The invention discloses a
rotary machine axis trajectory recognition method based on
deep learning. The method comprises the steps of firstly, collecting axis trajectory data under a fault of a
rotary machine, obtaining an axis trajectory diagram and a shape
label corresponding to the axis trajectory diagram, and forming a fault sample
library, secondly, performing data enhancement on the axis trajectory diagram in the sample
library, and then constructing an axis trajectory recognition model based on a deep neural network, and thirdly, collecting axis trajectory data during operation of the rotating
machine in real time, conducting comparison diagnosis based on the constructed axis trajectory recognition model, determining the shape of the axis trajectory online, and then determining the fault type. According to the method, automatic axis trajectory recognition can be realized without depending on complex description mathematical
feature extraction, and meanwhile, different view features of the axis trajectory are extracted by using
convolution kernels of different sizes, so that the recognition precision is improved; in addition, the fault sample
library can be updated according to real-
time data, the recognition model is continuously optimized, and the function of self-perfecting and upgrading is achieved.