This project is part of the predictive maintenance assignment focused on estimating the Remaining Useful Life (RUL) of aircraft engines using machine learning models, specifically LSTM (Long Short-Term Memory) and GRU (Gated Recurrent Unit) neural networks. The project uses NASA’s C-MAPSS dataset, which contains run-to-failure data for engines under various conditions.
The objective is to predict the RUL of engines to improve maintenance schedules and avoid unexpected failures.
-
code/
Starter_Notebook_Predictive_Maintenance_using_LSTM-1 (3).ipynb
: The main Jupyter notebook containing the entire workflow including data preprocessing, model training, evaluation, and visualization.main_rul.py
: A Python script designed to load and evaluate the saved LSTM and GRU models on test data.best_gru_model.keras
&best_lstm_model.keras
: These files contain the best versions of the GRU and LSTM models saved during training.
-
data/: Contains raw datasets in txt format required for training and testing models.
train_FD001.txt
,test_FD001.txt
, andRUL_FD001.txt
are sample files; make sure to adjust these based on the datasets you actually use.
-
visualizations/: This folder contains all the visuals I produced in my notebook. I have a total of 12 images. Each image that has an _n at the end represents the order of image creation.
CM_1.png
- Is the first image created- ....
Bonus_Attempt_12.png
- Is the last image created
-
README.md: Provides a detailed overview of the project, instructions for setup, and how to run the scripts.
If you download the jupyter notebook and run it in google colab, all will run accordingly.
To run the entire workflow, open Starter_Notebook_Predictive_Maintenance_using_LSTM-1 (3).ipynb in Google Colab or any Jupyter Notebook environment. This notebook includes:
- Data Preprocessing and Feature Engineering
- Hyperparameter Tuning for LSTM and GRU models
- Model Training and Evaluation
- Visualization of Results