Scholarlink

Background <p>Gear pumps play an important role in ensuring the reliable operation of hydraulic systems and actuators. However, their degradation is affected by the coupling of mechanical wear and hydraulic performance deterioration. Remaining useful life prediction based on a single sensor signal may therefore lead to large errors. In addition, the high reliability requirement of gear pumps makes life-cycle degradation data difficult to obtain.</p> Methods <p>To address these problems, a three-stress accelerated life test was designed for gear pumps to reduce the test time and obtain degradation data. Multidomain vibration features were extracted from the acquired signals in the time domain, frequency domain, and time-frequency domain. These features were then fused with volumetric efficiency data using kernel principal component analysis to construct degradation indicators. Finally, an Attention-Seq2Seq-BiGRU model was developed to capture long-term degradation information and predict the remaining useful life of gear pumps.</p> Results <p>The proposed method achieved accurate RUL prediction for gear pumps. The fused features based on vibration signals and volumetric efficiency data improved prediction stability. Compared with GRU, LSTM, BiGRU, and Seq2Seq-BiGRU models, the proposed Attention-Seq2Seq-BiGRU model showed the best prediction performance.</p> Conclusions <p>The proposed framework effectively combines accelerated life testing, KPCA-based feature fusion, and Attention-Seq2Seq-BiGRU modeling. The results demonstrate that this method can improve the accuracy and stability of gear pump RUL prediction.</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

Accelerated Life Testing and Remaining Useful Life Prediction Research of Gear Pumps

  • 期刊论文

  • 发表日期：

    2026年6月29日

查看全文
  • Changshu Li,
  • Jing Ma,
  • Wenyan Song,
  • Xifeng Chen

摘要

Background

Gear pumps play an important role in ensuring the reliable operation of hydraulic systems and actuators. However, their degradation is affected by the coupling of mechanical wear and hydraulic performance deterioration. Remaining useful life prediction based on a single sensor signal may therefore lead to large errors. In addition, the high reliability requirement of gear pumps makes life-cycle degradation data difficult to obtain.

Methods

To address these problems, a three-stress accelerated life test was designed for gear pumps to reduce the test time and obtain degradation data. Multidomain vibration features were extracted from the acquired signals in the time domain, frequency domain, and time-frequency domain. These features were then fused with volumetric efficiency data using kernel principal component analysis to construct degradation indicators. Finally, an Attention-Seq2Seq-BiGRU model was developed to capture long-term degradation information and predict the remaining useful life of gear pumps.

Results

The proposed method achieved accurate RUL prediction for gear pumps. The fused features based on vibration signals and volumetric efficiency data improved prediction stability. Compared with GRU, LSTM, BiGRU, and Seq2Seq-BiGRU models, the proposed Attention-Seq2Seq-BiGRU model showed the best prediction performance.

Conclusions

The proposed framework effectively combines accelerated life testing, KPCA-based feature fusion, and Attention-Seq2Seq-BiGRU modeling. The results demonstrate that this method can improve the accuracy and stability of gear pump RUL prediction.