Research On Intelligent Algorithms And Approximation Models To Multidisciplinary Design Optimization

Complex product (such as aircraft, ship, automotive vehicles and machine tools) design optimization contains multi-level and multidiscipline which it interact and couple together, which Leading the design optimization to be quite complex and hard to be computed. Multidisciplinary design optimization (MDO) is a frame with system synthetically optimization, for solving previous problems. Due to validity and practicability of decomposition and optimization, MDO has become focused on, gradually. Complex product design decomposition methods are the main ways to reduce design process and large amount of computation and have been attracted great attention in resent time. In the process of complex product design optimization, research on multidisciplinary design decomposition is interdisciplinary, thus, huge knowledge is necessary.In this dissertation, chapter three and based on the multidisciplinary design optimization of single level decomposition of complex product design optimization is studied by two case study. A research system of multidisciplinary design through all in one (AIO) method is constructed, including the design constraint. In remain of the dissertation and to fix the problem with multilevel decomposition, Analytical Target Cascading based on Physical Programming (ATC-PP), applied to solve the decomposition and equilibrium point of coupled variables in noncooperative element, stability analysis of equilibrium point, the construction of approximation model on analysis module,..etc. The above research system is well proved by a ship design optimization case study.Firstly, through (AIO) method based on Artificial Intelligence algorithm (Particle Swarm Optimization (PSO), Genetic Algorithm (GA)) is proposed. In this proposed experimental methods such as (Monte Carlo Sampling (MCS), Latin Hypercube Sampling (LHS)) is utilized to explore the design space and to sample data for covering the design space. Then, response surface method (RSM and Kriging) adopted as an approximation model for solving the system design problems and reducing the calculation and saving the develop time.Secondly, a new method named Analytical Target Cascading based on Physical Programming (ATC-PP) is proposed and its mathematical model and implement flows is researched. This method not only saves the merits of Analytical Target Cascading, decomposing the system based on components, but also utilized preference function to measure the errors of responses and linking variables between father and son elements. Engineers'design habits and preferences are thoroughly present by preference function. In the preference intervals, the problems of equilibrium point with coupled variables and stability of equilibrium point in noncooperative element are studied. The equilibrium points are solved by rational reaction sets presented by response surfaces.Thirdly, the construction method of Kriging model is proposed. Compared with other approximation models methods, Kriging model is more accurate. In ATC-PP, preference intervals, the means of design variables are unpredictived; as a consequence, a Blind Kriging model based on Bayesian Variable Selection Method is constructed and introduced. And for deeper research, corcibly remain linking variables in Blind Kriging model is proposed to avoid the linking variables deleting, which can make the whole system more feasible and stable.Fourthly, the proposed decomposition design system is employed in conceptual ship designs stage. A 50000DWT Handymax Bulk Carrier is taken as a case study to verify the proposed theory system in this dissertation is useful and ideal.Finally, the conclusions deduced from this dissertation...