Reliability evaluation of a multi-state system with dependent components and imprecise parameters: A structural reliability treatment

Lechang Yang, Xinyao Zhang, Zitong Lu, Yuqiang Fu, David Moens, Michael BeerReliab. Eng. Syst. Saf 2024 · DOI

Introduction

In modern engineering, particularly in high-stakes fields like satellite design, the traditional binary view of reliability—where a component is either "on" or "off"—is often insufficient. Real-world systems are frequently Multi-State Systems (MSS), capable of operating at various performance levels, such as "perfectly working," "partially working," or "failed." Evaluating the reliability of these systems is complicated by three major factors: component dependencies, data imprecision, and extreme computational complexity. This paper introduces a novel framework that addresses these challenges by integrating Survival Signatures, Vine Copulas, and an efficient Importance Sampling-based Monte Carlo (ISMC) algorithm.

Addressing the Complexity of Multi-State Systems

Traditional reliability models often assume that components fail independently. However, in complex structures, the failure of one part typically increases the stress on others, creating a chain of dependency. Furthermore, engineers rarely have perfect data; failure rates are often provided as intervals or "fuzzy" parameters rather than exact numbers. When these imprecise probabilities are combined with multi-state behavior and dependencies, the resulting mathematical models often require "double-loop" simulations that are too slow for practical use.

Illustrative bridge system and multi-state transition logic

As shown in the figure above, a system can transition through multiple states over time. Modeling these transitions requires a robust structural treatment that can handle the non-monotonic nature of system reliability when components are dependent.

Methodology: Survival Signatures and Vine Copulas

The proposed framework relies on two primary mathematical tools to decouple the system's structure from its probabilistic data:

  1. Survival Signatures: This approach separates the topological layout of the system from the failure distributions of its components. By calculating the signature once for a specific design, researchers can update failure data or imprecise parameters without re-evaluating the entire system structure.
  2. Vine Copulas: To model complex, high-dimensional dependencies, the authors employ C-vines and D-vines. Unlike simple linear correlations, Vine Copulas offer the flexibility to model how different components interact under various stress conditions, providing a more realistic picture of system behavior.

Computational Breakthrough: The ISMC Algorithm

The core innovation of this research is the Importance Sampling-based Monte Carlo (ISMC) algorithm. Standard approaches to imprecise probability require a Double-Loop Monte Carlo (DLMC) method—one loop to sample failure times and another to account for parameter uncertainty. This is computationally expensive and often prohibitive.

By treating imprecise parameters as part of an optimization problem within a single simulation loop, the ISMC algorithm uses importance weights to "collapse" the double loop. This allows for the generation of "p-boxes" (probability boxes) that define the upper and lower bounds of system reliability, providing engineers with a clear "safety corridor" for performance.

Experimental Validation and Results

The framework was validated using a numerical bridge system and a real-world Dual-Axis Drive Mechanism (DADM) from a satellite antenna system.

  • Efficiency Gains: In the bridge system benchmark, the ISMC algorithm achieved a 4x speedup over the conventional DLMC method (556 seconds vs. 2,192 seconds) while maintaining high accuracy (error rates < 0.01).
  • The Danger of Optimism: A critical finding was that models assuming component independence consistently produced over-optimistic reliability estimates. In aerospace applications, ignoring these dependencies could lead to underestimating the risk of catastrophic mission failure.
  • Practical Application: For the satellite DADM, the method successfully modeled the dependencies between stepping motors and brackets, providing a reliable p-box that accounts for both the backup nature of the motors and the imprecision of their failure data.

Conclusion

This research provides a significant step forward in structural reliability treatment for complex systems. By combining the structural flexibility of Survival Signatures with the dependency modeling of Vine Copulas and the efficiency of ISMC, the authors offer a practical tool for engineers to predict failure in multi-state systems. While currently optimized for systems where states are determined by working paths, future work aims to extend this logic to even more diverse architectures and complex failure modes.

Summary updated

BibTeX
@article{yang2024reliability,
  author = {Lechang Yang and Xinyao Zhang and Zitong Lu and Yuqiang Fu and David Moens and Michael Beer},
  title = {{Reliability evaluation of a multi-state system with dependent components and imprecise parameters: A structural reliability treatment}},
  journal = {Reliab. Eng. Syst. Saf.},
  year = {2024},
  doi = {10.1016/j.ress.2024.110240},
}
Figure