A noise reduction method based on adaptive weighted symplectic geometry decomposition and its application in early gear fault diagnosis

摘要

When the gear appears early fault, it will be accompanied by strong background noise, and the fault information is weak. Therefore, the result of noise reduction often determines whether the early gear fault can be accurately diagnosed. However, there are many defects in the existing methods of noise reduction. Wavelet decomposition (WT) requires setting parameters manually, and it is not adaptive. The ensemble empirical mode decomposition (EEMD) still has mode aliasing and endpoint effects. The singular spectrum analysis (SSA) and symplectic geometry mode decomposition (SGMD) select the useful components by energy size, which will delete the components with more fault information but less energy. Therefore, an adaptive weighted symplectic geometry decomposition (AWSGD) method is proposed for noise reduction in this paper. On the one hand, AWSGD is adaptive without setting parameters manually. On the other hand, AWSGD defines cycle kurtosis (CK) and periodic impact intensity (PII). CK is used to characterize the strength of periodic impact in the component, and PII is used to measure the fault information amount of the component. It can avoid the defect of the traditional noise reduction method by energy size. The noise reduction results of emulational and experimental signals show that AWSGD has excellent performance in noise reduction.