Scholarlink

<p>Rolling-element bearings play a crucial role in the reliable operation of rotating machinery, and their early fault detection is essential for preventing unexpected failures. This paper proposes an intelligent fault diagnosis framework designed to maintain high classification performance under varying operating conditions and limited fault data. The novelty of the proposed approach lies in a dual-condition cost-sensitive optimization strategy that jointly learns from full-condition and few-shot variable-condition data, thereby improving adaptability and generalization. The framework integrates deep feature learning with temporal and contextual modeling to capture complex fault characteristics from raw vibration signals. Extensive experiments conducted on the CWRU bearing dataset demonstrate that the proposed method achieves classification accuracies ranging from 99.64 to 100% for both three-class and six-class scenarios. The proposed framework was validated on another bearing dataset provided by the National Technical University “Kharkiv Polytechnic Institute”, achieving an accuracy of 99.89%, which confirms its robustness, precision, and strong generalization capability.</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-condition cost-sensitive optimization framework for intelligent bearing fault diagnosis

  • 期刊论文

  • 发表日期：

    2026年4月21日

查看全文
  • Sujit Kumar,
  • Bam Bahadur Sinha,
  • Pranab Das

摘要

Rolling-element bearings play a crucial role in the reliable operation of rotating machinery, and their early fault detection is essential for preventing unexpected failures. This paper proposes an intelligent fault diagnosis framework designed to maintain high classification performance under varying operating conditions and limited fault data. The novelty of the proposed approach lies in a dual-condition cost-sensitive optimization strategy that jointly learns from full-condition and few-shot variable-condition data, thereby improving adaptability and generalization. The framework integrates deep feature learning with temporal and contextual modeling to capture complex fault characteristics from raw vibration signals. Extensive experiments conducted on the CWRU bearing dataset demonstrate that the proposed method achieves classification accuracies ranging from 99.64 to 100% for both three-class and six-class scenarios. The proposed framework was validated on another bearing dataset provided by the National Technical University “Kharkiv Polytechnic Institute”, achieving an accuracy of 99.89%, which confirms its robustness, precision, and strong generalization capability.