Chemical manufacturing processes have the potential for catastrophic accidents – those involving loss of human life or major environmental impacts. The numerous ways through which a complex chemical process can fail is limited neither by our experience, nor by our imagination; consequently, socalled un-postulated safety-event scenarios pose a major concern. By combining extensive chemical process data, process modeling and state-of-the-art path sampling algorithms, this work aims to uncover rare and un-postulated process safety events, to identify secondary process variables that can portend such events (so they can be incorporated in alarm systems), and to characterize their rates of occurrence. Transition path sampling (TPS) has been a widely used by the nano-scale modeling community. Its goal is to identify rare-event trajectories that are worthy of investigation; such as the dissociation of a weak acid in aqueous solution or the flipping of phospholipids in a lipid bilayer. The types of rare-events that are best handled by transition path sampling are those in which the time-scale of the event is far shorter than the time-scale over which the event may take place. This ratio of time-scales is similar to a safety-event in a chemical manufacturing process; the event may occur once every five years of operation, while the event takes place over just a five-hour period. To simulate hundreds of such events, a typical process simulation with expected disturbances would have to be simulated for 500 years! It would be computationally beneficial to use a sampling technique that focuses on the safety trajectories of interest, and excludes the vast number of safe operating trajectories. The details of this Monte-Carlo approach are outlined, and the ability to discover un-postulated rare events demonstrated. The safety trajectories investigated with TPS are used to build event-specific alarms, whose simulated false-positive and false-negative rates are explored. Extensive process, control, and alarm data are essential and are highlighted throughout the presentation. A dynamic model of an air separation unit (ASU) is the test bed for the TPS technique.
展开▼
