Aerodynamic modeling method and system based on MAML

Abstract

The invention discloses an aerodynamic modeling method and system based on MAML. The aerodynamic modeling method comprises the following steps: S1, generating an aerodynamic data set for machine learning; s2, based on the aerodynamic force data set, establishing a base learning process; and S3, on the basis of the base learning process, obtaining a meta learning process. According to the method, a relatively good parameter of the base learner and the meta learner can be learned, and after the parameter is obtained, convergence in the model can be rapidly carried out by only needing a small number of learning samples for a nonlinear unsteady physical problem of a similar scene, so that the generalization ability under the condition of a small number of samples is improved, the precision and the learning efficiency are both considered, and the method is suitable for large-scale popularization and application. And the method has a relatively strong engineering application background.

Description

technical field [0001] The invention belongs to the field of aircraft aerodynamic design, in particular to an aerodynamic modeling method and system based on MAML. Background technique [0002] At present, there are mainly two types of nonlinear unsteady aerodynamic modeling methods widely used in engineering: one is to establish traditional mathematical aerodynamic models related to aerodynamic and flight physical quantities (such as algebraic models, step response models, etc.) , and the other is the pneumatic model of intelligent learning (such as fuzzy logic method, support vector machine (SVM), etc.). Among them, the traditional mathematical method is to carry out a large amount of aerodynamic data to carry out piecewise linear aerodynamic modeling. The accuracy of the model is low, and the parameter identification is difficult, which has gradually failed to meet the needs of existing projects. Intelligent learning methods can establish high-precision multi-input multi...

AI Technical Summary

Problems solved by technology

Among them, the traditional mathematical method is to carry out a large amount of aerodynamic data to carry out piecewise linear aerodynamic modeling. The accuracy of the model is low, and the parameter identification is difficult, which has gradually failed to meet the needs of existing projects.

Intelligent learning methods can establish high-precision multi-input multi-output nonlinear aerodynamic models, which are very suitable for nonlinear unsteady aerodynamic modeling. However, because such methods are mainly data-driven, there are generally large data requirements and long learning time , prone to overfitting, and weak generalization ability of small samples, which greatly limits the engineering application of intelligent learning aerodynamic modeling methods

Images