Reconstruction for Aircraft’s Pitch Angle Data via the Neural Network

Abstract

This paper proposes a method to reconstruct the pitch angle data of F15C aircraft through the data of roll angle, roll rate, yaw angle and yaw rate. It mainly includes data generation, data preprocessing, network training and network prediction. The flight data of F15C generated based on FlightGear software is divided into a training set and a test set. Through optimizing the calculation method of neural network learning rate, an adaptive learning rate calculation method is designed. Neural network is trained by roll angle, roll rate, yaw angle and yaw rate data of the training set, and its parameters are obtained which represent the weight between layers in the network. The data of roll angle, roll rate, yaw angle and yaw rate in the test set is imported into the network to get the reconstruction value of pitch angle data. This method can more accurately describe the complex nonlinearity of the sensor and has high accuracy of the reconstructed value without the model of the sensor. The simulation results explain the validity of the proposed method.