基于神经网络的山楂红酒香气成分色谱保留值的研究

Chromatographic retention value of aroma compounds from hawthorn red wine based on neural network

为研究山楂红酒中香气成分的性质,预测其色谱保留时间,采用MATLAB软件有关自编程序,运算得到了山楂红酒中香气成分的分子价连接指数(^(m)X_t^(V))和电拓扑状态指数(E_(i))。通过优化筛选,确定了分子价连接性指数的(0)^X_(p)^(V),(1)^X_(p)^(v)和4^X_(p)^(v)、电拓扑状态指数的E_(1)和E_(12)共5个参数,并建立了山楂红酒中香气成分色谱保留时间的定量构效关系模型,其相关系数为0.979,模型具有较强的稳定性和预测能力。将这5个参数作为反向传播算法(BP)神经网络的输入层变量,香气成分的色谱保留时间作为输出层变量,采用5∶5∶1的网络结构,获得了令人满意的神经网络预测模型,模型总相关系数为0.998,得到的色谱保留时间的预测值与实验值颇为吻合,相对平均误差为3.31%。结果表明,山楂红酒中香气成分的色谱保留时间与5种结构参数之间呈现良好的非线性关系,模型较好地揭示了香气成分色谱保留时间的递变规律。

In order to study the properties of aroma components in hawthorn red wine and predict their chromatographic retention time,the molecular valence connectivity index(^(m)X_t^(V))and electrotopological state index(E_(i))of aroma compounds from hawthorn red wine were operated by self-designed programs MATLAB.The molecular valence connectivity indexes(0)^X_(p)^(V),(1)^X_(p)^(v),4^X_(p)^(v),and electrotopological state indexes E_(1),E_(12) were determined after optimization and screening,a quantitative structure-retention relationship model of chromatographic retention time of aroma compounds from hawthorn red wine was established,the coefficient index was 0.979,and the model had strong stability and predictive ability.The five parameters were used as input variables of back propagation(BP)algorithm neural network and the chromatographic retention time was used as output variable,by using 5:5:1 network structure,a satisfactory neural network prediction model was obtained,and the total correlation coefficient of the model was 0.998.The predicted values by the model were in agreement with those of the experiment values,and the average relative error was 3.31%.The results showed that there was good nonlinear relationship between the chromatographic retention time and the five molecular structure parameters.The model could better elucidate the changing rule of chromatography retention time of aroma compounds.