Path Sampling and Dynamic Risk Analysis

点击翻译按钮获取中文摘要

Chemical manufacturing processes can pose serious hazards and so safety considerations play an important role in their design. Indeed, to minimize the risk of catastrophic accidents, which can result in loss of human life or major environmental damage, extensive instrumentation such as control systems, alarms, and safety interlocks, are routinely employed in chemical processes. While such efforts are generally successful in mitigating the most common and well-understood abnormal events, it is challenging to detect the onset and mitigate the effects of infrequent and unexpected abnormal events in real-time. Importantly, such rare safety events, which have not been considered in the plant design, can lead to the most severe consequences. This project extends the prior work of this research team in adapting computational strategies developed to study the motion of molecules to uncover conditions leading to plant shutdowns or accidents resulting from rare and highly abnormal safety events. Fortunately, expensive plant shutdowns or dangerous accidents rarely occur, but consequently, little (if any) data are available to alert plant operators in sufficient time to take safety actions to circumvent them. This proposal seeks to continue developing strategies for setting alarms and carrying out safety actions more reliably in the face of unanticipated abnormal events. These strategies have the potential to prevent large financial losses, prevent serious injuries, and save lives. To advance path-sampling strategies for identifying unlikely plant shutdowns and accidents, the research team will seek new strategies to set alarms and apply safety systems in a rigorous manner. The research program will begin with well-established strategies, such as dynamic risk-analysis (DRA – developed over the past 15 years by this project's PI), for estimating the failure probabilities associated with well-known postulated abnormal events. Using noise to represent an array of rare un-postulated abnormal events, computational experiments will be carried out to relate committor probabilities (functions predicting the probability of commitment to a path to failure) to process variables for which alarm thresholds and safety systems will be created - this will enable the application of DRA, for the first time, to evaluate online the effectiveness of the safety systems by estimating process failure probabilities in response to un-postulated rare events. In so doing, the researchers will undertake the design of new multi-variable real-time alarm systems, which not only will lead to safer operation by eliminating false-negatives, but also mitigate the nuisance of false-positive alarms. As these strategies are developed, they will be tested on well-known industrial processes (e.g., steam-methane reforming (SMR) to produce hydrogen) using historical operating data provided by the PI’s long-term collaborator, Air Liquide. These data are essential to understanding how, over several years, rare paths to plant shutdowns and accidents are initiated – with final safety system responses carried out in just minutes to few hours. Gradually, application of these strategies will move from exothermic, continuous-stirred tank reactors (CSTRs) to more complex polymerization reactors, and ultimately will scale-up to integrated chemical processes with recycle.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.