Optimal Alarm System Design and Implementation(ARC-16561-1)

The innovation described herein is an alarm system that can robustly predict a level-crossing event that is specified over a fixed prediction horizon. Given a specific modeling paradigm that can be learned using appropriate data-driven or machine learning techniques, an optimal alarm system can be designed to elicit the fewest false alarms for a fixed detection probability. When the modeling paradigm is a simple stationary linear dynamic system driven by white Gaussian noise, it is easy to use the well-known Kalman filter to enable value prediction of future process values. During implementation, these value predictions can subsequently be used to enable the prediction of associated level-crossing events that may occur in the future, using a parameter that was selected to achieve robustly optimal performance during the alarm system design stage.