Aimed at the problem that it is difficult to obtain a large number of fault samples for large rotating machinery and the recognition rate of moment invariance is low, an identification method for rotor shaft orbit based on combined moment and stochastic forest model was proposed.The measured shaft orbit was taken as the sample,Sobel operator was used to extract the contour of the shaft orbit,and the combined moment was constructed based on the geometric features of the shape and the moment invariant.The moment invariants and the combined moments were used as input of the random forest model to classify,and the classification accuracy of the combined moments was the highest.The classification results of random forest, support vector machine (SVM) and BP neural network were compared.The results show that the classification accuracy of random forest is higher than that of SVM and BP neural network, and the identification time is shorter, so it is a new method to diagnose the fault of rotating machinery