A Dynamic-Weighted Deep Transfer Learning Framework for Thermal Conductivity Prediction and Analysis

Abstract

To overcome the limitations of small-sample data in establishing microstructure–property linkages, this study introduces a deep transfer learning framework with a dynamic weighting mechanism. By transferring the perceptual capabilities of the ResNet-18 model and adaptively fusing them with material domain knowledge, the model effectively captures complex features such as phase distribution. Using silicon nitride ceramics as the primary research object, the framework achieves an average cross-validation prediction accuracy (R²) of 0.73, representing a 104.3% relative improvement compared to the traditional CNN framework, and the optimal model reaches an accuracy of R² = 0.89. Furthermore, this framework also demonstrates exceptional predictive accuracy on silicon carbide ceramics (R² = 0.84) and sintered nano silver (R² = 0.93), indicating its strong generalization capabilities. By employing multilevel gradient-weighted class activation mapping (Grad-CAM) and sliding occlusion analysis, the decision-making process of the model is elucidated, thereby validating the logical soundness of its predictions. Additionally, symbolic regression is utilized to identify the influence of different microstructural features on thermal conductivity and to establish their quantitative relationships. This research holds broad application prospects in the rapid development and design of thermal management materials, analysis of material microstructure images, and the establishment of structure–performance relationships between microstructural features and macroscopic properties.

Keywords

deep transfer learning / dynamic weighting mechanisms / microstructure-property relationships / symbolic regression / thermal management materials

References

[1] H. Y. He, W. X. Peng, J. B. Liu, et al., “Microstructured BN Composites With Internally Designed High Thermal Conductivity Paths for 3D Electronic Packaging,” Advanced Materials 34, no. 38 (2022): 2205120, https://doi.org/10.1002/adma.202205120.

[2] H. Tang, Y. Tang, Z. P. Wan, et al., “Review of Applications and Developments of Ultra-Thin Micro Heat Pipes for Electronic Cooling,” Applied Energy 223 (2018): 383–400, https://doi.org/10.1016/j.apenergy.2018.04.072.

[3] C. R. Eddy and D. K. Gaskill, “Silicon Carbide as a Platform for Power Electronics,” Science 324, no. 5933 (2009): 1398–1400, https://doi.org/10.1126/science.1168704.

[4] K. R , A. Srivastava, S. Midya, et al., “Enhanced Heat Dissipation and Reduced Power Consumption in Electronics Using 2D Hexagonal Boron Nitride,” Small 21, no. 32 (2025): 2502007, https://doi.org/10.1002/smll.202502007.

[5] M. Li, S. X. Li, Z. H. Zhang, C. J. Su, B. Wong, and Y. J. Hu, “Advancing Thermal Management Technology for Power Semiconductors Through Materials and Interface Engineering,” Accounts of Materials Research 6, no. 5 (2025): 563–576, https://doi.org/10.1021/accountsmr.4c00349.

[6] G. L. Guo, Y. J. Liu, Y. F. Ding, et al., “Flexible Yet Impermeable Composites With Wrinkle Structured BNNSs Assembling for High-Performance Thermal Management,” npj Flexible Electron 8, no. 1 (2024): 34, https://doi.org/10.1038/s41528-024-00320-4.

[7] P. J. Zhang, X. Ming, Y. J. Liu, et al., “An Improved Thermal Conductivity Measurement Scheme for Macroscopic Graphitic Films Using the Laser Flash Method,” Journal of Thermal Science 33, no. 4 (2024): 1480–1490, https://doi.org/10.1007/s11630-024-1997-x.

[8] S. K. Kim and Y. J. Kim, “Determination of Apparent Thickness of Graphite Coating in Flash Method,” Thermochimica Acta 468, no. 1–2 (2008): 6–9, https://doi.org/10.1016/j.tca.2007.11.012.

[9] F. Cernuschi, L. Lorenzoni, P. Bianchi, and A. Figari, “The Effects of Sample Surface Treatments on Laser Flash Thermal Diffusivity Measurements,” Infrared Physics & Technology 43, no. 3–5 (2002): 133–138, https://doi.org/10.1016/s1350-4495(02)00131-7.

[10] A. Shrivastava, R. Shukla, and P. Chaudhuri, “Effect of Porosity on Thermal Conductivity of Li₂TiO₃ Ceramic Compact,” Fusion Engineering and Design 166 (2021): 112318, https://doi.org/10.1016/j.fusengdes.2021.112318.

