Bearing fault diagnosis based on COCOA parameter optimization

To address the limitation of existing methods that often ignore periodic fault characteristics the fault periodicity characteristics resulting in fault extraction is not rich and the decomposition effect of the existing modal decomposition method is limited by the number of decomposition and the selection of penalty factors and other issues, this paper proposes a convolutional neural network fault diagnosis model based on COCOA parameter optimization. First, the filter parameters are dynamically adjusted by the Chaotic Fishhawk Co-optimization Algorithm (COCOA), and the signal is decomposed using the FMD, and then the decomposed modes are evaluated by the fitness function, and the modes containing fault information are selected as inputs to the subsequent diagnostic model; Then the bearing vibration signal is used as the input signal of both one-dimensional convolutional neural network and two-dimensional convolutional neural network, and the extracted fault information is fused in the convergence layer, and finally the fault is recognized by the classifier. Experiments show that the CNN model based on COCOA optimization shows obvious advantages in bearing fault diagnosis, and compared with CNN, VMD-CNN and FMD-CNN the model has excellent performance and fast processing speed.