Study on application of Principal Component Analysis to fault detection in hydraulic pump

This paper presents a method of squared prediction error changes based on Principal Component Analysis(PCA) of Q statistics to deal with real-time online fault detection of hydraulic pump. In this method, feature vector sample set expressed by frequency band energy information of wavelet packet decomposition is extracted by effective signal processing and feature. Then, establish main element model by normal samples and compare the samples with test samples achieved by Q statistics method to detect faults; Next, describe fault change characteristics with contribution diagram; lastly, test results of different fault types are researched through experimental data of center of spring failure, off-shoe, slipper and loose boot of axial piston pump.