 The proposed method uses the whale optimization algorithm, WOA, to adaptively determine the penalty factor and decomposition layers in the variational mode decomposition, VMD, algorithm. It then selects intrinsic mode function, IMF, components that have a high correlation with the original signal and reconstructs it to remove the noise. Finally, the K nearest neighbor, KNN, method is used to construct the fault diagnosis model of the graph attention network, JT. This method has been shown to be effective at reducing noise in the signal and achieving high accuracy in diagnosing rolling bearing faults. This article was authored by Yaping Wang, Xing Zhang, Rufan Chao, and others.