Supervised Prediction of Remaining Useful Life of Drills from Acoustic Signals

• Raphael Barbosa UFRPE
• Sergio Chevtchenko Western Sydney University
• Saeed Afshar Western Sydney University
• Naqib Ibnul Western Sydney University
• Gustavo Callou UFRPE
• Ermeson Andrade UFRPE

DOI: https://doi.org/10.5753/wtf.2026.24127

Resumo

Early failure detection in cutting tools is a critical challenge in automated manufacturing environments, where unplanned interruptions lead to significant operational costs. This study investigates the supervised prediction of imminent drill failures using acoustic signals collected during machining operations. Ultrasonic recordings obtained from microphones positioned inside a Computer Numerical Control (CNC) machine are segmented per hole and labeled according to failure occurrence within different prediction horizons (fail_in_1, fail_in_3, fail_in_5), indicating failures occurring within 1, 3, and 5 future drilling operations, respectively. Two approaches are evaluated: classical machine learning using statistical and spectral features with a Random Forest classifier, and deep learning using Convolutional Neural Networks (CNN) applied to log-Mel spectrograms. To ensure realistic generalization, data splitting is performed at the tool level, preventing leakage between training and test sets. Results show that acoustic signals exhibit discriminative patterns for early failure detection. The CNN achieves higher F1-scores across all horizons, while the Random Forest provides more stable performance and higher AUC values, highlighting the trade-off between sensitivity and robustness in predictive maintenance systems.

Referências

Barbosh, M., Ge, L., and Sadhu, A. (2024). Automated crack identification in structures using acoustic waveforms and deep learning. Journal of Infrastructure Preservation and Resilience, 5(1):10.

Breiman, L. (2001). Random forests. Machine learning, 45(1):5–32.

Byrne, G., Dornfeld, D., and Denkena, B. (2003). Advancing cutting technology. CIRP Annals, 52(2):483–507.

Carden, E. P. and Fanning, P. (2004). Vibration based condition monitoring: a review. Structural health monitoring, 3(4):355–377.

Chevtchenko, S. F., Rocha, E. D. S., Dos Santos, M. C. M., Mota, R. L., Vieira, D. M., De Andrade, E. C., and De Araújo, D. R. B. (2023). Anomaly detection in industrial machinery using iot devices and machine learning: A systematic mapping. IEEe Access, 11:128288–128305.

Ferrisi, S., Zangara, G., Izquierdo, D. R., Lofaro, D., Guido, R., Conforti, D., and Ambrogio, G. (2024). Tool condition monitoring for milling process using convolutional neural networks. Procedia Computer Science, 232:1607–1616.

Jardine, A. K., Lin, D., and Banjevic, D. (2006). A review on machinery diagnostics and prognostics implementing condition-based maintenance. Mechanical systems and signal processing, 20(7):1483–1510.

Lara de Leon, M. A., Kolarik, J., Byrtus, R., Koziorek, J., Zmij, P., and Martinek, R. (2024). Tool condition monitoring methods applicable in the metalworking process: Mal de leon et al. Archives of computational methods in engineering, 31(1):221–242.

LeCun, Y., Bengio, Y., and Hinton, G. (2015). Deep learning. nature, 521(7553):436–444.

Lei, Y., Li, N., Guo, L., Li, N., Yan, T., and Lin, J. (2018). Machinery health prognostics: A systematic review from data acquisition to rul prediction. Mechanical systems and signal processing, 104:799–834.

Piorkowski, P., Roszkowski, A., and Szabla, Z. (2025). Diagnostics of milling head using acoustic emission. Manufacturing Technology Journal, 25(2):222–229.

Si, X.-S., Wang, W., Hu, C.-H., and Zhou, D.-H. (2011). Remaining useful life estimation–a review on the statistical data driven approaches. European journal of operational research, 213(1):1–14.

Umar, M., Siddique, M. F., Ullah, N., and Kim, J.-M. (2024). Milling machine fault diagnosis using acoustic emission and hybrid deep learning with feature optimization. Applied Sciences, 14(22):10404.