droguzbektas/oguz-bektas-predictive-maintenance-101

Predictive Maintenance 101

Predictive Maintenance 101 is an educational resource prepared by Dr Oguz BEKTAS for introducing the main concepts of predictive maintenance, prognostics and health management, vibration analysis, anomaly detection, and explainable machine learning.

This repository is intended as a public learning-oriented dataset card and resource page. It is not designed as a benchmark dataset. Instead, it provides a structured entry point for understanding how industrial sensor data can be used to monitor machine health, detect abnormal behavior, and support maintenance decisions.

Author

Dr Oguz BEKTAS

Research interests include predictive maintenance, prognostics and health management, vibration analysis, time-series machine learning, condition monitoring, fault diagnosis, remaining useful life estimation, and explainable artificial intelligence.

Scope

Predictive maintenance aims to estimate the health condition of equipment before failure occurs. In industrial systems, this often requires the analysis of sensor data collected from machines, components, or operating environments.

Typical data sources include vibration signals, temperature measurements, pressure values, current signals, acoustic emissions, process telemetry, and operating condition variables.

This educational resource focuses especially on:

• vibration-based condition monitoring
• time-series feature extraction
• anomaly detection
• fault diagnosis
• machine health classification
• remaining useful life estimation
• explainable AI for maintenance decision support

Main Topics

Predictive Maintenance

Predictive maintenance uses data-driven and model-based methods to identify early signs of degradation. The aim is to support maintenance planning before functional failure occurs.

Prognostics and Health Management

Prognostics and Health Management, also known as PHM, focuses on diagnosis, prognosis, health assessment, and maintenance decision support. PHM methods are widely used in industrial machinery, aerospace systems, energy systems, transportation, and manufacturing.

Vibration Analysis

Vibration analysis is one of the most important tools for monitoring rotating machinery and mechanical systems. Vibration signals may contain useful information about imbalance, misalignment, looseness, bearing defects, gear faults, and structural degradation.

Feature Extraction

Sensor signals can be transformed into useful features before machine learning models are trained. Common feature groups include:

• time-domain statistical features
• frequency-domain features
• Welch power spectral density features
• envelope analysis features
• harmonic features
• Catch22 time-series features
• health indicators

Machine Learning

Machine learning methods can be used for fault classification, anomaly detection, health-state estimation, and remaining useful life prediction. Common algorithms include Random Forest, ExtraTrees, LightGBM, support vector machines, neural networks, and anomaly detection models.

Explainable AI

Explainable AI helps engineers understand why a model makes a specific prediction. Methods such as SHAP can be used to identify important sensors, features, and operating conditions.

Example Use Cases

This resource is relevant to the following use cases:

• bearing fault detection
• rotating machinery diagnostics
• vibration-based anomaly detection
• degradation trend analysis
• remaining useful life estimation
• industrial condition monitoring
• sensor-based maintenance decision support
• interpretable machine learning for PHM

Keywords

Predictive Maintenance; Prognostics and Health Management; PHM; Vibration Analysis; Time-Series Analysis; Machine Learning; Explainable AI; SHAP; Condition Monitoring; Fault Diagnosis; Remaining Useful Life; Industrial AI; Anomaly Detection; Health Indicator.

Related Links

• GitHub: https://github.com/droguzbektas
• GitHub Pages: https://droguzbektas.github.io/
• WarwickGrad: https://www.warwickgrad.net/users/droguzbektas/
• Google Sites: https://sites.google.com/view/doc-dr-oguz-bektas/home

Citation

If you use or refer to this educational resource, please cite it as:

@misc{bektas_predictive_maintenance_101,
  author       = {Oguz BEKTAS},
  title        = {Predictive Maintenance 101},
  year         = {2026},
  publisher    = {Hugging Face},
  note         = {Educational resource on predictive maintenance, PHM, vibration analysis, and explainable AI}
}