To address the limitations of traditional convolutional neural networks in learning key fault features, which affect the accuracy of rolling bearing fault diagnosis, this study proposes a multi-scale convolutional neural network fault diagnosis model fused with multi-attention mechanisms. The proposed method introduces a convolutional structure characterized by multi-channel and multi-scale properties, aiming to expand the receptive field of the network and effectively capture prominent features across different dimensions. Both the Self-Attention mechanism and the Convolutional Block Attention Module (CBAM) are enhanced and integrated into the multi-scale feature extraction model. These attention mechanisms work together to optimize the learning process of the network by reducing the influence of irrelevant signal components and adaptively amplifying the response to fault-related features. The model is validated using the CWRU dataset and the JUN dataset, demonstrating its superior performance and generalization ability.