In this paper, a classification method of flutter test signals based on convolutional neural network (CNN) and Hilbert- Huang transform (HHT) is established, which can be effectively applied to flutter boundary prediction. This method combines convolutional neural network and time-frequency analysis. Firstly, the flutter test signal is preprocessed by Hilbert-Huang transform and labeled according to the actual signal source. The label contains channel information and flutter information. All the signals and labels are composed of the dataset, the dataset is randomly scrambled and 80% of the number is taken as the training set, and the features are extracted, and the classification model is trained by convolutional neural network. The remaining 20% of the dataset is taken as the test set. The test set is used to test the classification model and verify the reliability and accuracy of the model. The accuracy of the final test set is above 90%, which indicates that the model trained by this method can effectively identify the channel information and the flutter information of the signal.
|