A Real-World IIoT Dataset for Predictive Maintenance Based on Machine Degradation

Overview

This dataset provides authentic industrial sensor data from CNC machining centers to support predictive maintenance and machine health monitoring research. It was published in Nature Scientific Reports in 2025 and covers continuous machine operation until failure or maintenance events.

Data Collection

• Data acquired from Computer Numerical Control (CNC) centers in an active production environment over six months.
• Sensor suite includes tri-axial vibration sensors (10-25 kHz), thermocouples, current/power meters, and acoustic emission sensors.
• Captures transition from healthy state to early degradation, advanced wear, and final failure.

Use Cases

• Training supervised and unsupervised models for failure forecasting and Remaining Useful Life (RUL) estimation.
• Benchmarking multi-sensor fusion and anomaly detection algorithms on clean vs. noisy industrial data.
• Analyzing high-frequency mechanical signatures for specific component fault diagnosis (e.g., bearings, tools).

Access

Documented in Scientific Reports and available via the publisher's linked repositories (e.g., PMC or Mendeley). Users must follow the provided citation and licensing terms.