The Multidisciplinary Design and Structural Optimization Analysis for Complex System Modeling

The major techniques of Multidisciplinary Design Optimization (MDO) include MDO for modeling, approximation methods, design for analysis, system decomposition, system sensitivity analysis, search algorithms of design space, optimization algorithms, MDO computing environment, etc. It is purposed to make most of the collaborative effect interacted among various disciplines to reach the global optimal solution systematically. This plan is mainly aimed to use structural dynamic analysis and performance improvement of complex product as a carrier for insightful exploration on the techniques of Multidisciplinary Design Optimization of practical modeling, approximation theories, sensitivity analysis methods and multi-objective optimization. Finally, the solid foundation for promotion and researches in new MDO fields can be well established. Currently, the approximation techniques frequently used by MDO include Design of Experiment, Polynomial Response Surface Methodology, Kriging, Radial Basis Function and Back-Propagation Neural Network. Among them, Kriging is an approximation method with higher statistical features. This method is featured with local prediction. There is better continuity and derivatives in correlation functions. In view of solving high non-linear problems, it is also featured with good curve-fitting results. It is the reason why we adopted this method in this study. Originally, this plan has been positively approved by NSC review members since last year. It had reached the standard past one plan, but did not pass the high threshold for multi-plan application. After this plan is further revised with more contents, it is now proposed to the first priorities of the two-year application plan (99/08/01~101/07/31), we look forward to continue to receive the support of the review members. The subject focuses on a cylindrical grinder. It is mainly based on creditable 2D lump mass model after experimental validation. Fist year the system dynamic equations can be automatically deducted by the past research software of transfer matrix method for multi-body system. After the dynamic characteristics and modal sensitivity analysis are conducted then refined the model, the elastic behavior of the spindle, workpiece and other modules is considered. The dynamic characteristics and the influence on workpiece surface quality caused by grinding stiffness are also explored due to coupling effects. Meanwhile, it is also compared with those models. Finally, sample points are selected by using the design of experiment to construct the patterns for Kriging. A detailed comparison is also used to explore the features and feasibility for some approximation techniques like RSM and BPNN which had studied in the past research with the reference available for engineering designs. In the 2nd years, the research is further planned to make a comparison with the surrogate model. Based on the optimal model after evaluation, integrate techniques of the Genetic Algorithms or other Intelligence Optimization Algorithms to Parallel Computing efficiency with the exploration on the analysis of multi-objective optimization. The foundation can be established to enter the realms of Multidisciplinary Design Optimization. Finally, the overall research results are well summarized thereafter.

Project ID:PB9907-10774
External Project ID:NSC99-2221-E182-022