Neural network prediction of wind field and landing point in target area of return capsule

A new method is studied to forecast the wind field and landing point based on neural network aiming at the problem of landing point calculation of spacecraft return module in this paper. Firstly, the wind field data are preprocessed, and the wind field on a specific date is predicted by long short-term memory artificial neural network (LSTM); then, by taking the parachute opening parameters and landing point coordinates as the input and output of the error back propagation (BP) neural network, a double-layer implicit neural network is established. Compared with the traditional method of establishing the physical model and dynamic differential equation of the parachute-payload, it can shorten the solving time by more than 90%, and the result is reliable. It is convenient to intuitively obtain the corresponding relationship between height, velocity and landing point. The influence of different layers of network and training sample number on the accuracy is compared, indicating the rationality of the network structure. The feasibility of using neural network to predict wind field and landing point of re-entry capsule is illustrated.