Parameterized Representations And Reuse Of Analysis Models For Thin-walled Structure Design

Simulation-based design optimization is an emerging paradigm for product development,and a thruster for deep fusion of industrialization and informationization.Currently,structural shape optimization is faced with two major challenges.One of the challenges is to propose effective shape parameterizations for structural design,and the other is to develop efficient methods for reproducing high-fidelity analysis models for simulation.Focusing on these two challenges,this thesis develops parameterization techniques and mesh deformation methods to reuse thin-walled structure models in carbody design,aiming to provide theoretical foundation and technical supports for simulation-driven design optimization.Specifically,following aspects are studied in this thesis:1)Since thin-walled beam components are ubiquitously used in engineering structure design,it is of significance to efficiently and reliably reuse their FE models.A skeleton-section parameterization is proposed based on advantages of suitable parameterizations in shape description and manipulation and structural properties of thin-walled beams.The radial basis functions(RBF)deformation method is used to globally and locally adapt legacy finite element(FE)models.The RBF-based method is extended by incorporating an anisotropic distance metric and a skeleton-embedding space.This allows the deformation method to obtain local modifications of complex thin-walled beam components using only sparse cross-sections.Numerical results demonstrate that the proposed method is able to achieve high-quality FE model adaptations,providing an efficient geometric processing tool for structural design of thin-walled beam components.2)The previously proposed method is applied to shape optimization design of an S-shaped frame with respect to its crashworthiness performance.The optimization task takes as design parameters the global shape of the structure and local geometries around its plastic hinges,and studies the influences of these parameters on the component’s crashworthiness.Optimal results show that improvements on crashworthiness are observed from both global and local model variations.With the proposed parametrized deformation method,the design space is efficiently sampled and a large set of quality FE model variations are obtained,which demonstrates the integrability between the proposed method and the surrogate-based optimization framework,enhancing the automatic level of design optimization of structural shape.3)A line-facet simplified representation and an algorithm for extracting this representation are proposed for general thin-walled components based on the observation that these components are often composed of beams and plates.Beam-like regions and plate-like ones are discriminated automatically based the sweeping approximation,and segmented via a region-growing strategy.The line-facet simplified representation is then extracted from the segmentation result.Numerical results show that the proposed algorithm is fast and able to extract simplified representations corresponding to the given components.Comparisons with manually extracted models show the effectiveness of the proposed algorithm.4)In order to address the need of associating styling and structural designs,a mechanism is proposed to concurrently edit styling and analysis models,and to overcome the challenge posed by the heterogeneity between the two models.A pair of wireframes each representing the styling or structural models are proposed,and deformation transfer between the pair of wireframes is achieved to associate the two heterogeneous models.The free-form deformation technique is extended by using the structural wireframe as driving handles to derive FE model deformation of body-in-white structures for carbody design.Desirable wireframe deformation results are obtained,and FE analyses of the deformed models show effectiveness of the proposed concurrent editing mechanism.