Aiming at the improvement of Multipoint optimal minimum entropy deconvolution adjusted (MOMEDA)

an enhanced MOMEDA algorithm is proposed and applied to the weak fault diagnosis of rolling bearings. Firstly

a frequency domain index that can reflect the filtering effect is constructed. It is found that the filtering effect of MOMEDA depends on the fault period rather than the filtering length. Then

a method combining autocorrelation

envelope demodulation and multipoint kurtosis (MKurt) is proposed to determine the optimal fault period. The rolling bearing fault and normal data of the external casing measuring point are used for verification. The results show that the method can effectively overcome the problem that the traditional MOMEDA algorithm is difficult to select the fault period

and can adaptively extract the more significant fault period. Finally

compared with the MED-based collaborative diagnosis method

it is found that MOMEDA focuses on determining the fault cycle

while the MED-based collaborative diagnosis method focuses on selecting the filter length. After the fault cycle is correctly extracted

the MOMEDA filtering effect is better.