Scholarlink

The intelligent development of machinery and equipment places higher demands on the prediction of the remaining useful life (RUL) of bearings. In this paper, a bearing RUL prediction method fusing the Gray Wolf optimization algorithm and Multiscale Temporal Convolutional Network-Informer (MsTCN-Informer) is proposed. Specifically, a multiscale dilated causal convolution (MsDCC) unit is first designed to extract local features in different dimensions of the vibration signal. Secondly, the Informer structure is introduced to reduce the complexity of model computation and enhance the modelling capability of the model through sparse attention mechanism with hierarchical distillation strategy. Subsequently, the Grey Wolf Optimisation (GWO) algorithm is introduced into the model to automatically tune and optimise the key hyperparameters of the model. Finally, the proposed GWO-MsTCN-Informer model is experimentally validated on the XJTU-SY dataset and its performance is compared with multiple methods. The method is experimentally verified to exhibit superior performance in both RUL prediction accuracy and stability, proving its effectiveness and application potential.

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

Bearing Remaining Useful Life Prediction Method Based on the GWO Algorithm and MsTCN-Informer Model

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Zhengqi Zhang,
  • Fan Zhang,
  • Qi Ouyang,
  • Shuaishuai Yan,
  • Xingyue Guo,
  • Aohang Pei,
  • Hongjin Wu

摘要

The intelligent development of machinery and equipment places higher demands on the prediction of the remaining useful life (RUL) of bearings. In this paper, a bearing RUL prediction method fusing the Gray Wolf optimization algorithm and Multiscale Temporal Convolutional Network-Informer (MsTCN-Informer) is proposed. Specifically, a multiscale dilated causal convolution (MsDCC) unit is first designed to extract local features in different dimensions of the vibration signal. Secondly, the Informer structure is introduced to reduce the complexity of model computation and enhance the modelling capability of the model through sparse attention mechanism with hierarchical distillation strategy. Subsequently, the Grey Wolf Optimisation (GWO) algorithm is introduced into the model to automatically tune and optimise the key hyperparameters of the model. Finally, the proposed GWO-MsTCN-Informer model is experimentally validated on the XJTU-SY dataset and its performance is compared with multiple methods. The method is experimentally verified to exhibit superior performance in both RUL prediction accuracy and stability, proving its effectiveness and application potential.