Fault Identification of Rolling Bearing Based on Improved EMD-Kurtosis Criterion and DBN

In view of the difficulty in extracting and recognizing fault features of rolling bearings, an improved KMI) algorithm is presented, which combines SVD dimension reduction, kurtosis criterion and DBN. Firstly, the original sampling signal is decomposed by the modified EMD, and a set of internal mode functions (IMFS) is obtained; then, Based on the maximum kurtosis criterion, the sensitive component with the highest kurtosis is chosen as the intrinsic mode function (IMF), and the singular value decomposition (SVD) is performed. Then, the singular value of the decomposition is used as the characteristic parameter to construct the characteristic vector. Finally, this paper applies the Deep Belief Network (DBN) to recognize the fault type of bearing. The simulation results show that the fault identification of rolling bearings by this method reaches more than 95.0 %.