基于系统聚类-神经网络的塑料饮料瓶差分拉曼光谱检验

Differential Raman Spectral Inspection of Plastic Drink Bottles Based on Hierarchical Clustering Analysis-Neural Networks

建立塑料饮料瓶物证快速准确检验鉴别方法。利用差分拉曼光谱法检验42个塑料饮料瓶样品,优化积分时间并进行重现性检验。在40 s最优积分时间条件下采集光谱,任选41个样品作为建立模型的数据集,剩余样品作为盲样,对41个样品材质初步定性分为聚对苯二甲酸乙二醇酯(PET)和聚乙烯(PE)两类。建立基于系统聚类(HCA)、多层感知器神经网络和径向基神经网络的PET样品鉴别模型,确定最优鉴别模型及样品最佳分类。结果表明,系统聚类-多层感知器神经网络为最优鉴别模型,PET样品最佳分类为2类。差分拉曼光谱法结合系统聚类和神经网络可实现塑料饮料瓶有效鉴别。

The purpose of this study was to establish a rapid and accurate method for inspection and identification of physical evidence of plastic beverage bottles.Firstly,42 plastic drink bottle samples were tested by differential Raman spectroscopy to investigate the effects of different integration times on the spectra of the samples,and the reproducibility was tested.Secondly,under the optimal integration time of 40 s,spectra of samples were collected,and 41 samples were selected randomly as the model training set,and one sample was selected as the blind sample of the model prediction set.Qualitative analysis was made on the materials of 41 samples,which were preliminarily divided into two categories:polyethylene terephthalate(PET)and polyethylene(PE).Finally,sample identification models for PET based on hierarchical clustering analysis,multilayer perceptron neural network and radial basis function neural network was established to determine the optimal identification model and the optimal classification of samples.The results show that the identification model of hierarchical clustering analysis and multilayer perceptron neural network is the best model.The samples made of PET are best classified into 2 categories.Differential Raman spectroscopy combines with hierarchical clustering analysis and neural networks can effectively identify plastic beverage bottles.