Scholarlink

<p>Traditional reliability-based robust design optimization (RBRDO) methods often require large amounts of sample data to ensure high confidence in structural parameters. However, in practical engineering applications, obtaining sufficient data is often difficult and costly. To address this issue, this paper proposes a Bayesian statistics-based approach to RBRDO. This approach integrates prior knowledge with limited sample data through Bayesian inference, constructs a Bayesian reliability assessment model, and combines reliability sensitivity analysis with multi-objective optimization strategies to achieve high reliability and low sensitivity in data-scarce conditions. The effectiveness of the proposed method is demonstrated through four examples involving vehicle components. With only 10 samples, the Bayesian reliability for all examples exceeds 0.9999, while reliability sensitivity is reduced to the order of 10<sup>–5</sup> to 10<sup>–7</sup>, and the error compared to Monte Carlo simulation (MCS) is less than 0.002%. These results demonstrate that the design exhibits excellent robustness and stability under parameter uncertainty, providing a practical solution for the reliability-based design of vehicle components in high-cost testing environments.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Bayesian Reliability-Based Robust Design Optimization of Vehicle Components

  • 期刊论文

  • 发表日期：

    2026年7月13日

查看全文
  • Zhaozhan Li,
  • Zhaowang Li,
  • Jufang Jia,
  • Xiangdong He

摘要

Traditional reliability-based robust design optimization (RBRDO) methods often require large amounts of sample data to ensure high confidence in structural parameters. However, in practical engineering applications, obtaining sufficient data is often difficult and costly. To address this issue, this paper proposes a Bayesian statistics-based approach to RBRDO. This approach integrates prior knowledge with limited sample data through Bayesian inference, constructs a Bayesian reliability assessment model, and combines reliability sensitivity analysis with multi-objective optimization strategies to achieve high reliability and low sensitivity in data-scarce conditions. The effectiveness of the proposed method is demonstrated through four examples involving vehicle components. With only 10 samples, the Bayesian reliability for all examples exceeds 0.9999, while reliability sensitivity is reduced to the order of 10–5 to 10–7, and the error compared to Monte Carlo simulation (MCS) is less than 0.002%. These results demonstrate that the design exhibits excellent robustness and stability under parameter uncertainty, providing a practical solution for the reliability-based design of vehicle components in high-cost testing environments.