Evaluation Of Infrared Air-to-air Missile Anti-jamming Performance Based On Bayesian Neural Network

With the development of infrared technology,countermeasure environment faced by air-to-air missile is becoming increasingly serious.The anti-interference ability has been a key performance indicator of air-to-air missile.This paper is devoted to research anti-jamming evaluation of air-to-air missile.Hybrid prior,power prior approach and Bayesian neural network(BNN)are introduced to this evaluation problem,and it is evaluated successfully.The main work is described as follows:The application of hybrid prior and power prior approach on anti-jamming evaluation of air-to-air missile is researched.As for hybrid prior approach,three determination methods of hybrid parameter are put forward,and several relevant features of the third method are discussed.As for power prior approach,a new prior distribution of power parameter is reconstructed by use of prior and field-test data.An amendatory power prior approach is put forward,in addition to certain relevant features are studied,and the analysis result of mean square error(MSE)demonstrates that the estimation effect of amendatory approach is better.This paper deduces detailedly BP neural network and BNN training formula.These two networks are applied to the same matching example,and the result explains the better matching performance of BNN.An anti-jamming index system of air-to-air missile is constructed,and then three evaluation methods based on BNN are presented.The essence of these three methods is to construct the nonlinear mapping between index system and anti-jamming performance,and thus replace the expert scoring method by BNN.In the end,Analytic Hierarchy Process(AHP)which is a common method on engineering is used to solve the same evaluation problem,and the result is compared with that of the three evaluation method based on BNN.The analysis result shows that evaluation conclusions based on BNN are more similar and stable.Two solutions based on different evaluation model assumption are given,as for the evaluation problem of air-to-air missile under complicated infrared countermeasure environment.The first is evaluation model based on environment factor assumption,and it is solved by power prior approach combined with MCMC method.The second is evaluation model based on environment complexity assumption.BNN is used to construct mapping between environment complexity and miss distance distributed parameter of air-to-air missile.The post density function is established with the help of hybrid prior approach and thus this evaluation problem is solved.The simulation results of these two problems demonstrate the feasibility of the solutions.