Automatic identification of flaws is very important for ultrasonic nondestructive testing and evaluation of large shafts. A novel automatic identification model of defects is presented. Empirical Model decomposition (EMD) is applied to feature extraction of ultrasonic signals, and possibility Support Vector Machine (SVM) with Dempster-Shafer (DS) theory to perform the identification task. Meanwhile, comparative study on convergent velocity and classified effect is done among SVM and artificial neural network (ANN) with DS models. To validate the method, some experiments are performed and the results show the proposed system has very high identification performance for large shafts and the possibility SVM processes better classification performance and spreading potential than ANN with DS model under the small study sample condition.