An adaptive multiscale fully convolutional network for bearing fault diagnosis under noisy environments

•A novel end-to-end adaptive multiscale fully convolution network for bearing fault diagnosis under noisy environments is proposed.•The proposed AMFCN significantly enhances the feature extraction ability, noise immunity and robustness of CNN.•The AMFCN performs state-of-the-art under various noise environments on CWRU and Paderborn bearing fault datasets.