Fault diagnosis of train axle box bearing based on HMI-POA-VMD
To address the challenges associated with the harsh operating conditions of train axle box bearings,where fault signals are often obscured by noise and extracting fault features remains difficult,this study proposes a Variational Mode Decomposition(VMD)parameter optimization method.The method integrates envelope entropy and kurtosis into a Harmonic Mean Index(HMI)fitness function.The algorithm is validated using fault data from a proportional test bench.First,to ensure that the syn-thetic function effectively captures both the periodicity and impulsiveness of signals at comparable mag-nitudes,the harmonic mean index is introduced.This index,combining kurtosis and envelope en-tropy,serves as the fitness function,and the Pelican Optimization Algorithm(POA)is employed to perform a global search for optimal values.Second,the crucial parameters of VMD are optimized and determined using the HMI-POA algorithm,including the optimal decomposition layer number K and punishment factor α.These crucial parameters are then applied to decompose fault signals into K Intrin-sic Mode Function(IMF),with the optimal component identified based on the Weighted Kurtosis(WK)index.Finally,the envelope demodulation of the optimal component signal is performed to ex-tract the fault characteristic features of the rolling bearings.The proposed HMI-POA-VMD algorithm is validated using fault data from a proportional test bench.Its superiority is further demonstrated through comparison with traditional methods,using the Fault Feature Coefficient(FFC)as the evalua-tion criterion.Experimental results show that the proposed method significantly enhances the extrac-tion of fault frequencies.Compared to single fitness function optimization and traditional VMD,the FFC improves by 49.1%and 62.5%respetively.This highlights the method's capability to extract richer fault frequency information and accurately identify features in noisy environments.
rolling bearingvariational mode decompositionPOAharmonic mean indexfault fea-ture extraction
万方数据
维普