[11] M. Potenza, P. Coppa, S. Corasaniti, and G. Bovesecchi, “Numerical Simulation of Thermal Diffusivity Measurements With the Laser-Flash Method to Evaluate the Effective Property of Composite Materials,” Journal of Heat Transfer 143, no. 7 (2021): 72102, https://doi.org/10.1115/1.4050995.

[12] X. H. Lu, J. Liu, C. Shu, et al., “Densifying Conduction Networks of Vertically Aligned Carbon Fiber Arrays With Secondary Graphene Networks for Highly Thermally Conductive Polymer Composites,” Advanced Functional Materials 35, no. 11 (2025): 2417324, https://doi.org/10.1002/adfm.202417324.

[13] J. Zhang, C. Liu, X. Li, et al., “Enhance Thermal Conductivity via Constructing a Low Heat Resistance Interface With Quaternized Graphene,” Composites Communications 45 (2024): 101800, https://doi.org/10.1016/j.coco.2023.101800.

[14] Y. Zhou, H. Hyuga, D. Kusano, Y. Yoshizawa, and K. Hirao, “A Tough Silicon Nitride Ceramic With High Thermal Conductivity,” Advanced Materials 23, no. 39 (2011): 4563–4567, https://doi.org/10.1002/adma.201102462.

[15] X. W. Zhu, Y. Sakka, Y. Zhou, K. Hirao, and K. Itatani, “A Strategy for Fabricating Textured Silicon Nitride With Enhanced Thermal Conductivity,” Journal of the European Ceramic Society 34, no. 10 (2014): 2585–2589, https://doi.org/10.1016/j.jeurceramsoc.2014.01.025.

[16] A. Merchant, S. Batzner, S. S. Schoenholz, M. Aykol, G. Cheon, and E. D. Cubuk, “Scaling Deep Learning for Materials Discovery,” Nature 624, no. 7990 (2023): 80–85, https://doi.org/10.1038/s41586-023-06735-9.

[17] C. B. Zhang and X. Chen, “FerroAI: A Deep Learning Model for Predicting Phase Diagrams of Ferroelectric Materials,” npj Computational Materials 11, no. 1 (2025): 282, https://doi.org/10.1038/s41524-025-01778-0.

[18] H. Xiao, R. Li, X. Y. Shi, et al., “An Invertible, Invariant Crystal Representation for Inverse Design of Solid-State Materials Using Generative Deep Learning,” Nature Communications 14, no. 1 (2024): 7027, https://doi.org/10.1038/s41467-023-42870-7.

[19] T. S. Vu, M. Q. Ha, D. N. Nguyen, et al., “Towards Understanding Structure–property Relations in Materials With Interpretable Deep Learning,” npj Computational Materials 9, no. 1 (2023): 215, https://doi.org/10.1038/s41524-023-01163-9.

[20] E. M. D. Siriwardane, Y. Zhao, I. Perera, and J. J. Hu, “Generative Design of Stable Semiconductor Materials Using Deep Learning and Density Functional Theory,” npj Computational Materials 8, no. 1 (2022): 164, https://doi.org/10.1038/s41524-022-00850-3.

[21] W. Q. Huang, Y. C. Jin, Z. M. Li, et al., “Auto-Resolving the Atomic Structure at Van der Waals Interfaces Using a Generative Model,” Nature Communications 16, no. 1 (2025): 2927, https://doi.org/10.1038/s41467-025-58160-3.

[22] C. H. Fan, H. Luo, Q. C. Zhao, X. F. Wang, H. X. Cheng, and Y. Chang, “Deep Learning Accelerated the Discovery of Multi-Principal Element Alloys With Various Strength-Toughness Trade-Offs,” Materials Genome Engineering Advances 3, no. 3 (2025): e70020, https://doi.org/10.1002/mgea.70020.

[23] S. Y. Han, C. C. Wang, Y. Zhang, W. Xu, and H. S. Di, “Employing Deep Learning in Non-Parametric Inverse Visualization of Elastic–Plastic Mechanisms in Dual-Phase Steels,” Materials Genome Engineering Advances 2, no. 1 (2024): e29, https://doi.org/10.1002/mgea.29.

[24] D. Yan, A. D. Smith, and C. C. Chen, “Structure Prediction and Materials Design With Generative Neural Networks,” Nature Computational Science 3, no. 7 (2023): 572–574, https://doi.org/10.1038/s43588-023-00471-w.

