Research On Multi Objective Evolutionary Algorithm And Its Application In The Flight Vehicle Dynamics System

With the extension of application requirement in aeronautical and aerospace engineering and the improvement of the computational simulation ability, the requirement of design level for aerospace vehicle is increased. Duo to the design indeces for flight vehicle being conflict and multi-modal, the simulating being complex and timing costly, and the constraints being nonlinear, it becomes a great challenge for solving the optimization problem based on flight vehicle dynamics. Therefore, it is desired to develop efficient multiobjective and expensive optimization algorithms.In this paper, the multiobjective evolutionary algorithms based on decomposition(MOEA/D) and expensive optimization algorithms is deeply researched, and the ellipsoid decomposition method, the sequential optimization method with multiple-point sampling based on MOEA/D is proposed, which are applied in the multiobjective trajectory optimization for hypersonic glider vehicle, the control-structure integrated multiobjective optimization of flexible spacecraft, and the multiobjective optimization for the design of Mars explorer.The part of optimization algotithms research is organized as follows.Firstly, MOEA/D is studied. As the basis, the general decomposition approach and the framework of MOEA/D are introduced and analysized. Aiming to solve the difficulty that boundary intersection(BI) method is hard to convergence for the MOP with nonlinear constraints, a novel ellipsoid decomposition method is proposed. The proposed method are combined with the gradient-based optimization algorithms and MOEA/D, and compared with BI and Tchebycheff method respectively. The simulation results validate the efficacy of the aforementioned method.Secondly, the expensive optimization methods are studied. For improving the tradeoff between the local and global search in expensive optimization, the multiobjective optimization algorithms is proposed to balance the conflict search direction, and a novel sequential optimization method with multiple-point sampling method is proposed. The simulation results validate the advantage of the proposed method in both optimization efficiency and effectiveness over the traditional method. For the expensive multiobjective optimization problem, an improvement for MOEA/D-EGO is proposed. The method considers both the accuracy of Kriging model and the diversity and convergence of PF, which are not well balanced in the original algorithm. The simulation results validate the efficacy of the proposed method.Thirdly, the mesh refinement in the direct method for trajectory optimization is studied. For the purpose of improve the accuracy of trajectory optimization for aerospace vehicle, a mesh adaptive refinement method based on the second-generation wavelets is proposed. In the method, the original trajectory optimization problem is transformed into a nonlinear programming(NLP) problem by Runge-Kutta scheme, and then the NLP is solved by standard codes. After the optimization was completed, the wavelet transformation of control or state is performed to get the wavelets coefficients. The new nodes used to next iteration optimization could be determined based on the magnitude of the coefficients. The simulation results validate the accuracy and the efficacy of the proposed method.The part of optimization application research is structured as follows.Firstly, the multiobjective trajectory optimization problem(MTOP) for hypersonic glider vehicle is studied. The hybrid optimization method for solving MTOP is proposed, which combines the MOEA/D and the collocation method. The numerical simulation of MOP for maximum cross range and minimum peak heat of a glider vehicle is carried out. The simulation results show that the feasibility and efficacy of the proposed method.Secondly, the control-structure integrated multiobjective optimization of flexible spacecraft is studied. The motion equations of flexible spacecraft are derived from the Lagrange principle and the assumed modes method. The attitude control is based on PID controller. Then, the MOP for CSID is defined. A modified version of MOEA/D with our proposed hybrid constraint handling method is proposed for optimization. It has been applied to a spacecraft with symmetrically installed flexible appendages to find optimal tradeoffs. The numerical results shown that the method could give improvement of structural and control designs.Thirdly, the multiobjective optimization for the design of Mars explorer is studied. The discipline analysis model of trajectory, aerodynamic, heat and the condition of parachute stretch for Mars explorer is developed. The MOPs for both aeroshell shape optimization and entry trajectory optimization is defined, the objectives and constraints are based on the designing requirement. The two MOPs are solved by MOEA/D with the hybrid constraint handling method. The approximate PF obtained can make better tradeoffs between the objectives than the baseline design.Overall, in order to solve the difficulty that the MOEA/D solves the MOP aring from the design of flight vehicle based on dynamics, four issues, i.e. decomposition method, constraints handling method, expensive optimization method and mesh refinement for trajectory optimization, are studied. The proposed methods have been applied to the MTOP for hypersonic glider vehicle, CSID, and aeroshell shape optimization and entry trajectory optimization of Mars explorer. The research work can greatly advance the development of multiobjective evolutionary algorithms and enhance the system design quality of flight vehicles.