Aircraft full-automatic pneumatic optimization method based on reinforcement learning and transfer learning

Abstract

The invention discloses an aircraft full-automatic aerodynamic optimization method based on reinforcement learning and transfer learning. The method is used for solving the problem that an existing pneumatic optimization method is prone to falling into local optimization or low in convergence speed, manual intervention is excluded in the final high-precision optimization stage through the optimization method, and the optimization efficiency is further improved. According to the technical scheme, firstly, a reinforcement learning environment based on semi-empirical estimation and high-precisionfluid simulation is established; a reinforcement learning neural network is constructed; a reward function is set, the global optimization capability of reinforcement learning is utilized; in the network training process, optimization experience is extracted from a semi-experience estimation method and stored in network parameters; then, another reinforcement learning neural network is constructed, migration learning is used for migrating the extracted optimization experience to the network, then the network is applied to aerodynamic optimization based on high-precision fluid simulation, andfinally, high-precision design parameters with excellent aerodynamic performance are obtained by training the network. Compared with a background technology method, the method has the advantages thatthe convergence speed is increased, the strong global optimization capability is realized, and the engineering value for high-precision pneumatic optimization is very high.

Description

technical field [0001] The invention belongs to the technical field of aircraft engineering, in particular to an aircraft global aerodynamic optimization method based on reinforcement learning and migration learning. Background technique [0002] Aerodynamic optimization refers to the design of the shape and relative position of the main components of the aircraft. It is necessary to obtain the design with the best aerodynamic performance under given constraints. The mainstream aerodynamic optimization methods can be divided into two categories: gradient-based methods and traditional intelligent methods. The optimization efficiency of the single extreme value problem of the gradient-based method is very high, but most of the aerodynamic optimization is a complex multi-extreme value problem, and the gradient-based method is generally easy to fall into the local optimum, which cannot meet the needs of the global aerodynamic optimization. Traditional intelligent methods mainly...

AI Technical Summary

Problems solved by technology

[0004] In order to solve the problem that the existing aircraft aerodynamic optimization algorithm cannot have both the global optimization ability and the convergence speed, the present invention proposes an aircraft global aerodynamic optimization method based on reinforcement learning and transfer learning, and this method does not require Manual intervention can realize fully automatic optimization, further improving the efficiency of optimization

Images