Research on CNN-BiLSTM Flight Training Slope Prediction Based on Attention Mechanism

Abstract

Today's civil aviation flight training has a large demand and high safety requirements. In order to better study the issues about flight training safety, the aircraft slope in flight training is studied. Specifically speaking, a CNN-BiLSTM model based on the attention mechanism is proposed to predict the flight slope problem in flight training. The roll angle of flight in training data is the object of study. Firstly, the parameters with greater influence on the roll angle are selected by gray correlation analysis. Secondly, the time window technique is used to convert the time series prediction problem into a supervised learning problem, and the feature vectors are extracted by convolutional neural network, and the extracted feature vectors are input into the BiLSTM network based on the attention mechanism. Finally, the optimal parameters of the model are determined by grid search, and the roll angle in the training data of a training model is predicted experimentally and the generalization ability of the model in other attitudes of flight training is investigated. The results of experiments show that the prediction model has better performance, high accuracy, small error and good stability in the flight training slope prediction problem compared with the traditional RNN prediction model and LSTM prediction model.