Improved Multi-Bandwidth Mode Manifold for Enhanced Bearing Fault Diagnosis

Abstract Variational mode decomposition (VMD) has been proved to be useful for extraction of fault-induced transients rolling bearings. Multi-bandwidth manifold (Triple M, TM) is one variation the VMD, which units multiple fault-related modes with different bandwidths by a nonlinear learning algorithm named local tangent space alignment (LTSA). The merit TM method that bearing extracted contain low level in-band noise without optimization VMD parameters. However, determination neighborhood size LTSA time-consuming, and may have problem asymmetry in up-and-down direction. This paper aims improve efficiency waveform symmetry method. Specifically, multi-bandwidth consisting are first obtained repeating recycling (RVMD) bandwidth balance Then, performed on extract their inherent structure, natural nearest neighbor N, TN) adopted efficiently reasonably select neighbors each data point modes. Finally, weight-based feature compensation strategy designed synthesize low-dimensional features alleviate problem, resulting symmetric can represent real fault transient components. major contribution improved diagnosis pure symmetrical as real. One simulation analysis two experimental applications validate enhanced performance over traditional methods. research proposes advantages high efficiency, good removal capability.