神经网络法研究圆叶葡萄挥发性成分保留指数  

Research on the Chromatographic Retention Index of the Volatile Flavoring Compounds in Vitis rotundifolia by Neural Network

作　　者：堵锡华[1] 

机构地区：[1]徐州工程学院化学化工学院,江苏徐州221111

出　　处：《酿酒科技》2015年第7期14-16,20,共4页Liquor-Making Science & Technology

基　　金：国家自然科学基金项目(No.21472071)

摘　　要：为了研究圆叶葡萄挥发性组分的香气特征,基于分子结构和邻接矩阵,计算了49个圆叶葡萄挥发性成分的分子形状指数、电性拓扑状态指数和电性距离矢量,建立了这些有机化合物包括醇类、醛类、酯类、酸类、酮类等分子的色谱保留指数与4K、I8、I13、M9、M155种指数的定量结构-保留相关性(QSRR)模型。将上述5种结构参数作为神经网络的输入神经元,采用5∶4∶1的网络结构,利用BP算法获得了令人较为满意的QSRR预测模型,模型的总相关系数R为0.967,利用模型得到的色谱保留指数预测值与实验值的相对平均误差为4.87%,两者吻合度较好。结果表明,圆叶葡萄挥发性成分的保留指数与5种结构参数之间有良好的非线性关系,模型能较好解释圆叶葡萄挥发性组分香味性质的递变规律。In order to study the flavoring characteristics of the volatile components in Vitis rotundifolia, we calculated the molecular shape index, electrotopological state index and electronegativity distance vector of 49 volatile components in Vitis rotundifolia based on the location of molecular structure and conjugation matrix. Then we developed the quantitative structure-retention relationship （QSRR） between the chromatographic retention index and 4K, I8, 113, Mg, M15 of 49 organic compounds including alcohols, aldehydes, esters, acids and ketone. And the five structural parameters were used as the input neurons of the neural network, we constructed satisfying QSRR models with back-propagation algorithm, the network structure of which was 5：4：1. The total correlation coefficient R of the model was 0.967. The relative mean deviation between the experimental and the predicted values of RI was 4.87 %. The results showed that there was good nonlinear relationship between RI and the five structural parameters, and the model could well describe the changing rule of aroma properties for the volatile components in Vitis rotundifolia.

关 键 词：圆叶葡萄 色谱保留指数 神经网络 分子形状指数 电性拓扑状态指数 葡萄酒 

分 类 号：S663.1[农业科学—果树学] TP183[农业科学—园艺学]