Scholarlink

<p>The vibration level of aero-engine is the key index to evaluate its stability and reliability. As the main excitation source, the rotor unbalance excitation directly affects the vibration characteristics of the whole engine. Therefore, timely and effective identification of rotor imbalance can significantly improve the working performance and service life of the engine. However, the aero-engine rotor faces the technical bottleneck of insufficient identification accuracy under complex working conditions. In this paper, the power turbine rotor of aero-turboshaft engine is taken as the research object, and an unbalance identification method of rotor system based on CNN and BiLSTM is proposed. The experimental data are filtered to realize the effective analysis and processing of the vibration signal of the rotor system. The fusion database of simulation and experiment is constructed and trained by CNN-BiLSTM network. The results demonstrate that the proposed fusion model attains an MAE of 0.0499, an RMSE of 0.0983, and an R² of 0.982 on the test set. The maximum error in unbalanced mass is 3.28%, while that in unbalanced phase reaches 5.52%. These evaluation metrics confirm that the fusion model outperforms single models in prediction accuracy. This study enhances the precision of rotor unbalance identification and provides a theoretical basis for vibration control in aero-engines.</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

Deep learning-based multi-parameter unbalance identification for turbine rotor systems

  • 期刊论文

  • 发表日期：

    2026年5月8日

查看全文
  • Hui Li,
  • Changzhi Shi,
  • Xuejun Li,
  • Weihua Li,
  • Dongming Xiao

摘要

The vibration level of aero-engine is the key index to evaluate its stability and reliability. As the main excitation source, the rotor unbalance excitation directly affects the vibration characteristics of the whole engine. Therefore, timely and effective identification of rotor imbalance can significantly improve the working performance and service life of the engine. However, the aero-engine rotor faces the technical bottleneck of insufficient identification accuracy under complex working conditions. In this paper, the power turbine rotor of aero-turboshaft engine is taken as the research object, and an unbalance identification method of rotor system based on CNN and BiLSTM is proposed. The experimental data are filtered to realize the effective analysis and processing of the vibration signal of the rotor system. The fusion database of simulation and experiment is constructed and trained by CNN-BiLSTM network. The results demonstrate that the proposed fusion model attains an MAE of 0.0499, an RMSE of 0.0983, and an R² of 0.982 on the test set. The maximum error in unbalanced mass is 3.28%, while that in unbalanced phase reaches 5.52%. These evaluation metrics confirm that the fusion model outperforms single models in prediction accuracy. This study enhances the precision of rotor unbalance identification and provides a theoretical basis for vibration control in aero-engines.