Integrating Fuzzy Set Theory with Pandora Temporal Fault Trees for Dynamic Failure Analysis of Complex Systems

TL;DR

This paper introduces a novel method combining fuzzy set theory with Pandora Temporal Fault Trees to improve dynamic failure analysis of complex systems under uncertain failure data conditions.

Contribution

It develops fuzzy logic gates and importance measures for Pandora TFT, enabling analysis with imprecise failure data, validated through a practical aircraft system case study.

Findings

Effective handling of uncertain failure data using fuzzy logic.

Enhanced failure analysis capability for complex systems.

Validation against Petri net and Bayesian network results.

Abstract

Pandora temporal fault tree, as one notable extension of the fault tree, introduces temporal gates and temporal laws. Pandora Temporal Fault Tree(TFT) enhances the capability of fault trees and enables the modeling of system failure behavior that depends on sequences. The calculation of system failure probability in Pandora TFT relies on precise probabilistic information on component failures. However, obtaining such precise failure data can often be challenging. The data may be uncertain as historical records are used to derive failure data for system components. To mitigate this uncertainty, in this study, we proposed a method that integrates fuzzy set theory with Pandora TFT. This integration aims to enable dynamic analysis of complex systems, even in cases where quantitative failure data of components is unreliable or imprecise. The proposed work introduces the development of Fuzzy AND, Fuzzy OR, Fuzzy PAND, and Fuzzy POR logic gates for Pandora TFT. We also introduce a fuzzy importance measure for criticality analysis of basic events. All events in our analysis are assumed to have exponentially distributed failures, with their failure rates represented as triangular fuzzy numbers. We illustrate the proposed method through a case study of the Aircraft Fuel Distribution System (AFDS), highlighting its practical application and effectiveness in analyzing complex systems. The results are compared with existing results from Petri net and Bayesian network techniques to validate the findings.