To start the app.py file, you should use the Streamlit command:

streamlit run app.py

---

Prognos AI

NASA C-MAPSS Turbofan Engine RUL Prediction

This repository contains a machine learning project focused on predicting the Remaining Useful Life (RUL) of aircraft turbofan engines using the NASA C-MAPSS dataset. The project includes a comprehensive research phase where multiple deep learning architectures were evaluated to find the best-performing model for real-time health monitoring.

📊 About the Dataset

The C-MAPSS (Commercial Modular Aero-Propulsion System Simulation) dataset is a standard benchmark for predictive maintenance. It consists of four sub-datasets (FD001, FD002, FD003, FD004) reflecting different operating conditions and fault modes:

• Data: Multivariate time-series data from 21 sensors and 3 operational settings.
• Target: Remaining Useful Life (RUL) in flight cycles.
• Complexity: Ranges from a single operating condition/fault mode (FD001) to six operating conditions and two fault modes (FD004).

🚀 The Multi-Model Approach

My approach involved an extensive research phase where I trained and compared several deep learning architectures to capture the temporal dependencies in sensor data:

1. CNN (1D Convolutional Neural Networks): Used for feature extraction from the sensor sequences.
2. LSTM (Long Short-Term Memory): To capture long-term dependencies in the time-series data.
3. Stacked LSTM: Multiple LSTM layers for deeper temporal feature learning.
4. Bi-LSTM (Bidirectional LSTM): To process sequences in both forward and backward directions.
5. GRU (Gated Recurrent Units): A computationally efficient alternative to LSTM.
6. CNN + LSTM Hybrid: A combination model using CNN for local feature extraction followed by LSTM for temporal modeling.

📈 Evaluation Metrics

The models were evaluated using the following metrics to ensure high precision in safety-critical aviation environments:

• RMSE (Root Mean Squared Error): To measure the average prediction error magnitude.
• MAE (Mean Absolute Error): To understand the average absolute deviation from true RUL.
• R² Score (Coefficient of Determination): To assess how well the models explain the variance in the engine's degradation.

🛠️ Local Installation & Setup

To run the AeroHealth Manager dashboard on your local machine, follow these steps:

1. Clone the Repository

git clone https://github.com/your-username/prognos-ai-cmapss.git
cd to the folder

2. Install Dependencies

Ensure you have tensorflow, streamlit, and pandas installed:

pip install streamlit pandas numpy tensorflow joblib plotly scikit-learn

3. Run the Dashboard

Launch the Streamlit application to visualize individual engine inspections and fleet health heatmaps:

streamlit run app.py

🖥️ Project Structure

• research.ipynb: Initial data exploration and preprocessing.
• FD001(1).ipynb to FD004model.ipynb: Comprehensive training scripts for the various model architectures.
• app.py: Streamlit-based dashboard for real-time RUL prediction and fleet management.
• *.h5: Saved trained models for different fleets.
• scaler_*.pkl: Pre-trained scalers for data normalization.