A novel manifold learning denoising method on bearing vibration signals

Bearing failures are a major source of problem in rotating machines. These faults appear as impulses at periodic intervals resulting in form of specific characteristic frequencies. However, the characteristic frequencies are submerged in noise causing by a result of small imperfections in the balance or smoothness of the components of the bearing. To retrieve the characteristic fault frequencies of the vibration signal, signal denoising is an essential processing step in fault diagnosis of the bearings. This paper presents time-frequency analysis and nonlinear manifold learning technique for denoising vibration signals corrupted by additive white Gaussian noise. According to keeping the computing time acceptable, a novel manifold learning denoising method is put forward combining data compression and reconstruct operations. Simulation and experiments are employed to verify the feasibility and effectiveness of the proposed method on bearing vibration signals. Furthermore, this method can be used in other fault detection fields, such as engine, suspension device, and vehicle structures.