Defect prevention with predictive maintenance

Solution Idea

If you'd like to see us expand this article with more information (implementation details, pricing guidance, code examples, etc), let us know with GitHub Feedback!

Learn how to use Azure Machine Learning to predict failures before they happen with real-time assembly line data.

This solution is built on the Azure managed services: Azure Stream Analytics, Event Hubs, Machine Learning Studio, Azure Synapse Analytics and Power BI. These services run in a high-availability environment, patched and supported, allowing you to focus on your solution instead of the environment they run in.

Architecture

Azure SQL DW Machine Learning (Real time predictions) Power BI ALS test measurements (Telemetry) Event Hub Stream Analytics (Real time analytics) Dashboard of predictions/alerts Realtime data stats, Anomaliesand aggregates Realtime event and predictions

Components

  • Azure Stream Analytics: Stream Analytics provides near real-time analytics on the input stream from the Azure Event Hub. Input data is filtered and passed to a Machine Learning endpoint, finally sending the results to the Power BI dashboard.
  • Event Hubs ingests raw assembly-line data and passes it on to Stream Analytics.
  • Machine Learning Studio: Machine Learning predicts potential failures based on real-time assembly-line data from Stream Analytics.
  • Azure Synapse Analytics: Synapse Analytics stores assembly-line data along with failure predictions.
  • Power BI visualizes real-time assembly-line data from Stream Analytics and the predicted failures and alerts from Data Warehouse.

Next Steps