[25] S. Gong, K. Q. Yan, T. Xie, et al., “Examining Graph Neural Networks for Crystal Structures: Limitations and Opportunities for Capturing Periodicity,” Science Advances 9, no. 45 (2023): eadi3245, https://doi.org/10.1126/sciadv.adi3245.

[26] Y. X. Zhang, J. L. Zhou, N. Wang, et al., “Spatiotemporal Prediction and Characterization of Microstructure Evolution During In-Situ Heat Treatment Using a Deep Learning Algorithm,” Materials Today Nano 26 (2024): 100471, https://doi.org/10.1016/j.mtnano.2024.100471.

[27] B. T. Zong, J. S. Li, P. Wang, W. J. Liao, T. Lookman, and R. H. Yuan, “Spatiotemporal Microstructure Evolution During Martensitic Transformation in Titanium Alloys Using Deep Learning,” Acta Materialia 301 (2025): 121603, https://doi.org/10.1016/j.actamat.2025.121603.

[28] P. Wu, H. W. Huang, Q. C. Yang, et al., “SimGate: A Deep Learning Surrogate Model for Predicting Microstructure Evolution Using the Phase-Field Method,” Computational Materials Science 256 (2025): 113883, https://doi.org/10.1016/j.commatsci.2025.113883.

[29] D. Jha, K. Choudhary, F. Tavazza, et al., “Enhancing Materials Property Prediction by Leveraging Computational and Experimental Data Using Deep Transfer Learning,” Nature Communications 10, no. 1 (2019): 5316, https://doi.org/10.1038/s41467-019-13297-w.

[30] S. Muroga, Y. Miki, and K. Hata, “A Comprehensive and Versatile Multimodal Deep-Learning Approach for Predicting Diverse Properties of Advanced Materials,” Advanced Science 10, no. 24 (2023): 2302508, https://doi.org/10.1002/advs.202302508.

[31] I. P. Malashin, D. Martysyuk, V. Nelyub, A. Borodulin, A. Gantimurov, and V. Tynchenko, “Deep Learning for Property Prediction of Natural Fiber Polymer Composites,” Scientific Reports 15, no. 1 (2025): 27837, https://doi.org/10.1038/s41598-025-10841-1.

[32] X. Jiang, H. D. Fu, Y. Bai, et al., “Interpretable Machine Learning Applications: A Promising Prospect of AI for Materials,” Advanced Functional Materials 35, no. 41 (2025): 2507734, https://doi.org/10.1002/adfm.202507734.

[33] Y. Liu, K. P. Kelley, R. K. Vasudevan, H. Funakubo, M. A. Ziatdinov, and S. V. Kalinin, “Experimental Discovery of Structure–Property Relationships in Ferroelectric Materials via Active Learning,” Nature Machine Intelligence 4, no. 4 (2022): 341–350, https://doi.org/10.1038/s42256-022-00460-0.

[34] A. Bhaduri, A. Gupta, and L. Graham-Brady, “Stress Field Prediction in Fiber-Reinforced Composite Materials Using a Deep Learning Approach,” Composites Part B: Engineering 238 (2022): 109879, https://doi.org/10.1016/j.compositesb.2022.109879.

[35] K. Choudhary, B. DeCost, C. Chen, et al., “Recent Advances and Applications of Deep Learning Methods in Materials Science,” npj Computational Materials 8, no. 1 (2022): 59, https://doi.org/10.1038/s41524-022-00734-6.

[36] D. Ren, C. C. Wang, X. L. Wei, Q. Q. Lai, and W. Xu, “Building a Quantitative Composition-Microstructure-Property Relationship of Dual-Phase Steels via Multimodal Data Mining,” Acta Materialia 252 (2023): 118954, https://doi.org/10.1016/j.actamat.2023.118954.

[37] P. C. Xu, X. B. Ji, M. J. Li, and W. C. Lu, “Small Data Machine Learning in Materials Science,” npj Computational Materials 9, no. 1 (2023): 42, https://doi.org/10.1038/s41524-023-01000-z.

[38] K. M. Li, D. Persaud, K. Choudhary, B. DeCost, M. Greenwood, and J. Hattrick-Simpers, “Exploiting Redundancy in Large Materials Datasets for Efficient Machine Learning With Less Data,” Nature Communications 14, no. 1 (2023): 7283, https://doi.org/10.1038/s41467-023-42992-y.

[39] R. Chang, Y. X. Wang, and E. Ertekin, “Towards Overcoming Data Scarcity in Materials Science: Unifying Models and Datasets With a Mixture of Experts Framework,” npj Computational Materials 8, no. 1 (2022): 242, https://doi.org/10.1038/s41524-022-00929-x.

