Citation

Zarei E, Mohammadfam I, Azadeh M, Mirzaei-Aliabadi M. Safety Promot. Inj. Prev. (Tehran) 2018; 5(4): 201-212.

Copyright

(Copyright © 2018, Shahid Beheshti Medical University)

DOI

unavailable

PMID

unavailable

Abstract

Background and Objectives: Accidents of the industrial processes have caused irreparable economic, social, environmental and even political loses in the country. To prevent such accidents, identifying, evaluating and analyzing the causes of these incidents with new approaches are required for designing preventive strategies is a necessity. Therefore, the objective of the present study was directed toward the identifying and dynamic analyzing of the root causes of the process accidents. The Bowtie (BT) model and Bayesian Network (BN) were implemented for analyzing the accidents.

Materials and Methods: First, the accidents' scenarios were modelled quantitatively and quantitatively using the BT model, and then, the cause-consequence model of the accident scenarios was modelled in the BN using the proposed algorithm. Capabilities of the BN including, deductive, abductive reasoning and updated probability was used for dynamic analysis of the accident scenarios.

Results: The results showed that deductive reasoning for estimating the occurrence probabilities of a scenario and its consequences is more accurate by BN than BT. BN model is capable of doing probability updating of root events using the precursor accident data through abductive reasoning, taking into account conditional dependency among root events, safety barriers and modelling of common cause's failures. However, BT model does not have such capabilities.

Conclusion: In the present study, a novel, dynamic and quantitative model was introduced that allows the continuous identification and monitoring of the safety risks in the process industries. Implementing the proposed model in the process industries can significantly reduce the risk of the industrial accidents and improve the level of safety.

Language: en

Keywords

Accident; Analysis; Bayesian network; Bow tie; Dynamic; Modeling; Process; Process Industry