Traditional fault diagnosis of aircrafts rudder surface has poor effects and weak generalization ability.To solve the problem,a fault diagnosis model suitable for civil aircrafts elevators is built based on Convolutional Neural Network (CNN) algorithm and Support Vector Machine (SVM) classifier.The model learns the original fault data layer by layer to extract fault features and identify fault types,and uses SVM instead of Softmax function to classify the faults.The model is compared with the traditional models of CNN network and Deep Belief Network (DBN).The experimental results show that the proposed model has the highest accuracy for elevator fault recognition,which can be above 99%.In order to directly observe the differences of the three models in feature representation,the recognition results are visualized through dimensionality reduction (T-SNE).Through the visualized graphs,it can be seen that the CNN-SVM model has significant clustering effects.Finally,noise is added to the data set,and it is verified that the proposed model has better anti-noise ability,generalization ability and reinforcement learning ability than the other two models.