In many ways you could say that IT + OT = IoT
Raw data on its own is just that — for example , a giant table of temperature readings recorded every second that say 107 degrees . Not very useful on its own . Certainly a blip that suddenly shows 200 degrees and then returns back to 107 degrees might be a concern , or it could be a false positive . But what happens if 107 becomes 130 for a few hours or is trending upward ? In general , the answer requires looking across the data to determine what it means . In other words , deriving context from the data .
Predictive maintenance is the art and data science associated with ingesting these massive pools of raw data , using them to train machine learning algorithms to derive context , and anticipating part failures with high statistical probability , and making recommendations on when parts should be proactively replaced before downtime occurs .
And now that the four-walled barriers between factory floors are coming down , we are beginning to see the migration of both historical and real-time data to the cloud for analysis -- which in turn will enable better maintenance predictions in the future . And all those years or decades of “ dark data ” languishing in the data historian suddenly becomes useful in the cloud , where it is finally cost-effective to analyze .
Autonomy and Control
One of the greatest perceived barriers to adoption of IIoT on the factory floor or production line is based on security concerns ( anyone remember Stuxnet ?). At the same time , there is a practical concern that is based on latency .
8 | THE DOPPLER | FALL 2017