Research On Adaptive Hybrid Response Surface Optimization To Weapon Delivery Planning

This thesis focuses on weapon delivery planning that is an important aspect in the tactical aircraft mission planning. The weapon delivery planning can be regarded as a constraint optimization problem with multi-variables, highly nonlinear and discontinuous. The decision-goals and constraints based on actual systems are costly to be evaluated in these problems. To reduce mentioned evaluation cost, response surface model is proposed and developed as a surrogate-model for planning. A response surfaces based optimization framework is constructed. Afterwards, hybrid response surface models are designed. Furthermore, a response surface optimization method is proposed and applied into applications in weapon delivery planning. The main achievements and progress are as follows:Firstly, an optimization framework based on response surface model is established. Response surface technique can quickly create an approximation between a set of independent variables and the system's response. Therefore, an optimization framework is constructed by integrating the merits of response surfaces and the surrogate-model optimization. The design of experiments, establishment and solving of optimization problems, and updating of response surface are then discussed, respectively. This provides a feasible fundamental framework for further design of response surface optimization method.Secondly, a hybrid response surface (HRS) is presented. The polynomial and radial basis functions are used to estimate the response surface approximation accuracy that may affect the optimization efficiency specifically. And the HRS is modeled to speed up the whole optimization process. It is significant that the HRS model can be free of the coefficient matrix that may appear singular to ensure the uniqueness of the model.Thirdly, an adaptive hybrid response surface optimization (AHRSO) method is proposed. The AHRSO method is put forward on the basis of symmetric Latin hypercube design (SyLHD) and circle dynamic constrained (CDC) strategy. A restart mode is added to prevent the optimization process dropping into a local search, which improves the performance of the proposed method. Experimental results indicate that AHRSO is competitive for global optimization, and stable for global optimization.Finally, an AHRSO based problem-solving method is presented for the weapon delivery planning. The key points of a typical Pop-up attack in the weapon delivery process are analyzed. Thus, the process parameters are regarded as decision variables. Weapon delivery planning problems are modeled as a standard optimization model with goals and constraints in the planning. A weapon delivery planning problems-solving method is then presented by using the response surface optimization. Compared with the original planning approaches, the presented method can get better planning results with less evaluation cost.