Research On Optimization Methods Of Weapon Equipment System Of Systems Based On Surrogate Model

To optimize the deployment of Weapon Equipment System of Systems ("SoS"for short thereafter ) aiming at maximization of the combat effectiveness under such constraints as operational requirements and expense, is one of the core issues in argumentation of SoS. To tackle the problem of low efficiency and high cost in traditional optimization methods of SoS based on simulation, we propose an optimization method of SoS based on surrogate model, in which the simulation model is replaced by surrogate model. The main achievements of the thesis are as follows:(1) An optimization framework of SoS based on surrogate model is proposed. We investigate the concepts and characteristics of the SoS and analyze the advantages and disadvantages of existing optimization methods. Then, an optimization framework of SoS based on surrogate model is proposed.(2) An optimization method of SoS based on polynomial surrogate model is presented. To solve the problem of deficiency of polynomial model in describing high-dimensional and nonlinear problem and to improve the efficiency of optimization, an optimization method of SoS based on polynomial surrogate model is presented. First, factor screening is conducted through a sensitivity analysis. The original points are generated using LHS and pre-optimized. Then, the surrogate model is updated using CCD and minimized response surface. Finally, local space searching is made using a gradient descent strategy.(3) An optimization algorithm of SoS based on Kriging surrogate model is proposed. Based on a thorough analysis of the mathematical mechanism, characteristics and applicability of the Kriging surrogate model, we propose an optimization algorithm of SoS based on Kriging surrogate model and uniform design. In the proposed method, the original points are generated from uniform design. Then, the surrogate model is updated using Maximization Expected Improvement method. A hybrid searching strategy of genetic algorithm and gradient descent is used for space searching at last. Experiments show that the proposed method is of higher precision and higher speed of convergence compared with methods based on polynomial surrogate model and simulation.(4) A demonstration study is conducted. We solve the optimization of penetration capability in air combat equipment system of systems using the proposed methods in this thesis.