Abnormal events involving potential safety hazards and near misses are used as early warnings and signs for the escalation of minor accidents to major accidents, which can be used to establish accident models to identify source events and correct unsafe factors in the protection system. Tailored to the process characteristics and accident features of liquefied natural gas(LNG) storage areas, the system hazard identification, prediction and prevention(SHIPP) model was improved, and a novel risk assessment modeling method integrating fault trees, Bayesian networks, and the A-star algorithm was proposed. Firstly, based on expert experience and abnormal events in the accident alarm database, a safety barrier model and fault tree were established. Then, following the chain rule, the fault tree was mapped to a Bayesian network. Finally, the improved A-star algorithm was integrated to determine the accident occurrence pathways. Research based on the LNG accident alarm database indicates that this method, compared to the traditional SHIPP model, can achieve dynamic forward risk assessment and quantify the conditional probabilities between accidents, as well as simulate the accident occurrence process when safety barriers fail in reverse. The research results can provide reasonable design and decision-making for the system safety and risk avoidance of LNG storage areas.

abnormal events  /  fault tree  /  bayesian network  /  A-star algorithm  /  SHIPP model