In nuclear power plant system, pump is the key equipment to maintain the flow of the primary loop coolant and the secondary loop heat transfer fluid. The main coolant pump and the feed water pump are in long-term operation status. Bearings are the key components to ensure stable operation of the pump, and which could be damaged in abnormal conditions. Once the failure occurred in the bearings, pumps would exhibit periodic vibration, which might cause the flow pulsations of coolant and heat transfer fluid; gradually, these situations could reduce the control accuracy and the stability of pump. Therefore, the detection and diagnosis of pump bearings are significant to improve the safety and stability of reactor system. We proposed an approach combined with signal processing and machine learning to extract the signal features and recognize the signal samples automatically. The proposed approach consists of three main steps: firstly, empirical mode decomposition (EMD) is applied to decompose the signals into several intrinsic mode functions (IMFs) which are corresponding to the different components of the original signals; secondly, calculating the correlation coefficient between each IMF and the original signal, the correlation coefficient sequence imply the components distribution of the signal which can be applied to recognize the signal samples; finally, extracting a part of correlation coefficient sequences to train the support vector machine (SVM), and then an classifier can be obtained and use to recognize the other signal samples automatically. Experimental results show that this method can effectively detect the pump bearing operating conditions and failures, and can provide a reference for the safe and stable operation of reactor pumps.
- Nuclear Engineering Division
Pump Bearing Fault Detection Based on EMD and SVM
Feng, Y, Li, X, Ke, Z, Chen, Z, & Tao, M. "Pump Bearing Fault Detection Based on EMD and SVM." Proceedings of the 2018 26th International Conference on Nuclear Engineering. London, England. July 22–26, 2018. V001T01A008. ASME. https://doi.org/10.1115/ICONE26-81584
Download citation file: