Clustering of frequency-based vibration signal for bearing fault detection
Bearing is one of the vital parts in any rotating machinery. Failure of this particular part can affect the machinery performance and in time will cause major failure to the machinery. Due to this crucial problem, on-line monitoring has become an alternative in prevention maintenance. The objective...
| Summary: | Bearing is one of the vital parts in any rotating machinery. Failure of this particular part can affect the machinery performance and in time will cause major failure to the machinery. Due to this crucial problem, on-line monitoring has become an alternative in prevention maintenance. The objective of this project is to study the trend of frequency spectrum from different bearing defects and to apply clustering approach using Principle Component Analysis, PCA on frequency domain signals. A set of good condition bearing is used along with four types of defective bearing which are inner race defect,corroded defect, contaminated defect and lastly roller defect. The signals are acquired using a PCB piezoelectric accelerometer and a National Instrument Data Acquisition System (NI-DAQ). The bearing will be run on three speed rotation which is 440, 1480 and 2672 RPM. The data is acquired by using DASYLab software, for both the time domain and frequency domain signals. The data then analyzed using PCA method through MATLAB software. Data is then plotted on scatter plot. After that, the data will be clustered using Agglomerative Hierarchical Clustering where a dendrogram is used to show a cluster of data in which the respective data for all types of bearing tested remain in their cluster. Finally, this method is suggested as an alternative in bearing fault detection, especially online monitoring. |
|---|