ABB IRB 1100 Robot Synthetic Data for Predictive Maintenance

2,000 simulated instances with 215 features monitor electrical, mechanical, thermal, and vibration parameters of an industrial robot. Mohamed Lamine Soudani generated this synthetic dataset to address the scarcity of real-world failure data, achieving an F1-macro score of 0.7625 with LightGBM. The dataset enables fault identification and severity-level validation for three defect types.

Use Cases

• Training fault detection models based on simulated multisensory signals for six robot axes.
• Evaluating feature selection methods based on a dimensionality reduction analysis that cut features by over 50%.
• Benchmarking ensemble tree methods against SVM approaches for industrial defect classification.
• Developing severity-level prioritization systems based on three severity levels per defect type.
• Studying the challenge of classifying gear wear defects based on partial feature overlap and imbalance effects.

Strengths

• Dataset comprises 2,000 instances with 215 explanatory characteristics.
• Simulation model incorporates sensors for electrical, mechanical, thermal, and vibration parameters.
• Results include detailed performance metrics (F1-macro = 0.7625) and class-based analysis (F1 ≈ 0.97–1.00 for some defects).
• Framework supports both fault identification and severity-level validation for three defect types.

Limitations

• Data is synthetic, generated via simulation, and requires validation on real-world industrial datasets.
• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.

Provenance

Source

Mohamed Lamine Soudani via figshare.

Collection Method

Generated via a sophisticated simulation model of the ABB IRB 1100 industrial robot.

Freshness

Last updated 2026-05-22 12:48:58; freshness should be verified.