外文翻译学习运用ProENGINEER几何模型建立有限元模型的过程

毕业设计(论文)外文资料翻译系 部： 机械工程系 专 业： 机械工程及自动化 姓 名： 学 号： (用外文写)外文出处： Kistler B L, Technical ReportR. SAND-97-8239,1997. 附 件： 1.外文资料翻译译文；2.外文原文。 指导教师评语： 外文资料内容与课题相关，译文能正确地表达原文的意文，语言表述基本符合汉语的习惯，语句较通顺，层次清晰。 签名： 年 月 日注：请将该封面与附件装订成册。附件1：外文资料翻译译文学习运用Pro / ENGINEER几何模型建立有限元模型的过程摘 要建立Pro/ENGINEER允许结构一体化模型的方法和生成热网格和无需重新几何图形计算的分析软件。本学习的目的不是要深入学习Pro/ENGINEER的力学或者生成网格或者分析软件，而是首次尝试对将产生有益的分析模型的时间比分析师需要创建一个单独的模型的时间更短的桑迪亚职员提供建议。该研究评价了运用Pro/ENGINEER建立各种各样的几何形状和对设计师、绘图员、分析师提供一般建议。致 谢绘图员Mark Mickelsen和Dennis Fritts直接支持这项研究；设计师Dave Neustel以及有限元分析师Hal Radloff Mike kanouff和Bruce Kistler。 此外，Arlo Ames的见解和运用Pro/ENGINEER 的能力是非常宝贵的。引 言有限元分析系统的执行过程或者组成部分一般分为四个步骤：1)定义几何形状，2)创建几何形状网格，3)应用网格的性能和边界及载荷条件，和4)执行有限元计算，和5) 检验分析结果。本研究检验前两步骤之间的关系，本研究的原因是电子设计（几何图形）定义的能力在过去十年取得了巨大的增长，像用Pro / ENGINEER这种电脑软件一样，1 现在能够经常定义实体几何。这些电子数据库不但可以创建传统的制造目的蓝图，而且还可以对计算机制造工艺和有限元分析传输电子信息。可能的话 ，这种电子信息的传输可以节省分析师在1和2 两个步骤的大量时间。此外，在过去十年中生成网格代码也有着显著的改善。许多不同的代码，现在有着在一般表面上自动生成壳网的能力。有些人具有（或接近了）将一般网格状实体要么自动生成四面体或六面体单元的能力。由于事情正在发生如此之快的变化以至于分析师和绘图员在如何最好地运用这些新工具上可能没有经验。这项研究以了解一些用当今的电子工具提高创建有限元模型过程和使用电子设计定义作为输入的分析机制。自桑迪亚已选择的Pro/ENGINEER作为其设计标准来定义计算机程序，审查Pro/ENGINEER某些细节。为了了解详情和本研究不同阶段的重大意义，我们相信，读者需要对几个领域有一个基本的了解。这些领域包括简短的有限元分析过程背景，生成网格能力（包括目前的问题领域）的目前状况，不同Pro/ENGINEER功能的说明和它们如何运用于有限元分析过程中，和了解这些技术变革可能怎样影响绘图员，分析师，以及制造商在设计工作之间的相互关系，鼓舞读者去深入学习这些章节的介绍。目的是为了了解主要研究这样做的意义。背 景在过去对系统或组件进行有限元分析是有难度的，需要给出一个关于结构或热方面效应的正确估算或者预测，为了进行分析，分析师需要画出或者以电子文档的形式建立几何模型，并提供相关的材料属性和负载条件。分析师利用这些信息，并对问题进行必要的假设后建立一个有限元模型。这个模型能够在一定的时间内得出一个近似的求解。因此，分析求解时间的长短是第三个需要考虑的因素。通常，几何图形信息是以图纸的形式提供给分析师的。分析师需要根据经验对一些细节（如螺栓孔，切断槽等）进行取舍，以便保证分析结果的近似准确度。然后，分析师将初始的几何图形以相对简单的形式重建。对几何图形的重建和在此基础上建立有限元模型的过程将耗费分析师80%的时间和精力。随着实体模型设计软件，如Pro/ENGINEER，和更加强大的计算机编码以及用于分析计算用的计算机的推广，我们可以更加方便地对电子实体模型直接进行分析求解。由于结构上的细节（如螺栓孔，切断槽等）对计算结果没什么影响，设计师仍然愿意舍弃它们，尽管在建立Pro/ENGINEER模型可以将它们考虑进去。也有些情况下，对于分析师使用实体几何图形来建模它可能会无效的一些理由。这有两个例子，1）薄结构，它可以准确地分析，使用三维壳单元比当用实体单元时更能降低计算成本和模型尺寸大小，和2）轴对称结构，可充分分析利用二维轴对称模型代表横截面。在这两种情况下或者重新创建几何模型或者使用Pro / ENGINEER的实体模型，分析师一定必须提前知道分析什么样的类型。这就依赖于当前的生成网格和分析技术了。例如，目前的生成网格技术只允许接受使用四面体单元（四环素）的一般实体几何图形的自动生成网格，即使六面体单元（六环素）通常用更少的单元提供一个更好的方案。因此，如果需要六面体单元的话，该分析师将不得不修改Pro / ENGINEER提供的几何模型，以适应非自动生成网格。此外，四面体单元往往有问题，甚至超越他们的最低精度。低价四环素要素往往表现出剪切闭锁和过度的刚度，而高阶四环素要素中不能使用明确的分析（动态分析需要非常小的时间间隔）。因此，分析师必须基于分析类型来选择生成网格类型部分。另一个考虑是模型的尺寸大小。有的3-D模型可以非常迅速地变的太大以至于无法运行，可能的原因或者是计算时间或者内存容量大小，这两者都是目前计算机所限制。“小”100x100x100单元的3-D网格产生一百万单元的模型尺寸，而迄今为止传统有限元模型已低于十万单元。因此，谨慎的做法是在有可能的情况下，以2-D为模型结构，即使3-D计算方法可能会产生更准确的结果。这又可能需要修改来自Pro / ENGINEER提供的实体几何模型。最后一个考虑是，通常绘图员“创建”一个Pro / ENGINEER模型比分析师“创建” 一个分析模型花费更少的时间和精力。因此，它是合理的（从整体设计到分析过程）以首先集中于可以用Pro / ENGINEER便于分析师的建模来做的事情。尽管这是一个事实，即制图者几乎总是由设计师，而不是分析师。因此除非设计师的同意，分析师可能会感到不太愿意对任何模型做出修改。目前存在的生成网格问题目前生成网格技术和啮合过程中有一些已知的障碍。这些障碍包括1）带有小功能大的几何模型的啮合问题，2）复杂的非标准几何形状的啮合问题，3）使用实体几何模型来创建壳模型的问题，4）连接不同部件的集合建模之间的问题，5）处理公差的问题，和6）如Pro / ENGINEER实体模型代码转移生成网格代码传输几何图形信息的问题。传统的生成网格技术可以自动生成低阶网格形状，具体地说，点、线、四曲面、和六面实体。在2-D中，目前的铺平技术现在仍然存在着一般性三角形和四边形单元几何图形网格，这些技术相对强劲。在3-D中，目前的技术现已存在的一般四面体（四环素）单元形状自动啮合，但没有更理想的六面体（六环素）单元。然而，这些3-D生成网格代码是不足以让每个几何模型网格总是成功，并且他们已越来越难以增加几何模型的复杂性。具体来说，在大型复杂几何模型上有许多小特征往往造成生成网格代码失败，因为他们无法完成从小单元（约小功能）到大单元和再一次回到（下一个小功能）的过渡 。同样的问题也可能会发生在没有“小”功能部分，但有很多复杂性功能。也就是说，从特征转换特征将最终失败，因为生成网格代码通常在一个起点和“扫描”走向另一个点。一位分析师可能需要分解单一的3-D部分将其分成若干“子部分”以便于部分网格能够成功。目前生成网格的另一个领域问题是由3-D几何模型来创建一个薄壳模型。一个薄壳有限元是没有厚度的，但假设任何单元有一半的厚度的刚度（即，它是被假定为处于中平面的厚度）。因为他们没有厚度，薄壳单元零件建模目的是为了与其他部分接触将现在的几何间隙隔开。确定这些新的界面往往是困难的。此外，实体模型设计定义代码（如Pro/ENGINEER）不容易或自动提供这种中平面曲面位置的生成网格代码摆在首位。因此，分析师可能创建几何模型可用于薄壳单元模型的决策，也可能创建几何模型定义壳单元模型的界面。