Research on Feature Extraction and Pattern Recognition of Tiny Internal Leakage of Check Valve
Check valves are widely used in hydraulic systems of construction machinery,agricultural machinery and military vehicles,the leakage is a common fault of check valves.This paper proposes a fault diagnosis method of check valve tiny internal leakage based on multi-source,multi-domain,multi-scale feature extraction and machine learning.First of all,the empirical mode decomposition(EEMD)is performed on the vibration signals and pressure signals of the four types of leakage failures.Secondly,the singular value,form factor,entropy and other methods from time domain,frequency domain and time-frequency domain are used to extract features and construct fault feature vector.Finally,the particle swarm-support vector machine algorithm are adopted to classify the leakage fault patterns.Experimental results show that the method can effectively detect leakage and the pattern recognition accuracy of leakage is over 90%.This paper laid a foundation for the research on the leakage rate prediction of the internal leakage of check valves,which has a good engineering application prospect.
国家科技期刊平台
NSTL
万方数据
