Scholarlink

<p>Aiming at the problem of few-sample fault diagnosis in industrial scenarios, a dualbranch inverted residual network guided by mixed loss (DBIRN-ML) is developed in this paper. First, a multi-depth feature fusion network is established based on inverted residual modules, an attention mechanism and a backward-forward feature fusion strategy. Its paired input mechanism expands learning tasks to alleviate the few-shot limitation. Then, a contrastive loss is constructed between the two feature extraction branches, while classification losses are separately connected after each branch of the dual-branch network. The weighted sum of the above loss functions forms a mixed loss, enabling the network to synergistically extract contrastive information from sample pairs and categorical information from individual samples, thereby further enhancing the fault identification capability of DBIRN-ML under few-shot conditions. Two case studies validate the diagnosis effectiveness of the method in the limited sample situation.</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

A dual-branch inverted residual network guided by mixed loss for the bearing fault diagnosis with few samples

  • 期刊论文

  • 发表日期：

    2026年1月8日

查看全文
  • Jie Liu,
  • Ting Wang,
  • Hai Wang,
  • Jingsong Xie,
  • Kai Zhang

摘要

Aiming at the problem of few-sample fault diagnosis in industrial scenarios, a dualbranch inverted residual network guided by mixed loss (DBIRN-ML) is developed in this paper. First, a multi-depth feature fusion network is established based on inverted residual modules, an attention mechanism and a backward-forward feature fusion strategy. Its paired input mechanism expands learning tasks to alleviate the few-shot limitation. Then, a contrastive loss is constructed between the two feature extraction branches, while classification losses are separately connected after each branch of the dual-branch network. The weighted sum of the above loss functions forms a mixed loss, enabling the network to synergistically extract contrastive information from sample pairs and categorical information from individual samples, thereby further enhancing the fault identification capability of DBIRN-ML under few-shot conditions. Two case studies validate the diagnosis effectiveness of the method in the limited sample situation.