Research On Prediction Of Aero-optical Imaging Deviation Based On Improved Optimization Algorithm

Aero-optical imaging deviation is one of the important factors that affect the development of precision guided weapons.Airborne computer can quickly obtain the current state of the aero-optical imaging deviation value,which has a direct application value to improve the hit rate.Aiming at the problems of high calculation cost,long implementation period,and high cost of wind tunnel experiments and numerical simulations,using the obtained typical working condition data,an improved optimization algorithm is used to establish a prediction model,and aero-optical imaging deviation are quickly estimated in actual engineering.Firstly,Based on the Extreme Learning Machine(ELM)algorithm,a new algorithm is proposed.Aiming at the problem of network instability caused by randomly given initial weights and thresholds of ELM,particle swarm optimization(PSO)is used to optimize the parameters of ELM.In order to improve the search efficiency of PSO,an imaging deviation prediction model based on improved PSO optimized ELM is established by using the method of dynamically changing the inertia weight.The simulation results show that the improved PSO optimized ELM model effectively overcomes the shortcomings of the instability of ELM network,the long search time of PSO algorithm and easily falling into local optimum,and can be effectively applied to imaging deviation prediction.Secondly,based on the Beetle Antenna Search(BAS)algorithm,a BAS-ELM imaging migration prediction model is implemented.The simulation experiments show that the prediction model based on BAS-ELM has fast convergence speed and high prediction accuracy.It can be used as a supplementary algorithm for improving PSO optimized ELM model in practical engineering,and validates the improved PSO optimized ELM model prediction results.Then,based on Support Vector Regression(SVR),an imaging deviation prediction model based on Improved Genetic Algorithms(IGA)optimized SVR is constructed.The parameters of kernel function and penalty function of SVR are optimized by adaptive genetic algorithm.The crossover probability and mutation probability of GA are adjusted dynamically in the process of evolution.The simulation results show that the IGA-SVR prediction model has higher accuracy and shorter prediction time than the SVR model using GA,and can predict the imaging migration well.Finally,a set of imaging migration analysis software is designed and established.Through the analysis software,the CFD software FLUENT can be called to perform actual calculation of the imaging offset,or the prediction model can be used for rapid engineering estimation.Within a given range,by inputting the values of height,Mach number,angle of attack,and line of sight angle,the imaging deviation can be predicted by fitting a good prediction model.