Rolling bearings, as the core components of mechanical operation, their failures can lead to the deterioration of rotating machinery's operating conditions. Convolutional networks, as a method for diagnosing rolling bearing faults, address their fixed window limitations by leveraging the advantages of 1D convolutional neural network(1D-CNN) in processing one-dimensional data. Utilizing the multiscale concept, different-sized windows were used simultaneously at the same layer to extract signal features. Considering the impact of time dimension information on anomaly detection methods, the pooling layer of the neural network was combined with the Koopman model to obtain higher-order dynamic features. Finally, the fault features obtained were inputted into a fully connected layer for fault diagnosis. To verify the advantages of the model, the proposed initial model and two improved models were compared under the same working conditions, alongside a comparative analysis with algorithms such as support vector machines (SVM) and back propagation neural network(BPNN) . The results show that the proposed model has a better recognition effect, with the accuracy of rolling bearing fault reaching 99.99%.