Received 27.08.2025, Revised 12.11.2025, Accepted 23.12.2025

Methods of signal processing and data interpretation for detecting microdefects in industrial materials

Kanan Mikayilov, Latafat Gardashova

The prompt and accurate detection of microdefects in industrial materials is a priority for improving product quality, production safety, and process optimisation. The purpose of this study was to create an automated inspection system that uses artificial intelligence to identify microdefects in industrial materials. The study was conducted on laboratory and industrial samples with microdefects using a multi-sensor system consisting of visual cameras, ultrasound, thermography, and X-rays. The data was pre-processed by filtering, normalising, and extracting contours and analysed using Convolutional Neural Network (CNN), Vision Transformer, and 3D CNN deep learning models with multimodal integration, transfer learning, augmentation, and weight optimisation, with the system performance evaluated by accuracy, precision, recall, and F1-score metrics. A comprehensive analysis showed that the individual use of visual cameras with an accuracy of 92.3%, ultrasonic sensors with an accuracy of 89.5%, thermography with an accuracy of 85.1%, and an X-ray scanner with an accuracy of 95.6% provided high results, and their combination increased the integrated index to 97.8%, which confirms the advantages of the multichannel approach. The use of pre-processing methods (Gaussian and median filters, normalisation, histogram alignment) and augmentation increased the accuracy to 94.1% and the F1-score to 92.6% (compared to the initial 85.2%), while transfer learning increased accuracy by 12-15% and reduced training time, reducing the number of false positives. The system maintained an accuracy of over 90% in noise and variations in production conditions, and at least 80% in extreme scenarios. Practical tests on a server with NVIDIA A6000 GPUs showed an average sample processing time of 120-180 ms (5-8 FPS) and linear scalability with the number of GPUs, which confirmed the system’s suitability for integration into real-time industrial systems. The findings of this study can be used by quality control specialists and developers of industrial information and measurement systems to improve the accuracy and efficiency of microdefect detection

neural networks; deep learning; computer vision; transfer learning; data augmentation
113-124
Mikayilov, K., & Gardashova, L. (2025). Methods of signal processing and data interpretation for detecting microdefects in industrial materials. Information Technologies and Computer Engineering, 22(3), 113-124. https://doi.org/10.31649/vitce/3.2025.113

References

[1] Alamuru, S., Reddy, G.S., & Raju, M.J. (2024). Artificial intelligence and machine learning for defect detection in castings. Journal of Physics: Conference Series, 2837, article number 012079. doi: 10.1088/1742-6596/2837/1/012079.

[2] Ashebir, D.A., Hendlmeier, A., Dunn, M., Arablouei, R., Lomov, S.V., Di Pietro, A., & Nikzad, M. (2024). Detecting multi-scale defects in material extrusion additive manufacturing of fiber-reinforced thermoplastic composites: A review of challenges and advanced non-destructive testing techniques. Polymers, 16(21), article number 2986. doi: 10.3390/polym16212986.

[3] Deepak, A., & Rao, P. (2024). Real time defect identification using advanced artificial intelligence based material strain sensors for environmental safety. Journal of Electrical Systems, 20(1), 33-40. doi: 10.52783/jes.660.

[4] Dubaish, A.A., & Jaber, A.A. (2023). State-of-the-art review into signal processing and artificial intelligence-based approaches applied in gearbox defect diagnosis. Engineering and Technology Journal, 42(1), 157-172. doi: 10.30684/ etj.2023.142462.1535.

[5] Ebayyeh, A., & Abd Al Rahman, M. (2022). Deep learning for automatic optical inspection and quality evaluation of semiconductor and optoelectronic manufacturing. London: Brunel University London.

[6] Ercetin, A., Der, O., Akkoyun, F., Gowdru Chandrashekarappa, M.P., Şener, R., Çalışan, M., Olgun, N., Chate, G., & Bharath, K.N. (2024). Review of image processing methods for surface and tool condition assessments in machining. Journal of Manufacturing and Materials Processing, 8(6), article number 244. doi: 10.3390/jmmp8060244.

[7] He, Z., Lin, S., Xiao, Y., Fang, H., & Sun, L. (2025). A novel micro-defects imaging system for high-reflective material surfaces. IEEE Transactions on Instrumentation and Measurement, 74, article number 4501814. doi: 10.1109/ TIM.2025.3541809.

[8] Jiao, S., Yang, W., Wu, C., Li, Y., & Xue, B. (2025). Mixed-type micro-defect detection in semiconductor wafers: A dual-modal feature real-time detection approach via optical topography and lightweight classification network. Engineering Applications of Artificial Intelligence, 160, article number 111838. doi: 10.1016/j.engappai.2025.111838.

