Abstract: Deep learning (DL) models for bearing fault diagnosis often struggle to preserve discriminative features under variable operating conditions and are unable to effectively capture multiscale ...