Abstract

Gearboxes   serve   as   critical   components   in   power transmission  systems,  and  their  reliable  fault  diagnosis  is essential   for   ensuring   operational   safety  and   durability. Although blind deconvolution (BD) techniques are widely used to extract periodic impulsive features from vibration signals, most existing methods rely on accurate phase information— typically   requiring   additional   encoder   installation,   which increases   system   complexity   and   limits   their   practical application.  To  address  this  limitation,  we  propose  a  new 

framework  that  integrates  BD  with  dynamic  time  warping (DTW), enabling robust signal alignment and feature extraction under significant phase estimation errors. The proposed method enhances  impulsive signatures  while  addressing  phase misalignment  by  DTW-based  sample-wise  signal  alignment, enabling  the effective extraction of periodic fault signatures without precise phase information. This approach also facilitates the separation of fault-induced impulses from random noise, thereby  addressing  a  key  limitation  of  conventional  phase- dependent techniques. The proposed method is experimentally validated using a planetary gearbox testbed, where vibration data  were  acquired  under  controlled  fault  and  normal conditions. Compared to conventional methods, the proposed method  demonstrates  superior  sensitivity  in  capturing  weak periodic faults.  This work presents a practical, encoder-free solution for the  condition monitoring of rotating machinery, extending  the  applicability of  vibration-based  diagnostics  to environments where encoder use is infeasible or cost-prohibitive.