Dynamic Risk Analysis: A Case Study On An Industrial Scale Chemical Operation at a Major Petroleum Refinery

In the chemical process industries (CPIs), the extents of human and financial losses due to accidents are staggering.  Incident investigations have revealed that every major accident is preceded by many incidents or minor accidents with low or limited severity.  These are often indicators of potential accidents known as “near-misses”.  While many near-misses do not lead to serious consequences, they always deteriorate the operational productivity of the process.  Also, recent studies have pointed to the importance of identifying near-misses to predict the probability of accidents (Meel and Seider, 2006) and also managing them to reduce the potential of accidents (Phimister et al., 2003; Cooke and Rohleder, 2006). 

The key premise of this work is that patterns in abnormal events (which occur when process variables drift out of their normal operating ranges) can be viewed as near-misses and provide information on both successful and unsuccessful actions by safety systems (installed to prevent abnormal behavior), directly or indirectly associated with the process.   To obtain data on the abnormal events, which have different criticalities, for all of the measured variables, the dynamic and extensive databases associated with their distributed control systems (DCSs) and emergency shutdown systems (ESDs), and incident and accident records, are utilized.  These data are recorded to monitor the dynamics of the process.   The DCS database contains alarm frequency data; that is, alarm identity tags for the variables, alarm types (low, high, high-high, etc.), times at which the variables cross the alarm thresholds (in both directions), and variable priorities, among others.  The ESD database, of greater consequence, contains trip (shutdown) data, etc. 

New incident-tracking and learning methods, based on the above databases, are proposed to develop leading indicators that help identify an escalation in the probability of occurrence of critical, undesirable events like trips and accidents – permitting management and plant operators to take appropriate actions to prevent upset situations.  They are also used to estimate and predict the failure probabilities of the safety systems, and more importantly, the probabilities of the occurrence of trips, using Bayesian theory (Pariyani et al., 2009).  Furthermore, new key performance indicators over time, which effectively utilize the abundant information, are proposed to assess and improve the safety and operational performance of the processing units.

These methods have been successfully applied to an industrial scale chemical operation at a major overseas petroleum refinery that processes more than 250,000 barrels of oil per day.  Results and conclusions from this analysis are presented.    

 

References

1.   Phimister J. R., U. Oktem, P. R. Kleindorfer, and H. Kunreuther, “Near-miss incident management in the chemical process industry,” Risk Analysis, 23, 445-459 (2003).

2.   Cooke D.L., and T. R. Rohleder,  “Learning from incidents: From normal accidents to high reliability.” System Dynamics Review, 22(3), 213 -239 (2006).

3.   Meel, A., and W. D. Seider, “Plant-specific dynamic failure assessment using Bayesian theory,” Chem. Eng. Sci., 61, 7036-7056 (2006).

4.   Pariyani A., W. D. Seider, U. G. Oktem, and M. Soroush, “Synergistic enhancement and assessment of process safety and product quality,” in preparation (2009).