Scholarlink

<p>The gearbox is a crucial component for the operation of machinery. There is an imbalance in the data of various faults that occur during its operation. Therefore, this paper proposes an ensemble learning method from the perspective of feature extraction. Firstly, an iterative wavelet packet decomposition algorithm is designed. Through multi-level signal decomposition, deep time-frequency features are gradually extracted, and a multi-head attention mechanism is combined to adaptively learn the importance weights of features, thereby constructing an enhanced feature representation. Secondly, an attention-weighted ensemble learning classification framework was developed, integrating six base classifiers such as random forest, support vector machine, gradient boosting decision tree, K-nearest neighbor, multi-layer perceptron, and decision tree. The weights of the classifiers were dynamically allocated based on the cross-validation performance to achieve intelligent weighted voting decision-making. Finally, the effectiveness of the proposed model in addressing the feature extraction problem of imbalanced data was verified on two datasets.</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

Imbalance fault diagnosis based on iterative attention wavelet packet decomposition and weighted ensemble classification algorithm

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年7月27日

查看全文
  • Mingdong Zhao,
  • Ren Wang,
  • Xibin Guo,
  • Zhen Guo

摘要

The gearbox is a crucial component for the operation of machinery. There is an imbalance in the data of various faults that occur during its operation. Therefore, this paper proposes an ensemble learning method from the perspective of feature extraction. Firstly, an iterative wavelet packet decomposition algorithm is designed. Through multi-level signal decomposition, deep time-frequency features are gradually extracted, and a multi-head attention mechanism is combined to adaptively learn the importance weights of features, thereby constructing an enhanced feature representation. Secondly, an attention-weighted ensemble learning classification framework was developed, integrating six base classifiers such as random forest, support vector machine, gradient boosting decision tree, K-nearest neighbor, multi-layer perceptron, and decision tree. The weights of the classifiers were dynamically allocated based on the cross-validation performance to achieve intelligent weighted voting decision-making. Finally, the effectiveness of the proposed model in addressing the feature extraction problem of imbalanced data was verified on two datasets.