Inclusion and holes detection in 3D objects using numerical modeling and neural networks

Authors

DOI:

https://doi.org/10.15587/2706-5448.2026.352493

Keywords:

YOLO, steady-heat thermography, non-destructive testing, synthetic data generation, numerical modeling

Abstract

The object of research is a stationary heat-conduction process in three-dimensional heterogeneous media which components are in ideal thermal contact.

The addressed problem is the difficulties in detecting, localizing, and classifying inclusions in 3D objects. Most methods rely on employing active thermography which produces data as time-dependent sequences. Both capturing and processing it is quite costly and resource required.

The paper proposes a hybrid approach that relies on steady-heat thermograms, which are significantly easier and cheaper to capture. The developed approach uses one of the variants of the indirect near-boundary element method (INBEM), SOTA detectors (YOLO 11n and 12n), and a custom depth head based on ConvNeXt V2. Additionally, the paper showed that attention-centric architecture is promising for processing steady heat images.

INBEM with near-boundary elements in the form of families of points achieves an execution time of approximately 50 seconds per sample, with a general maximum error of approximately 0.08. This enabled the creation of a large dataset, comprising approximately 130K samples. Additionally, a testing dataset with a size of 7K and slightly different variance is obtained. Both YOLO 11n and 12n showed mAP50:95 metric results on the testing dataset of 85.2% and 89.4%, respectively. The precision/recall for both models are the following: 92.0/92.8 and 92.3/96.3. The depth head showed a MAPE of about 2%.

The proposed method focuses on detecting inclusions and holes using steady heat images, so it is suitable for relatively low-cost analysis, as obtaining such data is easier and quicker than collecting time-dependent data. It may be useful to screen slab-like structures, such as photovoltaic panels. Wall diagnostics is one possible future application area, as the work can be extended to semi-infinite objects. Thus, the results may serve as a basis for a low-cost inspection tool.

