Anticipatory Alarm System

Large industrial plants usually contain complex systems of interdependent alarms, and one fault can trigger a flood of alarms in a domino effect. But too many alerts can be detrimental to diagnosing the root cause by generating undue confusion, which often leads to unnecessary shutdowns and loss of production.

A new intelligent system that borrows from the dynamic models used in extreme weather forecasting could reduce the number of flood alarms by predicting when they might occur. By combining process models with historical data, it can estimate the likely rates of change so that operators can assess each variable and take evasive action before an alarm occurs.

The system has been tested in a depropanizer plant by simulating a variety of faults including loss of cooling water and fouling of the condenser. In every case, the alarms were predicted accurately, allowing operators to diagnose problems more quickly.