The Study Of Kriging Model And The Application In The Optimization Of Reusable Launch Vehicles

As a new kind of launch machine,Reusable Flight Vehicle(RLV)has the potentials of reducing the cost of transportation in aerospace and increasing the ability of delivery,launch frequency and launch reliability.As a consequence,RLV owns great values in military affairs and commerce.This paper takes RLV as studying objects and researches the Double Extreme Algorithm and N-Kriging model.The multidisciplinary modle of RLV is also set in this paper.Considering the uncertainty in each subjuct,the UMDO of RLV is performed and the system solution that meets the demand of constrain reliability is obtained.Firstly,a new global optimization algorithm called Double Extreme Algorithm(DEA)is presented.The algorithm uses local optimization algorithms to search local optimum.Then search the wrost solutions around the optimial one to jump out of local optimum.Use Householder transformation and line search to expand searching area and get the highest probability of finding the global optimum.In the function tests,DEA has a higher probability of finding the optimum than GA and PSO.Secondly,when the sampling values contain measuring errors,detailed derivation processes of N-Kriging models and variances of prediction errors are presented in this paper.The results of simulation experiments indicate that the variance of predicting error obtained using the method in this paper accords with experimental results very well,which means this method can compute the variance of predicting error and constitute the basis of uncertainy optimization.Thirdly,the wing geometry of RLV is modeled by parameters.Use Latin Hypercube Sampling to build the geometry of RLV with different wing parameters.The aerodynamic force of RLV is computed by Cart3 D simulation software and the aerodynamic model is set by N-Krigng model.The models of structure,trajectory,mass and control subject are also set.Finally,UMDO is applied in RLV.Take the solution of MDO as the baseline and using sensitive analysis to compute the the influence of each uncertain factor on constraint.Choose 6 factors with the biggest influence to perform the UMDO and obtain the solution that meets the demand of constrain reliability.This paper studies the optimization algotithms and metamodels in UMDO and applies them in the UMDO of RLV.The solution that meets the reliability demand is obtained and this confirms the validity of UMDO in the preliminary design of RLV,which can provide references for further research in RLV.