Gearbox fault diagnosis process based on maximum kurtosis entropy deconvolution of lpso

When multiple faults coexist, different fault features need to be decomposed into different intrinsic mode functions. Therefore, EEMD is needed to decompose the signal. It can not only eliminate the high noise component independent of the original signal, but also decompose different time scales into different intrinsic mode functions. Considering the existence of mode mixing in EEMD, the idea of mode function reconstruction is introduced to improve the energy of shock signal and eliminate mode mixing. In order to determine the length of the filter adaptively, the filter bank is used to optimize the length of the filter, which not only avoids the subjectivity of artificial setting, but also improves the accuracy of parameter selection. Finally, the envelope spectrum of the de-noising signal is analyzed to determine the final fault characteristics. The process steps of fault diagnosis are as follows:

Firstly, the vibration signal is decomposed by EEMD.

(2) determine whether there is modal mixing, and remove the weak correlation between high-frequency noise component and eigenfunction.

(3) if there is mode mixing, the same mode function is reconstructed to obtain CMF1, cmf2, etc.

(4) solve the envelope spectrum of the above eigenmode function, and calculate the kurtosis and ES value.

(5) calculate the kurtosis spectral entropy KSE.

(6) in the lpso algorithm, set the number of particles M and the number of iterations n, set the maximum and minimum inertia factor, and initialize L. In order to accurately cover the whole frequency band of the fault, l should satisfy the inequality l > 2fm / Fs, where FM is the fault characteristic frequency and FS is the sampling frequency.

(7) signals are input to the lpso. The improved lpso algorithm is used to optimize the length of mksed filter, and the optimal solution l0 is obtained.

(8) the optimal length l0 is input to the mksed filter, and the kurtosis spectral entropy of the fifth step is used as the objective function to denoise the mksed of CMF1 and cmf2. The optimal filter is obtained and the de-noising signal is output.

(9) the output signal is demodulated by the envelope spectrum to determine the fault characteristics.