Fault Diagnosis Methods Based On Rotating Speed Estimation For Bearings Under Variable Working Condition

Mechanical transmission systems are prone to failure due to the long-term operation under variable working conditions such as speed increase,speed reduction or variable load.Therefore,it is significance to develop feature extraction and fault diagnosis methods for variable speed rolling bearings.In this study,The parameterized time-frequency analysis(PTFA)methods are used to extract the instantaneous rotational speed from the vibration signal directly,and then combined with a series of high-efficiency signal processing methods to feature extraction.Some simulation signals and practical experimentations validate the effectiveness of the proposed method.The specific work arrangements are as follows:Sometimes,it is inconvenient to install the tachometer or encoder because of cost or design reasons.Without key-phase signal,conventional computer order tracking will be unavailable for varying speed rolling bearing fault diagnosis.To avoid this problem,kernel regression residual decomposition enhanced polynomial chirplet transform method is proposed to detect faults with time varying rotation speed.PCT extracts the speed curve form vibration signal itself and KSSD decoupling the complex original signal to detect fault feature.It’s applications on simulation and experimental signals prove that the KSSD enhanced PCT diagnostic method is powerful for rolling bearings fault diagnosis under variable conditions without speed sensors.In view of the excessive fluctuation or even sudden change of the speed,the conventional time-frequency analysis methods can’t provide high concentrated time-frequency representation and difficult to extract the speed curve accurately.An angle synchronous averaging method based on spline-kernelled chirplet transform to detect faults at variable speed is developed to solve the problem.ASA transforms the non-stationary time-domain signal into stationary angular-domain signal.Then it eliminates the influence of environment noise and nonsynchronous periodic impulses.Owing to the interference of noise and the operations of other components in the machine,rolling bearing fault characteristic is submerged and hardly to be detect.Hence,we propose a PCT-based maximum correlated kurtosis deconvolution approach to detect bearing faults under variable speed conditions.Kurtosis is easily affected by single or a small number of high amplitude pulses in the signal.MCKD overcomes the shortcomings of kurtosis,fully considers the periodic characteristics of the impact component,and effectively extracts fault characteristics.