AI/ML based Python App and a Dashboard for Predictive Maintenance


What?

Development of an incubation AI/ML app that will predict the moment of degradation/failure of an aircraft gas turbine engine well in advance based on historical data emanating from multiple sensors.


How?

By developing machine learning strategies in order to effectively build a machine learning model involving multiple time-series data. The following key technologies were used to achieve the objectives:

  • Streamlit for UI/Dashboard
  • Heroku cloud for app deployment
  • Python for the development of predictive maintenance app
  • Explainable AI technologies to boost confidence and trust
  • DeepAR Neural Networks for building time-series models
  • Probabilistic Forecasts to make decisions with confidence intervals
  • MxNet & GluonTs frameworks for experimenting/building machine learning models
  • Azure Blob Data Storage for maintaining hundreds of time-series sensor data & AI artifacts
  • Docker for application isolation towards deployment
  • Gitlab CI/CD for continuous integration & continuous deployment


May I see it ?

The following video shows a minimal version of the predictive maintenance dashboard. The application is deployed in Heroku, and can be accessed with a credential "guest_pm" (both username and password). As the application sleeps while it is not actively used, the first load often takes some time.