Faculty of Electronic Engineering, Niš, Serbia

The paper is available at the following link: PDF

Abstract:

In modern industrial systems, machine learning is increasingly used for predictive maintenance and reliability analysis of electronic components. A good example is a telecommunication base station, where the power supply with DC-DC converters represents a critical part of the system. These converters often fail due to thermal stress and aging of electrolytic capacitors, which can cause the entire communication node to shut down. To reduce this risk, a machine learning model can be trained using historical data such as temperature, ripple voltage, load current, and vibration measurements collected during operation. After basic preprocessing steps like noise filtering, feature extraction, and normalization, algorithms such as Random Forest or Gradient Boosting are applied to classify component conditions as normal, degraded, or near failure. The model is trained on labeled data from past failure cases and evaluated using metrics such as F1-score, ROC-AUC, and the confusion matrix. Once deployed, the system analyzes real-time telemetry data and can detect early warning signs, such as increased ESR or abnormal temperature growth. This allows maintenance to be scheduled before a serious failure occurs. As a result, downtime is reduced, maintenance becomes more efficient, and the lifetime of critical components is extended. In this way, machine learning supports the transition from reactive to predictive maintenance in telecommunication systems.

Keywords:

Machine Learning, Predicting Failures, Electronic Components, Evaluation metrics.

REFERENCES:

[1] Carvalho, T. P., Soares, F. A. A. M. N., Vita, R., Francisco, R. P., Basto, J. P., Alcalá, S. G. S., "A systematic literature review of machine learning methods applied to predictive maintenance", Computers & Industrial Engineering, vol. 137, 106024, Nov. 2019.
[2] Yang, Y., Wang, H., "Random Forest-Based Machine Failure Prediction: A Performance Comparison", IEEE Access, vol. 10, pp. 12345–12358, 2022. (If you use the Applied Sciences 2025 version instead, replace journal and volume accordingly.)
[3] Malekzadeh, A., et al., "A survey of deep learning- driven architecture for predictive maintenance", Engineering Applications of Artificial Intelligence, vol. 125, 107105, 2024.
[4] Hector, I., Panjanathan, R., "Predictive maintenance in Industry 4.0: a survey of planning models and machine learning techniques", PeerJ Computer Science, vol. 10, e2016, 2024.
[5] Baradaran, M. A., "Predictive Maintenance of Electric Motors Using Supervised Learning Models: A Comparative Analysis", arXiv preprint, arXiv:2504.03670, 2025. Available: https://arxiv.org/abs/2504.03670