[40] Y. Y. Yu, J. Xiong, X. Wu, and Q. Qian, “From Small Data Modeling to Large Language Model Screening: A Dual-Strategy Framework for Materials Intelligent Design,” Advanced Science 11, no. 45 (2024): 2403548, https://doi.org/10.1002/advs.202403548.

[41] M. Fathidoost, Y. Yang, N. Thor, et al., “Thermal Conductivity Analysis of Polymer-Derived Nanocomposite via Image-Based Structure Reconstruction, Computational Homogenization, and Machine Learning,” Advanced Engineering Materials 26, no. 17 (2024): 2302021, https://doi.org/10.1002/adem.202302021.

[42] M. Stiefel, M. Müller, B. I. Bachmann, M. A. Guitar, U. P. Nayak, and F. Mücklich, “Enhancing Machine Learning Classification of Microstructures: A Workflow Study on Joining Image Data and Metadata in CNN,” MRS Communications 14, no. 3 (2024): 363–371. https://doi.org/10.1557/s43579-024-00549-0.

[43] L. Alzubaidi, J. Bai, A. Al-Sabaawi, et al., “A Survey on Deep Learning Tools Dealing With Data Scarcity: Definitions, Challenges, Solutions, Tips, and Applications,” Journal of Big Data 10, no. 1 (2023): 46, https://doi.org/10.1186/s40537-023-00727-2.

[44] Y. Liu, Z. W. Yang, X. X. Zou, et al., “Data Quantity Governance for Machine Learning in Materials Science,” National Science Review 10, no. 7 (2023): 234–250, https://doi.org/10.1093/nsr/nwad125.

[45] M. Müller, M. Stiefel, B. I. Bachmann, D. Britz, and F. Mücklich, “Overview: Machine Learning for Segmentation and Classification of Complex Steel Microstructures,” Metals 14, no. 5 (2024): 553, https://doi.org/10.3390/met14050553.

[46] B. T. Zong, J. S. Li, T. H. Yuan, J. Wang, and R. H. Yuan, “Recent Progress on Machine Learning With Limited Materials Data: Using Tools From Data Science and Domain Knowledge,” Journal of Materiomics 11, no. 3 (2025): 100916, https://doi.org/10.1016/j.jmat.2024.07.002.

[47] W. J. Guo, F. Y. Li, L. Y. Wang, et al., “Accelerated Discovery of Near-Zero Ablation Ultra-High Temperature Ceramics via GAN-Enhanced Directionally Constrained Active Learning,” Advanced Powder Materials 4, no. 3 (2025): 100287, https://doi.org/10.1016/j.apmate.2025.100287.

[48] Y. B. Dan, Y. Zhao, X. Li, S. B. Li, M. Hu, and J. J. Hu, “Generative Adversarial Networks (GAN) Based Efficient Sampling of Chemical Composition Space for Inverse Design of Inorganic Materials,” npj Computational Materials 6, no. 1 (2020): 84, https://doi.org/10.1038/s41524-020-00352-0.

[49] B. Murgas, J. Stickel, L. Brewer, and S. Ghosh, “Modeling Complex Polycrystalline Alloys Using a Generative Adversarial Network Enabled Computational Platform,” Nature Communications 15, no. 1 (2024): 9441, https://doi.org/10.1038/s41467-024-53865-3.

[50] Y. W. Mao, Q. He, and X. H. Zhao, “Designing Complex Architectured Materials With Generative Adversarial Networks,” Science Advances 6, no. 17 (2020): eaaz4169, https://doi.org/10.1126/sciadv.aaz4169.

[51] B. Murgas, J. Stickel, and S. Ghosh, “Generative Adversarial Network (GAN) Enabled Statistically Equivalent Virtual Microstructures (SEVM) for Modeling Cold Spray Formed Bimodal Polycrystals,” npj Computational Materials 10, no. 1 (2024): 32, https://doi.org/10.1038/s41524-024-01219-4.

[52] D. Fokina, E. Muravleva, G. Ovchinnikov, and I. Oseledets, “Microstructure Synthesis Using Style-Based Generative Adversarial Networks,” Physical Review E 101, no. 4 (2020): 43308, https://doi.org/10.1103/physreve.101.043308.

[53] Y. Jiang, J. S. Li, X. Yang, and R. H. Yuan, “Applications of Generative Adversarial Networks in Materials Science,” Materials Genome Engineering Advances 2, no. 1 (2024): e30, https://doi.org/10.1002/mgea.30.