Author Biographies

Roman Havdulskyi , Lviv Polytechnic National University

PhD Student

Department of Software

Liubov Zhuravchak, Lviv Polytechnic National University

Doctor of Technical Sciences, Senior Researcher, Professor

Department of Software

References

  1. Rezayiye, R. K., Ibarra-Castanedo, C., Maldague, X. (2024). Methods for Corrosion Detection in Pipes Using Thermography: A Case Study on Synthetic Datasets. Algorithms, 17 (10), 439. https://doi.org/10.3390/a17100439
  2. Salah, M., Werghi, N., Svetinovic, D., Abdulrahman, Y. (2025). Multi-Modal Attention Networks for Enhanced Segmentation and Depth Estimation of Subsurface Defects in Pulse Thermography. ArXiv. https://doi.org/10.48550/arxiv.2501.09994
  3. Sels, S., Bogaerts, B., Verspeek, S., Ribbens, B., Steenackers, G., Penne, R. et al. (2020). 3D Defect detection using weighted principal component thermography. Optics and Lasers in Engineering, 128, 106039. https://doi.org/10.1016/j.optlaseng.2020.106039
  4. Dudzik, S. (2015). Two-stage neural algorithm for defect detection and characterization uses an active thermography. Infrared Physics & Technology, 71, 187–197. https://doi.org/10.1016/j.infrared.2015.03.003
  5. Szymanik, B., Psuj, G., Hashemi, M., Lopato, P. (2021). Detection and Identification of Defects in 3D-Printed Dielectric Structures via Thermographic Inspection and Deep Neural Networks. Materials, 14 (15), 4168. https://doi.org/10.3390/ma14154168
  6. Kumar, R., Puranik, V. E., Gupta, R. (2024). Unveiling the Potential of Infrared Thermography in Quantitative Investigation of Potential-Induced Degradation in Crystalline Silicon PV Module. Solar Energy Advances, 4, 100049. https://doi.org/10.1016/j.seja.2023.100049
  7. Spajić, M., Talajić, M., Mršić, L. (2024). Using CNNs for Photovoltaic Panel Defect Detection via Infrared Thermography to Support Industry 4.0. Business Systems Research Journal, 15 (1), 45–66. https://doi.org/10.2478/bsrj-2024-0003
  8. Di Tommaso, A., Betti, A., Fontanelli, G., Michelozzi, B. (2022). A multi-stage model based on YOLOv3 for defect detection in PV panels based on IR and visible imaging by unmanned aerial vehicle. Renewable Energy, 193, 941–962. https://doi.org/10.1016/j.renene.2022.04.046
  9. Ta, Q. T., Mac, V. H., Huh, J., Yim, H. J., Lee, G. (2025). Automatic detection of subsurface defects in concrete structures using state-of-the-art deep learning-based object detectors on the infrared dataset. Engineering Structures, 329, 119829. https://doi.org/10.1016/j.engstruct.2025.119829
  10. Jeon, M., Yoo, S., Kim, S.-W. (2022). A Contactless PCBA Defect Detection Method: Convolutional Neural Networks With Thermographic Images. IEEE Transactions on Components, Packaging and Manufacturing Technology, 12 (3), 489–501. https://doi.org/10.1109/tcpmt.2022.3147319
  11. Liu, J., Long, X., Jiang, C., Liao, W. (2024). Multi-feature vision transformer for automatic defect detection and quantification in composites using thermography. NDT & E International, 143, 103033. https://doi.org/10.1016/j.ndteint.2023.103033
  12. Wu, Z., Chen, S., Feng, F., Qi, J., Feng, L., Tao, N. et al. (2023). Automatic defect detection and three-dimensional reconstruction from pulsed thermography images based on a bidirectional long-short term memory network. Engineering Applications of Artificial Intelligence, 124, 106574. https://doi.org/10.1016/j.engappai.2023.106574
  13. Resendiz-Ochoa, E., Trejo-Chavez, O., Saucedo-Dorantes, J. J., Morales-Hernandez, L. A., Cruz-Albarran, I. A. (2024). Application of Thermography and Convolutional Neural Network to Diagnose Mechanical Faults in Induction Motors and Gearbox Wear. Applied System Innovation, 7 (6), 123. https://doi.org/10.3390/asi7060123
  14. Era, I. Z., Zhou, F., Raihan, A. S., Ahmed, I., Abul-Haj, A., Craig, J. et al. (2024). In-Situ Melt Pool Characterization via Thermal Imaging for Defect Detection in Directed Energy Deposition Using Vision Transformers. ArXiv. https://doi.org/10.48550/arxiv.2411.12028
  15. Zheng, S., Zhang, J., Zu, R., Li, Y. (2024). Vision transformer-enhanced thermal anomaly detection in building facades through fusion of thermal and visible imagery. Journal of Asian Architecture and Building Engineering, 24 (4), 2854–2868. https://doi.org/10.1080/13467581.2024.2379866
  16. Campione, I., Lucchi, F., Santopuoli, N., Seccia, L. (2020). 3D Thermal Imaging System with Decoupled Acquisition for Industrial and Cultural Heritage Applications. Applied Sciences, 10 (3), 828. https://doi.org/10.3390/app10030828
  17. Sabathiel, S., Sanchis-Alepuz, H., Wilson, A. S., Reynvaan, J., Stipsitz, M. (2024). Neural network-based reconstruction of steady-state temperature systems with unknown material composition. Scientific Reports, 14 (1). https://doi.org/10.1038/s41598-024-73380-1
  18. Zhuravchak, L. M., Zabrods’ka, N. V. (2010). Nonstationary thermal fields in inhomogeneous materials with nonlinear behavior of the components. Materials Science, 46 (1), 36–46. https://doi.org/10.1007/s11003-010-9261-9
  19. Zhuravchak, L., Kruk, O. (2015). Consideration of the nonlinear behavior of environmental material and a three-dimensional internal heat sources in mathematical modeling of heat conduction. Mathematical Modeling and Computing, 2 (1), 107–113. https://doi.org/10.23939/mmc2015.01.107
