Scholarlink

<p>To enhance the prediction precision of gear remaining useful life (RUL), this study proposes a novel RUL prediction approach based on multi-head attention (MHA) and hybrid variational auto encoder (HyVAE) optimized by an improved Arctic puffin optimization (IAPO) algorithm. In the study, the time-domain and frequency-domain characteristics are firstly extracted from the original vibration signals, and then these characteristics are weighted and fused through the MHA mechanism. Subsequently, bidirectional gated recurrent units (BiGRU) are integrated into the variational auto encoder (VAE) framework, forming a HyVAE network, which is then utilized to perform predictive analysis on the fused features. Moreover, to enhance the HyVAE network’s performance, we propose the IAPO algorithm that integrates tent chaotic mapping and Cauchy mutation perturbation, addressing the limitations of conventional APO. The optimized IAPO is subsequently employed for key parameter optimization in the HyVAE network, ultimately yielding the superior IAPO-MHA-HyVAE prediction model. The effectiveness and advantages of the proposed approach in predicting gear RUL are validated through the analysis of two different gear life-cycle vibration signal datasets.</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

Remaining useful life prediction of gears: An IAPO-MHA-HyVAE-based approach

  • 期刊论文

  • 发表日期：

    2026年1月10日

查看全文
  • Xiangmin Chen,
  • Yonghui Li,
  • Hanlin Lei,
  • Kang Zhang,
  • Shilong Hu,
  • Yang Su

摘要

To enhance the prediction precision of gear remaining useful life (RUL), this study proposes a novel RUL prediction approach based on multi-head attention (MHA) and hybrid variational auto encoder (HyVAE) optimized by an improved Arctic puffin optimization (IAPO) algorithm. In the study, the time-domain and frequency-domain characteristics are firstly extracted from the original vibration signals, and then these characteristics are weighted and fused through the MHA mechanism. Subsequently, bidirectional gated recurrent units (BiGRU) are integrated into the variational auto encoder (VAE) framework, forming a HyVAE network, which is then utilized to perform predictive analysis on the fused features. Moreover, to enhance the HyVAE network’s performance, we propose the IAPO algorithm that integrates tent chaotic mapping and Cauchy mutation perturbation, addressing the limitations of conventional APO. The optimized IAPO is subsequently employed for key parameter optimization in the HyVAE network, ultimately yielding the superior IAPO-MHA-HyVAE prediction model. The effectiveness and advantages of the proposed approach in predicting gear RUL are validated through the analysis of two different gear life-cycle vibration signal datasets.