Feature Extraction Method Of Burst Signal For Bearing Fault Of Liquid Propellant Rocket Engine

Turbo pump is a key component of liquid rocket engine,which works under high temperature,low temperature,high pressure,low pressure and high speed,and has complex structure.Rolling bearing is a key component of turbo pump that maintain high-speed operation for long periods of time,while withstanding various forms of stress extrusion,friction,and working in ultra-low temperature environments.Compared with ordinary civil bearing,the bearing of the rocket engine turbine pump is self-lubricated by solid,which can easily lead to structural damage when the bearing is subjected to external forces in an instant,or ablation caused by instantaneous dry friction.The characteristics of its vibration signal are:(1)The vibration data contains the low frequency component,and also the high frequency component.(2)The magnitude of the fault characteristic frequency of the bearing is relatively lower,which is easy to be obliterated under the complicated background noise.(3)Once the bearing surface is failing,its deterioration will be rapidly.These result that the fault characteristic frequency is abruptly appeared and lasted short time.For the fault diagnosis of cryogenic high-speed bearing of liquid rocket engine,to extract the fault features effectively is the key step.The existing vibration signal analysis methods are not suitable for rocket engine turbine pump bearings completely.Therefore,how to extract the fault feature of liquid rocket engine bearing quickly and accurately has become an urgent problem to be solved in related field.In this doctoral thesis,the theory,method and key technology of feature extraction of burst signal for bearing fault of liquid propellant rocket engine are studied in depth,the main research work is as follows:(1)For the problem that hidden mutation signal is difficult to pinpoint in time-domain,a new algorithm for getting the appearing and the disappearing moments of a specified signal from vibration data is proposed.Firstly,the characteristic frequency of the signal is extracted by discrete Fourier transform,and then the time-lapse Fourier analysis of the frequency is executed.The appearing and disappearing moments of the frequency can be shown by the inflection points of amplitude-time curve which is determined by time-lapse Fourier analysis.Finally,the amplitude of the frequency can be obtained by the Fourier analysis of the appearing-disappearing period.Through theoretical analysis,theory simulation and engineering simulation,the algorithm can accurately capture the appearing and disappearing moments of mutation signal in the vibration data.(2)Aiming at the problem of extracting the characteristic frequency of abrupt signal in the vibration data of turbine pump bearing of liquid rocket engine,an abrupt parameter analysis method is proposed.Five numerical parameters of the model are constructed based on the analysis of amplitude-time inflection and the abrupt feature of signal can be amplified by these parameters.Through the verification of theoretical analysis,engineering simulation and high-speed bearing vibration data engineering analysis,the algorithm can capture the abrupt frequency.(3)Aiming at the problem that the fault position of turbine pump bearing in rocket engine is difficult to determine accurately,the vibration source coordinate analysis method is proposed by using the phase information of the amplitude-time curve.Four vibration sensors are placed in the right-angled coordinate system with known coordinates,and the amplitude-time curves are used to obtain the appearing moments of the same frequency reaching four sensors,and then the vibration source coordinate is calculated according to the position coordinates of four sensors and four appearing moments corresponding four sensors.The vibration source coordinate is judged whether it located in the bearing area and which specific part of the bearing it located in.The method of vibration source coordinate analysis is verified by using engineering simulation data,and the results show that it can be used to determine the specific part of the bearing fault.4)In order to determine the time interval of sudden signal occurrence for further amplitude-time analysis,three kinds of entropy planes based on multi-time scale are constructed:First,the traditional statistical complexity measure and power spectral entropy are combined to construct the complexity-entropy causality plane in frequency domain;Secondly,the normalized displacement entropy used in the original method is extended to Tsallis entropy and Renyi entropy,and the generalized displacement Entropy plane model is obtained;Thirdly,combined with displacement entropy and distributed entropy,the generalized displacement distribution entropy plane is constructed.Through the verification of mathematical simulation data and bearing test engineering data,the results show that all three entropy planes can find the window with abnormal value well.5)A low-temperature high-speed bearing fault diagnosis system is developed,which is an open and extensible research Platform.The platform integrates traditional analysis methods such as time domain,frequency domain,short time FFT,wavelet transform,wavelet threshold de-noising,EEMD,hilbert-transform(HHT),and also integrates my promoted the analysis methods of amplitude-time curve analysis,mutation parameter and entropy Plane.Based on this platform,mathematical simulation,engineering simulation and the construction of the time domain signal of bearing fault can be carried out.The all data analysis in this paper is done on the bearing fault diagnosis system.