Research On Turbopump Fault Diagnosis Method Based On Fusion Of Acoustic And Vibration Signals

As a key component of the liquid propellant rocket engine,the turbopump works in extreme environments of high temperature,high pressure,and severe vibration.Once a defect occurs,it will quickly develop into a malfunction,leading to turbopump failure or even engine damage.Due to the limited installation position of the sensor on the turbopump,from the perspective of fault diagnosis technology,it is impossible to obtain the test data of multiple measuring points,it is impossible to fully analyze its failure mode,and it is difficult to perform failure analysis.In the case of limited sensor measurement points,combined with the characteristics of non-contact measurement of acoustic signals,the study of fault diagnosis methods for acoustic-vibration information fusion is realized.In this paper,the turbopump system of the liquid propellant rocket engine is simplified into a rotor-bearing system,combined with a method based on vibration signal,an acoustic signal,and acoustic-vibration signal fusion,and the turbopump fault diagnosis method is studied.The main research contents include:(1)According to the structure of the rolling bearing and its characteristic frequency calculation formula,combining the structural characteristics of the bearing and comprehensively considering the effects of amplitude modulation and frequency modulation,a model of the vibration signal of the bearing received by the vibration sensor is established;combined with the characteristics of the model,a simpler rolling bearing is derived The characteristic frequency calculation method,and the calculation formula is given.(2)Aiming at the instability of acoustic signal analysis based on the frequency domain,and time-frequency domain,a one-third octave method is proposed.This method distinguishes the spectral difference of each state and combines the K-nearest neighbor value Classification,support vector machine,and decision tree methods to improve the instability of acoustic signal recognition and realize stable fault diagnosis.(3)Given the goals and strategies of fault information fusion diagnosis,the commonly used information fusion methods are summarized,and the advantages and disadvantages of these methods are analyzed.Aiming at the situation that DempsterShafer’s evidence theory is not applicable when there is high conflict between evidence bodies,combined with Dempster-Shafer’s evidence fusion method and Jousselme distance to measure evidence conflict,an improved DS evidence theory is proposed.(4)Combining the K-nearest neighbor value and the improved DS evidence theory,a multi-information fusion fault diagnosis model based on KNN and improved DS evidence is proposed.The information of multiple types of sensors is successfully fused and verified by experiments.The test results of laboratory rotor-bearing test platform data show that the proposed bearing vibration model,one-third frequency doubling method,and the multi-information fusion fault diagnosis model based on KNN and improved Dempster-Shafer’s evidence are all effective for fault diagnosis.This research is of great significance to the fault diagnosis of electromechanical systems and turbopumps.