xuejun1108 / predictive_maintenance

Predictive Maintenance have been done on the NASA's Turbofans to predict the Remaining useful life of the engines based upon the input from several sensors.

Predictive Maintenance

Table of Content

• Demo
• Synopsis
• Appendix
• Links
• Directory Tree
• Deep Learning Model Structure
• Features
• Run Locally
• License
• Technology Used

Demo

Synopsis

Predictive Maintenance have been done on the NASA turbofan engine dataset to predict the Remaining useful life of the engine based upon the input from several sensors, several machine learnings models have been used to predict the RUL of the engine, and the best model which have performed well on the training data have been selected for hyperparameter tuning to increase the accuracy of the model, and finally the pre-trained model have been used for the web application to predict the real world data from the user.

So based on the given data,

Independent variable (X) : All the sensor values

Dependent variable (Y) : Remaining useful life (RUL)

Based on the given data we can identify this as a regression problem, so we can use various machine learning problems to solve this problems which are as follows:

• Linear regression
• Lasso regression
• Ridge regression
• Decision tree regressor
• Random forest regressor
• Gradient boost
• Support Vector regressor
• Artificial Neural Network
• Recurrent Neural Network
• Long short term memory

I have declared this problem with random forest regression, since other models doesn't gave much accuracy.

Machine learning model : Random forest regressor (sklearn)

Data preprocessing : Pandas

Data visualization : Matplotlib, Seaborn

Web framework : Streamlit

Model deployment : Streamlit server

Appendix

The requirement for developing this model is present in the requirements.txt file.

The development of the model is present in the main.ipynb file.

The streamlit framework for the web app development is made in the app.py file.

Links

• Dataset link : https://www.kaggle.com/c/predictive-maintenance/data

• Github link : https://github.com/Vedakeerthi/Predictive_Maintenance

• Web Application link : https://vedakeerthi-predictive-maintenance-app-zuy0jx.streamlit.app/

Directory Tree

├── LICENSE
├── README.md
├── ann.png
├── ann2.png
├── ann_architecture.png
├── app.py
|── lstm.png
├── main.ipynb
|── model_result.gif
├── requirements.txt
|── rnn.png
├── y_pred.csv

Deep Learning Model Structure

Model	Structure
ANN
RNN
LSTM

Features

• Live prediction analysis.
• Fullscreen mode supports in mobile, pc.
• Cross platform can be used on multiple operating system.

Run Locally

Clone the project

  git clone https://github.com/Vedakeerthi/Predictive_Maintenance.git

Install dependencies

  pip install -r requirements.txt

Start the server

  streamlit run app.py

Run the app on server by the local link provided

License

Technology Used