摘要
针对菜籽和大豆毛油储藏期间油罐上、中、下层毛油氧化情况的差异问题,提出一种电子鼻技术结合化学计量学对毛油氧化情况进行判别的方法,为改善毛油储藏措施提供技术参考.基于常规理化分析,通过电子鼻结合偏最小二乘判别(PLS-DA)、支持向量机(SVM)、BP神经网络(BPNN)、人工神经网络(ANN)对特征样本进行特征筛选和分类识别.结果表明,上、中、下层毛油的理化差异不显著,且中、下层呈现高度相似性,上层劣变稍快.PLS-DA和BPNN模型在大豆和菜籽毛油上层和中下层混样氧化程度的分类识别上效果不佳,而优化后的SVM和ANN模型测试集准确率分别达到了96.7%和98.3%.因此,电子鼻结合ANN或SVM可以有效识别上层和中下层的大豆和菜籽毛油.
Abstract
Because of the difference in the oxidation of the upper,middle,and lower layers of the oil tank of rapeseed and soybean crude oil during storage,an electronic nose technology combined with stoichiometry is proposed to distinguish the oxidation of crude oil,which could provide technical reference for improving the storage measures.The feature samples are screened and classified,through electronic nose combined with partial least squares discrimination(PLS-DA),support vector machine(SVM),BP neural network(BPNN),and ar-tificial neural network(ANN),based on conventional physical and chemical analysis.The results show that the physical and chemical differences of the upper,middle,lower oils are not significant,and the middle and lower layers are highly similar,and the upper layer deteriorats slightly faster.The PLS-DA and BPNN model are not effective in classifying and identifying the oxidation degree of the upper and middle-lower mixed samples of soybean and rapeseed oil,but the accuracy of optimized model test set with SVM and ANN achieves 96.7%and 98.3%,respectively.Therefore,the electronic nose combined with ANN or SVM can effectively identify the upper and mid-dle-lower layers of soybean and rapeseed crude oils.
维普
万方数据