A New Method for Quantitative Estimation of Rolling Bearings Under Variable Working Conditions

Quantitative estimation of fault severity in rolling bearings is crucial for making proper maintenance decisions. However, bearings usually operate under variable working conditions in actual industrial applications, which significantly affects the extraction of fault severity information. For addre...

Full description

Saved in:
Bibliographic Details
Published in:IEEE/ASME transactions on mechatronics 2024-02, Vol.29 (1), p.41-51
Main Authors: Yu, Yaoxiang, Gu, Xi, Ma, Weipeng, Guo, Liang, Gao, Hongli, Zhang, Guoli
Format: Article
Language:English
Subjects:
Data mining
Datasets
Estimation
Fault diagnosis
Feature extraction
Health indicator (HI)
Industrial applications
Labels
neural network
Neural networks
quantitative estimation
relative similarity
Roller bearings
Rolling bearings
Training
Working conditions
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Quantitative estimation of fault severity in rolling bearings is crucial for making proper maintenance decisions. However, bearings usually operate under variable working conditions in actual industrial applications, which significantly affects the extraction of fault severity information. For addressing this issue, a new health indicator construction method based on a quantitative estimation neural network is proposed. First, a dataset is constructed, where each sample datum corresponds to two labels: a severity label and a speed label. The former represents the fault severity, while the latter denotes the running speed. Then, the quantitative estimation neural network is designed to extract speed-invariant and severity-variant features from the sample data. This network consists of three modules: a feature extractor that extracts features from the sample data, a speed discriminator that ensures the speed-invariance of the features, and a severity classifier that ensures the severity-variance of the features. Finally, relative similarity is used to measure health indicators for rolling bearings. Two bearing datasets collected under variable speeds are used to validate the effectiveness of the proposed method. The results demonstrate that the proposed method can effectively reduce the impact of variable speeds and quantitatively estimate the fault severity in rolling bearings.
ISSN:1083-4435
1941-014X
DOI:10.1109/TMECH.2023.3267699