Research and application of improved adaptive MOMEDA fault diagnosis method

Abstract

In a strong noise environment, vibration signals are easily submerged by noise. In recent years, many scholars have studied a large number of noise reduction methods. In 2017, Multipoint Optimal Minimum Entropy Deconvolution Adjusted (MOMEDA) is applied to the fault diagnosis of gearbox. Although MOMEDA overcomes Maximum correlation kurtosis deconvolution (MCKD) defects and it can extract continuous impulse signal, but it still has the following problems: 1) It can only extract single periodic pulse. If we want to extract the characteristics of multiple periodic pulse signals, we need to further update the algorithm; 2) In a strong noise environment, MOMEDA can also search for a fixed periodic signal, but most of the information is false component, it is easy to cause misdiagnosis, therefore, the signal needs to be preprocessed; 3) The accuracy of MOMEDA noise reduction is affected by the search interval and filter size, and McDondald did not reasonably explain them, so an adaptive selection method is needed. Considering these problems. Firstly the article preprocesses the composite fault with ensemble empirical mode decomposition (EEMD) and then reconstructs the intrinsic mode function with the same time scale. Further, proposing kurtosis spectral entropy as the objective function, the grid search method is used to search the filter length of MOMEDA, and the reconstructed intrinsic mode function is further denominated by MOMEDA. Finally, the proposed method is used to search the complex fault pulse signals in strong noise environment. It proves its reliability.