A dropout-based bilstm network wing drag coefficient prediction method

Abstract

The invention discloses a BiLSTM network wing drag coefficient prediction method based on Dropout. A wing parameter data set is constructed by acquiring the wing shape parameters and the wing drag coefficient, and a feature engineering data set is obtained according to the wing parameter data set. For the training of constructing the Dropout-BiLSTM network model, the prediction of the wing drag coefficient is completed based on the Dropout-BiLSTM network model after training. The network is convenient for sequence modeling, and can solve the problem of gradient explosion or gradient disappearance in the optimization process to a certain extent. Combined with the Dropout mechanism, it can avoid the advantages of over-fitting in the model and obtain high accuracy while greatly reducing the operation. The time scale of the process.

Description

technical field [0001] The invention relates to the field of wing drag coefficient prediction, in particular to a Dropout-based BiLSTM network wing drag coefficient prediction method. Background technique [0002] High-altitude and long-endurance UAVs have broad application prospects in the military and civilian fields, and airfoil is the key to affecting the aerodynamic performance of this type of aircraft. Therefore, the prediction of the aerodynamic performance of the laminar airfoil is a key technology in the research of high-altitude and long-endurance UAVs, and the aerodynamic drag coefficient is an important part of it. [0003] The current prediction methods for the aerodynamic drag coefficient of wings are mainly divided into traditional numerical simulation methods and artificial neural network methods. [0004] The traditional method to obtain the wing drag coefficient usually uses computational fluid dynamics (CFD) software for numerical simulation, establishes ...

AI Technical Summary

Problems solved by technology

[0006] In view of the above-mentioned shortcomings in the prior art: the aerodynamic drag coefficient cannot be efficiently and accurately predicted, and the local optimum and over-fitting problems in the calculation process are not avoided, the present invention provides a dropout-based BiLSTM network wing drag coefficient prediction method

Images