✈️ Jet Engine Health Monitoring System

My Original Kaggle Notebook

📖 Overview

In aerospace maintenance, predicting the exact "time to failure" is often less valuable than knowing the current health state of an engine. This project builds a predictive maintenance system that classifies NASA turbofan engines into three actionable risk categories based on sensor data.

Instead of raw RUL regression, this system calculates a Life Ratio (LR) to normalize engine wear across different units and predicts the safety status.

🚦 Operational Status Classifications

The model categorizes engine health into the following system for maintenance decision-making:

Label	Status	Criteria (Life Ratio)	Operational Action
🟢 0	Good	LR <= 0.6	Safe to fly. No action needed.
🟡 1	Moderate	0.6 < LR <= 0.8	Flag for upcoming maintenance.
🔴 2	Warning	LR > 0.8	CRITICAL: Immediate Grounding/Overhaul.

📊 Methodology & Engineering

• Data Cleaning: Removed 11 low-variance features (e.g., setting3, s01, s18, s19) that offered no predictive signal.
• Target Engineering: Calculated Life Ratio = Current_Cycle / Total_Cycles to normalize wear.
• Model: CatBoost Classifier chosen for its robustness with categorical data and tabular performance.
• Optimization: Used RandomizedSearchCV to tune hyperparameters (learning_rate=0.01, depth=10, l2_leaf_reg=0.01).

🚀 Key Results

• Generalization: Hyperparameter tuning increased Test Set accuracy from 67% to 69% and nearly doubled the R2 score.
• Safety First: The model successfully minimizes "False Negatives" (predicting 'Good' when an engine is actually in 'Warning' state), which is the most critical metric for aviation safety.

(Figure 1: Confusion Matrix showing high recall for 'Warning' state to prevent false negatives)

🛠️ Usage

1. Clone the repo:

   git clone [https://github.com/szxivk/jet-engine-health-monitoring.git](https://github.com/szxivk/jet-engine-health-monitoring.git)
   cd jet-engine-health-monitoring
   

2. Install dependencies: (Note: This project strictly requires scikit-learn 1.5.2 to avoid compatibility issues with the latest version)

   pip install -r requirements.txt
   

3. Run the notebook:

   jupyter notebook notebooks/turbofan_prediction.ipynb
   
   

🔗 Data Source

NASA C-MAPSS (Commercial Modular Aero-Propulsion System Simulation). Download Data Here

👨‍💻 Author's Note

This project applies concepts from the Machine Learning Specialization by DeepLearning.AI to a real-world aerospace scenario, focusing on the practical engineering implementation of predictive maintenance.

---

Created by Shahroz Khan - Aerospace Engineering & Data Science