Information and communication and chemical technologies

Vol. 3 No. 28 (2025) / 2025-09-30 / Number of views: 178

PREDICTIVE MAINTENANCE OF SPINDLE BEARINGS BASED ON VIBRATION DIAGNOSTICS (A CASE STUDY OF THE CKE 6136/6150 LATHE)

Authors

A.Baitursynuly Kostanay Regional University
https://orcid.org/0000-0002-8681-4552
A.Baitursynuly Kostanay Regional University
https://orcid.org/0000-0001-5347-7206
A.Baitursynuly Kostanay Regional University
https://orcid.org/0000-0002-8681-4552
A.Baitursynuly Kostanay Regional University
https://orcid.org/0000-0003-2834-1010
M.Dulatov Kostanay Engineering and Economics University
https://orcid.org/0000-0002-1415-5244
A.Baitursynuly Kostanay Regional University
https://orcid.org/0000-0002-6157-4652

Keywords

predictive maintenance, vibration diagnostics, machine learning, spindle bearings, spectral analysis, failure prediction, gradient boosting, intelligent equipment monitoring

Link to DOI:

https://doi.org/10.58805/kazutb.v.3.28-873

How to quote

Madin, V., et al. “PREDICTIVE MAINTENANCE OF SPINDLE BEARINGS BASED ON VIBRATION DIAGNOSTICS (A CASE STUDY OF THE CKE 6136 6150 LATHE)”. Vestnik KazUTB, vol. 3, no. 28, Sept. 2025, doi:10.58805/kazutb.v.3.28-873.

Abstract

Modern predictive maintenance (Predictive Maintenance, PdM) methods enhance equipment reliability and reduce repair costs by enabling early fault detection. This study presents a comprehensive approach to predictive maintenance of spindle bearings in the CKE 6136/6150 lathe, based on vibration diagnostics, spectral analysis, and machine learning techniques. A thorough analysis of vibration, temperature, and load time series was conducted, incorporating both statistical and spectral analysis. The results indicate that the front support’s radial vibration is a key wear indicator.

During the study, machine learning models, including linear regression and gradient boosting (XGBoost), were tested. The XGBoost model demonstrated the best performance with an accuracy of R² = 0.999. A SHAP analysis confirmed the significance of vibration and temperature in predicting wear levels. Spectral analysis identified characteristic bearing defect frequencies (BPFO, BPFI, BSF, FTF), but also highlighted the limitations of the method when data sampling frequency was insufficient. This underscores the importance of selecting an appropriate measurement frequency, as low sampling rates may lead to the loss of diagnostically significant information.

The obtained results confirm the effectiveness of the proposed approach for spindle bearing wear diagnostics. The conclusion outlines future research directions, including increasing data collection frequency, validating the model on real machines, and developing adaptive predictive maintenance algorithms.