Bearing fault recognition based on wavelet frequency band division and multi-level clustering_21_program_code

Citation Author(s):
Central South University of Forestry and Technology
Central South University of Forestry and Technology
Central South University of Forestry and Technology
Central South University
Central South University of Forestry and Technology
Submitted by:
Yuanyuan Liu
Last updated:
Sat, 01/22/2022 - 08:50
0 ratings - Please login to submit your rating.


 It is the key problem of machine condition monitoring to judge whether the rolling bearing has a fault or not and judge the fault location according to the noise signal. Aiming at this problem, a rolling bearing fault identification method is proposed based on Wavelet Frequency Band Subdivision (WFBS), Principal Component Analysis (PCA) and Multi-Level Clustering (MLC). Firstly, the original signals are divided into different Wavelet Packet Energy Spectrum indicators based on wavelet frequency bands, then they are combined with time-frequency domain indicators to form a parameter set. After PCA processing, the dimension of the parameter set is reduced to an appropriate dimension and the useful information is retained as much as possible. In this paper, the concept of WFBS and MLC are proposed to deal with the parameter set. The fault location and fault severity are determined in the form of confusion matrix graph through MLC. The experimental results show that this method can identify different fault types of rolling bearings with high accuracy and strong applicability, and displays practical significance in engineering applications.



Running instructions:

Procedure steps:

Required matlab version: Matlab 2018 b (or above).

There are four files (Package_wavlet_JSX.m,Package_wavlet_JSX.fig,three_cluster.m,three_cluster.fig) are important. All files are needed if you want to run this project correctly.

1. Open package_ wavlet_ JSX m file

2. click menu: Run

3. First,you can click "Copy xls delete txt data" to clear the cache

4. Select the imported data ("All data ") on the far right. It will take some time to load the data, and the time has passed when the main program command window appears Description (A messagebox shows 'Data loading completed').

5. Click the " PCA analysis"  to get the contribution rate and recommendation dimension.

6. Click the "Multi- level Clustering"  can help you to  anthor GUI interface.

7. Select data of different dimensions according to require and click "Ffirst clustering", "Second clustering" and "Third clustering" in turn


1. This is an original project on Bearing Fault Recognition Based on Wavelet Frequency Band Subdivision and Multi-Level Clustering.
2. A fusion method of energy spectrum indicators based on Wavelet Frequency Band Subdivision and time-frequency domain indicators are proposed. This method can make use of the difference between fault features and the complementary characteristics of information to separate the fault to the greatest extent.
3. The Principal Component Analysis reduce the data dimension significantly. It is convenient for later cluster analysis and improves the computing ability of whole system.
4. A Multi-Level Clustering system is proposed and constructed, which can distinguish different fault types and fault degrees. Then the clustering results are displayed through the confusion matrix, and the accuracy of fault classification is 99.3%.
5. If the number of principal components is 6 in the multi-level cluster analysis system according to the program steps, the classification and recognition accuracy can be 100%.

Submitted by Yuanyuan Liu on Sat, 01/22/2022 - 09:05