Research on fault diagnosis of rolling bearings based on multi-method fusion

To address the limitation that Variational Mode Decomposition (VMD) relies on empirical settings for the mode decomposition number <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML"> <mml:mi>K</mml:mi> </mml:math> and penalty factor <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML"> <mml:mi>α</mml:mi> </mml:math> , this paper proposed the RIME-VMD-KNN method for bearing fault diagnosis. Specifically, the RIME algorithm was used to intelligently optimize <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML"> <mml:mi>K</mml:mi> </mml:math> and <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML"> <mml:mi>α</mml:mi> </mml:math> of VMD, breaking the reliance on experience; Pearson Correlation Coefficient (PCC) was adopted to screen Intrinsic Mode Functions (IMFs) with high fault correlation for signal reconstruction, preserving key features; and the sample entropy of the reconstructed signal was input into KNN for fault identification. Experiments show that the optimization performance of RIME is superior to that of GA, GWO and AOA; the generalization ability is verified by supplementary tests on the XJTU-SY dataset; KNN is simpler and more efficient than SVM, proving the rationality of its selection; the confusion matrix and multiple random cross-validation confirm stability; and computing time and resource data are provided to verify the feasibility of embedded deployment. This method improves the reliability and real-time performance of diagnosis and has engineering value.