|
|
|
|
|
|
| Comparison Study of Four Transfer Learning Architectures for Degree of Polymerization Assessment of Insulating Paper Across Different
Instruments |
| LI Han, SUN Wei-zhe, CHEN Xi-yuan, ZHANG Guan-jun, LI Yuan* |
School of Electrical Engineering, Xi'an Jiaotong University, Xi'an 710049, China
|
|
|
|
|
Abstract Near-infrared spectroscopy (NIRS) has become an important nondestructive technique for assessing the degree of polymerization (DP) of insulating paper as an alternative to traditional chemical methods. As a typical data-driven chemometric approach, NIRS often suffers from amplitude shifts in spectral data collected across different instruments, which severely limit its generalizability and large-scale deployment in engineering applications. In this work, we construct a model transfer scenario involving four representative spectrometers to analyze inter-instrument spectral discrepancies. We systematically compare the transfer learning fine-tuning performance of four mainstream deep neural network architectures under different parameter configurations, investigate the effect of layer-freezing strategies on transfer performance, and identify the optimal network structure and hyperparameter combination within a predefined search space. The results show that significant spectral amplitude differences exist among instruments, leading to drastic degradation of predictive performance when a source-domain model is directly applied to the target domain, rendering the model nearly ineffective. Freezing strategies improve fine-tuning performance by stabilizing the network; specifically, freezing the front-end feature extraction layers while fine-tuning the higher-level decision layers enhances transferability without compromising stability. Among the four tested architectures—MLP, 1D-CNN, EOT, and ResNet—EOT achieved the lowest error in the source domain but performed worse after fine-tuning in the target domain, whereas ResNet exhibited higher source-domain error than EOT but achieved better fine-tuning performance. This indicates that source-domain training error is not strongly correlated with transfer effectiveness. Overall, a three-branch ResNet network incorporating multi-scale Inception modules achieved the best target-domain performance, with an RMSE of 78.5 and a MAPE of 8.6% after fine-tuning, significantly outperforming the other models. These findings provide theoretical support for constructing NIRS modeling frameworks with cross-instrument generalizability.
|
|
Received: 2025-07-09
Accepted: 2025-08-27
|
|
|
|
Corresponding Authors:
LI Yuan
E-mail: liyuan8490@xjtu.edu.cn
|
|
[1] Li Y, Li H, Zhang W, et al. IEEE Transactions on Industry Applications, 2023, 59(3): 2750.
[2] National Energy Administration(国家能源局). DL/T 2410—2021 Determination Method of Degree of Polymerization of Insulating Paper (Pressboards) in Power Transformer (Near Infrared Spectroscopy Method)[DL/T 2410—2021 变压器绝缘纸(板)聚合度测定法(近红外光谱法)], 2022.
[3] Wang H P, Chen P, Dai J W, et al. TrAC Trends in Analytical Chemistry, 2022, 153: 116648.
[4] Mishra P, Klont R, Verklerj T, et al. Infrared Physics & Technology, 2021, 119: 103918.
[5] CHU Xiao-li，LU Wan-zhen (诸小立，陆婉珍). Spectroscopy and Spectral Analysis(光谱学与光谱分析), 2014，34(10)：2595.
[6] Folch-Fortuny A, Vitale R, de Denoord O E, et al. Journal of Chemometrics, 2017, 31(3): e2874.
[7] Mishra P, Nikzad-Langerodi R, Marini F, et al. Trends in Analytical Chemistry, 2021, 143: 116331.
[8] Mishra P, Nikzad-Langerodi, Marini F, et al. TrAC Trends in Analytical Chemistry, 2021, 143: 116331.
[9] Mishra P, Passos D. Infrared Physics & Technology, 2021, 117: 103863.
[10] Padarian J, Minasny B, Mcbratney A B. Geoderma, 2019, 340: 279.
[11] DENG Di, LI Hao, MU Hong-bo(邓 迪，李 昊，牟洪波). Technology Innovation and Application(科技创新与应用), 2025, 15(14): 76.
