Chapter 7
Prediction of Bearing Remaining Useful
Life Based on LSTM Network
Xin Wang, Jiazheng Guo, Jian Wang, Changying Liu, and Chuang Du
Abstract Bearing is one of the important components in the mechanical system, and
the reliability of the bearing is significant. The paper aimed to simulate remaining
useful life (RUL) prediction process in the engineering project. In the process of
predicting the RUL of the bearing, the noisy in the data leads to the over-fitting
phenomenon. In order to solve the problem, it is necessary to train a Variational
Auto-Encoder (VAE) for denoising short time series before the prediction starts. In
the paper, the VAE was selected to denoise the bearing performance degradation index
sequence. The denoised data curve was better than the original, which was smooth.
The denoised data were used to solve the bearing RUL prediction problem based
on LSTM network, and the over-fitting phenomenon did not occur. Compared with
the prediction results of directly using Support Vector machine Regression (SVR)
and Artificial Neural Network (ANN), the deviation of the prediction based on Long
Short-term Memory (LSTM) results from the observation values were smaller, and
the method can be applied in the engineering project.
Keywords Variational Auto-Encoder · LSTM network · Bearing remaining ·
Useful life prediction
7.1 Introduction
Bearing RUL prediction is one of the important components in Prognostic and Health
Management (PHM) system. In recent decades, varieties of methods have been
proposed to study bearing RUL prediction. The earliest appeared was to predict
the RUL of bearings based on statistical analysis. On the one hand, the formula for
X. Wang · J. Guo · C. Du
Key Laboratory of CNC Equipment Reliability, Ministry of Education, School of Mechanical and
Aerospace Engineering (SMAE), Jilin University, Changchun, People’s Republic of China
J. Wang · C. Liu (B)
School of Instrument Science and Electrical Engineering, Jilin University, Changchun, People’s
Republic of China
e-mail: liuchangy@jlu.edu.cn
© The Author(s), under exclusive license to Springer Nature Switzerland AG 2021
J. Xu and K. M. Pandey (eds.), Mechanical Engineering and Materials,
Mechanisms and Machine Science 100,
https://doi.org/10.1007/978-3-030-68303-0_7
81
Prediction of Bearing Remaining Useful
Life Based on LSTM Network
Xin Wang, Jiazheng Guo, Jian Wang, Changying Liu, and Chuang Du
Abstract Bearing is one of the important components in the mechanical system, and
the reliability of the bearing is significant. The paper aimed to simulate remaining
useful life (RUL) prediction process in the engineering project. In the process of
predicting the RUL of the bearing, the noisy in the data leads to the over-fitting
phenomenon. In order to solve the problem, it is necessary to train a Variational
Auto-Encoder (VAE) for denoising short time series before the prediction starts. In
the paper, the VAE was selected to denoise the bearing performance degradation index
sequence. The denoised data curve was better than the original, which was smooth.
The denoised data were used to solve the bearing RUL prediction problem based
on LSTM network, and the over-fitting phenomenon did not occur. Compared with
the prediction results of directly using Support Vector machine Regression (SVR)
and Artificial Neural Network (ANN), the deviation of the prediction based on Long
Short-term Memory (LSTM) results from the observation values were smaller, and
the method can be applied in the engineering project.
Keywords Variational Auto-Encoder · LSTM network · Bearing remaining ·
Useful life prediction
7.1 Introduction
Bearing RUL prediction is one of the important components in Prognostic and Health
Management (PHM) system. In recent decades, varieties of methods have been
proposed to study bearing RUL prediction. The earliest appeared was to predict
the RUL of bearings based on statistical analysis. On the one hand, the formula for
X. Wang · J. Guo · C. Du
Key Laboratory of CNC Equipment Reliability, Ministry of Education, School of Mechanical and
Aerospace Engineering (SMAE), Jilin University, Changchun, People’s Republic of China
J. Wang · C. Liu (B)
School of Instrument Science and Electrical Engineering, Jilin University, Changchun, People’s
Republic of China
e-mail: liuchangy@jlu.edu.cn
© The Author(s), under exclusive license to Springer Nature Switzerland AG 2021
J. Xu and K. M. Pandey (eds.), Mechanical Engineering and Materials,
Mechanisms and Machine Science 100,
https://doi.org/10.1007/978-3-030-68303-0_7
81