End-to-End Solution For IoT devices Predictive Maintenance and Management

Goal

• Our ultimate goal is to build a non-exclusive end-to-end IoT device predictive maintenance and management solution, from on-field devices all the way to a web console. The solution should be easy-to-use, flexible for all devices, and scalable to accommodate future growth.

Live

• http://www.devxia.com/crouching_tigers
• Please use http instead of https. AWS EC2 does not provide SSL certificates.

Video Presentation

• https://youtu.be/beyPiOhJXwg

Report

• report.pdf

Repository structure

• code/iot_project.ipynb: the main code for our project, including modules for:
  1. Data ETL/EDA
  2. Model Training
  3. Model Testing
  4. Anomaly Detection Fitting/Testing
  5. Plot
• code/project_iot.py: old models using pyspark machine learning pipeline and DecisionTree classification algorithm for the project. Discarded in the latest modules
• code/LinearRegression.py/LogisticRegression.py/options_regression.py: similar to project_iot.py, but using different regression algorithms
• code/AnomalyDetection.ipynb: Test code for anomaly detection using different dataset
• code/FeatureSelection.ipynb: Test code for feature selection using different dataset
• model: saved model exported from Tensorflow, deployed on Google Cloud Machine Learning Engine.
• dataset/demo_data.zip: The raw dataset we actually used for final result.
• web/node_server: Proof of concept Node.js server that simulate on-field device uptime and heartbeat
  • See more instruction at web/node_server/README.md
  • Deployed on AWS
  • Example REST API endpoint: GET http://ec2-18-236-179-13.us-west-2.compute.amazonaws.com:5000/machines
• web/frontend: Proof of concept React frontend dashboard
  • See more instruction at web/frontend/README.md
  • Deployed on http://www.devxia.com/crouching_tigers
  • Template starter by Creative Tim (https://www.creative-tim.com)
  • Commnunicate with model deployed on Google Cloud Machine Learning Engine and Node.js server on AWS.