Study On Low Frequency Acoustic Feature Extraction Method Of Hydraulic Rotating Machinery Based On Principal Component Analysis

For hydraulic rotating machinery,such as pump,hydroturbine and propeller,accurate identification of the vibration and noise characteristic signal parameters is the key problem of condition monitoring,fault early warning and target recognition.The inspired typical features of hydraulic rotating machinery operation process,such as shaft and blades frequency,have low frequency characteristic,which is difficult to directly monitoring the features under environmental interference.For the low frequency characteristics features extraction of hydraulic rotating machinery,the mechanism of low frequency radiation acoustic signal and noise signal analysis model of centrifugal pump and propeller are studied in this paper.The high anti-noise,high precision,high efficiency methods of low frequency acoustic feature extraction are proposed based on resonance demodulation method,cyclostationary analysis method,time-frequency analysis method and principal component analysis technology.In this paper,the main research contents are as follows:(1)The signal characteristics of hydraulic rotating machinery radiated noise signal are studied.Amplitude modulation and amplitude-frequency modulation signal models are established.Based on the component analysis of radiated acoustic signals,the time domain average method and signal preprocessing method based on autoregressive model are studied respectively.The high efficiency performance of signal preprocessing method based on autoregressive model is validated by simulation analysis,which provides a theoretical basis for the preprocessing of noise reduction and demodulation of hydraulic rotating machinery radiated acoustic signals.(2)Based on multiple frequency bands demodulation and principal component analysis technology,this paper proposed a high anti-noise demodulation method,called DMP.The modulation signal components in multi-band can be enhanced and extraction by principal component analysis technology,which improves the anti-noise performance of DMP method.Through the demodulation experiment of centrifugal pump under different working conditions and simulation analysis,the demodulation performance are verified.The low frequency acoustic features are extracted by DMP method.The DMP method shows a better anti-noise demodulation performance than classical kurtogram method.(3)Based on cyclostationary analysis and principal component analysis technology,a high precision demodulation algorithm is proposed,called DCP.The modulation signal components in second order cumulant matrix can be enhanced and extraction by principal component analysis technology,which improves the anti-noise performance and precision of DCP method.Through the demodulation experiment of propeller under different working conditions and simulation analysis,the low frequency features are extracted by DCP method under low signal to noise ratio condition.The DCP method shows a better demodulation precision performance than classical kurtogram and cyclostationary methods.(4)Based on time-frequency analysis and principal component analysis technology,a high efficientency demodulation method is proposed,called DTP.The modulation signal components in time-frequency distribution matrix can be enhanced and extraction by principal component analysis technology,which improves the anti-noise performance of DTP method.Through the demodulation experiment of propeller and centrifugal pump under different working conditions and simulation analysis,the low frequency features are extracted by DTP method under low signal to noise ratio condition.The DTP method shows a better demodulation efficient than classical kurtogram and cyclostationary methods.(5)For the proposed DMP method,DCP method and DTP method,the applicability of the demodulation methods is analyzed.Through the demodulation experiment of propeller under different working conditions and simulation analysis,the applicability of the demodulation methods is verified.DMP method is suitable for the low frequency characteristic signal extraction under low signal to noise ratio condition.DCP method is suitable for the low frequency characteristic signal extraction with high precision.DCP method is suitable for the low frequency characteristic signal extraction of rapid and on-line extraction.In this paper,the low frequency acoustic feauture extraction methods of typicall hydraulic rotating machinery are studied in the case of centrifugal pump and underwater propeller.The high anti-noise,high precision and high efficiency low frequency acoustic feature extraction methods are proposed,which can be used for low frequency acoustic features extraction under multi conditions and complex environment interference.The proposed methods provide the research foundation for the condition monitoring,fault diagnosis and target recognition of hydraulic rotating machinery.