A Design Method of Aero-Engine Labyrinth Seal Parameters Based on AGA-BP Neural Network

In order to improve the adaptability of the standard genetic algorithm (GA) in the dynamically changing environment, this paper proposes an adaptive GA (AGA), which can optimize BP neural network more efficiently. Furthermore, a design method of aero-engine labyrinth seal parameters is proposed, which aims to solve the problems of difficulty in obtaining experience parameters and low accuracy in the process of aero-engine labyrinth seal design. Better neural network parameters are attainable using the proposed AGA. Crossover and mutation operators in the proposed AGA have adaptively modified crossover and mutation probabilities depending on the change of fitness value across algorithm iterations. Experimental results on a certain type of civil aero-engine demonstrate that the proposed method has the ability to design the parameters of aero-engine labyrinth seal efficiently and precisely. In comparison to the conventional BP neural network, the proposed method achieves 0.91% MAPE outperforming BP with 1.22% MAPE. The proposed method achieves 0.06% MAE, which superior to GA-BP with 0.08% MAE. Moreover, the proposed method can release human resources and improve the success rate of civil aero-engine labyrinth seal parameter design.