Scholarlink

<p>An essential technique for ensuring the dependability and security of rotating machinery is the capacity to predict the remaining useful life (RUL) of bearings. Nonetheless, some significant issues with RUL prediction still need to be solved, such as learning multi-dimensional feature data and efficiently screening multi-domain characteristics. To handle above issues, an adaptive multi-domain feature selection and nested long short-term memory (AMF-NLSTM) based RUL prediction method is put forth. The first step is to detect the bearings’ degradation information by extracting the multi-domain features of the bearing monitoring signals. Then, the degradation-sensitive features are screened using an adaptive feature selection technique. Finally, an NLSTM model is established to learn the long-timescale information between the feature time series and RUL to achieve RUL prediction. Experimental verification is conducted using the XJTU-SY and the PHM 2012 bearing datasets. The suggested method outperforms conventional RUL prediction methods, making it useful for predictive maintenance of rotating machinery.</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

Remaining Useful Life Prediction of Rotating Machinery Bearings Based on Adaptive Multi-domain Feature Selection and Nested LSTM

  • 期刊论文

  • 发表日期：

    2025年4月28日

查看全文
  • Ming Zhao,
  • Jinxin Wu,
  • Xianwang Li

摘要

An essential technique for ensuring the dependability and security of rotating machinery is the capacity to predict the remaining useful life (RUL) of bearings. Nonetheless, some significant issues with RUL prediction still need to be solved, such as learning multi-dimensional feature data and efficiently screening multi-domain characteristics. To handle above issues, an adaptive multi-domain feature selection and nested long short-term memory (AMF-NLSTM) based RUL prediction method is put forth. The first step is to detect the bearings’ degradation information by extracting the multi-domain features of the bearing monitoring signals. Then, the degradation-sensitive features are screened using an adaptive feature selection technique. Finally, an NLSTM model is established to learn the long-timescale information between the feature time series and RUL to achieve RUL prediction. Experimental verification is conducted using the XJTU-SY and the PHM 2012 bearing datasets. The suggested method outperforms conventional RUL prediction methods, making it useful for predictive maintenance of rotating machinery.