We’re simulating a distributed real-world production/industrial environment. Our clients run the gamut of virtually every industry and environment. While there are many different sensors and data types that we monitor, some directly, some through PLC’s, etc.
That reality also means software that runs locally at the edge, as well software running in the cloud.
While the cloud may be used to enhance and refine analytics, the real world requires that analytics run independently and locally at the edge, close to the real-time prediction streams of monitored equipment. In the cloud we see actionable alerts, which take many paths, and may integrate with many other enterprise systems.
In other words, the edge is telling you “this is everything I know right now,” while the cloud gives the aggregate view and lets management know what to be addressed.
In this demo we’re using video analytics to look for fire, smoke, and monitoring an otherwise unconnected analog gauge (“hmi”) for anomalies and events.
Google Cloud Core and Open Source Technologies in Use
- For preprocessing and training:
- Cloud Storage
- Cloud ML Engine
- Inception v3
- For runtime predictions and alerts:
- Tensorflow serving (on the edge)
- Cloud Pub/Sub
- Cloud Functions
- Stackdriver Monitoring
- Google Compute Engine Hosting
- Node.js Express with OAuth 2 Authentication