Diesel engine continuous working condition decoupling identification method based on graph self-attention network

The invention discloses a diesel engine continuous working condition decoupling recognition method based on a graph self-attention network, and belongs to the technical field of equipment state monitoring and diagnosis. Firstly, experimental working conditions and working condition labels are designed, and a training set and a test set are divided according to working condition distribution. Secondly, establishing a self-attention network layer, calculating similarity and succession among angular domain signal fragments, establishing a graph convolution network layer, aggregating neighbor node information, then establishing a working condition decoupling space, and mapping signal features to the working condition decoupling space in a dimensionality reduction manner; and finally, forming, training and storing a working condition identification network model which ingeniously combines a self-attention mechanism and a graph convolution algorithm. When a new sample is input into the stored working condition reco