Impact of Thermal Image Preprocessing Methods on a ResNet50 Model Baseline Performance for Ball Bearing Fault Diagnosis

Notice

This is an unedited manuscript accepted for publication and provided as an Article in Press for early access at the author’s request. The article will undergo copyediting, typesetting, and galley proof review before final publication. Please be aware that errors may be identified during production that could affect the content. All legal disclaimers of the journal apply.

Year : 2026 | Volume : 16 | 03 | Page :
By

Tanmay Garg,

Aditya Singh Tomar,

Vaibhav Shivhare,

  1. Student, Department of Mechanical Engineering, Madhav Institute Of Technology and Science, Deemed University, Gwalior, Madhya Pradesh, India
  2. Student, Department of Mechanical Engineering, Madhav Institute Of Technology and Science, Deemed University, Gwalior, Madhya Pradesh, India
  3. Professor, Department of Mechanical Engineering, Madhav Institute Of Technology and Science, Deemed University, Gwalior, Madhya Pradesh, India

Abstract

Ball bearings are an important part for the operational rotating machinery. To prevent it from failing, traditional engineering uses many methods. One of these is infrared thermography (IRT). IRT is a useful tool that helps monitor the situation without physical contact. But in the current world, various fields use artificial intelligence to do it more accurately & for fast detection. Deep learning models play an important role in fault detection with the help of thermal images, but background noise is also part of these thermal images that contributes nothing to bearing fault analysis; even deep learning models become confused due to this noise in thermal images. These noises distract the model, which is not really good. This paper presents a comparative analysis effect of the three processing methods within the baseline of the ResNet50 model. These three preprocessing methods are used to resolved this problem, they three are the dispersion degree-based region selection method (DDRS), the local binary pattern (LBP) texture preprocessing method, and the wavelet transform denoising method (WTD/WT) Result of the work we found ResNet50 baseline work well with 98.34 % accuracy. The study also shows that proper preprocessing increases feature stability during training, speeds up model convergence, and improves classification accuracy. The comparative analysis offers insightful information about the connection between deep feature learning and thermal image enhancement techniques, showing that the choice of preprocessing techniques is crucial to optimizing the efficacy of transfer learning models for condition monitoring applications. All things considered, this study develops a thorough assessment of thermal image preprocessing techniques for bearing failure diagnosis using ResNet50 and advances the creation of dependable, non-invasive predictive maintenance systems. In addition to providing a solid basis for future research involving advanced attention mechanisms, hybrid preprocessing frameworks, explainable artificial intelligence (XAI), and multimodal sensor fusion to further enhance the robustness and generalization of intelligent fault diagnosis systems, the findings provide useful guidance for researchers and industry practitioners in choosing appropriate preprocessing techniques for thermal image-based fault diagnosis.

Keywords: Bearing faults, Thermal image analysis, Infrared Thermography, Dispersion Degree, Machine Learning, Condition Monitoring

How to cite this article: Tanmay Garg, Aditya Singh Tomar, Vaibhav Shivhare. Impact of Thermal Image Preprocessing Methods on a ResNet50 Model Baseline Performance for Ball Bearing Fault Diagnosis. Trends in Mechanical Engineering & Technology. 2026; 16(03):-.
How to cite this URL: Tanmay Garg, Aditya Singh Tomar, Vaibhav Shivhare. Impact of Thermal Image Preprocessing Methods on a ResNet50 Model Baseline Performance for Ball Bearing Fault Diagnosis. Trends in Mechanical Engineering & Technology. 2026; 16(03):-. Available from: https://journals.stmjournals.com/tmet/article=2026/view=260289

