Gearbox fault diagnosis method based on semi-supervised dynamic graph attention

The invention discloses a gearbox fault diagnosis method based on semi-supervised dynamic graph attention, and relates to the technical field of fault diagnosis, and the method comprises the steps: obtaining a fixed-length diagnosis sample from an original vibration signal of a gearbox, and taking the fixed-length diagnosis sample as a KNN graph node inputted by a graph attention network through FFT; a fuzzy distance between graph nodes is calculated by adopting a pooling strategy, a dynamic attention mechanism is introduced to solve the problem that a static graph attention network distributes weights of different types of nodes similarly, a dynamic multi-head graph attention gearbox fault diagnosis model is constructed, and a Softmax function is used as a classifier. Semi-supervised learning is realized through a label propagation algorithm under the condition of few label samples; and performing model training by using an Adam optimizer through a back propagation method, and storing the trained fault diagn