Axle box bearing fault diagnosis via an adaptive feature mode decomposition based on dual-index fusion

Considering that the early fault characteristics contained in the vibration signal of axle box bearings are easily affected by complex transmission paths and wheelset impact noise, a parameter optimization feature mode decomposition (POFMD) method based on dual-index fusion is proposed. Firstly, the square envelope of the time-domain signal autocorrelation function is introduced into the calculation of traditional Gini index and defined as ACSESGI. Secondly, the weight values of ACSESGI and the weight values of squared envelope kurtosis of different mode components obtained from FMD decomposition are calculated respectively. Finally, the above two weight values are added together and utilized as the objective function for parameter optimization. The effectiveness of presented method is verified through early fault diagnosis results of scaled-down axle box bearings and full-size axle box bearings, and its superiority was further validated by comparing with variational mode decomposition and Autogram.