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.
Key words：Near-infrared spectroscopy; Insulating paper; Transfer learning; Neural network; Fine-tuning
李 含，孙伟哲，陈希源，张冠军，李 元. 四种迁移学习架构在跨仪器绝缘纸聚合度评估中的效果对比[J]. 光谱学与光谱分析, 2025, 45(11): 3145-3152.
LI Han, SUN Wei-zhe, CHEN Xi-yuan, ZHANG Guan-jun, LI Yuan. Comparison Study of Four Transfer Learning Architectures for Degree of Polymerization Assessment of Insulating Paper Across Different
Instruments. SPECTROSCOPY AND SPECTRAL ANALYSIS, 2025, 45(11): 3145-3152.
[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.
