A novel algorithm for model uncertainty reduction in trapezoidal fuzzy fault tree risk assessment

Intelligent risk assessment in complex systems increasingly relies on methods like trapezoidal fuzzy fault trees. However, conventional techniques often struggle with accurately calculating top-event probabilities and handling model uncertainty, which undermines the reliability of risk evaluations. To address this, we propose a novel algorithm that effectively reduces model uncertainty in trapezoidal fuzzy fault tree analysis. By leveraging the cut-set theorem and operating directly on trapezoidal fuzzy numbers without defuzzification, our approach preserves the completeness of fuzzy information throughout the calculation. The algorithm accommodates three common fault tree logics: OR-only, AND-only, and mixed OR/AND gates. In a case study on liquid ammonia leakage, the method achieved a 45.25% reduction in model uncertainty—surpassing existing approaches, which reached only 36.65%. Results showed 99.40% consistency with benchmark literature, affirming the algorithm’s accuracy. Additional analysis using OREDA field data demonstrated the method’s stability across different system structures and risk levels. To further verify robustness under input uncertainty, we conducted 60 perturbation tests on representative basic events with high, medium, and low failure probabilities. The algorithm exhibited exceptional stability, with output variations remaining below ±1% even under ±15% input perturbations, confirming its suitability for real-world applications where data uncertainties are common. This work offers a reliable and scalable solution for risk assessment in high-stakes industries such as nuclear energy and chemical processing.