By using Stacking ensemble strategy, a fault diagnosis method CNN-RF was proposed by integrating convolutional neural network (CNN) and random forest (RF) method. By using this method, not only the data features in the data set could be accurately extracted, but also an effective method could be provided to balance the errors of the data set for the problem of insufficient fault data in the data set, so the diagnostic accuracy could be improved. The collected real-time running data of embedded CNC system were analyzed and processed by using the separate model and the integrated model.The results show that by using the CNN-RF model, the accuracy of the fault diagnosis is high, by which the correctness of the model is verified.