Weak Crack Detection for Gearbox Using Sparse Denoising and Decomposition Method

The localized faults of rotating machinery can be diagnosed by the extraction of the periodic transient impulses (PTIs); however, the PTI is generally weak and may get submerged in strong noise. To address this issue, a novel approach based on nonconvex sparse regularization denoising and adaptive sparse decomposition is proposed. The main work can be divided into two areas: 1) raw signal denoising and 2) repetitive impulses isolation. Specifically, for the raw signal denoising, the augmented Huber function is proposed as penalty function, and the convexity of the objective cost function (OCF) can be maintained; meanwhile, the solution of the proposed OCF can be solved using forward-backward splitting algorithm. For the repetitive impulses isolation, due to the commonly used resonance-based signal sparse decomposition (RSSD) method lacks adaptability and flexibility in practical application, the genetic algorithm (GA) is introduced to optimize the decomposition parameters of the RSSD that are selected adaptively in the desirable range according to the denoised signal in terms of the global optimization characteristic of GA. As an example, a pinion gear with weak root-crack failure is investigated based on the proposed approach. Compared to some state-of-the-art methods such as L1-norm fused lasso optimization (LFLO) and maximum correlated kurtosis deconvolution method, the results demonstrate that the proposed approach can effectively extract the weak fault frequency and its harmonics, and the shortcoming of the systematic underestimation of LFLO method has also been greatly improved.