附件2：外文原文A Study of the process of using pro/ENGINEER Geometry models to Create finite Element ModelsAbstractMethods for building pro/ENGINEER models which allowed integration with structural and thermal mesh generation and analyses software without recreating geometry were evaluated. This study was not intended to be an in-depth study of the mechanics of pro/ENGINEER or of mesh generation or analysis software, but instead was a first cut attempt to provide recommendation for Sandia personnel which would yield useful analytical models in less time than an analyst would require to create a separate model. The study evaluated a wide variety of geometries built in pro/ENGINEER and provide general recommendations for designers, drafters, and analysts.AcknowledgmentsThis study was directly supported by Mark Mickelsen and dennis fritts ,drafters ; Dave neustel and Hal Radloff , designers ;and Mike kanouff and bruce kistler finite element analysts. Also ,Arlo Ames was invaluable for his insight into the behavior and capabilities of pro/ENGINEER .IntroductionThe process of performing finite element analysis of systems or components consists generally of four steps :1) geometry definition ,2) mesh creation from the geometry,3) application to the mesh of properties and boundary and load conditions, and 4) performing the finite element calculations ,and 5) examining the result of the analysis . This study examines the link between the first two steps. The reason for the study is that the past decade has seen a tremendous growth in the capabilities of electronic design (geometry) definition, with such computer software as pro/ENGINEER . 1 now being able to routinely define solid geometries. These electronic databases can create traditional blueprints for manufacturing purposes, but can also transfer information electronically to computerized manufacturing processes and to finite element analysts. Potentially, this electronic transfer of information can save the analyst a significant amount of time in both steps 1 and 2.In addition , the mesh generation codes have also improved significantly in the last decade . Many different codes now have the capability of automatically generating shell meshes on general surfaces .and some have (or are close to having ) the ability to mesh general-shaped solids automatically with either tetrahedral or hexahedral elements.Because things are changing so quickly analysts and drafters may not have experience in how to best use these new tools .This study was undertaken to understand some of the mechanisms which would enhance the process of creating finite element models using todays electronic tools and using electronic design definition as input to the analyst. Since sandia has chosen pro/ENGINEER as its standard design definition computer program, pro/ENGINEER was examined in some detail.In order to understand the details and the significance of the different phases of this study, we believe that the reader needs to have a basic understanding of several areas. These areas include a brief background of the finite element process, a current status of mesh generation capabilities(including current problem areas), a description of different pro/ENGINEER capabilities and how they apply to the finite element analysis process, and an understanding of how these technological changes might affect the interrelationship between the work the designer, the drafter, the analyst, and the manufacturer ,the reader is encouraged to thoroughly study these introductory sections .in order to understand the significance of things that were done in the main study. BackgroundThe process of performing finite elements analysis of systems or components has in the past been challenging .the analyst could be call on to give either a very preliminary estimate of a structural or thermal