  20. Zhuravchak, L. (2019). Computation of pressure change in piecewise-homogeneous reservoir for elastic regime by indirect near-boundary element method. 2019 IEEE 14th International Conference on Computer Sciences and Information Technologies (CSIT), 141–144. https://doi.org/10.1109/stc-csit.2019.8929820
  21. Zhuravchak, L.; Shakhovska, N., Medykovskyy, M. O. (Eds.) (2019). Mathematical Modelling of Non-stationary Processes in the Piecewise-Homogeneous Domains by Near-Boundary Element Method. Advances in Intelligent Systems and Computing IV. Cham: Springer, 64–77. https://doi.org/10.1007/978-3-030-33695-0_6
  22. Havdulskyi, R., Zhuravchak, L. (2025). Recognition of inclusion characteristics using neural network methods in stationary process modeling. Visnyk Natsionalnoho universytetu “Lvivska politekhnika”. Seriia Informatsiini systemy ta merezhi, 17, 75–92. https://doi.org/10.23939/sisn2025.17.075
  23. Havdulskyi, R., Zhuravchak, L., Yakovyna, V.; Hovorushchenko, T., Savenko, O., Popov, P. T., Lysenko, S. (Eds.) (2025). Inclusion parameters estimation using INBEM and CNNs with gradient-based feature extraction. 2nd International Workshop on Intelligent & CyberPhysical Systems (ICyberPhyS 2025). CEUR Workshop Proceedings, 4013. Available at: https://ceur-ws.org/Vol-4013/paper2.pdf
  24. Khanam, R., Hussain, M. (2024). YOLOv11: An Overview of the Key Architectural Enhancements. ArXiv. https://doi.org/10.48550/arxiv.2410.17725
  25. Tian, Y., Ye, Q., Doermann, D. (2025). YOLOv12: Attention-Centric Real-Time Object Detectors. ArXiv. https://doi.org/10.48550/arxiv.2502.12524
  26. Jaeger, B. E., Schmid, S., Grosse, C. U., Gögelein, A., Elischberger, F. (2022). Infrared Thermal Imaging-Based Turbine Blade Crack Classification Using Deep Learning. Journal of Nondestructive Evaluation, 41 (4). https://doi.org/10.1007/s10921-022-00907-9
  27. Zhuravchak, L. M., Zabrodska, N. V. (2020). Using of partly-boundary elements as a version of the indirect near-boundary element method for potential field modeling. Mathematical Modeling and Computing, 8 (1), 1–10. https://doi.org/10.23939/mmc2021.01.001
  28. Zhuravchak, L. M. (2024). Potential field modeling by combination of near-boundary and contact elements with non-classical finite differences in a heterogeneous medium. Mathematical Modeling and Computing, 11 (2), 373–384. https://doi.org/10.23939/mmc2024.02.373
  29. Zhuravchak, L., Zabrodska, N. (2024). Algorithm for determining inclusion parameters in solving inverse problems of geoelectrical exploration using the profiling method. Geodynamics, 1 (36), 98–107. https://doi.org/10.23939/jgd2024.01.098
  30. Ren, S., He, K., Girshick, R., Sun, J. (2015). Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks. ArXiv. https://doi.org/10.48550/arxiv.1506.01497
  31. Sharma, A., Kumar, V., Longchamps, L. (2024). Comparative performance of YOLOv8, YOLOv9, YOLOv10, YOLOv11 and Faster R-CNN models for detection of multiple weed species. Smart Agricultural Technology, 9, 100648. https://doi.org/10.1016/j.atech.2024.100648
  32. Russakovsky, O., Deng, J., Su, H., Krause, J., Sanjeev Satheesh, Ma, S. et al. (2014). ImageNet Large Scale Visual Recognition Challenge. ArXiv. https://doi.org/10.48550/arxiv.1409.0575
  33. Woo, S., Debnath, S., Hu, R., Chen, X., Liu, Z., Kweon, I. S. et al. (2023). ConvNeXt V2: Co-designing and Scaling ConvNets with Masked Autoencoders. ArXiv. https://doi.org/10.48550/arxiv.2301.00808
  34. Shazeer, N. (2020). GLU Variants Improve Transformer. ArXiv. https://doi.org/10.48550/arXiv.2002.05202
  35. Arora, I. (2024). Improving Performance of Data Science Applications in Python. Indian Journal Of Science And Technology, 17 (24), 2499–2507. https://doi.org/10.17485/ijst/v17i24.914
  36. Zhuravchak, L. (2023). Computational Aspects of the Use of Different Types of Near-boundary elements in Modeling the Non-stationary Heat Conduction Process. 2023 IEEE 18th International Conference on Computer Science and Information Technologies (CSIT). Lviv: IEEE, 1–5. https://doi.org/10.1109/csit61576.2023.10324216
  37. Sapkota, R., Harsha, C. R., Sharda, A., Karkee, M. (2025). YOLO26: Key Architectural Enhancements and Performance Benchmarking for Real-Time Object Detection. ArXiv. https://doi.org/10.48550/arxiv.2509.25164
Inclusion and holes detection in 3D objects using numerical modeling and neural networks

Downloads

Published

2026-02-28

How to Cite

Havdulskyi , R., & Zhuravchak, L. (2026). Inclusion and holes detection in 3D objects using numerical modeling and neural networks. Technology Audit and Production Reserves, 1(2(87), 113–122. https://doi.org/10.15587/2706-5448.2026.352493

Issue

Vol. 1 No. 2(87) (2026): Information and control systems

Section

Mathematical Modeling

License

Copyright (c) 2026 Roman Havdulskyi , Liubov Zhuravchak

Creative Commons License

This work is licensed under a Creative Commons Attribution 4.0 International License.

The consolidation and conditions for the transfer of copyright (identification of authorship) is carried out in the License Agreement. In particular, the authors reserve the right to the authorship of their manuscript and transfer the first publication of this work to the journal under the terms of the Creative Commons CC BY license. At the same time, they have the right to conclude on their own additional agreements concerning the non-exclusive distribution of the work in the form in which it was published by this journal, but provided that the link to the first publication of the article in this journal is preserved.