Anyone can visualize sensor data in Predix and APM, but often business value comes from compound metrics derived from multiple sensor values. Our inspiration was the GE LM 6000 Marine Gas Turbine, a state of the art power turbine. Our customer research led us to understand the importance of turbine efficiency, defined as a combination of fuel flow, power output and fuel quality, in understanding downtime both in realtime and historically.

What it does

Our suite of Predix/APM extensions creates a realtime and historical dashboard within APM to monitor low efficiency events, and determine which of the factors caused them.

DEMO Username: cream-user Password: eba0d34a6f9a

How I built it

  1. Intel Edison running Predix Machine
  2. Edge Analytics streaming to Predix Timeseries
  3. Edge Analytics streaming alerts to Predix APM
  4. Cloud metrics calculating turbine efficiency in realtime through websocket
  5. Cloud metrics calculating historical event frequency via a node.js microservice
  6. APM microapp providing interactive heatmap and timeseries visualizations
  7. Coca-Cola

Challenges I ran into

  1. Spinning up APM example apps (documentation and seed apps)
  2. Timeseries storage
  3. Rendering realtime and historical d3.js within angular/polymer framework
  4. Abstraction of Predix Machine code
  5. Response time of APM REST calls often ~ 10s or more

Accomplishments that I'm proud of

  1. 100% live data - no fake data!!
  2. Creating Edge Analytics and streaming to Timeseries/APM
  3. Leveraging snazzy d3.js heatmap for visualization
  4. Including pre-existing px-components

What I learned

  1. Re-using components is great in theory but introduces constraints
  2. Frontend dependency management can get complex mixing multiple frameworks
  3. Focus on the customer's problem to drive success
  4. Move fast and pivot when needed

What's next for Predix Decoded

  1. Sleep
  2. Eat
  3. Pool
Share this project: