Abstract：Fast and accurate identification of insulation material types can prevent the mixing and misuse of raw materials, which is a key part of quality control in cable production. The traditional method using Fourier Transform Infrared Spectroscopy (FTIR) for sampling inspection has several drawbacks, including high cost, low efficiency, and poor adaptability to field conditions. It is difficult to meet the need for fast and full inspection of raw materials. This paper proposes a method for fast identification of cable insulation material types by combining near-infrared spectroscopy with a one-dimensional convolutional neural network (1D-CNN). Six common cable insulation materials were used as the research objects. Their spectral data were collected using a near-infrared spectrometer, and a 1D-CNN model with two convolution-pooling units was built. The model leverages its ability to extract local features from high-dimensional near-infrared spectral data, thereby identifying spectral differences between various material types. Based on this, various spectral preprocessing methods were applied to remove unwanted interference. The modeling performance under each strategy was compared systematically, and the second derivative of the spectrum using Savitzky-Golay smoothing was found to be the best preprocessing method. Bayesian optimization was introduced to adjust key model parameters and improve recognition accuracy automatically. The optimized model achieved an identification accuracy of 95.00%, a weighted average precision of 95.07%, a weighted average recall of 95.00%, and a weighted average F1 score of 0.950 4, which are significantly better than traditional machine learning models. The results show that combining the 1D-CNN model with near-infrared spectroscopy enables fast, accurate, and non-destructive identification of cable insulation material types, with strong potential for field application. This study provides a reliable solution for the rapid screening and testing of cable insulation materials on a large scale, offering strong technical support for building an intelligent quality control system throughout the entire insulation material process.
孙伟哲，李 含，陈希源，张冠军，李 元. 电力电缆绝缘材料型号的近红外光谱快速识别方法研究[J]. 光谱学与光谱分析, 2025, 45(11): 3153-3159.
SUN Wei-zhe, LI Han, CHEN Xi-yuan, ZHANG Guan-jun, LI Yuan. Study on Spectral On-Site Rapid Identification Method of Cable Insulation Material Models. SPECTROSCOPY AND SPECTRAL ANALYSIS, 2025, 45(11): 3153-3159.
[1] DUAN Xiao-li, LIU Yu-feng, LIU San-wei, et al(段肖力，刘雨丰，刘三伟，等). Transactions of China Electrotechnical Society(电工技术学报)，2025，40(5)：1540.
[2] YANG Ya-di, DENG Xian-bo(杨亚迪，邓显波). Advanced Materials for Power Cables: More Energy-Efficient, Cost-Effective, and Low-Consumption(电力电缆用上新材料 节能降本更低耗). State Grid News(国家电网报), 2024-04-09（6）.
[3] LI Jia-ming, LIU Jian, ZHENG Jian-kang, et al. Research on the Weight of Influence Factors in 10kV Cable Network Based on Data Analysis and Information Theory. 2018 12th International Conference on the Properties and Applications of Dielectric Materials (ICPADM), Xi'an, China, 2018, 451.
[4] JIA He-feng, ZHU Xiang-hua, XIE Bo(贾贺峰，朱翔华，谢 波). China Standardization(中国标准化), 2024, (1): 178.
[5] HUI Bao-jun, FU Ming-li, LIU Tong, et al(惠宝军，傅明利，刘 通，等). Southern Power System Technology(南方电网技术)，2017，11(12): 44.
[6] ZHANG Ya-chun, LIU Shao-jiang(张亚春，刘绍江). China Plastics Industry(塑料工业)，2025，53(4)：116.
[7] ZHENG Yu, PENG Ying-jie, LIN Qian-yin, et al(郑 钰，彭颖杰，林倩茵，等). Industry Control Computer(工业控制计算机)，2025，28(5)：92.
[8] TANG Jie, LUO Yan-bo, LI Xiang-yu, et al(唐 杰，罗彦波，李翔宇，等). Spectroscopy and Spectral Analysis(光谱学与光谱分析), 2024, 44(3): 731.
[9] YAN Xu-mei, CHEN Chao, WANG Nan, et al(严旭梅，陈 超，王 楠，等). Journal of Materials and Metallurgy(材料与冶金学报), 2023, 22(5): 430.
[10] Bergstra J, Bengio Y. Journal of Machine Learning Research, 2012, 13: 281.
[11] CHAI Qin-qin, ZENG Jian, ZHANG Xun(柴琴琴，曾 建，张 勋). Acta Agriculturae Zhejiangensis(浙江农业学报), 2022, 34(2): 391.
[12] CUI Jia-xu, YANG Bo(崔佳旭，杨 博). Journal of Software(软件学报), 2018, 29(10): 3068.
