tortorish/cmapss-release

Code implementation for "A dual attention LSTM lightweight model based on exponential smoothing for remaining useful life prediction"

This project provides all the code for data-preprocessing, model construction and model evaluation. Checkpoints for the trained models are also included for further study. Pytorch is used to construct the DA-LSTM model.

File Structure

├───dataset
│   ├───__init__.py
│   └───preprocessing.py
├───model
│   ├───__init__.py
│   ├───Attention_modules.py
│   └───LSTM_Attention.py
├───scipt
│   ├───parametric_statistics.py
│   ├───test_model.py
│   └───train_model.py
├───trials
│   ├───model_FD001.pkl
│   ├───model_FD002.pkl
│   ├───model_FD003.pkl
│   └───model_FD004.pkl
├───utils
│   ├───__init__.py
│   └───functions.py

• The CMAPSSData/ folder contains the official CMAPSS-dataset.
• The dataset/ folder and utils/ folder contain the functions for data preprocessing and model training respectively.
• The trials/ folder contains the checkpoints for the trained models on the four sub-datasets.
• The model/ folder contains the components for the DA-LSTM model.

Running Guide

You can either load the models for testing or training your own models from scratch. Take notice that there are four sub-datasets in the CMAPSS Dastset. Consequently, certain parameters need to be changed according to the subdatasets.

Running Tests

You can load the trained checkpoints directly by running test_model.py in the scipt/ folder.

cd [parent folder of the project]/CMAPSS-release/scipt/
python .\test_model.py

Training Models

If you need to retrain the models, you can run train_model.py in the scipt/ folder. The trianed checkpoints are save in the \trials directory.

cd [parent folder of the project]/CMAPSS-release/scipt/
python .\train_model.py

Evaluating Models

If you need to evaluate number of parameters in the model, you can run parameter statistics.py in the scipt/ folder.

cd [parent folder of the project]/CMAPSS-release/scipt/
python .\parameter statistics.py