近红外光谱的不同牌号聚乳酸识别方法

塑料牌号是塑料生产公司根据原料性质、用途的差异而内部制定的编号。通过检测材料的物理化学性能能间接识别其牌号，但速度慢且具有破坏性。因此，利用了近红外光谱(near infrared spectroscopy, NIR)技术对不同牌号的聚乳酸(polylactic acid), PLA)进行识别。采用主成分分析法(principle component analysis, PCA)分别与马氏距离(mahalanobis distance, MD)、人工神经网络(artificial neural network, ANN)和支持向量机(support vector machine, SVM)结合的模型进行分析预测。在900～1 700 nm的波长范围，采用三种不同牌号的聚乳酸共90个样本的光谱进行建模，另取这3种牌号共90个样本进行识别，比较三种预测模型对PLA牌号的识别能力。结果表明，在对样品的光谱数据做主成分分析后，以验证集的前两个主成分做散点图，发现明显的聚类现象，以前9个主成分得分作为输入变量所建立的马氏距离判别、人工神经网络判别、支持向量机判别均能够对不同牌号的聚乳酸有效识别。最好的判别方法——马氏距离判别正确率能够达到98.9%。因此，近红外光谱能够对不同牌号的PLA进行无损、快速、准确的识别。

Discrimination of Different Types of PLA with Near Infrared Spectroscopy

The type of plastics is the serial number that manufacturing companies formulated based on thenature and application of raw materials. Detecting the physical and chemical properties of plasticscan indirectly identify their types, but these test methods are time-consuming and destructive. In this work, near-infrared spectroscopy technology was used to identify different types of Poly(lactic acid)(PLA). In addition, three models，PCA-MD，PCA-ANN and PCA-SVM, were applied for the analysis and prediction of the sample. In the wavelength range of 900～1 700 nm, a total of 90 samples of three different types of PLA were used to establish the model and another 90 samples of these three types of PLA were taken for prediction and identification. Comparing the identification ability of three prediction models to PLA types, we can find that the scatter plot of the first two principal components scores of the validation set had an obvious clustering phenomenon after the PCA of the spectral data. The first nine principal component scores were taken as the input variables of Mahalanobis distance, ANN and SVM discriminants, and these discriminants effectively identified the type of PLA, among which the accuracy of the best discriminant——Mahalanobis distance can reach 98.9%. Therefore, near infrared spectroscopy can be used for nondestructive, fast and accurate identification of different types of PLA.