/Predictive-Maintenance

Developed at the IEEE CS SPIT Aeravat 1.0 AI Hackathon

Primary LanguageJavaScript

AI Kavach: Predictive Maintenance for Industrial Equipment using Machine Learning

Project Overview

AI Kavach is a predictive maintenance solution developed for industrial equipment utilizing machine learning techniques. The project aims to predict equipment failures and maintenance needs in advance, enabling proactive maintenance scheduling and minimizing downtime. This project was developed as part of the IEEE CS SPIT AERAVAT 1.0 AI Hackathon, where it won 1st prize in the machine learning domain.

Problem Statement

Predictive Maintenance for Industrial Equipment Using Machine Learning.

Objectives

  • Develop a predictive maintenance solution for industrial equipment.
  • Predict equipment failures and maintenance needs in advance.
  • Enable proactive maintenance scheduling to minimize downtime.

Dataset Source

The dataset used for this project is sourced from NASA CMAPSS Jet Engine Simulated Data. It consists of multiple multivariate time series, each representing data from a different engine in a fleet of engines of the same type.

Data Preprocessing

  • Feature Reduction
  • Handling Missing Values
  • Applying Min-Max Scaler

Model Training (RUL Prediction)

  • XG Boost
  • Random Forest Regressor
  • Decision Tree Regressor

React Interactive Web App Features

  • User Authentication and Access Control
  • Real-time Sensor Data Visualization and RUL Prediction
  • Alerting and Notification System through Web Browser and Telegram
  • Maintenance Scheduling through Personalized Calendar
  • Static Plots and Dashboard
  • Feedback and Reporting

Technology Stack

  • React.js
  • Flask
  • MongoDB
  • Node.js
  • Telegram
  • React Chart.js

Model Performance

XG Boost achieved the highest R2 score of 0.65, indicating strong predictive capability.

Conclusion

  • XG Boost demonstrated the highest predictive capability.
  • The React Interactive Web App offers comprehensive features for real-time monitoring, prediction, scheduling, and notification.
  • Continuous refinement and integration of user feedback enhance the predictive maintenance system for industrial applications.

For more details, refer to the respective folders in the repository.

Contributors:

  • Aditya Potdar
  • Pushkar Waykole
  • Harshit Singh
  • Sarvagya Singh

Note: Please find the detailed code, documentation, and results in the repository. Contributions and feedback are welcomed.