Scholarlink

Background <p>Tool wear monitoring is a critical aspect of machining process reliability as it directly affects productivity, surface quality, and maintenance planning. Accurate estimation of the Remaining Useful Life (RUL) of cutting tools enables predictive maintenance and reduces unexpected downtime in manufacturing systems.</p> Methods <p>This study presents a vibration-based, data-driven framework for estimating the Remaining Useful Life of face milling cutters operating on a vertical machining center during mild steel machining. Tri-axial spindle vibration signals were acquired and represented using physically interpretable time-domain, statistical, and frequency-domain descriptors. To capture both instantaneous signal characteristics and their temporal evolution, a hybrid Convolutional Neural Network–Bidirectional Long Short-Term Memory (CNN–BiLSTM) regression model was developed. The convolutional layers extract localized feature patterns, while the bidirectional recurrent layers model sequential dependencies across successive machining passes. The performance of the proposed model was compared with several baseline regression approaches.</p> Results <p>Across repeated evaluations, the CNN–BiLSTM model demonstrated improved prediction accuracy compared to baseline models. It achieved a Mean Absolute Error (MAE) of 7.2 ± 0.5 passes, a Root Mean Square Error (RMSE) of 9.5 ± 0.6 passes, and an R² score of 0.96 ± 0.01 on the test dataset.</p> Conclusion <p>The results indicate that combining vibration-based features with temporal deep learning models provides reliable and consistent RUL estimation under the studied machining conditions. The proposed framework offers a practical approach for data-driven tool condition monitoring and supports predictive maintenance strategies in machining environments.</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 Face Milling Cutter via Vibration Signatures and CNN–BiLSTM Architecture

  • 期刊论文

  • 发表日期：

    2026年3月10日

查看全文
  • Sangeeta Mundra,
  • Manisha Yadav,
  • Sadhana Jadhav

摘要

Background

Tool wear monitoring is a critical aspect of machining process reliability as it directly affects productivity, surface quality, and maintenance planning. Accurate estimation of the Remaining Useful Life (RUL) of cutting tools enables predictive maintenance and reduces unexpected downtime in manufacturing systems.

Methods

This study presents a vibration-based, data-driven framework for estimating the Remaining Useful Life of face milling cutters operating on a vertical machining center during mild steel machining. Tri-axial spindle vibration signals were acquired and represented using physically interpretable time-domain, statistical, and frequency-domain descriptors. To capture both instantaneous signal characteristics and their temporal evolution, a hybrid Convolutional Neural Network–Bidirectional Long Short-Term Memory (CNN–BiLSTM) regression model was developed. The convolutional layers extract localized feature patterns, while the bidirectional recurrent layers model sequential dependencies across successive machining passes. The performance of the proposed model was compared with several baseline regression approaches.

Results

Across repeated evaluations, the CNN–BiLSTM model demonstrated improved prediction accuracy compared to baseline models. It achieved a Mean Absolute Error (MAE) of 7.2 ± 0.5 passes, a Root Mean Square Error (RMSE) of 9.5 ± 0.6 passes, and an R² score of 0.96 ± 0.01 on the test dataset.

Conclusion

The results indicate that combining vibration-based features with temporal deep learning models provides reliable and consistent RUL estimation under the studied machining conditions. The proposed framework offers a practical approach for data-driven tool condition monitoring and supports predictive maintenance strategies in machining environments.