Automotive and Turbofan Engine Sensor Data for Predictive Maintenance

Mahamudul Hasan's dataset supports a 2026 study on a unified AI-driven predictive maintenance framework. It contains two processed datasets: 19,535 real-world OBD-II automotive engine sensor observations and the NASA C-MAPSS FD001 turbofan engine dataset labeled for remaining useful life. The repository includes model outputs, feature importance values, and threshold optimization results to enable replication of the research.

Use Cases

• Training supervised learning models for failure prediction based on multivariate sensor data.
• Evaluating model performance (Logistic Regression, Random Forest, SVM, Gradient Boosting) in a predictive maintenance context.
• Analyzing feature importance for sensor signals using permutation importance and standardized coefficients.
• Optimizing decision thresholds under a cost-sensitive framework for maintenance scheduling.

Strengths

• Includes 19,535 real-world automotive engine sensor observations from OBD-II measurements.
• Provides processed model outputs, evaluation results, and multiple feature importance metrics (permutation, Random Forest, Logistic Regression coefficients).
• Documentation covers all preprocessing steps, model configurations, and evaluation procedures to ensure reproducibility.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count for the NASA C-MAPSS component is unknown, which may limit suitability assessment.
• The dataset is small (2.8 MB), which may limit the complexity of models it can support.

Provenance

Source

figshare, author Mahamudul Hasan

Collection Method

Includes real-world OBD-II measurements and a processed version of the public NASA C-MAPSS FD001 dataset.

Freshness

Last updated 2026-04-26 10:29:34; freshness should be verified.