Turbofan RUL Predictor

Predict Engine Life Before Failure

Upload a turbofan engine's sensor history and get a data-driven estimate of how many operating cycles remain before failure — with an AI explanation of what's driving the prediction.

XGBoost SHAP Interpretability NASA CMAPSS FD001 21 Sensors

Upload Sensor Data

CSV with columns: cycle, sensor_2, sensor_3, sensor_4, sensor_7, sensor_8, sensor_9, sensor_11 … sensor_21

📂

Drop CSV file here

or click to browse

Load sample engine data to try in-browser, or download the sample CSV to test the upload flow.

📡

Results will appear here

Upload sensor data and run prediction

How It Works

Sensor Ingestion

21 sensors per cycle — temperatures, pressures, fan/core speeds, fuel flow — from the NASA CMAPSS FD001 dataset.

Feature Engineering

Rolling mean and standard deviation over 30-cycle windows. Constant-variance sensors dropped. Degradation indicators derived from domain knowledge.

XGBoost Model

Gradient boosted regression trained on 100 run-to-failure engine trajectories. Engine-grouped cross-validation prevents data leakage.

SHAP Explanation

TreeSHAP explains each prediction — which sensors are most responsible and in which direction. Tied back to known turbofan degradation physics.