Dynamic Quantitative Risk Analysis Of Petrochemical Plants

During the production,risk continues to change.Therefore,it is necessary to develop dynamic quantitative risk analysis system to monitor and predict risk,then enterprise could depend on analysis result to make safety-related decision.Precursor events means the sum of abnormal events including incidents,near-miss and alarms,which indicate the incident probability.Compared with other quantitative risk analysis methodology,Bayesian network has advantages of dynamic updating and representing fuzzy logic.Therefore,Bayesian network was used to quantify and update incident probability.Identify incident cause and consequences in Bow-tie model first,and transfer it into Bayesian network.Precursor incident information and prior probability is imported to calculate incident posterior probability,which represented the probability change trend over time.When using Bow-tie model to analyze human error,deficiency is usually caused by fuzzy logic,time-consuming quantitative analysis,so this paper categorized three human error types based on their contribution role in bow-tie.SPAR-H,SRK cognitive framework and maximum entropy model are the basis of structuralized human reliability quantitative method.Analysts are able to calculate human reliability in bow-tie model with high efficiency.Dynamic quantitative risk system turns to be more comprehensive.Considering equipment parts and human error,comprehensive dynamic risk assessment system was developed with risk matrix as assessment principle.Isomerization rectifying column overpressure scenario was selected as a case to be applied in this methodology.The results showed throughout the life time of the process unit,the increase of precursor incidents leads to significant increase of facility failure likelihood and incident risk.Moreover,at the 5~thh year of since first startup,the probability of vapor cloud explosion exceeded acceptable level reaching to high risk level.Sensitive analysis indicated that column pressure remote meter and high liquid level alarm had an impact on incident occurrence.For this reason,it is recommended that industry managers should pay attention to precursor incident information and take appropriate actions to reduce the number of precursor incidents,for example,improving inspection and maintenance plan,replacing key facilities and investigating incidents thoroughly and etc.