02-05-2023 дата публикации
Номер: CN116049638A
Принадлежит:
The invention discloses a transformer vibration detection method, a transformer vibration detection system, transformer vibration detection equipment and a storage medium. On one hand, a global-local cross comparison attention network is proposed to enhance interaction between a global image and a local highlight area; on the second aspect, a fault-normal cross comparison attention network is proposed to establish comparison between a fault image and a normal image, local features of the fault image are further determined, more complementary parts are found and recognized, and layering and classification of faults with smaller fine grit are evaluated more effectively; and on the third aspect, an offline training mode is adopted for the double cross comparison attention network, the attention network is further optimized, meanwhile, the operation speed of online fault detection is increased, finally, short-time fault prediction is performed on images without faults, and the data images after ...
Подробнее