J-2X Rocket Engine: Simulated Leak Detection Data Using Decision Trees

NASA Ames Research Center, Pratt & Whitney Rocketdyne, and NASA Stennis Space Center researchers created simulated data for fault detection in the J-2X rocket engine. The dataset includes 56 simulations from a high-fidelity model with varying leak sizes, locations, and onset times, plus sensor noise and turbine degradation. A C4.5 decision tree was trained on 11 simulations and tested on 45, achieving low false alarm and missed detection rates.

Use Cases

• Training interpretable fault classifiers based on simulated sensor data with noise and degradation.
• Benchmarking decision tree performance for fault isolation based on labeled examples of nominal and leak conditions.
• Researching data-driven health management systems for rocket engines based on high-fidelity simulation outputs.

Strengths

• Data includes 56 high-fidelity simulations with controlled variations in leak size, location, and onset time.
• Simulations incorporate realistic elements like sensor noise and gradual turbine efficiency degradation.
• Decision tree model demonstrated low false alarm and missed detection rates on 45 test simulations.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• The model showed confusion between two of the five leak locations, indicating potential similarity in fault signatures.

Provenance

Source

National Aeronautics and Space Administration (NASA)

Collection Method

Data generated from 56 simulations using the Detailed Real-Time Model (DRTM), a high-fidelity simulator developed at Pratt & Whitney Rocketdyne.

Time Range

null

Freshness

Last updated 2026-03-13 20:28:59.974735; freshness should be verified.

Geography

null