Model-predictive Safety System for Predictive Detection of Operation Hazards

This paper highlights the capability and potential of a recently-introduced method of designing modelpredictive safety (MPS) systems (Mohseni Ahooyi et al., 2016). For the first time, the method proposed a systematic utilization of process models to generate predictive alarm signals (alerts) for the detection of present and future operation hazards (OHs) in real time. An MPS system uses a process model to project in real-time the process operability status and to generate alarm signal(s) indicating the presence of a current or future OH. It triggers alarm(s) in real time when the process is unable to satisfy an operability constraint over a moving time-horizon into the future. Unlike typical existing functional safety systems that generate reactive, non-interacting alarm signal(s) when a process variable exceeds a threshold, an MPS system generates predictive alarm signals that systematically account for process nonlinearities and interactions, and alert the process personnel to imminent and potential, present and future OHs. Although the method uses the concepts of moving-horizon, model-based prediction and state estimation, it does not deal with control at all. The performance of the system is shown using a polymerization reactor example.