Research On Weak Signal Detection Technology Of Rolling Bearing Early Failure

Rolling bearing is one of the most widely used components in mechanicaldevices. As the machineries and equipments becoming more and more sophisticatedand complex, the requirements of precision and reliability for rolling bearing arehigher and higher. Bearing damage, even minor damage, may affect the normalfunction of the devices. So the detection of rolling bearing fault information,especially weak signal of early fault, is particularly important. This article applied“resonance-based sparse signal decomposition”(RSSD) into weak signal detectionof rolling bearing early fault, and got good results.We analysed the waveform characteristics of rolling bearing fault impulseresponse based on the single degree of freedom vibration model of the rollingbearing system, and proposed the two degrees of freedom vibration model which iscloser to the actual situation. We studyed the principle of selecting demodulationfrequency band and modulation characteristics of fault signal. Combined with theactually measured rolling bearing fault vibration signal, we analyzed thecomposition of rolling bearing early fault signals, as well as the causes of thesecomponents and the methods to hand them.The rule of the main parameters’s influence on RSSD results were investigated,and the simulation results verified that using RSSD to extract fault impulse responseis feasible. The sustained oscillations provoked by failure impact are fitted by thehigh-resonance component, and the abrupt peaks in vibration signal (mainly are theabrupt part of the random noise and the sharp peak in fault impact response) arefitted by the low-resonance component, while the uniform part of the noise ismainly included in the residual component.We studied the parameter selection principle of RSSD point at the rollingbearing fault signal, and proposed a fault information extraction method based onRSSD. The specific processes is: Firstly, determine the natural frequency to bedemodulated according to the vibration signal spectrum, then do RSSD on bearingvibration signal and extract the subbands near natural frequency in the tworesonance components. Then sum up these subbands respectively to get the mainsubbands of the two resonance components, and sum up these two main subbands toget the original signal’s main subband. Finally envelope demodulate these threemain subbands and obtain the autopower spectrum of the envelope signal s. Thepeaks in the autopower spectrum will reflect the degree and location of the fault inrolling bearing.In order to verify the effectiveness of the proposed method to extract faultinformation, we studyed the vibration signal of rolling bearing with different fault types and fault sizes, and in different working conditions. Then we compared thismethod with envelope analysis and wavelet analysis. The results show that thespectrum obtained by this method has less burrs, and a higher signal-noise ratio, andthis method performs much better in revealing the amplitude modulation frequencyof the fault impact response. The proposed method in this paper provides a newsolution for weak signal detection of rolling bearing early failure.