Scholarlink

In order to realize the mechanical fault diagnosis of disconnector, a mechanical fault diagnosis method for disconnector based on sparrow search algorithm (SSA)—variational mode decomposition (VMD) and convolutional neural network—support vector machine (CNN-SVM) is proposed. Firstly, in the process of optimizing the values of VMD’s parameters K and α, SSA is employed, after decomposition, select intrinsic mode functions (IMF) with correlation coefficients greater than 0.2; Then, the composite multi-scale weighted permutation entropy (CMWPE) and nine time-frequency domain index values of the selected IMF are calculated innovatively from the entropy domain and the time-frequency domain respectively, and the feature vector is constructed, and the t-SNE algorithm is used to reduce the dimension; Finally, using SSA-SVM instead of the Softmax layer, the trained network is used for fault recognition. Verified and analyzed through on-site experimental data, this algorithm has better accuracy in diagnosing mechanical faults of disconnector compared to other algorithms, reaching 98.33%.

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

Mechanical Fault Diagnosis Method of Disconnector Based on SSA-VMD and CNN-SVM

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Chi Zhang,
  • Hongzhong Ma,
  • Zhenghong Tian

摘要

In order to realize the mechanical fault diagnosis of disconnector, a mechanical fault diagnosis method for disconnector based on sparrow search algorithm (SSA)—variational mode decomposition (VMD) and convolutional neural network—support vector machine (CNN-SVM) is proposed. Firstly, in the process of optimizing the values of VMD’s parameters K and α, SSA is employed, after decomposition, select intrinsic mode functions (IMF) with correlation coefficients greater than 0.2; Then, the composite multi-scale weighted permutation entropy (CMWPE) and nine time-frequency domain index values of the selected IMF are calculated innovatively from the entropy domain and the time-frequency domain respectively, and the feature vector is constructed, and the t-SNE algorithm is used to reduce the dimension; Finally, using SSA-SVM instead of the Softmax layer, the trained network is used for fault recognition. Verified and analyzed through on-site experimental data, this algorithm has better accuracy in diagnosing mechanical faults of disconnector compared to other algorithms, reaching 98.33%.