Research On Arithmetic Of MDO

Multidisciplinary Design Optimization (MDO) is a new engineering discipline which rose in the end of 1990s, and it can solve the problem of large-scale complex engineering system's design. Collaborative Optimization (CO) and Multimethod Collaborative Optimization (MCO) are of the most potential methods in MDO. As an initial research of CO within China, this thesis includes the whole architecture of CO, the emphasis are on the analysis and improvement of CO and MCO technology.The main contents of this thesis are as follows:1. The mathematic foundation, calculational flow and migratory strategy of distributed parallel genetic algorithm (DPGA) are researched starting with modern optimization algorithm. Distributed parallel genetic algorithm is proposed to design the parameter optimization of Proportional Integral Derivative (PID) controller. Appling it into simulator controls, PGA improves the micro adjusting of local search area, reduces the sensitivity of initial values. Traditional penalty function was often used to deal with nonlinear constrained optimization problems, but the effect was not perfect. In this article, fuzzy theory is introduced in optimization and a new fuzzy penalty function based on feasible degree is presented. This new algorithm integrates the optimums of different genetic operator, and uses parallel operation to improve the optimization efficiency. Then a satisfying performance is obtained.2. The conception of surrogate model is introduced based on the embedded analysis of CO framework, and a new fuzzy penalty function based on feasible degree is presented. The computation results indicate that multi-level surrogate model has the capacity of good approximations, and improves the local search ability of evolutionary optimization.3. Improving Collaborative Optimization frame based on former research on distributed parallel genetic algorithm, fuzzy penalty function and multilevel-surrogate models. The mathematical framework based on surrogate model is established, and is demonstrated with typical mathematic examples.4. A MCO algorithm is combined with genetic algorithms, Simulated Annealing methods and Powell's methods. Then the MCO algorithm is used on the plane design optimization. The results show that the MCO is effective for advancing the rate of the global optimal solution and saving the operation time.