Reciprocating compressor bearing fault feature extraction method

Abstract

The invention provides a reciprocating compressor bearing fault feature extraction method, and belongs to the field of fault diagnosis. The reciprocating compressor sliding bearing vibration acceleration signal is decomposed by using an optimized smooth prior analysis model method, the de-trending term and trend term data of the vibration signal are obtained, effective noise reduction is performed on the reciprocating compressor signal, and by solving the sample quantile permutation entropy of the de-trending term data, the reciprocating compressor sliding bearing vibration acceleration signal is obtained. The nonlinear behavior of the bearing vibration signal is quantitatively described, and the fault features of the reciprocating compressor can be more accurately diagnosed by using the feature vector formed by the sample quantile permutation entropy. According to the method, the problem that the trend term removing effect is not obvious due to the fact that the smooth prior analysis model parameter lambda is selected according to experience is solved, the parameter lambda is optimized through the sparrow search algorithm, effective noise reduction of the reciprocating compressor signal is achieved, feature extraction of the reciprocating compressor vibration signal is conveniently achieved, and the method is suitable for popularization and application. And reliable data support is provided for characterization and identification of fault signals of the reciprocating compressor.

Description

technical field [0001] The invention belongs to the field of fault diagnosis, in particular to a method for extracting fault features of a reciprocating compressor bearing. Background technique [0002] Reciprocating compressor, as the product with the widest pressure range, the largest quantity and the longest history in the compressor industry, occupies an irreplaceable position in the industry. As a large-scale mechanical equipment, the reciprocating compressor, once a failure occurs, will lead to a shutdown of production, affecting the production efficiency, and serious damage to the mechanical equipment, and even cause casualties. Therefore, the fault diagnosis of the reciprocating compressor is of great economic significance. and project value. [0003] The reciprocating compressor will be affected by various factors such as impact, friction, etc., so its signal has nonlinear and non-stationary characteristics, resulting in inconspicuous fault characteristics, which i...

AI Technical Summary

Problems solved by technology

As a large-scale mechanical equipment, reciprocating compressors, once they fail, will lead to stoppage of work and production, affecting production efficiency, and serious damage to mechanical equipment, and even casualties. Therefore, fault diagnosis of reciprocating compressors is of great economic significance and engineering value

[0003] The reciprocating compressor will be affected by various factors such as impact and friction during work, so its signal has nonlinear and non-stationary characteristics, which makes its fault characteristics not obvious, which is not conducive to extraction

The characteristics of the reciprocating compressor signal determine that the traditional linear signal processing method is no longer applicable, so many nonlinear signals have been widely used, and the first application is wavelet analysis, but wavelet analysis is too large due to the selection of wavelet basis functions. Complex, and the results obtained by different wavelet basis functions vary greatly

Subsequently, a series of adaptive decomposition methods appeared: EMD, LMD, VMD, etc., but these methods are all affected by the number of decomposition layers K. If the number of decomposition layers is too large, modal aliasing will occur. small, the decomposed signal cannot completely reflect the vibration characteristics of the signal. Based on this, this paper proposes to use the smooth prior analysis model method to adaptively decompose the vibration signal of the reciprocating compressor. The advantage of this method is that the signal is decomposed into trend items and decomposed Two items of the trend item can fully describe the vibration characteristics of the signal, but the smoothing prior analysis model decomposition method will be affected by the regularization parameter λ

Images