Adaptive sparse-tree structure noise reduction method of very noisy vibration signal of main reducer

Abstract

The invention discloses an adaptive sparse-tree structure noise reduction method of a very noisy vibration signal of a main reducer. The method comprises a first phase in which position points relatedto a significant feature are detected from signals, and serve as basis of adaptive regular weight adjustment, and a wavelet coefficient related to the signal feature is reserved; and a second phase in which a tree-structure wavelet coefficient is estimated on the basis of the adaptive weight. The correlation structure between wavelet coefficients is used, the tree structure sparsity of noise reduction estimation can be improved by regular least square regression, noise components can be filtered from the very noisy vibration signal, the frequency modulated component related to a fault featureis reserved, and effective characteristic information in the signals is avoided from smooth transition.

Description

technical field [0001] The invention belongs to the technical field of vibration reduction and noise reduction, and in particular relates to an adaptive sparse tree structure noise reduction method for strong noise vibration signals of a main reducer. Background technique [0002] The vibration signal under the normal state of the final drive mainly includes the meshing frequency and its harmonic components, but due to the presence of strong noise components, the collected vibration signal is a non-stationary signal and contains some irrelevant frequency components. [0003] In the background of strong noise, due to the interaction and influence between the noise frequency and the natural frequency (mesh frequency and rotation frequency) of the final drive, the presented vibration signal components are relatively complex, and the modulation sidebands are not obvious. Comparing the time domain waveform and the frequency domain waveform in the normal state, the vibration ampli...

AI Technical Summary

Problems solved by technology

However, due to the presence of noise components in the vibration signal, the modulation frequency components in the spectrogram are complex and cannot clearly reflect the sideband components of the meshing frequency and its harmonics

Especially in the frequency domain waveform of bumping faults, some interference frequency components caused by noise appear. Therefore, it is difficult to judge the type of fault by analyzing the frequency modulation characteristics of the frequency domain waveform

When the main reducer is in the early stage of failure, the fault pulse is relatively weak, and it is easy to be submerged in strong noise, which increases the difficulty of feature extraction and reduces the accuracy of fault diagnosis

[0005] It can be seen that based on the low signal-to-noise ratio of the vibration signal collected in the strong noise environment, it is difficult to extract effective fault features for fault diagnosis

In addition, due to the fact that the vibration amplitudes of various faults are different from each other under the background of strong noise, some fault features are often considered as noise components and filtered, and excessive smoothing of high-frequency features occurs

Images