Bearing fault diagnosis based on optimized variational mode decomposition and 1D convolutional neural networks

Abstract

Due to the fact that measured vibration signals from a bearing are complex and non-stationary in nature, and that impulse characteristics are always immersed in stochastic noise, it is usually difficult to diagnose fault symptoms manually. A novel hybrid fault diagnosis approach is developed for the denoising signals and fault classification in this work, which combines successfully the variational mode decomposition (VMD) and one dimensional convolutional neural network (1-D CNN). VMD is utilized to remove stochastic noise in the raw signal and to enhance the corresponding characteristics. Since the modal number and penalty parameter are very important in VMD, a particle swarm mutation optimization (PSMO) as a novel optimization method and the weighted signal difference average (WSDA) as a new fitness function are proposed to optimize the parameters of VMD. The reconstructed signals of mode components decomposed by optimized VMD are used as the input of the 1-D CNN to obtain fault diagnosis models. The performance of the proposed hybrid approach has been evaluated using the sets of experimental data of rolling bearings. The experimental results demonstrate that the VMD can eliminate signal noise and strengthen status characteristics, and the proposed hybrid approach has a superior capability for fault diagnosis from vibration signals of bearings.

Variational mode decomposition for surface and intramuscular EMG signal denoising

“…To avoid mode mixing in VMD, the components of each VMF are obtained concerning the frequency domain of the signal. VMD-based denoising has been successfully applied for seismic time-frequency analysis, for the detection of gearbox fault diagnosis [18], wind speed forecasting [19], bearing fault diagnosis [20], denoising of biomedical images [21], ECG and EMG signal denoising [22,23]. proposed a new SOFT iterative thresholding-based VMD (SIT-VMD) to denoise sEMG signals in comparison with wavelet and EMD-based denoising methods [23].…”

A Width Multi-Scale Adversarial Domain Adaptation Residual Network with AConvolutional Block Attention Module

“…Different from the traditional machine learning methods, deep learning utilizes the end-toend idea to adaptively extract the deep fault information and features in the bearing vibration signals, which does not require the prior knowledge of experts, so it can better meet the requirements of big data development [19][20][21][22]. Wang et al combined variational mode decomposition (VMD) and CNN to improve the fault classification effect of the network by eliminating the influence of signal noise [23]. Zhang et al combined the STAC hyperbolic tangent network and the residual network to adaptively extract the bearing fault features [24].…”

Rolling Bearing Fault Diagnosis Based on Successive Variational Mode Decomposition and the EP Index

“…Although these two parameters can be directly pre-set by experience or experiments, the method has the drawback of blindness and is difficult to obtain excellent performance of VMD. Consequently, researchers usually utilize some intelligent optimization algorithms to determine the values of the two parameters, such as the genetic algorithm [ 24 , 25 ], particle swarm optimization [ 16 , 26 ], differential search algorithm [ 27 ], Archimedes optimization algorithm [ 28 ], grey wolf optimization [ 29 , 30 ], whale optimization algorithm [ 31 ], cuckoo search algorithm [ 32 ], sparrow search algorithm [ 33 ], and so on.…”