Study On Sequential Approximate Optimization And Its Application

Sequential Approximate Optimization(SAO) has been shown to be an effective approach for the design of computationally expensive models, e.g. the aerospace systems, and it is an important methodology to reduce costs and enhance efficiency. The success or failure of a sequential approximate optimization rests on the approximate accuracy and in?ll criteria. In order to improve the performance of sequential approximate optimization, the construction of a high accuracy surrogate model and selection of infill point is studied. The improved SAO algorithm is applied in the optimization design of Boost-glide Missile. The main study contents are shown as follow:As intelligent optimization algorithm has shown its shortages when long running expensive computer simulations is used in optimization design, the advantage of SAO, a high efficiency to get a global optimum solution, has been shown in this dissertation.By comparative analyses of some surrogate model, the radial basis function(RBF) is chosen to construct the surrogate model in SAO. How the shape parameter affects the accuracy of RBF surrogate model is studied. By analyzing the physical meaning of shape parameter, an calculation of local density of sampling points was proposed, based on which an determination of shape parameter has been established. Through the examination of standard test function, the accuracy is improved when the new methodology is applied.The infill criteria of SAO has been comparatively studied and analyzed. The shortage of single objective optimization based infill criteria has been shown, which focus on improve one of the global accuracy and solve efficiency, and cannot obtain a better balance between exploration and exploitation. A multi-objective optimization based infill criteria were proposed, which can balance exploration and exploitation, has ability to both local search and global search. The improved method is more efficient to predict the optimum solution by locating the infill points discriminatingly.The enhanced RBF surrogate model and improved infill criteria are employed in the SAO algorithm, and a calculation flow is proposed. A “2-step determination of convergence” was applied to determinate whether the algorithm convergent. 1): determinate whether the surrogate model is updated; 2): determinate whether the global approximate accuracy is enhanced. The algorithm also has been validated by standard test function.Lastly, the proposed methodology is applied to the system optimization design of boost-glide missile, and obtains an approving solution, which validates the proposed algorithm again.