Multimodal unified generalization and translation network for intelligent fault diagnosis under dynamic environments

Multimodal data fusion can generate reliable fault representations for intelligent fault diagnosis. However, simple data fusion strategies often introduce fault-irrelevant information, thereby reducing robustness against unknown domain shifts. Moreover, traditional methods generally lack adaptive mechanisms to address missing modalities, leading to considerable performance degradation under sensor failure conditions. To address these problems, …