response, or a very detailed prediction of that same response. To perform the evaluation, the analyst was typically given a geometry definition , either in paper or electronic form ,some materials information , and some load information .the analyst took this information and made enough assumptions about the problem to allow a finite element modal to be built which would result in an acceptable answer within the available amount of time. Thus, a limited time to perform an analysis was a third constraint.Often, the geometry information was given to the analyst in paper form . The analyst needed to make decisions based on experience to determine how much of the detail (such as bolt holes ,cut-outs, etc.) to include in order to have an acceptable level of accuracy in the analysis .then the analyst recreated, in some form , a simplified version of the geometry which had already been created by a drafter, this process , of reconstructing the geometry for the finite element model, and then of creating the finite element model , took up to 80% of the analysts time and efforts. With the more prevalent use of solid modeler design definition programs, such as pro/ENGINEER 1, and the more powerful codes and computers used by the analyst, it is now more feasible to attempt an analysis which directly utilizes an electronic solid model definition of the design. however, this is only beneficial if the analyst does not have to recreate or significantly modify the geometry to be compatible with the required analysis typically, the analyst would still like to ignore much of the detail (such as bolt holes ,cut-outs, ect) because that detail does not contribute to the accuracy of the solution ,even though that detail may be integrated into the pro/ENGINEER model。There are also instances where it may be inefficient for the analyst to work with a solid geometry for some reason . Two examples of this are 1) thin structures, which can be accurately analyzed using 3-dimensional shell elements at a lower computational cost and model size than when using solid elements; and 2) axisymmetric structures, which may be adequately analyzed using a 2-dimensional axisymmetric model representing the cross-section.In either recreating a geometry or using a solid geometry form Pro/ENGINEER, the analyst must know ahead of time what types of analyses are going to be required .This is dependent on the current state of mesh generation and analysis technology . for instance , current mesh generation technology only allows acceptable automatic mesh generation of general solid geometries using tetrahedral (tet) elements , even though hexahedral (hex) elements typically provide a better answer with fewer elements . Thus , if hex elements are required ,the analyst will have to modify the geometry provided from Pro/ENGINEER, to accommodate the non-automatic mesh generation. In addition, tet elements tend to have problems even beyond their lower accuracy. Low order tet elements tend to exhibit shear locking and excessive stiffness, while higher order tet elements cannot be used in explicit analyses (dynamic analyses requiring very small time steps). So the analyst must choose the type of mesh generation based partly on the type of analysis.Another consideration is model size. There-dimensional models can very quickly become too large to run either because of calculation time or memory size, both of which are limitations of the current generation of computers. A “small” 3-dimensional mesh of 100x100x100 cells result in a model size of a million elements, while traditional finite element models to date have been less than 100,000 