间80 min,改性剂用量25 mL,夹带剂浓度72%,静态平衡时间27 min;测试样本的网络预测值和实际测量值的相对误差小于4%。结论遗传算法优化的BP网络模型可对中药药效物质基础的超临界萃取结果进行预测,GA优化的萃取工艺比常规最小二乘法的优化结果优越。

Study on Supercritical Extraction Process of Curcumol in Curcuma phaeocaulis Valeton Based on Back-Propagation Neural Network and Genetic Algorithm

OBJECTIVE To optimize the supercritical extraction process for the active components in Curcuma phaeocaulis valeton with back-propagation neural network and genetic algorithm.METHODS Gas chromatography was used to determine the contents of curcumol in the extract.BP neural network was established and optimized with genetic algorithm to forecast the supercritical extraction of Curcuma phaeocaulis valeton.Uniform design and genetic algorithm were used to optimize the trained BP network to obtain optimum SFE process.RESULTS The optimum process was established as follows:20 MPa as extracting pressure,45 ℃ as extracting temperature,80 min as dynamic extracting time and 27 min as static extracting time,25 mL of 72% alcohol as modifier.The relative error between the predicted value from BP network and observed value was lower than 4%.CONCLUSION GA-optimized BP neural network can be employed to forecast SFE extraction of active components in Chinese herb medicines.GA-optimized SFE processes are better than the process optimized by nonlinear regress.