Abstract In health condition monitoring of mechanical equipment, the signal is the key source of information. However, signal resolution is often degraded due to factors such as equipment vibration and electromagnetic interference. To address this issue, an Efficient Sub-pixel Convolutional Attention Residual Network (ESPCARN) built on the idea of signal resolution improvement is proposed in this paper. Firstly, the original low-resolution samples are input into a CBAM-ResNet to obtain more feature information of the channels and space within the residual connection and a multi-feature mapping with four channels was generated. Subsequently, the four-channel low-resolution features are aligned periodically through sub-pixel convolution layer, resulting in a set of high-resolution samples and the feature dimension of the data was increased to four times that of the original low-resolution data, thereby realizing the resolution enhancement. Finally, two experiments with different working conditions are established to evaluate the performance of the proposed fault diagnosis method, and the experimental results verified the efficacy of the ESPCARN framework.