The Feature Extraction and Diagnosis of Rolling Bearing Based on CEEMD and LDWPSO-PNN

The vibration signals of rolling bearing are often highly nonstationary and nonlinear, and consequently it is not accurate to extract and identify the characteristics of these signals by the traditional methods. In order to improve the performance on the feature extraction from bearing signals and t...

Full description

Saved in:
Bibliographic Details
Published in:IEEE access 2020, Vol.8, p.19810-19819
Main Authors: Liu, Fuzheng, Gao, Junwei, Liu, Huabo
Format: Article
Language:English
Subjects:
Adaptive algorithms
Algorithms
CEEMD
Decomposition
Fault diagnosis
Feature extraction
global and local search ability
intelligent fault diagnosis
LDWPSO-PNN
Neural networks
Particle swarm optimization
Roller bearings
Rolling bearing
Signal processing
Vibration
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:The vibration signals of rolling bearing are often highly nonstationary and nonlinear, and consequently it is not accurate to extract and identify the characteristics of these signals by the traditional methods. In order to improve the performance on the feature extraction from bearing signals and the accuracy of the diagnosis, it requires effective signal processing and diagnose algorithms. In this paper, a new fault diagnosis algorithm which combines complementary ensemble empirical mode decomposition (CEEMD), probabilistic neural network (PNN) and particle swarm optimization (PSO) algorithm optimized by improved linear decreasing weight (LDW) algorithm is proposed. In this method, firstly the vibration signals are decomposed into a number of Intrinsic Mode Functions (IMFs) by the CEEMD algorithm since it has good adaptive ability to nonstable signals and can effectively extract fault features. Then the improved LDWPSO algorithm is introduced to solve the problem that the selection of smoothing factor in PNN model is arbitrary and uncertain. Finally, train and diagnose the fault types of rolling bearing using the LDWPSO-PNN model. The proposed method is verified by the experimental datasets. The results indicate that the method can extract the feature vectors of the vibration signals and distinguish them effectively.
ISSN:2169-3536
2169-3536
DOI:10.1109/ACCESS.2020.2968843