[9] Kim, H.S., Park, D.W., Kim, S.I., Oh, G.H., & Kim, H.S. (2023). Non-destructive detection of thin micro-defects in glass reinforced polymer composites using a terahertz electro-magnetic wave based on a convolution neural network. Composites Part B: Engineering, 257, article number 110694. doi: 10.1016/j.compositesb.2023.110694.

[10] Mahamud, Z.H., Khan, M.R., Amin, J.M., & Islam, M.S. (2025). AI for defect detection in additive manufacturing: Applications in renewable energy and biomedical engineering. Strategic Data Management and Innovation, 2(1), 1-20. doi: 10.71292/SDMI.V2I01.8.

[11] Malakar, K.D., Roy, S., & Kumar, M. (2025). Digital imaging: Processing and analysis. In K. Das Malakar, S. Roy & M. Kumar (Eds.), Geospatial technologies in coastal ecologies monitoring and management (pp. 257-289). Cham: Springer. doi: 10.1007/978-3-031-92017-2_8.

[12] Mohammed, A., & Hussain, M. (2025). Advances and challenges in deep learning for automated welding defect detection: A technical survey. IEEE Access, 13, 94553-94569. doi: 10.1109/ACCESS.2025.3574083.

[13] Mustafayev, E., & Azimov, R. (2024). Computer system of evaluation of the mass exam results based on recognition of handprinted Azerbaijani characters. In G. Mammadova, T. Aliev & K. Aida-zade (Eds.), Information technologies and their applications (pp. 171-183). Cham: Springer. doi: 10.1007/978-3-031-73420-5_15.

[14] Na, D., Choi, M., Yuan, F.G., & Kim, H. (2025). Detection of microdefects in multilayer ceramic capacitors using the instantaneous frequency in the electromechanical response. Measurement, 244, article number 116528. doi: 10.1016/j. measurement.2024.116528.

[15] Prottasha, N.J., Sami, A.A., Kowsher, M., Murad, S.A., Bairagi, A.K., Masud, M., & Baz, M. (2022). Transfer learning for sentiment analysis using BERT based supervised fine-tuning. Sensors, 22(11), article number 4157. doi: 10.3390/ s22114157.

[16] Qiu, M., & Ge, X. (2025). Nondestructive evaluation of aging failure in insulation coatings by ultrasonic guided wave based on signal processing and machine learning. Coatings, 15(3), article number 347. doi: 10.3390/coatings15030347.

[17] Ren, Z., Fang, F., Yan, N., & Wu, Y. (2022). State of the art in defect detection based on machine vision. International Journal of Precision Engineering and Manufacturing – Green Technology, 9(2), 661-691. doi: 10.1007/s40684-021-00343-6.

[18] Saberironaghi, A., Ren, J., & El-Gindy, M. (2023). Defect detection methods for industrial products using deep learning techniques: A review. Algorithms, 16(2), article number 95. doi: 10.3390/a16020095.

[19] Tai, J.L., Sultan, M.T., Łukaszewicz, A., Józwik, J., Oksiuta, Z., & Shahar, F.S. (2025). Recent trends in non-destructive testing approaches for composite materials: A review of successful implementations. Materials, 18(13), article number 3146. doi: 10.3390/ma18133146.

[20] Tran, K.S., Shirinzadeh, B., & Smith, J. (2024). Eddy current-based identification and depth investigation of microdefects in steel filaments. Sensors, 24(16), article number 5101. doi: 10.3390/s24165101.

[21] Wei, B., Ma, Q., Wen, S., Zhang, J., Peng, L., & Huang, S. (2025). Aluminium foil micro-defect detection method based on motion-induced eddy current using differential TMR sensor spatial position compensation. Nondestructive Testing and Evaluation. doi: 10.1080/10589759.2025.2456674.

[22] Xiao, Q., Huang, J., Huang, Z., Li, C., & Xu, J. (2023). Transparent component defect detection method based on improved YOLOv7 algorithm. International Journal of Pattern Recognition and Artificial Intelligence, 37(14), article number 2350030. doi: 10.1142/S0218001423500301.

[23] Yan, X., & Fu, H. (2024). Opportunities and challenges for predicting the service status of SLM metal parts under big data and artificial intelligence. Materials, 17(22), article number 5648. doi: 10.3390/ma17225648.

[24] Yusubov, E. (2023). Design of an intelligent information measurement system for photovoltaic dc microgrids. Proceedings of Azerbaijan High Technical Educational Institutions, 34(11), 54-64. doi: 10.36962/PAHTEI34112023-54.

[25] Zhao, Z. (2021). Review of non-destructive testing methods for defect detection of ceramics. Ceramics International, 47(4), 4389-4397. doi: 10.1016/j.ceramint.2020.10.065.

[26] Zheng, X., Zheng, S., Kong, Y., & Chen, J. (2021). Recent advances in surface defect inspection of industrial products using deep learning techniques. International Journal of Advanced Manufacturing Technology, 113(1), 35-58. doi: 10.1007/s00170-021-06592-8.