[54] X. Lyu and X. D. Ren, “Microstructure Reconstruction of 2D/3D Random Materials via Diffusion-Based Deep Generative Models,” Scientific Reports 14, no. 1 (2024): 5041, https://doi.org/10.1038/s41598-024-54861-9.

[55] E. Azqadan, H. Jahed, and A. Arami, “Predictive Microstructure Image Generation Using Denoising Diffusion Probabilistic Models,” Acta Materialia 261 (2023): 119406, https://doi.org/10.1016/j.actamat.2023.119406.

[56] I. D. Mienye and T. G. Swart, “A Comprehensive Review of Deep Learning: Architectures, Recent Advances, and Applications,” Information 15, no. 12 (2024): 755, https://doi.org/10.3390/info15120755.

[57] C. Liu, X. F. Li, J. R. Ge, et al., “A Deep Learning Framework Based on Attention Mechanism for Predicting the Mechanical Properties and Failure Mode of Embedded Wrinkle Fiber-Reinforced Composites,” Composites Part A Appl Sci Manuf 186 (2024): 108401, https://doi.org/10.1016/j.compositesa.2024.108401.

[58] M. Biswas, R. Pramanik, S. Sen, A. Sinitca, D. Kaplun, and R. Sarkar, “Microstructural Segmentation Using a Union of Attention Guided U-Net Models With Different Color Transformed Images,” Scientific Reports 13, no. 1 (2023): 5737, https://doi.org/10.1038/s41598-023-32318-9.

[59] A. Y. T. Wang, S. K. Kauwe, R. J. Murdock, and T. D. Sparks, “Compositionally Restricted Attention-Based Network for Materials Property Predictions,” npj Computational Materials 7, no. 1 (2021): 77, https://doi.org/10.1038/s41524-021-00545-1.

[60] J. M. Wang, H. J. Sun, and P. Jia, “Adaptive Kernel Convolutional Stereo Matching Recurrent Network,” Sensors 24, no. 22 (2024): 7386, https://doi.org/10.3390/s24227386.

[61] A. Uppal, B. Kakkar, P. Johri, Y. Kumar, and A. Koul, “Enhancing Lymphoma Cancer Detection Using Deep Transfer Learning on Histopathological Images,” Scientific Reports 15, no. 1 (2025): 38042, https://doi.org/10.1038/s41598-025-21888-5.

[62] M. Tabassum, A. Di Ieva, and S. D. Liu, “Meta Transfer Learning for Brain Tumor Segmentation Using nnUNet in Meningioma and Metastasis Cases,” Scientific Reports 15, no. 1 (2025): 37599, https://doi.org/10.1038/s41598-025-20957-z.

[63] N. L. Yadav, S. Singh, R. Kumar, and D. K. Nishad, “Transfer Learning With Fuzzy Decision Support for Multi-Class Lung Disease Classification: Performance Analysis of Pre-Trained CNN Models,” Scientific Reports 15, no. 1 (2025): 35127, https://doi.org/10.1038/s41598-025-19114-3.

[64] H. E. Kim, A. Cosa-Linan, N. Santhanam, M. Jannesari, M. E. Maros, and T. Ganslandt, “Transfer Learning for Medical Image Classification: A Literature Review,” BMC Medical Imaging 22, no. 1 (2022): 69, https://doi.org/10.1186/s12880-022-00793-7.

[65] X. Yu, J. Wang, Q. Q. Hong, R. Teku, S. H. Wang, and Y. D. Zhang, “Transfer Learning for Medical Images Analyses: A Survey,” Neurocomputing 489 (2022): 230–254, https://doi.org/10.1016/j.neucom.2021.08.159.

[66] M. A. Morid, A. Borjali, and G. Del Fiol, “A Scoping Review of Transfer Learning Research on Medical Image Analysis Using ImageNet,” Computers in Biology and Medicine 128 (2021): 104115, https://doi.org/10.1016/j.compbiomed.2020.104115.

[67] C. J. Du, G. S. Zou, Z. W. A , et al., “Highly Accurate and Efficient Prediction of Effective Thermal Conductivity of Sintered Silver Based on Deep Learning Method,” International Journal of Heat and Mass Transfer 201 (2023): 123654, https://doi.org/10.1016/j.ijheatmasstransfer.2022.123654.

RIGHTS & PERMISSIONS

2026 The Author(s). Materials Genome Engineering Advances published by Wiley-VCH GmbH on behalf of University of Science and Technology Beijing.