SoyVitou/NASA-C-MAPSS-Turbofan-Engine

----------------------------------------------------------------------------------------------------------------------------------------------------

Remaining useful life prediction  |  Predictive maintenance  |  Digital twin research  |  Turbofan engine degradation

----------------------------------------------------------------------------------------------------------------------------------------------------

1. Project Introduction

This dataset repository provides a cleaned Hugging Face version of the NASA C-MAPSS Jet Engine Simulated Data for Remaining Useful Life (RUL) prediction.

The dataset contains multivariate time-series data from a fleet of simulated turbofan engines. each engine starts in a normal operating condition, then develops degradation over time. the main task is to predict how many operational cycles remain before engine failure.

This format is useful for building real-world style projects such as:

• Predictive maintenance systems
• Aircraft engine health monitoring
• Digital twin dashboards
• RUL prediction models
• LLM-assisted maintenance explanation systems
• Anomaly detection and degradation analysis

2. Main use cases

Use case	Description
Remaining useful life prediction	predict the number of cycles an engine can still operate before failure
Predictive maintenance	support maintenance planning before engine failure occurs
Digital twin research	create a virtual engine twin that tracks sensor state, degradation, and health status
Time-series forecasting	model multivariate sensor trends over engine life cycles
Tabular regression	train models using operational settings and sensor measurements to predict RUL
LLM maintenance assistant	use an LLM to explain sensor changes, predicted RUL, and maintenance recommendations

3. Repository Structure

NASA-C-MAPSS-Turbofan-Engine/
├── README.md
├── data/
│   ├── train_FD001.txt
│   ├── test_FD001.txt
│   ├── RUL_FD001.txt
│   ├── ...
│   └── RUL_FD004.txt
└── processed/
    ├── FD001/
    │   ├── train.parquet
    │   └── test.parquet
    ├── FD002/
    │   ├── train.parquet
    │   └── test.parquet
    ├── FD003/
    │   ├── train.parquet
    │   └── test.parquet
    └── FD004/
        ├── train.parquet
        └── test.parquet

• data/ contains the original raw .txt files.
• processed/ contains cleaned .parquet files for the Hugging Face Dataset Viewer.
• each processed file includes readable sensor names and an added RUL column.

4. Dataset Description

The original NASA C-MAPSS dataset consists of multiple multivariate time series. each time series represents one engine from a fleet of engines of the same type.

each engine has unknown initial wear and manufacturing variation. this is treated as normal variation, not as a fault. during operation, the engine begins normally and later develops a fault. in the training set, the fault grows until failure. in the test set, the time series ends before failure, and the true remaining useful life is provided separately.

each row is a snapshot from one operational cycle.

5. Dataset Configurations

Config	Train trajectories	Test trajectories	Operating conditions	Fault modes	Fault type
FD001	100	100	1	1	HPC degradation
FD002	260	259	6	1	HPC degradation
FD003	100	100	1	2	HPC degradation, fan degradation
FD004	248	249	6	2	HPC degradation, fan degradation

6. File Types

File type	Description
train_FDxxx.txt	full run-to-failure engine trajectories
test_FDxxx.txt	partial engine trajectories ending before failure
RUL_FDxxx.txt	true RUL values for the final cycle of each test engine
train.parquet	cleaned training split with column names and computed RUL
test.parquet	cleaned test split with column names and computed RUL

7. Column Dictionary

The original files contain 26 numeric columns: unit ID, cycle, 3 operational settings, and 21 sensor measurements.

Column Description Unit unit_id engine unit ID - cycle time cycle cycles setting_1 operational setting 1 - setting_2 operational setting 2 - setting_3 operational setting 3 - T2 total temperature at fan inlet °R T24 total temperature at LPC outlet °R T30 total temperature at HPC outlet °R T50 total temperature at LPT outlet °R P2 pressure at fan inlet psia P15 total pressure in bypass-duct psia P30 total pressure at HPC outlet psia Nf physical fan speed rpm Nc physical core speed rpm epr engine pressure ratio P50/P2 -

Column Description Unit Ps30 static pressure at HPC outlet psia phi ratio of fuel flow to Ps30 pps/psi NRf corrected fan speed rpm NRc corrected core speed rpm BPR bypass ratio - farB burner fuel-air ratio - htBleed bleed enthalpy - Nf_dmd demanded fan speed rpm PCNfR_dmd demanded corrected fan speed rpm W31 HPT coolant bleed lbm/s W32 LPT coolant bleed lbm/s RUL remaining useful life cycles dataset dataset subset name - split train or test split -

8. Remaining Useful Life Target

8.1. Training split

In the training files, each engine runs until failure. the RUL value is computed as:

RUL = max_cycle_for_engine - current_cycle

example:

unit_id	cycle	max_cycle	RUL
1	1	192	191
1	2	192	190
1	191	192	1
1	192	192	0

8.2. Test split

In the test files, each engine stops before failure. the final true RUL is provided in the corresponding RUL_FDxxx.txt file.

the row-level test RUL is computed as:

RUL = max_test_cycle_for_engine - current_cycle + final_RUL

9. Loading the dataset

from datasets import load_dataset

dataset = load_dataset(
    "SoyVitou/NASA-C-MAPSS-Turbofan-Engine",
    "FD001"
)

print(dataset)
print(dataset["train"][0])

available configs:

configs = ["FD001", "FD002", "FD003", "FD004"]

10. Example project: LLM + digital twin

This dataset can be used to create an engine digital twin with an LLM assistant.

sensor data
    ↓
RUL prediction model
    ↓
engine digital twin state
    ↓
LLM explanation assistant
    ↓
maintenance recommendation

example digital twin state:

{
  "engine_id": 1,
  "current_cycle": 120,
  "predicted_rul": 38,
  "health_status": "warning",
  "main_degradation_signals": ["T50", "P30", "Nc"]
}

example user question:

why is engine 1 in warning state?

example LLM answer:

engine 1 is in warning state because the predicted remaining useful life is low, and several sensor trends indicate engine degradation. maintenance inspection is recommended before the next operating window.

11. Suggested machine learning tasks

Task	Target
RUL regression	predict RUL
health stage classification	classify normal, warning, or critical
anomaly detection	detect abnormal sensor behavior
degradation trend analysis	identify sensor changes over engine cycles
cross-condition generalization	train on one condition and test on another
digital twin monitoring	update virtual engine health state over time

12. Original source

this dataset is derived from the NASA C-MAPSS Jet Engine Simulated Data.

• source: CMAPSS Jet Engine Simulated Data
• publisher: NASA Prognostics Center of Excellence, PCoE
• theme: aerospace
• resource type: dataset

13. Reference

A. Saxena, K. Goebel, D. Simon, and N. Eklund,
“Damage Propagation Modeling for Aircraft Engine Run-to-Failure Simulation,”
Proceedings of the 1st International Conference on Prognostics and Health Management, PHM08, Denver, CO, October 2008.

14. Citation

if you use this dataset, please cite the original NASA C-MAPSS dataset and the reference paper above.

15. Note

this repository provides the dataset in a Hugging Face friendly structure. the raw .txt files are preserved, and the processed .parquet files are added to make the dataset easier to preview, load, and use in machine learning workflows.