AI4I 2020 Predictive Maintenance — Milling Machine Sensor Failures [10,000 Records]

Overview

The AI4I 2020 Predictive Maintenance Dataset is a synthetic dataset carefully designed to reflect real industrial predictive maintenance data encountered in milling machine operations. It was created because actual industrial maintenance datasets are rarely publicly available due to proprietary constraints, yet there is strong demand for labeled IIoT failure data in machine learning research.

The dataset models five distinct failure modes — Tool Wear Failure (TWF), Heat Dissipation Failure (HDF), Power Failure (PWF), Overstrain Failure (OSF), and Random Failure (RNF) — alongside a combined machine failure label. Each record includes air and process temperature, rotational speed, torque, and tool wear duration, making it a compact but highly structured multi-label classification benchmark.

With 10,000 records, no missing values, and clear feature-label relationships, AI4I 2020 is widely used to benchmark and compare ML classifiers including Random Forest, XGBoost, SVM, and neural networks for IIoT fault detection and root-cause analysis.

Column Schema

Key Statistics

• Total Records: 10,000
• Features: 14 columns
• Failure Types: 5 individual modes + 1 combined label
• Missing Values: None
• File Format: CSV
• Donated: August 2020

Use Cases

• Multi-label failure mode classification for CNC milling machines
• Benchmarking ML/DL classifiers on IIoT predictive maintenance tasks
• Feature importance analysis for industrial sensor-driven fault detection
• Root cause analysis of manufacturing equipment failures

Source & Attribution

The dataset was created by Stephan Matzka at the Berlin School of Economics and Law and donated to the UCI Machine Learning Repository. It is available via UCI and widely referenced in Industry 4.0 and IIoT machine learning papers.