[12] WU Yun-fei, LUAN Xiao-li, LIU Fei(邬云飞, 栾小丽, 刘 飞). CIESC Journal(化工学报), 2022, 73(2): 782.
[13] Zhang Z, Li Y, Li Y. Frontiers in Plant Science, 2022, 13: 1006292.
[14] Mishra P, Passos D. Chemometrics and Intelligent Laboratory Systems, 2021, 212: 104283.
[15] Wang Y, Yin K, Hu B, et al. Geoderma, 2025, 456: 117257.
[16] LI Chang-sheng, GAO Shu-hui(李昌盛，高树辉). Journal of Instrumental Analysis(分析测试学报), 2025, 44(5): 781.
[17] JU Wei-liang, YANG Wei, SONG Ya-mei, et al(剧伟良, 杨 玮, 宋亚美, 等). Transactions of the Chinese Society for Agricultural Machinery(农业机械学报), 2025, 56(6): 268.
[18] Yu Y, Huang J, Wang L, et al. Food Control, 2025, 167: 110823.
[19] El Idrissi El-Bouzaiei Y, Hibbi F Z, Abdoun O. Optimizing Convolutional Neural Network Impact of Hyperparameter Tuning and Transfer Learning, in: Innovations in Optimization and Machine Learning (Pages 301-326), IGI Global, 2025.
[20] Lelekakis N, Martin D, Wijaya J. IEEE Transactions on Dielectrics and Electrical Insulation, 2012, 19(6): 2009.
|
| [1] |
ZHANG Ying1, 2, ZHANG Chi1, 2, SHI Jian-bo1, 2, LIU Si-si3*. Advances in the Application of Machine Learning to the Spectrum
Detection of Gases[J]. SPECTROSCOPY AND SPECTRAL ANALYSIS, 2025, 45(11): 3001-3010. |
| [2] |
ZHANG Sheng-jie1, ZHAI Chen2*, XING Wei-hai2, FENG Xiao-hui2, YANG Ying-kang2, YANG You-you2, SHI Chao3, YANG Zhou2, WANG Wen-xiu1*. Development of Predictive Models for Fatty Acids Content in Multiple Pork Muscle Sites Using Spectroscopic Techniques[J]. SPECTROSCOPY AND SPECTRAL ANALYSIS, 2025, 45(11): 3122-3129. |
| [3] |
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[J]. SPECTROSCOPY AND SPECTRAL ANALYSIS, 2025, 45(11): 3153-3159. |
| [4] |
ZHANG Shuo1, 2, 3, LI Yu1, JIANG Tong1*, HUANG Jian-han1, HUANG Yin-wei1, 4, TIAN Jing-chun3, SHAN Mao-yu1, LI Pei-yao1. SWIR Spectral Characteristics and Mineral Phase Transformations: Multi-Indicator Synergistic Response for Granite Weathering Classification[J]. SPECTROSCOPY AND SPECTRAL ANALYSIS, 2025, 45(11): 3235-3245. |
| [5] |
ZHANG Xu1, XIE Zhuo-jun1, QIN Zi-quan1, ZHAO Rui-jie1, LIU Wen-zheng1, BAI Xue-bing1, XIONG Xiao-lin2*, LIU Xu1. Research Progress of Miniaturized Vis/NIR Spectrometers in Leaf and Fruit Nondestructive Detection[J]. SPECTROSCOPY AND SPECTRAL ANALYSIS, 2025, 45(10): 2720-2729. |
| [6] |
SUN Xu-dong1, LONG Tao1, WANG Jia-hua2, FENG Shao-ran3, ZENG Ti-wei4, XIE Dong-fu1, FU Wei1. Near-Infrared Spectroscopy Prediction of the Optimal Harvest Date for Autumn Moon Pear: Considering the Correction of Ambient Light Changes and Inter-Instrument Differences[J]. SPECTROSCOPY AND SPECTRAL ANALYSIS, 2025, 45(10): 2774-2782. |
