Central AsianResearch Index

A Model for Predicting the Remaining Life of Rolling Bearings Based on a Vibration Signal Using XGBoost, Random Forest, and LSTM Algorithms

The paper presents an approach to predicting the remaining useful life of rolling bearings based on a single-channel accelerometric vibration signal. A Random Forest classifier is used to distinguish healthy and faulty operating conditions, while XGBoost and LSTM models are employed to estimate RUL from a set of time‑ and frequency‑domain statistical features of the vibration signal. The models are trained and validated on the NASA Bearing Dataset, which contains run-to-failure recordings of bearing vibration. The proposed method achieves a high classification accuracy (up to 98-99%) with low computational complexity and requires only a limited number of input features and a single sensor channel. Visualization of the predicted RUL in time and SHAP‑based feature importance analysis confirm the applicability of the developed model to technical diagnostics, condition monitoring and predictive maintenance of bearing assemblies.

Сведения

Журнал
Trudy Universiteta
Опубликовано
2026 · Т. 2, № 103, С. 443
Язык
Английский
ISSN
1609-1825 · 2710-3382
Издатель
Karaganda Technical University named after Abylkas Saginov
Аффилиации
Akhmet Baitursynuly Kostanay Regional University