Research On Fault Diagnosis Methods For Rotating Machinery Under Variable Speed Conditions

With the development of rotating machinery in the direction of large-scale,complex and intelligent,its service conditions are becoming more and more severe,and core components such as bearings and rotors often fail,which seriously affects the safe and reliable operation of equipment.Timely and accurate diagnosis of core component failures has become the "killer card" to ensure the safe and stable operation of rotating machinery,thus,attracting many researchers and equipment engineers at home and abroad to carry out theoretical research and technology development.As we all know,due to frequent fluctuations in variable speeds,rotating machinery is essentially working under non-stationary variable rotational speed conditions,and the frequency of the vibration signal is expressed as a time function under the modulation of amplitude,frequency,and phase,which makes the fault feature extraction difficulty.Therefore,for the failure of core components such as bearings and rotors,this study takes the tacholess order tracking method as the main line,analyzes in detail the factors that affect the accurate extraction of fault features under variable speeds,and conducts in-depth time-frequency analysis,time-frequency ridge smoothing and filtering theory methods and applications.This paper summarizes the development history of fault diagnosis technology,analyzes,summarizes,and summarizes the domestic and foreign status quo of the four algorithms of time-frequency analysis,order tracking,and spectrum analysis and filtering,and points out that the current mechanical fault diagnosis methods under variable speed conditions.The inability to accurately extract fault features is the bottleneck and crux of current health monitoring and fault diagnosis.The speed signal is one of the important factors in the realization of the tacholess order tracking method.In order to accurately track the change trend of the non-stationary signal,the time-frequency analysis method is used to localize the signal.Aiming at the problem that the traditional linear time-frequency analysis method cannot match the changing chirprate,this paper proposes an adaptive chirprate matching method based on iterative synchrosqueezing generalized linear chirplet transform,which improves the readability of the time-frequency representation of the signal and realizes instantaneous frequency estimation.Aiming at the problem that the transient impact signal changes quickly and is difficult to capture,a second-order transient extraction S-transform method is proposed.This method accurately matches the impact characteristics of the signal through the derived second-order group delay operator.Simulation and experimental results show that this method has strong robustness.The discretization window function in the time-frequency analysis method will lead to an increase in the error between the estimated instantaneous frequency value and the actual value,and then affect the interpolation effect of the resampling technology in the tacholess order tracking method.Therefore,this paper establishes an adaptive weighted smoothing model to perform secondary processing on the fitted curve to reduce the error with the real value.The model does not require human intervention and prior knowledge to explore the criteria for setting model parameters.The tacholess order tracking method abandons the dependence of the key-phase signal,takes the time-frequency analysis method as the starting point,accurately estimates the instantaneous frequency of the rotating part,and finally transfers the non-stationary signal into a stationary signal by interpolation at equal angle intervals.Although the resampled could eliminate the frequency modulation effect of the signal,the fault characteristics are still seriously affected by the interference component.Based on this,the Teager energy operator enhanced Gaussian-Laplace filter and adaptive threshold filter methods are proposed to feature enhancement and eliminate the interference of background noise and inherent components.Vibration signals are affected by the transmission path and background noises during transmission and easily entagled into the unrelated interference components,which are resulting in serious signal redundancy consequences.Usually,these results will lead to a serious waste of resources in processing process and difficulty in identifying fault characteristics.Aiming at the above problems,using the advantages of(order)cepstrum,a model filtering method based on a combined penalty is proposed.This method eliminates the components in the signal that are irrelevant to the fault characteristics through the sparse representation of the signal and extracts the fault characteristics of the key parts of the rotating machinery.A new approach is provided.Combining iterative synchrosqueezing-based general linear chirplet transform with the above three filtering methods to form a signal feature extraction technology under constant/variable speed conditions.Based on the above principles,a high-end equipment remote monitoring platform has been developed.This system diagnoses the failure of key components for Garrett Ebara Pump Industry Co.,Ltd.and Tiande Pump Industry,improves the safety of pump products,and provides an effective tool for equipment health status monitoring and fault diagnosis.