elements. Therefore, it is prudent wherever possible to model structures as 2-dimensional, even when a 3-dimensional calculation may yield more accurate results. This again may require modification of solid geometry provided from Pro/ENGINEER.A final consideration is that typically it takes much less time and effort for a drafter to “build” a Pro/ENGINEER model than it does for an analyst to “build” an analysis model. Therefore, it is reasonable (from an overall design-to-analysis process) to focus first on things which can be done in Pro/ENGINEER to facilitate the analysts model building. This is despite the fact that the drafter is almost always funded by the designer rather than the analyst, and therefore might feel reluctant to do any model modification for the analyst unless agreed to by the designer.Problems with Current Mesh GenerationCurrent mesh generation technology and the meshing process have some known obstacles. These include 1) problems meshing large geometries with small features, 2) problems meshing complex non-standard geometric shapes, 3) problems using solid geometries to create shell models, 4) problems modeling the connectivity between different parts of assemblics, 5) problems handling tolerances, and 6) problems transferring geometry information from solid modeler codes such as pro/ENGINEER into mesh generation codes.Traditional mesh generation technology can automatically mesh low order shapes, specifically, points, lines, four-sided surfaces, and 6-sided solids. In two dimensions, current paving techniques now exist to also mesh general geometries with three and four-sided elements. These techniques are relatively robust. In three dimensions, current techniques now exist for automatically meshing general shapes with tetrahedral (tet) elements, but not the more desirable hexahedral (hex) elements. However, there three-dimensional mesh generation codes are not robust enough to always successfully mesh every geometry, and they have increasing difficulty with increased geometry complexity. Specifically, having many small features in a large complex geometry often causes mesh generation codes to fail because they cannot complete the transitions well from small elements (around the small features ) to large elements and back again ( to the next small feature). The same problem may also occur in a part with no “small” features, but with lots of complexity. That is, transitioning from feature to feature can eventually fail because the mesh generation codes usually start at one location and “sweep” toward another location. An analyst may need to break up a single three-dimensional part into several “sub-parts” in order to successfully mesh the partAnother area where current mesh generation has problems is in creating a shell model from a three-dimensional geometry. A shell finite element has no thickness, but assumes the stiffness of something which has half of the thickness on either side of the element ( that is, it is assumed to be positioned at the midplane of the thickness). Because they have no thickness, parts modeled with shell elements which are supposed to physically interface with other parts will now be geometrically separated by a gap. Defining these new interface can often be difficult. Futhermore, solid model design definition codes (such as pro/ENGINEER) do not easily or automatically provide this midplane surface location to the mesh generation codes in the first place. Thus, an analyst may have to create geometry to be used in making the shell element model, and may also have to create geometry to define the interfaces in the shell element model.