Process accident prediction using Bayesian network based on IT2Fs and Z-number: A case study of spherical tanks

Area of Science:

• Risk Assessment and Management
• Fuzzy Logic in Engineering
• Bayesian Networks for Safety Analysis

Background:

• Traditional risk assessment methods often struggle with uncertainty inherent in expert judgments and data.
• Dynamic risk assessment is crucial for evolving industrial environments and accident prevention.
• Integrating fuzzy logic and Bayesian networks can enhance the handling of imprecise and uncertain information.

Purpose of the Study:

• To propose a novel dynamic risk assessment methodology using Bayesian networks (BN) with fuzzy data.
• To decrease uncertainty in risk assessment by integrating Interval Type-2 Fuzzy Sets (IT2FS) and Z-numbers.
• To provide a more reliable framework for accident prevention performance in industrial settings.

Main Methods:

• Construction of a bow-tie diagram using the System Hazard Identification, Prediction, and Prevention (SHIPP) approach, Top Event Fault Tree, and Barriers Failure Fault Tree.
• Quantification of expert opinions on prior probabilities using IT2FS and Z-numbers to reduce uncertainty.
• Calculation of posterior probabilities of critical basic events and barrier failures using Bayesian updating with beta distribution and recorded data.

Main Results:

• The methodology successfully calculated the posterior probability of barrier failure and consequences over a 5-year period.
• The IT2FS-Z approach demonstrated a shallower upward trend in consequence probability compared to IT2FS alone, due to expert confidence levels.
• Observed differences in results were more pronounced with higher variance (10-4 vs. 10-5), highlighting the impact of uncertainty quantification.

Conclusions:

• The proposed BN-based fuzzy dynamic risk assessment method effectively reduces uncertainty compared to traditional approaches.
• The integration of IT2FS and Z-numbers enhances the reliability of risk assessment by incorporating expert confidence.
• This study offers industry managers a more comprehensive and dependable tool for optimizing accident prevention strategies.