Scholarlink

<p>To enhance the accuracy and robustness of bearing fault diagnosis, this paper proposes a novel diagnostic method that optimizes the weighted k-nearest neighbor (WKNN) classifier using a multi-strategy improved particle swarm optimization (MPSO) algorithm. First, singular spectrum decomposition (SSD) is employed to decompose bearing vibration signals, and the decomposed component signals are reconstructed via the Pearson correlation coefficient method. Subsequently, features reflecting the complexity of the reconstructed signals—including permutation entropy, sample entropy, and energy entropy—are extracted to construct the fault diagnosis feature vector. To strengthen the global search capability and population diversity of the particle swarm optimization (PSO) algorithm, and thereby avoid premature convergence to local optima, this study introduces Tent chaotic mapping and dynamic opposition-based learning (DOBL) for algorithm improvement. The enhanced PSO algorithm is then utilized to optimize two key parameters of the WKNN classifier: the number of nearest neighbors and inverse distance weights, which effectively improves the fault classification accuracy. Experimental results demonstrate that the proposed method exhibits excellent diagnostic accuracy and robustness under complex operating conditions, achieving accuracies of 95.63 % and 96.25 % on two distinct datasets. This performance outperforms that of traditional optimization algorithms and conventional diagnostic methods, fully validating the effectiveness of the proposed approach in bearing fault diagnosis.</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

Wind turbine bearing fault diagnosis based on multi-strategy improved particle swarm optimization weighted k-nearest neighbor method

  • 期刊论文

  • 发表日期：

    2026年7月10日

查看全文
  • Pengcheng Zhao,
  • Jianjian Zhao,
  • Jie Sun,
  • Junjiao Shi,
  • Angang Wei,
  • Xinli Zhang

摘要

To enhance the accuracy and robustness of bearing fault diagnosis, this paper proposes a novel diagnostic method that optimizes the weighted k-nearest neighbor (WKNN) classifier using a multi-strategy improved particle swarm optimization (MPSO) algorithm. First, singular spectrum decomposition (SSD) is employed to decompose bearing vibration signals, and the decomposed component signals are reconstructed via the Pearson correlation coefficient method. Subsequently, features reflecting the complexity of the reconstructed signals—including permutation entropy, sample entropy, and energy entropy—are extracted to construct the fault diagnosis feature vector. To strengthen the global search capability and population diversity of the particle swarm optimization (PSO) algorithm, and thereby avoid premature convergence to local optima, this study introduces Tent chaotic mapping and dynamic opposition-based learning (DOBL) for algorithm improvement. The enhanced PSO algorithm is then utilized to optimize two key parameters of the WKNN classifier: the number of nearest neighbors and inverse distance weights, which effectively improves the fault classification accuracy. Experimental results demonstrate that the proposed method exhibits excellent diagnostic accuracy and robustness under complex operating conditions, achieving accuracies of 95.63 % and 96.25 % on two distinct datasets. This performance outperforms that of traditional optimization algorithms and conventional diagnostic methods, fully validating the effectiveness of the proposed approach in bearing fault diagnosis.