BP神经网络结合正交试验优化苦参方中荆芥挥发油的提取工艺

Extraction optimization for volatile oil from Schizonepetae Herba in Kushen Recipe by back propagation neural network and orthogonal design

目的使用BP神经网络结合正交试验优化苦参方(苦参和荆芥)中荆芥挥发油的提取工艺。方法采用水蒸气蒸馏法提取挥发油,以挥发油得率为评价指标,正交试验设计考察浸泡时间、提取时间、加水量对提取的影响,筛选挥发油提取工艺。正交试验实验数据作为反向传播神经网络输入,对主要影响因素进行仿真优化。结果优化得到的提取工艺条件为加入12倍量水,浸泡1 h,水蒸汽蒸馏提取6 h,检验样本的网络预测值和实际测量值的相对误差小于1%。结论 BP神经网络结合正交试验不需要增加试验次数,就能分析苦参方中荆芥挥发油提取因素变化规律和寻找最佳参数。

AIM To optimize extraction of volatile oil from Schizonepetae Herba in Kushen Recipe（ Sophorae flavescentis Radix, Schizonepetae Herba） by back propagation（ BP） neural network and orthogonal design.METHODS Steam distillation was for the extraction of volatile oil. With the yield rate of volatile oil as the index,the orthogonal design was applied to determining the time for maceration and extraction,and the water consumption as well. The data from orthogonal test served as input of BP neural network to figure out the main influence factors in the optimization. RESULTS The optimal extraction conditions were verified to be 12 volumes of water,1 h maceration,and 6 h steam distillation,which demonstrated a less than 1% difference between the predictive and detected values. CONCLUSION BP neural network with orthogonal test can be used to obtain the most optimal extraction parameters for volatile oil in Kushen Recipe with no more experiment times.