IoT Analytics for predictive maintenance

Summary

Predictive maintenance is the killer IoT big data application for large manufacturing powerhouses. Companies ranging from GE to Siemens to Hitachi all incur significant costs caused by unscheduled machinery maintenance. Most of these devices (which includes construction equipment, medical equipment, aircraft engines, windmill farms etc.) are now equipped with sensors which produce large streams of data which are sent to the cloud for both stream analytics that allow immediate actions to be taken as well as deep analytics on historical data to improve asset performance and help with predictive maintenance

Challenge

  • Ingest data at a very high rate, millions of events per second from a large number of sensors
  • Build counters and intelligent KPIs and store them into tables
  • Identify faults in the network in real time
  • Perform root cause analysis on outage events
  • Isolate and notify upstream systems of outage events
  • Stream analytics on ingested data including joins across streams
  • Maintain interactive latency on time windows to allow actions to be taken

Solution

SnappyData is being used to receive stream data from a number of medical devices and then inspected to figure out machine fatigue information. After stream processing completes, the data is written to column tables which are exposed as Spark data frames and these data frames are run through Spark machine learning libraries

Snappy Capabilities in Use

  • Optimized data ingestion
  • Use of both row and column tables
  • Stream processing
  • Machine Learning

The Spark Database

SnappyData is Spark 2.x compatible and open source. Download now