References

  1. Santos LFD, Canuto JLS, Souza RCT, Aylon LBR. Thermographic image-based diagnosis of failures in electrical motors using deep transfer learning. Engineering Applications of Artificial Intelligence. 2023;126.
  2. Xu F, et al. A review of bearing failure modes, mechanisms and causes. Engineering Failure Analysis. 2023;152.
  3. Atif M, et al. AI-driven thermography-based fault diagnosis in single-phase induction motor. Results in Engineering. 2024.
  4. Gao Y, et al. Convolution neural network fusion high power halogen lamp induced thermal-wave diffusion multidimensional features. Infrared Physics & Technology. 2023;133.
  5. Janssens O, et al. Data-driven imbalance and hard particle detection in rotating machinery using infrared thermal imaging. Infrared Physics & Technology. 2017.
  6. Souza RM, et al. Deep learning for diagnosis and classification of faults in industrial rotating machinery. Computers & Industrial Engineering. 2021;153.
  7. Glowacz A, Glowacz Z. Diagnosis of the three-phase induction motor using thermal imaging. Infrared Physics & Technology. 2016.
  8. Jeffali F, et al. Diagnosis of three-phase induction motor using infrared thermography. Infrared Physics & Technology. 2019.
  9. Kannojia SP, Jaiswal G. Effects of varying resolution on performance of CNN based image classification. International Journal of Computer Sciences and Engineering. 2018.
  10. Chang L, et al. Experimental study on infrared radiation and crack evolution characteristics. Infrared Physics & Technology. 2024;138.
  11. Singh G, et al. Induction motor inter-turn fault detection using infrared thermographic analysis. Infrared Physics & Technology. 2016.
  12. He Y, et al. Infrared machine vision and infrared thermography with deep learning: a review. Infrared Physics & Technology. 2021;116.
  13. Singh G, Naikan VNA. Infrared thermography based diagnosis of inter-turn fault and cooling system failure. Infrared Physics & Technology. 2017.
  14. Nasiri A, et al. Intelligent fault diagnosis of cooling radiator based on deep learning. Applied Thermal Engineering. 2019.
  15. Janssens O, et al. Thermal image based fault diagnosis for rotating machinery. Infrared Physics & Technology. 2015;73.
  16. Osemudiamhen AE, et al. Machine learning-driven strength prediction using infrared radiation indices. Infrared Physics & Technology. 2025.
  17. Mian T, Choudhary A, Fatima S. Vibration and infrared thermography based multiple fault diagnosis using deep learning. Nondestructive Testing and Evaluation. 2022.
  18. Sharma K, Goyal D, Kanda R. Intelligent fault diagnosis of bearings based on convolutional neural network using infrared thermography. Proc Inst Mech Eng Part J J Eng Tribol. 2022;236(12). doi:10.1177/13506501221082746.
  19. Alvarado-Robles G, et al. Non-contact intelligent fault diagnosis based on thermography and deep-feature learning. Infrared Physics & Technology. 2025.
  20. Takabi J. On the thermally induced failure of rolling element bearings [dissertation]. Louisiana: Louisiana State University; 2015.
  21. Shivhare V, Kumar R. A lightweight CNN using smoothed pseudo-Wigner–Ville distribution for bearing fault severity classification. Journal of Vibration Engineering & Technologies. 2026;14:68.
  22. Khan H, Upadhyay N, Shivhare V. Bearing defects classification using wavelet time scattering features and machine learning techniques. Maintenance, Reliability and Condition Monitoring. 2025;5(1):98-108.
  23. Khan H, Sharma A, Upadhyay N, Shivhare V. Bearing defects classification using wavelet time scattering features and ensemble techniques. In: Singh M, et al., editors. Proceedings of the unified conference of DAMAS, IncoME VIII and TEPEN Conferences: UNIfied 2024. Cham: Springer; 2026. (Mechanisms and Machine Science; vol. 181).
  24. Aher H, Ghuge N. Performance comparison of machine learning algorithms for condition monitoring of tapered roller bearings. Tribology and Materials. 2025;4(2):100-115.
  25. Huang S, Wang F, Pan T, Wang Z, Zhou H. Research on a monitoring model of revolute pair clearance based on dynamic features and thermal imaging fusion. Infrared Physics & Technology. 2023;135.
  26. Duan L, et al. Segmented infrared image analysis for rotating machinery fault diagnosis. Infrared Physics & Technology. 2016;77:267-276.
  27. Mehta A, Goyal D, Choudhary A, Pabla BS, Belghith S. Machine learning-based fault diagnosis of self-aligning bearings using infrared thermography. Mathematical Problems in Engineering. 2021.
  28. Al-Musawi AK, Anayi F, Packianather M. Three-phase induction motor fault detection based on thermal image segmentation. Infrared Physics & Technology. 2020;104.
  29. Mian T, Choudhary A, Fatima S. Vibration and infrared thermography-based multiple fault diagnosis of bearing using deep learning. Nondestructive Testing and Evaluation. 2023;38(2):275-296.
Ahead of Print Subscription Original Research
Volume 16
03
Received 17/07/2026
Accepted 18/09/2026
Published 13/10/2026
Publication Time 88 Days

Login

My IP

PlumX Metrics

Support