APPLICATION OF QUANTITATIVE ANALYSIS OF GEAR FAULT IN μ-SVD NOISE REDUCTION ALGORITHM DIAGNOSIS

Abstract

 In order to extract the fault features of mechanical equipment submerged by strong background noise, a general subspace denoising algorithm based on Singular value decomposition (SVD) is used to process the signal, that is, μ-SVD denoising algorithm. This algorithm overcomes the disadvantage of traditional wavelet threshold denoising algorithm, which only deals with the wavelet coefficients point by point and ignores the whole structure of the wavelet coefficients. The traditional SVD denoising algorithm is a special case when Lagrange multiplier μ = 0 in μ-SVD denoising algorithm. μ-SVD denoising algorithm includes filter factor, which can suppress the contribution of singular value dominated by noise contribution to the signal after denoising. The parameter selection method of μ-SVD denoising algorithm is discussed, and the influence of denoising order and Lagrange multiplier on denoising effect is emphatically studied. The test results of gear fault simulation signal and gear fault vibration signal at the early stage show that μ-SVD denoising algorithm is better than traditional SVD denoising algorithm in denoising effect. It can extract gear fault features better under strong background noise.