• Spectroscopy and Spectral Analysis
  • Vol. 41, Issue 9, 2759 (2021)
Yuan FANG, Zhang-ping HE, Shi-chao ZHU, Xian-rong LIANG, and Gang JIN*;
Author Affiliations
  • National Engineering Research Center of Novel Equipment for Polymer Processing, Guangdong Provincial Key Laboratory of Technique and Equipment for Macromolecular Advanced Manufacturing, Key Laboratory of Polymer Processing Engineering of Ministry of Education, South China University of Technology, Guangzhou 510640, China
  • show less
    DOI: 10.3964/j.issn.1000-0593(2021)09-2759-05 Cite this Article
    Yuan FANG, Zhang-ping HE, Shi-chao ZHU, Xian-rong LIANG, Gang JIN. In-Line Identification of Different Grades of GPPS Based on Near-Infrared Spectroscopy[J]. Spectroscopy and Spectral Analysis, 2021, 41(9): 2759 Copy Citation Text show less
    Schematic diagram of in-line near-infrared spectral measurement system
    Fig. 1. Schematic diagram of in-line near-infrared spectral measurement system
    (a) In-line NIR spectra of different grades of GPPS, (b) Changes in characteristic peaks of in-line NIR spectra of different grades of GPPS
    Fig. 2. (a) In-line NIR spectra of different grades of GPPS, (b) Changes in characteristic peaks of in-line NIR spectra of different grades of GPPS
    Explained variance contribution of principal components of in-line NIR spectra
    Fig. 3. Explained variance contribution of principal components of in-line NIR spectra
    K-means clustering result of in-line NIR spectra of different grades of GPPS
    Fig. 4. K-means clustering result of in-line NIR spectra of different grades of GPPS
    Relationship between PRESS and the number of principal components
    Fig. 5. Relationship between PRESS and the number of principal components
    Identification results for different grades of GPPS in validation set(a): PLS-DA model; (b): RF model
    Fig. 6. Identification results for different grades of GPPS in validation set
    (a): PLS-DA model; (b): RF model
    Relationship between maximum depth of CART and classification accuracy of different grades of GPPS
    Fig. 7. Relationship between maximum depth of CART and classification accuracy of different grades of GPPS
    分类
    变量
    牌号验证集
    个数
    识别正确
    个数
    识别正确
    率/%
    1158K503876
    25250503876
    35255050100
    4PG-335050100
    5GP-1505050100
    Table 1. Identification results of different grades of GPPS based on the PLS-DA algorithm
    分类
    变量
    牌号验证集
    个数
    识别正确
    个数
    识别正确
    率/%
    1158K5050100
    25250504692
    35255050100
    4PG-335050100
    5GP-150504386
    Table 2. Identification results of different grades of GPPS based on the RF algorithm
    Yuan FANG, Zhang-ping HE, Shi-chao ZHU, Xian-rong LIANG, Gang JIN. In-Line Identification of Different Grades of GPPS Based on Near-Infrared Spectroscopy[J]. Spectroscopy and Spectral Analysis, 2021, 41(9): 2759
    Download Citation