Research On Load Identification Algorithm Of Aircraft Based On Neural Network

This thesis studies the theory of dynamic load identification for aircraft structure.Nowadays,the techniques of dynamic load identification are mainly separated into two classes:time domain identification method for dynamic load and frequency domain identification method.This article propose a method that Mapping relationship between strain response and load by intelligence algorithm,avoid the ill posed problem of time domain method in recognition process and improved the accuracy.Firstly,the application of multiple linear regression in load identification is studied.Find the more suitable node to build regression equation based on an optimal approach for load equations selection,establish a more robust regression model by removing the anomaly points through residual analysis.It is tested on cantilever beam model,cylindrical shell model and missile model respectively.Experimental results show that the algorithm can achieve high recognition accuracy.Secondly,this thesis also studies the dynamic load identification algorithm based on neural network,optimization based on the combination of genetic algorithm and BP neural network is executed to avoid the problem of local optimum.Compare the convergence speed and load identification precision of two kinds of networks on the cylindrical shell model,The results show that the convergence speed of GA-BP network is faster and the precision is higher.At the same time,we use a missile's simulation data to verify by adding different levels of noise,the results show that the model can still achieve high recognition accuracy under the condition of noise interference.Finally,studies the method of identification of parametric flight loads.Filter the input independent variable by MIV algorithm,Then train the model through BP network to obtain the identification result of flight load.Taking a missile model as the research object,we get flight parameters and flight load data by flight mechanics and shell internal force formula.The results show that the model running time can be significantly shortened while ensuring the accuracy of the model.