| [7] |
ZHAO Xiao-yan1*, ZHENG Shao-wen1, WU Xian-hao1, SUN Zhi-yan2, 3, TAO Rui2, 3, ZHANG Tian-yao1, YUAN Yuan1, LIU Xing4, ZHOU Da-biao2, 3, ZHANG Zhao-hui1, YANG Pei2, 3*. Prediction of EGFR Amplification Status of Glioma Based on Terahertz Spectral Data With Convolutional Neural Networks[J]. SPECTROSCOPY AND SPECTRAL ANALYSIS, 2025, 45(10): 2856-2862. |
| [8] |
ZHANG Xin-zhi1, SAIMAITI Patiguli1, ZHANG Lu-wen1, YANG Ya-fei2*, BIAN Xi-hui1, 3*. Research Progress on Geographical Origin Discrimination of Traditional Chinese Medicines Based on Near-Infrared Spectroscopy[J]. SPECTROSCOPY AND SPECTRAL ANALYSIS, 2025, 45(09): 2410-2416. |
| [9] |
ZHANG Xue-li1, 2, YANG Hao1, 2, LI Chen-fei1, 2, SUN Yi-le1, 2, LIU Zong-lin1, 2, ZHENG De-cong1, 2, SONG Hai-yan1, 2*. Origin Discrimination and Soluble Protein Content Prediction of Dried Daylily Based on Near Infrared Spectroscopy[J]. SPECTROSCOPY AND SPECTRAL ANALYSIS, 2025, 45(09): 2491-2495. |
| [10] |
BU Zi-chuan, LIU Ji-hong*, REN Kai-li, LIU Chi, ZHANG Jia-geng, YAN Xue-wen. Raman Spectral Sample Data Enhancement Method Based on Voigt
Function[J]. SPECTROSCOPY AND SPECTRAL ANALYSIS, 2025, 45(09): 2502-2510. |
| [11] |
ZHU Li-wei, DU Qian-xi, TANG Guo-hong, LI Hong-you, ZHANG Xiao-na, CHEN Qing-fu, SHI Tao-xiong*. Near-Infrared Modeling for Total Flavonoid and Protein Contents in
Buckwheat Leaves Based on CARS Feature Extraction[J]. SPECTROSCOPY AND SPECTRAL ANALYSIS, 2025, 45(09): 2585-2589. |
| [12] |
HU Wei-cheng, LI Yun-peng*, WANG Hong-wei, DAI Xue-jing, WANG Hua-peng. Hyperspectral Image Visualization and Recognition of Bloodstains and Their Analogues Based on Attention Mechanism[J]. SPECTROSCOPY AND SPECTRAL ANALYSIS, 2025, 45(09): 2625-2631. |
| [13] |
FAN Xiao-yong1, LI Heng-kai1*, LIU Kun-ming2, WANG Xiu-li3, YU Yang1, LI Xiao-yu1. A Multi-Layer Attention Convolutional Neural Network Model for Fine Classification of Hyperspectral Images in Rare Earth Mining Areas[J]. SPECTROSCOPY AND SPECTRAL ANALYSIS, 2025, 45(09): 2666-2675. |
| [14] |
ZOU Liang1, KOU Shao-ping1, REN Ke-long1, YUAN Guang-fu2, XU Zhi-bin3, XU Shi-fan1, WU Jing-tao4*. A Collaborative Detection Method for Coal Ash Content and Calorific
Value Based on A-Unet3+ and Portable NIR Spectrometer[J]. SPECTROSCOPY AND SPECTRAL ANALYSIS, 2025, 45(08): 2210-2217. |
| [15] |
HUANG Tian-yu1, YANG Hui-hua1, 2, LI Ling-qiao1*. InvertResNet: A Qualitative Analysis Method for Drug Products Based on Deep Learning and Near-Infrared Spectroscopy[J]. SPECTROSCOPY AND SPECTRAL ANALYSIS, 2025, 45(08): 2218-2227. |
|
|
|
|
