Scholarlink

<p>Multi-modal information fusion technology is active in the field of mechanical equipment health state diagnosis (HSD), with its effective capabilities for feature collaboration and complementarity. For complex mechanical systems, such as manipulators, effectively discriminating mode-invariant representations from coupled modal information remains a significant challenge. To address this, a novel intelligent HSD framework, named the modal inversion contrast reinforcement-enabled fusion network (MICR-FN), is developed for the multi-modal information fusion. Specifically, the intrinsic representation features contained within each independent modal information are sequentially mined and strengthened through a unique dilated collaborative residual Transformer (DCRFormer) module, where a novel self-attention perception mechanism and a weight residual connection mechanism effectively prevent the loss of critical information. Subsequently, modality-invariant and modality-specific features in each modality are further separated, with a designed inverse contrast reinforcement strategy (ICRS) ensuring semantic consistency and credibility among modality-disentangled representations, thereby facilitating the diagnosis task. Ultimately, the feasibility, advantages, robustness, and anti-interference performance of the proposed fusion method are comprehensively assessed and analyzed through a classical 3-prismatic-revolute-revolute (3-PRR) planar parallel manipulator under multiple scenario cases.</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

Modal inversion contrast reinforcement-enabled multi-modal information fusion network for intelligent diagnosis of planar parallel manipulators

  • 期刊论文

  • 发表日期：

    2026年5月20日

查看全文
  • Bo Zhao,
  • Qiqiang Wu,
  • Haidong Shao,
  • Weihua Li,
  • Zijun Zhang,
  • Xianmin Zhang

摘要

Multi-modal information fusion technology is active in the field of mechanical equipment health state diagnosis (HSD), with its effective capabilities for feature collaboration and complementarity. For complex mechanical systems, such as manipulators, effectively discriminating mode-invariant representations from coupled modal information remains a significant challenge. To address this, a novel intelligent HSD framework, named the modal inversion contrast reinforcement-enabled fusion network (MICR-FN), is developed for the multi-modal information fusion. Specifically, the intrinsic representation features contained within each independent modal information are sequentially mined and strengthened through a unique dilated collaborative residual Transformer (DCRFormer) module, where a novel self-attention perception mechanism and a weight residual connection mechanism effectively prevent the loss of critical information. Subsequently, modality-invariant and modality-specific features in each modality are further separated, with a designed inverse contrast reinforcement strategy (ICRS) ensuring semantic consistency and credibility among modality-disentangled representations, thereby facilitating the diagnosis task. Ultimately, the feasibility, advantages, robustness, and anti-interference performance of the proposed fusion method are comprehensively assessed and analyzed through a classical 3-prismatic-revolute-revolute (3-PRR) planar parallel manipulator under multiple scenario cases.