Quantitative Dynamic Resilience Assessment of Chemical Process Units Using Dynamic Bayesian Network

Chemical process systems are becoming more and more sophisticated and complex. This makes it more challenging to identify the causes of system failures and perform the process safety analysis. In most cases, accidents happen at the level of socio-technical interactions, and the emerging hazards of these systems cannot be wholly identified and are highly uncertain. Resilient process systems can better handle uncertain hazards and failure scenarios. The dynamic resilience assessment facilitates the identification of the critical factors affecting resilience during the pre- and post-failure phases in a temporal manner. This, in turn, facilitates the identification of the root causes of the accident, timely prevention of it, and employment of useful and specific safety measures. This study has made a first attempt to use a Bow-Tie (BT) model as a tool to perform accident scenario analysis, and then the BT was converted to DBN for dynamic resilience assessment. This process facilitates the identification of the functionality state of the system and the critical factors affecting the resilience state of the system. Quantitative resilience assessment should be further enhanced for identification of the root causes of the accident on the level of socio-technical interactions and development of the specific resilience attributes to withstand or recover from the highly probable disruption factors. This approach is believed to ensure complex process system safety and functionality. The current study also investigates the opportunity of integrating Functional Resonance Analysis Method (FRAM) and Dynamic Bayesian Network (DBN) for quantitative resilience assessment to identify the highly probable disruption factors and to develop the corresponding resilience attributes. The proposed method is demonstrated through case studies on a two-phase separator of the acid gas sweetening unit: operating at standard ambient conditions( Case Study 2) and operating at harsh cold conditions (Case Study 1). The analysis of the resilience state of the process system at the worn-out conditions is also done for each case study. The study also integrates Aspen Hysys simulation for the probability of failure (POF) generation. The outcomes of this research provide a rigorous dynamic quantitative resilience analysis approach for complex process systems on the level of socio-technical interactions and a tool for identification of the critical factors or safety measures that enhance the resilience state of the chemical process system