Telemetry from industrial assets is captured by historians, read by the AWS IoT SiteWise connector running on a local AWS IoT Greengrasscore, and published to the cloud in AWS IoT SiteWise. Telemetry is exported to an Amazon Simple Storage Service (Amazon S3) bucket, where it is prepared for machine learning training and then live inferencing. Custom code deployed to an AWS Lambda function prepares the data to be read by Amazon Lookout for Equipment for training and inference. Asset performance models are trained per asset in Amazon Lookout for Equipment. Models are made available for batch inferencing using a scheduler. Combined with raw telemetry, batch inference results are stored as a new property in AWS IoT SiteWise. Using Amazon QuickSighton the combined exported data, the operations teams can triage underperforming assets in a dashboard. Operations teams perform human-in-theloop validation on alerts to
