Predix Analytic Correlation Engine - PACE

PACE

Inspiration

GE has brought the concept of industrial internet and brilliant manufacturing to us all, however, IT assets are still using rudimentary monitoring to identify possible issues. We must become more proactive in detection before issues cause an incident.

What it does

PACE - Predix Analytic Correlation Engine proactively correlates IT incidents with analytics based on root cause analysis. Using Predix micro-services it traces the root cause of failures using Time Series data correlated with both the Asset model and with the Service NOW Incident management to both create new, and group existing Service Now tickets to simplify support. PACE provide a direct roadmap to technicians by weighting the overlap of devices on multiple incident paths, saving investigation time and duplication of efforts.

Data points from sensors are streamed using the Predix Machine to the Predix APM Time Series database. The Asset data was loaded into the Predix APM Asset Service based upon the Asset Model in the Service Now Configuration Management Database. The Gateway application allows us to query data from the Asset Service and as needed correlate with the Time Series Database to identify possible problem devices. raise a Service now request if a problem is identified.

How we built it

We simulated the asset environment using data from the Edison Machines connected to various sensors. Ingesting the data into APM time series. We produced code that mirrored the Service Now Asset model into APM asset. We created a rest API that allowed Service Now to push new incidents into our Predix micro-service. APM asset allowed us to quickly iterate the relationships of the asset provided on the Service Now incident. With each ingestion incident we correlate the relationships of current incident assets to previous known problem devices. When a threshold of incidents are realized for a particular asset, a health check is performed on the asset. If an asset fails a health check the information is fed into Service Now through a custom API to create a parent ticket and alerts all child tickets of the issue, and initiates a workflow to begin fixing the identified problem.

Challenges we ran into

Like many others discovered, Predix APM was difficult due to bugs on the platform. APM time-series ingestion was down for periods. There was heavy latency using the APM dashboard that often impeded progress.