Surrogate-Assisted Unified Optimization Framework for Investigating Marine Structural Design Under Information Uncertainty.

surrogate model;optimization;structural design;Naval Architecture and Marine Engineering;Engineering;Naval Architecture and Marine Engineering

Liu, YanSinger, David Jacob ;

University of Michigan

关键词: surrogate model;    optimization;    structural design;    Naval Architecture and Marine Engineering;    Engineering;    Naval Architecture and Marine Engineering;   

Others  :  https://deepblue.lib.umich.edu/bitstream/handle/2027.42/133365/yanliuch_1.pdf?sequence=1&isAllowed=y

瑞士|英语

来源: The Illinois Digital Environment for Access to Learning and Scholarship

【 摘 要 】

Structural decisions made in the early stages of marine systems design can have a large impact on future acquisition, maintenance and life-cycle costs. However, owing to the unique nature of early stage marine system design, these critical structure decisions are often made on the basis of incomplete information or knowledge about the design. When coupled with design optimization analysis, the complex, uncertain early stage design environment makes it very difficult to deliver a quantified trade-off analysis for decision making. This work presents a novel decision support method that integrates design optimization, high-fidelity analysis, and modeling of information uncertainty for early stage design and analysis. To support this method this dissertation improves the design optimization methods for marine structures by proposing several novel surrogate modeling techniques and strategies. The proposed work treats the uncertainties that are sourced from limited information in a non-statistical interval uncertainty form. This interval uncertainty is treated as an objective function in an optimization framework in order to explore the impact of information uncertainty on structural design performance. In this examination, the potential structural weight penalty regarding information uncertainty can be quickly identified in early stage, avoiding costly redesign later in the design. This dissertation then continues to explore a balanced computational structure between fidelity and efficiency. A proposed novel variable fidelity approach can be applied to wisely allocate expensive high-fidelity computational simulations. In achieving the proposed capabilities for design optimization, several surrogate modeling methods are developed concerning worst-case estimation, clustered multiple meta-modeling, and mixed variable modeling techniques. These surrogate methods have been demonstrated to significantly improve the efficiency of optimizer in dealing with the challenges of early stage marine structure design.