Development of risk assessment model for petrochemical storage using bayesian approach

Abstract

Petrochemical products are inherently hazardous. Accidents related with these products are very frequent these days in Bangladesh. If the storage of this highly flammable and explosive hazardous chemical is improper and unsafe, it can cause events like Boiling Liquid Expanding Vapor Explosion, Vapor Cloud Explosion, Pool Fire etc. resulting in casualties, property damage and environmental pollution. An appropriate and effective method of hazard identification, evaluation and prevention is required to avoid any catastrophic incident in the process industries. Numerous approaches i.e. qualitative methods, quantitative methods and combination of two or more methods had been being proposed to facilitate the risk management in petrochemical storage. For this research a quantitative methodology: Bayesian Network (BN) Analysis was chosen on priority basis over the qualitative ones for the case study of i-Butane storage tank. The objective ofthis research was to identify and analyze hazards and assess risks in i-Butane storage facilities using Bayesian Network Analysis. Prior to this, a systematic HAZOP was performed with the help of piping and instrumentation diagram and risk had been also assessed quantitatively by Bow-Tie method. Bow-Tie diagram was constructed by combining fault tree and event tree. The top event was i-Butane release from the storage tank. Four protective barriers were introduced here: Release prevention barrier, Dispersion prevention barrier, Ignition prevention barrier and Escalation prevention barrier. Based on the failure or success of safety barriers, five types of consequences were considered: Safe, Near Miss, Mishap, Incident and accident. Frequency of i-Butane release from storage tank was found 9.985x10-7/year. Bayesian network analysis was performed using Bayesian theorem with the help of a trial version of a software named: Agena Risk. In this model, the fault trees and event trees from previous Bow-Tie were put into the Bayesian inference based software to update the event frequencies. The fault trees and event trees had been updated by implementing additional causes for the occurrence of top event and barrier failure. As a result frequency of i-Butane release from storage tank was increased to 1.71001x10-5/year. Larger frequency of the top event contributed to more consequences such as: Catastrophe. It was compensated by two more barriers: 1) Emergency Management and Damage Control Barrier and 2) Human Factor Barrier. Therefore, this study proposed to develop a Bow-Tie model and fault trees and map them into a Bayesian Network which produces more reliable results and an easy approach to dynamic update of risks.