Scholarlink

<p>This paper proposes an innovative health status assessment model based on a stacked ensemble method. In air compressor health status assessment. A single model performs poorly when faced with complex multi-source information. To overcome this challenge, the interval hierarchical analysis method (IAHP) was first combined with the entropy weight method (EWM) to calculate parameter weights, and a dynamic decay factor was introduced to quantify the degree of degradation, thereby constructing a comprehensive health status model (HDM). Subsequently, a multi-level comprehensive evaluation model based on fuzzy-DS was designed, utilizing membership functions and D-S evidence theory to handle uncertainty and recursively fuse multi-source evidence. Finally, the HDM and multi-level Dempster-Shafer evidence model were used as base learners, and a stacked ensemble method with a random forest as the meta-learner was employed for collaborative optimization. Taking an oil-immersed screw air compressor as an example, the effectiveness of this method was validated using five-fold cross-validation based on performance metrics. The accuracy of the proposed Stacking ensemble model reached 99.87 %, significantly outperforming single models, providing higher robustness and accuracy for assessing compressor health under complex conditions.</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

Health state assessment of air compressor based on stacking ensemble learning

  • 期刊论文

  • 发表日期：

    2025年12月10日

查看全文
  • Dali Hou,
  • Zhengrong Yang,
  • Xiaoran Wang

摘要

This paper proposes an innovative health status assessment model based on a stacked ensemble method. In air compressor health status assessment. A single model performs poorly when faced with complex multi-source information. To overcome this challenge, the interval hierarchical analysis method (IAHP) was first combined with the entropy weight method (EWM) to calculate parameter weights, and a dynamic decay factor was introduced to quantify the degree of degradation, thereby constructing a comprehensive health status model (HDM). Subsequently, a multi-level comprehensive evaluation model based on fuzzy-DS was designed, utilizing membership functions and D-S evidence theory to handle uncertainty and recursively fuse multi-source evidence. Finally, the HDM and multi-level Dempster-Shafer evidence model were used as base learners, and a stacked ensemble method with a random forest as the meta-learner was employed for collaborative optimization. Taking an oil-immersed screw air compressor as an example, the effectiveness of this method was validated using five-fold cross-validation based on performance metrics. The accuracy of the proposed Stacking ensemble model reached 99.87 %, significantly outperforming single models, providing higher robustness and accuracy for assessing compressor health under complex conditions.