Reliability Analysis of Complex Multi-State System Based on Universal Generating Function and Bayesian Network

Xu Liu, Wen Yao, Xiaohu Zheng, Yingchun XuProceedings of the Institution of Mechanical Engineers Part O Journal of Risk and Reliability 2022 · DOI · OpenReview

TL;DR

This paper introduces a novel UGF-BN method to analyze the reliability of complex multi-state systems (MSS). By combining the efficiency of Universal Generating Functions (UGF) for bottom-level components with the inferential power of Bayesian Networks (BN) for higher-level system structures, the method significantly reduces computational cost, especially for systems with many components. The approach is demonstrated through an aircraft reliability-based design optimization case study.

What Problem Does This Paper Address?

Reliability analysis is crucial for designing, manufacturing, and maintaining complex systems. Multi-state systems (MSS) are particularly challenging because their components can operate at various performance levels, and their failures can lead to system degradation rather than complete failure. Traditional methods like Universal Generating Functions (UGF) are efficient for calculating system reliability but struggle when component interdependencies are complex or not explicitly defined. Conversely, Bayesian Networks (BN) excel at handling such uncertainties but become computationally prohibitive for systems with a large number of components. This paper addresses the need for a method that combines the strengths of both UGF and BN to efficiently and accurately analyze the reliability of complex MSS, especially those with a large number of low-level components and intricate higher-level relationships.

Methods

The core contribution is the UGF-BN method, a hybrid approach designed to overcome the limitations of UGF and BN. The methodology proceeds in a hierarchical manner:

  1. UGF for Bottom Components (Level 1): For the numerous bottom-level components, the UGF method is employed. This leverages UGF's ability to efficiently compute the probability distributions of these components, even when they have multiple states and complex performance rates.
  2. BN for System Structure (Levels 2 to n): The probability distributions obtained from the UGF analysis of the bottom components are then used as input for a Bayesian Network. The BN is constructed to model the relationships between subsystems and higher-level components, particularly where these relationships are not easily expressed by explicit functions. This allows for robust uncertainty inference.

The UGF-BN method is applied in a two-stage process: first, UGF is used to analyze the initial set of components, and then BN takes over for the subsequent system levels. This strategy aims to balance computational efficiency with the ability to handle complex dependencies.

The paper also extends this to Aircraft Reliability-Based Design Optimization. This involves defining design variables (number of components), constraint conditions (mass, power, cost), and an optimization objective (maximizing system reliability). The UGF-BN method is integrated into this optimization process to efficiently calculate system reliability under various design scenarios.

Datasets

The paper does not introduce new datasets. Instead, it uses two case studies to demonstrate and validate the proposed UGF-BN method:

  • Case 1: An unmanned aerial vehicle (UAV) dynamical system with five levels, 15 subsystems, and 68 bottom components. The paper provides descriptions of subsystems and probability distributions of bottom components.
  • Case 2: An autopilot software (PX4) system with five levels. This case focuses on reliability-based design optimization, using component data (mass, power, cost, lifetime distribution parameters) provided in tables.

Key Insights and Ideas

  • Synergistic Power of UGF and BN: The UGF-BN method effectively bridges the gap between UGF's computational efficiency for component-level analysis and BN's strength in handling complex, uncertain system relationships. This hybrid approach is key to tackling large-scale MSS.
  • Computational Efficiency Gains: The primary advantage highlighted is a significant reduction in computational cost compared to using BN alone, especially as the number of bottom components increases. This makes reliability analysis of very large systems tractable.
  • Hierarchical Modeling is Key: The paper emphasizes the importance of a hierarchical system structure where lower levels are analyzed with UGF and higher levels with BN. This decomposition is crucial for managing complexity.
  • Practical Application in Design Optimization: The successful application of UGF-BN to aircraft reliability-based design optimization demonstrates its real-world utility, enabling engineers to make informed decisions under budget and performance constraints.

Key Results

  • Accuracy Validation: The UGF-BN method's results for system reliability were shown to be identical to those obtained by the traditional BN method in Case 1, validating its accuracy.
  • Time Efficiency: In Case 1, Figure 10 clearly illustrates that the computational time for the UGF-BN method increases much more slowly than for the BN method as the number of bottom components grows. Figure 11 shows the time ratio of BN to UGF-BN, which increases rapidly with component count, indicating substantial speedups. For instance, with 168 bottom components, the time ratio is over 60.
  • Design Optimization Success: In Case 2, the UGF-BN method was used to optimize an autopilot system. The optimum scheme significantly reduced mass, power, and cost while substantially improving reliability compared to a baseline scheme. For example, total mass decreased by 39.08%, and reliability increased from 0.745 to 0.913.
  • Time Ratio in Optimization: Figure 16 shows that the UGF-BN method is consistently 5-7 times faster than the BN method for the optimization process across different computers.

Limitations and Future Work

  • UGF Application Scope: The paper notes that in the proposed UGF-BN framework, UGF is primarily used for level-1 components. The authors suggest that extending UGF analysis to more levels could further improve computational efficiency.
  • Further Optimization: While the paper demonstrates optimization for mass, power, and cost, future work could explore incorporating other design objectives or more complex constraint scenarios.
  • Generalizability: The method is demonstrated on specific system types (UAV, autopilot). Further research could explore its applicability to a broader range of complex multi-state systems.

Summary updated

Abstract

In the complex multi-state system (MSS), reliability analysis is a significant research content, both for equipment design, manufacturing, usage and maintenance. Universal Generating Function (UGF) is an important method in the reliability analysis, which efficiently obtains the system reliability by a fast algebraic procedure. However, when structural relationships between subsystems or components are not clear or without explicit expressions, the UGF method is difficult to use or not applicable at all. Bayesian Network (BN) has a natural advantage in terms of uncertainty inference for the relationship without explicit expressions. For the number of components is extremely large, though, it has the defects of low efficiency. To overcome the respective defects of UGF and BN, a novel reliability analysis method called UGF-BN is proposed for the complex MSS. In the UGF-BN framework, the UGF method is firstly used to analyze the bottom components with a large number. Then probability distributions obtained are taken as the input of BN. Finally, the reliability of the complex MSS is modeled by the BN method. This proposed method improves the computational efficiency, especially for the MSS with the large number of bottom components. Besides, the aircraft reliability-based design optimization based on the UGF-BN method is further studied with budget constraints on mass, power, and cost. Finally, two cases are used to demonstrate and verify the proposed method.
BibTeX
@article{liu2022reliability,
  author = {Xu Liu and Wen Yao and Xiaohu Zheng and Yingchun Xu},
  title = {{Reliability Analysis of Complex Multi-State System Based on Universal Generating Function and Bayesian Network}},
  journal = {Proceedings of the Institution of Mechanical Engineers Part O Journal of Risk and Reliability},
  year = {2022},
  doi = {10.1177/1748006x231173301},
}
Figure